The Princeton Guide to Ecology - PDF Free Download (2024)

The Princeton Guide to


The Princeton Guide to

associate editors

Stephen R. Carpenter University of Wisconsin Madison

H. Charles J. Godfray

Ecology Simon A. Levin editor Princeton University

University of Oxford

Ann P. Kinzig Arizona State University

Michel Loreau McGill University

Jonathan B. Losos Harvard University

Brian Walker CSIRO Sustainable Ecosystems

David S. Wilcove Princeton University

princeton university press managing editor

Christopher G. Morris

Princeton & Oxford

Copyright Ó 2009 by Princeton University Press Published by Princeton University Press, 41 William Street, Princeton, New Jersey 08540 In the United Kingdom: Princeton University Press, 6 Oxford Street, Woodstock, Oxfordshire OX20 1TW All Rights Reserved Library of Congress Cataloging-in-Publication Data The Princeton guide to ecology / Simon A. Levin, editor. p. cm. Includes bibliographical references and index. ISBN 978 0 691 12839 9 (hardcover : alk. paper) 1. Ecology. 2. Ecology Economic aspects. I. Levin, Simon A. QH541.p742 2009 577 dc22 2008049649 British Library Cataloging in Publication Data are available This book has been composed in Sabon and Din Printed on acid free paper. ? Printed in the United States of America 10 9 8 7 6 5 4 3 2 1

Contents Preface Contributors

vii ix

Part I Autecology I.1 Ecological Niche I.2 Physiological Ecology: Animals I.3 Physiological Ecology: Plants I.4 Functional Morphology: Muscles, Elastic Mechanisms, and Animal Performance I.5 Habitat Selection I.6 Dispersal I.7 Foraging Behavior I.8 Social Behavior I.9 Phenotypic Plasticity I.10 Life History I.11 Remote Sensing and Geographic Information Systems I.12 Geographic Range I.13 Adaptation I.14 Phenotypic Selection I.15 Population Genetics and Ecology I.16 Phylogenetics and Comparative Methods I.17 Microevolution I.18 Ecological Speciation: Natural Selection and the Formation of New Species I.19 Adaptive Radiation

1 3 14 20

Part II Population Ecology II.1 Age-Structured and Stage-Structured Population Dynamics II.2 Density Dependence and SingleSpecies Population Dynamics II.3 Biological Chaos and Complex Dynamics II.4 Metapopulations and Spatial Population Processes II.5 Competition and Coexistence in Plant Communities II.6 Competition and Coexistence in Animal Communities II.7 Predator–Prey Interactions II.8 Host–Parasitoid Interactions

27 38 45 51 59 65 72 79 87 93 101 109 117 126 134 143 153 155 166 172 177 186 196 202 213

II.9 Ecological Epidemiology II.10 Interactions between Plants and Herbivores II.11 Mutualism and Symbiosis II.12 Ecology of Microbial Populations II.13 Coevolution


Part III Communities and Ecosystems III.1 Biodiversity: Concepts, Patterns, and Measurement III.2 Competition, Neutrality, and Community Organization III.3 Predation and Community Organization III.4 Facilitation and the Organization of Plant Communities III.5 Indirect Effects in Communities and Ecosystems: The Role of Trophic and Nontrophic Interactions III.6 Top-Down and Bottom-Up Regulation of Communities III.7 The Structure and Stability of Food Webs III.8 Spatial and Metacommunity Dynamics in Biodiversity III.9 Ecosystem Productivity and Carbon Flows: Patterns across Ecosystems III.10 Nutrient Cycling and Biogeochemistry III.11 Terrestrial Carbon and Biogeochemical Cycles III.12 Freshwater Carbon and Biogeochemical Cycles III.13 The Marine Carbon Cycle III.14 Biodiversity and Ecosystem Functioning III.15 Ecological Stoichiometry III.16 Macroecological Perspectives on Communities and Ecosystems III.17 Alternative Stable States and Regime Shifts in Ecosystems III.18 Responses of Communities and Ecosystems to Global Changes


227 233 239 247

257 264 274 282

289 296 305 312 320 330 340 347 358 367 376 386 395 407


Contents III.19 Evolution of Communities and Ecosystems

Part IV Landscapes and the Biosphere IV.1 Landscape Dynamics IV.2 Landscape Pattern and Biodiversity IV.3 Ecological Dynamics in Fragmented Landscapes IV.4 Biodiversity Patterns in Managed and Natural Landscapes IV.5 Boundary Dynamics in Landscapes IV.6 Spatial Patterns of Species Diversity in Terrestrial Environments IV.7 Biosphere–Atmosphere Interactions in Landscapes IV.8 Seascape Patterns and Dynamics of Coral Reefs IV.9 Seascape Microbial Ecology: Habitat Structure, Biodiversity, and Ecosystem Function IV.10 Spatial Dynamics of Marine Fisheries

414 423 425 431 438 445 458

VI.5 VI.6 VI.7 VI.8 VI.9 VI.10 VI.11 VI.12

464 474


Forests Grasslands Marine Ecosystem Services Provisioning Services: A Focus on Fresh Water Regulating Services: A Focus on Disease Regulation Support Services: A Focus on Genetic Diversity The Economics of Ecosystem Services Technological Substitution and Augmentation of Ecosystem Services Conservation of Ecosystem Services

606 614 619 625 634 642 652 659 670

482 488 501

Part V Conservation Biology V.1 Causes and Consequences of Species Extinctions V.2 Population Viability Analysis V.3 Principles of Reserve Design V.4 Building and Implementing Systems of Conservation Areas V.5 Marine Conservation V.6 Conservation and Global Climate Change V.7 Restoration Ecology


Part VI Ecosystem Services VI.1 Ecosystem Services: Issues of Scale and Trade-Offs VI.2 Biodiversity, Ecosystem Functioning, and Ecosystem Services VI.3 Beyond Biodiversity: Other Aspects of Ecological Organization VI.4 Human-Dominated Systems: Agroecosystems


514 521 529 538 548 557 566

579 584 591 597

Part VII Managing the Biosphere VII.1 Biological Control: Theory and Practice VII.2 Fisheries Management VII.3 Wildlife Management VII.4 Managing the Global Water System VII.5 Managing Nutrient Mobilization and Eutrophication VII.6 Managing Infectious Diseases VII.7 Agriculture, Land Use, and the Transformation of Planet Earth VII.8 The Ecology, Economics, and Management of Alien Invasive Species VII.9 Ecological Economics: Principles of Economic Policy Design for Ecosystem Management VII.10 Governance and Institutions VII.11 Assessments: Linking Ecology to Policy Milestones in Ecology Glossary Index

679 683 689 695 701 712 718 724 731 740 748 754 761 775 793

Preface One can argue about when ecology was born as a science, although surely the writings of Charles Darwin and Alfred Russell Wallace created the essential context for the emergence of a new study of the interrelationships of species with each other and with their environments. The term ‘‘oekologie,’’ combining the Greek words for ‘‘household’’ and ‘‘knowledge,’’ was coined in 1866 by the remarkable German scientist, philosopher, and physician Ernst Haeckel and first was developed in scientific depth in the 1895 textbook by the Danish botanist Johannes Eugenius Buelow Warming, Plantesamfund—Grundtræk af den økologiske Plantegeografi [Plant Communities: An Introduction to Ecological Plant Geography]. Ecology has come a long way as a subject, from Eugen Warming to global warming. Ecology has its roots in natural history and, indeed, in evolutionary thinking. But ecology itself has evolved considerably since its birth, building bridges to mathematics, to the physical sciences and engineering, to molecular biology, and, increasingly, to the social sciences. Just as we are beginning to appreciate not only the beauty of natural systems but also their essential role in providing an infinite range of goods and services on which humanity depends, we are reluctantly also learning that we are destroying those life-support systems and threatening the sustainability of the biosphere as we know it. Ecology, the unifying science in integrating knowledge of life on our planet, has become the essential science in learning how to preserve it. This volume is an effort to present, in one readable collection, the diversity of ecology, from the basic to the applied. It is meant to serve both as a reader for anyone interested in learning more about the subject and as an essential reference for college and university courses on ecology and sustainability as well as for advanced high school students and the interested lay public. As such, it builds on the basic principles of autecology, population biology, and community and ecosystems science, which form the foundation for discussions regarding current threats to sustainability and how we can manage the biosphere responsibly. The Princeton Guide to Ecology is organized into seven sections tightly integrated with one another. The

core textual material is supplemented by suggestions for further reading at the end of each article, by a glossary of key terms, and by a chronology that traces landmark events in ecology. Ecology views biological systems as wholes, not as independent parts, while seeking to elucidate how these wholes emerge from and affect the parts. Increasingly, this holistic perspective, rechristened as the theory of complex adaptive systems, has informed understanding and improved management of economic and financial systems, social systems, complex materials, and even physiology and medicine—but essentially this means little more than taking an ecological approach to such systems, investigating the interplay among processes at diverse scales and the interaction between systems and their environments. In many colleges and universities where ecology has flourished, botany and zoology have vanished as separate departments and been replaced by more integrative ones. Ecologists tend to organize their thinking across scales, from cells to organisms, from organisms to populations, from populations to communities, ecosystems, landscapes, and the biosphere. This view also dictates the organization of this volume, which begins with autecology, the study of the physiology, behavior, and life history of the primary integrative unit of ecology, the organism. From the organismal level, the next natural levels of organization are the population, then the community and ecosystem, and then finally landscapes and the biosphere. With this basic foundation, the Guide then turns to more applied issues: understanding what biodiversity and the ecological systems in which they reside mean to us, as captured in the concept of ‘‘ecosystem services’’; exploring the scientific basis for managing our natural systems and the resources we extract from them; and developing the theoretical principles underlying the conservation of natural resources. These chapters naturally reach out to other disciplines, including economics and the social sciences, for the partnerships that are essential in achieving a sustainable future for humanity. If this ambitious effort has been successful, it is because of the exceptional quality of the authors and their contributions, and especially the remarkable set



of associate editors who have cheerfully integrated their sections and worked closely with one another to assure transitions as seamless as could be imagined. Anne Savarese at Princeton University Press and managing editor Chris Morris have assured smooth logistics throughout and added their own keen insights at appropriate times. I also am delighted to acknowledge the inspiration of Sam Elworthy, former editor-in-chief of Princeton University Press, who conceived the idea of the Guide and convinced me to take on the project.

As always, I am grateful for the unwavering support of Carole Levin, my wife and friend. As we go to press, our happiness at the completion of this effort is mixed with sadness because of the untimely death on March 22, 2008, of our distinguished contributor, Robert Denno. Simon A. Levin Princeton, New Jersey May 17, 2008

Contributors David D. Ackerly, Department of Integrative Biology, University of California, Berkeley I.3 PHYSIOLOGICAL ECOLOGY: PLANTS; I.16 PHYLOGENETICS AND COMPARATIVE METHODS

Eldridge S. Adams, Department of Ecology and Evolutionary Biology, University of Connecticut I.8 SOCIAL BEHAVIOR

Joseph Alcamo, Center for Environmental Systems Research, University of Kassel VII.4 MANAGING THE GLOBAL WATER SYSTEM

Priyanga Amarasekare, Department of Ecology and Evolution, University of Chicago II.6 COMPETITION AND COEXISTENCE IN ANIMAL COMMUNITIES This research was funded by a grant from NSF (DEB-0717350).

Darren Bade, Department of Biological Sciences, Kent State University III.12 FRESHWATER CARBON AND BIOGEOCHEMICAL CYCLES

Victoria J. Bakker, Division of Physical and Biological Sciences, University of California, Santa Cruz V.2 POPULATION VIABILITY ANALYSIS

Marissa L. Baskett, Department of Environmental Science and Policy, University of California, Davis VI.7 MARINE ECOSYSTEM SERVICES

Michael Begon, School of Biological Sciences, University of Liverpool II.9 ECOLOGICAL EPIDEMIOLOGY

Michael A. Bell, Department of Ecology and Evolution, Stony Brook University I.17 MICROEVOLUTION

Thomas Bell, Department of Zoology, Oxford University II.12 ECOLOGY OF MICROBIAL POPULATIONS

E. T. Borer, Department of Zoology, Oregon State University III.6 TOP-DOWN AND BOTTOM-UP REGULATION OF COMMUNITIES This work was completed as part of the Trophic Structure Comparisons Working Group supported by the National Center for Ecological Analysis and Synthesis, a Center funded by NSF (Grant #DEB-0072909), the University of California at Santa Barbara, and the state of California.

Mark S. Boyce, Department of Biological Sciences, University of Alberta VII.3 WILDLIFE MANAGEMENT

Corey J. A. Bradshaw, Research Institute for Climate Change and Sustainability, University of Adelaide V.1 CAUSES AND CONSEQUENCES OF SPECIES EXTINCTIONS

Cheryl J. Briggs, Department of Ecology, Evolution and Marine Biology, University of California, Santa Barbara II.8 HOST–PARASITOID INTERACTIONS

Judith L. Bronstein, Department of Ecology and Evolutionary Biology, University of Arizona II.11 MUTUALISM AND SYMBIOSIS

Barry W. Brook, Research Institute for Climate Change and Sustainability, University of Adelaide V.1 CAUSES AND CONSEQUENCES OF SPECIES EXTINCTIONS

Joel S. Brown, Department of Biological Sciences, University of Illinois at Chicago I.7 FORAGING BEHAVIOR

Ragan M. Callaway, Division of Biological Sciences, University of Montana III.4 FACILITATION AND THE ORGANIZATION OF PLANT COMMUNITIES

Stephen R. Carpenter, Department of Zoology, University of Wisconsin Madison VII MANAGING THE BIOSPHERE

Just Cebrian, Dauphin Island Sea Lab and Department of Marine Sciences, University of South Alabama III.9 ECOSYSTEM PRODUCTIVITY AND CARBON FLOWS: PATTERNS ACROSS ECOSYSTEMS

Je´roˆme Chave, CNRS (Centre National de la Recherche Scientifique), Laboratoire Evolution et Diversite´ Biologique III.2 COMPETITION, NEUTRALITY, AND COMMUNITY ORGANIZATION

Ryan Chisholm, Department of Ecology and Evolutionary Biology, Princeton University VII.8 THE ECOLOGY, ECONOMICS, AND MANAGEMENT OF ALIEN INVASIVE SPECIES

Scott L. Collins, Department of Biology, University of New Mexico IV.5 BOUNDARY DYNAMICS IN LANDSCAPES This research was supported by National Science Foundation support to the Sevilleta Long-term Ecological Research Program at the University of New Mexico (DEB 0620482).

Robert K. Colwell, Department of Ecology and Evolutionary Biology, University of Connecticut III.1 BIODIVERSITY: CONCEPTS, PATTERNS, AND MEASUREMENT

Molly S. Cross, Wildlife Conservation Society North America Program V.6 CONSERVATION AND GLOBAL CLIMATE CHANGE

Peter Daszak, Consortium for Conservation Medicine VI.9 REGULATING SERVICES: A FOCUS ON DISEASE REGULATION

Diane M. Debinski, Department of Ecology, Evolution and Organismal Biology, Iowa State University V.6 CONSERVATION AND GLOBAL CLIMATE CHANGE



Robert F. Denno, late Professor of Entomology, University of Maryland II.7 PREDATOR–PREY INTERACTIONS

Daniel F. Doak, Department of Zoology and Physiology, University of Wyoming V.2 POPULATION VIABILITY ANALYSIS

Martha Downs, Environmental Change Initiative, Brown University VI.6 GRASSLANDS

Laurie E. Drinkwater, Department of Horticulture, Cornell University VI.4 HUMAN-DOMINATED SYSTEMS: AGROECOSYSTEMS

Ray Dybzinski, Department of Ecology, Evolution, and Behavior, University of Minnesota II.5 COMPETITION AND COEXISTENCE IN PLANT COMMUNITIES

Stephen P. Ellner, Department of Ecology and Evolutionary Biology, Cornell University II.1 AGE-STRUCTURED AND STAGE-STRUCTURED POPULATION DYNAMICS

J. J. Elser, School of Life Sciences, Arizona State University III.15 ECOLOGICAL STOICHIOMETRY

Paul Falkowski, Department of Geological Sciences and Institute of Marine and Coastal Sciences, Rutgers University III.13 THE MARINE CARBON CYCLE

Myra E. Finkelstein, Division of Physical and Biological Sciences, University of California, Santa Cruz V.2 POPULATION VIABILITY ANALYSIS

Joern Fischer, Fenner School of Environment and Society, Australian National University IV.2 LANDSCAPE PATTERN AND BIODIVERSITY

Jonathan A. Foley, Center for Sustainability and the Global Environment and Department of Environmental Studies and Atmospheric and Oceanic Sciences, University of Wisconsin Madison VII.7 AGRICULTURE, LAND USE, AND THE TRANSFORMATION OF PLANET EARTH

Kevin J. Gaston, Department of Animal and Plant Sciences, University of Sheffield I.12 GEOGRAPHIC RANGE

Rosemary Gillespie, Department of Environmental Science, Policy and Management, University of California, Berkeley I.19 ADAPTIVE RADIATION

H. Charles J. Godfray, Department of Zoology, University of Oxford II POPULATION ECOLOGY


Indur M. Goklany, Office of Policy Analysis, U.S. Department of the Interior VI.12 TECHNOLOGICAL SUBSTITUTION AND AUGMENTATION OF ECOSYSTEM SERVICES

James R. Gosz, Biology Department, University of New Mexico IV.5 BOUNDARY DYNAMICS IN LANDSCAPES This research was supported by National Science Foundation support to the Sevilleta Long-term Ecological Research Program at the University of New Mexico (DEB 0620482).

Catherine Graham, Department of Ecology and Evolution, Stony Brook University I.11 REMOTE SENSING AND GEOGRAPHIC INFORMATION SYSTEMS

D. S. Gruner, Department of Entomology, University of Maryland III.6 TOP-DOWN AND BOTTOM-UP REGULATION OF COMMUNITIES This work was completed as part of the Trophic Structure Comparisons Working Group supported by the National Center for Ecological Analysis and Synthesis, a Center funded by NSF (Grant #DEB-0072909), the University of California at Santa Barbara, and the state of California.

Nick Haddad, Department of Zoology, North Carolina State University V.3 PRINCIPLES OF RESERVE DESIGN

Benjamin S. Halpern, National Center for Ecological Analysis and Synthesis, University of California, Santa Barbara VI.7 MARINE ECOSYSTEM SERVICES

Ilkka Hanski, Department of Ecology and Systematics, University of Helsinki II.4 METAPOPULATIONS AND SPATIAL POPULATION PROCESSES

Alan Hastings, Department of Environmental Science and Policy, University of California, Davis II.3 BIOLOGICAL CHAOS AND COMPLEX DYNAMICS

Andrew Hector, Institute of Environmental Sciences, University of Zurich III.14 BIODIVERSITY AND ECOSYSTEM FUNCTIONING

Philip Hedrick, School of Life Sciences, Arizona State University I.15 POPULATION GENETICS AND ECOLOGY

Nicole Heller, Department of Biology, Franklin & Marshall College III.18 RESPONSES OF COMMUNITIES AND ECOSYSTEMS TO GLOBAL CHANGES

Justin P. Henningsen, Biology Department, University of Massachusetts I.4 FUNCTIONAL MORPHOLOGY: MUSCLES, ELASTIC MECHANISMS, AND ANIMAL PERFORMANCE

Ray Hilborn, School of Aquatic and Fishery Sciences, University of Washington VII.2 FISHERIES MANAGEMENT

Richard J. Hobbs, School of Plant Biology, University of Western Australia IV.2 LANDSCAPE PATTERN AND BIODIVERSITY V.7 RESTORATION ECOLOGY

Robert D. Holt, Department of Zoology, University of Florida III.3 PREDATION AND COMMUNITY ORGANIZATION


Terry P. Hughes, Centre for Coral Reef Biodiversity, James Cook University IV.8 SEASCAPE PATTERNS AND DYNAMICS OF CORAL REEFS

Contributors Duncan J. Irschick, Biology Department, University of Massachusetts I.4 FUNCTIONAL MORPHOLOGY: MUSCLES, ELASTIC MECHANISMS, AND ANIMAL PERFORMANCE

Anthony R. Ives, Department of Zoology, University of Wisconsin Madison II.2 DENSITY DEPENDENCE AND SINGLE-SPECIES POPULATION DYNAMICS

Jeremy B. C. Jackson, Smithsonian Tropical Research Institute V.5 MARINE CONSERVATION


David M. Karl, Center for Microbial Oceanography: Research and Education, and Department of Oceanography, University of Hawai’i at Manoa IV.9 SEASCAPE MICROBIAL ECOLOGY: HABITAT STRUCTURE, BIODIVERSITY, AND ECOSYSTEM FUNCTION

A. Marm Kilpatrick, Consortium for Conservation Medicine VI.9 REGULATING SERVICES: A FOCUS ON DISEASE REGULATION

Joel G. Kingsolver, Department of Biology, University of North Carolina at Chapel Hill I.14 PHENOTYPIC SELECTION

Ann P. Kinzig, School of Life Sciences, Arizona State University VI ECOSYSTEM SERVICES

Allan Larson, Biology Department, University of Washington I.13 ADAPTATION

Julien Lartigue, Office of Oceanic and Atmospheric Research, National Oceanic and Atmospheric Administration III.9 ECOSYSTEM PRODUCTIVITY AND CARBON FLOWS: PATTERNS ACROSS ECOSYSTEMS

M. A. Leibold, Section of Integrative Biology, University of Texas III.8 SPATIAL AND METACOMMUNITY DYNAMICS IN BIODIVERSITY

Ricardo M. Letelier, College of Oceanic and Atmospheric Sciences, Oregon State University IV.9 SEASCAPE MICROBIAL ECOLOGY: HABITAT STRUCTURE, BIODIVERSITY, AND ECOSYSTEM FUNCTION


David B. Lindenmayer, Center for Resource and Environmental Studies, Australian National University IV.2 LANDSCAPE PATTERN AND BIODIVERSITY

Nicolas Loeuille, Universite´ Paris 6 (Laboratory of Ecology) III.19 EVOLUTION OF COMMUNITIES AND ECOSYSTEMS


Michel Loreau, Department of Biology, McGill University III COMMUNITIES AND ECOSYSTEMS

Jonathan B. Losos, Department of Organismic and Evolutionary Biology, Harvard University I AUTECOLOGY

John A. Ludwig, CSIRO Tropical Research Center IV.1 LANDSCAPE DYNAMICS

Pablo A. Marquet, Center for Advanced Studies in Ecology and Biodiversity and Ecology Department, Catholic University of Chile III.16 MACROECOLOGICAL PERSPECTIVES ON COMMUNITIES AND ECOSYSTEMS

Pamela A. Matson, School of Earth Sciences, Stanford University III.10 NUTRIENT CYCLING AND BIOGEOCHEMISTRY

Brian A. Maurer, Department of Fisheries and Wildlife and Department of Geography, Michigan State University IV.6 SPATIAL PATTERNS OF SPECIES DIVERSITY IN TERRESTRIAL ENVIRONMENTS

Kevin McCann, Department of Zoology, University of Guelph III.7 THE STRUCTURE AND STABILITY OF FOOD WEBS

Evelyn H. Merrill, Department of Biological Sciences, University of Alberta VII.3 WILDLIFE MANAGEMENT

Clark A. Miller, Consortium for Science, Policy, and Outcomes and Department of Political Science, Arizona State University VII.11 ASSESSMENTS: LINKING ECOLOGY TO POLICY

Chad Monfreda, Center for Sustainability and the Global Environment, University of Wisconsin Madison VII.7 AGRICULTURE, LAND USE, AND THE TRANSFORMATION OF PLANET EARTH

Paul R. Moorcroft, Department of Organismic and Evolutionary Biology, Harvard University IV.4 BIODIVERSITY PATTERNS IN MANAGED AND NATURAL LANDSCAPES

Rebecca J. Morris, Department of Zoology, University of Oxford II.10 INTERACTIONS BETWEEN PLANTS AND HERBIVORES

William F. Morris, Department of Biology, Duke University I.10 LIFE HISTORY

William Murdoch, Department of Ecology, Evolution, and Marine Biology, University of California, Santa Barbara VII.1 BIOLOGICAL CONTROL: THEORY AND PRACTICE

Shahid Naeem, Department of Ecology, Evolution, and Environmental Biology, Columbia University VI.2 BIODIVERSITY, ECOSYSTEM FUNCTIONING, AND ECOSYSTEM SERVICES

Jon Norberg, Department of Systems Ecology, Stockholm University VI.3 BEYOND BIODIVERSITY: OTHER ASPECTS OF ECOLOGICAL ORGANIZATION



Patrik Nosil, Department of Zoology, University of British Columbia I.18 ECOLOGICAL SPECIATION: NATURAL SELECTION AND THE FORMATION OF NEW SPECIES

Sarah H. Olsen, Center for Sustainability and the Global Environment, University of Wisconsin Madison VII.6 MANAGING INFECTIOUS DISEASES

Megan O’Rourke, Department of Ecology and Evolutionary Biology, Cornell University VI.4 HUMAN-DOMINATED SYSTEMS: AGROECOSYSTEMS

Elinor Ostrom, Department of Political Science, Indiana University VII.10 GOVERNANCE AND INSTITUTIONS

Guayana I. Pa´ez Acosta, Department of Global Ecology, Carnegie Institution for Science VI.5 FORESTS

Margaret A. Palmer, Department of Entomology, University of Maryland VI.8 PROVISIONING SERVICES: A FOCUS ON FRESH WATER

Jonathan A. Patz, Center for Sustainability and the Global Environment and Department of Population Health Sciences, University of Wisconsin Madison VII.6 MANAGING INFECTIOUS DISEASES VII.7 AGRICULTURE, LAND USE, AND THE TRANSFORMATION OF PLANET EARTH

Daniel Pauly, Fisheries Center, Aquatic Ecosystems Research Laboratory, University of British Columbia IV.10 SPATIAL DYNAMICS OF MARINE FISHERIES

Oliver R. W. Pergams, Department of Biological Sciences, University of Illinois at Chicago VI.10 SUPPORT SERVICES: A FOCUS ON GENETIC DIVERSITY

Nicholas Perrin, Department of Ecology and Evolution, University of Lausanne I.6 DISPERSAL

Charles Perrings, School of Life Sciences, Arizona State University VI.11 THE ECONOMICS OF ECOSYSTEM SERVICES

Debra P. C. Peters, USDA Agricultural Research Service, Jornada Experimental Range, New Mexico State University IV.5 BOUNDARY DYNAMICS IN LANDSCAPES This research was supported by National Science Foundation support to the Sevilleta Long-term Ecological Research Program at the University of New Mexico (DEB 0620482).

David W. Pfenning, Department of Biology, University of North Carolina at Chapel Hill I.14 PHENOTYPIC SELECTION

Alison G. Power, Department of Ecology and Evolutionary Biology, Cornell University VI.4 HUMAN-DOMINATED SYSTEMS: AGROECOSYSTEMS

Robert L. Pressey, School of Biological Sciences, University of Queensland V.4 BUILDING AND IMPLEMENTING SYSTEMS OF CONSERVATION AREAS

Navin Ramankutty, Department of Geography and Earth System Science Program, McGill University VII.7 AGRICULTURE, LAND USE, AND THE TRANSFORMATION OF PLANET EARTH

Mark Rees, Department of Animal and Plant Sciences, University of Sheffield II.1 AGE-STRUCTURED AND STAGE-STRUCTURED POPULATION DYNAMICS

David C. Richardson, Marine Estuarine Environmental Sciences Program, University of Maryland VI.8 PROVISIONING SERVICES: A FOCUS ON FRESH WATER

Jon Paul Rodrı´guez, Center for Ecology, Venezuelan Institute for Scientific Investigations (Instituto Venezolano de Investigaciones Cientı´ficas IVIC) VI.13 CONSERVATION OF ECOSYSTEM SERVICES


Osvaldo E. Sala, Department of Ecology and Evolutionary Biology, Center for Environmental Studies, and Environmental Change Initiative, Brown University VI.6 GRASSLANDS

Marten Scheffer, Aquatic Ecology and Water Management Group, Wageningen University III.17 ALTERNATIVE STABLE STATES AND REGIME SHIFTS IN ECOSYSTEMS

D. W. Schindler, Department of Ecology, University of Alberta VII.5 MANAGING NUTRIENT MOBILIZATION AND EUTROPHICATION


Thomas W. Schoener, Section of Evolution and Ecology, College of Biological Sciences, University of California, Davis I.1 ECOLOGICAL NICHE

R. J. Scholes, CSIR Division of Water, Environment, and Forest Technology, South Africa VI.1 ECOSYSTEM SERVICES: ISSUES OF SCALE AND TRADEOFFS

Anthony R. E. Sinclair, Department of Zoology, University of British Columbia VII.3 WILDLIFE MANAGEMENT

Navjot S. Sodhi, Department of Biological Sciences, National University of Singapore V.1 CAUSES AND CONSEQUENCES OF SPECIES EXTINCTIONS

Luis A. Solo´rzano, Andes Amazon Initiative, Gordon and Betty Moore Foundation VI.5 FORESTS

Judy Stamps, Section of Evolution and Ecology, College of Biological Sciences, University of California, Davis I.5 HABITAT SELECTION

R. W. Sterner, Department of Ecology, Evolution, and Behavior, University of Minnesota III.15 ECOLOGICAL STOICHIOMETRY

Stephanie A. Stuart, Department of Integrative Biology, University of California, Berkeley I.3 PHYSIOLOGICAL ECOLOGY: PLANTS

Contributors John N. Thompson, Department of Ecology and Evolutionary Biology, University of California, Santa Cruz II.13 COEVOLUTION

David Tilman, Department of Ecology, Evolution, and Behavior, University of Minnesota II.5 COMPETITION AND COEXISTENCE IN PLANT COMMUNITIES

David J. Tongway, Fenner School of Environment and Society, Australian National University IV.1 LANDSCAPE DYNAMICS

Joseph Travis, Department of Biological Science, Florida State University I.9 PHENOTYPIC PLASTICITY

Will R. Turner, Center for Applied Diversity Science, Conservation International V.4 BUILDING AND IMPLEMENTING SYSTEMS OF CONSERVATION AREAS

Peter M. Vitousek, Department of Biological Sciences, Stanford University III.10 NUTRIENT CYCLING AND BIOGEOCHEMISTRY

Brian Walker, CSIRO Sustainable Ecosystems IV LANDSCAPES AND THE BIOSPHERE

Reg Watson, Fisheries Centre, University of British Columbia IV.10 SPATIAL DYNAMICS OF MARINE FISHERIES


Martin Wikelski, Department of Ecology and Evolutionary Biology, Princeton University I.2 PHYSIOLOGICAL ECOLOGY: ANIMALS

Andy Wilby, Department of Biological Sciences, Lancaster University III.14 BIODIVERSITY AND ECOSYSTEM FUNCTIONING

David S. Wilcove, Department of Ecology and Evolutionary Biology, Princeton University V CONSERVATION BIOLOGY

F. I. Woodward, Department of Animal and Plant Sciences, University of Sheffield IV.7 BIOSPHERE–ATMOSPHERE INTERACTIONS IN LANDSCAPES

Jianguo Wu, School of Life Sciences, Arizona State University IV.3 ECOLOGICAL DYNAMICS IN FRAGMENTED LANDSCAPES


Erika Zavaleta, Environmental Studies Department, University of California, Santa Cruz III.18 RESPONSES OF COMMUNITIES AND ECOSYSTEMS TO GLOBAL CHANGES

I Autecology Jonathan B. Losos Autecology refers to how a single species interacts with the environment; its counterpart is synecology, which refers to how multiple species interact with each other. This latter term is mostly congruent with the field of community ecology, the subject of part III of this volume. Integral to any discussion of autecology is the concept of the niche. This concept has a long and checkered history in the field of ecology, and the term itself has taken on different meanings through time (chapter I.1). In the most general sense, however, we may think of the niche of a population as the way members of that population interact with their environment, both biotic and abiotic. In other words, the term ‘‘niche’’ refers to where organisms live and what they do there. The first step in considering how organisms interact with their environment is investigating how the specific phenotypic characteristics of members of a population allow them to exist in a particular environment. The environment poses a wide variety of challenges to organisms: for example, they must be able to obtain and retain enough water, withstand high or low temperatures, and obtain enough nutrients to survive. More than a century of research has revealed that species, and even populations of species, are often finely tuned to the specific conditions in the environment in which they live. In recent years, increasingly sophisticated approaches and instrumentation have allowed an exquisitely detailed understanding of the physiological basis of organismal function (chapters I.2–I.4). Animals—and, in some sense, fast-growing plants— also can influence the way they interact with their environment through behavioral means. For example, animals can choose the habitat in which they occur and thus can determine, to some extent, the environment they experience throughout their lives (chapter I.5). Many organisms move from their birth site at a particular stage in life; although for plants and some animals, dispersal is passive, other species actively choose where to settle (chapter I.6). Behavior, of course, is a key component of how most animals interact with their environment. Almost

all aspects of the natural history of animals have a behavior component. In part I, we consider foraging (chapter I.7) and social behavior (chapter I.8). Other topics are included in parts II and VI of this volume. Most plants have relatively little ability to determine the environmental conditions they experience. But plants often have another option available—they frequently exhibit substantial phenotypic plasticity, which allows a plant to alter its phenotype in an advantageous way to be better suited to its environment. Scientists have long appreciated this ability in plants, and zoologists have come to realize relatively recently that many animal species exhibit adaptive phenotypic plasticity as well (chapter I.9). Organisms adapt in yet another way, by molding their life cycle—what is termed ‘‘life history’’—to the particular environment in which they live (chapter I.10). Thus, species in environments in which resources are abundant and threats are common may have short generation times and early reproduction. Conversely, in environments in which resources are more scarce but threats are not as severe, a more successful strategy may be to defer reproduction and to invest in becoming better competitors for resources, delaying reproduction and ultimately producing fewer, but better provisioned, offspring. No species occurs everywhere in the world. The behavior and physiological capabilities of a species determine where a species can and cannot occur. In the last few years, advances in remote sensing technology have provided the capability to visualize the distribution of environmental conditions with great precision over large spatial scales (chapter I.11). Combined with records of species occurrences and, ideally, an understanding of species’ physiological capabilities, these geographic information systems approaches have opened new vistas for understanding how and why species occur where they do; these approaches are also of great importance in predicting how species will respond to rapidly changing environmental conditions



(see parts IV and V). Of course, the distribution of a species is not only a function of its physiological capabilities and other aspects of its ecology. Rather, Earth geography and history also are important—a species cannot occupy an area that it has never had the opportunity to colonize. Consequently, biological and historical factors combine to determine the geographic range of any species (chapter I.12). Integral to an understanding of how organisms interact with their environment is the concept of adaptation, the idea that natural selection has molded the characteristics of populations so that they are well suited to the particular circ*mstances in their environment (chapter I.13). Of course, this is not to say that organisms are optimally adapted to their current conditions, nor that every feature exhibited by a population represents an adaptation for some aspect of the environment. Quite the contrary, natural selection is only one of many processes that affect how populations evolve (chapters I.14 and I.15); in some circ*mstances, processes other than natural selection will predominate, leading populations to be less well adapted to their current circ*mstances. Ecologists are increasingly interested in the evolutionary time scale. On one hand, it has become clear that, in many cases, we can understand the current state of species and of entire communities only by considering their history. Species are not blank slates, to be molded by selection to the optimum configuration for their environment; rather, they have a historical starting point, and selection can work to modify species

only from this point (chapter I.13). Similarly, communities, too, have histories—the current state of a community is a result of which species have managed to get to a given locality and how those species interact once there. Methods to incorporate evolutionary information, in the form of phylogenies (or evolutionary trees), are now widely utilized and becoming increasingly sophisticated (chapter I.16). Conversely, evolutionary biologists have clearly demonstrated over the last several decades that evolutionary change can occur very rapidly (chapter I.17). Consequently, ecologists ignore evolution at their own peril—populations can adapt quickly enough that evolution can have effects even on ecological time scales. Evolution is important in another respect. The components of ecological interactions are species. The study of speciation—how new species arise—has long been the province of evolutionary biologists, but in recent years it has become clear that ecology may play an important role in affecting rates of speciation. In particular, the concept of ecological speciation—the idea that speciation is intimately tied to ecological divergence—has gathered great support (chapter I.18). Hence, in this respect as well, ecological and evolutionary perspectives are strongly intertwined. Finally, over larger time scales, certain groups of organisms diversify greatly, producing not only a large number of species but also occupying a great variety of ecological niches. Some scientists consider this phenomenon, known as adaptive radiation, to be responsible for the majority of life’s diversity (chapter I.19).

I.1 Ecological Niche Thomas W. Schoener OUTLINE

1. Three concepts of the ecological niche 2. The recess/role niche and seeking ecological equivalents 3. The population-persistence niche and mechanistically representing competition 4. The resource-utilization niche and understanding the evolution of species differences 5. Environmental niche modeling and analyzing niches on a macroscale 6. Conclusion It may come as something of a surprise that ecological niche, a term so common in the popular media, has three distinct meanings among scientists, each with an associated conceptual basis: these are the recess/role niche, the population-persistence niche, and the resource-utilization niche.

GLOSSARY character displacement. The situation in which two

species are more different in geographic locations where they overlap than between locations where they occur alone community. Those species populations occurring at some location competition. Ecological interaction in which two or more species negatively affect one another by consuming common resources or by other harmful means convergence. Development of increasing similarity over time, usually applied to species somewhat unrelated evolutionarily niche dimension. Environmental variable along which a species’ niche is characterized, e.g., food size, and typically represented as the axis of a graph polymorphism. The existence of two or more forms, differing in morphology or some other way, in the same population

population. Those individuals of a species occurring at

some location population growth rate r. The per capita rate at which

a population changes size, typically computed as the birthrate minus the death rate 1. THREE CONCEPTS OF THE ECOLOGICAL NICHE The Recess/Role Niche

The first use of ‘‘ecological niche’’ appeared in a report on ladybugs written by R. H. Johnson nearly a century ago, although the term was used shortly thereafter by the zoologist Joseph Grinnell, who is generally given credit for its original development. The meaning was very close to figurative usage: the ecological niche of a species is its ‘‘role,’’ ‘‘place,’’ or more literally ‘‘recess’’ (in the sense of a ‘‘nook’’ or ‘‘cubbyhole’’) in an ecological community. Thus, the California thrasher, one of Grinnell’s major examples, is a bird of the chaparral community that feeds mostly on the ground by working over the surface litter and eating both animal and plant items of a suitable size. Escape from predators is similarly terrestrial, with the well-camouflaged bird shuffling off through the underbrush on the rare occasions when it is threatened. The idea that there exists a set of characteristic habitat and food types with accompanying behavioral, morphological, and physiological adaptations leads to the notion of ecological equivalents. These are defined as two or more species with very similar niche characteristics that occur in completely different localities. An example from Grinnell’s writings is the kangaroo rat of North America, which ‘‘corresponds exactly’’ to the jerboa (another desert rodent species) of the Sahara. The existence of ecological equivalents would imply that rather invariant rules determine the niches available for occupancy in a particular kind of environment, e.g., a desert. Moreover, niches can be empty in the sense that a suitable species does not occur within a


locality, perhaps because it never got there or was unable to evolve in situ. But to what extent do ecological equivalents really exist? Decades after Grinnell’s work, we now know (section 2, below) that although some examples can certainly be found, perhaps more commonly, species of similar environments (e.g., deserts) among distant localities are neither identical nor often even similar. Perhaps such considerations helped to engender the two other meanings of ecological niche, each with its accompanying set of ideas about how the ecological world works. The Population-Persistence Niche

The population-persistence niche has its roots in papers written in the mid-twentieth century by the ecologist and limnologist G. E. Hutchinson. This concept focuses on the species, in this case its population, rather than on the environment. Hutchinson formulates the ecological niche as a quantitative description of the range of environmental conditions that allow a population to persist in some location; the term persist means having a positive or at least zero (break-even) population growth rate, r (if r is negative, the population dwindles away to extinction). An example of an environmental condition is temperature; a second example would be humidity (for organisms on land) or salinity (for organisms in water). If we represent an environmental condition by the axis of a graph, a range is an interval along that axis, e.g., temperature from 08C to 308C (figure 1). A second interval, say for relative humidity, might range from 20% to 80% along the humidity axis. We can have as many different environmental axes as necessary to characterize the population growth rate. If r for a given axis is uncorrelated with the values of variables of the other axis (e.g., if the range of temperatures allowing r 0 is the same for any value of humidity), then the niche is rectangular (as in figure 1); otherwise it will have other shapes. Hutchinson labeled his concept the cumbersome ‘‘ndimensional hypervolume’’ (imagine three or more environmental axes). The more succinctly labeled fundamental niche is that portion of niche space where the species population can persist. The fundamental niche is visualized as being in the absence of other species that compete with the given species for resources and thereby affect its persistence. To account for this latter circ*mstance, Hutchinson defined the ‘‘realized niche’’ as that portion of the fundamental niche not overlapping the fundamental niches of competing species, plus that portion overlapping the competing species’ niches where the given species can still persist (have r 0).

Relative humidity


100 90 80 70 60 50 40 30 20 10 0

Niche (r >0)


10 20 Temperature



Figure 1. Example of Hutchinson’s population persistence niche. Rectangle encloses the ranges of temperature and humidity in which the species’ population can persist (where r 0).

Hutchinson’s concept is important for several reasons. First, it provides a precise, quantitative way to characterize the ecological niche. Second, it focuses on what the species itself does rather than on the opportunity for a species to exist or not in a community (the latter being the ‘‘recess’’ concept of Grinnell). Thus, ecological equivalents are not necessarily expected and, if they do not occur, are not troubling to the concept: for Hutchinson, there are no ‘‘empty niches.’’ Such a precise formulation of the niche is not without its drawbacks, however. Chief among them perhaps is the difficulty of finding out what the population-persistence niche of a species actually is in nature. Presumably, for each point of the n-dimensional hypervolume—say for each value of temperature and humidity—one needs to culture populations or otherwise determine their population growth rate r; and one repeats this for different points until one has all combinations of temperature and humidity for which the population can persist. The difficulty of so doing for all but microorganisms (at best) is easy to imagine. A second problem is that certain niche characteristics as conceptualized by Grinnell are not easily ordered along an environmental axis. An example is food size: at any given real location, food comes in a variety of sizes (rather than there being one food size for each location). Of course, one can use average food size, but such a concept is not as plausible as using average temperature because animals come across a variety of food sizes on a daily basis. Animals of a particular body size (and therefore a particular size of feeding apparatus, e.g., mouth) have limitations on the extreme values of food size that can be consumed: items too large cannot be swallowed, and those too small cannot be handled deftly (or eaten in an energetically profitable way). Hence, a more detailed

Ecological Niche

Although this model represents a vast improvement in the concept of population-persistence niche, the operational difficulty of measurement still exists: determining the niche for figure 2 (Chase and Leibold) is not much easier than for figure 1 (Hutchinson).


Predator A

Resource A


The Resource-Utilization Niche

Resource B

Predator B D.

Predator (P)

Stress intensity (S)


Resource (R)


Resource (R)

Figure 2. Chase Leibold concept of the population persistence niche. Each panel has two regions, a shaded region where r 0 and an unshaded region where r < 0. The niche is the shaded region. (A) A species with two substitutable resources (the axes measure re source density); (B) a species with two predators; (C) a species with a predator and a resource; (D) a species with a stress and a re source. (Figure courtesy of J. M. Chase.)

description than the average food size available at a location is desirable. Third, Hutchinson’s niche is onesided in the sense that it assumes a rather passive species that does not affect other species in the community in a way that eventually feeds back onto the given species. Fourth, Hutchinson focuses almost exclusively on one type of ecological interaction, competition between species; for example, his distinction between the fundamental and the realized niche. In this way, his concept was not as inclusive as that of Grinnell. In part as a reaction to the latter two drawbacks, Jonathan Chase and Mathew Leibold have substantially extended the population-persistence niche. In a recent but already very influential book, they define the niche as a joint specification of environmental conditions or variables that allow a species to have r 0 along with the effects of that species on those environmental variables. Niche axes are quite broadly construed and can include a variety of factors that impact populations (and vice versa); examples include amount of a given resource, abundance of a given predator, and degree of a physical stress such as wind speed (figure 2). Thus, one can incorporate effects of species on environmental conditions, and one can specify a given region of niche space where a species has r 0 (figure 2).

An eminently operational concept of the ecological niche, formulated by two evolutionary ecologists, Robert MacArthur and Richard Levins, is the resourceutilization niche, our third meaning. Like the population-persistence niche, the resource-utilization niche is quantitative and multidimensional, but it focuses entirely on what members of a species population in some locality actually do—in particular, how they use resources. The relative use (¼ utilization) of resources along a given niche axis can be described as a frequency distribution or histogram. Take, for example, the axis food size. We can (figure 3, top) draw a histogram showing the fraction of food of different sizes consumed by all members combined of a given population; e.g., the fraction of the total population’s foods between 5 and 6 mm. If we have a second dimension, say feeding height, we can graph the fraction of food items eaten at different heights in the vegetation. The two can be combined as a joint distribution or threedimensional histogram (figure 3, bottom), and this can be further generalized (although not easily graphed) for as many dimensions as ecologists find important to describe the population’s resource use. A broad classification of the kinds of niche axes used for utilizations consists of habitat, food type, and time. Within habitat, microhabitat and macrohabitat are distinguished, whereby microhabitat has a smaller spatial scale (e.g., height in vegetation) than does macrohabitat (e.g., vegetation zone such as tropical rainforest or desert). Within food type, food size and hardness can be distinguished. Within time, daily and seasonal activity can be distinguished. The resource-utilization niche immediately frees us from the problem with Hutchinson’s formulation that certain environmental variables cannot be meaningfully described using only the average. Indeed, the resource-utilization niche is nothing more than a precisely formulated description of the natural history of a species: its habitat, food types, and activity times, among other things. Such natural history can include nonfeeding habitats and activity times for behaviors such as predator escape and mating, all characterizable on its niche axes. Thus, we have a niche concept that precisely encapsulates what ecologists measure anyway. Indeed, Grinnell, the originator of the recess/role niche concept, measured such things in his study



unlike both the recess/role niche of Grinnell and the population-persistence niche of Chase and Leibold. We now review seriatim the kinds of research engendered by the three concepts of the niche as well as a very recent research trend called ecological niche modeling that includes elements of all three.


Frequency of use (P,h)


1 2 3 4 5 6 7 8 9 10111213

1 2 3 4 5 6 7 8 9 10111213

Prey-size category

Prey-size category



g in ed Fe

ht ig e h

Pr ey s


Figure 3. An example of the resource utilization niche. (A) A one dimensional niche, where the dimension is prey size. Numbers give prey size categories, indexed by h; (B) the same utilization smoothed; (C) utilization of two resource dimensions, prey size and feeding height. (Redrawn from Schoener, 1986.)

organisms but with the assumption that, in so doing, he was discovering something about the availability of niches in the community—an availability or opportunity to which the species more or less had to conform. The resource-utilization niche, in contrast, assumes nothing about rigidly determined niche recesses in a community, nor about the necessity of ecological equivalents, nor about the existence of empty niches. The resource-utilization niche was formulated a decade or so after the population-persistence niche but, unlike the latter, has remained rather unchanged up to the present. This is despite the fact that, by emphasizing resources, it is seldom extended beyond resource use,


In an early study of grassland birds inhabiting far-flung locations—Kansas, Chile, and California—Cody found that each community contained about the same number of species and the same ecological types: three or four passerines (small ‘‘perching’’ birds), a larger vegetarian ‘‘grouse-like’’ species, both a long- and a shortbilled wader, and two or three raptors. Twenty pairs of ecological equivalents were identified between the two Mediterranean systems: Chile and California. However, later studies by Cody in other Mediterranean systems including Sardinia and South Africa showed a weaker pattern, especially for the latter, whose floras were very different. In contrast to birds, plants in Chilean and Californian systems showed little convergence at the community level; for example, woody vegetation in Chile comprises less of the total cover but more total species and has a greater diversity of height layers than in California. Nonetheless, the major growth forms (e.g., broad-leaved evergreen, broad-leaved deciduous) are similar, even with regard to number of species, although several forms present in Chile (e.g., spinosestemmed shrubs) are absent from California—an apparent empty niche. Major resemblances between plant growth forms among plants with very different evolutionary lineages occur rather commonly among plants; a striking example is given by American cacti and African euphorbs. Perhaps the least evidence for ecological equivalents after systematic search is among colubrid snakes of North, Central, and South America. Cadle and Greene find few ecological equivalents (and little evidence for community similarity); instead, a number of types (fossorial earthworm eaters, nocturnal arboreal lizard/frog eaters) in some communities are conspicuously absent in others. Probably the most extensive work on convergence and ecological equivalents has been done on lizards. An initial study by Fuentes, again comparing Chile and California, found convergences in community characteristics as well as in individual niche traits— microhabitat, daily activity time, and food type. In a second major study, Pianka found less evidence for

Ecological Niche

Figure 4. An example of ecological equivalents: the horned toad (Phrynosoma platyrhinos) of North American deserts and the thorny devil (Moloch horridus) of Australian deserts. (From Pianka, E. R. 2000. Evolutionary Ecology. San Francisco: Harper & Row. Used by permission of Pearson Education, Inc.)

similarity in community characteristics than difference among lizards of the three warm-desert systems of North America, Australia, and Africa. Nonetheless, striking ecological equivalents sometimes exist, such as the amazing resemblance between the horned toad of North America and the thorny devil of Australia (figure 4). Examples of ecological equivalents are most impressive when the species from widely different localities are relatively unrelated in terms of evolutionary descent: convergent evolution toward the same morphology and behavior would seem to support the idea of the niche as a functional optimum characteristic of particular types of communities (e.g., those in deserts) into which species repeatedly evolve. Nonetheless, a plausible hypothesis for lack of convergence is that major evolutionary stocks are so different that evolution is too constrained to produce much convergence. Melville, Harmon, and Losos recently examined two lizard families, the Iguanidae and Agamidae, of North America and Australia, respectively, which are closely enough related to belong to the same clade (Iguania) even though they have been geographically separate for as long as 150 million years. Using an approach that takes into account evolutionary relatedness, they found convergence in habitat use and locomotor morphology, including pairs of ecological equivalents, between the two deserts. Another example of convergence among relatively closely related species is provided by the Anolis lizards of large West Indian islands: Cuba, Hispaniola, Jamaica, and Puerto Rico. Here, various ecomorphs— species occupying the same microhabitat—have independently evolved on the separate islands. Harmon, Kolbe, Cheverud, and Losos found that five functionally distinct morphological characters—body size, body shape, head shape, lamella (ridges on toes) number, and sexual size dimorphism—converge among the different islands as a function of habitat similarity. For example, lizards living on the ground and low trunks are more similar between Cuba and Hispaniola than either is to other ecomorphs (e.g., those living in tree


crowns) co-occurring on the same island and to which they are more closely related. A final recently discovered example of convergence occurs in a completely different group: orb-weaving spiders of the genus Tetragnatha of the Hawaiian islands. Blackledge and Gillespie found that spiders inhabiting different islands constructed remarkably similar webs. These convergences toward ecological equivalency, which they called ‘‘ethotypes’’ (ethology is the study of behavior, and this emphasizes the behavioral similarity), occurred independently in evolution. Like the Australian Iguania discussed above, the group as a whole consists of relatively closely related species. In conclusion, although the evidence for ecological equivalents is certainly mixed, more and more examples are coming to light that make Grinnell’s rather old concept seem alive if not completely well. As Schluter has suggested, to the extent that ecological equivalents exist and are independently evolved, morphology, physiology, and behavior must constrain the efficiencies with which resources and other factors characteristic of particular kinds of ecosystems (e.g., deserts) can be dealt with—ecological equivalents mark peaks in the adaptive landscape. 3. THE POPULATION-PERSISTENCE NICHE AND MECHANISTICALLY REPRESENTING COMPETITION

Maguire in 1973 may have been first to plot population growth rate r for real species as a function of niche dimensions and to make predictions about the competitive outcomes among them. In the l950s, Birch had studied several species of beetle infesting stored grain in Australia; figure 5 shows Maguire’s plot of Birch’s data with respect to temperature and moisture. Isoclines of positive values of r down to zero (no population growth) show different patterns for the two species, such that Calandra oryzae has a higher r for lower temperatures and somewhat greater moistures than Rhizopertha dominica. The dashed line in figure 5 separates regions of niche space where one versus the other species has the higher r. Assuming no complications, an environment on one or the other side of the line will favor one or the other species of beetle in competition. To illustrate their ideas about the populationpersistence niche, Chase and Leibold replot data of Tilman for two species of diatoms, Asterionella and Cyclotella (figure 6). The situation is somewhat more complex than that shown in figure 2 because resources are not substitutable (which would mean that the populations can survive on either resource alone or on some combination) but rather are essential: figure 6A shows the general case, where a species must have a

Autecology .1 .005

.2 .0


.140 7










.605 .4




.7 .6 .6


.7.3 .501



0 .7 .6 .5 .4 .2 0 -









7 6 8



.3 5








.001 -

















12 2
















.105 3








4 .013















8 15







Temperature Figure 5. Maguire’s illustration of the population persistence niche for the dimensions temperature (C) and percentage moisture of wheat, using the beetles Calandra oryzae (small strain) and Rhizo pertha dominica. Numbers in squares indicate average monthly

conditions in Bourke, New South Wales, Australia; numbers in circles give same for Adelaide, South Australia. (Redrawn from Maguire, B., Jr. 1973. Niche response structure and the analytical potentials of its relationship to the habitat. American Naturalist 107: 213 246)

minimal amount of each resource in order that r 0. For two such species, coexistence is possible if each species can just survive (r ¼ 0) for a different one of the two resources. In Tilman’s experiment, the resources are the nutrients silicate (SiO2) and phosphate (PO4), and the levels of each can be controlled in the laboratory. Asterionella is a specialist on SiO2, and Cyclotella on PO4. From the individual species growth curves on the separate resources, one can predict regions of niche space (plots of SiO2 versus PO4 concentration) where each species has a lower r ¼ 0 and so is limited by a different resource. In that region (figure 6B), the species can coexist. Outside that region, one or the other species wins, depending on which resource is more abundant. Such empirical studies are impressively successful in the highly controlled setting of the laboratory, but they are very difficult indeed to perform in the field. Chase and Liebold could find only one such field study, again by Tilman (and Wedin), in which several plant species

vary in their ability to utilize nitrogen from the soil. These relative abilities were used rather successfully to predict competitive outcomes along a natural nitrogen gradient. Probably, practical difficulties largely explain why the population-persistence niche is a concept with mostly theoretical development. It seems most likely that it will be easiest to apply to organisms with the size and behavior that enable their populations to persist in small spatial units (sometimes called microcosms). 4. THE RESOURCE-UTILIZATION NICHE AND UNDERSTANDING THE EVOLUTION OF SPECIES DIFFERENCES

How similar can species be and still coexist? An answer was obtained in the last section for species having a small number of ecological requirements or resource types. What if species fed on a wide variety of resources, such as foods of different sizes found at different

Ecological Niche A.


B. clo ella Cyclotella Wins ins

4 PO4 (M)

Resource A


Stable Coexistence

3 2 1

Asterionella Wins 0 Resource B



60 SiO2 (M)



Figure 6. (A) Chase Leibold population persistence niche (shaded area) for essential resources. (B) Ranges of coexistence and ex clusion for two species of diatoms competing for two essential resources in chemostats. Circles give nutrient ratio treatments where the two species coexist; stars give treatments where

Asterionella excludes Cyclotella; diamonds give treatments where Cyclotella excludes Asterionella. (Redrawn from Chase and Lei bold, 2003; their figure 4.1 in turn redrawn from Tilman, D. 1977. Resource competition between planktonic algae: An experimental and theoretical approach. Ecology 58: 338 348)

vegetation heights and preferring different temperatures? This situation applies to predators, such as Grinnell’s California thrasher, that eat a great variety of insects and other arthropods that in turn have their own populations with their own niche characteristics. The 1967 paper in which MacArthur and Levins promoted the resource-utilization niche has as its main objective the understanding of how similar competing species can be and yet still coexist. It is sometimes said that species cannot coexist if they occupy the same niche, but the theory of MacArthur and Levins also posits that if the niches of the species are too similar (too much niche overlap), they still cannot coexist. To illustrate, imagine two species with the one-dimensional niche in figure 7; this dimension might be food size, and one species tends to eat larger food on average than the other. If the niches are too close (figure 7A, left), they are too similar (the niche overlap [shaded area] is too great), and the better competitor will eliminate the other from the community. That degree of closeness at which the species can just coexist (any closer and one is eliminated) is called the limiting similarity (figure 7A, middle); the niches can, of course, be farther apart and still allow coexistence (figure 7A, right). Limiting similarity is measured in units of d/w, where d is the distance between peaks and w is the width of the niche (usually computed as the standard deviation of the utilization distribution; figure 7B). The larger the w, the more generalized the species; a specialist has a thin niche (small w; figure 7C). In MacArthur and Levins’s theory, a d/w slightly larger than 1.0 is the limiting similarity; much subse-

quent work has shown limiting similarity to vary greatly yet be about 1 (certainly to an order of magnitude). Indeed, sometimes real species differ by almost exactly this theoretical value. A sensational example is provided by two mud snails (Hydrobia) studied by Fenchel in Denmark. The snails ingest particles: diatoms and inorganic pebbles covered with minute sessile organisms. About 150 years before the study, a fjord collapsed, and one species invaded the other’s range. The resource-utilization niches of the species displaced away from one another, apparently independently, numerous times, to d=w 1 (figure 8, top left). Corresponding to this niche difference is a difference in body (shell) size such that larger species ingest larger particles (figure 8, top right), and the body sizes of the species had diverged (in a process called character displacement) to a ratio of 1.3–1.5 (figure 8, bottom). Consistent with the theory of Taper and Case (see below), this ratio is higher than the ratio of d’s for the two resource utilizations of 1.2. So far we have represented the resource-utilization niche as a distribution summing together the food-size or other niche characteristics of all individuals in a population. However, individuals may differ in their niche characteristics, sometimes just by chance opportunity (e.g., what they happen to come across to eat), but sometimes because they have different morphologies and behaviors that make them specialized for a certain portion of a niche axis (just as species can be specialized). Figure 9 shows the two extreme possibilities for such component individuals; note that each individual can be a generalist (figure 9, left) or a




No coexistence: species are too similar (overlap is large)

Limiting similarity: (overlap is medium-sized)


More different than limiting similarity (overlap is small)

C. Small w d

Large w w

Figure 7. (A) One dimensional resource utilization niches of two species showing no coexistence because niches are too similar (left), limiting similarity species can just coexist (middle), and

greater than limiting similarity also allowing coexistence (right). (B) Niche distance d and niche width w for two species. (C) Niches of generalist (large w ) and specialist (small w ) species.

specialist (figure 9, middle), in either case producing the same utilization for all individuals combined (figure 9, right). What difference does it make which of the two situations one has? A series of specialist individuals may eventually allow the population as a whole to be more generalized in the absence of competing species, and this ‘‘polymorphism’’ might even lead to speciation (see chapter I.18). Such polymorphism, when measured in terms of those morphological characters corresponding to position on the niche axis (e.g., shell size corresponding to mean food-particle size), was uncommon in the literature at the time of Taper and Case’s paper, and this was consistent with their theoretical model in which the proportion of different kinds of individual niches evolves once the competing species meet geographically. Recently, however, Bolnick, Sva¨nback, Ara´go, and Persson looked at the resourceutilization niches themselves rather than the morphological characters that reflect them. They found that the bigger the w for the total population, the bigger the between-phenotype niche width, measured as the standard deviation of the d’s of the niches of the component individuals. It remains to be seen exactly how these apparently somewhat contradictory trends will be reconciled.

concepts. The method characterizes the macrohabitat niche of a species by quantitatively summarizing geographic-information-system (GIS) information on climatic and similar variables at stations throughout the species’ geographic range. Such macrohabitat niche information is then used to predict the potential geographic range of the subject species. Because of its focus on macrohabitat, the scale is similar to Hutchinson’s version of the population-persistence niche. However, the method specifies the ‘‘empty niches’’ of Grinnell’s recess/role niche as those localities having the niche characteristics of the subject species but where that species does not, in fact, occur. Finally, it allows quantification of niche similarity between species via measures of niche overlap used for the typically finer scale of resource-utilization niches of MacArthur and Levins. One of the most successful applications of ENM so far examines the question of whether the more closely related species are, the more similar are their niches. The question is important because if the answer is yes, evolutionary history must have a major influence in determining niche characteristics relative to the influence of the community in which the species now occurs. A study by Knouft, Losos, Glor, and Kolbe on the 11 species of the Anolis sagrei group in Cuba found no evidence that niches were more similar, the more closely related the species (evolutionary relatedness is assessed using molecular genetics). A second study, by Warren, Glor, and Turelli showed along with the previous study that the most recently diverged species


A recent set of techniques, called environmental niche modeling (ENM), combines elements of all three niche

Ecological Niche



40 100 Median food particle diameter (µm)

20 0

Volume %

60 40 20

Volume %


40 20 0

10 4


Length (mm)





16 32 64 128 256 512 Diameter (µm)




















3 2 1 0

3 4 Shell length (mm)





9 10


Length (mm)

Volume %














70A 70E GI 70B 13

3 2 1 0

Figure 8. (top left) Resource utilization niches for prey size among two species of gastropods (Hydrobia ulvae, gray circles; H. ven trosa, black circles) where the species overlap (top), where H. ulvae is alone (middle), and where H. ventrosa is alone (bottom). (top right) Median diameters of ingested food particles of four species of Hydrobia plotted against shell length. (bottom) Average lengths of

Hydrobia ulvae (gray circles) and H. ventrosa (black circles) from 15 localities where the species co occur (left) and 17 localities where one of the two species occurs alone. All samples from the Limfjord during summer 1974. [Redrawn from Fenchel, T. 1975. Character displacement and coexistence in mud snails (Hydrobiidae). Oeco logia 20: 19 32]

had the greatest climatic-niche differences. The second study, however, gave somewhat more support for the hypothesis in general, in that niche similarity between closely related species of birds, butterflies, and mammals separated by the Isthmus of Panama was greater than expected by chance. However, somewhat contrary to the founding ENM study by Peterson, Soberon, and Sachez, niches were rarely identical, so the

overall answer is in fact mixed, as is so often the case in ecology. 6. CONCLUSION

The research trends discussed in relation to the three niche concepts are summarized as an evolutionary tree in figure 10. In this diagram, the thicker arrows



Figure 9. (left) Resource utilization niche of a species population. (middle) Population of generalist individuals whose niches sum to the curve at top. (right) Population of specialist individuals whose


niches sum to the curve at top. (After Klopfer, Peter H. 1962. Be havioral Aspects of Ecology. Princeton, NJ: Princeton University Press)

Population-persistence and impacts (Chase, Leibold)

Ecological niche modeling (ENMs)

Resource-utilization niche (MacArthur-Levins)




Species coexistence

Population-persistence niche (Hutchinson)



1930 Figure 10. Timeline showing the de velopment of niche concepts and the research programs stemming from them.


indicate a greater influence of one concept or research program on the next. Note that all three niche concepts, despite sometimes rather early beginnings, have stimulated research that is being actively pursued at the present time. FURTHER READING Chase, Jonathan M., and Mathew A. Leibold. 2003. Ecolog ical Niches. Chicago and London: University of Chicago Press. A recent major revision of the population persistence niche concept. Grinnell, Joseph. 1917. The niche relationships of the Cali fornia thrasher. Auk 34: 427 433. One of the founding papers of the recess/role niche concept.

Recess/role niche (Grinnell)

Harmon, Luke J., Jason J. Kolbe, James M. Cheverud, and Jonathan B. Losos. 2005. Convergence and the multi dimensional niche. Evolution 59: 409 421. A very recent study of convergence and ecological equivalents that em phasizes the niche. Hutchinson, G. Evelyn. 1957. Concluding remarks. Cold Spring Harbor Symposia on Quantitative Biology 22: 415 427. The founding paper of the population persistence niche concept. Knouft, Jason H., Jonathan B. Losos, Richard E. Glor, and Jason J. Kolbe. 2006. Phylogenetic analysis of the evolution of the niche in lizards of the Anolis sagrei group. Ecology 87: S29 S38. A recent exemplar of the environmental niche modeling technique. MacArthur, Robert H., and Richard Levins. 1967. The lim iting similarity, convergence and divergence of coexisting

Ecological Niche species. American Naturalist 101: 377 385. The founding paper of the resource utilization niche concept. Mooney, Harold A., ed. 1977. Convergent Evolution and Chile and California. Stroudsburg, PA: Dowden, Hutch inson and Ross. Contains major, detailed papers com paring two climatically similar regions. Schoener, Thomas W. 1986. Resource partitioning. In J. Kikkawa and D. J. Anderson, eds. Community


Ecology Pattern and Process. Oxford: Blackwell Scien tific Publications, 91 126. A review of how the resource utilization niche is used in ecological research. Schoener, Thomas W. 1989. The ecological niche. In J. M. Cherrett, ed. Ecological Concepts: The Contribution of Ecology to an Understanding of the Natural World. London: Blackwell Scientific Publications, 79 113. A de tailed history of the development of niche concepts.

I.2 Physiological Ecology: Animals Martin Wikelski


1. Guiding concept: Trade-offs 2. Guiding concept: Performance as integrative measure of individual fitness 3. Process I: Acquisition of environmental information 4. Process II: Internal communication and regulation of physiological function 5. Process III: Energy expenditure as one central hub for trade-offs 6. Process IV: Key innovations 7. Process V: Self-defense: Immunoecology 8. Application: Conservation physiology 9. Future challenges Physiological ecologists study how animals live and function within environments that are constantly changing. Key guiding concepts in physiological ecology are that (1) individual animals are subject to trade-offs such that all (physiological) actions cannot be performed maximally at the same time. Trade-offs underlie the fact that ‘‘a jack of all physiological trades is a master of none,’’ which in turn is the basis of the generalist-specialist continuum that brings about much of the niche differentiation in ecology. (2) A second guiding concept is that whole-organism performance provides an integrative measure of individual success in life. Quantifying individual performance allows physiological ecologists to assess the integration of traits within an organism and to determine how natural selection orchestrates not just one but all characteristics of an organism at the same time. Whereas in the past, physiological ecologists have also often studied animals in laboratory situations, technological advances now allow researchers to ‘‘go wild’’ and address individual physiological functions in the very environment where such functions have evolved. The importance of studying animal function in the wild cannot be overestimated because many organismal tradeoffs are expressed only when food is scarce or predators are abundant.

GLOSSARY constraints. These can absolutely limit certain actions

of an organism. Even if all efforts in a trade-off scenario are devoted toward a particular action, this action is not sufficient to satisfy an organism’s current needs. energy. In biology, energy, which is essential for life, is gathered from the breaking of chemical bonds during metabolic processes. Energy is often stored by cells in the form of substances such as carbohydrate molecules (including sugars) and lipids, which release energy when reacting with oxygen. hormones. These substances are chemical messengers that carry information from one part of the organism (e.g., the brain) to another (e.g., the gonads) often via the blood transport system. Hormones bind to receptors on target cells and thus regulate the function of their targets. Various factors influence the effects of a hormone, including its pattern of secretion, transport processes, the response of the receiving tissue, and the speed with which the hormone is degraded. metabolic rate. Energy expenditure per unit time. Metabolic rate is normally expressed in terms of rate of heat production (kilojoules per time). performance. This refers to whole-organism performance capabilities (e.g., how fast an organism can sprint) that are determined by physiological traits (e.g., composition of muscle fibers). trade-offs. These attributes refer to the loss of one quality or aspect of something in return for gaining another quality or aspect. Physiological ecology occupies a central role in the biological sciences and has a long tradition of integrating other biological disciplines. Physiological systems provide the interface between genomics at the lowest mechanistic level to organismal life history and evolution at the highest level of biological integration.

Physiological Ecology: Animals Every biological process linking genes to behavior will ultimately have to be understood mechanistically on the physiological level to truly provide a picture of how organisms function. There are many levels at which physiological ecologists attempt to discern how organisms work. On the lowest level, physiological ecology meets genomics and proteomics. For example, Chi-Hing Chris Cheng and Art DeVries from the University of Illinois, working on the antifreeze protein in Antarctic fish, discovered that the protein is coded by a simple but frequent DNA repeat derived from a snippet in a trypsinogen-like protein gene, initially presumably by chance. This protein appeared to have just the right structure to recognize the surface structure of ice crystals that enter into the blood of the fish. Working up the physiological levels, because ice that enters into the fish’s circulation always end up in the spleen, Cheng and DeVries hypothesized that the immune system, perhaps macrophages, of these fish living at subfreezing temperatures would take care of the nascent ice crystals encapsulated or presented by the antifreeze protein. Perhaps not unlike a pathogen, the immune system then either ‘‘kills’’ or lyses or excretes the nasty foreign body—a spiny ice crystal that would otherwise serve as a crystallization hotbed for more ice. What followed showed the true heuristic power of the physiological ecology approach. When Cheng and DeVries compared different antifreeze proteins among unrelated species of Antarctic and Arctic fish, they found that all of them use the same mechanism to deal with nascent ice in their blood and body fluids. It turned out that most fish can survive within the subfreezing, icy polar waters only if they have enough ‘‘antifreeze’’ in their circulation. Thus, Cheng and DeVries were able to integrate from a simple physiological innovation to explain a major ecological question: why there exist almost exclusively notothenioid, antifreeze fish around the Antarctic continent. Moreover, Cheng recently discovered that an unrelated innovation provides Arctic cod fishes with a near-identical antifreeze protein as the Antarctic notothenoids to brave the cold in the North. However, organismal innovations rarely if ever come without a cost. It is not entirely clear what the cost is for Antarctic fish to have antifreeze protein, but we may soon find out if the Antarctic ocean circulation changes with global warming and the waters around the icy continent warm up. Such conditions could allow other, ‘‘nonantifreeze’’ fish to invade and challenge the old survivors, perhaps by bringing pathogens into a system that is not optimized to deal with anything else invading cells but ice crystals. If so, we may yet again see how physiological trade-offs govern ecological processes.



Physiological trade-offs are truly ubiquitous in nature. Everybody can immediately and intuitively understand them. If an organism puts too much energy into detoxifying ice crystals, other functions—perhaps predator defense, pathogen killing, or sperm maturation— lag behind (in fact, many notothenioid fish species are infested, often heavily, by parasites). Ecologists have discovered many pervasive life history trade-offs whose physiological underpinnings are currently under intensive investigation. For example, the more an animal reproduces, the more likely it is to lead a shorter life. The faster an animal grows, the more resources it needs, and again the more likely it is to lead a shorter life. However, there are circ*mstances when such trade-offs are not observed. In one, animals come in different qualities, with high-quality individuals within a species sometimes ‘‘living harder and dying older’’ than lowquality individuals. Such exceptions to the trade-off rule present considerable challenges and research opportunities for physiological ecologists. What mechanism(s) allow—at least in the short run of one or several generations—one individual to be more likely to survive or to live longer than others? Another challenge to the trade-off rule is presented by laboratory, domestication, and generally captive conditions. Under such circ*mstances, animals often appear to escape trade-offs. Again, it is yet unclear how animals can become ‘‘masters of all trades.’’ The most likely physiological scenario is that the abundance of energy and nutrients provided in captivity allows individuals to obtain everything they need and thus to override physiological trade-offs. If confirmed and analyzed on the mechanistic level, this important distinction between feast and famine in the wild and almost pure feast in the laboratory could shed significant light onto one of the most pervasive principles in physiological ecology. The question of how trade-offs come about immediately leads us to question how a multitude of organismal functions can be integrated and optimized. Physiological ecologists have found a simple, perhaps ingenious, way to ascertain how individual animals can deal with their environment. 2. GUIDING CONCEPT: PERFORMANCE AS INTEGRATIVE MEASURE OF INDIVIDUAL FITNESS

Instead of analyzing each physiological trait on its own in isolation, physiological ecologists resort to quantifying whole-organism function. Imagine the different ways in which one could answer whether lizard muscle fibers work well at low or at high temperatures. A valid reductionist approach could be to isolate each muscle



fiber type, cultivate them all in vitro, expose the fibers to different temperatures, stimulate them electrically, and measure their energy expenditure and contraction rate and speed. However, what matters for individual animals is how they use their entire complement of muscles to perform certain common tasks such as fast running. Maximum running speed may be related to male fighting ability, female nest-digging ability, insect-catching capacity, and agility to escape predators. Thus, all individuals in a lizard population are expected to rely on fast sprint speed. Ray Huey of the University of Washington made use of this experimental paradigm and showed in comparative studies of individual wholebody performance that most ectotherms are able to cope with a large range of low environmental temperatures. However, as individual performance reaches its maximum, it rapidly drops off toward even higher ambient temperatures. The physiological basis for this performance asymmetry is presumably found in temperature sensitivities of physiological or molecular processes. Interestingly, individual performance is also subject to strong trade-offs. For example, although some species of ectotherms have large temperature ranges under which they can perform well, others have very narrow performance breadths (see below). 3. PROCESS I: ACQUISITION OF ENVIRONMENTAL INFORMATION

One of the most survival-relevant tasks of animals is to gather environmental information. Again, this task is subject to physiological trade-offs. Physiological ecologists working on bat echolocation determined that producing the ultrasound that bounces back from objects, i.e., provides bats with environmental information, is costly both in immediate energetic costs and in associated physiological costs. In addition to energetic costs, bats face the costs of producing organs and brain structures that enable them to expend energy on echolocation calls in the first place. Biologists actually exploit the fact that environmental information gathering is expensive. Bats spare the costs of echolocation when flying in known habitat and often do not echolocate there, allowing researchers to trap them with fine nylon nets. High physiological costs of maintaining functioning tissue may also explain why juvenile migratory songbirds start out with a small hippocampus, a brain area involved in spatial memory and thus long-term information gathering. As individuals conduct their first transcontinental journeys, they add additional cells and

connect their cells in more complex ways. However, because space in the brain capsule is presumably limited, the physically and physiologically expanded spatial memory for a life on the move may again be traded off against other brain functions that in turn deteriorate. Energetic trade-offs between form and physiological function are particularly prominent in long-distance migratory songbirds that had to evolve streamlined foreheads for aerodynamic reasons, compared with their short-distance migrating relatives. Physiological ecologist Melissa Bowlin recently learned by studying heart rate in naturally migrating New World thrushes (songbirds) that even small morphological differences significantly affect costs of transport in the air. 4. PROCESS II: INTERNAL COMMUNICATION AND REGULATION OF PHYSIOLOGICAL FUNCTION

Once environmental information is gathered, it needs to be communicated most efficiently throughout the organism. Again it appears that cost minimization and trade-offs are key guiding physiological principles. Quick, practically immediate transfer of environmental information is achieved by costly electrical (neuronal) connections. However, for many types of information that either need to be communicated continuously or at least on the long term, electrical connections are by far too costly. Instead, animals use small and ‘‘cheap to produce’’ messenger chemicals (hormones) that bind to receptors in target tissues. The main advantage of a hormonal communication system is that it is inherently flexible at many levels, i.e., rates of physiological processes can be altered at production, at the chemically supported transport of hormones to target tissues, at the possible breakdown of messenger chemicals, and with respect to the number of receptors expressed at and by target tissues. Thus, for example, if a cell does not need (much) stimulation, it can degrade particular types of incoming hormone molecules (indicating particular, general environmental messages) in its periphery and/or provide only very few receptor sites as ‘‘mailboxes.’’ Cells can also destroy the ‘‘mail’’ immediately so that it has no long-lasting effect. Physiological ecologist John Wingfield showed that this cheap hormonal messenger system conveys both long-term and short-term environmental information and prepares the individual organism for certain activities. Many animals reproduce seasonally and grow reproductive organs in response to changes in day length, often mediated by the light-sensitive hormone melatonin. Because of physiological trade-offs, individuals do not allocate maximum efforts toward certain

Physiological Ecology: Animals reproductive activities such as territorial defense from the outset. Instead, organisms often use behavior– physiology feedback loops to allow them to carefully regulate their efforts in response to environmental factors, in this case the actions of other members of the population. Thus, if population density is high in songbirds, individuals interact with others of the same species more often. Wingfield showed that individuals can ramp up reproductive hormones such as testosterone in response to a social, particularly reproductive, challenge. It is yet unclear whether this feedback is via increased physical activity (i.e., energy expenditure), increased neuronal stimulation (e.g., visual density), or a combination. In any case, it is clear that animals use hormones as a cheap means to communicate environmental information throughout the body. 5. PROCESS III: ENERGY EXPENDITURE AS ONE CENTRAL HUB FOR TRADE-OFFS

All along it has become obvious that organismal tradeoffs can be expressed to a significant extent in terms of allocations in energy turnover. Energy is probably one of the physiological factors that are most limited under natural circ*mstances. It is thus not astonishing that physiological ecologists cast many of their discussions in energetic terms and consider energy as the central hub for physiological trade-offs. Life follows the laws of thermodynamics, i.e., energy can neither be created nor destroyed (First Law). Furthermore, the disorder of a system (its entropy) increases over time as its energy content degrades to unusable heat. The only way animals can compensate for ever-increasing entropy is by constantly acquiring energy via food. However, foraging is again costly as well as time consuming, i.e., poses opportunity costs and is risky. The food then has to be broken down into chemicals usable by the organism, again a costly, damaging, time-consuming process. Because animals will do anything to minimize costs, it should be obvious that environmental temperature is one of the most important habitat factors. Temperature has a hump-shaped influence on molecular processes such as enzyme activity. Coming from the low side, increasing temperatures enhance the rate of physiological processes and thus energy expenditure. Higherthan-optimum temperatures often show destructive effects and can result in serious structural damage. Organisms incur costs at low environmental temperatures either because they are less agile (many ectotherms) or because they have to produce more internal heat (endotherms). Some animals have special tissues that help them produce heat very efficiently,


such as brown fat in bats, which produces heat without shivering. Higher-than-optimum temperatures often become dangerous because organisms very rapidly lose performance and expend much energy in thermoregulatory activities, both behaviorally and physiologically (panting, activation of heat shock proteins). Although most animals attempt to minimize energy expenditure for nonessential tasks, it has become clear that, across various types of animals, high energy expenditure has evolutionary benefits. Increased energy expenditure involving constantly high body temperatures with an associated constant interior milieu has perhaps been one of the key innovations in physiology. 6. PROCESS IV: KEY INNOVATIONS

Evolutionary key innovations give organisms access to new resources and cause rapid, sometimes spectacular adaptive radiation, as seen above in the case of antifreeze proteins. It has been postulated that a long sequence of key physiological innovations is responsible for the diversity of life forms present today. For example, Michael Berenbrink and colleagues discovered that a key physiological innovation underlies the large adaptive radiation of fish. It is the unique ability of fish to secrete molecular oxygen into the swimbladder—a seemingly simple physiological process that had already been invented some 100 million years earlier in the eye. However, because certain fish were later able to regulate swimming behavior very cheaply using their new oxygen-filled swimbladders, they diversified hugely in form and function. The physiological key in this process was a change in the Naþ/ Hþ exchange activity of red blood cells and a change in the content of surface histidine of hemoglobin (histidine is one of the 20 most common amino acids). Another common key innovation—and again a highly efficient way of organisms to economize on physiological expenses—is to use special chemical components of other organisms. May Berenbaum demonstrated such a system very nicely in the interaction between the parsnip webworm and wild parsnips. Throughout the parsnip plant there exist a group of toxic chemicals called furanocoumarins that are the favorite food of the parsnip webworm. Furanocoumarins are so toxic that only very few herbivores can deal with them. However, webworms possess a highly efficient detoxification system involving cytochrome P450s, a very large and diverse superfamily of hemoproteins (iron-containing proteins) that simply insert one atom of oxygen into an organic substrate. Webworms use the toxic furanocoumarins to strongly deter predators from eating them.



Although animals can engage the help of others, perhaps through their chemicals, to defend themselves against predators, there are also more direct ways to fight pathogens and parasites. 7. PROCESS V: SELF-DEFENSE: IMMUNOECOLOGY

The study of the physiological ecology of immune reactions is a relatively new but fast growing and highly important field. In the past, immune biology has largely focused on very specific, fine-scale mechanisms of the immune defense. Immunoecology adds the systemic component to such detailed studies by addressing the integration of various immune responses on the individual level. In a key contribution, Kelly Lee and Kirk Klasing showed that the relative immune defense effort spent on either the innate or the adaptive arm of the immune system may be ecologically important. For example, such a differential allocation of efforts into different arms of the immune system may distinguish highly from poorly invasive species, such as the house sparrow and the tree sparrow, respectively. Along the idea of whole-organism performance tests (see above), immunoecologists assess the reaction of individuals toward various immunological challenges simultaneously and as a composite measure. Physiological immune responses can be mediated by essentially two arms, the innate and the adaptive part of immune systems. The first line of defense is usually the innate arm. Specialized cells patrol tissues and have a superb ability to recognize an invader as foreign. As soon as the foreigner-recognition process starts, the first innate cells release signal molecules (cytokines) attracting bacteria- and virus-eating cells (scavenger macrophages, natural killer cells). Subsequently, the cells of the innate immune system send specialized signal molecules to the second (adaptive) arm of the immune system. The adaptive part of the immune system activates its machinery to produce antibodies that bind to and neutralize the foreign invaders. Whereas the innate system is costly to maintain and to activate, the adaptive system is costly to grow in the first place— once it is established, it appears fairly cheap to maintain. It is important to note that organisms differ strongly in how much emphasis they put on the two arms of the system. Again it appears that because of omnipresent trade-offs, a jack of all immunological traits is a master of none. It is important to note in this context that some biomedical experimental subjects such as the house mouse do not necessarily provide systems that reflect the immune allocation in humans. Whereas humans are long-lived and invest heavily in the adaptive arm of the immune response (costly to develop but

cheap to run), house mice are generally so short-lived and dependent on fighting each disease immediately that they invest much more strongly in the innate arm of the immune response. It will remain a challenge in physiological ecology to understand exactly how organisms allocate resources toward immune responses. 8. APPLICATION: CONSERVATION PHYSIOLOGY

Animals have always been sentinels for environmental changes and catastrophes. For example, when the causal (reproductive) effects of dichloro-diphenyltrichloroethane (DDT) on top predators became clear, DDT-like substances were prohibited in large parts of the world. For conservation strategies to be successful, it is important to understand the physiological responses of organisms to their changed environment. Perhaps one of the most useful tools in conservation physiology is the rapid assessment of environmental stress via the measurement of glucocorticoid ‘‘stress’’ hormones. These steroid hormones are ubiquitous in vertebrates and occur at low (baseline) levels in all individuals. In many cases when individuals are experiencing increased environmental demands such as inclement weather or predation, glucocorticoids increase in the circulation and, subsequently, in the feces. Conservation physiologists often experimentally induce mild stress (capture and handling) to assess the capacity of an individual to react to environmental stress. The usefulness of conservation physiology is that it can reduce the complexity of conservation problems to highlighting a single set or small number of the most important stressors for organisms. New physiological techniques can enable a rapid assessment of the causes of conservation problems and the consequences of conservation actions. 9. FUTURE CHALLENGES

The biggest challenges in the future of physiological ecology will be to monitor, understand, and ultimately predict what animals do during their often long lives. Advanced biologging techniques of physiological parameters are at the brink of enabling field researchers to conduct studies that a few years ago were possible only in a laboratory situation. Furthermore, even small animals can perhaps soon be followed over large temporal and spatial scales in the wild. Such new data on physiological state and overall individual space use may ultimately allow researchers to understand the animal mind. Once we know in (almost) real time how individuals process environmental information (via hormonal mechanisms), and we know the environmental conditions in the vicinity of an individual (via

Physiological Ecology: Animals animal-borne location loggers) in combination with the individual’s physiological state, we may be able to predict decisions of animals mechanistically. FURTHER READING Cheng, Chris C., and Luise Chen. 1999. Evolution of an antifreeze glycoprotein. Nature 401: 443 444. Janzen, Daniel H. 1967. Why mountain passes are higher in the tropics. American Naturalist 101: 233 249.


Lee, Kelly A., and Kirk C. Klasing. 2004. A role for immu nology in invasion biology. Trends in Ecology and Evo lution 19: 523 529. Martin, Lynn B., Zac M. Weil, and Randy J. Nelson. 2007. Immune defense and reproductive pace of life in Per omyscus mice. Ecology 88: 2516 2528. Wikelski, Martin, and Stephen Cooke. 2006. Conservation physiology. Trends in Ecology and Evolution 21: 38 46. Wingfield, John C. 2003. Control of behavioural strategies for capricious environments. Animal Behaviour 66: 807 815.

I.3 Physiological Ecology: Plants David D. Ackerly and Stephanie A. Stuart OUTLINE

1. 2. 3. 4. 5.

Introduction Resource acquisition Resource allocation and growth Responses to environmental conditions Ecophysiology, distributions, and global climate change

Plant physiological ecology addresses the physiological interactions of plants with the abiotic and biotic environment and the consequences for plant growth, distributions, and responses to changing conditions. Plants have three unique features that influence their physiological ecology: they are autotrophs (obtaining energy from the sun), they are sessile and unable to move, and they are modular, exhibiting indeterminate growth. Plant growth depends on acquisition of four critical resources: light, CO2, mineral nutrients, and water. Light together with nitrogen-rich enzymes in the leaf drive photosynthetic assimilation of CO2 into carbohydrates. Uptake of nitrogen and phosphorus, the elements most often limiting growth, is facilitated by symbiotic associations on plant roots with bacteria and fungi, respectively. Most water acquired by plants is lost in transpiration in exchange for CO2 uptake through stomata. Water moves through a plant by cohesiontension, drawn upward as a result of evaporation from leaves. Excessive tension can lead to embolism, in which air bubbles enter the water column and block water transport. Within the plant, allocation of resources to alternative functions creates important trade-offs that critically influence plant responses and performance in contrasting environments. Physiological ecology plays a critical role in understanding the distributions of individual species and of major biomes at a global scale and is vital to understand the potential impacts of global climate change on vegetation and biodiversity.

GLOSSARY acquisition. The processes of acquiring resources from

the environment, such as photosynthesis in leaves and nutrient uptake by roots.

allocation. The partitioning of resources among alter-

native structures or functions within a plant. The principle of allocation states that resources used for one purpose will be unavailable for other purposes, creating trade-offs that strongly influence plant growth and life cycles. conditions. Factors of the environment that influence an organism but cannot be consumed or competed for (e.g., temperature, pH). embolism (or cavitation). The blockage of water transport by air bubbles in the xylem (water-transporting cells), causing reduced water transport and, potentially, plant death. leaf energy balance. The balance of energy inputs and outputs that influence leaf temperature. Solar radiation is the most important input, and transpirational cooling and convective heat loss are the most important outputs. photosynthetic pathway. Plants exhibit three alternative photosynthetic pathways (C3, C4, and CAM) that differ in underlying biochemical and physiological mechanisms, resulting in contrasting performance depending on temperature and the availability of light, water, and nutrients. resources. Aspects of the environment that are consumed during growth and that plants compete for. The most important are light, water, nutrients, and space. water and nutrient use efficiency. The efficiency of photosynthesis relative to investment of water or nutrients, respectively. 1. INTRODUCTION

Physiological ecology examines how plants acquire and utilize resources, tolerate and adapt to abiotic conditions, and respond to changes in their environment. The study of physiological ecology considers plant physiology in relation to the physics and chemistry of the abiotic world on one hand, and a broad ecological and evolutionary context on the other. Plant physio-

Physiological Ecology: Plants logical ecology provides the basic sciences with essential information about plant evolution, biodiversity, ecosystem productivity, and carbon and nutrient cycling. It also plays an instrumental role in a wide range of applied sciences, including agriculture, forestry, management of invasive species, restoration ecology, and global change biology. In its early years, plant ecophysiology addressed two broad themes. One was the effort beginning in the midnineteenth century to understand the global distribution of major biomes and vegetation types, led by pioneering plant geographers such as A. von Humboldt, A.F.W. Schimper, and their followers. These workers recognized that similar vegetation types arise under similar climates in different parts of the world, and they developed basic principles of plant form and function that could explain these global patterns. This led to a subsequent emphasis, in the early twentieth century, on the question of how plants survive in extreme environments. Principles of physiology and biophysics were applied in natural settings to understand how plants can tolerate and even thrive from the heat of the desert to the extreme cold of the high arctic and the upper limits of vegetation on high mountains. Both of these traditions combined the mechanistic view of the physiologist with the idea of evolutionary adaptation to understand why species with different physiological characteristics dominate under contrasting environmental conditions. In the United States, plant physiological ecology played a key role in the development of ecology as a discipline. The Plant World, published until 1919, was the forerunner of Ecology, the flagship journal of the Ecological Society of America. Ecology in the early twentieth century emphasized physiological, functional, and ecosystem ecology. Population and community ecology as we now know them had not yet emerged. Three Important Things about Being a Plant

Plants share three important features that have profound consequences for their ecology and evolution, including physiological ecology. (1) Plants are autotrophs, converting sunlight to stored chemical energy that is the basis for terrestrial food webs and ecosystems. (Nonphotosynthetic and parasitic plants are an exception to this rule.) Photosynthesis is one of the outstanding products of evolution and is still more efficient than any photovoltaic mechanism for the capture and conversion of solar energy. (2) Plants are sessile—once a seed germinates and the seedling is established in the soil, plants cannot move. They cannot hide or escape from abiotic conditions or biotic ene-


mies, and they cannot seek out mates, at least directly, for reproduction. Plants exhibit an enormous diversity of seed germination mechanisms that control the time and place of germination, thus shaping the environment the seedling and adult plant subsequently occupy. (3) Plants exhibit modular, indeterminate growth. They grow by cell division in regions known as meristems, located at the tips of growing branches, in axils at the base of leaves, beneath the bark of trees, and at the tips of roots. Meristematic cells are undifferentiated throughout the life of a plant, and most plants never reach a fixed, mature size. The combination of indeterminate growth and immobility means that growth and development are important mechanisms through which plants respond to the environment, and in this way growth in plants plays an analogous role to behavior in animals. Conditions and Resources

At the core of physiological ecology is the study of how organisms respond to and are affected by the abiotic environment. In the case of plants, it is useful to divide the environment into conditions and resources. Conditions are factors that cannot be consumed or depleted by organisms, such as temperature, pH, or salinity. Resources are substances (or sources of energy) that are captured or consumed, can be depleted, and can be the focus of competitive interactions among individuals. The following sections address the acquisition and allocation of resources and the mechanisms by which plants respond to and tolerate a wide range of environmental conditions. 2. RESOURCE ACQUISITION

All plants require the same basic resources for growth and reproduction: carbon, light, mineral nutrients, and water. The essential challenge for terrestrial plants is that these resources are located in different places (above versus below ground) and have very different modes and rates of supply in the environment. Carbon and Light

Carbon, in the form of atmospheric carbon dioxide, is available at a relatively constant concentration. CO2 enters leaves through microscopic pores known as stomata. Stomata are formed by pairs of cells, known as guard cells, which are joined at either end, like two elongated balloons. When fluids move into the cells, they swell and bend, opening a small pore that allows gases to diffuse in and out of the leaf. The regulation



of pressure within these cells is an intricate process influenced by chemical signals from the leaves and roots that depend on soil moisture availability and internal water status of the plant. When stomata open, CO2 diffuses from the atmosphere into the pores, where it crosses an air–liquid interface and dissolves in the interior fluids of the leaf as carbonic acid. At the same time, however, water evaporates from inside the leaves and diffuses through the stomata into the surrounding atmosphere. This exchange of water for CO2 is one of the most fundamental trade-offs governing photosynthesis and plant growth. The concentration gradient driving the diffusion of water out is much steeper than the gradient for CO2 coming in. As a consequence, plants lose 100 to 500 molecules of water for each molecule of CO2 they absorb. Most of the water taken up by plants (see below) is used for this purpose. The ratio of water loss to CO2 uptake is known as water use efficiency and represents a critical physiological trait that influences plant growth and distribution in contrasting climates. Photosynthesis is a biochemical reaction that uses energy from sunlight to combine CO2 and water to make carbohydrates (glucose, starch, and other sugars), releasing oxygen in the process. Photosynthesis involves two coupled processes, known as the light reactions and carbon reactions. The light reactions use solar energy to reduce NADPþ to NADPH and phosphorylate ATP. These provide energy for the carbon reactions of the Calvin cycle. The most important of these reactions is the fixation of a CO2 molecule to a five-carbon carbohydrate chain, followed by a rapid split into two three-carbon molecules, which gives this process the name C3 photosynthesis. The CO2 fixation step is regulated by the enzyme RUBISCO, which, because of its relatively low efficiency, is present in very high levels in the leaf. RUBISCO represents up to 50% of all proteins in a leaf and is thought to be the most abundant protein on the planet. The C3 pathway is dominant in plants of temperate and cool climates as well as in most trees. In deserts, grasslands, and other dry environments, two alternative photosynthetic pathways are found, each of which has evolved many times independently. C4 photosynthesis, common in grasses (including crops such as corn), utilizes the enzyme PEP-carboxylase instead of RUBISCO for the initial fixation of CO2. This is a more efficient alternative, which can operate at lower internal CO2 concentrations, so the plant can have fewer, smaller, or less open stomata and therefore lose less water. The first fixation step creates four-carbon compounds (hence the name), which are then shuttled to cells deeper inside the leaf where they are broken down, releasing the CO2 for incorporation into the

Calvin cycle. The extra steps of the C4 pathway require additional energy, so C4 plants occur primarily in warm and high-light environments. The third pathway is known as CAM photosynthesis (Crassulacean acid metabolism), named for the plant family Crassulaceae where it was first discovered. CAM photosynthesis is widespread in cactus, tropical euphorbias, yuccas, and other succulents. CAM also utilizes PEP-carboxylase for the initial fixation step, but the stomata are opened only at night, allowing CO2 to diffuse into the leaf with minimal water loss because of lower temperatures and higher relative humidity. The carbon is stored in carbon acids (hence the name) until daylight, when they are broken down and passed to the Calvin cycle. Almost all plants described as ‘‘succulents’’ have CAM photosynthesis, and the swollen and fleshy leaves or stems contain the expanded cells that are used to store the four-carbon compounds through the night. The nighttime uptake of carbon results in greatly enhanced water use efficiency because CAM plants lose only as few as 10 water molecules per CO2 molecule acquired. However, overall photosynthetic rates are very low, limiting growth rates. Photosynthesis in sun versus shade also presents trade-offs that are important for plants growing in heterogeneous light environments such as the forest understory. Plants with C3 photosynthesis exhibit a characteristic light response of photosynthesis. In complete darkness, photosynthesis is shut down, and leaves have a net loss of CO2 as a result of background respiratory processes. With slight increases in light, photosynthetic rates increase until the light compensation point is reached, when photosynthesis balances respiration and there is no net loss or gain of CO2 by the leaf. Net photosynthetic rates become positive above this light level and increase rapidly until they reach a point where the concentration and activity levels of RUBISCO and other enzymes become more limiting than the availability of light energy. At this point, photosynthetic rates reach a plateau known as the light-saturated photosynthetic rate. In shade, photosynthesis is primarily limited by light energy rather than enzyme levels. As a result, shade leaves have lower nitrogen concentrations per unit leaf area, a lower saturated photosynthetic rate, and lower background respiration. The result is that the light compensation point is lower, and shade plants can maintain zero or positive carbon balance at lower light levels. The differences between sun and shade leaves are generally observed both within and between species. Competition for light is largely asymmetric or onesided: the highest leaves in the canopy capture the most light, and leaves lower down, on the same or other

Physiological Ecology: Plants plants, receive much less. It has been argued that the evolution of plant height can be understood as an evolutionary game: if all the plants in a community ‘‘agreed’’ to reduce their height equally, they would all still receive the same amount of light. However, the community would be easily invaded by a taller ‘‘cheater’’ that received a disproportionate share of this critical resource. Taller strategies will continue to invade until the costs of additional height (in structural support, movement of water, etc.) outweigh the benefits, and an equilibrium is reached. This equilibrium will vary, depending on the availability of light, water, and nutrients and is thought to explain variation in the height of forests and other vegetation around the world. Water

Water is central to the life of a plant. In addition to the water lost in exchange for carbon uptake (see above), water is needed for tissue hydration, nutrient uptake, long-distance signaling, and as a source of pressure for structural support and cell expansion. Water transport begins in the soil, where root hairs provide a large surface area for water uptake. Water moves through and around cells until it reaches the endodermis, a root layer in which the cell walls are impermeable to water because of a waxy inclusion in the cell membrane known as the Casparian strip. To move beyond the Casparian strip, water must pass through living cells. This allows the plant to regulate how much water enters the active root tissue and can be used to generate root pressure. The main water transport tissue inside roots, trunk, branches, and leaves is the xylem, composed of hollow cells that are dead at maturity. Water travels through the xylem to the leaves and evaporates from air/water interfaces within the stomata; evaporation from the leaves is known as transpiration. Movement along this path occurs as water moves from areas of high water concentration (high water potential) to areas of lower water concentration (lower water potential). Under all but the most humid conditions, the concentration of water (water potential) is much lower in the atmosphere than it is within the plant. The difference in concentration drives evaporation into the atmosphere. Within the xylem, water is connected by cohesion to form a single column between leaf and root. As a result, evaporation at the leaves’ surface effectively pulls water out of the soil. Cohesion, which is the result of hydrogen bonding between water molecules, gives the water column the ability to withstand stretching, also known as tension or negative pressure. This scenario, first proposed in 1893, is known as the tension–cohesion theory. The underlying mechanisms have been questioned


from time to time, but in general, it is considered to be well supported by empirical evidence. Like a supersaturated or supercooled solution, water under tension is in a metastable state. As a result, it is vulnerable to disruption and can vacuum boil, spontaneously forming an air bubble or embolism. This process is known as cavitation. Because embolisms break the water column, they block the transport of water from root to leaf. MRI studies show that cavitation is constantly occurring, and being repaired, during transpiration. Stresses, such as drought and freezing, can cause more damaging levels of embolism. Xylem architecture is highly redundant, and plants can survive, and sometimes repair, many of these losses. However, if a sufficient proportion of the xylem is blocked, embolism can cause the death of distal branches and leaves. When a plant dies in a drought, embolism is likely the proximal cause of death. Nutrients

Mineral nutrients, like water, are primarily acquired below ground. Roots are the essential foraging organs for below-ground resources. Many roots also sustain colonies of mycorrhizal fungi, which are important for the uptake of nutrients. The two soil nutrients that are most often limiting to plant growth are nitrogen, in the form of nitrate or ammonium, and phosphorus, usually taken up as phosphate. Other macronutrients required in relatively large quantities are potassium, calcium, magnesium, and sulfur; micronutrients, required in much smaller quantities, include chlorine, iron, manganese, boron, zinc, copper, nickel, and molybdenum and (in some plants) sodium, cobalt, and silicon. Nitrogen is one of the most abundant elements in the biosphere but one that is often available in limited supplies. The primary source of nitrogen is dinitrogen gas from the atmosphere; however, plants are unable to assimilate this nitrogen directly. Instead, atmospheric nitrogen is assimilated by nitrogen-fixing bacteria; many of these bacteria enter into symbiotic relationships with plants and occupy nodules on plant roots. N-fixing symbioses are widespread among thousands of species in the legume family (Fabaceae) as well as at least nine other related groups. Once nitrogen has entered an ecosystem, decomposition of litter and mineralization of organic nitrogen by soil microbes makes nitrogen available for uptake by plants. Photosynthetic nitrogen use efficiency (PNUE) is the ratio of photosynthetic productivity to the concentration of nitrogen within the leaves (or the whole plant). On a short-term basis, PNUE is higher in faster-growing plants with short leaf lifespan, high tissue nitrogen concentrations, and high photosynthetic rates. However, over



the lifespan of the leaf, PNUE tends to be higher in slower-growing plants with lower instantaneous rates of photosynthesis but long-lived leaves. Phosphorus is primarily derived from weathering and soil formation and then is cycled within an ecosystem in parallel with nitrogen. Phosphorus is relatively immobile and diffuses very slowly in soil water. Symbiotic relationships between plants and mycorrhizal fungi play a critical role in phosphorus uptake, as the hyphae of the mycorrhizas greatly extend the foraging area of the root system. As in the N-fixing symbioses, plants provide carbohydrate as an energy source for the fungi. Plants may also leach organic acids into soil, and the reduction in soil pH increases the availability of phosphorus as cations are exchanged on clay particles and other surfaces. Recent studies have shown that 10% to 30% of net carbon gained by a plant may be lost into the soil either as leachates or carbohydrate supplied to symbiotic fungi or microorganisms. 3. RESOURCE ALLOCATION AND GROWTH

Resource acquisition is only a part of the story. To understand how plants grow, respond to the environment, and differ from each other, we must examine the allocation of resources. Allocation refers to the partitioning of acquired resources among different structures and functions within the plant. The principle of allocation underlies many of the fundamental tradeoffs involved in plant growth: energy or materials can be allocated to only one structure or function at a time, so investment in one process will invariably entail trade-offs in others. Carbohydrates synthesized in the leaf by photosynthesis are loaded into the phloem and can be transferred to other parts of the canopy, to the branches and trunk, to flowers and fruits, and down into the roots. Nutrients, taken up in the roots, move upward in the xylem sap and are also divided among different parts of the canopy and utilized in the production of new leaves, stems, and roots; the provisioning of seeds; and the synthesis of enzymes and proteins throughout the plant. Construction and maintenance of plant tissues require a significant amount of energy as well. For each gram of biomass used to construct new leaf tissue, approximately 0.5 g of carbohydrate will be required to provide the energy for biosynthesis. Biochemical reactions and maintenance of enzyme pools also require a continual input of energy. Leaves in particular, where the photosynthetic machinery are at work, will burn off 5–10% of their photosynthetic uptake as maintenance respiration. Patterns of allocation are critically important for plant growth. In particular, allocation to leaves creates

a positive feedback, as leaves can then capture more carbon, which can be used for additional growth, etc. Investment in leaves is like investment in a savings account, with the benefits of compound interest over time. For example, wild radish plants invest approximately twice as much of their carbon gain in new leaves, compared to the domestic radish, which has been selected for high allocation to the tuber. Although the leaves of the two types have identical photosynthetic rates, the wild type grows to three times larger than its domestic relative over the course of a season. On the other hand, there must be limits to the benefits of investing in leaves. A plant with too few roots would have lower growth rates as a result of insufficient water or nutrients or might simply fall over if it were not well rooted in the soil. A plant with too little above-ground structures (stems and petioles) would have a canopy packed with leaves, all shading each other, with very inefficient light capture. The principle of optimal allocation captures these trade-offs, as it is clear that there is an intermediate optimum in the allocation of resources to leaves at which the growth rate of the plant is maximized. In practice, it can be difficult to determine whether a plant is actually at its optimum, but the idea plays a central role as a guiding principle of ecology. Over the entire life cycle, natural selection favors those genotypes that maximize their fitness, usually thought of in terms of lifetime reproductive output (see chapter I.14). In short-lived plants such as annuals, allocation must shift rapidly from vegetative growth to the production of flowers and the maturation of seeds and fruits. At the end of the life cycle, 50% or more of the biomass in many annual plants is allocated to reproductive structures. In longer-lived plants, a key component of lifetime fitness is survival through periods of adversity, including cold and drought, or recovery from disturbance or herbivory. Regrowth after disturbance relies on the mobilization of reserve energy in the form of nonstructural carbohydrates that can be stored in stems and roots. For example, shrubs of Mediterranean-type climate regions that are adapted to regrowth after fire have increased allocation to these energy reserves, with ensuing trade-offs in growth rate and annual reproduction. 4. RESPONSES TO ENVIRONMENTAL CONDITIONS

Because they are sessile and exothermic, plants must tolerate a wide range of conditions. They cannot seek shelter or migrate to a more hospitable habitat except through reproduction. As a result, almost anything that moves a plant away from its optimal conditions may be considered ‘‘stressful,’’ including extremes of light levels (too much or too little), temperature (both hot

Physiological Ecology: Plants and cold), water supply (drought or flooding), and soil composition. However, it is important to remember that what is ‘‘stressful’’ depends on what conditions the genotype has adapted to. For instance, salty conditions that would kill most crop or house plants may be those under which a salt-marsh or mangrove plant grows best and reproduces most efficiently. Nonetheless, it appears that plants that tolerate generally stressful habitats (the very cold, hot, dry, wet, salty, toxic, or nutrient-poor) do so through conservative, slow-growth strategies. The study of leaf energy balance demonstrates the insights gained by combining ecophysiology with first principles of biophysics. The temperature of a leaf, like that of any other object, will reach an equilibrium when energy inputs and outputs are balanced. The primary input for plants is solar radiation, although the amount of radiation absorbed by a leaf may be reduced by reflective coverings such as hairs or a steeply angled leaf surface. The most important energy output is the heat loss that accompanies the evaporation of water lost in transpiration. As the temperature of a leaf increases, heat loss also goes up as a result of the steeper temperature gradient between the leaf and its surroundings, and this will eventually bring a leaf to equilibrium. Depending on radiation, wind, humidity, and other conditions, leaf temperatures may range from 58C below to 158C, or more, above ambient air temperatures. The size of a leaf has an important effect on leaf energy balance, as smaller leaves interrupt air circulation less, and are thus more closely coupled to air temperatures and less prone to overheat. Both low- and high-temperature stresses are important determinants of photosynthesis, growth and distribution, and, by extension, of vegetation type and community composition. There is no one optimal temperature range for all plants; instead, optimal temperature ranges vary. In addition, the lowest or highest temperatures that a plant can tolerate can frequently be expanded by prior exposure to sublethal cold or hot temperatures, a process known as acclimation. Adaptations to heat stress include decreases in leaf size, increases in leaf reflectance, the production of molecular chaperones that stabilize proteins and membranes, and a shift toward saturated lipids in cell membranes. Adaptations to cold stress include narrow vessel diameters, the production of molecular chaperones that stabilize proteins and membranes, and a shift toward unsaturated lipids in cell membranes. Drought stress occurs when the water potential of the soil drops below that of the plant and the atmosphere, and the plant cannot isolate itself from the soil or draw enough water to facilitate carbon gain. Flooding stress occurs when roots are deprived of oxygen and can no


longer perform necessary functions such as water and nutrient uptake. The range of drought or flooding that can be tolerated varies widely across both clades and habitats. The composition of soil can also have important effects on the uptake of water and nutrients. Soil pH, salt concentration, and heavy metal concentration can all limit the uptake of water and nutrients and inhibit root growth. Some plants are adapted specifically to these stresses and thrive in alkaline, salty, or contaminated soils. Phytoremediation, the effort to remove soil contaminants through specially adapted vegetation, relies on such plants. 5. ECOPHYSIOLOGY, DISTRIBUTIONS, AND GLOBAL CLIMATE CHANGE

The science of ecology is often defined as the study of distribution and abundance of organisms. Ecophysiology clearly plays a central role in these broad questions, particularly in explaining distributional limits of species along environmental gradients. Species distributions often reflect intrinsic tolerance limits related to physiological traits. One particularly well-studied case is the distribution of the saguaro cactus in the southwestern United States, where the northern limits of the geographic range closely parallel the –78C winter isotherm. More generally, the traits of species tend to change as one moves across environmental gradients because distribution patterns reflect the adaptations of plants to contrasting environments. A well-studied example is the relationship between the resistance to xylem embolism and water deficit and distributions, where less-resistant species either live closer to water sources or have deeper roots to maintain access to water through dry periods. However, species may employ very different mechanisms to survive in any particular environment, so there is no simple one-toone relationship between any particular physiological trait and the environmental conditions where species live. Understanding the physiological basis of species distributions is more important than ever in relation to global climate change. Paleoecological data demonstrate that plant distributions can track changing climate over centuries and millennia. Since the last glacial maximum, 10–20 thousand years ago, tree species in eastern North America and Europe have expanded their northern range limits by 1000 km or more. Rising CO2 levels from burning of fossil fuels, deforestation, and other factors are expected to cause sharp increases in temperature in the next century, coupled with spatially and temporally variable shifts in precipitation. These changes are occurring much more rapidly than postglacial climate changes, raising significant concerns



about whether plants and animals will be able to track favorable climates. Mechanistic models that incorporate physiological tolerances, as well as biotic interactions and dispersal capacity, are critical to improve these forecasts, especially for invasive species that may not occupy the full extent of their potential range in many parts of the world. Physiological information has also been used to model the distribution of the world’s major biomes. Vegetation modeling uses the idea of plant functional types, an idealized representation of a small number of physiological strategies. Carbon gain and growth of these life forms are simulated under mean climate characteristics of large grid cells that span the globe; the mix of types that prevail is then used to infer typical vegetation types, such as temperate deciduous forest, evergreen tropical forest, etc. These models have been calibrated with great success and are able to predict the broad patterns of global vegetation. Within a region, vegetation type can be a critical determinant of energy, water, and nutrient cycles. Recent work suggests that understanding these cycles at an organismal level may be critical to understanding fluxes and cycles at the scale of landscapes, regions, or ecosystems. For instance, grasslands process, consume, and convert resources in different ways, and at different rates, than forests. This occurs, in part, because of the

many physiological differences between grasses and trees. Combining information about the physiology and behavior of plants with an understanding of ecosystemscale patterns and processes provides essential data for models of global climate, biogeochemistry, and atmospheric circulation. Interaction with these disciplines is essential to scaling up to landscape, biome, and global levels.

FURTHER READING Ackerly, D. D., and R. Monson, eds. 2003. Evolution of plant functional traits. International Journal of Plant Sciences, 164, no. 3 (Special Issue). Collection of twelve papers. Lambers, H., F. S. Chapin III, and T. L. Pons. 1998. Plant Physiological Ecology. New York: Springer Verlag. Mooney, H. A., R. W. Pearcy, and J. Ehleringer, eds. 1987. How plants cope: Plant physiological ecology. BioScience 37, No. 1. Collection of six papers. Pearcy, R., J. Ehleringer, H. A. Mooney, and P. Rundel. 1989. Plant Physiological Ecology: Field Methods and Instru mentation. London: Chapman & Hall. Tyree, M. T., and M. H. Zimmerman. 2003. Xylem Structure and the Ascent of Sap, 2nd ed. New York: Springer Verlag. Woodward, F. I. 1987. Climate and Plant Distribution. Cam bridge: Cambridge University Press.

I.4 Functional Morphology: Muscles, Elastic Mechanisms, and Animal Performance Duncan J. Irschick and Justin P. Henningsen OUTLINE

1. Techniques and history 2. Examples 3. Future directions Functional morphology is the study of relationships between morphology and organismal function. A simple inspection of animal diversity reveals a remarkable array of phenotypes and concomitant functions. For example, even within a single mammalian group (bats), one observes organisms consuming food of all types, such as blood, fruit, leaves, nectar, insects, and other animals. Accompanying this diversity in diet is a remarkable diversity in morphological structure ranging from vampire bats with fangs for making sharp incisions for drawing blood to leaf-eaters specialized for grinding and mastication. One also observes similar variation for different kinds of animal locomotion. Whereas some organisms have evolved wings for flight, such as in birds, bats, and flying insects, other species have evolved elongated hindlimbs for running or jumping, such as in some lizards and kangaroos. This diversity in form and function forms an essential template for functional morphologists because it provides the ‘‘menu’’ from which researchers can address how function relates to form.

GLOSSARY biomechanics. A subfield of functional morphology that

applies mathematical and biophysical theory to understand animal movement function. The use, action, or mechanical role of phenotypic features kinematics. Animal movement; the angles, velocities, and rates at which different body parts move throughout space and the study thereof

kinetics. Forces produced by organisms during dynamic

movements and the study thereof morphology. The descriptive features of the external

and internal (anatomical) phenotype performance. A quantitative measure of the ability of

an organism to conduct an ecologically relevant task such as sprinting, jumping, or biting structure. The configuration of muscles, bones, tendons, and other tissues that allow animals to achieve dynamic movements Functional morphology is inherently mechanistic in that it seeks to understand the basic mechanical principles that explain organismal function. Thus, rather than focus purely on descriptive patterns of organismal function (i.e., the frog jumped 20 cm), functional morphology aims to understand the underlying physiological and morphological principles that allow organisms to conduct physical tasks such as swimming, running, flying, and feeding, among others. In contrast to reductionist research that studies living organisms from the biochemical or biophysical perspective (e.g., cell biology), functional morphology generally focuses on emergent functional properties arising from the whole organism. Whole-organism functional capacities represent the end output from integrated morphological, physiological, and behavioral attributes of organisms, and hence their study requires an integrative approach. For example, cheetahs are known for their remarkable sprinting capacities, and one can study how different aspects of their internal anatomy (i.e., lung and heart function, limb muscular morphology) allow cheetahs to sprint so quickly. However, functional morphology is less focused on functional capacities below the organismal level, such as the effectiveness of an



enzyme at catalyzing reactions. This rule is not absolute, as, for example, many researchers study the function of individual muscle fibers to understand how larger muscle units function. The field of functional morphology is built around several key ideas. First and foremost, the morphology of animals provides the foundation for all movement, such as the use of muscles and bones during locomotion. However, although descriptive studies of morphology are essential, by themselves they provide an incomplete picture of animal movement. Consequently, functional morphologists also aim to quantify animal function, such as feeding or locomotion. Before the advent of techniques for quantifying animal function, researchers assumed that function followed directly from morphology, but in fact, this relationship is complicated. Although the dimensions and configuration of nerves, bones, and muscle limit certain features of animal function (i.e., how fast an animal can run), they are rarely directly predictive. This is because the level of animal function (performance, see below) is driven not only by morphology but also by behavior, which is poorly understood in terms of anatomical bases. Consider the example of Dick Fosbury, who pioneered the ‘‘Fosbury Flop’’ (jumping head first, with the back to the ground) to win the gold medal for the high-jump event at the 1968 Olympics. Before the advent of the technique, high-jumpers took off from their inside foot and swung their outside foot up and over the bar. By altering the ‘‘behavior’’ of jumping, Fosbury was able to achieve a significantly higher jumping capacity. This example also highlights an important concept in functional morphology, namely, whole-organism performance capacity, which is how well an organism completes an ecologically relevant task, such as maximum sprint speed or maximum bite force. Of course, performance capacities such as the high-jump today rarely matter to modern humans but were likely highly important during the evolution of modern humans, and similar performance traits remain important for animals. In the case of the high-jump, although the ‘‘function’’ is jumping, the performance metric is jump height. Therefore, alternations in behavior (how an athlete jumps) can greatly affect performance at a given function. A subtle aspect of this view is that performance as defined here is measured at the level of the whole organism and not at a level below the organism, such as in the case of an enzyme catalyzing a reaction. The reason for this distinction is that the actions of the whole organism are those that interact with the environment, and therefore, by understanding the dual nature of morphology and performance, we can acquire a reasonably complete view of how organisms operate in a natural environment.

We provide an overview of the state of functional morphology by first describing some of the techniques used along with a brief history of functional morphological studies. We then explain some general principles of functional morphological studies that have arisen from research over the last few decades. To provide an overview of the range of techniques used in functional morphological studies, we also provide four key examples of cutting-edge functional morphological research. We conclude by describing some promising future directions for this field. 1. TECHNIQUES AND HISTORY

The complex nature of functional morphology makes a comprehensive list of techniques impossible. Rather, we provide a brief and historical outline of common techniques and also describe recently emerging technologies that offer promise for the future. Until relatively recent times, anatomical studies have been the mainstay of functional morphology. For centuries, scientists have used dissection to examine form–function relationships. Although form can be a poor predictor of function, the anatomical knowledge gained from these studies provided a foundation for many fields, including medicine and modern functional morphology. In the eighteenth and nineteenth centuries, scientists such as Dumeril, Couvier, and Mivart recorded many detailed anatomical drawings of a variety of vertebrates, many of which are still widely used today. The modern study of anatomy continues to employ basic dissection and description techniques but has remained robust by the incorporation of new imaging techniques. For example, the practice of clearing and staining enables researchers to examine patterns of bone, muscle, and tendon simultaneously and in situ. The advent of x-ray technology provided new insights into bone structure, which has now been enhanced with computerized tomography (CT) scans that enable the creation of three-dimensional models. Magnetic resonance imaging (MRI), first pioneered in 1977, has enabled unparalleled images of internal anatomy. Finally, increasingly sophisticated computer software capable of creating finite element models enables detailed reconstructions of internal structural components. Quantification of movement has been a central goal of functional morphologists. The study of kinematics has progressed from the time of still cameras to modern high-speed cameras that can operate at framing rates of 1000 frames/sec. Before the beginning of the twentieth century, the study of movement was confined to still images and simple observation (e.g., stride length based on footprints from a horse in the dust). A major

Functional Morphology advance in imaging occurred when an artist, Eadweard Muybridge, took photographs of animal movement in rapid succession to allow visualization of basic gaits and locomotor patterns. The development of celluloid film and video technology has further improved the ability of functional morphologists to capture the movement of animals for quantification. Most recently, high-speed cameras have enabled biologists to digitally capture extremely rapid events that heretofore had remained largely mysterious. Examples include the snapping of appendages in snapping shrimp and the ballistic movements of the tongue in salamanders, both of which occur over a matter of milliseconds. A relatively new technological and conceptual advance has been the use of high-speed imaging equipment in the field, allowing biologists to capture behaviors in their natural environment. Such technology has been especially useful for gaining data on large marine mammals, which cannot be easily studied in a captive environment. Studies of muscle anatomy and function have been another major focus of functional morphology. As with advances in imaging techniques for studying animal movement, technological innovations have continued to improve our ability to understand muscle function. Beginning in the 1940s, electromyography (EMG) has been used to detect and measure muscle activity. When muscles are activated, there is a measurable change in the electrical charge of the tissue. By placing wire electrodes into a muscle and measuring the current produced, one can determine when a muscle is activated. One of the early proponents of EMG work was Carl Gans, who wrote several books explaining how this method could be used to study muscle function in animals. A more recent development, sonomicrometry, enables researchers to study how individual muscle fibers change in length during movement. By inserting two tiny crystals into a muscle fiber at different points, one can measure the change in length of muscle fibers during contraction. When combined with EMG, sonomicrometry is particularly informative because these two techniques simultaneously provide information on the timing of muscle activation and length changes in individual muscle fibers. For vertebrates, bones serve as important anchors for movement. The process of locomotion, particularly in large animals (e.g., horses), imposes tremendous strains on bones, and relatively new methods using strain gauge technology enable researchers to measure the degree of torque (twisting force) and strain imposed on bones during locomotion. In a similar vein, force platforms, which amount to extremely sensitive threedimensional scales, can effectively measure the forces that animals exert as they move across the ground.


These kinetic techniques enable researchers to measure quantities such as power, force, acceleration, and energy use during locomotion. Finally, recent technological innovations have provided unparalleled ability to visualize animal movement in fluids. Unlike terrestrial movement, in which forces can be quantified by the use of force platforms, movement in aquatic environments is far more challenging to study. Methods such as digital particle image velocimetry (DPIV) use small reflective particles that interface with a laser beam to reconstruct the force vectors produced during aquatic movements of aquatic animals. General Principles

Here we describe some general principles and recent findings for functional morphology based on three important areas of research, namely anatomy, energetics, and neuromuscular function. Anatomy: Problems of Stress and Loading Morphology provides a foundation for all movement. Many morphological structures are designed for withstanding peak stresses during normal activities, such as feeding, walking, or jumping, because a failure in morphological structure (e.g., a broken bone) could be catastrophic. Therefore, throughout the course of evolution, animals have experienced strong selection for morphological structures to withstand high peak stresses. An oft-used measure of the ability of bone to withstand stresses is the safety factor, which is calculated by dividing the maximum stress a bone can withstand without failing by the peak stress of normal locomotor activites. The issue of withstanding high peak loads is particularly relevant during locomotion, when animals can exert high peak forces on limb joints, such as the knee or ankle. This problem is especially acute for very large animals (>300 kg) because as animals become larger, the ability of the musculoskeletal system to produce force and sustain stresses increases by a factor of 2, whereas body mass increases by a factor of 3. In other words, as animals become larger, the potential for catastrophic injuries (e.g., broken bones) increases dramatically. Consequently, one might also expect strong selection for constant ‘‘safety factors’’ for bones as animal size increases, and surveys and experiments with a variety of mammals show that safety factors are roughly similar (2–4) among mammals of different sizes. Large mammals have compensated for increased forces by adopting a more upright posture during movement, most obviously exhibited in elephants and horses, for example. This upright posture provides



large mammals with greater mechanical advantage and decreases stress relative to smaller mammals, which adopt a more crouched limb posture. The issue of peak stresses on morphological structures is also relevant during feeding, especially for animals that consume very hard prey. In this regard, animals that consume hard prey with morphological structures that seem ill suited for the task (e.g., cartilage) are especially intriguing. Cartilaginous animals, such as sharks and rays, consume large and hard prey without the benefit of hardened bones. One of the most spectacular examples of this phenomenon is the stingray, which can crush extremely hard clams and crabs in its jaws. Closer inspection of the internal anatomy of the jaws shows a device that operates much like a nutcracker, with prey (i.e., the ‘‘nut’’) being inserted on the ‘‘open’’ side; the triangle-like jaw structure then closes and effectively crushes the prey. This ‘‘nutcracker’’ design is far more effective at crushing prey than if the stingray used its parallel jaw parts to bite down ‘‘equally’’ (i.e., using the jaws in a parallel fashion). Further, the internal anatomy of stingray jaws also shows some intriguing convergent features with bone. Normal bone, especially the end of long bones where much of the weight loading occurs, is strengthened by the presence of a series of internal struts. MRI images of stingray jaws show a similar set of strengthening struts that enable stingrays to produce high forces without damaging their jaws. Energetics: Elastic Elements as Drivers of Movement Movement requires energy. For animals without muscles, movement is generated by a variety of mechanisms, such as hydrostatic pressure, but for most animals with muscles, a widely accepted model is that muscles drive locomotion using molecular motors, such as myosin, which converts chemical energy (in the form of ATP) to potential energy. According to this view, the speed or intensity of movement should correlate closely with the amount of energy spent. For the majority of movements driven by muscle, this simple prediction has been confirmed. For example, as reptiles run faster, their energetic expenditure increases linearly until the animal reaches muscular exhaustion, at which point muscle function rapidly decreases because of fatigue. However, this basic principle of ‘‘move faster, work harder’’ is not universal. When researchers began calculating levels of energy expenditure in large mammals during locomotion, they noticed that in some cases, expenditure would increase linearly with speed at low to modest speeds, but at higher speeds, energetic expenditure often increased at a far slower rate if at all.

A notable example is jumping in kangaroos; when hopping at slow speeds, energetic expenditure increases linearly, but at high speeds, kangaroos can move as cheaply (from an energetic perspective) as if they were moving at slower speeds. In other words, these animals seem to be cheating; they increase speed without consuming additional energy. Extensive research into the anatomy of large mammals such as kangaroos and other large ungulates (deer, gazelles) provided a potential mechanism for this energetic savings. Many large ungulates possess elongated and enlarged tendons that act as springs during locomotion. When the animal’s foot contacts the substrate during locomotion (particularly rapid locomotion), the tendon or ligament is compressed, storing elastic energy much like a compressed spring. As the foot leaves the ground, the pressure on the compressed tendons and ligaments is released, and elastic recoil from these springlike structures provides additional force to propel the animal, thus resulting in energetic savings. The idea that locomotion is driven, at least in part, by compressive springlike mechanisms represents one example of a class of elastic mechanisms that can produce force for locomotion at relatively low energetic cost. Highly elastic structures such as tendons and ligaments are the primary culprits, but muscle also exhibits some elastic capacity. The use of elastic elements for saving energy appears to be ubiquitous in the animal kingdom and often enables higher performance than would be predicted based on simple calculations of muscle power alone. For example, some frogs can ‘‘store’’ energy in tendons and ligaments before jumping. On jumping, which is initiated by their hindlimb muscles, this stored elastic energy is released, providing additional force that enables frogs to jump longer distances than based on power output from their muscles alone. Other examples include a ‘‘catapult’’ mechanism in tiny leafhopper insects, in which elastic energy is stored and subsequently released by a catch mechanism during jumping. Fleas present perhaps the most spectacular example of elastic mechanisms for increasing performance; before a flea jumps, resilin, a highly elastic protein in the legs, is stretched, and this elastic energy is then released rapidly, enabling fleas to produce extremely high accelerations. This form of energy storage is energetically beneficial because the resilin is stretched slowly at low power outputs, yet released quickly, producing high power outputs over a very short time period, much like a catapult. One well-studied form of elastic energy savings has been formulated into a predictive model, called the mass-spring model. The mass-spring model is most applicable to large terrestrial animals (e.g., kangaroos), and models a large mass (e.g., the body) attached to a

Functional Morphology large springlike structure (the limb, with its compressive tendons and ligaments). During locomotion, the mass compresses the spring during the middle of the stance phase, and then the body recovers the elastic energy at the end of the stance phase. The mass-spring model is especially useful for making energetic distinctions between walking and running in large mammals, such as humans. During walking, the hip follows an inverted-pendulum motion and is literally ‘‘vaulted’’ over the knee, with little compression of the elastic structures in the knee joint. By contrast, during running, the hip dips during midstance, and the elastic elements in the knee joint are compressed, allowing some elastic storage and recoil. This distinction implies that walking in humans is driven largely (if not entirely) by muscles, whereas running makes greater use of elastic elements. However, walking also derives some energetic savings from the manner in which the hip is first raised (to the ‘‘top’’ of the pendulum, which requires energy), and then lowered. The lowering of the hip is energetically cheaper than the raising of the hip because of the conversion of potential energy into kinetic energy. An important consideration is that the energetic ‘‘benefits’’ of such spring-like structures are probably most apparent for larger animals (e.g., horses) and are likely less important for very small animals such as insects. This is results from the simple fact that larger animals, because of their large mass, can exert much higher forces on tendons and ligaments during locomotion compared to smaller animals. Neuromuscular Function: High-Frequency and Ballistic Movements

Despite the widespread influence of elastic mechanisms across the animal kingdom, muscles remain the primary driver of most movement, such as chewing, jumping, and running. The network of muscles, attached to bones via tendons, and the nerves innervating them make up the neuromuscular system. The neuromuscular system can be thought of as two primary parts: the muscles themselves, with their various structural components, and the nerves that innervate them and provide the wiring for effective control. Under ‘‘normal’’ conditions, the movement of most joints can be reasonably modeled, but certain extreme kinds of movement pose a challenge to conventional views of muscle dynamics. High-frequency movements, such as tail rattling in rattlesnakes (*90 Hz) and vocalizations in some fish (e.g., the toadfish, *200 Hz) pose a significant challenge to the neuromuscular system. At a first pass, one might guess that nature has already provided the solution to this challenge in the form of different muscle


fiber types, each uniquely suited for different contraction speeds and force production. For example, fasttwitch muscles enable animals to produce rapid and powerful (high-force) movements over relatively short time periods, such as explosive jumps, whereas slowtwitch muscle fibers enable more sustainable and less powerful movements, such as maintaining a stationary posture. However, even fast-twitch muscle fibers would be hard-pressed to accomplish such high-frequency movements, such as observed in toadfish, especially over long time periods. Beyond the necessary coordination among the brain, nerves, and muscles, such high-frequency movements are challenging for other reasons, such as rates of calcium exchange (the driver of muscle function), ATP, and potential for rapid muscular fatigue. Inspection of the muscles used in highfrequency movements, such as the ‘‘boatwhistle’’ call of the toadfish and the rattle of rattlesnakes, show several key adaptations for rapid movements. These muscles exhibit improved release and sequestration of calcium from the sarcoplasmic reticulum, resulting in high rates of calcium movement within muscle fibers. These muscle fibers also show high rates of attachment and detachment of myosin cross-bridges and rapid removal of calcium from helper proteins. The rattle of rattlesnakes is especially confounding, as rattlesnakes can rattle at extremely high frequencies at relatively low energetic cost. Further, muscles involved in rattling show an interesting property: as the force of rattling increases, the energetic cost of muscular twitches remains constant. The rattler muscles seem to show a unique energy-saving property, namely, as the frequency of rattling increases, rattling force also increases, but the energetic cost of each twitch of the muscle remains constant. A second significant challenge to the neuromuscular system comes in the form of extremely explosive movements. For example, many frogs rely on extremely rapid projections of their tongue (which can take place over time periods as short as 30 msec, or 30 one-thousandths of a second), to capture prey. In most animals that use tongues to feed, the tongue is moved by active muscle recruitment, and the overall length of projection is modest (< 2% of body length or less). In some salamanders, such as those within the genus Hydromantes, the tongue can be projected at up to 100% of body length over just a few milliseconds. Investigations of the anatomy surrounding this elongate tongue show some muscle fibers that essentially constrict the tongue, projecting it from the mouth at high speeds, much like a watermelon seed being squeezed between fingers. Other retractor muscles then ‘‘reel in’’ the long tongue after use. The result is an explosive ballistic projection of the tongue in which the tongue


Autecology radiography to observe locomotion in savannah monitors (V. exanthematicus) and green iguanas. This technique produces video x-rays that allow observation of internal structures, in this case the gular, or throat, cavity. Pressure transducers implanted in the lungs and gular cavity provided information about breathing cycles. The results were consistent with the axial constraint hypothesis in an interesting way. In green iguanas, lung ventilation decreased as locomotion speeds increased, as predicted. In the savannah monitors, however, lung ventilation was not impaired at higher speeds. The videoradiographs showed that the monitor lizards used gular pumping to force air into the lungs. That is, the hyobranchial apparatus compresses the gular cavity, creating positive pressure and causing air to be moved to the lungs (figure 1). This work provides an elegant example of how the proper application of technology allows functional morphologists to test hypotheses of organismal function in live animals. In this case, the research provides support for the axial-constraint hypothesis while showing that savannah monitors have circumvented the constraint through the evolution of a gular pump.

moves of its own accord, without any additional input of energy beyond the initial squeezing. A perhaps more spectacular example comes from chameleons, which also exhibit ballistic properties of their tongues, which can extend up to twice their body length, and which also adhere to prey via suction. 2. EXAMPLES

To provide an overview of the breadth of techniques and questions used by functional morphologists, we provide four brief examples of integrative research that spans a range of different animal taxa. Gular Pumping during Locomotion in Monitor Lizards

The diversity of locomotor modes in animals is one of the most striking features of animal evolution. An oft-cited feature of this diversity is the dichotomy between reptiles and mammals in their aerobic capacity. Whereas mammals have substantial aerobic capacity and can run at relatively high speeds for long periods of time (> 30 min), reptiles quickly switch from aerobic to anaerobic locomotion during fast movements. However, the mechanism for this dichotomy has not been entirely well understood, but one possible explanation can be derived from the axial constraint hypothesis, which states that lizards face a respiratory trade-off when moving at moderate to high speeds. Lizards with a sprawling posture locomote with lateral undulations of the body trunk that are produced with alternating unilateral contractions of the intercostal muscles. The intercostal muscles are also used bilaterally to ventilate the lungs during resting respiration. Hence, the axial constraint hypothesis predicts that as the animal increases its speed, respiration will be inhibited. However, previous work with monitor lizards (Varanus spp.) and green iguanas (Iguana iguana) provided mixed support for this hypothesis. Researchers used video-



Fluid Dynamics during Swimming in Bluegill Sunfish

Many animals move in aquatic environments, such as fish, marine mammals, and invertebrates, among others. Unlike terrestrial environments, in which the quantification of forces is relatively straightforward (i.e., using force platforms; see below), detailing the forces involved during aquatic locomotion is far more difficult. Aquatic organisms typically move through fluids by use of specialized appendages (e.g., fins) that push against the fluid. The development of DPIV was an important advance for understanding aquatic locomotion because it allowed visualization of these more sophisticated force vectors. Research with sunfish has shown that complex fluid dynamics produced by a fish could be quantified in


Figure 1. X ray negative video images of a savannah monitor lizard (V. exanthematicus) walk ing on a treadmill at 1 km/hr, with corresponding drawings of the body, lungs, and gular cavity. The lizard is shown at three dif ferent stages of a single breath cycle: (A) end of exhalation, (B) end of costal and gular inspira tion, and (C) end of gular pump. (From Owerkowicz et al., 1999)

Functional Morphology Frontal






Figure 2. Water velocity vector fields calculated for orthogonal pla nar sections of the pectoral fin wake during swimming at 0.5 body length/sec. (Frontal Plane) Ventral view, anterior and upstream to the left. (Parasagittal Plane) Left lateral view, anterior and upstream to the left. (Transverse Plane) Posterior view showing the left side of body and the left fin; free stream flow passes perpendicularly through the plane of the page toward the viewer. Flow patterns are shown at the time of (A) the stroke reversal, following fin downstroke in the direction of the upper curved arrow, and (B) mid to late up

stroke in the direction of the lower curved arrow. The last three digits in the numerical code in each panel denote time (in milliseconds). Mean free stream flow velocity (10.5 cm/sec from left to right) has been subtracted from the frontal and parasagittal plane vector matrices to reveal wake structure. Note the spanwise component of flow in the frontal plane. By the middle to end of the upstroke, discrete pairs of counterrotating starting and stopping vortices are visible in each plane, each with a central jet of relatively high velocity flow (B). Scales: arrow, 20 cm/sec; bar, 1 cm. (From Drucker and Lauder, 1999)

three dimensions. To accomplish this goal, researchers have used DPIV to measure the dynamics of the wake produced by swimming bluegill sunfish (Lepomis macrochirus). The water of a flow tank was seeded with small (approximately 12 mm diameter) silver-coated glass beads. These beads remain suspended in the water column in a relatively uniform pattern and are illuminated with a laser focused in a plane. As the fish swims, the beads move with the water. High-speed digital video records the movement of the particles, and these videos allow quantification of the forces exerted by the fish on the fluid medium. Measuring the change in position of a

bead between consecutive frames allows calculation of a vector that can be used to determine quantities such as vorticity (rotational movement) and velocity. Filming from frontal, parasagittal, and transverse perspectives permits the quantities to be analyzed in three dimensions (figure 2). At slow speeds (0.5 body lengths/sec), when the bluegill sunfish are using labriform (with the pectoral fins only) swimming, the fins create vortices during the downstroke and stroke reversal. The propulsive forces generated by these fin strokes approximately balance the magnitude of drag and weight experienced by the fish. Researchers have also detected



considerable forces directed medially, which may aid in maneuverability. This approach was significant not only in its individual discoveries but also in providing a roadmap for future research in locomotion in aquatic animals. Jumping Performance in Anoles

One of the most exciting innovations in functional morphology has been the implementation of evolutionary principles, such as comparative methods. Because many functional morphological approaches are detailed in nature, studies are typically restricted to one or two species. However, improvements in technology have enabled researchers to capture detailed functional measurements on sets of species that differ in ecology and behavior. This approach was illuminated in recent research on jumping in Caribbean anole lizards. Caribbean anole lizards are both speciose and ubiquitous, and one of the primary features of their evolutionary radiation has been variation in their ability to jump and run. Some species are outstanding jumpers and runners, whereas other species display only modest capacities. Researchers examined the kinetics of jumping performance in 12 species of Anolis lizards that varied dramatically in morphology and jumping behavior. Anoles are particularly useful for examining evolution because they have undergone independent adaptive radiations on islands throughout the Greater Antilles, and the same morphological ‘‘types’’ (ecomorphs) have repeatedly and independently arisen. Some key questions addressed were (1) whether ecomorphs differ in jumping kinetics; (2) whether morphology correlates with jumping kinetics; and (3) whether anoles use takeoff angles that maximize jumping performance. To answer these questions, the authors measured the forces generated during jumping in 12 anole species using a force platform. This device uses a series of transducers (flexible and hollow metal rods) that are attached to a metal plate. When the lizard jumps from the plate, force is transmitted through the plate into the transducers. Bending in the transducers is then transmitted via attached wires to an amplifier and eventually a computer. When used in concert with high-speed digital cameras, force platforms allow researchers to examine the interaction between the forces used to propel locomotion (kinetics), and the movements used during locomotion (kinematics). One surprising result was the lack of difference among morphologically divergent ecomorphs in jumping kinetics. This lack of a difference occurred because there were two ways to be a good jumper, namely having long hindlimbs versus short and stocky hind-

limbs. Long hindlimbs accelerate the body over a longer time period, thus enhancing jump distance. By contrast, short and stocky hindlimbs provide extra muscle that provides high power outputs during jumping. As some anole ecomorphs have long hindlimbs, whereas others have shorter, more muscled hindlimbs, this twopronged approach to jumping results in a lack of difference among ecomorph types. A second surprising result was that anole lizards used a simple method to perform relatively long, yet short (in duration) and shallow (in height) jumps. Using simple jump equations, researchers have calculated the ‘‘expected’’ jump angles that anoles should use to maximize horizontal jump distance given their hindlimb length and takeoff velocity (figure 3). Interestingly, anoles jump using slightly lower takeoff angles (on average 2–48) than the predicted angles. However, this reduction in angle had only a very slight effect (*1% on average) on horizontal distance traveled but greatly diminished jump duration (*7% on average) and jump height (*15% on average). The ecological significance of this biomechanical perturbation remains unresolved, but these jumps may allow anoles to jump more elusively in a cluttered arboreal environment, as shorter jump durations might allow an individual to change directions more rapidly when fleeing from predators. A nice biomechanical trick! Flight Performance in Carpenter Bees

Despite the previous examples, the field of functional morphology is not limited to vertebrates. Invertebrates offer a nearly infinite palette of morphological form and behavior for functional morphologists to study, ranging from rapidly running spiders to slow and methodical sea worms. One exciting example of a functional morphological approach as applied to invertebrates is the hovering performance in carpenter bees (Xylocopa varipunctata). Functional morphologists are often interested in how morphology changes during the growth of an individual and how these changes affect function. In most animals, body shape changes dramatically from birth to adulthood, which is termed allometric growth. Because shape (morphology) is likely to dictate function and performance to some extent, one might expect differing selection pressures at different life stages. Carpenter bees are large, robust bees that excavate nest cavities in wood. These insects have relatively high wing loading, meaning that they have a greater mass to wing area ratio than other bees. Researchers have used different mixtures of three gases to change the density of the air in a chamber in which a bee hovered. Oxygen

Functional Morphology


0.10 36 37 Eq



39 Hindlimb length (m)

0.08 40 41 0.06

Gu Cr


42 Ev

Gr 0.04


Di Pu 1.4


Ca 1.6

43 1.8



Takeoff velocity (m/s)

levels were kept constant (normoxic) so metabolic rates would remain unaffected while nitrogen (N2), and helium (He) concentrations were varied. Because N2 is denser than He, changing the proportion of these two gases in the chamber results in differing gas densities. By decreasing the density until the bee is unable to hover, the researchers were able to calculate maximum flight performance. Bees were filmed as they flew, and audio recordings of wingbeats were made simultaneously. Additionally, metabolic rates were measured with flow-through respirometry. Using bees with a wide range of body sizes allowed an ontogenetic examination of flight performance, kinematics, and energetics. In these bees, abdominal mass, thoracic mass, relative thoracic muscle mass, and wing loading all scaled allometrically with respect to body mass. The thoracic muscles are responsible for flight, and their relative mass decreases as body mass increases. Perhaps unsurprisingly, those bees with the highest relative thoracic muscle mass were able to hover in pure heliox. Interestingly, both wingbeat frequency and stroke amplitude correlated positively with body mass during hovering in normodense air but were mass independent during maximal flight. Power output and metabolic rate were significantly elevated during flight in hypodense air. In this case, large bees hover at near-maximal performance, whereas smaller bees have more power and kinematic reserves during hovering flight.

Figure 3. Landscape showing the theoretical optimal (i.e., distance maximizing) takeoff angle as a function of hindlimb length (H ) and takeoff velocity (V ). Each species (N ¼ 12) is represented by a single point noted by letters (e.g., Ev ¼ A. evermanni ) with its H and V coordinates. All pre dicted ‘‘optimal’’ takeoff angles fall within the 38.88 41.88 range. (From Toro et al., 2004)


The field of functional morphology has emerged and evolved during the past century, and prospects for the future are exciting. We offer several areas that we feel constitute stimulating new frontiers for functional morphology. A conceptual advance that has arisen over the past several years has been the idea that researchers should aim to study organisms not only under controlled laboratory settings but also in natural surroundings. Ecological physiologists have long recognized the importance of understanding organismal function in an ecological setting, but the assumption that behaviors and functions obtained under laboratory settings are reasonable surrogates of natural behaviors persists. Recent work has shown that this is not always the case because animals will adjust their behavior as a function of their immediate environment. Technological advances are rapidly enabling researchers to take the ‘‘laboratory’’ into the field, thereby diminishing, or in some cases obviating, the traditional dichotomy between ‘‘field’’ and ‘‘laboratory’’ biology. For example, new advances in high-speed imaging now provide researchers with relatively inexpensive, lightweight, and portable high-speed cameras capable of filming events at up to 1000 frames/sec. Improved wireless technology enables researchers to use remote sensing devices to study changes in an animal’s internal



body temperature, hormone levels, or many other physiological parameters, all in a natural setting. An excellent example of the value of this approach is the study of movement in aquatic marine mammals. Unlike smaller aquatic animals, in which behavior and locomotion can be effectively studied in seminatural settings, such as large aquaria or tanks, human beings are not yet capable of recreating the habitat scale for marine mammals such as elephant seals or sea lions, many of which regularly travel dozens of kilometers in a single day. Further, many marine mammals regularly dive to depths greater than 100 m, and much of their behavior is closely tied to this daily rhythm of diving and returning to the surface for breaths of air. Advances in remote sensing technology have enabled researchers to measure swimming speeds, body temperatures, heart rates, and many other variables on freeswimming marine mammals. In the next several years, improvements in nanotechnology promise the exciting possibility of using similar remote-sensing devices on much smaller organisms, such as invertebrates, birds, reptiles, or small fish, for example. Some of the most exciting recent developments have come from improvements in imaging. Until recently, researchers were able to gather only still shots of internal structures, such as bones and muscle, during their use, which begged the question as to how internal structures are used in normal circ*mstances. Researchers at Brown University, spearheaded by Elizabeth Brainerd, have developed an integrated 4-D MRI system that effectively visualizes the movement of internal structures in three dimensions (X, Y, and Z) in real time. Software being developed in conjunction with this MRI system allows the accurate reconstruction of bones and muscles during feeding, locomotion, and many other kinds of movement. Using this software, one can reconstruct movements of the skeleton in conjunction with key muscles, providing unparalleled views of dynamic movements. In addition to providing valuable basic information on organismal anatomy and function, these approaches offer exciting possibilities for sports medicine and rehabilitation after traumatic injuries (e.g., anterior cruciate ligament tears). A surprising unification of efforts recently explored relationships between the morphology of sexual traits, such as peaco*ck feathers, lizard horns, etc., and functional and physiological variables. Many animals exhibit elaborate and colorful sexual structures, and the reasons for which animals have evolved such seemingly deleterious structures constitute a point of contention within evolutionary biology. Functional morphological approaches offer promise for addressing this debate because the size, shape, and color of sexual signals may be ‘‘honest’’ signals of underlying functional variables

that are valuable measures of male quality. Indeed, recent work has shown that the size and shape of such sexual structures often exhibit strong correlations with performance traits. For example, lizards with enlarged throat fans, or dewlaps, also exhibit high bite forces, despite no obvious functional link between the anatomy of head musculature (which dictates bite forces) and dewlap size and shape. In some species, male lizards with high bite forces are typically more dominant when competing against males with lower bite forces, which is reflected in their higher reproductive success. The imprint of sexual selection appears to be very strong in the realm of functional biology, and the subject promises increasing collaborations over the next decade between functional morphologists and researchers interested in sexual selection. A final emerging area of growth concerns the use of animal models for the design of robots. Much has been written on the ability of organisms to execute difficult tasks with seeming alacrity, such as ‘‘cheap’’ (energetically) and fast jumping in kangaroos or fast and maneuverable swimming in fish and shark species. By collaborating with functional morphologists, engineers have constructed a wide variety of ‘‘animal robots’’ modeled after snakes, lobsters, and fish, among others, based on underlying principles of muscle anatomy and function. In an era of declining funding for functional morphologists from the federal government, many researchers are turning to alternative sources for funding, and although we applaud the applied use of functional morphological principles in general, we caution against overzealous devotion toward such applied purposes, as researchers should be aware of potentially destructive uses for the robots that they are helping to design.

FURTHER READING Alexander, R. M. 1975. Biomechanics. Outline Studies in Biology. London: Chapman & Hall. Arnold, S. J. 1983. Morphology, performance, and fitness. American Zoologist 23: 347 361. Biewener, A. A. 2003. Animal Locomotion. Oxford: Oxford University Press. Drucker, E. G., and G. V. Lauder. 1999. Locomotor forces on a swimming fish: Three dimensional vortex wake dy namics quantified using digital particle image velocimetry. Journal of Experimental Biology 202: 2393 2412. Dumont, E. R., J. Piccirillo, and I. R. Grosse. 2005. Finite element analysis of biting behavior and bone stress in the facial skeletons of bats. The Anatomical Record 293: 319 330. Irschick, D. J., and T. Garland, Jr. 2001. Integrating function and ecology in studies of adaptation: Investigations of locomotor capacity as a model system. Annual Reviews of Ecology and Systematics 32: 367 396.

Functional Morphology Irschick, D. J., A. Herrel, B. Vanhooydonck, and R. Van Damme. 2007. A functional approach to sexual selection. Functional Ecology 21: 621 626. Lauder, G. V. 1996. The argument from design. In M. R. Rose and G. V. Lauder, eds., Adaptation. San Diego: Academic Press, 55 91. Loeb, G. E., and C. Gans. 1986. Electromyography for Ex perimentalists. Chicago: University of Chicago Press. Owerkowicz, T., C. Farmer, J. W, Hicks, and E. L. Brainerd. 1999. Contribution of gular pumping to lung ventilation in monitor lizards. Science 284: 1661 1663.


Roberts, S. P., J. F. Harrison, and R. Dudley. 2004. Allometry of kinematics and energetics in carpenter bees (Xylocopa varipunctata) hovering in variable density gases. Journal of Experimental Biology 207: 993 1004. Roberts, T. J., R. L. Marsh, P. G. Weyand, and C. R. Taylor. 1997. Muscular force in running turkeys: The economy of minimizing work. Science 275: 1113 1115. Toro, E., A. Herrel, and D. J. Irschick. 2004. The evolution of jumping performance in Caribbean Anolis lizards: Solu tions to biomechanical trade offs. The American Nat uralist 163: 844 856.

I.5 Habitat Selection Judy Stamps OUTLINE

1. Habitat and habitat selection at different spatial and temporal scales 2. Habitat selection: The behavior 3. Implications of habitat selection for basic and applied ecology Separately and in combination, the terms habitat and selection mean different things to different audiences. This chapter focuses on habitat selection behavior at the level of individuals and considers how the processes that affect the choices made by organisms at different spatial scales affect the distributions at the population level. Because we initially focus on habitat selection at the level of individuals, habitat can be defined as a location in which a particular organism is able to conduct activities that contribute to survival and/or reproduction. That is, habitat is organismspecific rather than being determined by features that may be obvious to humans (e.g., vegetation type). Selection can be defined as a behavioral process by which an organism chooses a particular habitat in which to conduct specific activities. Hence, habitat selection implies that individual organisms have a choice of different types of habitat available to them and that they actively move into, remain in, and/or return to certain areas rather than others.

GLOSSARY conspecific attraction. Attraction of individuals to

conspecifics during the process of habitat selection habitat selection. The process by which individuals

choose areas in which they will conduct specific activities heterospecific attraction. Attraction of individuals to other potentially competing species during the process of habitat selection indirect cues. Stimuli that are produced by factors that are correlated with other factors with direct effects on intrinsic habitat quality

intrinsic habitat quality. The expected fitness of an in-

dividual when it uses or lives in a given habitat, after controlling for any effects of conspecifics on fitness microhabitat. An area used for a specific type of activity (e.g., foraging, oviposition, nesting) natal habitat preference induction. Exposure to cues in an individual’s natal habitat increases the attractiveness of those cues during habitat selection 1. HABITAT AND HABITAT SELECTION AT DIFFERENT SPATIAL AND TEMPORAL SCALES

Habitats and habitat selection can occur at several different spatial and temporal scales. At larger scales, habitat refers to areas that are required for the long-term survival and reproduction of the members of a given population. In this case, habitat includes all of the areas required by all of the life stages of the members of that population, including areas that allow dispersers to travel among different patches of suitable habitat. For instance, from the perspective of a migratory bird, habitat includes breeding habitat, wintering habitat, and migratory stopovers that connect these venues. Habitat selection at large spatial and temporal scales occurs when individuals choose localities or regions that might be capable of supporting them, their offspring, and their descendents for an extended period of time. At intermediate spatial and temporal scales, habitat refers to an area capable of supporting an individual for a biologically significant, finite period of its lifetime. Examples of habitat at this spatial scale include the selection of a feeding territory by a juvenile salmonid or an area suitable for feeding and oviposition by a female butterfly. Habitat selection at this scale is particularly important for sessile organisms such as barnacles because in this case a decision made early in life affects an individual’s fitness for the rest of its life. In contrast, mobile organisms may select new habitats several times over the course of their lives as a result of changes in resource requirements, experience, or

Habitat Selection competitive ability during development, or as a consequence of seasonal movements from one area to another. Finally, at even smaller spatial and temporal scales, habitat refers to an area in which an organism is able to conduct specific activities, such as foraging, resting, courtship, oviposition, or parental care. The term microhabitat is often used to refer to such areas. With the exception of sessile species, microhabitat selection typically occurs multiple times and involves many different types of habitats over the course of an individual’s lifetime. Recently, it has become apparent that scale matters and that models that predict behavior and distributions at small spatial scales may do a less satisfactory job at predicting them at larger spatial scales. In order to appreciate how scale affects habitat selection, it is helpful to consider one of the most influential general models of habitat selection, the ideal free distribution (IFD) (Fretwell and Lucas, 1970). The IFD predicts the area that an animal will select under the assumption that animals have accurate estimates of the intrinsic quality of the different areas that are available to them. In turn, intrinsic quality indicates the net fitness payoff that an individual can expect when it is using an area, after controlling for any effects of conspecifics on fitness. This particular model also assumes that animals compete with one another when they are using a habitat, such that fitness is inversely related to population density. Finally, it assumes that animals incur no costs when they are searching for a habitat, so they are always free to choose the habitat that maximizes their expected fitness. Under this simplified scenario, the probability that an individual will select a given area will be positively related to the relative intrinsic quality of that area and inversely related to the density of other individuals in that area. Considerable empirical support for the IFD has been obtained in studies of habitat selection at small spatial scales, e.g., in studies of foraging patch selection conducted in tanks in the laboratory, or restricted areas in the field. This is not surprising because these are situations in which the assumptions of the ideal free distribution are most likely to be satisfied. Empirical support at microhabitat scales has also been obtained for modified versions of the IFD, e.g., models that assume that individuals differ with respect to competitive ability and sort themselves among habitat patches based on their competitive ability relative to the other individuals with whom they interact on a regular basis. In contrast, studies of habitat selection at larger spatial and temporal scales often yield results at odds with the predictions of the IFD. These discrepancies


have drawn attention to assumptions of this model that may not apply when animals select regions, localities, neighborhoods, home ranges, or territories for longterm use. One of the first assumptions of the IFD to be reevaluated was that individual fitness is inversely related to conspecific density. An alternative possibility is an Allee distribution, in which individual fitness increases as a function of density at low to intermediate densities and then declines at intermediate to high densities. Thus far, Allee distributions in the context of habitat selection have been documented for a wide array of taxa, including territorial animals, as well as species that live in colonies or groups. Of course, if individuals benefit from the presence of conspecifics at low to intermediate densities, then one would expect different patterns of habitat choice than if interactions between conspecifics were entirely competitive. For example, whereas the IFD predicts that empty patches of suitable habitat would be more attractive than a comparable habitat containing a moderate number of conspecifics, habitat selection under an Allee distribution predicts the reverse: newcomers should avoid an empty patch in favor of a patch that already contains members of their species. 2. HABITAT SELECTION: THE BEHAVIOR Assessment of Habitat Quality

Indirect Cues and the Effects of Conspecifics and Heterospecifics on Habitat Selection One of the assumptions of the IFD that is scaledependent is that organisms have accurate estimates of the relative intrinsic quality of all of the habitats that are available to them. This assumption is most likely to be valid when individuals are able to extensively sample and evaluate many different habitats. For instance, an animal that has lived in a home range for an extended period of time probably has reasonable estimates of the relative quality of different foraging patches within its home range. However, extensive sampling of potential habitats is less feasible when animals are choosing habitats at larger spatial and temporal scales. In this situation, sampling can be constrained by many factors, including limits on the amount of time or energy that is available to search for and investigate novel habitats or elevated risk of mortality as animals explore unfamiliar areas. In addition, because large-scale habitat selection involves areas that individuals will use for extended periods of time, there is no guarantee that factors with major impacts on fitness will even be present when individuals choose a habitat. For instance, the larval dispersers of some



benthic marine invertebrates settle in spring, when important attributes of the attachment site (e.g., exposure to hot temperatures in midsummer, or exposure to storm surges in winter) cannot be directly assessed. When direct assessment of habitat quality is not an option, organisms may rely on indirect cues of intrinsic habitat quality. Indirect cues are stimuli that can be reliably detected when organisms are searching for habitats and that are correlated with biotic or abiotic factors that affect fitness after they have chosen a habitat. For many years, biologists have focused on indirect cues involving structural features of the habitat, e.g., shapes, colors, odors, sounds, or other cues that are likely to be correlated with other factors (food, predators, parasites, etc.) with direct impacts on fitness when animals are using a habitat. More recently, researchers have expanded the notion of indirect cues to include stimuli produced by conspecifics and heterospecifics. Thus, the conspecific cuing hypothesis argues that the presence or number of conspecifics in an area can provide information about other factors (predators, food supplies, parasites, etc.) that affect the intrinsic quality of that area. Similarly, the conspecific performance hypothesis argues that cues related to the reproductive performance of conspecifics in an area may provide information about factors that affect breeding performance in that area. Even other species that are potential competitors of a focal species may provide indirect cues to habitat quality, as is outlined in the heterospecific cuing hypothesis. These hypotheses have broadened the notion of indirect cuing to include the possibility that conspecifics, successful conspecific breeders, and heterospecifics affect habitat selection behavior because they provide information about other factors with major impacts on intrinsic habitat quality. Indirect Cues versus Direct Benefits in the Effects of Cues on Habitat Selection The issue of the effects of conspecifics and heterospecifics on habitat selection raises an important general question, namely, whether organisms respond to particular cues when selecting habitats because those cues are correlated with other factors that affect intrinsic habitat quality (an indirect cue), or whether they respond to those cues because they are produced by factors that directly affect fitness after they choose a habitat (direct benefits). That is, individuals may be attracted to cues from conspecifics or heterospecifics because these cues are associated with other factors that affect habitat quality (conspecific cuing, heterospecific cuing) and/or because individuals directly benefit from the presence of conspecific or heterospecific neighbors after settling in a habitat. As a result,

two mutually nonexclusive hypotheses predict that individuals will be attracted to conspecifics (conspecific attraction) or to heterospecifics (heterospecific attraction) when choosing a habitat. In fact, a growing number of empirical studies indicate that individuals are attracted to conspecifics, to successful conspecifics, or to heterospecifics when choosing a habitat. However, it is not yet clear whether indirect cues, direct benefits, or both contribute to positive effects of cues from conspecifics or heterospecifics on habitat choice. Fortunately, this is a question that is currently under study, and we can expect more information on this topic in the coming years. The Development of Responses to Cues during Habitat Selection Over the years, it has become apparent that animals have two potential sources of information about associations between cues and intrinsic habitat quality: information from previous generations, via genes and maternal effects, and information from an individual’s immediate past, via learned associations between cues and factors with direct effects on habitat quality. Information from the past is reflected by preexisting biases, which in the context of habitat selection are expressed when naive individuals with no previous exposure to a natural cue are attracted to that cue. For instance, in nature, larval anemonefish disperse from their natal habitat (a sea anemone of a particular species) and then attempt to locate and settle on a new host anemone. Naive larvae raised in the absence of sea anemones are more strongly attracted to the odor of their usual host anemone than to the odor of other sea anemones. Similar examples of differential attraction by naive individuals to cues produced by ‘‘ancestral’’ habitats have been reported in many different taxa. An individual’s personal experience can modify or alter preexisting biases for the attractiveness of cues from natural habitats. For instance, exposure to cues in an animal’s natal habitat may increase the attractiveness of those cues to natal dispersers, a process termed ‘‘natal habitat preference induction’’ (NHPI). Thus far, NHPI has been reported for a number of taxa, including many insects and a scattering of vertebrates. The latter includes the anemonefish mentioned earlier, in which larvae raised in the presence of their typical host anemone are more strongly attracted to olfactory cues from their host anemone than are naive larvae. Preexisting biases and personal experience interact throughout development to affect preferences for cues for particular patches or types of habitat. As a result, nature and nurture both affect the ways that animals respond to indirect cues during the process of habitat selection.

Habitat Selection Adding Search to Habitat Selection

By definition, habitat selection at larger spatial scales involves longer travel distances and higher travel costs than habitat selection at smaller spatial scales. Travel costs include increased risk of mortality from predators, accidents, infection, or other adverse conditions individuals face when traveling through unfamiliar, often inhospitable, terrain. These costs are compounded when individuals have difficulty detecting suitable habitats from a distance or when suitable habitat is sparsely distributed in the landscape. In such cases, individuals may travel considerable distances without finding any suitable habitat. Moreover, habitat selection at larger spatial scales is often time- or energy-limited. For instance, natal dispersal in brush mice is restricted to a 1to 2-week period before sexual maturation, and natal dispersal times and distances in bark beetles are constrained by the fact that they do not feed en route and must rely on energy stored before they leave their natal habitat. Long travel distances, high travel costs, and time- or energy-limited search constrain habitat selection behavior in ways that can often be safely ignored in studies of microhabitat selection. In the absence of these constraints, it is reasonable to assume that preference and choice will map onto one another and that organisms will select the habitat that they perceive (correctly or incorrectly) will yield the highest fitness. However, when organisms choose habitats at larger spatial scales, preference is only one of several factors that affect habitat choice. For example, when energylimited individuals search for a habitat, theory predicts that individuals with higher energy reserves will be more selective than individuals with lower energy reserves. In that case, individuals in poor condition will be more likely to accept less-preferred habitats early in the search and less likely to end up settling in highly preferred habitats than individuals in good condition. This is an example of a ‘‘silver spoon effect’’ in which favorable conditions early in life increase the chances that an individual will be successful later in life. In fact, this type of silver spoon effect has been documented in studies of benthic marine invertebrates (bryozoans), in which larvae with large food reserves are more selective and more likely to settle in highly preferred habitats than larvae with lower food reserves. Adding search to habitat selection highlights a number of other reasons why animals should accept lesspreferred habitats, even though more-preferred habitats exist elsewhere in the same landscape. High travel costs, difficulty in detecting suitable habitats, long distances between suitable habitats, a shortage of time or energy available for search, and a scarcity of high-


quality habitats are all factors that will favor individuals who are relatively nonselective when they are searching for a new habitat. Nonselective individuals still prefer some habitats to others and can express these preferences if provided with a choice of habitats located directly next to one another. However, under natural conditions, nonselective individuals will be more likely to accept any suitable habitat rather than incur the added costs of continuing to search for a more-preferred habitat. In turn, reduced selectivity during search increases the proportion of individuals that end up choosing habitats in proportion to their availability. Hence, individuals that are highly selective at smaller spatial scales (e.g., when choosing foraging patches within a home range or territory) may be considerably less so when choosing a habitat at larger spatial scales (e.g., when selecting a region in which to settle, or establishing a home range or territory within that region). 3. IMPLICATIONS OF HABITAT SELECTION FOR BASIC AND APPLIED ECOLOGY

Most ecologists and conservation biologists are not nearly as interested in the behavioral processes that generate habitat selection as they are in the effects of these processes on animal distributions and population viability. Indeed, in the ecological literature, the term habitat selection usually does not refer to the behavior of individual animals but rather to differential patterns of habitat use. In this literature, habitat selection is inferred when the density of individuals in a particular type of habitat is higher than predicted on the basis of a null model that assumes that individuals use different types of habitat in proportion to their availability. By extension, it is assumed that higher-than-predicted densities in a given type of habitat occur because organisms preferentially settle in, use, or remain in that type of habitat. However, active habitat choice is only one of several factors that can produce differential habitat use, so this assumption need not always be valid. For instance, newly settled larvae of benthic marine fish and invertebrates are often strongly associated with certain types of habitat. In the past, researchers assumed that differential patterns of habitat use by new recruits were a result of active habitat choice, but recent studies indicate that habitat-specific predation in the hours to days immediately following settlement contributes to these patterns. Because most researchers had assumed that the factors affecting the mortality of new arrivals were the same as the factors affecting the mortality of settled larvae, habitat-specific mortality early in the settlement period in larval recruits went undetected for many years.



A major goal of studies of habitat selection in the ecological literature is to identify the types of habitat that are most suitable for the members of a population or species. The notion that differential habitat use reflects differences in intrinsic habitat quality rests on yet another assumption, namely that organisms accurately estimate the relative intrinsic quality of different types of habitat, so that preference and performance are positively correlated across different types of habitat. If this assumption is valid, then the relative abundance of organisms in a given type of habitat may provide useful information about the quality of that type of habitat, relative to the quality of other types of habitat in the same area. The assumptions outlined in the previous two paragraphs are actually quite similar to the underlying assumptions of the IFD. As a result, differential habitat use patterns are most likely to reflect habitat quality when the assumptions of the IFD are satisfied. Recall that the IFD assumes that individuals compete with one another while living in a habitat and that all of the individuals in a species are comparable with respect to their competitive ability. However, if individuals benefit from the presence of conspecifics after settling in a habitat, then the density of individuals in a given type of habitat need not reflect the relative quality of that type of habitat. For example, if individuals prefer to settle in the company of conspecifics, lower-quality patches that contain a moderate number of conspecifics may be more attractive to both local recruits and to potential immigrants than empty patches of higherquality habitat. Alternatively, if individuals differ with respect to their competitive ability, then highly competitive individuals may be able to exclude lesscompetitive individuals from higher-quality habitats. In this case, there is no guarantee that population densities will be higher in habitats of higher intrinsic quality, even if these habitats are preferred by every member of the species. The use of indirect cues in habitat selection can also disrupt relationships among population density, habitat preferences, and habitat quality. Even under the best of circ*mstances, the association between indirect cues and habitat quality is correlational rather than causal, so that organisms that rely on indirect cues will occasionally prefer lower-quality habitats by mistake. And because indirect cues provide only approximate estimates of habitat quality, organisms that rely on them are likely to have difficulty discriminating among habitats that do not differ very much with respect to habitat quality. Indeed, when organisms rely on indirect cues for habitat selection, ecologists with accurate estimates of habitat-specific mortality and reproductive rates probably have a better notion of the relative

quality of different types of habitat than do the organisms that are selecting those habitats. Although associations between indirect cues and habitat quality have always been imprecise, humans have contributed more than their share to the disruption of correlations between indirect cues and habitat quality. The recent literature on ‘‘ecological traps’’ considers cases in which a sudden environmental change (e.g., addition of a novel predator, altered habitat structure) has decoupled indirect cues from the true quality of the type of habitat that produces them. Most empirical studies of ecological traps have focused on situations in which humans are responsible for changing correlations between indirect cues and habitat quality. Examples include birds that preferentially settle in plantations of exotic trees rather than natural forests but that suffer lower nesting success in the former as a result of nest predation, or mayflies that prefer to lay their eggs on dry asphalt roads rather than ponds because asphalt reflects more of the polarized light that these animals use to choose oviposition sites. Hence, even if indirect cues used to be strongly correlated with habitat quality, there is no guarantee that this is still the case in today’s altered world. Another situation in which indirect cues can encourage mismatches between habitat preferences and relative habitat quality occurs when the attractiveness of indirect cues increases after exposure to those cues in the natal environment (NHPI). This is because NHPI encourages animals to select new habitats that are comparable to their natal habitat, even if other types of higher-quality habitats are available in the same landscape. Thus, NHPI may help explain situations in which animals raised in degraded habitats are reluctant to recruit to nearby patches of restored, high-quality habitat, or in which captive-raised or translocated animals fail to settle in habitats that are known to be of high quality for the members of their species. Even if we are willing to assume that organisms have perfect estimates of the intrinsic quality and the density of conspecifics at every habitat that is available to them, and that every individual prefers the same type of habitat, adding search to habitat selection further complicates relationships among habitat preferences, habitat quality, and population density. When organisms have to search for a habitat, preference is no longer the only factor affecting habitat choice. Instead, the optimal behavior for a given individual depends not only on the benefits of finding a high-quality habitat but also on the costs of searching for it. Thus, if patches of highquality habitat are rare and sparsely distributed, and if search is time- or energy-limited, then theory suggests that most of the individuals in a population will be relatively nonselective and, hence, likely to settle in

Habitat Selection habitats in proportion to their availability. As a result, habitat fragmentation and habitat degradation will not only reduce the amount of habitat that is available to support a population but also shift behavior in a direction that discourages individual selectivity and encourages individuals to accept habitats in proportion to their availability in the landscape. When this happens, an analysis of habitat use in relation to habitat availability might conclude (correctly) that individuals were not exhibiting habitat selection. However, it might also conclude (incorrectly) that individuals do not prefer some types of habitat to others or that all of the available habitats were of comparable intrinsic quality. A number of other factors that occur when animals search for habitats can affect relationships among habitat choice, habitat preference, and the distribution of individuals. For instance, species with small perceptual ranges may have difficulty detecting suitable habitats. If searching individuals run a strong risk of not finding any suitable habitat, and if low-quality habitats produce cues that can be detected at longer distances than the cues from high-quality habitats, then individuals should be differentially attracted to, and differentially settle in, low- rather than high-quality habitats. This scenario may help account for the fact that pest species such as aphids recruit to large expanses of agricultural crops rather than to isolated patches of their native host plants, even though those crops are less suitable for feeding and oviposition than the native hosts. The condition of the individuals who are selecting habitats may also affect relationships between preference and choice because, as was noted above, when time- or energy-limited animals are searching for habitats, individuals in poor condition are expected to be less selective during search than individuals in good condition. Hence, habitat degradation may not only reduce the survivorship and reproductive success of individuals who live in those lower-quality habitats but also produce individuals who lack the stamina or stored resources necessary to locate and settle in patches of higher-quality habitat. In conclusion, habitat selection behavior is scaledependent. Although simple habitat-selection models do a reasonable job of predicting individual behavior and spatial distributions involving habitat selection at smaller spatial scales, more complex models may be required to predict patterns of habitat selection at larger spatial scales. Recent theoretical and empirical studies of habitat selection at larger spatial scales have expanded traditional models to consider situations in which organisms benefit from the presence of conspecifics or heterospecifics after settlement, rely on indirect cues to assess habitat quality, or incur costs when searching for potentially suitable habitats. On the debit


side, this recent body of work provides a number of reasons why differential patterns of habitat use at larger spatial scales may not provide reliable estimates of either habitat preferences or intrinsic habitat quality. On the positive side, these recent studies have generated a number of new hypotheses about habitatselection behavior, some of which have already been supported in studies of habitat selection at larger spatial scales. This new body of work provides possible explanations for distribution patterns that have been observed in nature and offers suggestions that may help applied biologists manage the habitat selection behavior of species of concern to humans.

FURTHER READING The Ideal Free Distribution Flaxman, Samuel M., and Christina A. de Roos. 2007. Dif ferent modes of resource variation provide a critical test of ideal free distribution models. Behavioral Ecology and Sociobiology 61: 877 886. A recent study illustrating how the predictions of the IFD have been tested and validated for habitat selection behavior at small spatial scales. Fretwell, Steven D., and H. L. Lucas. 1970. On territorial behavior and other factors influencing habitat distribution in birds. I. Theoretical development. Acta Biotheoretica 19: 16 36. The classic paper that presented the initial model. Trengenza, T. 1995. Building on the ideal free distribution. Advances in Evolutionary Research 26: 253 307. A gen eral review of the topic.

Contributions that Consider Habitat Selection at Larger Spatial Scales Jones, Jason. 2001. Habitat selection studies in avian ecology: A critical review. Auk 118: 557 562. A review of habitat selection studies involving birds. Stamps, Judy A. 2001. Habitat selection by dispersers: Inte grating proximate and ultimate approaches. In J. Clobert, E. Danchin, A. A. Dhondt, and J. D. Nichols, eds., Dis persal. Oxford: Oxford University Press, 230 242. A review of factors affecting habitat selection behavior at larger spatial scales, with a focus on natal dispersers. Sutherland, William J. 1996. From Individual Behaviour to Population Ecology. Oxford: Oxford University Press. A book that illustrates the ways that IFD models and modifications of these models can be used to study habitat selection at small to intermediate spatial scales. Underwood, Anthony J., Gee M. Chapman, and Tasman P. Crowe. 2004. Identifying and understanding ecological preferences for habitat or prey. Journal of Experimental Marine Biology and Ecology 300: 161 187. A review of methods of studying habitat preferences with an emphasis on marine organisms.



The Use of Indirect Cues in Habitat Selection Johnson, Matthew D. 2007. Measuring habitat quality: A review. The Condor 109: 489 504. A review of methods for estimating habitat quality with a focus on birds. Robertson, Bruce A., and Richard L. Hutto. 2006. A frame work for understanding ecological traps and an evaluation of existing evidence. Ecology 87: 1075 1085. A recent review of empirical studies of ‘‘ecological traps’’: situa

tions in which indirect cues are no longer strongly corre lated with habitat quality. Stamps, Judy A., and V. V. Krishnan. 2005. Nonintuitive cue use in habitat selection. Ecology 86: 2860 2867. An overview of the use of indirect cues in habitat selection, with an emphasis of the use of indirect cues for habitat selection at larger spatial scales.

I.6 Dispersal Nicolas Perrin OUTLINE

1. Definition, patterns, and mechanisms 2. Evolutionary causes 3. Demographic and genetic consequences of dispersal 4. Measuring dispersal After a brief overview of the general patterns and the variety of mechanisms used for dispersal, this chapter delineates its evolutionary causes. Besides the spatial distribution and temporal dynamics of limiting resources, genetic structures resulting from mating or social systems play a role by affecting the potential for inbreeding and kin competition. Depending on conditions, however, dispersal may also have detrimental consequences at the population level, in terms of both demography and genetics. Finally, the chapter outlines recent developments in the way dispersal is measured.

GLOSSARY coancestry. Probability that two alleles sampled from

two different individuals are identical by descent. FST. A measure of genetic differentiation among populations, expressing the proportion of variance within a set of demes that results from the differentiation among them. genetic load. Decrease in average population fitness (relative to the fittest genotype) caused, e.g., by immigration of locally less-adapted immigrants (migration load), mating among relatives (inbreeding load), fixation of deleterious alleles (drift load), or any other population process. heterosis. Increase in fitness resulting from matings among individuals from different populations (as a result, e.g., of superdominance or drift-load effects). inbreeding depression. Drop in fitness resulting from the mating between relatives (caused, e.g., by recessive deleterious mutations).

local competition. Competition among relatives for

limiting resources (including mates). mass effects. Quantitative effects of dispersal on local

population dynamics. Emigration from a population may have negative effects on its demography, whereas immigration may have positive (rescue) effects. outbreeding depression. Drop in fitness resulting from the mating among distantly related individuals (from, e.g., the disruption of coadapted gene complexes). phoresis. Mechanism of dispersal by attachment of the propagule to another, actively dispersing organism. polygyny. Mating system in which a few males monopolize many females. propagule. Any part of an organism used for the purpose of dispersal and propagation. sink. Any population that consistently receives more immigrants than it sends emigrants. source. Any population that consistently sends more emigrants than it receives immigrants. 1. DEFINITION, PATTERNS, AND MECHANISMS

There are many ways to define dispersal. The simplest and possibly most appealing one might be to define it as the movement of organisms away from their place of birth. The crucial feature here is that dispersers do not reproduce where they were born. This opposes dispersal to ‘‘philopatry,’’ i.e., the tendency to reproduce at the natal place. Dispersal is referred to as ‘‘effective’’ when immigrants in a population contribute to local reproduction (i.e., when the rate of dispersal translates into a rate of gene flow among populations). Dispersal is a very ubiquitous feature throughout the living world, from bacteria to animals, including organisms that spend most of their life cycle in a sessile form (such as plants and fungi, but also many filterfeeding invertebrates). Dispersal patterns can be described by a ‘‘dispersal kernel,’’ which expresses settlement probability as a



function of distance from the source. The shapes of such kernels are obviously bound to depend on the dispersal mechanisms involved, which might be passive (e.g., transport by wind or water) or more active (e.g., flight). In the latter case, the kernel will also depend on behavioral strategies (e.g., random walk versus directed movement) and cognitive abilities in interaction with landscape features. In its simplest form, the kernel is an exponential negative function of distance from the source. However, even slight departures from this simple function might be of importance. Long-distance dispersers have a disproportionate impact on population processes, in particular during colonization events, by determining the rate of spread and the establishment of long-lasting genetic structures. Whether dispersal kernels have thin or fat tails (i.e., decrease faster or slower than an exponential) thus becomes an important theoretical issue. It is also one difficult to address empirically because long-distance dispersal events are rare and therefore often missed in mark-recapture experiments. Adaptations to dispersal are extremely diverse and often remarkably ingenious. Plants normally disperse passively by relying on currents (wind, water) or animals, both in the gametic (pollen) and zygotic (seed) dispersal phase. Pollen dispersal by animals involves complex interactions that are usually mutualisms (in which plants provide nectar to pollinators) but may include parasitism: more than one-third of orchid species do not provide their pollinators with either pollen or nectar rewards, relying on floral mimicry for pollination. Some species (e.g., the genus Ophrys) mimic the morphology and odors of female bees, inducing males to disperse their pollen through series of copulation attempts. As for the zygotic phase, adaptations to wind dispersal include seeds that resemble parachutes (e.g., the hairy expansions, or ‘‘pappus,’’ on dandelion seeds), helicopters (maple trees), or gliders. The tiny sizes of some propagules are also an adaptation allowing efficient dispersal by winds (e.g., the seeds of orchids or the spores of fungi). Adaptations to water dispersal are commonly seen in littoral plants. Flotation of the fruit allows seeds to be carried away on the tide or ocean currents. Examples include Rhizophora spp. (mangrove trees) and Cocos nucifera (coconuts), which may successfully colonize very remote tropical islands (note that this mechanism of dispersal also implies specific adaptation for the long-term maintenance of germination ability). Many plants have their seeds dispersed by animals and also build with them interactions that range from parasitism to comensalism to mutualism. The burs of burdock (Arctium lappa) and co*cklebur (Xanthium strumarium) are covered with stiff, hooked spines that

attach themselves to a passing animal’s fur so that the animal will carry them away. The American devil’s claw (Proboscidea louisianica) produces one of the largest hitchhiker fruits in the world, consisting of strange seed pods that attach to the feet of large herbivores. In mutualistic interactions, seeds are contained within a soft fruit adapted to animal consumption. Such fruits often present conspicuous bright colors when ripe (color vision in humans evolved as an adaptation to frugivory by our primate ancestors). These seeds have a tough protective outer coating so that although the fruit is digested, the seeds pass through their host’s digestive tract intact and grow wherever they fall. The viscid berries of mistletoe (Viscum album), a hemiparasite on trees, are deposited on potential host plants by the European mistle thrush (Turdus viscivorus) when cleaning its bill on branches after a meal. Such fruits attractive to animals are among the most successful adaptations related to seed dispersal. Other plant species have evolved mechanical means to overcome the tendency of a seed to drop close to its parent. Seedpods are built such that seeds are ejected away from the parent plant at maturation. Examples of explosive seed dispersal include cosmopolitan weeds such as Oxalis corniculata or Impatiens spp. As it dries, the capsule becomes sensitive to disturbance, ejecting tiny seeds in an explosive discharge. The fruit of the squirting cucumber (Ecballium elaterium) bursts when ripe, violently ejecting seeds together with a mucilaginous juice. Similar adaptations can be found in fungi: the spores of the dung-colonizing Pilobolus must go through a cow digestive system and come out with the dung to start the next generation. Because cows will not graze within a certain distance of dungs, spores have first to be dispersed away from their natal heap by an initial explosive event. Sessile or low-mobility animals have developed larval stages that function as dispersing propagules. Marine invertebrates display a great variety of planktonic larvae (planula in cnidarians, veliger in mollusks, trochophore in polychaets, zoe in decapods, pluteus in echinoderms, etc.) whose morphological adaptations and body expansions (e.g., setae) increase floatability and allow long-distance dispersal by marine currents. Specific behavioral mechanisms allow settlement in suitable places before developing into the adult stage. The aerial plankton similarly contains a diversity of dispersing propagules, including the juveniles of many orb-weaver spiders, ballooning via silk lines. The wind lifts lines along with the spider and floats it off to a new area. Phoretic behavior is also widespread. Pseudoscorpions, for instance, use their claws to grasp the hair of mammals or legs of insects for a lift. Dispersal is often quite active in many evolved invertebrates (cephalo-

Dispersal pods, crustaceans, insects) because of their good swimming, walking, or flying abilities. The best skills in terms of mobility are to be found among vertebrates, whose cognitive capacities also allow dispersal and settlement decisions to be fine-tuned to environmental or social conditions prevailing locally. Most animals, including vertebrates, tend to disperse first at the juvenile stage (natal dispersal), but many exceptions occur. A few others, such as aquatic insects, disperse as adults (the winged stage), flight being in this case the only way to disperse from pond to pond. Mayflies (Ephemeroptera), for instance, have very short adult lifespans (a few hours in some species), devoted entirely to mating and dispersal (adults do not feed). Dispersal is also often biased by sex. Males are usually the dispersing sex in mammals, whereas females tend to be the dispersing sex in birds. A few species have evolved a dispersal polymorphism, with morphologically distinct dispersing propagules. Within the Asteracea family, several genera (e.g., Leontodon, Heterotheca, Senecio) present two distinct types of seeds, the one aimed at long-distance dispersal being smaller, with a developed pappus. Some insects (mostly among Orthoptera and Hemiptera) also present a marked dispersal dimorphism with both long-winged and short-winged (or even wingless) individuals. This dimorphism is strongly marked in aphids (plant lice), which may display alternation of winged and wingless generations, depending on environmental conditions. Dispersal morphs have also been described in mammals (e.g., the naked mole rats Heterocephalus glaber). 2. EVOLUTIONARY CAUSES

Dispersal is costly. In addition to the fixed costs of building up specialized structures (including trade-offs involved), dispersers entail the energetic costs of crossing inhospitable habitats and associated mortality risks (up to 50% in small mammals). Settlers in new habitats may be at competitive disadvantage with residents, who benefit from a better knowledge of local areas and possibly help from relatives. Dispersal may also induce a migration load if immigrants are genetically maladapted to local conditions. Some benefits must therefore counteract these costs for dispersal to be evolutionary stable. Theoretical investigations in this field rely on the mathematical tools of game theory because the fitness returns of alternative strategies are frequency dependent (whether dispersing is the best decision also depends on what conspecifics are doing). Some evident benefits to dispersal accrue in dynamic landscapes, where resources display significant spatial and temporal variance. This is most obvious when patch qualities present negative autocorrelations, as


happens when suitable habitats have a determinate lifespan. Pilobolus fungi growing on a dung, or mycophagous Drosophila developing on the fruiting body of a mushroom, have to disperse their propagules because the patch currently exploited is bound to disappear soon. But dispersal may also evolve in the absence of such negative autocorrelations. Under random extinction of local habitats, a purely philopatric lineage will eventually disappear because the survival probability of an occupied patch declines asymptotically to zero. Thus, to ensure long-term survival, lineages have to send away a proportion of migrants, the optimal value of which depends on extinction rate. The same is actually true in any sort of dynamic landscape, even in absence of extinction. When new areas become open to colonization, dispersing lineages enjoy a higher fitness than purely philopatric ones. What about stable environments? Why leave a good local patch, even if crowded, if the other places you may reach are neither better nor less crowded, and you nevertheless have to endure the costs of dispersal? As it turns out, it may still pay to disperse in such conditions because of selective pressures imposed by local kin structures. Under complete philopatry, individuals would interact only with relatives. This might be disadvantageous in terms of inbreeding and also in terms of competition for mates or resources. First, mating with relatives often induces some inbreeding depression, expressed as a reduced offspring fitness, as a consequence of the load of deleterious mutations that accumulate in all populations. Gametheoretical models show that, were inbreeding avoidance the only reason for leaving a natal patch, then one sex only, either male or female, should disperse. Indeed, if males disperse, females can avoid the costs of inbreeding without having to pay those of dispersal. Avoidance of inbreeding (mostly self-fertilization) is certainly the only force behind gametic dispersal in plants, and observed patterns match these expectations (only male gametes disperse). In animals (which normally do not show distinct gametic and zygotic dispersal phases), dispersal is quite commonly sex biased, but both sexes usually disperse to some extent, implying that inbreeding avoidance is rarely the only cause of dispersal. Numerical simulations suggest that inbreeding avoidance may account for about one-third of the level of dispersal in stable landscapes. Second, a less obvious but potentially important cause of dispersal is the avoidance of local competition, which actually refers to competition with relatives. Competition is detrimental anyway, but competition with relatives is even worse. It affects not only the actor’s direct fitness but also that of the related



competitors, who share some of the actor’s genes. If one has to compete anyway, it is better to do so with unrelated individuals. Kin selection thus promotes dispersal as a way to increase inclusive fitness (by dispersing, the actor leaves one breeding opportunity to a relative). The relevant limiting factors can be trophic resources, territories, or mates. Local competition for breeding partners is thought to be responsible for the male-biased dispersal of many mammals, because their polygynous habits and female-defense behavior potentially induce a strong local mate competition. However, interactions with relatives may also bring benefits. Social structures usually emerge from cooperative interactions among kin. In many social mammals, including mice, helping among related females increases overall breeding success. These fitness benefits promote a strong female philopatry. In combination with polygyny, female philopatry boosts local relatedness because offspring within a group share the same father, and mothers are closely related. In turn, high relatedness increases the risks of inbreeding depression and local-mate competition, which promote male dispersal. Social species thus often display strong sex biases in dispersal. The combination of polygyny and heavily male-biased dispersal allows disentangling the dynamics of coancestry from that of inbreeding, which permits kin structures and social systems to develop without incurring inbreeding costs. 3. DEMOGRAPHIC AND GENETIC CONSEQUENCES OF DISPERSAL

When driven by the temporal fluctuations of local patch saturation, dispersal has the potential to positively affect regional demography, provided local fluctuations are not spatially correlated. This is truer when dispersal displays positive density dependence, being elicited by environmental cues linked to local saturation (e.g., resource depletion). In such a case, dispersal may dampen population fluctuations and prevent local extinctions through rescue effects. The positive effects of dispersal are obvious under metapopulation dynamics, when regional demographic equilibrium results from a balance between random extinctions and recolonizations of otherwise equivalent patches (see chapter II.4). Emigrating when resources become scarce also prevents local overexploitation of resources, with longlasting negative effects. Impeding emigration from resource-depleted patches may induce dramatic density cycles that might lead to extinctions, as observed in ungulate populations introduced on small islands. The positive effects of dispersal are less clear when local patches in the landscape show consistent and

predictable differences in quality. Dispersal then becomes asymmetric, with a dominant flow from good patches (sources) to bad ones (sinks). Although immigration may fuel low-quality patches and maintain local populations above carrying capacity through mass effects, emigration also imposes a load on sources that may threaten their long-term existence. Once sources are extinct, the whole system will rapidly collapse. The levels of dispersal normally favored by natural selection are unlikely to be optimal in terms of population survival. This is by no means surprising because selection operates at the level of individuals, not populations. The possibility actually exists, at least theoretically, that individual selection drives dispersal patterns that ultimately lead to the collapse of populations (evolutionary ‘‘suicide’’). As for genetic consequences, the main direct effect of dispersal is to hom*ogenize gene pools. Here also, consequences may be positive as well as negative. The positive side consists of a genetic rescue of isolated populations stemming from two causes. First, isolated populations are threatened by deleterious mutations, which occur in all natural populations at a rate estimated to be one mutation per genome and per generation (order of magnitude). Mutations with large effects are easily purged (particularly in large populations) and do not last for long. The main threats actually come from small-effect mutations, which may accumulate in small populations, where drift is strong enough to counterbalance selection. Deleterious mutations segregating within populations are responsible for the inbreeding load, revealed by a decrease in the fecundity or fitness of offspring from matings among relatives. Once fixed in a population, deleterious mutations do not contribute any more to inbreeding load (because they occur in all mating partners, independent of their relatedness) but to the so-called drift load, revealed by the enhanced fecundity or fitness of offspring from matings among partners from different populations (heterosis). Second, because of their small effective size and enhanced genetic drift, isolated populations are also threatened by a general loss of genetic variance, which jeopardizes their evolutionary potential. Connectivity with other populations may thus bring new genetic variance, thereby restoring their ability to respond adaptively to ecological changes. As a result of the interplay among drift, dispersal, and local selective pressures, a set of loosely interconnected populations might actually evolve more rapidly than a large panmictic population (Wright’s shifting balance). The negative genetic aspects of dispersal are referred to as migration load. This load may stem from the

Dispersal disruption of local adaptations, when migrants settle in habitats that differ from their natal habitat according to important environmental variables. This is more likely to happen when dispersal is passive and settlement is random (as opposed to active dispersal and targeting of favorable settlement sites). The migration load may also stem from the disruption of coadapted gene complexes (outbreeding depression). Small and isolated populations at the margin of species distribution have the potential to adapt to local conditions and, ultimately, to develop into new species. By hom*ogenizing gene pools, dispersal may thus prevent the local adaptation of these marginal populations and their evolution toward the exploitation of differential ecological niches and, so, ultimately counteract the dynamics of speciation and the building of biodiversity. The demographic and genetic consequences of dispersal obviously depend on specific features of both focal landscapes and study species, including their dispersal strategies and cognitive abilities. Individualbased modeling suggests that ‘‘blind’’ strategies (e.g., random walking) increase connectivity among patches, leading to low population structure (low FST), but, because of the high loss of propagules in the matrix, they are affordable only if propagules and dispersal are rather inexpensive. By contrast, ‘‘short-sighted’’ or ‘‘long-sighted’’ strategies (in which propagules are able to target favorable patches at some distance) are less costly in terms of matrix losses, but they induce stronger demographic (source-sink) and genetic structures because specific landscape features trap propagules into a restricted set of dispersal paths. Efforts to integrate genetic and demographic consequences of dispersal into a common framework are still scarce despite their potential importance for applied ecology and should certainly be pursued. The potential negative aspects of dispersal (both demographic and genetic), in particular, have to be borne in mind when devising management strategies for conservation biology (e.g., when building dispersal corridors or reinforcing local populations). 4. MEASURING DISPERSAL

Our ability to measure dispersal depends on the choice of appropriate methods to study its occurrence in natural populations. Ideally, a combination of both field observations and genetic methods is required to obtain a comprehensive picture of dispersal patterns and to make inferences about its proximate and ultimate causes. Field data provide valuable insights into the species social and reproductive behavior, which are essential to better understand the potential causes to


dispersal but usually do not allow quantification of how dispersal translates into gene flow because effective dispersal can be low even when there is high mobility. Moreover, for species that are particularly vagile, difficult to individually identify or to mark and recapture, estimating dispersal by direct observation is not always feasible. Genetic methods can be used to complement and reduce the invasiveness, effort, and expense of mark-recapture studies and give insights into how dispersal translates into effective dispersal and gene flow. An appreciation of the species life history is essential, at the very least to establish when dispersal is likely to occur. Because dispersal is often a juvenile trait, sampling juveniles provides access to predispersal individuals, whereas sampling adults provides a mixture of residents and immigrants. It is important to emphasize that these different cohorts should be analyzed separately. Doing otherwise would reduce our ability to gain insightful information from the contrast between pre- and postdispersal samples. Conventional genetic methods for measuring dispersal can be classified into those that measure either past gene flow or instantaneous dispersal. Both classically assume a population island model and simple population genetics framework. The simplest approach to estimate past gene flow builds on Wright’s formula for genetic differentiation between subpopulations, FST ¼

1 , 4Ne m þ 1

which provides an indirect measure of effective dispersal rate (Nem), a product of effective size (Ne) and dispersal rate (m). This, however, offers only an approximate solution because it relies on simplistic assumptions of island models, including stable populations of similar effective sizes and hom*ogeneous dispersal rate. Sex biases in gene flow can similarly be estimated using sexspecific markers. The mitochondrial DNA (mtDNA) is inherited maternally in most animals. Plants also usually transmit mitochondria and chloroplasts uniparentally, either paternally or maternally. Similarly, the Y chromosome in male-heterogametic species (and the W chromosome in female-heterogametic ones) is transmitted uniparentally. Because these markers are nonrecombining, information on historical patterns of gene flow is maintained in successive generations, and sexbiased gene flow can be inferred most simply through patterns of haplotype distribution or from the relative estimate of gene flow in males compared to females (obtained, e.g., by contrasting FSTY and FSTmt). It is important to stress, however, that the difference in FST



between Y and mtDNA may stem from differences in male and female effective population size as well as dispersal rate. A female-biased dispersal may combine with a highly polygynous mating system to generate a particularly small male Ne and a strong contrast in sexspecific genetic structure (FSTY FSTmt in hamadryas baboons, for instance). It is worth noting, however, that a large sampling variance is associated with Y and mtDNA markers because of their lack of recombination, so differences may partly stem from stochastic events. Recombining biparental markers (such as autosomal microsatellites) can also be used to estimate instantaneous dispersal, possibly in a sex-specific way. One approach builds on the contrast in FST between adults and predispersal juveniles. Assuming an island model of dispersal, the ratio of (sex-specific) FST estimated after dispersal over FST estimated before dispersal is a simple function of the (sex-specific) dispersal rate and can thus be used to estimate the proportion of immigrant individuals (males and females) in a subpopulation per generation. Alternative approaches rely on genetic assignment techniques, which calculate the probability of origin of a focal individual, given its genotype and the gene frequencies in potential source populations. Maximal power is usually achieved when dispersal rate is at an intermediate value (approximately 10% per generation). With high dispersal, a population will consist of a large proportion of immigrants, so that populations will not be differentiated enough. With low dispersal, immigrants constitute only a small proportion of the individuals sampled and may

not be detected at all. Individual-based assignment tests based on likelihood or Bayesian principles offer several advantages over summary statistics and should be more powerful because they do not average over the population, allow immigrant individuals to be readily identified, are more geographically explicit, and in the latter do not require populations to be predefined. Although individual assignment techniques based on Bayesian principles applied to multilocus genotypes are becoming standard tools in molecular ecology, their potential for studying dispersal has perhaps yet to be realized.

FURTHER READING Bowler, D. E., and T. G. Benton. 2005. Causes and conse quences of animal dispersal strategies: Relating individual behaviour to spatial dynamics. Biological Reviews 80: 205 225. Clobert, J., E. Danchin, A. A. Dhondt, and J. D. Nichols. 2001. Dispersal. Oxford: Oxford University Press. Clobert, J., R. A. Ims, and F. Rousset. 2004. Causes, conse quences and mechanisms of dispersal. In I. Hanski and O. Gaggiotti, eds., Ecology, Genetics, and Evolution of Me tapopulations. Amsterdam: Academic Press, 307 335. Levin, S. A., H. C. Muller Landau, R. Nathan, and J. Chave. 2003. The ecology and evolution of seed dispersal: A theoretical perspective. Annual Review of Ecology and Systematics 34: 575 604. Ronce, O. 2007. How does it feel to be like a rolling stone? Ten questions about dispersal evolution. Annual Review of Ecology, Evolution and Systematics 38: 231 253.

I.7 Foraging Behavior Joel S. Brown OUTLINE

1. 2. 3. 4. 5. 6. 7. 8.

Foraging behaviors, adaptations, and autecology Finding food Handling time To eat or not to eat? Patch use Social foraging Fear and foraging Coadaptations between foraging behaviors and morphology 9. Nutrient foraging in plants A need for energy and resources for survival, growth, and reproduction is a universal property of life. Hence, all organisms must forage. Even plants have noncognitive foraging behaviors. Life exhibits a wonderful diversity of feeding behaviors and associated morphological and physiological adaptations. Food must be found and handled. Letting the food come to the forager (sit and wait) or actively seeking food items (active pursuit) are two tactics for finding food items. Handling a food item may be as simple as absorption (endocytosis by a single cell organism) or a complex choreography of subduing, dismembering, and/or digesting a prey. Diet choice involves foragers deciding which food items to accept or reject. Patch use considers how thoroughly a forager should deplete the food from a spot before giving up and moving to a fresh spot. Foraging often occurs socially because groups permit sharing of information, scrounging, group hunting, task specialization, and, most often, safety in numbers. Predation risk and fear loom large in foraging, as animals balance the conflicting demands of finding food while avoiding becoming food themselves. All of these topics of foraging behavior become central to understanding an organism’s ecology and evolution.

GLOSSARY diet choice. The decisions made by foragers regard-

ing which encountered food items to consume and which to reject. The abundances of different food

types, their ease of finding and handling, and their value to the forager generally influence the decisions to eat or not to eat. foraging games. The behavioral challenges facing both predator and prey when the prey can perceive and respond to the hunting tactics of the predator, and the predator can perceive and respond to the antipredator tactics of its prey. These can be as straightforward as pursuit-evasion games; or complex sets of decisions summed up by when and where to forage; or levels of prey vigilance and predator boldness. Finally, foraging games such as producer– scrounger games or behaviors involving territoriality and interference may occur between members of the same species. nutrient foraging. The noncognitive foraging behaviors of plants as they adjust allocations to roots and shoots, alter uptake kinetics or growth forms to influence the uptake of water, light, nitrogen, and other nutrients. patch use. The behaviors of foragers regarding how to deplete the food items of a given spot. Most importantly, when should the forager leave an area with food before moving to a fresh area? A forager should leave a depleted food patch when the benefits of continuing to harvest the patch no longer exceed the sum of metabolic, predation, and missed opportunity costs of foraging. social foraging. When feeding occurs as groups of the same or different species. Social foraging may allow for information sharing, producer–scrounger games, group hunting, task specialization, and very often safety in numbers. Safety in numbers occurs through the many eyes, dilution, and confusion effects. 1. FORAGING BEHAVIORS, ADAPTATIONS, AND AUTECOLOGY

An animal’s ecology can be summed up as follows: where does it live, what does it eat, and who eats it? For instance, on the sand dunes of Bir Asluj in the Negev



Desert of Israel, afternoon winds blow in from the Mediterranean, redistributing the sand and uncovering seeds. At sunset, as the wind abates, Greater Egyptian sand gerbils (Gerbillus pyramidum) emerge from their burrows and move under shrubs or across open spaces to search for seeds. A knowledge of where seeds likely aggregate guides their search paths. With a keen sense of smell, gerbils hone in on patches of seeds or seeds buried under the sand. With its forepaws, the gerbil recovers seeds and lifts them to its mouth, either transporting them in internal cheek pouches or deftly husking them with practiced coordination of forepaws and incisors. While the gerbil seeks food, predators seek the gerbil. A gerbil’s ears and auditory system can detect the low-frequency sounds of a barn owl’s wingbeat. A gerbil’s quick reactions may save it from the strike of a horned viper, and erratic locomotion permits escape from a pursuing red fox. With cheek pouches full, the gerbil returns to its burrow and deposits the seeds underground in its larder, or it may save time by burying the seeds in a shallow depression, contributing another snack to its scatterhoard. Perhaps another gerbil will pilfer this cache before the owner returns for it. As the night draws on, the gerbils deplete the available seeds and conclude the night’s foraging. Most will return to their burrows to await another wind and another night. A few will have fed the predators. Central to the gerbil’s ecology are its foraging behaviors and the foraging behaviors of its predators. These behaviors have been engineered by natural selection through the circ*mstances of making a living as a seed-eating desert rodent. As a result, feeding behaviors are often the most frequent and tangible expression of an organism’s ecology. Feeding behaviors, as products of natural selection, emerge from all organisms’ need for energy and resources to sustain life, permit growth, and allow for reproduction. Foraging behaviors are as diverse and varied as life itself. We may describe them in broad brushstrokes such as a sperm whale diving 400 m beneath the waves and submerging for over 40 min as it somehow seeks giant squid. Or these behaviors can be described in fine detail as the exact path and number of steps that a browsing white-tailed deer takes as it moves from one specific shrub to another, stopping to nibble particular leaves from particular branches. Each bite may necessitate some number of chews before the masticated mouthful is swallowed. The precise choreography of step, bite, and chew continues for hours. But in each case, the foraging behaviors can be seen as contingent responses to environmental opportunities and hazards. As behaviors, we can describe the animal’s repertoire of specific actions that it uses to find and harvest food. As adaptations, we can ask why did

the whale dive to a particular depth and spend a particular time? Why did the deer favor some leaves over others? Was it the leaves’ nutrition, the presence of spines, or plant toxins? In what follows, we explore both of these aspects of foraging behaviors—the sequence of actions required to get food and the adaptive nature of feeding behaviors. Change the organism, and the suite of available behaviors likely changes too. The exact behaviors available to a single-cell Paramecium are literally worlds apart from that of a web-building spider. Change the environment, and the behaviors of a given animal may change dramatically. Bream, a common fish of northern European lakes, can opt to snatch zooplankton from the water when available, or they can probe for tasty detritus in the muck of the lake’s bottom. Here, the focus is on the categories of foraging behaviors and scenarios and the concepts that permit general understanding of foraging behaviors as adaptations. Animals decide where to forage. This topic is covered under Habitat Selection (chapter I.5). Once an animal is searching for food, the sequences of activities alternate between finding food and then handling food. The forager, as it finds and harvests food, faces decisions of ‘‘to eat or not to eat’’ and when to give up a food patch (patch use) as it depletes. One often imagines solitary foragers going about their business peaceably, but social foraging and fear and foraging recognize how foraging can occur in groups and occur under the threat of predation. Through coadaptations of foraging and morphology, there is a wonderful evolutionary feedback between adaptive feeding behaviors and the other physiological (see chapter I.2) and morphological traits of the species. Finally, plant nutrient foraging examines the noncognitive behaviors of plants to access nitrogen, phosphate, light, and water. Throughout, foraging behaviors emerge as adaptations (MacArthur and Pianka, 1966; Emlen, 1966) that permit organisms to acquire food and resources quickly, efficiently, and/or safely. 2. FINDING FOOD

The chemical reaction involving the polymerization of atoms and molecules is limited by the concentration of molecular building blocks and the rate at which these building blocks can be ‘‘found’’ and ‘‘consumed’’ by the growing polymer. Such was the foraging behavior of the first replicators at the dawn of life on Earth, as partially autocatalytic reactions built combinations of proteins and/or nucleic acids. Brownian motion within an aqueous solution allowed these protolife forms to find food. Three billion years later, finding food looms large in all sequences of foraging behaviors. Active-

Foraging Behavior pursuit and sit-and-wait tactics provide two evolutionary strategies for finding food. Single-cell archaebacteria and other prokaryotes likely evolved to be either free-floating or attached to stone surfaces. A sit-and-wait strategy demands less energetically and physiologically of an organism. It generally takes the form of filter-feeding small food items or ambushing large prey items. For it to work, the food must move, either passively as food particles within a water current or actively as mobile prey. Caddisfly larvae attached to stream pebbles extrude a mucus to which food particles become stuck. Many clams buried in the mudflats of the intertidal create their own water current by ‘‘siphoning’’ water through a tube, past a filterlike organ, and then back into the water column. Because the gerbils thoroughly scour the sand dunes each night for seeds, the horned viper can remain motionless, coiled, and ready to strike. Sit-and-wait foraging tends to promote foraging efficiency (reward per unit cost) over foraging speed (captures per unit time). Mobile foragers that actively seek and pursue their prey generally enhance speed at the expense of foraging efficiency. Sessile or slow-moving food strongly selects for actively moving and searching foragers. Gerbils must go find their seeds, as their seeds do not find them. The encounter probability (units of per time) is a key foraging parameter. It describes the likelihood of a searching forager encountering a given food item. The encounter probability depends heavily on the forager’s senses. Vision, smell, touch, pressure sensors, hearing, and even cuing in on electromagnetic distortions provide tools for encountering prey. Then there are the cues emitted by the food items or prey. Together, the senses of the forager and the cues of its food determine the forager’s detection radius for food and its likelihood of accurately sensing the food item. For gerbils, smell may provide a detection radius of *10 cm with touch concluding a successful encounter. Larger seeds are easier to detect than smaller seeds, and with humidity, seeds become more odiferous. Random search is the simplest case where a feeding animal’s likelihood of encountering any given food item is constant and independent of the total number of prey. This idealized condition becomes distorted when predators can observe the distribution and abundance of many food items in advance. A hummingbird moving among flowers or a black rhinoceros among acacia trees can be a ‘‘traveling salesman’’ and map a best route for collecting the food items—they can do better than random encounter. More prey items can enhance the encounter probability by drawing the forager’s attention, or it may challenge the forager with a confusion effect as multiple prey flee haphazardly at the predator’s approach. Finally, the prey themselves may


alter and distort encounter probabilities through camouflage, deception, and even direct signals to the forager that it has been detected. With aposematic coloration, dangerous or unpalatable prey (bees, monarch butterflies, coral snakes) communicate their unsuitablility as food items. Conversely, red flowers and intensely colored fruits attract the attention of hummingbirds and robins, respectively. 3. HANDLING TIME

Handling time describes the effort and activities required to harvest an encountered food item. In its simplest form, handling time can be the fixed time required for a gerbil to husk and consume a seed. More generally, it includes all of the effort required to subdue (if necessary), transport (if not consumed on the spot), prepare, and ingest the food. In animals such as a python or a ruminating antelope, handling time can also include a digestive pause that precludes searching for and handling additional food items. In gerbils, handling time may include caching behaviors. An emperor penguin’s handling effort can include marching to and from the colony to provision young. For predators, encountering the prey may be much easier than actually capturing it. Stanley Temple (1987) recorded how a red-tailed hawk had success rates of 28%, 18%, and 12% when initiating a strike on eastern chipmunk, cottontail rabbit, and eastern grey squirrel, respectively. Predators have additional foraging behaviors of stealth, pursuit, and tactics for killing the prey while avoiding injury themselves. Barn owls attacking gerbils appear to use hearing to encounter and initiate a strike while using vision to enhance the accuracy of the final impact. The handling behaviors of many foragers may include preparing the prey for consumption and choosing which bits of prey to consume. The gerbil can facilitate digestion by first husking and chewing each seed. Sparrows may ‘‘whirr’’ the wings off insects before ingesting them. If handling is time consuming, such as for a squirrel consuming a hazelnut, the forager may carry the food item to a safer, more comfortable setting. The animal recoups its preparation time by speeding digestion and increasing the efficiency of assimilation. Partial prey consumption, such as a scorpion consuming only the yummier parts of an isopod (sowbug), increases the quality of the ingested food. If small glass beads are mixed into a pile of seeds, a gerbil will harvest and pouch some of these beads along with the seeds. When licking up termites, an aardvark may consume more dirt than termites. Situations arise where there may be no advantage to taking the time to discriminate between good and bad food items.



Foragers will forgo recognition time when undesirable items are few and far between, relatively harmless to consume, hard to discriminate, and time-consuming to separate. Otherwise, foragers will invest time to distinguish among potential food items. Handling time can be as simple as the time taken to consume an item or a sophisticated choreography of time, effort, and risk. For a mosquito, handling time begins when the humming of its wings stops as it alights on your skin. She seeks a promising capillary bed within which to insert her proboscis. She injects a bit of anticoagulant (with luck, free of malaria!) and begins the process of gorging her stomach. All the while, she aims to avoid your wrath should you awake and claim her life. 4. TO EAT OR NOT TO EAT?

Diet choice is one of the fundamental consequences of adaptive feeding behaviors. Organisms do not consume different foods in direct proportion to their abundances in the environment. Feeding animals always appear more or less selective. Diet-choice studies show how foraging behavior results in a triaging of what is available to what is actually consumed. In all cases, the mapping of food availability into diet involves aspects of finding and handling food. Ronald Pulliam (1974) developed a classic model of diet choice based on the simplest assumptions of random search (constant and fixed encounter probability) and constant handling time. Search is undirected in the sense that the forager does not know what food type it will find until it stumbles on a food item. While searching, the forager cannot alter its encounter probabilities for one food relative to another (search images allow foragers to do this). To the forager, a food type can be characterized by the encounter probability, a, its abundance in the environment, R, its handling time, h, and its energetic value, e. Even this simple model suggests quite a bit. For instance, the likelihood that the next encountered food item is food 1 as opposed to food 2 is a1 R1 ⁄ (a1 R1 þ a2 R2 ). The forager should prefer the food with the higher energy-to-handling-time ratio. So, food 1 is preferred if e1 ⁄ h1 > e2 =h2 . To maximize its feeding rate, the forager should always consume its preferred item. But should it consume the less-preferred food? The answer is straightforward and simple. If the energy gain from handling an encountered item of the lesspreferred food, e2 ⁄ h2 , is less than what could be gained from searching for and handling a preferred item, e1 a1 R1 ⁄ (1 þ a1 h1 R1 ), then the forager should be selective. Otherwise, the forager should be opportunistic and consume all encountered items.

This model and its many variants suggest how animal diets represent a biased sample of availability. If the forager actually rejects consuming less-preferred food items, then diet choice is an extreme all or nothing. Increasing the abundance of its preferred food should cause the forager to reject less preferred items. A bountiful environment encourages picky eaters. When a forager is opportunistic and consumes all encountered food items, it will have a diet that appears to favor those foods that are easier to find (higher encounter probabilities). Cryptic foods will be underrepresented, conspicuous foods overrepresented in the diet. This is why gerbils will harvest a greater fraction of the large seeds than the small seeds from a given patch of sand. This effect of encounter probability on diet explains why flowers and fruits have evolved to be conspicuous (it is adaptive to be harvested) and why moths, stick bugs, and other prey have evolved camouflage (it is nonadaptive to be eaten). Biases in diets can result from foods occurring in separate patches or habitats. When foods occur apart, search is no longer random with respect to food type. It is now directed toward one food or the other, but not both. The forager may appear to favor one food over another simply because that food occurs in particularly rich patches or safe habitats. The state of the forager may alter its diet choice. For a mountain lion, mule deer are hard to encounter and successfully capture, but they pose minimal risk of injury to the mountain lion. Porcupines are the opposite. They are easier to encounter but they pose greater risks of injury to the mountain lion. Hence, a well-fed, successful mountain lion should eschew porcupines. But a down-and-out mountain lion should prefer to try its luck on capturing a porcupine rather than succumbing to the certainty of starvation. Bruce Patterson (2004) and others note this factor in the foraging behavior of large, man-eating cats. The man-eating lions of Tsavo likely switched diet as a consequence of prior crippling injuries. Nutritional relationships—substitutable, complementary, antagosnistic, and essential—among foods can loom large in diet choice. Foods may offer different essential or complementary combinations of carbohydrates, fats, proteins, minerals, and vitamins. Shy on salt, moose of the northern Great Lakes of North America favor a salt-concentrated plant. Moose along coastal Scandinavia lose interest in this plant because much of their diet automatically includes plants impregnated with Baltic sea salt. A balanced diet means that foragers appear to favor the rarer food type or the food type with the scarcer nutrient. This balancing of nutrients can apply to plant toxins as well. Different plant species defend themselves with different chemical

Foraging Behavior toxins such as tannins and oxalates. To an herbivore, it may be better to consume some tannins and some oxalates rather than a lot of just one—dose makes the poison, and feeding animals will often include this fact in their foraging behaviors. 5. PATCH USE

A jar of peanut butter or jam becomes increasingly frustrating and unsatisfying as the contents deplete. When full, a single swipe of the knife yields a bountiful spread. When mostly depleted, repeated strokes of the knife yield paltry returns. Eventually, we give up and discard the jar even though some contents remain. We share this dilemma of when to give up a depleted food patch and seek another with almost all feeding animals. Food items generally occur patchily, and the rate of food harvest declines as the patch becomes depleted. At what point should the forager abandon the patch, and how much unharvested food will it be leaving behind? Eric Charnov (1976) proposed the Marginal Value Theorem for how long to remain in a food patch. The forager should leave its current patch when its harvest rate within the patch no longer exceeds what the forager’s average harvest rate would be from leaving this patch, traveling to a fresh patch, and foraging that patch to the same quitting harvest rate. Put simply, leave a patch when the marginal rate of return (current harvest rate) drops to equal the forager’s average harvest rate from the environment at large. A forager should spend less time in a poor patch than a rich patch; a forager should spend less time in a patch of a rich environment than a poor environment; and a forager should leave patches sooner when travel time between patches is less. Foragers generally conform to these predictions, but with caveats. The costs and benefits of patch use may be more varied, and this has inspired variations and extensions of Charnov’s model. More generally, a forager should remain in a food patch until the benefits of harvesting resources, H, no longer exceed the sum of metabolic, C, predation, P, and missed opportunity, MOC, of foraging. Leave a patch when the harvest rate drops to H ¼ CþPþ MOC. If the forager’s harvest rate within the food patch is directly related to the remaining abundance of food within the patch, the animal’s patch use strategy also results in some amount of food being left behind. This remaining amount of food is referred to as the ‘‘giving-up density.’’ The size of this giving-up density should be proportional to the animal’s perceptions of foraging costs. The gerbils of Bir Asluj demonstrate how giving-up densities change with these foraging costs and benefits.


Gerbils have lower giving-up densities on foods that are more valuable, on foods that are easier to find, and within food patches that offer higher encounter probabilities on foods. Cold nighttime temperatures raise the gerbils’ metabolic rates (C), and consequently they have higher giving-up densities on cold nights than warm nights. Gerbils feel safer (P) and have lower giving-up densities when seeds are under shrubs than a few meters away in the open, and they have lower giving-up densities on nights with no moon than with full moon. When resources are abundant within the environment (MOC), or when the gerbil has large stores of food, it will forage to a higher giving-up density and exaggerate its avoidance of the risky, open microhabitat even more. Well-off animals have more to lose from being killed by predators than animals in low states of energy or well-being. The ability of foragers to detect and respond to variability in the distribution of food among patches is important for their ecology and their foraging behaviors. Perfect information on food availability allows the animal to perfectly balance its foraging time toward rich and/or safe food patches. Poor information on patch quality leaves the forager spending too much time in poor patches and too little time in rich patches. In reality, animals use sensory cues to ‘‘visualize’’ and assess patch qualities before investing time in the patch. Additionally, the foraging animal can use its experience within the patch to estimate patch quality. If the forager is having an easier time finding food than it expected, this may indicate a higher than average food abundance. Bayesian foraging studies how animals can use prior expectations and current experience to form an estimate of patch quality. The actual patch use behaviors of animals suggest that few have perfect information. Rather, foragers use a combination of preharvest sensory cues and sample information while foraging to form and update their estimate of patch quality. The gerbils deplete their food patches by actually harvesting the seeds. But for foragers that have prey that can run, hide, or become vigilant, patch use takes the form of behavioral resource depression. The mere presence of the predator causes the ‘‘patch’’ to become less valuable as prey flee or become more wary. For predators with fearful prey, the catchability of their prey becomes as important as the number of prey. Cows in a pasture enjoy a very different proximity to birds than does the Cooper’s hawk or Goshawk that aims to capture these birds. Patch use behavior, through the giving-up density, has important implications for the distribution and abundance of the forager’s food or prey. It may be that what we see in nature is simply the residue of feeding



behaviors. What we see may often be what the foragers care not to eat, cannot eat, or cannot catch. 6. SOCIAL FORAGING

Leaf-cutter ants coordinate foraging as ants in the tree canopy drop their harvest to the ground where others transport the leaf discs back to the colony. Hyraxes post a sentinel that allows the other hyraxes to forage less fearfully. To counter these social foragers, one black eagle of a pair may circle in one direction from the colony, permitting the other eagle to fly in from elsewhere and surprise the otherwise distracted colony. Pelicans, seagulls, and cormorants are famous for forming noisy aggregations around promising patches of schooling fishes. These are all facets of foraging in groups. They reveal the competing interests associated with task specialization, predator detection, group hunting, information sharing (or dissembling), and shameless scrounging. On first inspection, social foraging makes no sense. If a gerbil seeks to comb the sand dunes for seeds, doing so as a group simply means everyone has to walk farther for the same reward. When searching for food, better to divide the space and spread out. Hence, two critical factors loom large in social foraging—the forager’s prey is behaviorally responsive, and the foragers fear their own predator. Advantages to social foraging as an antipredator adaptation accrue from having alarm calls, sentinels, many eyes, the dilution effect (better to catch my neighbor than me), and the confusion effect (many fleeing foragers may distract the predator from capturing any one forager). When prey can flee or react, group foraging may permit task specialization (driving prey into an ambush), the ability to aggregate the prey (dolphins and whales corralling fish), beating the brush (banded mongooses moving abreast to scare up insects), or permitting the capture of large dangerous prey (army ants on vertebrates, wolves on a moose). Information sharing looms as a benefit and consequence of group foraging. Spotting where others have found food may reduce the entire group’s efficiency at finding food, but it may reduce the variance in food consumption. Less successful foragers join the feeding frenzy created by one forager stumbling on a particularly rich food patch. Some animals such as vampire bats and African hunting dogs will regurgitate and share food. Overly satiated members feed hungrier members. Such food sharing can even allow for task specialization where some individuals collect food even as others incubate a nest, protect a brood, or defend a territory from intruders. Of course, information sharing and gauging the successes of others introduces

conflicts of interest where an individual may prefer to scrounge rather than produce its own harvest. Social groups encourage freeloading and producer– scrounger games. It may be that in some groups where siblings help their parents raise offspring (Florida scrub jays, Arabian babblers, and other birds), the balance of the relationship rests on the willingness of the parents to tolerate their ‘‘adult’’ offspring so long as they contribute food for their newest sibs. Ant and bee colonies represent eusociality, the extreme of social foraging. These species exhibit caste systems, information sharing, group hunting or harvesting, and food sharing. What makes these systems special relative to a wolf pack or a naked mole rat colony? It may be the evolutionary objectives of the foragers that dictate the dividing line between a eusocial system and one that is merely a highly despotic social hierarchy. Individual worker ants and bees have been shown to forage in a way that completely subordinates themselves toward the fitness and success of the colony, whereas wolves and even individual naked mole rats seem to promote their own self-interests tempered by their need to be part of and treated well by the group. 7. FEAR AND FORAGING

Not a section of this chapter has gone by without some role for predators in shaping foraging behaviors. Foragers face a fundamental trade-off between food and safety. This trade-off becomes exacerbated and almost ensured by the adaptive behavior of having higher giving-up densities in risky than safe habitats. In most places and at most times, feeding animals face an environment in which background food abundance is high in risky habitats and low in safe places. A clever forager will use the tools of time allocation and vigilance to balance this trade-off. A clever predator will consider its prey’s behaviors when doing its own foraging. The reciprocal behavioral responses of prey and predators lead to studies of predator–prey foraging games. Games of fear and stealth abound in nature across all taxa and ecosystems. Steven Lima and William Mitchell have described the predator–prey shell game as prey seeking places free of predators and predators seeking to be where the prey are. The environment determines the form of the game. The afternoon winds at Bir Asluj ensure abundant seeds at dusk. This encourages gerbils to emerge early, which encourages clever owls to do the same. The responses of gerbils to seeds and of owls to gerbils create three temporal gradients. The seeds decline steadily as the gerbils deplete them. Gerbils start the night wary of owls and become increasingly less so as

Foraging Behavior the night draws on. The owls modulate their behavior to track the seeds—busy early and less so later. Foraging games can encompass several prey and predators. Owls encourage gerbils to forage more under shrubs than in the open. Snakes take advantage of this fear response by lying under shrubs in ambush. Owls and snakes create predator facilitation where the presence of one predator species makes it easier for the other to capture the shared prey. Furthermore, the nightly decline of seeds and risk promotes the coexistence of the Greater Egyptian sand gerbil with a smaller cousin, Allenby’s gerbil (G. andersoni allenbyi). The size, temperament, and behavior of the large gerbil suits it for early in the night, whereas the little gerbil has adaptations and behaviors more suited to the resource poor but safer periods of the night. Burt Kotler (1984), through ‘‘fear and foraging,’’ showed the role of predation risk in the foraging behaviors and coexistence of kangaroo rats and pocket mice at a desert site in Nevada. The behaviors of the prey may facilitate the coexistence of diverse predators, and the behaviors of predators may similarly promote diverse prey. Nowhere is this more likely than the reciprocal radiation of insects and plants. The feeding behaviors of herbivorous insects select for plant defenses. The evolution of additional insect species to overcome these defenses simply encourages the evolution of additional and more diverse defenses among an increasing number of plants species. So the game of feeding and defending promotes other morphological adaptations and perhaps even speciation and adaptive radiations. 8. COADAPTATIONS BETWEEN FORAGING BEHAVIORS AND MORPHOLOGY

Coral reef fishes offer a bedazzling array of sizes, colors, and shapes. Many of these species feed on corals or the algae that grow in them. Close inspection of these fish reveals delicate differences in the mouthparts, mandibles, and teeth. Like a tray of dental instruments, these varied mouthparts permit the different species to scrape algae from diverse surfaces, chew coral, and probe interstices within the coral for food. The body sizes, fin dimensions, and body forms of the fish serve to stabilize and maneuver the fish within the water column to permit access to food and escape from predators. We see a fine-tuned coadaptation of feeding behaviors, mouthparts, and other morphological attributes. But, what came first—the behavior, the mouthparts, or the body form? My doctoral advisor Michael Rosenzweig would tell us how ‘‘Natural selection can never adapt an organism to something it does not do.’’ A feeding be-


havior must then precede coadaptive changes in physiology and morphology. But the species’ prior physiology and morphology must at the very least allow for the behavior. This necessitates an important distinction between behaviors being selective versus opportunistic, and morphological adaptations as being specialist versus generalist (Rosenzweig, 1991). A feeding animal may be more or less picky in its selection of foods and/or places to feed. A North American robin may choose to feed selectively on insects or fruits, or it may opportunistically feed on both as they are encountered. When the robin is feeding just on insects, its gut modulates to enhance the digestion of insects at the expense of fruit, and vice versa when robins feed primarily on fruits. Finally, the body size and morphology of a robin make it adept at probing for insects in the soil and leaf litter, moderately apt at picking insects from branches and leaves, and quite unable to collect insects from under bark or by ‘‘flycatching’’ insects from midair. As natural selection engineers a fit between form and function, feeding behaviors or their absence can have profound consequences for the other traits of an organism. If a feeding opportunity arises, then a species previously nonadapted to this opportunity may acclimate by altering its foraging behavior. As this opportunity becomes an important part of its ecology, there will be selection on the species morphology and physiology to adapt. For instance, the cultivation of apples in the New World led to the apple-maggot fly evolving a new species. The precursor species inhabited native hawthorns. Those that switched to apples were now selected to fine-tune their breeding strategies to better match the flowering and fruiting phenology of apples. As the flip side of this same force, if a forager ceases to have a particular feeding opportunity, the absence of this behavior from its repertoire could lead to the loss of morphological and physiological adaptations aimed at improving the rewards from the now-absent behavior. Conserving a species may require us to preserve environments that maintain its full suite of feeding behaviors. 9. NUTRIENT FORAGING IN PLANTS

Plants forage too. They exhibit noncognitive behaviors and responses to light, nutrients, and water. Their ‘‘behaviors’’ represent allocation decisions and growth patterns. Their architecture and investment into roots contribute water and nutrients. Investment into aboveground leaves and stems influences carbon fixation. When viewed as nutrient foraging, most, if not all, of the principles and concepts of animal foraging behavior apply to plants—often with dramatic effect.



We often take wood for granted. Clearing trees created farmland and pastures. The wood itself could heat homes and power machines. As a building material it is sturdy, strong, and durable. The chair I sit in now is made from maple. Why is there wood? Competition for light. Nutrient foraging for light creates a special form of the tragedy of the commons. To be successful at having full sunlight, a plant need only be a bit taller than its neighbors. But if these neighbors respond in kind, an arms race ensues with ever greater and greater investment in sturdy, tall, woody trunks. What determines the canopy height? The costs and benefits of foraging. The benefit of being in the sunlight remains mostly constant because the available pool of light does not change with height (unless one gets demonstrably closer to the sun!). Yet the costs multiply with ever thicker trunks, greater surface area for pathogens and boring insect pests, greater mechanical challenges of transporting water to the canopy and photosynthates back to the roots, and greater chances of toppling over. As the trees play an evolutionary game of light competition, they achieve a canopy height at which no individual can benefit from being a bit taller and no individual is willing to concede light by being shorter. The environment-specific and tree species–specific adjustments of these costs and benefits produce 80-m-tall redwood forests and 30-m-tall European beech forests. Other strategies for light foraging abound. Light gaps encourage the lateral growth of branches and strange bends in stalks or stems. Maple trees will produce ‘‘sun-loving’’ leaves for their canopy and ‘‘shadetolerant’’ leaves for their subcanopy branches. Some plants will track the path of the sun with their leaves. Leaf size, morphology, greenness, and stem structure all contribute to the hugely diverse ways by which plants forage for light. A kind of producer–scrounger game happens when species of vines skip the investment in wood and simply achieve the canopy by growing up another’s trunk. The same holds for belowground nutrient foraging via roots. Plants may overproliferate roots with the goal of ‘‘stealing’’ nutrients from a neighbor. Of course, the neighbor is selected to respond in kind, and a belowground tragedy of the commons ensues. Roots show other varieties of noncognitive foraging behaviors. Plants will direct root proliferation toward areas of high nutrients. Plants may modulate root architecture (fineness of roots, density of root hairs) and root uptake

kinetics (ability to actively transport nutrients) in response to nutrient opportunities. The bargaining game between mycorrhizal fungi and plants presents an emerging frontier. Mycorrhizae are adept at concentrating nitrogen and phosphorus and then exchanging these with the roots of a plant for carbohydrates. To what extent is this symbiosis best modeled as a nutrient game? Elevated carbon dioxide levels in the atmosphere pose one of the greatest and most interesting challenges for the twenty-first century. Can nutrient foraging by plants play a role in understanding the dynamics of atmospheric CO2 and the concomitant climate change? This author thinks so. Whether animal or plant, universal aspects of feeding behaviors involve tactics for searching for and handling resources, foods, and prey. This process reaps rewards in terms of the value of the harvest and incurs costs that include the risk of injury or predation. The interplay between natural selection and the variety of environmental circ*mstances produces the myriad of foraging behaviors found among the millions of species inhabiting the planet. These behaviors allow foragers to seek and handle foods quickly, efficiently, and safely. The wind will blow, the seeds will redistribute, the sun will set, and the gerbils and owls will emerge to forage. FURTHER READING Charnov, E. L. 1976. Optimal foraging: The marginal value theorem. Theoretical Population Biology 9: 129 136. Emlen, J. M. 1966. The role of time and energy in food preference. American Naturalist 100: 611 617. Kotler, B. P. 1984. Predation risk and the structure of desert rodent communities. Ecology 65: 689 701. MacArthur, R. H., and E. Pianka. 1966. On optimal use of a patchy environment. American Naturalist 100: 603 609. Patterson, B. D. 2004. The Lions of Tsavo: Exploring the Legacy of Africa’s Notorious Man eaters. New York: McGraw Hill Professional. Pulliam, H. R. 1974. On the theory of optimal diets. Amer ican Naturalist 108: 59 75. Rosenzweig, M. L. 1991. Habitat selection and population interactions: A search for mechanism. American Natural ist 137: S5 S28. Stephens, D. W., J. S. Brown, and R. Ydenberg, eds. 2007. Foraging: Behavior and Ecology. Chicago: University of Chicago Press. Temple, S. A. 1987. Do predators always capture substan dard individuals disproportionately from prey popula tions? Ecology 68: 669 674.

I.8 Social Behavior Eldridge S. Adams OUTLINE

1. Ecological consequences of social behavior 2. The evolution of cooperation and altruism 3. Mechanisms of social behavior Social life is a mix of cooperation, altruism, and selfishness. In species as diverse as slime molds, army ants, and great apes, individuals coordinate actions to achieve common goals. Yet competition and conflict are common within social groups and may lead to lethal altercations. Consider, for example, a well-integrated, long-lived society such as a colony of the honeybee Apis mellifera. Cooperative foraging is organized by communication among the worker bees, allowing the colony to allocate effort flexibly over a large region surrounding the hive. By acting in concert, nestmates build combs, raise young, and maintain a comfortable nest temperature even through snowy winters. When the hive is threatened by a vertebrate predator, workers sacrifice their lives to protect the queen and her offspring, perhaps the most celebrated example of altruism among the insects. Yet the benefits of this collective activity are not evenly shared. Although there may be well over 15,000 females in the society, one of them—the queen—lays the vast majority of eggs while eggs laid by other females are quickly eaten. Other conflicts are evident. As winter approaches, males, which do no work, are dragged out of the hive and left to die. When a new queen is reared, she may sting to death other prospective queens still developing in their royal cells, bringing reproductive competition to a deadly conclusion. Like other social species, the honeybee prompts two central questions. How do organisms benefit from group living? What prevents conflicts within the group from undermining cooperative aspects of social life?

GLOSSARY Allee effect. An inverse relationship between popula-

tion density and per capita population growth rate. Allee effects can accelerate the decline of a shrinking population.

altruism. Behavior that is costly to the individual per-

forming it and is beneficial to one or more other individuals; costs and benefits are measured in terms of effects on fitness, which can be quantified by lifetime reproductive success. coefficient of relatedness. The probability that one animal shares an allele carried by another as a result of descent from a common ancestor. cooperation. Behavior that benefits two or more interacting individuals. kin selection. Selection resulting from the effects of an organism on the fitness of relatives, as well as through the organism’s own reproduction. policing. Actions by group members that suppress or punish selfish behavior by other group members. selfishness. Behavior that benefits the individual performing it at a cost to one or more other individuals. self-organization. In social species, this refers to phenomena in which group organization arises spontaneously, without central control, because of the actions and interactions of multiple individuals. 1. ECOLOGICAL CONSEQUENCES OF SOCIAL BEHAVIOR

Why do so many organisms live in groups? To behavioral ecologists, the abundance and diversity of social species suggest that in many environments the benefits of group living outweigh its costs. The forces driving sociality vary, but field studies have revealed a few principal advantages, which recur in diverse taxa. One set of advantages emerges in the context of foraging. By acting in groups, animals can improve foraging success through enhanced search, ability to overcome prey defenses, or ability to outcompete other groups or individuals. On the other side of the hunt, potential prey often seek to escape capture by clustering and moving together. Group activity makes it more difficult or more dangerous for a predator to attack, and potential prey can maneuver for positions in which they are less vulnerable than other group members. Shared vigilance



allows animals to spend more time foraging and less time watching for predators, increasing energetic efficiency. It is common for social predators to attack social prey, as when pods of dolphins hunt fish or squid, when packs of African wild dogs (Lycaon pictus) chase impala (Aepyceros melampus), or when columns of army ants overrun the nests of paper wasps. Other advantages of group living arise in interactions between parasites and their hosts. Parasites and pathogens can overwhelm host defenses by acting in concert, using chemical signals to synchronize attacks in space and time. Some ant species specialize on parasitizing other ants, organizing raids in which brood is stolen to augment the worker force of the raiders. Potential hosts can also employ social behavior to defend themselves. Social grooming, common in primates and social insects, helps to reduce parasitism and even to improve immunity. Coordinated groups are formidable competitors for limited resources, including food, nest sites, and opportunities to mate. Competition between social species is seen in faunas as diverse as social carnivores in African grassland and ant colonies in the canopies of tropical rainforests, where the number of allies may override individual fighting ability to determine who gains access to food. Moreover, by promoting repeated contacts among individuals competing for opportunities to mate, social life itself produces an environment favorable to establishment of competitive coalitions. Thus, male lions (Panthera leo), dolphins (Tursiops spp.), and wild stallions (Equus caballus), among others, form alliances to acquire or guard mates. Subordinate or bachelor males team up to gain access to mates defended by stronger individuals, and the dominant male in turn may recruit assistance to fend off challengers. In other contexts, group living benefits animals by improving the efficiency of movement (e.g., the Vformation of geese), because of thermal advantages of clustering (e.g., social hibernation in marmots), and by improving the efficiency of nest construction (e.g., paper wasps). Much social behavior occurs in the context of rearing young. Cooperative breeding, with some individuals playing a supportive role to others, is seen in some birds, mammals, fish, and snapping shrimp and in thousands of species of social insects. Similar principles underlie success in most of these examples: the collective effort of a group greatly exceeds that of a solitary animal; effectiveness of fighting is improved by outnumbering antagonists; sharing tasks permits animals to devote more time and energy to other needs; division of labor allows individuals to focus on complementary activities. In animal conservation, the population consequences of sociality create a special concern. Social life

can buffer a group against environmental change. Ants, for example, curtail colony growth when food is in short supply and can even eat their young, allowing the colony to live through the period of scarcity. However, species dependent on social strategies may be subject to Allee effects, which occur when the per capita rate of population growth is inversely related to density. When populations are sparse, animals may be unable to form cooperative associations of sufficient size, causing the population size to spiral downward. Field evidence shows that reduced group size can lead to colony failure. For example, in the highly social Damaraland mole rat (Cryptomys damarensis), small colonies perish during droughts because they lack the workforce needed to excavate underground tunnels leading to the storage roots that form the bulk of their diet. Allee effects can amplify the risk of extinction for social species when their populations suffer modest declines. 2. THE EVOLUTION OF COOPERATION AND ALTRUISM

Despite the advantages outlined above, group living often fosters competition for limited resources and opportunities to mate. There are five primary hypotheses for the evolution of helping behaviors, in which one animal acts to increase the survival or reproduction of others. Each hypothesis proposes a different viewpoint on the forces that keep competition in check. Mutualism

By helping another individual, an organism can help itself. When group activity allows an outcome that cannot be achieved by acting alone, then there is little temptation for an individual to cheat by declining to participate. Doing so dooms the entire enterprise. This form of cooperation is sometimes called ‘‘by-product mutualism’’ because cooperation results from each individual acting in its own best interest. Yet if the collective action requires coordination and communication among group members, it is not merely an accidental consequence of selfish actions. Furthermore, even though cooperative behavior may produce mutual advantages, this does not guarantee that the benefits are evenly shared. Suppose, for example, that cooperation is essential for capture of large prey. Conflicts can still arise over how the food is divided or because individuals that did not participate in the chase seek a share of the catch. Such discord is seen in the spider Amaurobius ferox, in which young must act in groups to capture crickets, which are larger than the spiders themselves. Kil Won Kim and colleagues showed that success is very unlikely without participation by several

Social Behavior spiders; nevertheless, conflicts and freeloading are common. When the cricket is comparatively small, the spiders are more likely to fight among themselves for opportunities to feed on captured prey, and when the cricket is much larger, spiders are more likely to join in feeding without taking part in the capture. Kin Selection

William D. Hamilton reasoned that an organism can promote the spread of its genes in two ways: directly, by producing its own offspring, and indirectly, by helping relatives to survive and reproduce. The sum of these two components is referred to as the organism’s inclusive fitness. The effectiveness of the indirect route is governed by the coefficient of relatedness, the probability that a randomly chosen allele carried by the helper is shared by the beneficiary as a result of descent from a common ancestor. Relatedness ranges from 0 for nonrelatives, to 1 for identical twins or members of the same clone. Changes in gene frequency resulting from both the direct and indirect pathways constitute kin selection. The most famous effect of kin selection is that it can lead to the evolution of altruism, behavior by which an individual helps another at a cost to itself. Consider, for example, a bird that must decide whether to expend effort raising a son or daughter, or to help its parents to produce one more offspring—that is, a brother or a sister. Foregoing breeding to help parents is common in dozens of species of cooperatively breeding birds, such as the Florida scrub jay (Aphelocoma coerulescens). In birds, as in most familiar vertebrates, the coefficient of relatedness between parents and offspring is 0.5, because each parent has a 50% chance of passing a particular allele to a random son or daughter. Yet the coefficient of relatedness between siblings in these species is also 0.5. Therefore, a bird produces as many copies of its own genes by raising a sibling as it would by raising one of its own offspring. Which option is favored depends on their relative effectiveness. A shortage of mating opportunities for young adults or greater efficiency of groups can tip the balance in favor of helping relatives. In Florida scrub jays, for example, suitable breeding habitat is very limited, so young birds, especially males, profit from remaining in their parents’ territories and helping to rear siblings. Eventually, a male may inherit all or part of the parental territory, so he derives both indirect benefits from raising siblings and direct benefits from acquiring a breeding territory. Kin selection is demonstrated by adaptive evolution of the sterile castes of social insects. In the termite family Termitidae, for example, the soldiers have evolved varied morphological adaptations for defense, some with sickle-shaped mandibles, others with a nozzle-like


extension at the front of the head used to spray chemicals onto ants or other attackers. But these soldiers do not reproduce, so how can their morphology or behavior evolve? Soldiers act to protect their parents, the queens and kings, which carry and transmit essentially all of the soldiers’ genes. More effective forms of defense are favored by kin selection, entirely through the indirect route of helping relatives. The calculus of kin-selected altruism is summarized by Hamilton’s rule. An animal is favored to perform an altruistic behavior if c, the cost to itself, is exceeded by b, the benefit to the recipient, multiplied by r, the coefficient of relatedness (c < br). This condition is more easily satisfied for close relatives than for distant relatives. Appreciation of the importance of the indirect route of gene transmission has transformed the way social behavior is studied. Estimating genetic relatedness among group members is now a standard part of the analysis of animal social behavior. Reciprocity

An animal may be favored to help another, at a cost to itself, if at a later time the roles are reversed. Robert Trivers termed this type of interaction ‘‘reciprocal altruism,’’ noting that the behavior is altruistic in the short run but cooperative in the long run. The chief requirements are that the benefit to the recipient exceeds the cost to the donor and that some mechanism protects against ‘‘cheaters’’ that accept help from others but then do not reciprocate. Game theoreticians have published hundreds of analyses of the evolution of reciprocity, yet there are few clear examples from nonhuman animals. Possible instances are seen in hermaphroditic species, such as the black hamlet, Hypoplectrus nigricans, a reef-dwelling fish. Eric Fischer argued that, in a mating between two hamlets, each individual is favored to supply the relatively inexpensive sperm rather than the larger and more costly eggs. However, if neither individual provides eggs, then reproduction cannot take place. Among black hamlets, this dilemma is solved by breaking mating into a series of bouts, in which the two fish alternate in the male and female roles. Black hamlets take longer to offer eggs to a mate that did not reciprocate in a previous encounter, a tendency that can protect against cheating. Coercion and Policing

In some societies, coercion suppresses selfish actions and promotes helping behaviors. Dominance hierarchies govern reproductive rates in many social groups, with aggression by those at the top inhibiting reproduction



by subordinates and sometimes inducing them to work. Helping behavior, in this case, is not entirely voluntary but rather is the best option remaining after choice of action has been restricted by dominant animals. In some social insects, attempted selfish behavior by a group member is kept in check by responses of other group members, actions known as ‘‘policing.’’ This has been best studied in honeybees. In a typical honeybee colony, the queen lays the vast majority of eggs. However, workers can lay unfertilized eggs, which, because of the unusual method of sex determination in honeybees, develop into reproductive males. Workers are more closely related to their own sons than to the queen’s sons but are more closely related to the queen’s sons than to the sons of other randomly chosen workers. Therefore, although each worker has an incentive to lay eggs herself, she should oppose egg-laying by other workers, preferring that reproduction be left to the queen. In fact, worker honeybees usually thwart attempts by other workers to produce male offspring. Worker eggs are distinguished from the queen’s eggs, probably on the basis of odor, and are eaten by other workers before they develop. Effective policing lowers the incentive for workers to lay eggs in the first place and so can promote the evolution of worker sterility. Group Selection

Cooperation and altruism can be promoted by the increased survival and reproduction of groups in which these behaviors are prevalent. This hypothesis invokes selection operating at the level of groups and in opposition to individual-level selection within groups. The group selection hypothesis fell into disfavor in large part because of George C. Williams’ influential book Adaptation and Natural Selection, published in 1966. Williams critically reassessed former claims that particular behaviors evolved for the good of a group or the good of a species. He emphasized that individual selection is more powerful than group selection and that most putative examples of group-selected traits turn out, on closer inspection, to be advantageous to the individuals performing them. However, further theoretical work showed that the conditions under which group selection can shape social behavior are not as restrictive as previously thought. Older models relied on differential survival of groups and required groups to be well separated, with very limited gene flow. The newer models rely more on differential reproduction and allow groups to be temporary. To highlight the differences in model structure, the newer models are said to represent ‘‘trait-group selection.’’ Although claims of group selection continue to stimulate objections and misunderstanding, the

controversy is resolved in part by recognizing that group selection can be formally equivalent to kin selection. In other words, in many cases where group selection works to promote altruism, altruists are on average related to the group members that they aid. The usefulness of the group selection perspective is that it provides a way to describe the effects of behaviors and selection at both the individual and the group level. For an example of group selection overriding selection for selfishness within groups, we can turn again to honeybees. Recall that most worker-laid eggs are eaten by other workers before they develop. Beekeepers have discovered a small number of colonies in which this arrangement breaks down. In these ‘‘anarchic’’ colonies, some worker bees have more highly developed ovaries and can lay male eggs that are much less likely to be consumed by nestmates, presumably because they smell like the queen’s eggs. These workers are able to escape policing and therefore to produce their own sons. This escape represents a selfish behavior by the anarchic bees, and one that could potentially spread rapidly because of the huge increase in individual reproduction that it allows. However, colonies with anarchic bees fare poorly, producing many males but few workers, and without active intervention by the beekeeper, most anarchic colonies perish. Thus, although evasion of policing leads to individual success within the colony, the colony itself is quickly doomed. 3. MECHANISMS OF SOCIAL BEHAVIOR

The mechanisms underlying sociality are as diverse as the animals and behaviors themselves, encompassing genes and development, endocrinology and neurobiology, communication and cognition. The ability to identify specific genes affecting behavior and to follow their action through the physiology and development of the animal has accelerated rapidly. To mention a single example, the behavior of mice is altered by insertion of the vasopressin receptor gene from a more social rodent, the prairie vole. Larry Young and colleagues showed that male transgenic mice respond to the hormone vasopressin by increasing social behavior toward females, a response normally seen in prairie voles but not in mice. From the standpoint of understanding the ecology and evolution of sociality, two categories of mechanisms are particularly important. Proximate Mechanisms for Coordinating Action

Coordinated social behavior originated among singlecelled organisms. Among some existing species of bacteria, combined action is triggered by ‘‘quorumsensing,’’ in which high densities are detected by the

Social Behavior buildup of signal molecules released into the external environment. When these signals reach a critical concentration, they stimulate profound changes in gene expression and cell behavior. This sensitivity to cell density allows the bacteria to secrete proteins only when their high concentration is likely to have a beneficial effect. For example, quorum-sensing in the pathogenic bacterium Staphylococcus aureus stimulates release of toxins at high population densities, allowing the bacteria to outcompete other strains and increasing their virulence toward the host. In large groups of animals, much of the communication needed to organize collective action is achieved by simple, anonymous responses. The characteristic movement patterns of flocks, schools, and herds result largely from simple rules by which individuals adjust their spacing, alignment, and speed relative to other nearby animals. If the group-level behavior arises without central control, from the local decisions and interactions of numerous individuals, the pattern is said to be ‘‘self-organizing.’’ Even when individual animals have incomplete information, self-organization can produce a collective intelligence allowing favorable decisions. Ant traffic, for example, can coalesce on the shortest of several available routes despite the fact that no individual directly compares alternative pathways. Use of the shortest route comes about as an automatic consequence of the way ants deposit and respond to chemical trails. The chemical signal is reinforced more rapidly on shorter routes simply because it takes ants less time to walk from one end to the other. Other ants are then drawn to the trail segments that are more strongly marked. At the other extreme, group processes rely on tight feedback between particular individuals playing different roles. For example, Redouan Bshary and colleagues showed that two species of fish use signals to coordinate cooperative hunts. One species, the grouper Plectropomus pessuliferus, hunts in the open water over coral reefs, causing prey to seek cover. The other, the giant moray eel, Gymnothorax javanicus, readily moves through crevices, causing prey to flee into the open. A grouper initiates cooperative hunts by a visual signal, shaking its head back and forth rapidly while facing a moray eel. The two fish may then hunt together for more than 30 min, and both benefit from an increased rate of capture as a result of their complementary hunting styles. Proximate Mechanisms Curtailing Cheating and Selfishness

The stability of some forms of social behavior, including reciprocity and altruism, requires that the


choice of partners be restricted. As discussed above, altruism can evolve by kin selection if the donor and the beneficiary are related. Directing altruism preferentially toward kin does not necessarily require any special cognitive abilities. Instead, the spatial or group structure of the population may ensure that animals interact primarily with relatives. Alternatively, animals may learn the characteristics of group members during early development and then offer helping behavior only to those animals that resemble this learned template. Because individuals who are close by during juvenile stages are likely to be relatives, this type of learning allows helping behaviors to be directed toward kin. Many studies have sought evidence of a more refined ability to distinguish degrees of relatedness among equally familiar group members, but the evidence is scant. In principle, relatedness can be detected by shared heritable tags. This phenomenon is known as the armpit effect if the animal develops the standard for comparison by learning its own phenotype (e.g., by sniffing its own armpit), or the greenbeard effect if individuals recognize which other individuals carry copies of the same gene for altruism because the gene also codes for an identifying label (e.g., a green beard). Reciprocal altruism does not require that animals be related, or even of the same species, but it does require protection against freeloaders, which accept help but do not offer it in return. Reciprocity is closely related to concepts of fairness, scorekeeping, and reputation building. The cognitive and emotional capacity to remember the past behavior of other animals and to respond with generosity or reprisal is well developed in some nonhuman primates. Frans de Waal and colleagues have documented these abilities in chimpanzees and capuchin monkeys, which share food with individuals other than offspring. Chimpanzees are more likely to share food with particular individuals from which they have recently received grooming and to respond aggressively toward those who have not groomed them. The evidence from capuchins goes even further. In controlled experiments, capuchins were offered food rewards for performing certain tasks, while also watching the rewards given to a paired monkey for the same performance. The monkeys were willing to perform the task for a low-value food item, such as a piece of cucumber, so long as both monkeys were given the same reward. However, when the partner received a food item of greater value, such as a grape, monkeys were less willing to perform the same task unless they too were given a grape. Social primates can base actions on the memory of previous encounters, comparing the reward received for a given effort to the rewards obtained by other group members. Like other social



animals, they can at once struggle to ensure that the group succeeds and to improve success relative to others within the group. FURTHER READING Barron, Andrew B., Benjamin P. Oldroyd, and Francis L. Ratnieks. 2001. Worker reproduction in honey bees (Apis) and the anarchic syndrome: A review. Behavioral Ecology and Sociobiology 50(3): 199 208. Camazine, Scott, Jean Louis Deneubourg, Nigel R. Franks, James Sneyd, Guy Theraulaz, and Eric Bonabeau. 2001. Self organization in Biological Systems. Princeton, NJ: Princeton University Press.

Courchamp, Franck, Tim Clutton Brock, and Bryan Grenfell. 1999. Inverse density dependence and the Allee effect. Trends in Ecology and Evolution 14(10): 405 410. Crespi, Bernard J. 2001. The evolution of social behavior in microorganisms. Trends in Ecology and Evolutionary Biology 16(4): 178 183. de Waal, Frans. 2006. Primates and Philosophers: How Mor ality Evolved. Princeton, NJ: Princeton University Press. Dugatkin, Lee A. 1997. Cooperation among Animals: An Evo lutionary Perspective. New York: Oxford University Press. Krause, Jens, and Graeme D. Ruxton. 2002. Living in Groups. New York: Oxford University Press. Seeley, Thomas D. 1995. The Wisdom of the Hive: The Social Physiology of Honey Bee Colonies. Cambridge, MA: Harvard University Press.

I.9 Phenotypic Plasticity Joseph Travis OUTLINE

1. 2. 3. 4.

Introduction The spectrum of phenotypic plasticity The evolution of adaptive plasticity The ecological importance of phenotypic plasticity 5. Horizons for future ecological research on phenotypic plasticity Phenotypic plasticity is the ability of an individual to express different features under different environmental conditions. Examples of plasticity surround us: plants have broader leaves when grown in shady conditions, and animals are smaller when they develop in crowded conditions. Although some of these changes reflect unavoidable consequences of adverse conditions, many of them are the product of natural selection molding an organism’s ability to survive and reproduce in a world whose conditions vary from time to time and from place to place. Put another way, many examples of phenotypic plasticity reflect the evolution of a developmental system that attempts to produce different traits under different conditions because no single trait is best suited for all conditions. Plasticity facilitates a species’ ability to occupy a variety of habitats, persist in uncertain environments, and stabilize its interactions with other species whose incidence and numbers change over time and across space.

GLOSSARY carapace. The hard outer shell surrounding the bodies

of small animals such as waterfleas and larger animals such as turtles. diapause. A state of arrested development in which the animal can survive long periods of challenging conditions such as low temperatures or drought by lying dormant. ectothermic animals. Animals that use external sources of heat for metabolism and whose rates of metabolism are closely linked to external temperatures, such as invertebrates, fish, amphibians, and reptiles.

fitness. The number of offspring an individual leaves

behind for the next generation; fitness has two major components, survival (or length of life) and reproductive rate. numerical stability. A steady-state equilibrium in population size, that is, numbers of individuals, to which a system will return if it is perturbed; stability in predator–prey systems refers to the numerical stability of both predator and prey that allows them to coexist indefinitely. phenotypic plasticity. The ability of an individual to express different features under different environmental conditions. 1. INTRODUCTION

Phenotypic plasticity is the ability of an individual to express different features under different environmental conditions. This ‘‘adaptive plasticity’’ is one of the most remarkable products of Darwinian evolution. For adaptive plasticity to emerge, the developmental machinery to build different traits must be integrated with a sensory system that detects reliable cues about the prevailing environmental condition so that suitable traits are expressed in a timely manner. Adaptive plasticity is an interesting topic for evolutionary biology, but it is also an important topic in ecology. One reason is that plasticity can enable a species to cope with highly seasonal environments or occupy diverse habitats. But more subtly, plasticity can have a substantial effect on a variety of ecological processes and thereby act as an important influence on which species we see where and at what population sizes. 2. THE SPECTRUM OF PHENOTYPIC PLASTICITY

Phenotypic plasticity can be either reversible or irreversible. The most obvious examples of reversible changes are behavioral responses to environmental conditions. For example, tadpoles change their foraging patterns in response to the presence of predators. When predators



are removed, the tadpoles adjust accordingly. Other well-known reversible responses include physiological changes such as the increase in mitochondrial density in terrestrial vertebrates in response to experiencing lower oxygen levels and the changes in specific fatty acids incorporated into animal cell membranes in response to changing thermal conditions. Morphological changes can also be reversible: the gills of aquatic salamanders increase or decrease in response to oxygen levels in the water, and vertebrate muscles change in form and density in response to the amount of use they receive. As one might expect, reversible plasticity appears when environmental conditions change, often within an individual’s lifetime. In most cases, individuals retain the ability to change their features for most of their lives. The exception to this rule is diapause in insects and other arthropods. Diapause is a state of arrested development in which the animal can survive long periods of challenging conditions such as low temperatures or drought by lying dormant. When conditions improve, the animal breaks diapause and resumes normal activity and development. A species can enter diapause in only one stage, for example, eggs in crickets and larvae in beetles, and once broken, diapause cannot be reentered. Irreversible changes occur trivially when an organism adjusts the timing of a life history transition in response to environmental circ*mstances. Once an annual plant initiates flowering in response to its lighting conditions, there is no going back. Less trivially, irreversible changes are reflected in features that, once expressed, are not altered regardless of how conditions may change. For example, waterfleas in ponds develop spines and a thicker carapace in response to the presence of a predatory fly larva in the water; once developed, the carapace is not altered appreciably even if the predators disappear. A species of African acacia develops long spines on its stems in response to being browsed by giraffes and elephants; these spines remain for the lifetime of the tree, even if it never suffers from additional browsing. Irreversible plasticity appears when environmental conditions are less volatile and less likely to change drastically within the lifetime of an individual. In many of these cases, there is a narrow window of development within which the individual is sensitive to the cues in the environment that trigger the expression of the feature. Outside of that window, the cues elicit no response. When these narrow windows of sensitivity exist, the individual is committing itself for the future in response to conditions in one relatively short period. Whether reversible or irreversible, plasticity is expressed in response to a wide range of environmental

factors. Some factors act ubiquitously; nearly all plants alter the expression of shoots, leaves, and flowers in response to variation in their lighting environments, and most animals alter development in response to variation in their thermal environments. Classes of biotic agents—predators, pathogens, potential competitors— also induce plastic responses. In some cases, the cue for the response is direct: the African acacia develops spines after it has been browsed. In others, the cue is indirect: waterfleas develop thicker carapaces in response to a chemical cue that alerts them to the presence of a larval midge predator, even before there is any attack on an individual waterflea. The many examples of plasticity in nature might suggest that just about any feature of an organism can be phenotypically plastic and just about any environmental condition can induce a plastic response. This is true if one looks at all of nature’s examples en masse; every trait responds to some environmental factor, and just about any environmental factor imaginable affects some trait in some species. But in a very important sense, it is not: plasticity can be quite specific. To be sure, there are general patterns of plasticity; nearly all ectothermic animals make larger eggs at lower temperatures. But the more striking observation is that the development of certain traits responds in specific species to specific cues; traits in a species that respond to one environmental agent may not respond to a different one, and the same features in different species may not respond to the same agent. Put another way, when one says ‘‘Trait X is plastic,’’ one needs to specify in which species and in response to variation in which environmental condition. There are several striking examples of this specificity. Damselfly species that coexist with fish behave differently in the presence of fish than in their absence, but species that do not coexist with fish fail to respond to their presence and are more likely to be eaten. Plasticity can even be specific at the population level; wild parsnip populations with a history of heavy herbivory respond to leaf damage by releasing compounds toxic to insect herbivores, whereas populations without a history of heavy herbivory do not. Even more subtly, plasticity can be quite precise. That is, a trait may respond only to a particular range of variation in an environmental factor, and the same trait in different species may respond to a different range of variation in that same factor. Insect diapause is a classic example: populations of the same species at different latitudes enter diapause in response to different combinations of temperature and day length. The specificity and precision of so much phenotypic plasticity suggest that it is not merely an ineluctable

Phenotypic Plasticity consequence of animal or plant physiology but a wellhoned evolutionary response to variable environments of a particular kind. 3. THE EVOLUTION OF ADAPTIVE PLASTICITY

Adaptive plasticity should evolve whenever individuals with the capacity to adjust their development to the prevailing conditions outperform, in the long run, individuals that express the same trait values or features constitutively, that is, regardless of condition. By ‘‘outperform’’ we mean ‘‘have a higher fitness,’’ that is, be more likely to survive or leave more offspring behind. The subtlety is in the phrase ‘‘in the long run.’’ In any single circ*mstance, the individual with the capacity to adjust its development to express the most suitable feature will perform just as well as the individual who expresses the same feature constitutively. But it will outperform all of the individuals who express unsuitable features constitutively. Individuals with the capacity to adjust development have high fitness in all conditions, whereas individuals with constitutive development patterns for the same set of features have high fitness in some conditions but low fitness in most conditions. In the long run, over many generations or many locations, individuals with the capacity to adjust development have the highest average fitness. To illustrate the argument, consider the waterfleas that develop a thicker carapace in response to the presence of a predatory fly larva. Developing a thicker carapace takes energy that would be used otherwise to accelerate maturation and reproduction. When predators are present, the thicker carapace repays the investment because it reduces the ability of the fly larva to capture and kill the animal before it reproduces. In the absence of the predator, the thicker carapace is a waste of energy because it detracts from the ability of the waterflea to get on with the business of maturing, mating, and reproducing. A waterflea that made a thin carapace regardless of conditions would do well in the absence of predators but poorly in their presence; conversely, a waterflea that made a thick carapace regardless of conditions would thrive in the presence of predators but do poorly in their absence. The waterflea with the plastic developmental system has the best of both worlds and, if predators are present at some times but not others, would, in the long run, have a higher average fitness than waterfleas that develop thick or thin carapaces constitutively. If plasticity is such an obvious advantage over constitutive development, why would developmental systems be anything but plastic when different features are suited to different conditions? The apparently


transparent advantage of phenotypic plasticity, as illustrated by the waterflea example, is based on three assumptions. The first assumption is that a reliable cue exists to inform the developing waterflea about the risk of predation from fly larvae. The second assumption is that there is no cost to plasticity; that is, the plastic developmental system produces a waterflea as fit as the constitutively thick carapace in the presence of fly larvae and as fit as the constitutively thin carapace in the absence of fly larvae. The third assumption is that each of the two conditions, presence or absence of flies, occurs with sufficient frequency that each constitutive development pattern often has the worse fitness. Clearly, adaptive plasticity cannot evolve if the assumptions are blatantly false. For example, if there were no cue about the presence of predators, then there is no way to ensure the morphology appropriate for the condition, and the waterflea may as well guess which morphology to express. But what if we relax but do not void the assumptions? Suppose that a cue exists but is not perfectly reliable. Suppose that there is a fitness cost to plasticity; that is, the plastic system makes a slightly thinner carapace in the presence of the predator than does the unconditional ‘‘thick’’ system (and so is not quite as fit as ‘‘thick’’ when flies are present) and a slightly thicker carapace in the absence of the predator than does the unconditional ‘‘thin’’ system (and so is not quite as fit as ‘‘thin’’ when flies are absent). And suppose that the two conditions, presence or absence of predatory fly larvae, do not occur with equal frequency. Now the prospects for the evolution of adaptive plasticity depend on complicated relationships among the reliability of the cue, the cost of plasticity, and the evenness in frequency of the two conditions. The waterflea example can illustrate this complexity. Consider what happens when only one condition is very common; perhaps predatory fly larvae are almost always abundant. In this case, the individuals expressing the thick carapace are likely to prevail because they are the fittest individuals nearly all of the time. For plasticity to persist, individuals carrying the plastic developmental system must have a tremendous fitness advantage over the individuals expressing thick carapaces constitutively when predatory flies are absent in order to make up for their comparative deficiency in fitness when flies are present. The greater the cost of plasticity when flies are present, and the more often flies are present, the greater the advantage the plastic waterfleas must have when flies are absent. For a specific set of fitness relationships, the higher the variability in environmental circ*mstances, the more likely that plasticity in development will emerge



as a successful adaptation to that variability. However, this rule of thumb is valid only to a point. When conditions change too quickly, cues become unreliable, and plasticity does not improve on constitutive development or even random expression of features. This is especially true when plasticity is irreversible and the sensitivity to cues is restricted to a short period during development. If the environment changes faster than the time between the sensitive period and the expression of the appropriate feature, then plasticity is actually deleterious because it will perform worse than random expression of features. Adaptive phenotypic plasticity enables individuals to cope with circ*mstances that vary from time to time and place to place but are not so variable as to preclude reliable cues to guide development. This enabling of individuals propagates upward to the level of the population and beyond to produce some important ecological consequences. 4. THE ECOLOGICAL IMPORTANCE OF PHENOTYPIC PLASTICITY

The obvious ecological consequence of phenotypic plasticity is that it allows a species to expand its range to seasonal environments and diverse habitats. A seasonal environment is the ideal situation for the evolution of plasticity; seasons change frequently enough to promote reversible plasticity but not too frequently compared to the time scale of trait expression, reliable cues abound, and many of the features of different seasons are predictable. Nearly everyone is familiar with the many adjustments that plants and animals make to the changes of season in temperate regions from the physiological changes underlying migratory behavior in birds to those underlying the onset of winter dormancy in trees. Phenotypic plasticity can also allow species to occupy very uncertain habitats. Temporary ponds offer an example; the regular drying of the pond precludes sustainable fish populations, but the duration of the pond is uncertain, depending on the amount and timing of local rainfall. Nonetheless, temporary ponds harbor a considerable diversity of aquatic animals. Ponds offer refuge from what would otherwise be devastating predation by fish. But the dry periods would seem to preclude continuous occupancy by completely aquatic animals, and a short pond lifetime can leave the aquatic stage of animals that spend only part of their time in the water, such as tadpoles and dragonfly nymphs, high and dry if they cannot metamorphose quickly enough. Species that inhabit temporary ponds show remarkable varieties of phenotypic plasticity in response to drying conditions. Some copepods produce diapausing eggs

that rest in the soil, many of the frog and salamander larvae can accelerate their development as waters recede, and sirens (large, completely aquatic salamanders) burrow into the soil, secrete a waterproof cocoon around their bodies to prevent desiccation, and enter estivation until the waters return. Habitats can also be uncertain in their biotic components, and plasticity in response to the risks of predation and parasitism enables a species to cope more effectively with varying levels of risk. Temporary ponds exemplify this situation as well. Not only is their duration uncertain, but so is the period between drying and refilling. When the pond refills soon after drying, it is colonized quickly by predaceous insects including dragonflies and backswimmers. The aquatic larvae can achieve very high densities by the time that tadpoles appear later in the season. But if the ponds are dry for a long time, tadpoles have little risk of predation because the insects are at very low densities and are very small in body size. Many tadpoles from temporary ponds display extensive phenotypic plasticity to the presence or absence of predators. Most species change their activity patterns to reduce their encounter rate with predators, and some species alter their tail coloration and morphology to avoid predator detection and escape predator attack. But an example like this one raises an interesting question: if an organism evolves adaptive plasticity in response to variation in predation risk, does the advantage conferred by that plasticity have a reciprocal effect on the predator? This general question is at the heart of the close scrutiny that ecologists have been giving many examples of phenotypic plasticity. Indeed, reciprocal effects on predators or other biotic agents that induce plastic responses have been found in many studies and can ramify through a community and an ecosystem, with far-reaching consequences. To visualize this point, consider the tadpoles and dragonflies again. If the dragonflies are less able to procure tadpoles as food, they will increase their consumption of other prey such as aquatic invertebrates and cause their densities to decrease. Other predators in the system, which had been using aquatic invertebrates as their principal food resource, may then be forced into other trophic pathways. In effect, the adaptive plasticity in the tadpoles, once established, might drive a substantial change in species diversity, community structure, and perhaps even ecosystem processes such as nutrient cycling. This kind of effect has been found in many cases, and the indirect effect of one species on another, mediated through the consequences of expressing a feature that is a response to a third species, is often called a trait-mediated interaction. In our example, the

Phenotypic Plasticity decreased density of aquatic invertebrates represents an indirect effect of the tadpoles as they express the tail morphology that reduces their mortality rate from dragonfly predation. Trait-mediated interactions have been shown to be responsible for some interesting patterns of species diversity. For example, the presence of spiders in a New England old field causes several of their potential insect prey species to find refuge and foraging substrate on different plants than they would exploit in the absence of spiders. The plant preferred in the presence of spiders is actually a dominant competitor, and grazing by the insects reduces its density sufficiently for a competitively inferior species to increase in its density. The end result is that the presence of the spider increases the species diversity of the plant community. A growing body of mathematical theory has elaborated on these basic ideas, indicating potentially profound effects of plasticity on species interactions. Much of this theory has been inspired by a particular type of adaptive plasticity, the inducible defenses of plants. Induced defenses are morphological or chemical responses by plants in response to herbivore attack. The production of toxic chemicals in some populations of wild parsnip in response to herbivore damage is an example of an induced chemical defense. Induced chemical defenses are known in a wide variety of plants, from freshwater algae to trees. Although the defensive compounds produced by plants can be synthesized and deployed relatively quickly, they can be costly to manufacture, diverting energy away from other functions. If the risk of herbivory is high, plants that produce them have higher fitness than those that do not; if the risk is low, chemical defense production is a waste of energy. Analogous to the argument for the carapace thickness of waterfleas, inducible defenses are favored when herbivory is sufficiently variable and a reliable cue is available (and being chewed is usually a reliable signal that herbivores are active). Models inspired by inducible defenses indicate that adaptive plasticity can stabilize the numerical relationship between predator and prey or herbivore and host. To see this without mathematics, remember that predator–prey systems are inherently unstable because predators tend to overconsume prey. Any feature that protects a minimum fraction of the prey population from the predator can stabilize the system and allow predator and prey to coexist. Consider a herbivore– host system in which a constitutive defense appears via mutation. When this defense is expressed in some of the plants, it will protect a minimum fraction of individuals and stabilize the system. But as it spreads so that nearly all plants are protected, the herbivore loses its food resource and is likely to suffer a serious drop in


population size and perhaps even extinction. Now consider an inducible defense that is expressed only when the risk of herbivory is high. Initially, when the inducible defense is present in only a few plants, it stabilizes the interaction. As more individuals express the defense, the herbivores become food-limited, and their density starts to decrease. But as herbivore densities decrease, so does the risk of predation; fewer individuals will express the defense, leading to a greater opportunity for the herbivores, whose density can then increase. Eventually, the herbivore and plant populations reach equilibrium, and the proportion of plants expressing the defense also attains equilibrium. An experimental study of algae with and without inducible defenses has confirmed that inducible defenses can stabilize herbivore– host systems and even stabilize a system with three trophic levels: host, herbivore, and predator. But theory shows that adaptive plasticity will stabilize a dynamic predator–prey or herbivore–host system only if prey respond to the cue—predation risk high or low—with just the right speed, compared to the rate at which predators or herbivores can change their consumption rate. Obviously, a response that is too slow will be ineffective at deterring predation. A response that is too fast introduces a time lag between the appearance of the defense and the effect on the predators that destabilizes the system. There are too many predators when prey are well defended and too few when they are not. Systems like this will start cycling in numbers to the point where either the prey or the predator becomes extinct. Whether rapid plastic responses actually destabilize species interactions is one of many empirical questions about adaptive plasticity that remain to be answered. 5. HORIZONS FOR FUTURE ECOLOGICAL RESEARCH ON PHENOTYPIC PLASTICITY

The most important of the longstanding unresolved issues is the cost of plasticity. This is a difficult problem. It is rare to find both constitutive and plastic expression of suitable features in one population, so it is usually not possible to make the appropriate comparisons of fitness. The most common experiments that attempt to measure the cost of plasticity compare families that differ in their levels of plasticity. The results have been equivocal; some experiments have detected apparent costs, but others have not. The tantalizing prospect of using genetic engineering to create constitutive expression offers considerable promise for resolving the magnitude of costs and whether those costs occur similarly in all environments. The enthusiasm for studying trait-mediated interactions has produced an extensive documentation of



their existence and immediate effects. But in most cases, we do not know enough about the precision with which the traits are expressed, the relative frequencies of the different circ*mstances that provoke different expressions, or the full extent of the indirect effects that emerge in the community. We know that plasticity can have profound effects, but we do not know whether the documented cases of profound effects are exceptional. Although we know a great deal about which factors induce plastic responses, we know far less about the actual cues that organisms exploit. Delineating those cues is important for illuminating their reliability, which is a critical feature governing plasticity’s evolution and persistence. But there is another reason to identify the cues. Global change, sensu lato, could make erstwhile reliable cues unreliable, perhaps by dissociating combinations of signals that had been serving as very reliable cues. There is some evidence that this is happening in diapausing insects and migratory animals that use combinations of temperature and day length as their cue. The mysteries of what we do not know about phenotypic plasticity should not detract from the marvel of what is well known. Through adaptive plasticity, an organism can remake itself, within limits, to suit its circ*mstances. And the organism that remakes itself to suit its circ*mstances can also remake the ecological circ*mstances around it, creating myriad possibilities for itself and for those who would understand the distribution and abundance of organisms. FURTHER READING Bradshaw, A. D. 1965. Evolutionary significance of pheno typic plasticity in plants. Advances in Genetics 13: 115 155. This article remains the single best essay on the entire subject. In this essay, Bradshaw describes phenotypic plasticity and distinguishes it from related ideas in the literature, tracing its intellectual history accurately from a letter of Charles Darwin in 1881 to the scientific literature of the early 1960s. Further, this article produced the technical terms still in use today, and Bradshaw’s con cluding section on research horizons helped determine the research on plasticity for several academic generations. Bradshaw cited a large number of examples, mostly but not entirely from plants, to support his claim that there were patterns in plasticity and that ‘‘plasticity is therefore a property specific to individual characters in relation to specific environmental influences.’’ He discussed the types of variable environments in which one would expect to find plasticity, and his reasoning presaged the results of more sophisticated mathematical theory that would emerge over two decades later. DeAngelis, D. L., M. Vos, W. M. Mooij, and P. A. Abrams. 2007. Feedback effects between the food chain and in duced defense strategies. In N. Rooney, K. McCann, and D. Noakes, eds., From Energetics to Ecosystems: The

Dynamics and Structure of Ecological Systems. New York: Springer Verlag, 213 236. This report is among the most recent mathematical investigations of how induc ible defenses can affect the stability of predator prey or herbivore host systems and, in a larger context, the re sponses of individual species and the ecosystem to nutrient enrichment. The discussion section of the article offers an excellent introduction to the literature on mathematical models of the consequences of plasticity for those inter ested either in further reading or, especially, initiating research on the subject. DeWitt, T. J., and S. M. Scheiner, eds. 2003. Phenotypic Plasticity: Functional and Conceptual Approaches. New York: Oxford University Press. This edited volume in cludes a broad range of papers that, together, cover every facet of the subject from the varieties of plasticity in nature to what we know (or knew in 2003) about the genetic control of plastic development. Readers who are consid ering initiating research in the broad area of phenotypic plasticity should use this volume as their road map to its current research horizons. A virtue of this collection is the significant number of essays by younger workers with fresh perspectives. Karban, R., and I. T. Baldwin. 1997. Induced Responses to Herbivory. Chicago: University of Chicago Press. This is a very readable monograph that reviews and synthesizes the literature on the varieties of inducible defenses in plants. The text brings theory, as it existed at the time, to bear on the diversity of ways in which plants respond to herbivory, and its wealth of examples still serves as a readable and effective introduction to the topic. Kats, L. B., and L. M. Dill. 1998. The scent of death: Che mosensory assessment of predation risk by prey animals. Ecoscience 5: 361 394. This is an underappreciated re view paper that is focused on the diversity of chemical signals used by animals to assess predation risk and cue antipredator plasticity in a variety of traits. It is one of the few reviews in the ecological and evolutionary literature devoted primarily to a serious, thoughtful examination of specific cues and the all important theoretical issue of their reliability. Miner, B. G., S. E. Sultan, S. G. Morgan, D. K. Padilla, and R. A. Relyea. 2005. Ecological consequences of phenotypic plasticity. Trends in Ecology and Evolution 20: 685 692. This short paper is one of the few review papers focused specifically on the ecological consequences of plasticity and argues for its importance as an ecological topic, not merely a topic in evolutionary biology. It is focused pri marily on the effects of plasticity on species interactions and less on how plasticity enables habitat breadth. The paper and its bibliography offer an introduction to the recent literature on the various aspects of trait mediated interactions and the effects of inducible defenses. Pigliucci, M. 2001. Phenotypic Plasticity: Beyond Nature and Nurture. Baltimore, MD: The Johns Hopkins University Press. This book is a recent synthesis of the evolution of plasticity, and Pigliucci’s advocacy for thinking about integrated developmental systems is, in some ways, a modern counterpart to Schmalhausen’s book. One of the

Phenotypic Plasticity book’s strengths is its treatment of modern theory for the evolution of plasticity; the text offers lucid explications of some very complicated ideas, many of which have their origins in sophisticated mathematical theory, and clarifies the relationships among different theoretical approaches and the results of individual papers. Readers interested in a comprehensive introduction to the theory for the evo lution of plasticity should read the treatment in this book. Schmalhausen, I. I. 1949. Factors of Evolution. Philadelphia: The Blakiston Company. Reprinted Chicago: University of Chicago Press, 1986. This classic monograph empha sizes the evolution of integrated development systems for organisms. Schmalhausen took a unified view of evo lutionary development, placing plasticity in the same conceptual context as its opposite, canalization, which is the process of minimizing the variation in development so as to produce the same features or trait values regardless of environmental conditions. He discussed how and when evolution might take each course and set these ideas firmly in the context of what were, at that time, modern ideas in evolutionary genetics. The book still offers a compelling argument that developmental systems are adaptive evo lution’s most breathtaking product. Shapiro, A. M. 1976. Seasonal polyphenism. Evolutionary Biology 9: 259 333. This underappreciated review is a very thoughtful treatment of seasonal variation in mor phology, coloration, and life history, with some close at tention to insects. The text touches on the major themes in the evolution of plasticity and, despite its age, remains an excellent source of ideas and a laudable example of how to synthesize natural history, conceptual issues, and data. Sumner, F. B. 1932. Genetic, distributional, and evolutionary studies of the subspecies of deer mice (Peromyscus). Bib liographica Genetica 9: 1 106. This is a classic paper that summarizes and synthesizes Sumner’s decades of study of deer mouse ecology, genetics, and development. Sumner took the integrated approach to ecology and evolution that is so often proclaimed but so rarely practiced. The paper discusses how local adaptation (genetic differences


produced by Darwinian adaptation to local conditions) and phenotypic plasticity combine to allow deer mice to occupy diverse habitats. His experimental dissections of phenotypic variation into its genetic, environmental, and interactive components remain models for modern emu lation. Tollrian, R., and C. D. Harvell, eds. 1998. The Ecology and Evolution of Inducible Defenses. Princeton, NJ: Princeton University Press. This volume offers a comprehensive look into its subject, and the papers included in the volume examine topics from the biochemistry of defensive com pounds to trait mediated interactions. Although the pa pers were written before the recent flowering of mathe matical theory for the consequences of plasticity, the ideas that those theories examine are set out in several of these papers, and the volume clearly played a role in acceler ating this area of research. For a reader interested in the variety of induced defenses, this volume offers a strong introduction to a very diverse literature. In addition, the authors of individual papers come from several schools of thought, and therefore the volume offers varied per spectives on its topic that some other edited volumes do not. Travis, J. 1994. Evaluating the adaptive role of morphologi cal plasticity. In P. C. Wainwright and S. M. Reilly, eds., Ecological Morphology: Integrative Organismal Biology. Chicago: University of Chicago Press, 99 122. This re view paper was written for the scientist who is not a specialist in evolutionary biology or ecology and wishes to learn about phenotypic plasticity. It offers a synthetic examination of phenotypic plasticity, reviewing the con clusions of mathematical theory but without the math ematics for its evolution and matching a large number of examples, primarily drawn from animals, to the classes of theoretical treatments to which those examples apply. Although theory has advanced considerably since it was written, it is still a lucid introduction to the literature, especially the terminology, and its strength is in describing clear patterns in the vast array of examples.

I.10 Life History William F. Morris OUTLINE

1. Variation in life history among species and the notion of trade-offs 2. Key life history patterns and associated trade-offs The term life history summarizes the timing and magnitude of growth, reproduction, and mortality over the lifetime of an individual organism. Important features of an individual’s life history include the age or size at which reproduction begins, the relationship between size and age, the number of reproductive events over the individual’s lifetime, the size and number of offspring produced at each reproductive event, the sex ratio of offspring, the chance that the individual dies as a function of age or size, and the individual’s lifespan or longevity (the time elapsed between the birth and death of the individual). Although all of these features (so-called life history traits) describe individuals, some are more easily understood when viewed as aggregate properties of a population of individuals. This is particularly true of mortality and lifespan. Each individual dies once, at a certain age. But in a population of identical individuals, some may die at a young age and some at an old age. By imagining that the fraction of this population that is still alive at a given age also represents the probability that an average individual survives to that age, we see that the chance of survival to a given age, which is the converse of the chance of dying, or mortality, is a property of an individual. Similarly, we can envision the average lifespan (or ‘‘life expectancy’’) even though each individual has a single age at death. All sunflowers and the vast majority of sequoia seedlings die before reaching one year of age. Yet in a sequoia population, individuals have the potential to live for several millennia, which distinguishes sequoias from sunflowers.

GLOSSARY fertility. The number of daughters to which a female

gives birth during a specified age interval geometric mean. The nth root of the product of

n numbers

iteroparity. A reproductive pattern in which individu-

als reproduce more than once in their lives life table. A table summarizing age-specific survivor-

ship and fertility used to calculate the net reproductive rate net reproductive rate. The average number of daughters to which a newborn female gives birth over her entire life semelparity. A reproductive pattern in which individuals reproduce only once in their lives survivorship. The probability that a newborn survives to or beyond a specified age 1. VARIATION IN LIFE HISTORY AMONG SPECIES AND THE NOTION OF TRADE-OFFS

As for the difference in life expectancy between sunflowers and sequoias, each of the key life history traits varies 1000-fold or more among species, as illustrated in figure 1. Life history traits also vary among individuals of the same species. The fundamental question in ecological and evolutionary studies of life history is: why is there so much variation in life history traits among and within species? To answer this question, we start by recognizing that life history features are traits just like any other (e.g., coloration, bill shape, cold tolerance, body size, etc.) that can be acted on by natural selection. Moreover, variation in life history traits among individuals in a population often has a genetic basis, so genotypes favored by natural selection can potentially increase in frequency from one generation to the next. If life history traits are genetically based and subject to selection, evolution of life history might be expected to lead to an organism that begins reproducing immediately after birth and produces a large number of wellprovisioned offspring in a series of reproductive events throughout an infinitely long life (such an organism has been termed a ‘‘Darwinian monster’’ because it would quickly displace all other species from Earth). The reason we do not see Darwinian monsters even though

Life History



Salmon X

Songbird Human




Sequoia Birth


X 1


10 30 100 Age (years – log scale)




Figure 1. Diversity of life histories for six representative species, three plants (sunflower, oak, and sequoia trees) and three animals (salmon, songbird, and human). Rectangles show reproductive events; height of each rectangle indicates the magnitude of re

productive effort (separate reproductive events are merged for oaks and sequoias). An ‘‘x’’ marks the age at death of an adult. Note that age is on a logarithmic scale; sequoias can live 3000 times longer than sunflowers.

life history traits evolve is that different life history traits are not independent. Because the resources that an organism has available to invest in maintenance and survival, in growth, and in reproduction are always limited, life history evolution is constrained by trade-offs: a greater investment in one life history trait must come at the expense of a smaller investment in one or more other life history traits. Trade-offs between many different pairs of life history traits have been documented, and we will see several examples in the following section of this article. In recognizing trade-offs, we no longer expect that evolution will produce Darwinian monsters, but rather that natural selection will balance, for example, improvements in reproduction with reductions in survival. On one hand, the optimal balance may depend on features of the environment the organism occupies. On the other hand, multiple combinations of life history traits may produce equally fit organisms. Both provide explanations for the diversity of life histories we see among Earth’s biota.

verse weather, or genetic defects, the organism may not survive for long, so delaying reproduction carries the risk of dying before reproducing. This advantage of early reproduction is easily illustrated with a basic demographic tool, the life table (table 1). A life table has two principal columns, survivorship, usually denoted lx, which is the probability that a newborn female survives to age x or older, and fertility, usually denoted mx, which is the average number of daughters a mother of age x produces over the next age interval (note that life tables typically track females only). One use of a life table is to compute the average number of daughters a female will produce over her entire life, which is called the net reproductive rate and is usually denoted R0. Natural selection can be expected to favor production of more daughters. As shown in table 1, for a fixed set of lx values, fertility skewed toward earlier ages will lead to a higher R0 simply because females will be more likely to survive to reproduce. The second reason why early reproduction is advantageous is that daughters produced earlier will themselves begin reproducing sooner than will daughters produced later in the mother’s life. If we think of a mother and her female descendents as a lineage, a lineage founded by an early-reproducing mother will grow faster than will the lineage of a later-reproducing founder, even if both lineages have the same R0 (figure 2). However, the production of offspring costs resources that the parent could use for other purposes, such as growth. Individuals that reproduce early in life


All else being equal, an organism should begin reproducing as soon as possible, for two reasons. First, because of factors such as predators and diseases, ad-


Autecology Lineage 1

Lineage 2

Figure 2. Growth of two female lineages. In Lineage 1, each mother produces two daughters in her first year and then dies. In Lineage 2, each mother produces one daughter in her first and one in her second year and then dies. For both lineages, R0 ¼2, but the first lineage grows faster. Circles are females, dashed vertical arrows show survival of the same female, and solid arrows show production of daughters.

may grow less rapidly, a trade-off between early reproduction and early growth that may prevent early reproducers from reaching a large final size. Moreover, because larger individuals often have both a greater chance of surviving and a greater reproductive potential, early reproduction may also be involved in tradeoffs with later survival and later reproduction. If reproductive capacity increases rapidly as the size of an organism increases, delaying reproduction in order to achieve a larger size may allow an individual to produce more offspring over its entire life despite the advantages of early reproduction illustrated in table 1 and figure 2. Delaying reproduction in order to grow more rapidly is especially important for males of species with territorial breeding systems. For example, the victors of fights between male elephant seals gain nearly exclu-

sive reproductive access to harems of females. Because small males have little chance of winning fights, young males invest energy in growing larger rather than in futile attempts to breed. In contrast, females do not need to win fights to breed, so they begin reproducing several years earlier in life than do males. Thus, males and females of the same species may face different trade-offs between early reproduction and growth. Interestingly, a breeding system in which males delay reproduction to grow large enough to win male– male contests may open the door for alternative mating strategies used by smaller or younger males. In many fish species, small males known as satellites may mimic females or use other methods to sneak into the territory of a mating pair in order to fertilize some of the female’s eggs when she releases them into the water. Satellites also occur in other animal groups (including lizards). Satellites can increase in a population when they are rare, because there are many territorial males to ‘‘parasitize,’’ but their success declines as they become a larger proportion of the population. They may achieve reproductive success similar to that of territorial males, or they may be making the best of a small size caused by genetic or environmental factors. Number of Lifetime Breeding Events

Constant Environments Once an individual becomes reproductively mature, it can breed only once (semelparous species), as in sunflowers and salmon, or breed multiple times (iteroparous species). Intuitively, we might think that if breeding once is good, breeding more than once is

Table 1. Hypothetical life tables

Daughters produced at ages 2 to 4 Age, x 0 1 2 3 4 5

Daughters produced at ages 1 to 3



lx m x

Age, x



lx m x

1 0.6 0.4 0.2 0.1 0.0

0 0 1 2 1 0

0 0 0.4 0.6 0.1 0

0 1 2 3 4 5

1 0.6 0.4 0.2 0.1 0.0

0 1 2 1 0 0

0 0.6 0.8 0.2 0 0

R0 = sum of lxmx values


R0 = sum of lxmx values


Note: Hypothetical life tables with the same survivorship (lx) schedules but with daughters produced at later (left) versus earlier (right) ages of the mother. The product lxmx is the number of daughters a female is expected to produce at age x, accounting for the fact that she might not survive to age x. The sum of the lxmx values over all ages is the net reproductive rate, R0, which is higher when daughters are produced earlier in the mother’s life.

Life History better. But Lamont C. Cole pointed out in 1954 that the lineage of an immortal organism that reproduces every year would grow at the same rate as the lineage of an organism that reproduces only once at 1 year of age and then dies, but produces just one more offspring than does the iteroparous organism at each of its breeding events (it is easy to modify figure 2 to see how Cole’s claim might be true). More than one additional offspring would give the advantage to the semelparous organism, and if the cost of investing in survival is high, an organism that reproduced once and then died might be able to achieve even higher reproduction (note that we have just assumed a survival–reproduction trade-off). Thus, Cole claimed that iteroparity was more paradoxical than semelparity. Other ecologists showed later that Cole’s result requires that adults and newborn offspring have the same chance of surviving each year, which we now demonstrate with a simple mathematical model. If It and St are the numbers of individuals in the iteroparous and semelparous lineages in year t, then Itþ1 and Stþ1, the numbers in the two lineages the following year, are predicted by It þ 1 ¼ FN BI It þ FA It ¼ (FN BI þ FA )It St þ 1 ¼ (FN BS )St , where FN and FA are the fractions of newborns and adults surviving the year, and BI and BS are the numbers of newborns produced by each iteroparous and semelparous organism each year. There is no FA term in the equation for the semelparous lineage because individuals die after breeding. The two lineages will grow at the same rate if the terms in parentheses in the two equations are equal, that is, if FN BI þ FA ¼ FN BS or, dividing both sides of the equation by FN, if BI þ FA ⁄ FN ¼ BS : If instead the left side of the preceding equation is larger than the right side, the iteroparous lineage will grow faster. Note that if the same fraction of adults and newborns survive the year (i.e., if FA ¼ FN), we obtain Cole’s result (because the semelparous organisms are producing one more newborn than are the iteroparous organisms). However, because newborns are smaller and often more vulnerable than adults, for many species and environments, a smaller fraction of newborns than adults will survive. Because FA/FN may then be substantially greater than 1, semelparous reproduction may need to be a good deal greater to achieve a fitness equal to that of the iteroparous line-


age. Thus, an important advantage of iteroparity is that it capitalizes on the greater value of adults, as measured by their higher survival rates, even at the expense of lower offspring production at each breeding event. Variable Environments In the preceding section, we assumed that newborn survival is the same every year. Long-lived adults are even more valuable in an environment in which newborn survival varies from year to year, as is likely to occur for many species and environments. Imagine that there are two kinds of years, good and bad, for newborn survival. Assume that in good years, FN ¼ 0.5þD, and in bad years, FN ¼ 0.5D; by increasing the number D, we increase the contrast between good and bad years. If D ¼ 0, 50% of newborns survive every year. If D ¼ 0.5, all newborns survive in good years, and none survive in bad years (higher values of D are meaningless because the survival fraction cannot be less than 0 or greater than 1). If we assume that good and bad years are equally frequent but occur at random, then regardless of the value of D, the average newborn survival across years is 0.5. If newborn survival varies from year to year, the terms in parentheses in the equations we used above to predict the growth of iteroparous and semelparous lineages, which represent the annual lineage growth rates, also vary from year to year. Note that to get Itþ2, the size of the iteroparous lineage in year tþ2, we would first compute Itþ1 by multiplying It by the annual growth rate for the iteroparous lineage in year t [which would be (0.5þD) BI þFA if year t were a good year and (0.5D) BI þFA if year t were a bad year] and then multiply the result by the annual growth rate for year tþ1 (which again might be a good or a bad year). Thus, over a period of years, the size of a lineage (either iteroparous or semelparous) is determined by the product of the annual lineage growth rates. If good and bad years are equally likely, then over a long period of years, close to half of the years will be good and half will be bad, and in a typical 2-year period one will be good and one will be bad. Thus, the growth of the iteroparous lineage over a typical 2-year period will be determined by the product of the good-year and bad-year growth rates, [(0.5þD)BI þFA][(0.5D)BI þFA], and the growth over a typical 1-year period will be the square root of this product. Similarly, the typical 1year growth rate of the semelparous lineage will be the square root of the product [(0.5þD)BS] [(0.5D)BS]. These typical growth rates represent the geometric means of the annual growth rates. (The geometric mean of two numbers is the square root of


their product, whereas the more familiar average or arithmetic mean is one-half of their sum.) Because lineage growth is a multiplicative process, the geometric mean is a more appropriate measure of typical annual growth. Now let us set FA to 0.9 (90% of adults survive a year) and BI and BS to 0.5 and 2.5, respectively (the semelparous organism produces on average two more offspring per year than does the iteroparous organism). If D ¼ 0, all years are the same, and the annual growth rate of the iteroparous lineage, BI FN þFA ¼ 0.5 0.5þ0.9 ¼ 1.15, is less than the annual growth rate of the semelparous lineage BS FN ¼ 2.50.5 ¼ 1.25. Thus, in a constant environment characterized by the survival and reproductive rates we have chosen, the semelparous lineage outperforms the iteroparous lineage. But what happens as we increase the contrast between good and bad years by increasing the value of D? As figure 3 illustrates, the semelparous lineage continues to grow faster than the iteroparous lineage when year-to-year variability in newborn survival (as determined by D) is low. However, the growth rates of both lineages decline as D increases, but much more so for the semelparous lineage, so that once D exceeds 0.2, the iteroparous lineage outgrows the semelparous lineage. Note that we have not changed the average newborn survival rate, so the switch in relative performance of the two lineages shown in figure 3 is driven entirely by the increase in variability of newborn survival. The presence of long-lived adults in the iteroparous lineage allows it to persist during years when few or no newborns survive. In contrast, persistence of the semelparous lineage requires that at least some newborns survive every year. That is why the geometric mean growth rate of the semelparous lineage is zero when D ¼ 0.5 and its bad-year annual growth is zero; a single zero will cause the product of annual growth rates to be zero because a single year in which all newborns die will cause extinction of the lineage. Thus, iteroparity, even at a cost of reduced annual reproduction, can be favored when newborn survival is low on average and/or is highly variable from year to year. Essentially, long-lived adults spread their reproductive efforts over multiple years, so their lifetime reproductive success is less sensitive to a single bad year. Although iteroparity is one life history adaptation to randomly varying environments, semelparous species also possess life history adaptations to environmental variability, namely dormancy and diapause, which we omitted from our simple model. Annual plants, by definition, reproduce only once, but many of them produce seeds of which a fraction lie dormant in the soil for one or more years. Because the offspring of a parent plant then germinate in different years, the parent is

1.4 Geometric mean lineage growth rate


1.2 1.0 0.8 0.6 0.4 0.2 0

0.1 0.2 0.3 0.4 Contrast between good and bad years, D


Figure 3. The typical annual growth rates of the semelparous lineage (shown as a black line) and of the iteroparous lineage (shown as a gray line) when newborn survival varies randomly from year to year. The typical growth rate is measured as the geometric mean, and the contrast in newborn survival between good and bad years increases as D increases.

effectively spreading its reproduction over several years, just as an iteroparous organism does, thus reducing its sensitivity to a single bad year in which all seedlings die and increasing its geometric mean fitness. Similarly, many insects and other invertebrates can remain in an inactive diapause state as adults for more than a year, allowing the set of offspring of a single mother to sample different years even though each offspring reproduces only once. We assumed above that only newborn survival varied from year to year, but in reality, the survival, growth, and reproduction of individuals of all ages are likely to vary. In a completely unpredictable environment, year-to-year variability in all of these life history traits will depress the geometric mean growth rate of a lineage, just as variability in newborn survival does in figure 3, but variability in traits such as adult survival that make large contributions to the lineage growth rate will be more detrimental than variability in less influential traits. Across many species, there is growing evidence that the most influential life history traits vary less from year to year than do less influential traits, suggesting that mechanisms have evolved that buffer the life histories of those organisms against the most detrimental types of variability. However, not all year-to-year variability in life history traits is detrimental. Many environments are only partly unpredictable. For example, in fire-prone ecosystems, it may be difficult to predict whether a fire will occur in a given year, but once a fire has occurred, conditions of abundant resources, low competition, and low likelihood of additional disturbance may be

Life History quite predictable until enough fuel has accumulated to allow the next fire to occur. Many species inhabiting such disturbance-dominated systems have evolved timing of the phases in their life histories to exploit this environmental predictability. For example, many fireadapted plants produce seeds that germinate only in the period soon after a fire, so their offspring can take advantage of abundant postfire resources. Conversely, reproduction of these plants is often restricted to late in the interfire interval, after plants have grown to reproductive size and when their seeds will be poised to exploit the next interfire interval. In these species, among-year variation in life history traits reflects adaptation to multiyear environmental cycles rather than the detrimental influence of environmental variability. Lifespan and Aging

Even iteroparous organisms eventually die. Moreover, for many species, the chance of dying in a given interval of time may initially decline after birth as newborns grow to a less vulnerable size or as those with developmental defects die, but it eventually increases as individuals reach more advanced ages. The process of aging is defined as an increase in mortality risk late in life. As we would expect that the ability to continue living and reproducing would be favored by natural selection, why does aging occur? Peter B. Medawar argued, and William D. Hamilton showed mathematically, that the ability of natural selection to weed out genes that increase mortality or decrease fertility declines with an organism’s age. The reason for this decline in the strength of selection is that even individuals with good genes are increasingly likely to have died from external causes, such as predators or bad weather, as age increases. Therefore, few individuals will still be alive to enjoy the advantage of decreased mortality risk or increased fertility at advanced ages, whereas most individuals will benefit from increases in early-life survival or fertility. Two types of genes may underlie an increase in mortality or decrease in fertility with age. Detrimental genes that are expressed only late in life would experience only weak selection against them and so would tend to accumulate in the genome. Genes that have beneficial effects on survival or fertility early in life but detrimental effects late in life would be maintained because positive selection for their early effects would overwhelm negative selection for their late effects. The former type of genes play a role in Peter Medawar’s mutation accumulation theory of aging, whereas the latter type of genes are central to George C. Williams’ antagonistic pleiotropy theory of aging. There is evidence that both types of genes may be present in the same organisms.


For many organisms, including fruit flies, some plants, and humans, the risk of mortality reaches a plateau rather than continuing to increase at very advanced ages. A mortality plateau could arise because individuals that are frail from genetic or environmental factors are increasingly likely to have already died as age increases. But as Hamilton’s theory shows that natural selection will be powerless to eliminate detrimental genes expressed at all ages past the age at which reproduction ceases, detrimental late-acting genes could simply be maintained at intermediate frequencies by a balance between the nonselective evolutionary forces of mutation and genetic drift. Although Hamilton’s work predicts that there will be no selection to reduce mortality once reproduction ceases, that work only accounts for offspring directly produced by a mother. However, mothers can also contribute to the growth of their lineage by providing direct care to their granddaughters or by providing information that improves the quality of care their daughters provide to their own offspring. These direct and indirect transfers across two generations may explain why, in social species such as primates, mortality does not increase rapidly once a female ceases to give birth. Number and Size of Offspring

Whether an organism reproduces once or more than once, the resources it invests in a single breeding event can be used to produce a single large offspring, or those resources can be divided to produce more than one, but smaller, offspring. The trade-off between the size and number of offspring is a fundamental constraint on life history. Producing more offspring will cause a lineage to grow faster, but not if each offspring is too small to have a good chance of surviving. Therefore, the best solution in the face of a size–number trade-off is to produce the number (and therefore size) of offspring at which the product of the offspring number and the sizedependent probability that each offspring survives is at a maximum. Other factors may skew the optimum solution toward more and smaller offspring. For example, trees with seeds dispersed by wind typically make many small seeds. Even though the small seedling emerging from each seed has a low chance of surviving, by increasing the area reached by its windblown seeds, a mother plant will be more likely to place at least some of its seeds in sites suitable for seedling growth and survival, even if those sites are few and far between. A large number of studies have addressed the question of why most female birds lay fewer eggs in each nest than the maximum number observed for the species. Why do lineages that produce more eggs per nest



not come to replace lineages that produce fewer eggs per nest? David L. Lack proposed that birds should maximize the number of offspring surviving to leave the nest rather than the number of eggs per nest. Because the parents can provide less food to each chick when a nest contains many chicks, the chance that each chick survives declines with the number of eggs (and therefore chicks) in a nest. As a result, the product of the number of eggs per nest and the probability that the chick hatching from each egg survives to leave the nest is usually highest at an intermediate egg number. However, this number is often higher than the average number of eggs actually observed in nest. Birds may lay fewer eggs than Lack’s argument predicts because excessive investment in one nest reduces the parents’ chance of survival or their future reproductive success. That is, trade-offs between current reproduction and future survival or future reproduction may constrain the amount invested in a single bout of reproduction. See also chapters I.13, I.14, and II.1 in this volume.

Iteroparity Cole, L. C. 1954. The population consequences of life history phenomena. Quarterly Review of Biology 103: 103 137. A classic paper that spurred research into reproductive patterns and longevity.

Aging Hamilton, W. D. 1966. Moulding of senescence by natural selection. Journal of Theoretical Biology 12: 12 45. Al though rather mathematical, this paper is the basis for most evolutionary theories of aging. Partridge, L., and D. Gems. 2006. Beyond the evolutionary theory of ageing, from functional genomics to evo gero. Trends in Ecology and Evolution 21: 334 340. A recent review of genetic causes of aging. Rose, M. R. 1991. Evolutionary Biology of Aging. New York: Oxford University Press. Rose, M. R., C. L. Rauser, L. D. Mueller, and G. Benford. 2006. A revolution for aging research. Biogerontology 7: 269 277. A review of the evidence for and causes of mortality plateaus.

Offspring Number FURTHER READING General Roff, D. A. 1992. The Evolution of Life Histories. New York: Chapman & Hall. Stearns, S. C. 1992. The Evolution of Life Histories. Oxford: Oxford University Press. Two excellent and comprehen sive overviews of the field.

Godfray, H.C.J., L. Partridge, and P. H. Harvey. 1991. Clutch size. Annual Review of Ecology and Systematics 22: 409 429. Emphasizes determinants of offspring number in organisms other than birds. Lack, D. L. 1954. Natural Regulation of Animal Numbers. Oxford: Oxford University Press. This book includes Lack’s argument for why birds should lay fewer than the maximum number of eggs per nest.

I.11 Remote Sensing and Geographic Information Systems Catherine H. Graham and Scott J. Goetz OUTLINE

1. Basic concepts of remote sensing and geographic information systems 2. Applications of RS and GIS in ecology Remote sensing (RS) and geographic information systems (GIS) provide data and tools that are used extensively across ecology, evolution, biogeography, and conservation biology. Some fields in particular, such as landscape ecology and biogeography, have relied heavily and increasingly on sophisticated analyses afforded by these data and tools.

GLOSSARY electromagnetic energy. Energy or radiation in a wave

in space with an electrical field that varies in magnitude in a direction perpendicular to the direction in which the radiation is traveling and a magnetic field oriented at right angles to the electrical field. geospatial. The distribution of information in a geographic sense such that entities can be located by some coordinate of a reference system (i.e., latitude and longitude), which places these entities at some point on the globe. global positioning system (GPS). This system is a set of 24 satellites that orbit the Earth and communicate their position to a ground receiving device providing the geographic location of that receiver. 1. BASIC CONCEPTS OF REMOTE SENSING AND GEOGRAPHIC INFORMATION SYSTEMS

Broadly, RS is the gathering and processing of data about the physical world by a device detecting electromagnetic energy that is not in contact with the object, area, or phenomenon under investigation. For example, the images in Google Earth (http://earth are products of RS that provide infor-

mation about vegetation and human uses of the land, such as agriculture, roads, and buildings. As such, RS is generally used to generate data that are often then imported into a GIS. A GIS is a collection of tools that provide the ability to capture, display, manage, and analyze most forms of spatial data that are geographically referenced to the Earth’s surface (i.e., identified according to location). An important feature in GIS is the ability to relate different types of information, such as remotely sensed vegetation images and maps of human population density, in a geospatial context to explore associations and relationships among these various types of information. Together, RS and GIS, along with other relatively recent tools, such as geographic positioning systems (GPS) and sophisticated spatial statistics, provide a powerful platform to advance our understanding of the natural world. In this chapter, we first describe RS and GIS in greater detail, and we then provide some examples of how they are used in ecology. Remote Sensing

RS data provide real-time information about what is happening on our planet and can be used in a wide range of ecological studies. For example, it is now relatively easy to quantify deforestation rates across different ecosystems or even detect a fire in a remote area. RS uses theory developed in physics to measure electromagnetic energy emitted or reflected from distant objects. Electromagnetic radiation/energy can be described in terms of a stream of photons, which are massless particles, each traveling in a wavelike pattern and moving at the speed of light. This radiation varies across a spectrum of different amounts of energy in photons and size and frequency of waves. RS applications typically use wavelengths that include the visible wavelengths (blue through red), the






700 nm Wavelength


600 nm



500 nm

400 nm

Frequency in Hz 104






Radio, TV waves





infrared, and microwave regions of the electromagnetic spectrum (figure 1). Different types of objects (such as grassland and forest canopies) reflect, absorb, or scatter electromagnetic energy differently and, as a result, emit electromagnetic waves at different magnitudes in different portions of the electromagnetic spectrum. The principle behind RS is to detect and identify these characteristic waves (i.e., varying energy levels) for different materials, which are often referred to as spectral signatures. This is done by documenting the reflectance across a range of electromagnetic wavelengths for one or more objects or vegetation types. This information is often displayed as a spectral reflectance curve (figure 2; for an interactive tool, visit remote sensing/widgets/spectral curve/index.html). One of the first and still widely used sources of RS data is imagery from the Land Satellite (Landsat) series of sensors, first launched in 1973. The most recent Landsat images have seven different bands that cover a spectrum of 0.450 mm to 2.35 mm. Information about the reflectances in each of these bands can be used to classify different regions of an image (generally referred to as picture elements or pixels) as a certain type of land cover (figure 2). The eighth Landsat satellite in the series is scheduled for launch in 2011 and is called the Landsat Data Continuity Mission. Satellite-based sensors have several unique characteristics, including how data reach the sensor, the number and width of spectral bands, and the spatial and temporal resolution of the data. There are two broad types of sensors that differ in how they sense and capture data. These are referred to as passive or active sensors. In passive sensors, radiation emitted from an object is simply measured as it reaches the sensor. In contrast, active sensors emit a pulse of energy and

1018 Gamma rays


Figure 1. The electromagnetic spectrum. (From http://cache rendTypeId=35)

measure the portion of the energy that is returned or bounced back to the detector (figure 3). For example, land cover is often determined using a passive sensor, such as the Landsat images described above. Vegetation vertical structure or surface topography is more commonly measured with an active sensor, where returns of the energy pulses sent by the sensor are influenced by the complexity of object—a measure of its structure. For example, RADAR (radio detection and ranging) or LiDAR (light detection and ranging) returns from a tropical forest canopy may be sensitive to the density of biomass or the branching structure of trees. The width of the bands of the electromagnetic spectrum that a sensor can detect is known as its spectral resolution. Some sensors, especially older ones, can detect information in only a few wide spectral bands (typically blue, green, red, and near-infrared), whereas other sensors can obtain information across a narrow range of spectra that are sensitive to absorption by atmospheric water vapor or other gases. If a sensor has many bands of narrow width (commonly known as hyperspectral), it may be easier to differentiate among different objects by using the more detailed information from its spectral signature. Finally, sensors gather data at different spatial and temporal resolutions. Spatial resolution refers to the pixel size at which data are collected, whereas temporal resolution is the ‘‘revisit’’ time or how often a satellite flies over a given location (high temporal resolution ¼ short revisit time). A high temporal resolution is important for some research questions, such as tracking algal blooms in the ocean or the leaf-out of trees in the northern hemisphere spring. Often there is a technological trade-off between spatial and temporal

0.9 Pine Wood

0.8 0.7




0.5 0.4

Iron Dolomite

0.3 0.2 0.1


0 0.4


1.4 Wavelength (microns)



Vegetation Spectra 0.9 0.8 0.7 Russian Olive Reflectance

0.6 Maple


Lawn Grass


Leafy Spurge

0.3 0.2 Spruce

0.1 0 0.4


1.4 Wavelength (microns)



Figure 2. Spectral reflectance curves for different kinds of mate rial. (From http://www.profc.udec .cl/ gabriel/tutoriales/rsnote/cp1/ 1 9 1.gif)

Figure 3. Passive (left) and active (right) sensors. (From .htm)


Autecology Table 1. Examples of ecological variables and data sources useful for ecological research

Ecological variable

Spatial resolution

Revisit time

Moderate resolution Imaging Spectrometers (MODIS) Landsat

250 1000 m

1 2 days

30 m

16 days

Sea viewing Wide Field of view Sensor (SeaWiFS) QuickBird

1000 m

1 day


2 4 days


250 1000 m

1 2 days

Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER)

10 m

4 16 days


Shuttle Radar Topography Mission (SRTM)

90 m


Digital elevation models derived from radar singles (SRTM)

Vertical canopy structure

Laser Vegetation Imaging Sensor (LVIS)

1 10 m


Provides 3D measurements via laser pulses; provides biomass estimation and information about vegetation structure

Land cover




Description Can discriminate different land surfaces

Measure reflectance to assess presence/absence of vegetation and relative greenness of ocean and land chlorophyll to calculate productivity Information on leaf phenology, and in some cases flowering/fruiting cycles

Source: Modified from Turner et al. (2003). There are many more data than listed here; for a more complete list, see Turner et al. (2003).

resolution in which sensors with a high revisit time have larger pixel sizes. An extreme example of this is a meteorological satellite that views the Earth every few minutes but only at very coarse spatial resolution. This variation in sensor type (i.e., passive/active) and characteristics of images different sensors provide (i.e., bandwidth, number, pixel size, and revisit time) results in a broad range of images available for research in ecology. Table 1 lists some of these images, what sensor they come from, and their uses in ecological research. The National Aeronautics Space Administration (NASA) is a U.S. government organization that provides many RS images, such as the Landsat images mentioned earlier, although data are distributed by the U.S. Geological Survey (USGS). Other countries have similar organizations, such as Japan (the Japanese Space Agency, JAXA) or France (National Institute for Agricultural Research, INRA). Further, there are several commercial satellites that have very high-spatialresolution (just 1–4 m pixels) multispectral sensors, such as the IKONOS system from Space Imaging and the QuickBird system from DigitalGlobe.

Geographic Information Systems

A GIS is a method for depicting relationships between different kinds of information, usually in the form of maps, each depicting different kinds of information. Because each layer has a specific geographic location, a GIS can be used to explore how these layers relate to each other or to a particular phenomenon, and this often provides new insights or analysis options for ecologists. For example, a researcher might want to study how the spatial pattern of American crow abundance might be influenced by a series of environmental variables such as rainfall, habitat type, and temperature, as well as socioeconomic variables such as housing type and density. Using GIS would allow the researcher to map all these variables at once in order to obtain a visual sense of potential relationships among them, generate hypotheses regarding their relationships, and run statistical tests to determine if the hypothesized correlations were statistically meaningful (plate 1). Currently, GIS software packages have an array of statistical tools, although in some cases GIS data are

Remote Sensing exported to statistical software packages for more complex analyses. Additional data, referred to as attributes, can be tied to the spatial data and used to address questions in ecology. For example, we might have information on how bird flu relates to American crow abundance. 2. APPLICATIONS OF RS AND GIS IN ECOLOGY

There are a myriad of ways that RS and GIS are currently used—and could be used—in ecology. These tools provide far more than the ability to make maps. Their uses are limited only by human imagination and technology. With technological capacity expanding at a phenomenal rate, it is our job as students and professionals to use these data to solve theoretical and applied problems in ecology. What follows are four examples of how RS and GIS are used in ecology. Bird Abundance Patterns and Vegetation Structure

It has long been known that the diversity of habitat is related to the diversity of species that occupy that habitat. RS has the ability to provide information on both the two-dimensional (horizontal) and three-dimensional (vertical) aspects of habitat diversity. Satellite sensors like those on Landsat have been used in many studies linking some aspect of biodiversity (either species richness or abundance) to land cover or vegetation maps. This works in areas where there is a diversity of cover types, and species specialize on each of those cover types. So, for example, it has been possible to estimate the diversity of birds in a temperate area where there is a mixture of forest, shrub, and grassland vegetation. If the number of bird species or the abundance (total number of individuals of a given species) associated with each of the vegetation types is known (has been measured in the field), then one can approximate the diversity of species by simply multiplying all the pixels of each given type by the number of species that use that habitat (a ‘‘paint by numbers’’ approach). This approach tends not to work very well in densely forested areas, where habitat is more likely to be ‘‘partitioned’’ vertically and many species may be present in any given vegetation type. In this case, an active sensor such as RADAR or LiDAR can provide information on the vertical structure or diversity of habitat. One such study was conducted in the dense temperate forest of the Patuxent Wildlife Refuge in Maryland. Wildlife biologists working at the refuge counted the number of birds they observed or heard within a 5-min period at a given location and then walked 400 m (1000 feet) and counted again. They did this 266 times over a regular grid, ultimately counting


over 5000 birds from 88 different species. The number of bird species at any given grid location ranged from 2 to 27 (averaging 12), and abundance ranged from as few as 3 to over 100 (averaging 19). RS scientists supported by NASA then flew over the refuge in an airplane equipped with an imaging LiDAR system. They flew at night because the instrument (known as the laser vegetation imaging system, or LVIS, pronounced ‘‘Elvis’’) is an active sensor that does not rely on light from the sun reflected by the vegetation canopy. The imagery derived from LVIS included an initial return from the top of the vegetation as well as multiple returns reflected from different elements in the canopy and ultimately a strong return from the ground surface. The difference between the first return (the top of the canopy) and the ground return was used to estimate canopy height (an oblique view of this canopy height is shown in plate 2). Other useful canopy metrics include the heights at which 25%, 50%, and 75% of the total summed energy was returned. The point at which half of the energy was returned is referred to as the height of median energy (HOME). This variable is of particular interest because, unlike canopy height, it provides some information on the vertical complexity of the canopy. When HOME and height are used together, one can get a sense for whether there is, for example, a dense understory or a shrub vegetation layer or very little of either. Because some birds specialize in foraging and/or nesting in these specific vegetation layers within the forest canopy, the LVIS metrics provide information that can help to estimate bird diversity (and possibly abundance, although that has yet to be tested because it can vary substantially from year to year for reasons unrelated to habitat diversity). The results of this analysis indicate that LiDAR measurements of vegetation vertical structure provide a useful proxy for habitat diversity, and this, in turn, was related to bird species diversity. A number of different statistical analysis techniques showed that the LVIS metrics of habitat heterogeneity performed significantly better than those based on optical RS (i.e., Landsat), and use of both Landsat and LVIS improved little on use of LVIS alone. This is an important finding because it demonstrates not only that unique information can be derived from LiDAR but that this information has utility to biodiversity research. If a LiDAR sensor were placed on a satellite orbiting the Earth on a daily basis, it would provide valuable information on biodiversity and other aspects of ecosystems, including carbon stored in biomass, which is of interest for estimating carbon emissions to the atmosphere from deforestation and biomass burning. Just such a system is now under development at NASA and is expected to



be launched in 2014. The scientists who conducted the bird diversity study described here are members of the science team in order to ensure that the data will be useful for both biodiversity and carbon-climate applications. Determining Species Persistence in Dynamic Landscapes

A major goal for researchers, government agencies, and, often, local residents interested in conservation is to maintain viable populations of all the different kinds of species that live in a given region. A viable population is one that is likely to persist over a long period of time. Defined in terms of demographics, a population is viable when the new births in the population, combined with individuals colonizing from other areas (i.e., immigration), are greater than or equal to the deaths in the population and those individuals leaving the population through emigration, given human activities, variation in weather, and other natural disturbances (floods, hurricanes). Scientists have developed a series of tools to predict if a population is likely to remain viable and how changes in a species habitat (defined as places where the intrinsic rate of population growth is greater than 0) will affect its persistence. Species habitats are continuously changing as a result of natural processes, such as storms or floods and human land use. More and more, humans have removed natural forests, meadows, or other vegetation types for housing developments, resources (wood harvesting), or recreation, and as a result they have reduced and changed the spatial configuration of habitat for species. GIS mapping and modeling can be combined with population viability analyses to determine how changes in land use might influence the probability that a species might persist for long periods of time. Further, different scenarios can be explored; for example, in a forest you could determine how different rates and types of forest extraction would influence the viability of a species dependent on forest. A recent example of combining GIS mapping and prediction with viability analyses was used to make management recommendations for the sharp-tailed grouse (Tympanuchus phasianells), a pheasant-sized bird native to open and shrubby vegetation (steppe grasslands) in the Midwest (Akcakaya et al., 2004). Grouse populations have steadily declined because of conversion of its habitat to agriculture, housing developments, and forest plantations. Currently, much of the grouse population in the United States is contained within the Pine Barrens region of northwest Wisconsin. The habitat in the Pine Barrens is highly fragmented with patches of suitable habitat separated from other

such patches. Further, the region is dynamic, mostly a result of natural fires and silviculture (forestry), so that the number, size, and location of patches change over time. This presents a modeling challenge because the only way to correctly predict population viability for the species is to do it in the context of a highly heterogeneous and dynamic landscape. To address this challenge, Akcakaya and his colleagues created a series of different landscapes in a GIS, each simulated based on different silviculture and burning regimes. Basically, at a given time step they could predict the area and spatial configuration of grouse habitat and combine this with grouse demographics, such as how many young grouse were predicted to be born (fecundity) based on the size and quality of each patch. At the next time step, the area and spatial configuration of habitat changed based on the management regime being modeled, and as a result, the demographic responses also had to be recalculated. Using this approach, Akcakaya and his colleagues showed that the population viability of the sharp-tailed grouse depended both on landscape dynamics and on demographic variables such as fecundity and mortality. Further, demographic modeling in a static landscape provided overly optimistic results about the long-term viability of the species when compared to those obtained using dynamic landscapes. For species like the grouse, which depends on temporary habitat patches in fragmented landscapes, and where the viability of the species is strongly influenced by the rate of appearance and spatial arrangement of patches in the landscape, complex GIS models are required to predict its persistence. This approach has also been used to determine the viability of Bell’s sage sparrow, a species of special concern in California. This species relies on early-successional shrubland (chaparral) on the coast in California. Chaparral is a habitat that was extensive in the coastal dunes in California, which have been heavily impacted by human development. Further, chaparral is maintained by burning and is highly flammable. This presents an obvious and difficult issue for coastal California, where burning also causes devastation to housing developments. GIS modeling combined the species habitat, human land use changes, and demographic information on the sage sparrow to evaluate its long-term persistence. Without an integrated model run on a GIS platform, studies of this sort would not be possible. Predicting Invasive Species (Weed) Spread Using Ecological Niche Modeling

Invasive species from many different taxonomic groups are increasing in abundance and distributional range at

Remote Sensing an alarming rate across many ecosystems on Earth. As humans and human products move around the globe at ever-increasing rates, nonnative species move with them, either intentionally or accidentally. Only a small proportion of nonnative species become invasive, but this small group of weeding species has had tremendous ecological effects on native species and ecosystems by outcompeting native species for resources, preying on native species, or changing cycling of important nutrients in ecosystems. Further, the estimated damage and control cost of invasive species in the United States alone amount to more than $138 billion annually (Pimentel et al., 2004). It is difficult to predict which nonnative species will become invasive, but once we have some reason to believe that a species will be invasive, we need tools to predict how and where it will spread. Anticipating future distributions of invasive species is essential for management prioritization, early detection, and control. One way to predict where a species might spread is to evaluate the environmental conditions where a species exists on its native range and then use this information to predict where it might spread. Ecological niche modeling (also referred to as species distribution modeling) is a GIS-based model that can be used to predict geographic patterns of invasive species. Niche modeling requires two kinds of data: georeferenced occurrence records of where the species is in its native range and GIS-based maps of the environmental variables (e.g., temperature, precipitation) that are likely to influence the suitability of the environment for that species (plate 1). Using a GIS, we can extract the environmental information from environmental layers (maps) for each occurrence record of the species in the native range, establish a statistical relationship between species occurrence and this environmental information, and use this statistical relationship to predict where the species might exist in a different region. These models can be refined to include maps of human land use, such as roads across which people (and potentially invasive species) move. The house crow, a common bird in Asia (India, Pakistan, Sri Lanka, southwest Thailand, and coastal southern Iran), has been expanding its range into new regions, such as East Africa, and it has been observed in Australia and parts of Europe. It is associated with human settlements in all of its range, from small villages to large cities. Recently, Nyari and colleagues (2006) used ecological niche modeling to determine globally where the species could exist. They used information on human land use, in this case a map of the human footprint (Sanderson et al. 2002), which includes information about cities, roads, population density, and satellite images of night lights (i.e., intensity of elec-


tricity use). Using GIS tools, maps, and RS images, they could identify regions that might be susceptible to an invasion by the house crow. A similar project was conducted by Broennimann and colleagues (2007) with spotted knapweed (Centaurea maculosa), an invasive plant well established in North America that is native to Europe. As with the crow study, they used climate information and occurrences from its native range to predict where the plant could live in North America. They found a large correspondence between the model prediction and the actual range of the plant in North America, but certain populations of knapweed were not well predicted by the model. This result indicated that knapweed exists under novel environmental conditions in North America; the plant occupies a distinct climatic niche on its invaded range. Although this is somewhat disappointing from the perspective of accurate and complete model prediction of invasive species, it does provide fascinating insights into phenotypic plasticity and the potential for species to evolve into new environmental niches. Ongoing research is evaluating these different possibilities.

FURTHER READING Akcakaya, H. R., J. Franklin, A. D. Syphard, and J. R. Stephenson. 2005. Viability of Bell’s sage sparrow (Amphispiza belli ssp. belli): Altered fire regimes. Ecolog ical Applications 15: 521 531. An example of combining GIS data and viability modeling to study population persistence. Akcakaya, H. R., V. C. Radeloff, D. J. Mladenoff, and H. S. He. 2004. Integrating landscape and metapopulation modeling approaches: Viability of the sharp tailed grouse in a dynamic landscape. Conservation Biology 18: 526 537. An example of combining GIS data and viability modeling to study population persistence. Broennimann, O., U. A. Treier, H. Muller Scharer, W. Thuiller, A. T. Peterson, and A. Guisan. 2007. Evidence of climatic niche shift during biological invasion. Ecology Letters 10: 701 709. An example of using ecological niche modeling to study how species niches might shift during an invasion. Goetz, S., D. Steinberg, R. Dubayah, and B. Blair. 2007. Laser remote sensing of canopy habitat heterogeneity as a predictor of bird species richness in an eastern temperate forest, USA. Remote Sensing of Environment 108: 254 263. An example of using remote sensing to study bird diversity. Graham, C. H., S. Ferrier, F. Huettman, C. Moritz, and A. T. Peterson. 2004. New developments in museum based informatics and applications in biodiversity analysis. Trends in Ecology & Evolution 19: 497 503. A review of spatial biodiversity data available and how to use it in a GIS.



Kerr, J. T., and M. Ostrovsky. 2003. From space to species: Ecological applications for remote sensing. Trends in Ecology & Evolution 18: 299 305. Nyari, A., C. Ryall, and A. T. Peterson. 2006. Global invasive potential of the house crow Corvus splendens based on ecological niche modelling. Journal of Avian Biology 37: 306 311. An example of using ecological niche modeling and human landuse data to predict the spread of invasive species. Pimentel, D., R. Zuniga, and D. Morrison. 2004. Update on the environmental and economic costs associated with alien invasive species in the United States. Ecological

Economics 52: 273 288. An evaluation of the economic cost associated with managing alien invasive species. Sanderson E. W., M. Jaiteh, M. A. Levy, K. H. Redford, A. V. Wannebo, and G. Woolmer. 2002. The human footprint and the last of the wild. Bioscience 52: 891 904. A quan tification of the influence of human behaviors on wild areas. Turner, W., S. Spector, N. Gardiner, M. Fladeland, E. Ster ling, and M. Steininger. 2003. Remote sensing for biodi versity science and conservation. Trends in Ecology & Evolution 18: 306 314. A review of remote sensing data available for ecological and conservation research.

I.12 Geographic Range Kevin J. Gaston OUTLINE

1. 2. 3. 4.

Range size Range edges Range structure Fundamental units

No species occurs everywhere. Indeed, most are absent from the vast majority of sites across the globe. Those areas in which a species does occur constitute its geographic range. As such, the geographic range is one of the fundamental units in ecology. The sizes and distribution of geographic ranges give rise to patterns of species richness and change in species composition from site to site, and combined with their abundance and trait structure give rise to other spatial patterns in assemblages. Likewise, temporal changes in assemblages on both short and long time scales follow from changes in the size, position, and structure of geographic ranges.

GLOSSARY area of occupancy. The area within the outermost geo-

graphic limits to the occurrence of a species over which it is actually found extent of occurrence. The area within the outermost geographic limits to the occurrence of a species intraspecific species-abundance distribution. The frequency of areas within a species’ geographic range in which it attains different levels of abundance range edge or limit. The outermost geographic occurrences of a species, usually excluding vagrant individuals species–range size distribution. The frequency of species with geographic ranges of different sizes 1. RANGE SIZE

The sizes of the geographic ranges of species vary dramatically and can be characterized in two fundamentally different ways. Extent of occurrence is the

area within the outermost limits to the occurrence of a species, and area of occupancy is the area over which the species is actually found. The latter will tend to be consistently smaller because no species is distributed continuously across space even within the broad geographic limits to its occurrence. The finer the spatial resolution and the shorter the time period over which area of occupancy is measured, the smaller will be the area over which the species is documented to occur, and the greater this disparity will be. At one extreme lie those, predominantly freshwater or terrestrial, species that are currently found occurring in a single small habitat patch (often with only a very small number of individuals), which are thus narrowly distributed in terms both of extent of occurrence and area of occupancy. At the other extreme lie some marine organisms. Species of microorganisms may be widespread across the oceans both in terms of extent of occurrence and area of occupancy, whereas some large-bodied species of vertebrate may have large oceanic distributions in terms of extent of occurrence but, because of the relatively low numbers of individuals, not area of occupancy. Species–Range Size Distributions

Both within and across major taxonomic groups, the geographic ranges of the majority of species are relatively small, and only a very few are widespread. Indeed, within such groups species–range size distributions, the frequency of species with ranges of different sizes, are almost invariably strongly right-skewed. One important consequence is that the vast majority of occurrence records result from a small number of species. For example, by one estimation, at a spatial resolution of approximately 100 100 km, the 10% most globally widespread extant species of birds account for 50% of occurrence records. Given that the ratio of extents of occurrence to areas of occupancy may often be proportionately larger for rare species than for widespread



ones, that is, they occupy their ranges less densely, the dominance of occurrence records by widespread species may increase when documented at finer spatial resolutions. This dominance may explain why it is the more widespread rather than, as often assumed, the restricted species that contribute disproportionately to spatial variation in species richness and related macroecological patterns. Phylogenetic Constraint

The average sizes of geographic ranges can vary markedly between species in different major taxonomic groups. Thus, among nonmarine vertebrates, species of fish and amphibians tend naturally to have smaller ranges than do mammals, and mammals smaller ranges than do birds. However, within taxonomic groups, the extent to which the geographic range sizes of species exhibit phylogenetic constraint is contentious. Certainly range size is not as strongly conserved as are body size and many life history traits. Even where significantly conserved, it typically remains impossible to predict with any accuracy the range size of a species from that of its sister species or other close relatives, suggesting that such heritability has limited practical value (e.g., in estimating the range sizes of species whose distributions have not been well documented). This would tend to follow if the range sizes of different species are determined by the variable outcomes that result from the combinations of individual traits and environmental conditions occurring at particular times and places. Spatial Dynamics

The mean size of the geographic ranges of the species within a higher taxon tends to vary spatially. Most obviously, ranges are typically smaller in situations in which dispersal and environmental conditions are geographically highly constrained, such as on islands and at high elevations, and in specialized habitats (e.g., desert springs, deep sea vents). However, more systematic spatial patterns have also been argued to occur, in particular, increases in the latitudinal extent of ranges from low to high latitudes, in their altitudinal extent from low to high elevations, and in their depth extent from shallow to deep waters. The first of these is a phenomenon termed Rapoport’s rule. The pattern appears to be most evident in the terrestrial northern hemisphere but may actually reflect a general trend for terrestrial ranges to increase from high southern to high northern latitudes. Although other factors may also have an influence, this trend is at least in part a result of changes in land area.

Temporal Dynamics

More difficult to establish than patterns of spatial variation in geographic range sizes are the long-term temporal trends. How the mean range sizes of the species in a higher taxon have changed over geologic time remains virtually unknown (although it is likely to have been marked, given changes in the distributions of land masses, water bodies, and climatic conditions). Little more is understood about how the range of an individual species changes in size between its origination and its extinction. However, best evidence suggests that geographic ranges typically undergo a rapid increase in size following speciation and then a slower subsequent, and perhaps prolonged, decline to extinction. This is supported by studies of species introduced into areas in which they previously did not occur, which have revealed that following an initial lag phase, during which a species tends to remain rather restricted to the locale of its introduction and densities there tend to build up, spread can occur across large areas very rapidly (both phases are extremely short in terms of evolutionary time). When we focus on the events at the outset and conclusion of a species’ lifespan, geographic range size influences both the likelihood of speciation and that of extinction. At least when allopatric, the likelihood of speciation appears to be related to range size by a hump-shaped function. As ranges increase from small to moderate sizes, the likelihood of speciation increases because the chance of the range being bisected by a barrier to dispersal increases. However, at some point ranges will become sufficiently large that they will tend to engulf all but the largest potential barriers, such that they do not engender speciation, and the probability of division will decline. In addition, widespread species may have well-developed dispersal abilities and greater numbers of individuals that both help to maintain range contiguity and reduce speciation rates. By contrast, the likelihood of extinction is strongly negatively correlated with geographic range size. Indeed, abundance and range size are in general the two best predictors of the probability that a species will go extinct in the near future (although there are examples of previously very widespread species that, usually as a consequence of anthropogenic pressures, have rapidly declined to extinction). Larger ranges typically comprise greater numbers of individuals and thus have a smaller probability of a random walk to extinction, and because of their greater areal coverage, they have a reduced risk that adverse conditions in one region will affect all individuals at the same time. This raises the possibility of species selection acting on range sizes.

Geographic Range Traits

Within taxonomic groups, interspecific variation in geographic range sizes is often correlated with such variation in other traits, including dispersal ability, breadth of resource use, or environmental tolerance, local abundance, and body size. In at least some cases, these relationships seem likely to be mechanistic, although the paths of causality may be variable. Thus, although it seems intuitive that a greater dispersal ability will tend to lead to a species becoming more widespread, other barriers may prevent this from occurring (see below), and if the structures associated with good dispersal abilities are costly to build or maintain, they may be reduced or lost. Indeed, there is evidence that when other limits are removed, species can show rapid acquisition of improved dispersal abilities. Species that are able to exploit a wider variety of resources or persist under a wider range of environmental conditions should, all else being equal, be able to attain larger geographic ranges. Of course, all else may not be equal (e.g., extent of different resource types, dispersal abilities), which will tend to weaken any correlations between the level of such generalism and range size. The variety of resources and the breadth of environmental conditions that species can use are influenced both by the variety and breadth that can be exploited by individual organisms and by the differences in resource usage and tolerance of environmental conditions among individual organisms. The latter is probably a much more important influence on the relative geographic range sizes of species in many taxonomic groups, given evidence for marked spatial variation in the realized and fundamental niches of individuals, particularly of those species that are more widespread. The geographic range sizes of species within taxonomic groups tend commonly to be positively correlated with their local density, such that widespread species not only have more individuals but disproportionately more so than restricted species. A number of plausible mechanisms have been proposed to explain such a pattern, including variation in niche breadth (the range of resources or conditions a species can exploit), niche position (how typical are the resources or conditions that a species can exploit), habitat selection (the tendency for species to use more habitats as they become more abundant), and metapopulation dynamics (in which dynamics in local populations depend on those in other such populations). It seems likely that a variety of potentially mutually reinforcing processes may be at work and that the pattern is an almost inevitable consequence of the aggregated spatial distributions of the individuals of most species.


Also within taxonomic groups, the body sizes and geographic range sizes of species tend to exhibit an approximately triangular form, such that although species of all body sizes may have large geographic range sizes (the upper limit normally being imposed by the size of the land mass or ocean mass), the minimum range size observed tends to increase with body size. A positive relationship likely occurs because, on average, larger-bodied species have larger-sized home ranges than smaller-bodied ones. They may thus also require larger total geographic range sizes if range-wide populations are to exceed some minimum viable size, which would tend to result in a positive interspecific range size–body size relationship, and indeed a triangular one because there is no necessary upper constraint on the range size of small-bodied species. One consequence of this mechanism is that the body sizes of the largest species tend to increase with the areas of the land masses on which they occur. 2. RANGE EDGES

Regardless of their extents of occurrence or areas of occupancy, the geographic ranges of few species are entirely congruent. Rather, the bounds fall in different places and shift position on both ecological and evolutionary time scales. Why at any given time, or averaged over a particular period, the range edges of a particular species fall quite where they do has been much debated, and for surprisingly few species is it well understood. Part of the difficulty is that the question has a variety of answers, depending on the terms in which it is couched. First, one can determine whether there are abiotic and/or biotic factors that prevent further spread, and if so what these are. Second, one can consider how, in response to these factors, the population dynamics of a species change such that it is unable to persist beyond this point. Third, one can establish the genetic mechanisms that prevent a species from evolving capacities that would enable it to overcome any limiting abiotic and biotic factors that prevent it from expanding the limits to its geographic range and becoming more widespread. Abiotic and Biotic Factors

Two principal groups of abiotic factors have been argued to limit geographic ranges, physical barriers and climate. In both cases, it is actually the interplay between these factors and the traits of the particular species that is of concern. Thus, given its dispersal abilities and behavioral tendencies, the spread of a species may be delayed or entirely prevented by expanses of inhospitable habitat, such as water for terrestrial



organisms and land for marine ones, high elevations for lowland organisms and shallow water for deep-sea ones, or grasslands for forest species and forests for grassland species. In some cases, the role of behavior may be more significant than that of dispersal ability per se, with even quite small disjunctions in the distribution of suitable habitat greatly restricting spread. In many cases, rare long-distance dispersal events effectively define what does and does not constitute a barrier. Climatic constraints on geographic ranges attract by far the majority of attention from ecologists, particularly because these may be modified by anthropogenic climate change, with implications for, among others, agriculture, forestry, fisheries, and human health. Climate doubtless limits the potential occurrence of all species because their physiological tolerances are constrained as a consequence of the costs of maintaining wide tolerances and the trade-offs associated with being adapted to particular conditions. However, demonstrating that climatic factors actually determine the position of the limits to the range of a species is more difficult. A variety of forms of evidence are strongly suggestive, including observations of systematic latitudinal variation in elevational and depth limits to species occurrences (suggesting elevational and depth responses to changes in climate with latitude), of spatial coincidence between the occurrence of range edges and particular climatic conditions, and of temporal covariation in the position of range edges and particular climatic conditions. However, such patterns could reflect covariation with some other factors, such as the distribution of a key resource, predator, or parasite, which itself is climatically limited. Rather more convincing are demonstrations of the more direct influence of climatic conditions at range edges on reproduction and mortality, perhaps most commonly reflected in the failure of species to be able to complete their life cycles beyond the range boundary. A wide variety of biotic factors have been argued to limit the geographic ranges of species, including the absence of essential resources (e.g., nutrients, prey) and the presence of competitors, predators, or parasites. The role of resources in limiting ranges is perhaps most clearly demonstrated by specialist consumers, whose distributions must be contained within that of their host. Almost invariably, such species do not occur throughout the distribution of this resource, which in the absence of other factors presumably reflects spatial variations in the abundance and quality of the host. The roles of competitors, predators, and parasites in limiting ranges frequently involve interaction among three or more species, such that predators and parasites have alternative resources to exploit and competitors

have their influence through predators and parasites (apparent competition). Although it is easiest to consider them separately, in practice the limitation of the geographic ranges of individual species by abiotic and biotic factors may often be complex. Combinations of these factors may act synergistically, and different factors may be limiting on different parts of the range boundary and at different times, on both ecological and evolutionary time scales. Population Dynamics

The edges of geographic ranges are formed at the point at which births and immigration in local populations are exceeded by deaths and emigration. If we assume that dispersal is sufficient for the establishment of a species in peripheral sites but does not otherwise influence numbers in those local populations, then the key issues are the factors that drive local extinction, which is commonly observed to be higher among populations at range edges. These factors can include demographic stochasticity (in the extreme, in small populations by chance during the same time interval all individuals may die, all may fail to breed, or sex ratios may become highly skewed), the mean environmental conditions (including resource availability, interspecific competition, predation), and the temporal variance in environmental conditions (even when conditions on average are suitable, high variance will increase the likelihood of local extinction). Although such simple scenarios at range edges may occur, immigration may often play an important role in the dynamics of local populations, enabling them to persist even when the death rates of individuals exceed the birth rates. This leads to the notion of source–sink dynamics, in which the geographic range of a species can embrace unfavorable niche conditions, such that it occurs in environments at the range edges under which in isolation any local population would rapidly become extinct. This emphasizes the potential importance of thinking about range limits in terms of the interactions among multiple local populations. Indeed, under such a scenario, limits can be formed simply because the proportion of the landscape occupied by local populations becomes sufficiently low that the influence of dispersal on local populations becomes insufficient, and the ratio of population extinctions to colonizations too high. Genetics

A number of constraints have been suggested that prevent populations at range edges from evolving the capacity to spread further. Most attention is, however, focused on the possibility that if gene flow is sufficient,

Geographic Range


the occurrence at range edges of alleles that would otherwise enable range expansion may be swamped by alleles from other populations. This is particularly likely to be the case if gene flow occurs predominantly to edge populations from the typically larger numbers of local populations and individuals that do not occur at range edges, pushing the latter away from adaptation to local optima. However, a diverse array of other mechanisms have been suggested that include low levels of genetic variation in peripheral populations, that traits show low heritability as a consequence of directional selection in marginal environments, that traits show low heritability because of environmental variability in marginal environments, that changes in several independent characters are required for range expansion and so favored genotypes occur too rarely, that genetic trade-offs between fitness in favorable and stressful environments prevent the increase of genotypes adapted to stressful conditions, that genetic trade-offs among fitness traits in marginal conditions prevent traits from evolving, and that the accumulation of mutations that are deleterious under stressful conditions prevents adaptation.

abundance data, it is difficult to target conservation or other activities at abundance hotspots. Although local abundances may not tend consistently to decline toward range edges, there may be a greater likelihood that levels of occupancy do so. That is, although the densities of individuals within local populations, and therefore of intraspecific interactions, may not do so, the densities of local populations may change systematically. This would be consistent with a scenario in which range boundaries were more often determined by reductions in the availability of suitable habitat patches rather than in the quality of those patches where they do exist. Just as there are interspecific abundance–range size relationships, there are intraspecific ones such that as the local density of a species increases through time, so does its occupancy. This has significant implications for understanding how the geographic ranges of species spread and decline and also for a variety of applied issues such as strategies for harvesting species. As geographic ranges change in size, they seem essentially to move back and forth along trajectories in abundance– range size space.



Aside from their size and boundaries, geographic ranges are structured in complex ways. This structure is determined by the distribution of the individuals of a species within the range boundaries and by variation in the traits exhibited by those individuals.

A number of systematic patterns of variation in the traits of individual organisms have been documented across the geographic ranges of species (not simply between the edges and the rest of ranges). These include spatial trends in morphology (principally body size), physiology, and life history. Indeed, some of these have been regarded as sufficiently general as to constitute ecogeographical rules. They include the neoBergmannian rule or James’s rule (an increase in the size of a species toward higher latitudes or lower temperatures), Foster’s or the island rule (smaller species become larger and larger species smaller on islands compared with mainland areas), Gloger’s rule (a tendency for endothermic animal populations in warm and humid areas to be more heavily pigmented than in cool dry areas), Jordan’s rule (fish species develop more vertebrae in cold environments than in warm ones), and one of Rensch’s rules (an increase in litter sizes of mammals and clutch sizes of birds in colder climates). In the main, such patterns in geographic range structure appear to be driven by broad spatial trends in environmental conditions or by spatial trends in the temporal variation (between seasons or years) in those conditions. They tend also to be observed predominantly among the more widespread species, whose geographic ranges extend over a greater range of environmental conditions. The trait structures of the ranges of the majority of species may thus be a good deal more complex.

Abundance and Occupancy

Across its geographic range, a species is almost invariably rare in most of the places in which it occurs and relatively abundant in only a few. That is, intraspecific species-abundance distributions are strongly rightskewed. The notion has long prevailed that areas in which a species attains higher densities tend to lie toward the center of its geographic range, with the range edge being an area of lower density. This would seem likely to follow if conditions were most favorable in the range center and declined in all directions away from that core. However, whereas this might be a useful model in the abstract, the empirical evidence to support such a pattern of abundance is limited, and there are ample examples in which it does not occur, including cases in which high abundances are found close to range edges and of marked latitudinal trends in abundance across ranges. Overall, it seems that just as they take a wide diversity of shapes, the spatial abundance structures of ranges are also very varied. This is important from an applied perspective, as without local





The study of geographic ranges has been revolutionized by dramatic increases in available data on the occurrences of species and on the environments in which they occur (particularly from remote sensing), and in the technology available to handle those data. The broad-scale perspective that these have enabled has served to highlight the significance of geographic ranges as fundamental units in ecology, and much of that understanding of population and community ecology can usefully be cast in terms of the size, distribution, and structure of geographic ranges.

Brown, James H. 1995. Macroecology. Chicago: University of Chicago Press. Gaston, Kevin J. 1994. Rarity. London: Chapman & Hall. Gaston, Kevin J. 2003. The Structure and Dynamics of Geographic Ranges. Oxford: Oxford University Press. Gaston, Kevin J., and Tim M. Blackburn. 2000. Pattern and Process in Macroecology. Oxford: Blackwell Science. Lomolino, Mark V., Brett R. Riddle, and James H. Brown. 2006. Biogeography, 3rd ed. Sunderland, MA: Sinauer Associates. Williamson, Mark. 1996. Biological Invasions. London: Chapman & Hall.

I.13 Adaptation Allan Larson OUTLINE

1. Adaptation and Darwinism 2. Adaptation as a hypothesis of evolutionary history 3. Molecular population genetics of adaptation 4. Adaptation and selfish genetic elements Darwin’s theory of natural selection explains how genetically variable populations gradually accumulate traits that enhance an organism’s ability to survive and to reproduce. Calling a particular character an adaptation denotes the hypothesis that the character arose gradually by natural selection for a particular biological role, which is called the character’s function. Any hypothesis of character adaptation is therefore a historical explanation that must specify the particular population, the interval of evolutionary time, the geographic conditions in which the relevant evolution occurred, and the nature of character variation that was sorted by natural selection. Empirical rejection of the hypothesis of character adaptation suggests the alternative hypotheses of exaptation (a character co-opted by natural selection for a biological role not associated with the character’s origin), nonaptation (a character not discriminated from alternatives by natural selection), or disaptation (a character disfavored by selection relative to alternative forms). I illustrate the contrast between adaptationist and anti-Darwinian theories of character origination using a longstanding debate concerning evolution of mimicry of wing patterns among butterfly species. I describe adaptation as a molecular population-genetic process using as an example the medical syndrome of sickle-cell anemia in African populations; depending upon its genetic and environmental contexts, hemoglobin S may constitute an exaptation, a nonaptation, a disaptation, or a component of an adaptive complex of epistatically interacting genes. Evolutionary developmental modularity and phenotypic accommodation may enhance the role of phenotypically discontinuous changes in evolution by natural selection. Selfish genetic elements likely underlie most organismal characters that arise as disaptations and nonetheless

persist despite natural selection against them. Suppression of selfish genetic elements is potentially a major source of evolution by natural selection. The explicitly historical approach to adaptation illustrated here contrasts strongly with a now largely discredited analogistic approach used in older ecological literature.

GLOSSARY adaptation (as a process). Evolution of a population

by natural selection in which hereditary variants most favorable to organismal survival and reproduction are accumulated and less advantageous forms discarded; includes character adaptation and exaptation. balanced polymorphism. Occurrence in a population of a selective equilibrium at which two or more different allelic forms of a gene each have frequencies exceeding 0.05. character adaptation. A character that evolved gradually by natural selection for a particular biological role through which organisms possessing the character have a higher average rate of survival and reproduction than do organisms having contrasting conditions that have occurred in a population’s evolutionary history; adaptation in this usage contrasts with disaptation, exaptation, and nonaptation. developmental constraint. A bias in the morphological forms that a population can express caused by the mechanisms and limitations of organismal growth and morphogenesis. disaptation. A character that decreases its possessors’ average rate of survival and reproduction relative to contrasting conditions evident in a population’s evolutionary history; a primary disaptation is disadvantageous within the populational context in which it first appears; a secondary disaptation acquires a selective liability not present at its origin as a consequence of environmental change or an altered genetic context.



exaptation. Co-option of a character by natural selec-

tion for a biological role other than one through which the character was constructed by natural selection. function. The biological role through which an adaptive character was constructed by natural selection. gradualism. Accumulation of individually small quantitative changes in a population leads to qualitative change; contrasts with saltation, in which a single genetic change induces a large qualitative change in phenotype. mimicry. Evolution by natural selection in which a character is favored because it closely resembles one present in a different species; the species whose character is copied by a ‘‘mimic’’ is called the ‘‘model.’’ modularity. Evolution of developmental constraints by which one of two or more alternative, qualitatively different suites of characters can be activated by particular genetic or environmental cues. nonaptation. A character not selectively distinguishable from contrasting conditions present in the evolutionary history of a population. saltation. Evolution of a large, qualitative change in phenotype in a single mutational step; contrasts with gradualism. selfish genetic element. Genes that spread at a cost to the organism; stretches of DNA that act narrowly to advance their own proliferation or expression and typically cause negative effects on nonlinked genes in the same organism (modified from Burt and Trivers, 2006). 1. ADAPTATION AND DARWINISM

Among the various meanings given to the term adaptation in evolutionary ecology, synonymy with evolution by natural selection is probably the most common one. Darwin’s theory of natural selection explains how genetically variable populations accumulate traits that enhance an organism’s ability to survive and to reproduce by making resources more accessible (see chapter I.14). Less-favorable alternative traits decline in frequency and are lost from the population because their possessors lose the struggle for survival and reproduction. A population produces variant forms at random with respect to an organism’s needs, and natural selection retains only the advantageous forms. Closely allied with Darwin’s theory of natural selection is his theory of gradual change. Darwin considered abrupt changes of organismal form or physiology likely to disrupt normal functioning and thereby to be discarded by natural selection. The favorable traits that natural selection accumulates across generations each

contribute only small phenotypic effects in the traditional Darwinian hypothesis. Evolution of a qualitative change in organismal form, such as the origin of a new anatomical structure or color pattern, occurs gradually across many generations as natural selection increases the populational frequencies of many small component parts so that they come to reside in the same individuals. Natural selection acting on incremental variation thus provides Darwin’s major explanation for evolution of novel organismal forms. To call a particular character an adaptation denotes the hypothesis that the character arose gradually by natural selection for a particular biological role, which is termed the character’s function (Gould and Vrba, 1980). Any hypothesis of adaptation is a historical explanation that must specify a particular population, interval of evolutionary time, and geographic conditions in which the relevant evolution occurred. Gould and Vrba (1980) use an extant population as the focal point for analysis and restrict the term adaptation to a character whose current utility matches the function for which the character arose by natural selection. They apply the contrasting term exaptation to a character co-opted by natural selection for a biological role not associated with the character’s origin. One need not restrict hypotheses of adaptation versus exaptation to extant populations, but the historical frame of reference must make explicit the temporal and spatial dimensions across which the relevant evolutionary processes occurred. I illustrate the contrast between adaptationist and anti-Darwinian theories of character origination using a longstanding debate concerning evolution of mimicry of wing patterns among butterfly species. Many cases are documented in which two or more butterfly species share the same potential avian predators and also share closely matched patterns of warning coloration on their dorsal wing surfaces. Because an avian predator learns to associate specific warning coloration with distastefulness, a distasteful ‘‘model’’ species often evolves characteristic warning coloration. Other species that share the same potential predators as the model can gain a selective advantage by ‘‘mimicking’’ the warning coloration of the model species. In some cases, the mimic species is a desired prey item that tricks its potential predator by adopting warning coloration deceptively (called Batesian mimicry). If the mimic is distasteful, sharing of the same warning coloration among species provides mutual benefit (called Mu¨llerian mimicry). In each case, evolution of warning coloration deters avian predators, all of which seek their prey visually and learn to associate particular wing patterns with distasteful prey. The adaptationist and anti-Darwinian explanations of Mu¨llerian mim-

Adaptation icry concur that natural selection for a shared warning pattern benefits members of each species because a predator needs to learn only one warning pattern to reduce mortality in each species. In the late 1800s, the anti-Darwinian orthogenetic evolutionist Theodor Eimer used butterfly mimicry among other empirical examples to support an argument that natural selection cannot construct complex morphological characters by accumulating gradual changes. He argued that butterfly species have inherited from their common ancestor similar mechanics of wing development and shared biases in production of new patterns; genetic changes that introduce pigments onto a wing surface are therefore likely to produce similar geometric patterns in all species that share a particular set of developmental mechanisms. Natural selection acts to preserve shared warning coloration in multiple species, but the specific pattern is formally an exaptation; it is a consequence of developmental mechanics, not something evolved gradually by natural selection acting on randomly produced variation in pigmentation. Ronald Fisher in 1930 used butterfly mimicry to support the opposite, adaptationist hypothesis: a mimic species gradually evolves a sequentially improved match to its model by natural selection acting on many genes whose variation exerts random and incremental effects on pigment deposition across the wing surface. The detailed matching of the model’s pattern and coloration by the mimic species therefore constitutes character adaptation. In the 1980s, John Turner reported detailed genetic analyses of Mu¨llerian mimicry among South American species of Heliconius butterflies to reconstruct the genetic histories of evolution of their mimicry patterns (figure 1). He concluded that genetic changes of major phenotypic effect were important for producing close matches in pigmentation pattern among geographically codistributed butterfly species and that subsequent improvement of the matching occurred by accumulating multiple genetic changes of smaller phenotypic effect. This interpretation supports Eimer’s general hypothesis that shared developmental constraints explain evolution of shared patterns and that mimicry evolves by exaptation; only the detailed fine-tuning of the matched patterns, as explained by Turner, constitutes character adaptation as argued by Fisher. 2. ADAPTATION AS A HYPOTHESIS OF EVOLUTIONARY HISTORY

Each specific case of butterfly mimicry involves separate evolutionary histories of at least two species, and hypotheses of character adaptation versus exaptation


therefore must be tested separately for each case. The prevailing pattern reported by John Turner for Heliconius populations might or might not prevail in other groups. In each separate test of a hypothesis of adaptation, one seeks evidence capable of rejecting the claim that a hypothetically adaptive character arose gradually through accumulation of many genetic variants, each of which gave its possessors a higher net rate of converting resources into survival, growth, and/or reproduction (¼ ‘‘Darwinian fitness,’’ see chapter I.14) than did the alternatives with which it formed population-level polymorphisms. I emphasize the importance of historical precision in formulating and testing hypotheses of adaptation because careless uses of adaptation have elicited condemnation of adaptationist studies. I agree with the critics that one must resist an analogistic tradition in which one equates as similar or equivalent the character variation and selection pressures described for distantly related species. For example, claims in sociobiological literature that one can use behavioral ecological studies of ‘‘helpers at the nest’’ in a bird species to explain analogous behaviors in human families must be rejected as having no historical equivalence. Evolution by natural selection depends as critically on the specific character variation produced in a population and the genetic structure of that variation as it does on environmental conditions. The kinds of phenotypic variation produced independently in different species are comparable only to the extent that hom*ologous developmental mechanisms channel the morphological expression to a few major alternative forms in each case, as appears to occur in wing patterns of Heliconius butterflies. Historical hypotheses of adaptation can be categorized as microevolutionary or macroevolutionary depending on the investigator’s vantage point with respect to the historical process being studied (Rose and Lauder, 1996). A microevolutionary study measures dynamics of populational polymorphisms on a generational time scale (see chapter I.17). At this scale, an investigator must distinguish natural selection per se from the genetic response of a population to natural selection. Because the relationship between genotype and phenotype is complicated by genetic dominance and epistasis and phenotypic plasticity, natural selection on phenotypic variation does not guarantee a particular genetic response of the population to selection. Agricultural geneticists are well aware that selecting for a favorable characteristic in a crop species does not always cause a corresponding genetic improvement of the population in the following generation. A macroevolutionary study, by contrast, begins with the knowledge that a particular evolutionary


6 7

6 1 11 6



5 3



12 6 1 11 6 10


2 8


7 9





3 7 5


12 1 1













8 9














B 1



B 8














B A Figure 1. Evolution of shared antipredatory warning patterns by multiple geographic races of Heliconius melpomene (left) and H. erato (right) as interpreted by John R. G. Turner. Numbered areas on each map (top) indicate geographic distributions of corre sponding numbered wing patterns below the map. Four inferred ancestral wing patterns (A D) also are shown for each species. Tree diagrams show genetic substitutions of major effect that transform one color pattern to another by adding or subtracting large areas of pigmentation. Shared developmental constraints by

A these species likely underlie their parallel, saltational origin of the same major patterns; subsequent fine tuning of the match between patterns of geographically codistributed races occurs by polygenic changes compatible with an interpretation of gradual adaptive evolution. (After Turner, J.R.G. 1981. Adaptation and evolution in Heliconius: A defense of neoDarwinism. Annual Review of Ecology and Systematics 12: 99 121. Ó 1981 by Annual Reviews, www Used with permission.)

Adaptation change, as inferred by phylogenetic analysis, has occurred (see chapter I.16). The unanswered question is whether the organismal variation and environmental contexts of a particular character transition are compatible with a specific selective explanation. For example, one can reject the macroevolutionary hypothesis that bird feathers evolved by natural selection for utility in flight because the fossil record shows that evolution of feathers preceded evolution of flight in birds. The utility of feathers for flight in living birds is therefore an exaptation, although details of the size and shape of wing feathers in particular species might constitute adaptations for flight evolved more recently in the species’ evolutionary histories. 3. MOLECULAR POPULATION GENETICS OF ADAPTATION

I illustrate adaptation as a process with a strong empirical example of evolution by natural selection in the microevolutionary mode. Adaptation of human populations to resist malarial infection is perhaps the best case study in terms of documenting evolutionary change at the both the molecular genetic and phenotypic levels and in measuring critical environmental variables. My discussion draws on Templeton’s (2006) synthetic analysis of relevant medical and epidemiological data with comments on the respective roles of character adaptation versus exaptation. The medical syndrome of sickle-cell anemia in central African populations is perhaps the best-known case of evolution by natural selection at the molecular population-genetic level. Epidemic malaria was likely established in Central Africa as a consequence of agricultural practices introduced there about 2000 years ago (Templeton, 2006). The gene whose variation illustrates selectively guided change is the gene encoding b-hemoglobin. The most common and inferred ancestral allelic form of b-hemoglobin possesses as its sixth amino acid glutamic acid and is called ‘‘hemoglobin A.’’ A single mutational change to hemoglobin A substitutes valine at this position to produce an alternative allele called ‘‘hemoglobin S.’’ The S allele could have arisen from A more than once by mutation in different human populations. Hemoglobin S has a genetically dominant phenotype of malarial resistance. Hemoglobin S molecules form sickle-shaped aggregations within an erythrocyte under low-oxygen conditions; erythrocytes distorted by these aggregations are promptly destroyed by the spleen. Infection of an erythrocyte by a malarial parasite deprives the cell of oxygen, causing sickling; the spleen then destroys the infected erythrocyte and its malarial parasite before the parasite has completed its life cycle. A person heterozygous for the


A and S alleles thereby gains resistance to malaria from hemoglobin S, and hemoglobin A has a genetically dominant phenotype for normal respiration under most environmental conditions. An individual hom*ozygous for the S allele suffers the severe respiratory disability called ‘‘sickle-cell anemia.’’ Severe anemia is therefore a recessive phenotype of hemoglobin S. Before the introduction of epidemic malaria into Africa, the S allele would have been kept rare by natural selection because the SS hom*ozygous individuals have greatly diminished chances of surviving to adulthood. The allele is preferentially removed from the gene pool by natural selection when it occurs in the SS genotype. If mating is random with respect to genotypic variation at the b-hemoglobin locus, the rare S allele occurs almost exclusively in heterozygous AS genotypes, and its selective consequences depend mainly on its phenotypic consequences in the AS genotype (a consequence of Hardy-Weinberg equilibrium; see chapter I.15). In a malarial environment, selection favors AS individuals, thereby increasing the frequency of the S allele. Occurrence of the S allele at frequencies exceeding rarity is strongly geographically coincident with epidemic malaria (figure 2). An increase in frequency of S leads to more common occurrence of the selectively disfavored SS genotype producing a ‘‘balanced polymorphism’’ in which selection maintains both alleles in the population and moves allelic frequencies toward selective-equilibrium frequencies. The relative Darwinian fitnesses of the three genotypes in a malarial environment are AA (0.9), AS (1.0), and SS (0.3); the genotype with highest fitness is usually denoted 1.0 so that relative fitnesses of the other genotypes are expressed as a fraction of the optimal one. At selective equilibrium, the expected frequencies of the alleles in the population are A (0.89) and S (0.11). Because epidemic malaria in central Africa is a relatively recent introduction, many of these populations have not attained selective equilibrium for the hemoglobin ßpolymorphism. Hemoglobin S arose by mutation before the establishment of epidemic malaria in central Africa; in the malarial environment, it was co-opted for fitness consequences in AS genotypes that are incidental to the mutational origin of hemoglobin S. Hemoglobin S is thus an exaptation in the evolutionary history of central African populations. The balanced polymorphism of alleles A and S constitutes a population-level adaptation. Eradication of malaria would convert the polymorphism from adaptation to what Baum and Larson (1991) call a secondary disaptation; an environmental change causes a character that formerly had a selective advantage over its evolutionary antecedent



Distribution of hemoglobin S Distribution of Plasmodium falciparum malaria Center of origin and spread of hemoglobin C polymorphism 400 km 400 mi

Figure 2. Evolution by natural selection at the b hemoglobin locus in human populations of Africa. Polymorphisms for the hemoglobin C (circle and arrows) and hemoglobin S (crosshatching) forms of b hemoglobin are associated geographically with occurrence of epi demic malaria (Plasmodium falciparum, gray shading). Hemoglobin A is the most common allelic form in all areas. The frequency of allele C is highest ( 0.10) in the circled area and declines grad ually with distance outside the circle through the region marked by arrows. Frequencies of C and S alleles are inversely correlated in West Africa because as C becomes more common, natural selection favors C and disfavors S. Where C is absent or rare, natural selection moves S toward an equilibrium frequency of ap proximately 0.11. Depending on its environmental and population genetic contexts, hemoglobin S can be an exaptation, nonaptation, or disaptation for survival to adulthood.

(allele A close to fixation in this example) to one that is selectively disfavored relative to that condition. A polymorphism for the A and S alleles at the malarial selective-equilibrium frequencies is selectively disadvantageous relative to a population fixed for the A allele in an environment from which malaria has been eradicated; when S occurs at a frequency of *0.1, approximately 1% of the polymorphic population suffers severe anemia (SS individuals), and the S allele no longer confers a selective advantage in AS genotypes. As the frequency of S drops to very low values in a nonmalarial environment, the S allele occurs strictly in heterozygous individuals, and presence of hemoglobin S ceases to generate natural selection. Alternative forms that have no selective consequences constitute nonaptation (a selectively ‘‘neutral’’ character lacking current utility; Vrba and Gould, 1986). I illustrate the importance of geographic variation in adaptive evolution by extending this example from central African populations to western African ones

(figure 2). A third form of b-hemoglobin allelic to hemoglobin A and hemoglobin S occurs in some African populations (Templeton, 2006): the hemoglobin C allele, derived by a single mutation from hemoglobin A, differs from both hemoglobin A and hemoglobin S in having lysine at the sixth amino-acid position. Allele C has a genetically recessive phenotype of malarial resistance and is associated with severe anemia only when heterozygous with allele S. In a malarial environment, the relative Darwinian fitnesses of the six genotypes formed by the various combinations of the A, C, and S alleles are AA (0.7), AC (0.7), AS (0.8), CC (1.0), CS (0.5), and SS (0.2). The C allele likely arose from A in Africa before the introduction of epidemic malaria, when the A allele was close to fixation and S was rare. Under these conditions, the C allele would occur entirely in AC genotypes and would be a nonaptation relative to the A allele. On malarial introduction, selection increased the frequency of S through the higher viability of AS genotypes over AA and AC genotypes. As the frequency of the S allele increases toward the selective equilibrium frequencies (A ¼ 0.89, S ¼ 0.11), the C allele would appear in CS genotypes as well as AC genotypes, and selection would decrease its frequency by disfavoring CS individuals. Selection acts to keep the C allele rare under these conditions, in which C is a disaptation relative to A. Although the CC genotype is the most favorable possible condition, the CC genotype is too rare under conditions of random mating to contribute a selective advantage. Because alleles A and C are selectively equivalent in a nonmalarial environment, random genetic drift (see chapter I.15) might increase the frequency of C in a local population to a frequency high enough that CC genotypes occur regularly. Introduction of epidemic malaria into such a population would act very differently than it does in the population described in the preceding paragraph. Over many generations, the net effect of natural selection would be to increase frequency of the C allele toward fixation and to decrease frequencies of A and S ultimately to zero. Allele C constitutes an exaptation for malarial resistance in some western African populations (figure 2) in which allele C had drifted to sufficiently high frequencies before epidemic malaria that CC individuals were produced and subject to selective retention in a malarial environment. The contrasting selective consequences of the A, C, and S alleles of b-hemoglobin under slightly different starting frequencies of C and S in African malarial environments show that adaptive evolution depends critically on particular historical conditions. Another dimension to adaptive evolution is gene exchange among different geographic populations of a species. As western African populations evolve by se-

Adaptation lection to increase the frequency of the C allele and central African populations evolve toward selectiveequilibrium frequencies of the A (0.89) and S (0.11) alleles of b-hemoglobin, gene exchange occurs by interbreeding among these populations. Because fixation of allele C is the superior adaptive condition with respect to these three alleles, preferential contribution of C alleles from the favored west African genotypes should enable all populations eventually to undergo adaptive evolution in the manner of the west African populations by fixing C and eliminating A and S. Further analysis of geographic variation in evolution of malarial resistance by Templeton (2006) reveals several cases in which genetic epistasis between the sickle-cell polymorphism at b-hemoglobin and variation at other loci produces different kinds of adaptive evolution. In Greek and Arabian populations, hemoglobin S occurs against a genetic background in which a mutation at a genetically linked locus causes fetal hemoglobin to be expressed throughout adulthood (called ‘‘persistence of fetal hemoglobin’’) rather than ceasing its expression after birth. This combination of alleles at two loci provides a simple example of a ‘‘coadapted gene complex’’; hemoglobin S provides malarial resistance, while persistence of fetal hemoglobin alleviates the severe anemia associated with SS hom*ozygotes. High fitness depends on a particular combination of alleles at these two genes. One expects evolution of many advantageous phenotypes to occur by natural selection increasing the frequencies of selectively favored alleles at many genes. A consequence of such selection is that new mutations arising in different individuals and at different times in a population’s history can be brought to high frequency independently by selection, thereby increasing their chances of occurring together in the same individuals by genetic recombination. If the combination of alleles thus achieved is favored by selection for the same biological role that brought the individual mutations to high frequency, selection gradually constructs an adaptive composite character. Such characters can become developmentally and genetically integrated into modules, whose expression or suppression during development can provide a store of potential exaptations, sometimes called a ‘‘toolkit’’ for constructing new organismal forms. A toolkit of adaptively evolved modules makes possible further evolution not confined to traditional Darwinian gradualism. A developmental module can be activated potentially by genetic or environmental factors or their interactions; ‘‘phenotypic accommodation’’ denotes a beneficial modification of organismal development made in response to a novel behavioral or environmental stimulus (West-Eberhard, 2005).


Hypotheses of developmental modularity and phenotypic accommodation share the expectation that evolutionary saltations are more likely than acknowledged by traditional Darwinism. A well-studied case of polymorphism for discrete developmental modules involves two constrasting feeding morphologies in tropical American fish of the genus Cichlasoma. The contrasting forms differ abruptly in the structures of jaws located in the pharynx and used to crush food (figure 3). The alternative states of the pharyngeal jaws are termed the ‘‘papilliform’’ morph versus the ‘‘molariform’’ morph; the molariform morph has hypertrophied skeletal and muscular components and greater ability to masticate hard food items, such as snails, which are often the less preferred food items. In some species, consuming snails early in life appears facultatively to trigger development of the molariform morph. The morphological contrast between these states in C. minckleyi of Mexico is so great that it has been called ‘‘intraspecific macroevolution.’’ Widespread occurrence of molariform and papilliform feeding morphs among cichlid species indicates that they likely represent alternative developmental modules first assembled in the ancient history of cichlid fishes. The evolutionary origin of the alternative morphs is a macroevolutionary question, testable using interspecific phylogenetic analyses of how one form was constructed, perhaps gradually, from the other one. Microevolutionary studies of the alternative conditions as polymorphisms within species involve saltational changes governed by developmental switches and a prominent role for character exaptation in the adaptive trophic evolution of polymorphic populations. 4. ADAPTATION AND SELFISH GENETIC ELEMENTS

An important extension of the Darwinian evolutionary framework is to recognize semiautonomous selective processes occurring at the levels of genomic elements and species lineages in addition to the traditional level of varying organisms within populations. The abstract concepts of character sorting and selection have been expanded to encompass these levels (Gould, 2002; Vrba and Gould, 1986). At the genomic level, characteristic structures of retrotransposons include long terminal repeats and a coding region for reverse transcriptase, whose biological role is integral to the operation of the transposable element but whose origin cannot be explained as an organismal-level character adaptation. Unlike the genes encoding b-hemoglobin, mutational changes in transposable elements cannot be interpreted as having reached high evolutionary frequency to serve an organismal-level function, although their consequences can be co-opted as exaptations for



Figure 3. Cineradiographic tracing of the upper and lower pha ryngeal jaws contacting to grind a snail in the molariform morph of Cichlasoma minckleyi. Arrows denote the directions of movement of the jaws as they make contact. This hypertrophied molariform morph and a much smaller papilliform morph represent alternative developmental modules in this and other species of Cichlasoma. The molariform morph can masticate hard food items, such as snails, not available to the papilliform morph. (After Liem, K. F., and L. S. Kaufman. 1984. Intraspecific macroevolution: Functional bi ology of the polymorphic cichlid species, Cichlasoma minckleyi. In A. A. Echelle and I. Kornfield, eds., Evolution of Fish Species Flocks. Orono: University of Maine Press, 196 217)

organismal-level roles. Burt and Trivers (2006) use ‘‘selfish genetic element’’ todenote ‘‘the minority of genes that spread at a cost to the organism’’ and ‘‘stretches of DNA . . . that act narrowly to advance their own interests’’ and ‘‘typically cause negative effects on nonlinked genes’’ in the same organism. For example, during spermiogenesis of male mice heterozygous for the t-allele, developing sperm containing the t-allele disable those containing the wild-type allele, permitting the t-allele to persist in populations despite selection against its detrimental effects on organismal phenotype, such as absence of a tail. Someone studying occurrence of the tailless phenotype in mice in a Darwinian context would conclude that this trait is a

primary disaptation (Baum and Larson, 1991), a character disadvantageous relative to its ancestral alternative condition in the environmental context of its origin. Identification of primary disaptation using the adaptationist methodology described above would lead one to hypothesize association of such a phenotype with evolution of selfish genomic elements. Selfish genomic elements that harm their ‘‘host’’ organism incur natural selection for their suppression by other genetic functions. One therefore expects to observe evolution of organismal-level adaptations whose function is to suppress selfish genomic elements. Given the phylogenetically widespread occurrence of the selfish genetic elements reviewed by Burt and Trivers (2006), mechanisms evolved to stabilize genomic structure and function probably constitute a large portion of the adaptive evolutionary diversity of life. The basic concepts of adaptation and exaptation and historical methods for testing specific hypotheses of adaptation as discussed above are directly applicable to studies of inherent conflicts between organismal character adaptation and the proliferative drives of selfish genomic elements. FURTHER READING Baum, D. A., and A. Larson. 1991. Adaptation reviewed: A phylogenetic methodology for studying character macro evolution. Systematic Zoology 40: 1 18. Burt, A., and R. Trivers. 2006. Genes in Conflict: The Biology of Selfish Genetic Elements. Cambridge, MA: Harvard University Press. Gould, S. J. 2002. The Structure of Evolutionary Theory. Cambridge, MA: Harvard University Press. Gould, S. J., and E. S. Vrba. 1982. Exaptation: A missing term in the science of form. Paleobiology 8: 4 15. Rose, M. R., and G. V. Lauder. 1996. Adaptation. San Diego: Academic Press. Templeton, A. R. 2006. Population Genetics and Micro evolutionary Theory. Hoboken, NJ: John Wiley & Sons. Vrba, E. S., and S. J. Gould. 1986. The hierarchical expansion of sorting and selection: Sorting and selection cannot be equated. Paleobiology 12: 217 228. West Eberhard, M. J. 2005. Phenotypic accommodation: Adaptive innovation due to developmental plasticity. Journal of Experimental Zoology 304B: 610 618.

I.14 Phenotypic Selection David W. Pfennig and Joel G. Kingsolver OUTLINE

1. 2. 3. 4. 5. 6.

Introduction How phenotypic selection works Measuring phenotypic selection Phenotypic selection in the wild Misunderstandings about phenotypic selection Future directions

In this chapter, we describe the strength and patterns of natural selection in the wild. We focus on phenotypic selection because natural selection acts on the phenotypes of individual organisms. We begin by explaining what phenotypic selection is and how it works. We then explore how scientists study phenotypic selection in natural populations and discuss general patterns that have emerged from such investigations. Finally, we address common misunderstandings about selection and identify profitable avenues for future research.

GLOSSARY fitness. The extent to which an individual contributes

its genes to future generations relative to other individuals in the same population; a good operational definition of fitness is an individual’s relative reproductive success. heritability. In the broad sense, the fraction of the total phenotypic variation in a population that can be attributed to genetic differences among individuals; in the narrow sense, that fraction of the total phenotypic variation that results from the additive effects of genes. natural (phenotypic) selection. A difference, on average, between the survival or fecundity of individuals with certain phenotypes compared with individuals with other phenotypes. phenotype. The outward characteristics of organisms, such as their form, physiology, and behavior. quantitative trait. A trait that shows continuous rather than discrete variation; such traits are determined

by the combined influence of many different genes and the environment. selection gradient. A measure of the strength of selection acting on quantitative traits: for selection on a single trait, it is equal to the slope of the best-fit regression line in a scatterplot showing relative fitness as a function of phenotype; for selection acting on multiple traits, it is equal to the slope of the partial regression in a scatterplot showing relative fitness as a function of all phenotypes. sexual selection. A difference, among members of the same sex, between the average mating success of individuals with a particular phenotype and that of individuals with other phenotypes. 1. INTRODUCTION

In the introduction to On the Origin of Species, Darwin wrote, ‘‘a naturalist, reflecting on the mutual affinities of organic beings, on their embryological relations, their geographical distribution, geological succession, and other such facts, might come to the conclusion that each species had not been independently created, but had descended . . . from other species. Nevertheless, such a conclusion, even if well founded, would be unsatisfactory, until it could be shown how the innumerable species inhabiting this world have been modified. . .’’ (emphasis added). Thus, Darwin recognized that no theory of evolution would be complete if it failed to provide a plausible mechanism that could explain how living things change over evolutionary time. Darwin’s theory of evolution by natural selection provided such a mechanism. Yet, Darwin’s theory goes beyond explaining how living things change over time; it also explains the important concept of adaptation: the tendency for living things to evolve traits that make them so apparently well designed for survival and reproduction. Because of this broad explanatory power, Darwin’s theory ranks among the most important ideas in the history of human thought.



Although the central concept of Darwin’s theory is natural selection, Darwin never attempted to measure selection in nature. Moreover, in the century following the publication of On the Origin of Species, selection was generally regarded as too weak to be observed directly in natural populations. Partly for these reasons, some early evolutionists even questioned selection’s efficacy in driving evolutionary change. This view that selection is weak and cannot be measured has changed dramatically. Beginning in the 1930s, evolutionists demonstrated mathematically that natural selection alone could power evolutionary change and adaptation. Moreover, in the past three decades, selection has been detected and quantified in hundreds of populations in nature. These data demonstrate that not only does selection occur routinely in nature, but that it is often sufficiently potent to bring about substantial evolutionary change in a relatively short time period. Indeed, selection is now viewed as the cause of adaptive evolution within natural populations. 2. HOW PHENOTYPIC SELECTION WORKS

Phenotypic selection takes place when individuals with particular phenotypes survive to reproductive age at higher rates than do individuals with other phenotypes, or when individuals with particular phenotypes produce more offspring than do individuals with other phenotypes. In either case, selection results in differential reproductive success, where some individuals have more offspring than others. Thus, phenotypic selection requires phenotypic variation, where individuals differ in some of their characteristics, and differential reproduction, where some individuals have more surviving offspring than others because of their distinctive characteristics. Those individuals that have more surviving offspring are said to have higher fitness (note that an individual’s fitness is measured as how well the individual performs relative to other individuals in the same population). Ultimately, phenotypic selection can lead to changes in the genetic makeup of populations over time—evolution. In particular, when the phenotypic characteristics under selection are heritable—that is, when the variations among individuals are, at least in part, passed from parents to offspring—selection will cause the population to change in these characteristics over time. Thus, evolution by natural selection requires three conditions: variation, differential reproduction, and heredity. Indeed, when these three conditions are satisfied, evolution by natural selection is a certain outcome. Numerous factors in the environment can cause selection, including biological agents (such as an individual’s competitors, predators, and parasites) and

nonbiological agents (such as the weather). The specific phenotypic traits on which agents of selection act are termed targets of selection. As we will see, however, selection often acts on multiple traits simultaneously in the same individual, making it a challenge to determine precisely which trait represents the actual target of selection. Although phenotypic selection always favors an increase in fitness, it does not invariably bring about the evolution of greater trait values. In particular, when selection acts on quantitative (i.e., continuously distributed) traits, three different modes of selection are possible, each of which produces a distinctive pattern of trait evolution (figure 1). With directional selection, fitness consistently increases (or decreases) with the value of the trait. When directional selection acts on a trait, it changes the value of that trait in the population. Directional selection also tends to reduce variation, although often not dramatically. With stabilizing selection, individuals with intermediate trait values have highest fitness. Stabilizing selection does not tend to change the mean trait value. It does, however, reduce variation by disfavoring individuals in the tails of the trait’s distribution. Finally, with disruptive selection, individuals with extreme trait values have highest fitness. As with stabilizing selection, disruptive selection does not tend to change the mean trait value. Unlike stabilizing selection, however, disruptive selection increases variation by favoring individuals in the tails of the trait’s distribution. All three modes of selection drive evolution by eliminating individuals with low fitness and preserving individuals with high fitness. Moreover, as noted earlier, if the trait of interest is heritable, then evolution will result, but the trait distribution in the evolved population will differ depending on the mode of selection (see figure 1). In particular, for traits under positive directional selection, the population will evolve larger trait values (illustrated in figure 1), whereas for those under negative directional selection, the population will evolve smaller trait values. For traits under stabilizing selection, the population will evolve a smaller range of trait values as the average trait value becomes more common in the population. Finally, for traits under disruptive selection, the population will evolve a wider range of trait values, possibly leading to the evolution of discrete, alternative phenotypes (see figure 1). Given this background, we now turn to the issue of how to measure the mode and strength of phenotypic selection. 3. MEASURING PHENOTYPIC SELECTION

Suppose we are interested in measuring possible selection acting on some trait in a population. The first


Frequency of individuals

Phenotypic Selection

Phenotypes (beak size)

Stabilizing selection

Disruptive selection


Directional selection

Frequency of individuals

Selection favors larger beaks Selection favors mid-sized beaks

Original population

Selection favors small and large beaks

Evolved population

Phenotypes (beak size) Figure 1. Three different modes of selection (directional, stabilizing, and disruptive) that may act on a quantitative trait (i.e., a trait that shows continuous rather than discrete variation). The top panel shows the distribution of beak sizes in a hypothetical population of birds before selection; the middle panels show fitness associated

with different beak sizes during different modes of selection; and the bottom panels show the distribution of beak sizes following each form of selection. Note that for different modes of selection, the shape of the line relating fitness to phenotype varies (middle panels), as does the resulting pattern of trait evolution (bottom panels).

step is to estimate the fitness associated with different trait values. Ideally, we would identify individuals with different trait values and measure their overall fitness. In practice, however, most investigators measure only one component of fitness, such as survival, mating success, fecundity, or (even less directly) a trait that correlates with these fitness components, such as body size. Once we estimate fitness, we then fit a regression line (i.e., the best-fit line) through the data points relating fitness to phenotype. From the slope and shape of this regression line, we can determine the strength and mode of selection acting on our trait of interest. When this fitness function is described by a straight line (indicating directional selection; figure 1), the fitness (w) of the trait (z) can be estimated by the simple linear regression equation:

where a is the y-intercept of the fitness function and b is the fitness function’s slope. In this case, b measures the strength of directional selection. By contrast, when the fitness function has curvature (indicating stabilizing and disruptive selection; figure 1), quadratic regression is required to estimate the strength of selection. Here, fitness is estimated by:

w ¼ a þ bz,

w ¼ a þ bz þ (g=2)z2 , where g measures the amount of curvature in the fitness function. In this case, g measures the strength of quadratic selection. When b ¼ 0 and g is significantly negative (i.e., when the fitness function contains an intermediate performance maximum), we conclude that stabilizing selection is acting on the trait of interest. By contrast, when b ¼ 0 and g is significantly positive (i.e., when the fitness function contains an intermediate


Autecology A.


Omnivore morph

Smooth beak, small jaw muscles, long gut

Fitness (log body size, mm)


Carnivore morph

Serrated beak, large jaw muscles, short gut


















1 -6


1 0 3 6 -6 -3 0 Phenotype (morphological index; more positive values are more carnivore-like) 0


1 6 -6




Figure 2. Spadefoot toad tadpoles (Spea bombifrons and S. mul tiplicata) are highly variable in resource use and feeding mor phology as represented by two extreme morphotypes: (A) an om nivore morph, which feeds mostly on detritus, and (B) a carnivore morph, which specializes on fairy shrimp. The mode of selection operating on feeding morphology varies for different species and

populations as revealed when the fitness of individual tadpoles is plotted on phenotype for (C) S. bombifrons from mixed species ponds, (D) S. multiplicata from mixed species ponds, and (E) S. multiplicata from single species ponds. Each panel (C E) shows cubic spline regression estimates bracketed by 95% confidence intervals.

performance minimum), we conclude that disruptive selection is acting. To illustrate how each mode of selection may be manifest in natural populations, consider a recent study of spadefoot toad tadpoles by Pfennig and colleagues (2007). Tadpoles of two species from the southwestern United States, Spea bombifrons and S. multiplicata, are highly variable in resource use and feeding morphology as represented by two extreme morphotypes: an omnivore morph (figure 2A), which feeds mostly on the pond bottom on detritus (decaying organic material), and a carnivore morph (figure 2B), which feeds mostly in the water column on fairy shrimp. In some ponds, there is a clear dimorphism in feeding morphology; in other ponds, individuals with intermediate phenotypes may be most common. The mode of selection operating on feeding morphology varies for different species and populations. In mixed-species ponds (i.e., ponds containing both species), the most carnivore-like S. bombifrons tadpoles are largest (figure 2C; body size serves as a suitable proxy for fitness because larger individuals have higher survival, mating success, and fecundity in this system).

Thus, directional selection favors more carnivore-like S. bombifrons. Presumably, this pattern reflects selection on S. bombifrons to express resource-use phenotypes that minimize their overlap with S. multiplicata for food; S. multiplicata tend to be more omnivore-like than S. bombifrons. A different mode of selection was detected among S. multiplicata in mixed-species ponds. In this species, stabilizing selection appears to favor individuals with intermediate phenotypes (figure 2D). Presumably, carnivore phenotypes in these individuals are selectively disfavored; earlier work had shown that S. multiplicata carnivores are competitively inferior to S. bombifrons. Yet why does selection not favor omnivores, which are as distinct as possible from S. bombifrons? Presumably, selection acts against S. multiplicata omnivores in mixed-species ponds because omnivores metamorphose later and at a smaller body size than carnivores. Because mixed-species ponds typically contain relatively high shrimp densities, those S. multiplicata that express an intermediate feeding morphology—and can thereby supplement their detritus diet with, but not specialize on, the more nutritious shrimp resource—may be

Phenotypic Selection selectively favored. Thus, in mixed-species ponds, selection appears to favor S. multiplicata that are as carnivore-like as possible, but that are not so carnivore-like that they overlap with S. bombifrons in resource use. Finally, a third mode of selection was detected among S. multiplicata in single-species ponds (figure 2E). Here, disruptive selection favors extreme feeding morphologies. In these ponds, individuals expressing extreme phenotypes would most likely have fewer (and, in the case of extreme omnivores, perhaps lowerquality) resources available. Nevertheless, compared with the majority of the population that may be intermediate in phenotype (and in resource use), individuals expressing extreme phenotypes would also most likely have fewer competitors with which to share those resources. Thus, relative to intermediate individuals, the fitness of extreme omnivores and carnivores may be high. Although the above example illustrates the general approach that is widely used for measuring phenotypic selection in the wild, a critical assumption behind this approach is that variation in the measured trait causes the observed variation in fitness. However, rather than acting directly on the trait of interest (through direct selection), selection may be acting on other, unmeasured traits that are correlated with the measured trait (through indirect selection), generating a spurious correlation between the focal trait and fitness. One way to reduce the problem of indirect selection is to experimentally alter the trait of interest and then evaluate the effects of the manipulation on subsequent fitness (phenotypic engineering). To illustrate the latter approach, consider the following example. Male long-tailed widowbirds, Euplectus progne, are endowed with a half-meter-long tail. Malte Andersson hypothesized that these extraordinary tail feathers are selectively favored because females find them attractive; i.e., long tail feathers are favored by sexual selection. To test this hypothesis, Andersson predicted that experimentally augmenting a male’s tail feathers should enhance the male’s fitness. Andersson captured male widowbirds and then shortened the tails of some by removing a segment of tail feathers, only to glue them onto another bird’s tail, thereby lengthening the latter bird’s tail. He also had two control groups: one in which the male’s tail feathers were cut off and glued back on, and another in which the males were handled in the same way but no tail feathers were removed. The results of this phenotypic manipulation were dramatic. The tail-lengthened males were much more attractive to females than those that had suffered the loss of a portion of their tail feathers. Moreover, the


tail-lengthened males also did better than controls. These data therefore indicate that tail length is a target of selection, with females acting as the selective agent. Because it can expand the range of phenotypic values and reduce the problem of correlated traits, phenotypic engineering is especially useful for determining whether a trait is under direct selection and what mode of selection might operate on it. However, because phenotypic engineering often involves altering trait expression beyond the range of trait values observed in natural population, such manipulations do not help researchers estimate the strength of selection on natural populations in the wild, which is the topic we turn to next. 4. PHENOTYPIC SELECTION IN THE WILD

Numerous studies have used the above approaches to measure phenotypic selection in natural populations. Moreover, many of these studies measured selection acting on multiple traits on the same individual. Such data are particularly valuable because they allow us to distinguish direct selection on traits from the indirect effects of correlated traits. To estimate direct selection, we use a statistical approach known as multiple regression analysis. Multiple regression resembles simple linear regression (introduced in the previous section) except that fitness is regressed on multiple traits simultaneously, allowing us to measure the strength of direct selection acting on each trait after statistically controlling for the effects of correlations among other traits. Specifically, fitness (w) is estimated by: w ¼ a þ b1 z1 þ b2 z2 þ b3 z3 bi zi , where bi is the partial regression slope associated with trait zi. This parameter, termed the linear selection gradient, measures the strength of direct selection acting on trait zi. (Note that for selection on a single trait, the linear selection gradient is equal to the slope of the simple linear regression, as described in the previous section.) To allow comparisons among different types of traits and organisms, we can standardize the linear selection gradient by the amount of variation in the trait (e.g., by the standard deviation) to obtain a standardized measure of selection, bs. Kingsolver and colleagues recently reviewed studies that used these approaches to measure selection gradients in natural populations. They identified 993 estimates of directional selection (bs), obtained from a diversity of organisms, ecological settings, and traits. Because positive and negative values of bs occur with similar frequency, they used the absolute values, |bs|, as



an index of the magnitude of directional selection. The median value (50% of the values above and 50% below) of |bs| was 0.16, with a small fraction of values greater than 0.50, indicating strong selection. To put this strength of selection in perspective, imagine a population that experiences persistent directional selection of this magnitude (bs ¼ 0.16) on a trait that has a typical heritability of 0.3. In less than 70 generations, the population mean would exceed the initial range of variation in the population. In other words, phenotypic selection in many natural populations is sufficiently strong to cause substantial evolutionary change in a fairly short period of time on an evolutionary time scale. Another important issue to resolve is the relative magnitude of natural selection (i.e., selection resulting from variation among individuals in survival or fecundity) compared to sexual selection (i.e., selection caused by variation among individuals in mating success). The available data on directional selection gradients suggest that sexual selection is typically stronger than natural selection. Indeed, the median magnitude of sexual selection is more than twice as great as that of natural selection. Thus, competition for mates may be important for rapid evolution in nature. What are the patterns of quadratic selection in the wild? Kingsolver and colleagues (2001, 2007) identified 574 measures of the strength of quadratic selection, g. They found that 50% of the values of g are between –0.1 and þ0.1, implying that the magnitude of quadratic selection is often modest. Moreover, the frequency distribution of g is symmetric about zero, with negative and positive values equally common, which suggests that stabilizing selection is not more common than disruptive selection. Because disruptive selection is generally thought to be relatively rare in nature, this result is particularly surprising. It is possible that this result reflects sampling bias: only 16% of the values of g in the literature are significantly different from zero. Thus, most studies do not have the sample size or statistical power to quantify quadratic selection of the magnitude that may be typical in natural populations. Alternatively, this result may reflect the true pattern of quadratic selection in nature; i.e., disruptive selection may actually be relatively common. The possible widespread occurrence of disruptive selection may reflect a ubiquitous agent of selection in nature: competition for resources, such as food. Because competition tends to decrease individual fitness, natural selection is generally thought to favor traits that lessen competition’s intensity. One way for selection to do so is to favor evolutionary divergence between initially similar phenotypes through densitydependent or frequency-dependent disruptive selection (e.g., see figure 2E).

Thus, to summarize, phenotypic selection is common in nature, and it is often sufficiently strong to cause substantial evolutionary change in a relatively short time period. Moreover, sexual selection tends to be stronger than natural selection. Finally, stabilizing selection appears to be no more common than disruptive selection. However, because few studies have focused on quadratic selection specifically, it is difficult to say how common or how strong disruptive selection is relative to stabilizing selection in natural populations. 5. MISUNDERSTANDINGS ABOUT PHENOTYPIC SELECTION

Phenotypic selection is often misunderstood. We therefore highlight and clarify four common misunderstandings. Misunderstanding 1: Phenotypic Selection Always Results in Evolution

Selection and evolution are not the same, although the two concepts are often incorrectly equated. Selection is a process that produces evolution, whereas evolution is the historical pattern of change through time. Phenotypic selection (the process) can lead to evolution (the pattern), but it is only one of several processes that can do so (the others are mutation, gene flow, nonrandom mating, and genetic drift). Moreover, if a trait lacks heritable variation, selection will not produce evolution. Misunderstanding 2: Phenotypic Selection Causes Individuals to Change

A common misconception is that individual organisms evolve following selection. It is true that phenotypic selection acts on the phenotypes of individual organisms. However, after the selection event, none of the selected individuals are expected to change in any way. What does change are characteristics of the population. Thus, populations evolve; individual organisms do not. Phenotypic selection may indirectly cause the phenotypes of individual organisms to change. Specifically, agents of selection often alter the developmental expression of traits through a process known as phenotypic plasticity. When phenotypes are plastic, individuals that are genetically identical may express radically different phenotypes if they develop in different environments. For example, the spadefoot toad tadpoles in figure 2 are born as omnivores but may develop into carnivores following a change in their diet. In many species, individuals often exhibit heritable variation in their tendency to respond to environ-

Phenotypic Selection mental cues through phenotypic plasticity, indicating that plasticity itself is subject to natural selection and evolutionary change. Indeed, adaptive phenotypic plasticity is thought to evolve because it enables organisms to produce the optimal phenotype for the various environments that they may experience during their lifetime. Thus, by favoring the evolution of phenotypic plasticity, agents of selection may indirectly change the phenotypes of individual organisms. Misunderstanding 3: Selection Favors Individuals That Act for the Good of the Species

A common misunderstanding about phenotypic selection acting on behavior is that individual organisms will perform actions for the good of their species. However, if altruists survive and reproduce at lower rates than other individuals in the same population, then the tendency to behave altruistically should not evolve, unless the altruists receive some other benefit. As it turns out, nearly every act of altruism that has been studied in detail increases the altruist’s fitness, either because beneficiaries reciprocate or because the beneficiaries are genetically related to the altruist. Helping nondescendant kin (relatives other than offspring) can increase an altruist’s fitness because relatives share genes. Moreover, fitness gained by personal reproduction (direct fitness) and fitness gained by helping nondescendant kin (indirect fitness) can both be expressed in identical genetic terms. We can sum up an individual’s total contribution of genes to the next generation, creating a quantitative measure called inclusive fitness. Thus, altruism may be adaptive if it ultimately results in more shared genes being transmitted to the next generation. In general, natural selection should always favor traits that maximize an individual’s inclusive fitness. Misunderstanding 4: The Evolutionary Response to Phenotypic Selection Is Slow

It is often assumed that the evolutionary response to selection is slow. We have already seen, however, that phenotypic selection is often sufficiently strong to cause substantial evolutionary change in a relatively short time. Moreover, phenotypic selection may even produce substantial evolutionary change in only one generation. Consider a population that contains abundant phenotypic variation. If this variation has high heritability, and if there is strong truncating selection, in which individuals with a trait value above a certain threshold value survive or reproduce while those below this value do not, then the population will evolve dramatically in only one generation.


For example, Peter and Rosemary Grant recently documented character displacement—evolution in resource-acquisition traits stemming from competition between species—in a species of Gala´pagos finch that recently (i.e., in the last 25 years) confronted a novel competitor (Grant and Grant, 2006). Remarkably, their data suggest that the focal species may have evolved away from its competitor in beak morphology in only one generation. Thus, paradoxically, evolution may happen so rapidly that we may actually fail to detect it. 6. FUTURE DIRECTIONS

As we have seen, numerous recent studies have measured phenotypic selection in the wild. Many interesting patterns have emerged from these studies. However, a number of questions remain unanswered. Here, we list four such questions. First, does phenotypic selection vary over time and space? In particular, does the fact that environmental conditions change frequently cause the magnitude and even the direction of selection to change also? Such fluctuating selection could explain why most organisms appear to be experiencing at least some directional selection. If environments vary frequently, then the organisms living in these environments will tend to possess trait values that are suboptimal for their particular environment. Consequently, directional selection would always be acting to drive the trait value toward the current optimum. We need many more long-term field studies of selection in the wild to determine if the magnitude, direction, or mode of selection varies in time and space. Second, is disruptive selection relatively common in nature, and, if so, what agents drive it? Specifically, is disruptive selection often mediated by density- or frequency-dependent processes, such as competition? Resolving this issue is vital for understanding the origins and maintenance of alternative phenotypes in populations (e.g., see figure 2), and, possibly, the origin of new species. Third, what measure of fitness provides the most complete picture of selection? An operational definition of fitness is that it is the total number of offspring that an individual produces in its lifetime. Yet, for practical reasons, most studies consider only components of fitness, such as survival. We need more studies that determine how reliably individual fitness components predict true lifetime fitness in natural populations. We especially need more studies that compare the relative magnitude of selection on survival or fecundity (natural selection) with selection on mating success (sexual selection). As noted in section 4 above, the available data indicate that sexual selection is



typically significantly stronger than natural selection. Does this result generally hold across diverse taxa? Moreover, to develop a truly comprehensive view of how phenotypic selection drives trait evolution, we need more selection studies that determine how trait expression influences an individual’s inclusive fitness. Finally, what is the relative importance of evolution versus phenotypic plasticity in mediating rapid phenotypic responses to changing environments? Many organisms are currently undergoing rapid phenotypic change in response to ongoing human-mediated change in their environment. To what extent does such rapid phenotypic change reflect phenotypic plasticity as opposed to rapid evolution? In sum, natural selection is the central organizing principle of evolutionary theory. This theory explains not only how living things diversify but also those features of living things that so wonderfully equip them for survival and reproduction. Although natural selection is a simple concept, modern research is only beginning to discover that it works in myriad and sometimes subtle ways. FURTHER READING Andersson, Malte. 1982. Female choice selects for extreme tail length in a widowbird. Nature 299: 818 820. A classic example of the use of manipulative experiments to docu ment phenotypic selection in the wild. Conner, Jeffrey K., and Daniel L. Hartl. 2004. A Primer of Ecological Genetics. Sunderland, MA: Sinauer Associates.

An excellent overview of ecological genetics with a clear summary of how to measure selection. Endler, John A. 1986. Natural Selection in the Wild. Prince ton, NJ: Princeton University Press. A seminal monograph that describes advantages and disadvantages of various approaches for measuring phenotypic selection in natural populations. Grant, Peter R., and B. Rosemary Grant. 2006. Evolution of character displacement in Darwin’s finches. Science 313: 224 226. An interesting example that describes rapid evolution in a classic system. Kingsolver, Joel G., Hopi E. Hoekstra, Jon M. Hoekstra, David Berrigan, Sacha N. Vignieri, Chris H. Hill, Anhthu Hoang, Patricia Gilbert, and Peter Beerli. 2001. The strength of phenotypic selection in natural populations. American Naturalist 157: 245 261. Reviews numerous studies of selection in natural populations. Kingsolver, Joel G., and David W. Pfennig. 2007. Patterns and power of phenotypic selection in nature. BioScience 57: 561 572. An overview of how phenotypic selection acts in natural populations. Portions of this chapter (es pecially parts of sections 3, 4, and 6) are adapted from this review. Losos, Jonathan B., Thomas W. Schoener, R. Brian Langer hans, and David A. Spiller. 2006. Rapid temporal reversal in predator driven natural selection. Science 314: 1111. Illustrates how directional selection can reverse direction rapidly. Pfennig, David W., Amber M. Rice, and Ryan A. Martin. 2007. Field and experimental evidence for competition’s role in phenotypic divergence. Evolution 61: 257 271. Describes how different modes of selection may act on the same species when confronted with different environ mental circ*mstances.

I.15 Population Genetics and Ecology Philip Hedrick OUTLINE

1. 2. 3. 4. 5. 6.

Introduction Genetic drift and effective population size Neutral theory Gene flow and population structure Selection Future directions

About 40 years ago, scientists first strongly advocated the integration of population ecology and population genetics into population biology (Singh and Uyenoyama, 2004). Even today these two disciplines are not really integrated, but there is a general appreciation of population genetic concepts in population ecology and vice versa. For example, the new subdiscipline molecular ecology, and many articles in the journal Molecular Ecology, use genetic markers and principles to examine both ecological and evolutionary questions. Although some aspects of population genetics have changed quickly in recent years, many of its fundamentals are still important for aspects of ecological study.

GLOSSARY coalescence. The point at which common ancestry for

two alleles at a gene occurs in the past. effective population size. An ideal population that in-

corporates such factors as variation in the sex ratio of breeding individuals, the offspring number per individual, and numbers of breeding individuals in different generations. gene flow. Movement between groups that results in genetic exchange. genetic bottleneck. A period during which only a few individuals survive and become the only ancestors of the future generations of the population. genetic drift. Chance changes in allele frequencies that result from small population size. Hardy-Weinberg principle. After one generation of random mating, single-locus genotype frequencies

can be represented as a binomial function of the allele frequencies. neutral theory. Genetic change is primarily the result of mutation and genetic drift, and different molecular genotypes are neutral with respect to each other. population. A group of interbreeding individuals that exist together in time and space. selective sweep. Favorable directional selection that results in a region of low genetic variation closely linked to the selected region. 1. INTRODUCTION

The primary goals of population genetics are to understand the factors determining evolutionary change and stasis and the amount and pattern of genetic variation within and between populations (Hedrick, 2005; Hartl and Clark, 2007). In the 1920s and 1930s, shortly after widespread acceptance of Mendelian genetics, the theoretical basis of population genetics was developed by Ronald A. Fisher, J.B.S. Haldane, and Sewall Wright. Population genetics may be unique among biological sciences because it was first developed as a theoretical discipline by these men before experimental research had a significant impact. The advent of molecular genetic data of populations in the late 1960s and DNA sequence data in the 1980s revolutionized population genetics and produced many new questions. Population genetics and its evolutionary interpretations provided a fundamental context in which to interpret these new molecular genetic data. Further, population genetic approaches have made fundamental contributions to understanding the role of molecular variation in adaptive differences in morphology, behavior, and physiology. A primary goal in determining the extent and pattern of genetic variation is to document the variation that results in selective differences among individuals, the ‘‘stuff of evolution.’’ The amount and kind of genetic variation in populations are potentially affected by a number of factors, but primarily by selection, inbreeding, genetic drift,



gene flow, mutation, and recombination. These factors may have general or particular effects; for example, genetic drift and inbreeding can be considered to always reduce the amount of variation, and mutation to always increase the amount of variation. Other factors, such as selection and gene flow, may either increase or reduce genetic variation, depending on the particular situation. Combinations of two or more of these factors can generate many different levels and patterns of genetic variation. In 1968, Motoo Kimura introduced the important ‘‘neutral theory’’ of molecular evolution that assumes that genetic variation results from a combination of mutation generating variation and genetic drift eliminating it (Kimura, 1983). This theory is called neutral because allele and genotype differences at a gene are selectively neutral with respect to each other. This theory is consistent with many observations of molecular genetic variation (see below). To understand the influence of these evolutionary factors, one must first be able to describe and quantify the amount of genetic variation in a population and the pattern of genetic variation among populations. In recent years, new laboratory techniques have made it possible to obtain molecular genetic data in any species, and a number of software packages have become available to estimate the important parameters in population genetics and related topics. In addition, the online Evolution Directory (EvolDir) is a source of information about different molecular techniques, estimation procedures, and other current evolutionary genetic information. Let us first define the evolutionary or genetic connotation of the term population. As a simple ideal, a population is group of interbreeding individuals that exist together in time and space. Often it is assumed that a population is geographically well defined, although this may not always be true. Below we discuss the concept of effective population size, which provides a more explicit definition of population in evolutionary terms. Many of the theoretical developments in population genetics assume a large, random-mating population that forms the gene pool from which the female and male gametes are drawn. In some real-life situations, such as dense populations of insects or outcrossing plants, this ideal may be nearly correct, but in many natural situations, it is not closely approximated. For example, there may not be random mating, as in selffertilizing plants, or there may be small or isolated populations as in rare or endangered species. In these cases, modifications of the theoretical ideal must be made. One of the basic concepts of population genetics is the Hardy-Weinberg principle (often called Hardy-

Weinberg equilibrium [HWE]). It states that after one generation of random mating, single-locus genotype frequencies can be represented by a binomial (with two alleles) or multinomial (with multiple alleles) function of the allele frequencies. This principle allows great simplification of the description of a population’s genetic content by reducing the number of parameters that must be considered. Furthermore, in the absence of factors that change allele frequency (selection, genetic drift, gene flow, and mutation), and in the continued presence of random mating, the Hardy-Weinberg genotype proportions will not change over time. 2. GENETIC DRIFT AND EFFECTIVE POPULATION SIZE

Since the beginning of population genetics, there has been controversy concerning the importance of chance changes in allele frequencies because of small population size, termed genetic drift. Part of this controversy has resulted from the large numbers of individuals observed in many natural populations, large enough to think that chance effects would be small in comparison to the effects of other factors, such as selection and gene flow. However, if the selective effects or amount of gene flow are small relative to genetic drift, then long-term genetic change caused by genetic drift may be important. Under certain conditions, a finite population may be so small that genetic drift is significant even for loci with sizable selective effects, or when there is gene flow. For example, some populations may be continuously small for relatively long periods of time because of limited resources in the populated area, low tendency or capacity to disperse between suitable habitats, or territoriality among individuals. In addition, some populations may have intermittent small population sizes. Examples of such episodes are the overwintering loss of population numbers in many invertebrates and epidemics that periodically decimate populations of both plants and animals. Such population fluctuations generate genetic bottlenecks, or periods during which only a few individuals survive and become the only ancestors of the future generations of the population. Small population size is also important when a population grows from a few founder individuals, a phenomenon termed founder effect. For example, many island populations appear to have started from a very small number of individuals. If a single female who was fertilized by a single male founds a population, then only four genomes (assuming a diploid organism), two from the female and two from the male, may start a new population. In plants, a whole population may be initiated from a single seed—only two genomes, if

Population Genetics and Ecology self-fertilization occurs. As a result, populations descended from a small founder group may have low genetic variation or by chance have a high or low frequency of particular alleles. Another situation in which small population size is of great significance occurs when the population (or species) in question is one of the many threatened or endangered species (Allendorf and Luikart, 2007). For example, all approximately 500 whooping cranes alive today descend from only 20 whooping cranes that were alive in 1920 because they were hunted and their habitat destroyed. All 200,000 northern elephant seals alive today descend from as few as 20 that survived nineteenth-century hunting on Isla Guadalupe, Mexico. Further, all the living individuals of some species are descended from a few founders that were brought into captivity to establish a protected population, such as Przewalski’s horses (13 founders), California condors (13 founders), black-footed ferrets (6 founders), Gala´pagos tortoises from Espan˜ola Island (15 founders), and Mexican wolves (7 founders). The population size that is relevant for evolutionary matters, the number of breeding individuals, may be much less than the total number of individuals in an area, the census population size, and is the appropriate measure for many population ecology studies. The size of the breeding population may sometimes be estimated with reasonable accuracy by counting indicators of breeding activity such as nests, egg masses, and colonies in animals or counting the number of flowering individuals in plants. But even the breeding population number may not be indicative of the population size that is appropriate for evolutionary considerations. For example, factors such as variation in the sex ratio of breeding individuals, the offspring number per individual, and numbers of breeding individuals in different generations may be evolutionarily important. All these factors can influence the genetic contribution to the next generation, and a general estimate of the breeding population size does not necessarily take them into account. As a result, the effective population size, or Ne, a theoretical concept that incorporates variation in these factors and allows general predictions or statements irrespective of the particular forces responsible, is quite useful. In other words, the concept of an ideal population with a given effective size enables us to draw inferences concerning the evolutionary effects of finite population size by providing a mechanism for incorporating factors that result in deviations from the ideal. The concept of the effective population size makes it possible to consider an ideal population of size N in which all parents have an equal expectation of being


the parents of any progeny individual. In other words, the gametes are drawn randomly from all breeding individuals, and the probability of each adult producing a particular gamete equals 1/N, where N is the number of breeding individuals. A straightforward approach that is often used to tell the impact of various factors on the effective population size is the ratio of the effective population size to breeding (or sometimes census) population size, that is, Ne /N. Sometimes, this ratio is only around 0.1 to 0.25, indicating that the effective population size may be much less than the number of breeding individuals. In general, the effect of genetic drift is a function of the reciprocal of the number of gametes in a population, 1/(2Ne), for a diploid population. If Ne is large, then this value is small, and there is little genetic drift influence. Or, if Ne is small, then this value is relatively large, and genetic drift may be important. 3. NEUTRAL THEORY

Neutral theory assumes that selection plays a minor role in determining the maintenance of molecular variants and proposes that different molecular genotypes have almost identical relative fitnesses; that is, they are neutral with respect to each other. The actual definition of selective neutrality depends on whether changes in allele frequency are primarily determined by genetic drift. In a simple example, if s is the selective difference between two alleles at a locus, and if s < 1/(2Ne), the alleles are said to be neutral with respect to each other because the impact of genetic drift is larger than selection. This definition implies that alleles may be effectively neutral in a small population but not in a large population. Neutral theory does not claim that the relatively few allele substitutions responsible for evolutionarily adaptive traits are neutral, but it does suggest that the majority of allele substitutions have no selective advantage over those that they replace. Kimura also showed that the neutral theory was consistent with a molecular clock; that is, there is a constant rate of substitution over time for molecular variants. To illustrate the mathematical basis of the molecular clock, let us assume that mutation and genetic drift are the determinants of changes in frequencies of molecular variants. Let the mutation rate to a new allele be u so that in a population of size 2N there are 2Nu new mutants per generation. It can be shown that the probability of chance fixation of a new neutral mutant is 1/(2N) (the initial frequency of the new mutant). Therefore, the rate of allele substitution k is the product of the number of new mutants per generation and their probability of fixation, or



1 k ¼ 2Nu ¼ u: 2N In other words, this elegant prediction from the neutral theory is that the rate of substitution is equal to the mutation rate at the locus and is constant over time. Note that substitution rate is independent of the effective population size, a fact that may initially be counterintuitive. This independence occurs because in a smaller population there are fewer mutants; that is, 2Nu is smaller, but the initial frequency of these mutants is higher, which increases the probability of fixation, 1/2N, by the same magnitude by which the number of mutants is reduced. This simple, elegant mathematical prediction and others from the neutral theory provide the basis for the most important developments in evolutionary biology in the past halfcentury. One of the appealing aspects of the neutral theory is that, if it is used as a null hypothesis, then predictions about the magnitude and pattern of genetic variation are possible. Initially, molecular genetic variation was found to be consistent with that predicted from neutrality theory. In recent years, examination of neutral theory predictions in DNA sequences has allowed tests of the cumulative effect of many generations of selection, and a number of examples of selection on molecular variants have been documented (see below). Traditionally, population genetics examines the impact of various evolutionary factors on the amount and pattern of genetic variation in a population and how these factors influence the future potential for evolutionary change. Generally, evolution is conceived of as a forward process, examining and predicting the future characteristics of a population. However, rapid accumulation of DNA sequence data over the past two decades has changed the orientation of much of population genetics from a prospective one investigating the factors involved in observed evolutionary change to a retrospective one inferring evolutionary events that have occurred in the past. That is, understanding the evolutionary causes that have influenced the DNA sequence variation in a sample of individuals, such as the demographic and mutational history of the ancestors of the sample, has become the focus of much population genetics research. In a determination of DNA variation in a population, a sample of alleles is examined. Each of these alleles may have a different history, ranging from descending from the same ancestral allele, that is, identical by descent, in the previous generation to descending from the same ancestral allele many generations before. The point at which this common an-

cestry for two alleles occurs is called coalescence. If one goes back far enough in time in the population, then all alleles in the sample will coalesce into a single common ancestral allele. Research using the coalescent approach is the most dynamic area of theoretical population genetics because it is widely used to analyze DNA sequence data in populations and species. 4. GENE FLOW AND POPULATION STRUCTURE

In most species, populations are often subdivided into smaller units because of geographic, ecological, or behavioral factors. For example, the populations of fish in pools, trees on mountains, and insects on host plants are subdivided because suitable habitat for these species is not continuous. Population subdivision can also result from behavioral factors, such as troop formation in primates, territoriality in birds, and colony formation in social insects. When a population is subdivided, the amounts of genetic connectedness among the parts of the population can differ. Genetic connection depends primarily on the amount of gene flow, movement between groups that results in genetic exchange that takes place among the subpopulations or subgroups. When the amount of gene flow between groups is high, gene flow has the effect of hom*ogenizing genetic variation over the groups. When gene flow is low, genetic drift, selection, and even mutation in the separate groups may lead to genetic differentiation. It is sometimes useful to describe the population structure in a particular geographic framework. For example, within a watershed, there may be separated fish or plant groups that have a substantial amount of genetic exchange between them. On a larger scale, there may be genetic exchange between adjacent watersheds but in smaller amounts than between the groups within a watershed. On an even larger scale, there may be populations in quite separated watersheds that presumably have little direct exchange but may share some genetic history, depending on the amount of gene flow among the adjacent groups or occasional long-distance gene flow. This hierarchical representation is useful in describing the overall relationships of populations of an organism and in documenting the spatial pattern of genetic variation. Recently, there has been increasing interest in landscape and geographic approaches to estimating historical and contemporary gene flow. In addition, phylogeography, the joint use of phylogenetic techniques and geographic distributions, has been used to understand spatial relationships and distributions of populations within species or closely related species (Avise, 2000).

Population Genetics and Ecology In general, the subdivision of populations assumes that the various subpopulations are always present. Another view assumes that individual population subdivisions at particular sites may become extinct and then later be recolonized from other subpopulations, resulting in a metapopulation. The dynamics of extinction and recolonization can make metapopulations quite different both ecologically and genetically from the traditional concept of a subdivided population. Gene flow is central to understanding evolutionary potential and mechanisms in several areas of applied population genetics. First, the potential for movement of genes from genetically modified organisms (GMO) into related wild populations—that is, the gene flow of transgenes into natural populations—can be examined using population genetics. Second, the invasive potential of nonnative plants and animals into new areas may be affected by hybridization (gene flow) between nonnative and native organisms as well as adaptive change. Finally, a number of endangered species are composed of only one, or a few, remaining populations with low fitness. Gene flow from other populations of the same species can result in genetic rescue or genetic restoration of these populations by introducing new variation that allows removal of detrimental variation and restoration of adaptive change. Estimating the amount of gene flow in most situations is rather difficult. Direct estimates of the amount of movement can be obtained in organisms where different individuals can be identified or individual marks are used. Many approaches have been employed to mark individuals differentially, such as toe clipping in rodents, leg banding in birds, coded-wire tags in fish, and radio transponders in many different vertebrates. However, both movement of individuals and their incorporation into the breeding population are necessary for gene flow. Using highly variable genetic markers, it is now possible to identify parents genetically and thereby determine the spatial movement of gametes between generations without direct movement information of the parents. Or, individuals can be assigned to specific populations using genetic markers, thereby determining whether they are migrants or not. Indirect measures of gene flow using genetic markers are useful to confirm behavioral or other observations or when these observations are inconclusive or impossible. Most commonly, the number of successfully breeding migrants between groups is measured using techniques to evaluate population structure. Theoretically, assuming finite subpopulations of size Ne and a proportion m migrants into each subpopulation each generation, then



1 : 4Ne m þ 1

When Nem is large, the measure FST approaches 0, and when Nem is small, FST can approach 1. The value of FST for a group of populations can be estimated using the amount and pattern of molecular genetic variation over subpopulations. The availability of molecular and DNA sequence data in many organisms provides a database to determine the relationships between populations or species that are not obvious from other traits. It is generally assumed that molecular genetic data better reflect the true relationships between groups than other data, such as morphology or behavior, because molecular data are less influenced by selective effects. Furthermore, differences between relationships generated from molecular data and from other traits provide an opportunity to evaluate the effect of selective effects on other traits. Maternally inherited mitochrondrial DNA (mtDNA) data have been the workhorse for phylogeographic research because mtDNA does not recombine in most organisms and, as a result, shows a clearer phylogenetic record than many nuclear genes (chloroplast DNA and Y chromosomes are similarly useful). In addition, the effective population size for mtDNA (as well as for chloroplast DNA and Y chromosomes) is only about one-fourth that of nuclear genes so that divergence occurs about four times faster than for nuclear genes. However, this faster rate of divergence and potentially differential gene flow for the two sexes may cause the signal for these uniparentally inherited genes to be different from the phylogenetic pattern for nuclear genes, which constitute a very large proportion of the genome.


In the past few years, with the availability of extensive DNA sequence data for a number of organisms (particularly humans), there has been an intensive search for genomic regions exhibiting a signal of adaptive (positive Darwinian) selection. Many of the genomic regions identified have undergone a ‘‘selective sweep’’ because of favorable directional selection, as indicated by low genetic variation in genetic regions closely linked to the selected gene or regions. An elegant example of a selective sweep is adaptive melanism in the rock pocket mouse of the southwestern United States (Hoekstra et al., 2004). The mouse is generally light-colored and lives on light-colored granite rocks, but it also has melanic forms that live on relatively recent black lava formations in several


Autecology Arizona

Arizona, USA




Tule West

Mid East O’Neill

Sonora, Mexico

Substrate color: Coat color: Xmas






Figure 1. Six sampling sites (three on dark volcanic rock and three on light colored substrate) and coat color frequencies (in pie dia

grams) in rock pocket mice across a transect in the Sonoran desert. (From Hoekstra et al., 2004)

restricted sites. Figure 1 shows the frequencies of the normal recessive and dominant melanic forms from a 35-km transect in southwestern Arizona. Here the frequency of melanics is highly concordant with substrate color, that is, high frequencies of melanics in the center of this transect that has approximately 10 km of black lava and lower frequencies of melanics on the light-colored substrate sites at either end of the transect. Investigation of molecular variation in the Mc1r gene, which is known to have variants that produce dark-colored house mice, was found to correlate nearly completely with the light and melanic phenotypes. The melanic and normal alleles were found to differ by four amino acids, and the nucleotide diversity for the melanic alleles was 1/20 that for light alleles. The lower variation among the melanic alleles is consistent with

the expected pattern if selection has recently increased its frequency by a selective sweep in the area of black lava. Some of best-documented examples of adaptive selection are those resulting from recent human changes in the environment (Hedrick, 2006), such as development of genetic resistance in insect pests to chemicals used to control them or genetic resistance in pathogens to antibiotics. The genetic basis of pesticide resistance may be the result of many genes, mutants at a single or a few genes, or expansion of gene families. Because the molecular basis of many of these adaptive changes is known, detailed genetic and evolutionary understanding is possible. For example, resistance to some insecticides among mosquitoes that are vectors for diseases such as malaria (Anopheles gambiae) and West Nile virus (Culex pipiens) is the result of a single amino-acid

Population Genetics and Ecology


1.0 0.9

Expected heterozygosity

0.8 0.7 0.6 0.5 0.4 0.3


0.2 0.1 -646 -623 -537 -460 -363 -252 -141 -89 -58 -30 -17 -7.6 -4.5 -4.2 -3.9 -1.2 -0.1 0.2 0.52 1.48 4.05 5.87 30.3 39.9 49.2 121 143 159 181 239 300 323 359

0 Marker distance (kilobases) from dhfr Figure 2. The observed (solid circles) and expected (lines) het erozygosity around the dhfr gene in the malaria parasite Plasmo

dium falciparum, which provides resistance to the antimalarial drug pyrimethamine. (From Nair et al., 2003)

substitution. In C. pipiens, a single nucleotide change, GGC (glycine) to AGC (serine) at codon 119 in the gene for the enzyme acetylcholinesterase (ace-1) results in insensitivity to organophosphates. A complete lack of variation within samples among resistant variants of this gene suggests that they have originated and spread quite recently. Another set of examples of adaptive selection include those resulting from the development of host resistance to pathogens. For example, malaria kills more than one million children each year in Africa alone and is the strongest selective pressure in recent human history. As a result, selective protection from malaria by sickle cell, thalassemia, G6PD, Duffy, and many other genetic variants in the human host provide some of the clearest examples of adaptive variation and diversifying selection for pathogen resistance (Kwiatkowski, 2005). Genomic studies have demonstrated that selection for malarial resistance is strong, up to 10%, and that variants conferring resistance to malaria are recent, generally less than 5000 years old, consistent with the proposed timing of malaria as an important human disease. Often the resistant variants are in different populations, probably mainly in part because of their recent independent mutation origin. Efforts to control the malaria parasite using antimalarial drugs have resulted in widespread genetic resistance to these drugs. For example, pyrimethamine is an inexpensive antimalarial drug used in countries

where there is resistance in the malarial parasite to the widely used drug chloroquine. Pyrimethamine was introduced to the area along the Thailand-Myanmar border in the mid-1970s, and resistance spread to fixation in approximately 6 years. Resistance is the result of point mutations at the active site of the enzyme encoded by the gene dhfr on chromosome 4. Examination of genetic variation at genes near dhfr as shown in figure 2 showed remarkable reduced heterozygosity near dhfr and normal variation further away (Nair et al., 2003). This pattern of variation is consistent with a selective sweep, and the theoretical pattern expected from a selective sweep is shown by the curve in figure 2. The major histocompatibility complex (MHC) genes are part of the immune system in vertebrates, and differential selection through resistance to pathogens is widely thought to be the basis of their high genetic variation (Garrigan and Hedrick, 2003). Variation in the genes of the human MHC, known as HLA genes, has been the subject of intensive study for many years because of their role in determining acceptance or rejection of transplanted organs, many autoimmune diseases, and recognition of pathogens. High HLA variation allows recognition of more pathogens, consistent with the fact that HLA-B is the most variable gene in the human genome. In recent years, there has been extensive research examining R (disease resistant) genes in plants, a system with similarities to MHC.




Because of the widespread availability of DNA sequence data in many organisms, the future application of population genetic data and principles in ecology appears almost unlimited. Many basic ecological questions, such as how many individuals there are in a population, what their relationships are, or whether they are immigrants, may be definitively answered in future years using genetic techniques. Such precise descriptions may then provide data to understand the evolutionary and ecological factors influencing population dynamics and distributions. FURTHER READING Allendorf, Fred, and Gordon Luikart. 2007. Conservation and the Genetics of Populations. Oxford: Blackwell Pub lishing. A recent summary of the application of population genetics to conservation. Avise, John. 2000. Phylogeography: The History and For mation of Species. Cambridge, MA: Harvard University Press. The joint use of phylogenetic relationships and geographic patterns to understand evolution. Garrigan, Daniel, and Philip Hedrick. 2003. Perspective: Detecting adaptive molecular evolution, lessons from the MHC. Evolution 57: 1707 1722. A perspective on the strongest example of balancing selection, MHC, and the gain and loss of balancing selection signals. Hartl, Daniel, and Andrew Clark. 2007. Principles of Popu lation Genetics, 4th ed. Sunderland, MA: Sinauer As sociates. A recent summary of the principles of population genetics. Hedrick, Philip. 2005. Genetics of Populations, 3rd ed. Bos ton: Jones and Bartlett Publishers. A recent and thorough summary of the principles of population genetics.

Hedrick, Philip. 2006. Genetic polymorphism in heteroge neous environments: The age of genomics. Annual Review of Ecology, Evolution, and Systematics 37: 67 93. A summary of the recent empirical and theoretical exam ples of genetic polymorphism maintained by ecological variation. Hoekstra, Hopi, Kristen Drumm, and Michael Nachman. 2004. Ecological genetics of adaptive color polymorphism in pocket mice: Geographic variation in selected and neutral genes. Evolution 58: 1329 1341. A discussion of adaptive melanism in pocket mice living on dark lava re sults from amino acid changes in the Mc1r gene. Kimura, Motoo. 1983. The Neutral Theory of Molecular Evolution. Cambridge: Cambridge University Press. A summary of the neutral theory from the view of its major architect, Motoo Kimura. Kwiatkowski, Dominic. 2005. How malaria has affected the human genome and what human genetics can teach us about malaria. American Journal of Human Genetics 77: 171 192. A review of the many human genes that confer resistance to malaria, the strongest selective factor in re cent human history. Nair, Shalini, Jeff Williams, Alan Brockman, Lucy Paiphun, Mayfong Mayxay, P. N. Newton, J. P. Guthmann, F. M. Smithuis, T. T. Hien, N. J. White, F. Nosten, and T. J. Anderson. 2003. A selective sweep driven by pyrimeth amine treatment in southeast Asian malaria parasites. Molecular Biology and Evolution 20: 1526 1536. An example of the fast development of antimalarial drug re sistance by a selective sweep of new resistant variants in Plasmodium. Singh, Rama, and Marcy Uyenoyama, eds. 2004. The Evo lution of Population Biology Modern Synthesis. Cam bridge: Cambridge University Press. A summary of the con tributions to population biology over recent decades.

I.16 Phylogenetics and Comparative Methods David D. Ackerly OUTLINE

1. 2. 3. 4. 5.

The role of phylogenetics in ecology Phylogenies and the analysis of trait correlations Phylogenetic signal: Pattern and significance Phylogenetics and community ecology Prospects for the future

The study of ecology frequently draws on comparative observations and experiments that rely on the similarities and differences among species and the correlations among species traits and the environment. In such studies, consideration of the phylogenetic relationships among species provides valuable information for statistical inference and an understanding of evolutionary history underlying presentday ecological patterns. From a statistical perspective, related species do not necessarily provide independent data points for hypothesis tests, due to inheritance of shared characteristics from common ancestors. This similarity can be addressed through a variety of statistical techniques, including the widely used method of phylogenetic independent contrasts. Independent contrasts play a particularly valuable role in the analysis of trait and trait–environment correlations and may point toward alternative interpretations of comparative data. In community ecology, measures of the phylogenetic clustering or spacing of co-occurring species provide a useful tool to test alternative processes underlying community assembly. Co-occurrence of close relatives most likely reflects ecological filtering, in which related species with similar traits share the ability to tolerate local conditions. The reverse pattern of phylogenetic spacing of cooccurring species may reflect a variety of processes, and additional observations of species traits in relation to environment and interacting taxa will be necessary to address underlying processes. Use of comparative methods has increased dramatically with the rapid growth in phylogenetic information and computing power and will continue to play an important role in ecological research.


See figure 1 for illustrations of main terms. branch lengths. These may indicate either the number of inferred character changes or a measure of relative or absolute time along any particular branch connecting two nodes. If the molecular data underlying a phylogeny do not violate a molecular clock, a single rate may be imposed such that branch lengths will represent relative time, and contemporaneous taxa will be placed at the same distance from the root (i.e., the same age). If a molecular clock is violated, rate-smoothing methods have been developed to obtain the best-supported estimate of relative time. Fossils and biogeographic or paleoecological information may then be used to calibrate these branch lengths and convert them to units of absolute time. Rate-smoothing and calibration methods are fraught with difficulty, and branch lengths should be treated with caution. (Note that branch lengths may also be set arbitrarily for convenience when one is drawing trees, in which case they have no intrinsic biological meaning.) character states. Phylogenetic trees are reconstructed based on analysis of a matrix of characters, where each character can take on one of two or more states (binary or multistate, respectively) for each taxon in the group. Phylogenies can be reconstructed from molecular and/or morphological data, although the former are now much more common. Analyses that include morphological data are advantageous as they make it possible to incorporate taxa or fossils for which molecular data are not available. lineage. This refers to a single line of ancestor–descendant relationship, connecting nodes within a phylogeny. most recent common ancestor (MRCA). The MRCA is the most recent node that is shared by any two taxa in a tree.


Autecology branch node


polytomy root or basal node

MRCA of F and G


terminal taxa (’tips’)


clade containing species E, F, G, and H Figure 1. Example of a phylogenetic tree for eight taxa (A H), illus trating some of the terms in the glossary. This tree is ultrametric, meaning that all terminal taxa are equidistant from the root of the tree.

phylogenetic distance. The phylogenetic distance be-

tween two nodes or taxa refers to the sum of branch lengths from one tip (or internal node) down to the MRCA and back up to another tip (or node) of a tree. The phylogenetic distance matrix is an nn matrix (for n taxa) of such distances among all pairs of taxa, with 0s in the diagonal. phylogeny. A phylogeny, or phylogenetic tree, is a branching diagram showing the hierarchy of evolutionary relationships among a group of taxa (extant and/or extinct). Terminal taxa or tips are connected by branches to internal nodes that indicate a hypothesized ancestor. A clade includes all of the taxa (extant and extinct) that descend from a node. Phylogenies can be either rooted or unrooted, where the root represents the hypothesized ancestor of all taxa on the tree. polytomy. This refers to a node with three or more daughter nodes. A soft polytomy indicates uncertainty, where the true bifurcating relationships among the daughters are unknown. A hard polytomy represents a hypothesis of near simultaneous divergence where the sequence of individual speciation events cannot be meaningfully resolved. Most phylogenetic comparative methods treat polytomies as either hard or soft but do not always make the distinction explicit. ultrametric. An ultrametric tree is one in which all terminal taxa are contemporaneous; more precisely, the sum of the branch lengths from the root to each tip is the same for all tips. Phylogenies of extant taxa will be ultrametric if branch lengths have been adjusted to represent relative or absolute time.

In Great Britain there are 32 indigenous trees[:] of these 19 or more than half . . . have their sexes separated—an enormous proportion compared with the remainder of the British flora: nor is this wholly owing to a chance coincidence in some one Family having many trees & having a tendency to separated sexes: for the 32 trees belong to nine Families, & the trees with separate sexes to five Families. —Charles Darwin, manuscript for Natural Selection (unpublished) In the quote above, Darwin observes an interesting pattern among plant species of Great Britain. He notes that among trees, the proportion of species that have individuals of separate sexes (as in humans and most vertebrates) is much higher than among the flora as a whole, most of which is composed of shrubs and herbaceous plants. He explained the high frequency of separate sexes as an adaptation to promote crossfertilization in trees: because trees are large and have many flowers, the chance that an insect would carry pollen from one flower to another of the same individual is quite high. If all the flowers on a tree are of the same sex, these repeated visits by pollinators will not lead to high levels of self-fertilization. Darwin’s observations provide a nice example of what we now call comparative biology, which draws on comparisons of the similarities and differences among species to test ecological and evolutionary hypotheses. In addition, what Darwin recognized intuitively is that a simple count of the number of species exhibiting different characteristics might not be adequate to support his argument. If many of the species are drawn from the same family (that is, closely related in evolutionary terms), they are likely to share many ecological characteristics. Thus, a group with many tree species may also contain many species with separate sexes, reflecting their descent from a common ancestor. But if the evolutionary argument is sound— that trees should evolve separate sexes because of the problem of self-fertilization—then this combination of traits should evolve independently in many different taxonomic groups, and this is indeed what Darwin observed. Throughout the past 150 years, since the publication of Darwin’s On the Origin of Species, comparative biology has played a central role in ecology and evolutionary biology. In essence, each species alive today (or in the past) represents the outcome of a long, natural experiment. The results reflect the contemporary ecology of a species—interactions with the abiotic environment and with other forms of life—as well as the cumulative legacy of the past. Evolution works slowly,

Phylogenetics and most features are passed down from ancestor to descendant with little change. A penguin appears beautifully adapted to the challenges of surviving and reproducing under the extreme conditions of Antarctic life. But these adaptations must be understood in historical context: penguins are birds, and this experiment in polar living started with very specific initial conditions, including egg-laying, a feathered pelt, forelimbs modified into wings, and so on. Comparative research, placing penguins in the broader context of other birds and viewing them side by side with their closest relatives (loons, albatrosses, petrels, and shearwaters) is critical to an understanding and appreciation of their contemporary ecology and behavior. In the past 30 years, comparative biology has grown rapidly as a new generation of methods emerged, combining the historical perspective outlined above with the quantitative tools of experimental statistics. The emergence of modern phylogenetics triggered these developments. The word phylogeny refers to the evolutionary relationships among a group of organisms, illustrated as a branching tree where the tips (or leaves) may represent individuals, populations, species, or groups of species, and the internal branching points are their common ancestors. The study of phylogenetics has been revolutionized by the combination of molecular biology (providing a trove of data), conceptual advances (the theory of cladistics), and the availability of high-speed computers. Together, these advances have made it possible to infer highly resolved phylogenies for many groups of organisms. With continuous improvements in methods and the availability of data, the tree of life is taking shape and revealing the hierarchy of evolutionary relationships among living (and extinct) organisms. 1. THE ROLE OF PHYLOGENETICS IN ECOLOGY

The science of ecology studies the interactions of organisms with their environment and the consequences of these interactions for where species live and how they interact. To address these questions, it is often useful to compare different species, either through observations or experiments. The similarities and differences in how species respond to their environment or interact with each other can provide important ecological insights. When data are gathered on different species, understanding how they are related to each other (i.e., their phylogenetic relationships) contributes valuable information that can affect data analysis and interpretation. In this chapter, I focus on two areas of ecological research where phylogenies play a particularly important role: the analysis of correlations among species traits and environmental conditions (like Dar-


win’s example above) and the study of community ecology. In addition, I provide a brief discussion of the concept of phylogenetic signal, a general term for the similarity among close relatives. In the discussion below, it is assumed that a phylogeny is available for each group under consideration. Most phylogenies are based on molecular data, particularly DNA sequences, sometimes combined with morphological or other characteristics. The computational methods used to search for the best-supported phylogeny are continually being improved and are beyond the scope of this chapter. Regardless of the method used, it is important to recognize that every phylogeny is a hypothesis of relationships, and like any scientific hypothesis, it is subject to revision and improvement. Phylogenies may also contain different degrees of uncertainty, both in terms of the topology (the pattern of who is related to whom) and the lengths of the branches, which represent the amount of evolutionary change or the amount of time elapsed between different nodes of the tree. This uncertainty can be incorporated into comparative analyses; in many cases, the results are quite robust across a range of possible alternatives, so strongly supported and fully resolved phylogenies are not a prerequisite for comparative analysis. An overview of some terminology used to describe phylogenies is provided in the Glossary. 2. PHYLOGENIES AND THE ANALYSIS OF TRAIT CORRELATIONS

Research in functional ecology, life history strategies, and related areas of ecology often addresses questions of interspecific trait–trait and trait–environment associations, such as: Do mammals with larger body sizes have larger home ranges? Do plant species of open habitats tend to have smaller seeds? How are the traits of invasive species different from those of native species in a community? The answers to these questions help us to understand how species traits influence distribution, abundance, and interactions with other species in a community. They also have important applications in conservation biology, restoration ecology, and the management of invasive species. A variety of statistical methods can be applied to test hypotheses of trait associations, depending on the type of data available and the nature of the hypothesis. These include correlation, regression, analysis of variance, contingency table analysis, and others. One of the basic assumptions of virtually all statistical tests is that each data point represents an observation that is independent with respect to the underlying null hypothesis. This assumption is not required in order to calculate the various statistics; rather, it is essential to



deriving the statistical significance of the outcome. For conventional statistics, this significance value (or pvalue) represents the probability of observing the data if the underlying null hypothesis is true. When that probability is too low (conventionally, we use a cutoff value of 5%), we reject the null hypothesis and accept that there is a significant effect or relationship. For maximum-likelihood tests, which are playing an increasingly important role in ecology and comparative methods, the assumption of independence is used to assess the likelihood of the best-fit model relative to alternative models or hypotheses, given the observed data. The fundamental argument underlying the development of many comparative methods arises from the observation that related species are ecologically and phenotypically similar to one another. This will not hold for every trait, as instances of rapid divergence and of convergent evolution are widespread and important. But on average, species resemble their close relatives more than they do more distant taxa, and this similarity reflects descent from recent common ancestors. Because of this inherited similarity, it is argued that in statistical terms species do not represent independent data points, violating this basic assumption of significance testing. One can also approach this problem in terms of the underlying historical processes. Trait associations among extant species arise through a historical sequence of correlated changes occurring along each branch of the phylogeny; ideally, we would like to estimate the correlation between these changes to more directly measure evolutionary linkages between the traits. It is now well established that the correlations observed among living species (at the tips of the phylogeny) do not provide a reliable estimate of this historical pattern of correlated evolutionary changes that have occurred along the branches of the phylogeny. Although some researchers are strongly motivated by the statistical arguments, and others more by the historical questions, both perspectives lead one to the use of phylogenetic comparative methods. The selection of a comparative method to conduct associational analyses depends on the nature of the data and the hypothesis. One of the most common problems is the correlation (a measure of the strength of association) between two traits measured on a continuous scale (e.g., body size or seed size). Correlation coefficients range from 0 for two traits with no association up to 1 for traits that are very tightly linked (–1 if it is a negative association). In 1985, Joseph Felsenstein introduced the method of phylogenetic independent contrasts (often referred to as PICs) to address this question in a phylogenetic context; more than 20 years later, his method remains one of the most robust and

widely used of all comparative methods. The method of independent contrasts rests on the assumption that the evolutionary change in a trait that occurs along each lineage leading up to present-day species represents an independent event with respect to the changes occurring in other branches. Independence, in this context, refers to the statistical notion that the changes are independent manifestations of underlying processes, although the same processes (e.g., natural selection as a result of climate change) may be affecting multiple lineages in a group. If the trait changes that occur in two lineages arising from a common ancestor are independent, then, as Felsenstein demonstrated based on statistical theory, the difference between the trait values of the two descendants will also represent a statistically independent observation. These differences are calculated by subtracting the trait value of one species from the value of its closest relative, and they are referred to as PICs (there is an additional step involving the branch lengths on the phylogeny, which I do not describe here). In addition, Felsenstein showed that one can continue to calculate contrasts at deeper nodes of the tree, based on an iterative process of averaging the trait values at successively deeper nodes. In a fully resolved phylogeny, N species will be connected by N 1 common ancestors, so N trait values measured on the species will provide N 1 contrasts; these contrasts can be used as the variables in correlation, regression, and multivariate statistical analyses. A study that I conducted with Peter Reich in 1999 illustrates the application of independent contrasts and how they can impact the analysis of trait associations. We examined correlations among several functional attributes of leaves, including leaf size, leaf lifespan (the length of time a leaf persists on a plant), and specific leaf area (SLA, the ratio of leaf area to leaf dry mass; higher values indicate thinner or less dense leaves). Global studies of leaf function have found that leaves with higher SLA tend to have faster metabolic rates and shorter leaf lifespan, and this strategy is favored in more fertile habitats. The opposite set of traits is observed in leaves with low SLA. In addition, it is sometimes observed that leaves with low SLA and long leaf lifespan are smaller, and small leaves are often viewed as an adaptation to low-water or hightemperature environments. In particular, the needles of conifers (pines, spruces, etc.) are smaller in area and have a longer lifespan than the leaves of most flowering plants. In a data set of about 100 species, including both conifers and flowering plants, there are negative correlations of leaf lifespan with both SLA and leaf size. However, when we apply independent contrasts, the results change dramatically. The evolutionary correlation

Phylogenetics A.


Species data 3.0


Independent contrasts 0.4 Angiosperms Conifers Angiosperm/conifer contrast



0 2.0 -0.2 1.5

-0.4 R = -0.75



R = -0.64

-0.6 2





0.4 0.6 0.8 1.0 Leaf lifespan contrasts





D. 3

2.0 1.5 Leaf size contrasts

Leaf size (cm2, log)

2 1 0

1.0 0.5 0 -0.5

-1 -1.0 R = -0.42

-2 0

1 Leaf lifespan (mo., log)

R = 0.0

-1.5 2


Figure 2. Analysis of interspecific correlations among leaf traits, using independent contrasts. Panels A and B show the correlations of leaf lifespan with leaf size and specific leaf area, respectively. Black circles are data for flowering plant species; gray circles are for conifers. The strength of the associations is indicated by the correlation coefficients in the lower left corner of each panel. Pa nels C and D show the corresponding relationships analyzed with independent contrasts. Black circles are contrasts between nodes

within the flowering plant phylogeny, and gray circles are contrasts among conifers. The white circles represent the contrast at the basal node between the two groups. For convenience, the sub traction at each node is arranged such that the contrast for leaf lifespan is positive, and then the contrast for the other trait is positive or negative, depending on the trait values (subtraction must be in the same direction for both traits). (From Ackerly et al., 2000, Bioscience; copyright American Institute of Biological Sciences)

between SLA and leaf lifespan, based on contrasts, is similar to the pattern observed without using independent contrasts. But the evolutionary correlation between leaf lifespan and leaf size is essentially zero (figure 2). Why does this result shift so dramatically? As noted above, most of the correlation observed between leaf lifespan and size results from the marked difference between these traits in conifers and flowering plants, the deepest split in the phylogeny for this group of plants. Independent contrasts capture this shift as a single contrast. The rest of the contrasts, calculated among species of flowering plants or among species of conifers, exhibit no correlation in the shifts

occurring in these two traits. In essence, the pattern observed if each species is treated as an independent data point reflects the influence of a single event deep in the evolution of these groups; when this single event is represented as one data point in the analysis (based on the one contrast), its influence is diminished, and we see that there is not a consistent evolutionary tendency for correlated changes between these two traits. Other lines of evidence are consistent with this result: there is no evidence that leaf lifespan and leaf size are functionally or evolutionary linked to each other, so the result from independent contrasts proves more reliable. We are still left with an important pattern in the



present day: it is true that conifers have small, longlived needles, which differ on average from the leaves of flowering plants. These differences may be important to understanding the ecological differences between these two groups of plants, but they should not be taken as evidence of an ongoing functional and evolutionary linkage between these traits. As shown in this example, the method of independent contrasts addresses both the statistical and historical issues associated with the analysis of interspecific trait correlations. The contrasts are statistically independent, so significance values are reliable. The correlation or regression coefficients between the contrasts provide a much more precise measure of the underlying evolutionary pattern compared to a correlation of trait values from present-day species. However, like all statistical methods, independent contrasts invoke key assumptions, and these assumptions have been the source of some controversy. The most important assumption is that trait evolution conforms to a pattern of change known as a constant-variance random walk or Brownian motion. This model assumes that the changes occurring in each unit of time are equally likely to be positive or negative and are drawn from a normal distribution, such that small changes are more likely than large ones. Because these changes accumulate across multiple time steps, the total change along a branch is also expected to be proportional to the length of the branch. On the one hand, simulations have shown that statistical tests based on independent contrasts are quite robust to a variety of deviations from these basic assumptions, particularly if appropriate steps are taken to transform data or branch lengths in advance of analysis. In addition, the Brownian motion model is a reasonable first approximation of a model of evolutionary change based on our knowledge of quantitative genetics and the inheritance of continuous traits. On the other hand, a variety of other models of trait evolution may be considered; under some of these alternatives, species trait values are relatively independent of each other, and independent contrasts (or other comparative methods) do not necessarily provide a reliable measure of historical patterns. There are several other classes of comparative methods that can be used for questions of trait associations. One of the most important is known as the phylogenetic regression, introduced by Alan Grafen in 1989, or phylogenetic general linear models. These approaches utilize statistical methods in which the user can specify the degree of independence among observations. The phylogeny is used to generate what is known as a variance–covariance matrix, which captures the expected degree of dependence among each

pair of species in a study. This then opens up the full power of linear models, including multifactorial analysis of variance or covariance, with appropriate adjustment of significance tests reflecting the phylogeny. Although this facilitates a much broader range of hypothesis tests, one drawback is that the interpretation of results in terms of underlying historical processes is generally not as straightforward. A related class of methods uses maximum-likelihood approaches to find the best-fit model for a given set of interspecific trait data, given the phylogeny and alternative hypotheses of how the traits may be associated with each other. Maximum-likelihood approaches (and related Bayesian methods) have the general advantage that it is easier to invoke alternative underlying models of trait evolution. Further discussion of these methods, and the issues of branch lengths and evolutionary models, is beyond the scope of this chapter; researchers who will be using contrasts or other methods discussed here are well advised to seek a deeper understanding of these issues. It is important to note that discrete characters, such as presence/absence of a trait or different states of a morphological character, usually require different approaches. Traditional tests of association for discrete characters involve chi-square or G-tests, based on contingency tables showing the frequency of different pairs of states. Phylogenetic approaches can be used to reconstruct historical transitions from one state to the other and then to test for associations between these transitions or between transitions in one character and the background state of the other character. Maximum-likelihood models, such as the DISCRETE program introduced by Mark Pagel, provide powerful solutions to this problem by testing whether the probabilities of transitions in different characters are associated with each other (see box 1).

BOX 1. SOFTWARE FOR PHYLOGENETIC COMPARATIVE METHODS Phylogenetic comparative methods are computationally intensive, and a variety of software packages have been introduced that implement different tests. A few of the most important are briefly summarized here.

MacClade, first introduced by David and Wayne Maddison in 1987, set the standard for graphical elegance and ease of use in phylogenetic software. It is primarily used for reconstructing the evolution of discrete characters, based on parsimony methods, and also has limited capabilities for continuous characters.

Phylogenetics Mesquite, also developed by the Maddisons, is a cross-platform and open-source program (http:// with most of the features of MacClade plus a broader array of methods, including independent contrasts. R is a freely distributed program for statistical analysis and programming; individual users develop and contribute libraries that implement different methods ( Several libraries are now available (ape, ade4, geiger, PHYSIM, PHYLOGR) that implement numerous phylogenetic comparative methods. R is a very powerful program that is being adopted by many researchers in ecology (although it is difficult to learn at first). COMPARE, written by Emilia Martins, is a Web-based program that implements independent contrasts, phylogenetic linear models, and related methods ( Phylocom is a freely distributed program (http:// that is widely used for phylogenetic analysis of community structure and also conducts independent contrasts and analyses of phylogenetic signal. DISCRETE and Continuous, both written by Mark Pagel and colleagues, implement several maximum-likelihood methods for the analysis of trait correlations, modes of trait evolution, and related methods. Both of these programs are now included in the BayesTraits program (http://www


The fact that closely related species resemble each other —in ecological, morphological, behavioral, and other attributes—comes as no surprise to students of natural history. Evolution is generally a conservative process, and traits will usually change slowly, if at all, from one generation to the next. Adaptive radiations, in which species may diverge rapidly and take on novel adaptive traits and ecological lifestyles, are of interest precisely because they are unusual: at moments of ecological opportunity, following mass extinctions or the arrival of colonists on uninhabited islands, we see the potential for rapid evolutionary change. But most of the time, evolution is slow, and few changes accumulate, even over long periods of time. The lack of change is referred to as evolutionary stasis. The importance of understanding stasis in evolution has been highlighted by paleontologists, especially Steven Jay Gould, based on their study of the fossil record. When stasis, or at least a slow rate of change, plays out across the phylogeny, the result is that close relatives will be very similar.


Many terms have been used to describe this pattern of slow change: phylogenetic inertia, phylogenetic constraint, and phylogenetic effects. Often, these terms convey a sense that the phylogeny itself is the cause of ancestor–descendant resemblances. I find it useful to use the term phylogenetic signal, advocated in a recent essay by Simon Blomberg and Ted Garland, to emphasize that the similarity among relatives is a pattern and by itself does not reveal the underlying processes. An understanding of the causes of phylogenetic signal, and why it may vary in different groups and for different traits, draws on genetics, developmental biology, and ecology. We know that evolutionary change requires heritable, genetically based variation in a trait for selection to act on. Recent advances in the field of ‘‘evo-devo’’ are shedding light on how the process of development can influence the expression of genetic mutations, explaining why some traits vary more than others and why certain attributes may appear repeatedly in different lineages. On the other hand, even if ample genetic variation is available, natural selection may act to maintain traits in their current condition if an organism is well adapted to its current conditions. This process is known as stabilizing selection and may be pervasive in nature, although for a variety of technical reasons it can be quite hard to detect. The ability of plants and animals to migrate during episodes of climate change and track the environments to which they are well adapted may also be a process that reduces the rate of evolutionary change. There is no general consensus on the relative importance of these different factors that contribute to the phylogenetic signal in different traits, and it is very difficult to obtain all the relevant data in any particular case study. In the context of ecological research, it can be useful to quantify the pattern of phylogenetic signal and compare observed patterns to those expected under alternative evolutionary models. The Brownian motion model, in particular, provides an important point of comparison because it is the foundation of many comparative methods. Although Brownian motion represents a random model of evolutionary change, it does generate a fairly high degree of phylogenetic signal, as sister taxa diverge gradually from their common ancestors. In contrast, null models in which trait values are randomly rearranged among the species in a study provide a baseline measure for the complete absence of phylogenetic signal. Two closely related measures, Pagel’s l and Blomberg’s K statistic, are particularly useful, as they take on a value of 1 when patterns of trait similarity conform to expectations of Brownian motion and greater than or less than 1 when close relatives are more or less similar than expected, respectively. Another class of methods known as Mantel tests



is based on the correlation between the phylogenetic distances between species (the distance down the branches of the phylogeny to the common ancestor and back up to another species) and the ecological or phenotypic differences between them. These methods are useful for ecological characteristics such as niche overlap and co-occurrence where the degree of similarity or dissimilarity between species is quantified directly. Phylogenetic information can play an important role in the prediction of ecological traits when there is strong phylogenetic signal. For example, in a recent study, Je´roˆme Chave and colleagues demonstrated that wood density tends to be very similar among closely related tree species. Wood density is important for carbon storage, a critical factor in the global carbon cycle, but it has only been measured on a small proportion of tree species in the tropics. Knowledge that close relatives have similar wood density will allow more accurate prediction of carbon storage in diverse tropical forests, even for species for which wood density has not been measured directly. 4. PHYLOGENETICS AND COMMUNITY ECOLOGY

Phylogenetics is playing an increasingly important role in community ecology as a tool to gain insight into the processes that influence community structure. One of the earliest theoretical principles of ecology was the competitive exclusion theorem, formalized by Gause in the 1930s, which states that two species that utilize identical resources cannot coexist in a community. In the 1950s, this idea, together with the knowledge that closely related species are usually ecologically similar and therefore utilize similar resources, led to the prediction that species from the same genus should cooccur infrequently. This prediction was tested by calculating the average number of species per genus in isolated communities, such as islands, compared to the overall biota of the surrounding region. In the past 10 years, phylogenetic approaches to community ecology have been revitalized by the availability of highly resolved phylogenetic trees and new methods. In addition, developments in community assembly theory have emphasized an alternative view that co-occurring species may be more similar to each other than expected because similar traits may promote ecological success under particular environmental conditions. These two perspectives provide contrasting predictions regarding whether communities will be composed of more or less closely related species. Three steps are required to quantify the phylogenetic structure of ecological communities and test hypotheses about whether this structure is significantly

different than may be expected. First, the degree of relatedness among co-occurring species needs to be quantified, based on the best available phylogeny. Cam Webb and others have introduced several related methods to accomplish this. The simplest approach is simply to calculate the average phylogenetic distance between all pairs of species within the community. Other approaches take into account species abundance or measure the distance between each species and its closest relatives in the community, as opposed to more distant relatives. The second step is to specify a broader pool of species from which a particular community has been assembled. This provides the source pool to construct hypothetical communities that serve as a point of comparison with observed patterns. Ideally, the spatial scale defining this pool is large enough so that it includes all of the species that could, in a reasonable span of time, arrive at the community of interest. However, in practice, it is very difficult to determine exactly what this scale should be, and researchers rely on a variety of practical solutions to address this problem. Finally, one needs to construct an appropriate null model by which random communities can be drawn from this regional pool to determine whether the observed communities diverge from random expectations. Simple null models include a random draw of species, where each species is equally likely to be chosen. More complex models can be constructed, in which the probability of a species being chosen is proportional to its frequency of occurrence in the landscape. The construction and analysis of these null models are continuing points of discussion and development in this field. Many studies of phylogenetic community structure have appeared in recent years, and some generalizations are beginning to emerge. First, empirical and theoretical studies suggest an asymmetry in the interpretation of phylogenetic community data. It appears that clustering of close relatives within a community arises primarily from an ecological filtering process, in which similar species are favored as they share adaptations that are appropriate for the particular conditions. On the other hand, many different processes can lead to the opposite pattern in which communities are composed of more distant relatives than expected. These include competition, small-scale habitat heterogeneity, facilitative interactions among functionally disparate species, and even a filtering process when the traits that promote success have evolved independently in different clades. Theoretical studies also suggest that it is much harder to detect patterns in which coexisting species are distantly related, compared to the opposite pattern. A second result is the realization that communities will not be structured either by filtering or by competition or by any other single process. Many processes

Phylogenetics are likely at work, mediated by different sets of traits. For example, Jeannine Cavender-Bares and colleagues studied the composition of oak-dominated forests in Florida and found that local communities were generally composed of distantly related species. These species tended to share physiological traits affecting their water relations, with drought-adapted species occurring together on drier sites. Moreover, these hydraulic traits exhibited low phylogenetic signal, so similar species tended to be distantly related for these characteristics. On the other hand, co-occurring species displayed a high disparity of trait values related to acorn maturation and wood density. These traits exhibited a high degree of phylogenetic signal, but closely related species with similar values were distributed across different communities. Thus, it is critical to specify the traits that may be relevant to community assembly and examine their distribution on the phylogeny carefully before interpreting patterns of phylogenetic community structure in terms of particular underlying processes. Finally, there is a fascinating pattern in plant communities of a shift from the co-occurrence of more distant relatives when studies focus on a narrow clade (e.g., oaks) to a pattern of clustering of close relatives in broader studies that encompass the full spectrum of flowering plants or all seed plants. A similar shift occurs moving from smaller to larger spatial scales. Both of these patterns are consistent with a stronger role for resource partitioning among closer relatives and at smaller spatial scales, whereas habitat filtering becomes more apparent at larger spatial and phylogenetic scales. 5. PROSPECTS FOR THE FUTURE

The potential role of phylogenetics in ecology was heralded by several articles and books published in the late 1980s to mid-1990s. In the relatively short interval since then, many methods have been introduced or improved, and growth in research has been rapid. The number of citations in the scientific literature under the keywords phylogen and ecology rose from 4 in 1990 to 87 in 1995, 130 in 2000, and 275 in 2006. An important engine of this growth has of course been the constantly expanding availability and improved resolution of phylogenies for diverse groups of taxa, accompanied by new methods, fast computers, and easyto-use software. This chapter highlights two areas that


are most relevant to ecological research. Measures of phylogenetic diversity are also used as criteria to help prioritize taxa and habitats in conservation biology, and a wide variety of comparative methods are in use in evolutionary biology, including the study of adaptation, diversification, adaptive radiations, and related topics. Several important areas of challenge and opportunity lie ahead. One is the improved resolution of branch lengths and node ages on phylogenies, which will be provided by including more species and more genes and improvements in fossil calibration. Timecalibrated phylogenies are opening the door to linkages between comparative methods and paleoecology and will facilitate investigation of a new generation of questions. A second area is the development of global databases for ecological traits. This will allow us to assess questions of phylogenetic signal and ecological trait correlations across the entire phylogeny of major clades and to understand how the assembly of local floras and faunas relate to global patterns of ecological diversity. Third, phylogenetic methods are providing new insights into ecology and biogeography of microbes, fungi, and other groups that are difficult to study directly in the field. These are but a few of the growth areas at the intersection of phylogeny and ecology—the most exciting advances will be those that at this point are not even anticipated.

FURTHER READING Blomberg, S. P., and T. Garland, Jr. 2002. Tempo and mode in evolution: Phylogenetic inertia, adaptation and com parative methods. Journal of Evolutionary Biology 15: 899 910. Felsenstein, J. 1985. Phylogenies and the comparative method. American Naturalist 125: 1 15. Harvey, P. H., and M. Pagel. 1991. The Comparative Method in Evolutionary Biology. Oxford: Oxford Uni versity Press. Maddison, W. P., and D. R. Maddison. 1992. MacClade: Analysis of Phylogeny and Character Evolution. Sunder land, MA: Sinauer Associates. Pagel, M. D. 1999. Inferring the historical patterns of bio logical evolution. Nature 401: 877 884. Webb, C. O., D. D. Ackerly, M. McPeek, and M. J. Dono ghue. 2002. Phylogenies and community ecology. Annual Review of Ecology and Systematics 33: 475 505.

I.17 Microevolution Michael A. Bell OUTLINE

1. Evolution: Micro versus macro 2. ‘‘The ecological theater and the evolutionary play’’ 3. Microevolutionary mechanisms 4. Contemporary microevolution 5. The unintended consequences of human technology 6. Geographic variation 7. Phylogeography 8. Genomics and microevolution 9. Prospects Microevolution occurs within and among populations of a species and usually involves changes in the mean value or relative frequencies of alleles and phenotypes that are shared by most populations of the species. Divergence among populations of a species (i.e., conspecific populations) is often associated with habitat differences, and such divergence often has important ecological consequences. Population genetics deals with evolution in terms of allele and genotype frequencies within populations, so it provides the theoretical foundation to study microevolution. Widespread species typically exhibit geographic variation, which has generally been thought to take thousands of generations to evolve. However, recent research on contemporary evolution suggests that geographic variation can evolve within a few generations after species colonize new habitats or experience environmental change. The high rate at which microevolution can occur is important because it means that pathogens, pests, and harvested natural populations can rapidly evolve traits that adversely affect people. DNA variation within and among conspecific populations can be studied as a product of microevolution, and it also provides powerful tools to tease apart the contributions of common ancestry and local adaptation to the evolution of geographic variation. Thus, previously intractable problems in microevolution and its applications to natural resource management can now be studied using the emerging technologies of molecular biology and genomics.

GLOSSARY character displacement. This is the evolution of en-

hanced differences between species where they occur together as a result of selection against members of one or both species that use the same resources as members of the other species (i.e., ecological character displacement) or against individuals that tend to hybridize with members of the other species (i.e., reproductive character displacement). cline. A cline is a geographic gradient in the frequency or mean value of a phenotype or genotype. monophyletic group. This is a group of species that are more closely related to each other than any is to species outside the group. phenotypic plasticity. A change in an individual phenotype that does not alter its genetic constitution and is not inherited by its offspring. random walk. In population genetics, this is a change in allele frequencies from their initial values as a result of repeated episodes of genetic drift. taxon. A taxon (including higher taxon) is any named group (e.g., Vertebrata, Mammalia, hom*o sapiens) at any taxonomic rank (e.g., Kingdom, Class, Species); higher taxa are more inclusive. 1. EVOLUTION: MICRO VERSUS MACRO

Biological evolution is change through time in the heritable properties of a lineage or monophyletic group (clade). Microevolution is generally confined to evolution within and among conspecific populations, and it occurs within relatively short time spans. In contrast, macroevolution involves changes in the number or characteristic properties (e.g., average body size) of the species of a clade. It depends on the variation among species generated by microevolution and unfolds over longer periods. Nevertheless, the definitions of microevolution and macroevolution have been controversial, and there is disagreement about their mechanistic relationships and even the value of the terms.

Microevolution The division between microevolution and macroevolution is usually placed at speciation because members of different species do not routinely interbreed, and the evolutionary fates of separate species are largely independent. Microevolution involves changes in the frequencies of alleles and genotypes and of interactions between different genes. These changes are manifested as recognizable changes in the mean values or frequencies of biochemical, physiological, behavioral, developmental, and morphological phenotypes. A separate set of macroevolutionary mechanisms influences the probability of speciation and extinction. Thus, properties of species that promote speciation or impede extinction will tend to increase in a monophyletic group over time. Both microevolution and macroevolution contribute to biodiversity, but microevolution affects individuals and changes the properties of populations, whereas macroevolution alters the relative frequencies of species with different properties. There are also practical reasons to distinguish microevolution and macroevolution. Microevolution can be studied using comparative, observational, or experimental methods to study individuals and populations over a few generations in the laboratory and field. Existing genetic properties and ecological conditions can be used to interpret microevolution. In contrast, macroevolutionary studies focus on differences among species. Careful species description, characterization of clades, and investigation of phylogenetic relationships among taxa are paramount in macroevolutionary research. The environmental factors and genetic properties that influenced speciation and extinction have typically been lost in the dim past and are hard to infer. Consequently, microevolution and macroevolution are generally studied using different methods. 2. ‘‘THE ECOLOGICAL THEATER AND THE EVOLUTIONARY PLAY’’

G. Evelyn Hutchinson’s famous 1965 book, from which the title of this section was borrowed, emphasized that evolution occurs within an ecological context. Although existing genetic properties of a population (e.g., presence of an advantageous allele) influence its microevolutionary response to natural selection, ecology is a major factor in microevolution and a crucial source of information to interpret it. Furthermore, if environmental differences cause microevolutionary divergence among conspecific populations, they will exhibit differences that must be taken into account in ecological studies. Studies of microevolution and ecology are intimately associated and reciprocally illuminating.



Because microevolution involves changes in the relative frequencies of heritable traits within populations of a species, it can be analyzed in terms of the behavior of alleles and genotypes within populations. This is the subject of population genetics, and the Hardy-Weinberg equilibrium is the starting point to develop the genetic theory of microevolution. The Hardy-Weinberg equilibrium describes the distribution of alleles among diploid genotypes in a population in the total absence of evolution. It will be sketched here only briefly, but most textbooks on evolutionary biology develop it in detail (see chapter I.15). The Hardy-Weinberg Equilibrium

If no evolutionary forces impinged on a population, the relative frequencies of alleles and genotypes in the population would reach equilibrium values that would never change after one generation of random mating. Genotype frequencies under these Hardy-Weinberg equilibrium conditions can be calculated using the simple equation, 1 ¼ (p þ q)2, where p and q are the relative frequencies of two alleles of a gene and must sum to 1. Of course, no real population ever conforms to Hardy-Weinberg equilibrium conditions, although deviations from equilibrium frequencies are often so small that they are undetectable. Detectable deviations from equilibrium frequencies, however, indicate that microevolution is taking place and may suggest its causes. Potential causes for deviations from equilibrium frequencies include mutation, meiotic drive, assortative mating, gene flow, genetic drift, and natural selection, the most important of which are mutation, gene flow, genetic drift, and natural selection. Mutation

Mutations are heritable changes in DNA and are the ultimate source of variation for microevolution. However, mutation rates are so low (10–4–10–9 mutations/ generation/trait/individual) that they do not usually produce measurable departures from expected HardyWeinberg equilibrium frequencies. Phenotypic changes caused by mutation are not biased (i.e., random) to produce adaptation. Gene Flow

Gene flow occurs when an individual is born in a source population and reproduces after migrating to a recipient population. Its effects depend on the magnitude of genetic differences and rate of migration between



the donor and recipient populations. Gene flow is frequently a more important source of genetic variation than mutation, but it also tends to hom*ogenize populations that otherwise would evolve differences. Genetic Drift

Genetic drift is the change of allelic frequencies between generations just by chance (i.e., sample error). The magnitude of genetic drift is inversely related to effective population size (Ne), which, in turn, increases with the number of breeding individuals and evenness of the sex ratio and decreases as variation in the number of offspring per pair increases. When Ne is large, genetic drift causes very small differences between successive generations, but if Ne is small for even one generation, it can cause major changes in the genetic composition of a population. Even if Ne is consistently large, there will always be some drift, and its effects will accumulate, causing a ‘‘random walk’’ of small deviations that can add up to major changes in allelic frequencies over many generations. In populations with small Ne, rare alleles tend to be lost by drift, and even traits that are disfavored by natural selection can drift to high frequencies. Because there is initially only one copy of a new mutant allele, even advantageous mutations tend to be lost by drift. Natural Selection

Natural selection depends on three components: survival selection, fecundity selection (ability to produce offspring), and sexual selection (mating success of individuals compared to other members of the same sex). Each component results from differences in the relative rates of success of different phenotypic classes, and selection can be quantified as Darwinian fitness or a selection coefficient. Darwinian fitness depends strongly on the interaction of the phenotype with the environment, and environmental changes can cause changes in fitness associated with a phenotype. If a phenotype with high fitness is heritable, alleles that produce it will tend to increase through time (i.e., evolve). New phenotypes may appear in a population by means of gene flow, mutation, and sexual recombination of existing alleles, and natural selection can increase their frequencies and cause a population’s phenotype to evolve beyond its previous range of variation. Thus, natural selection is the major cause for evolutionary adaptation and phenotypic divergence among conspecific populations. Genetic drift also causes divergence among conspecific populations, and it is necessary to distinguish the effects of drift and selection. Phenotype–environment

correlations indicate selection but are not sufficient to identify the environmental variable that causes it. Further evidence based on differences in function or Darwinian fitness of phenotypes is necessary to confirm inferences based on phenotype–environment correlations. It is surprisingly difficult to establish that natural selection has caused microevolution of a specific trait. Nonheritable Change

Not all phenotypic variation among conspecific populations represents microevolution. Phenotypic plasticity may result from conditions experienced by the individual. Genetically identical individuals or the same individual at different times may differ because of phenotypic plasticity. For example, muscles may grow larger from exercise, skin may become darker from exposure to sunlight, and learning may alter behavior, but such changes do not affect the genetic constitution of the individual that experiences the phenotypic plasticity or that of its progeny. Similarly, maternal effects, phenotypic differences caused by a female’s nongenetic contributions (e.g., messenger RNA, yolk, parental care) to its progeny, may influence the offspring’s phenotype but not be inherited. However, the individual’s ability to exhibit phenotypic plasticity (i.e., show a phenotypic response to environmental variation) may be heritable, and thus plasticity may evolve. Phenotypic plasticity may cause nonheritable but ecologically important phenotypic variation among conspecific populations. Much of the phenotypic change caused by human environmental disturbance apparently results from phenotypic plasticity and not microevolution. 4. CONTEMPORARY MICROEVOLUTION

It is widely believed that microevolution is rarely rapid enough to be observed in progress. For many years, industrial melanism in the UK stood as the lone wellconfirmed example of contemporary evolution. The peppered moth, Biston betularia, and other moths and beetles evolved dark pigmentation where soot from industrial pollution darkened tree bark and killed lightcolored lichens on which the moths rest during the day. Although the speed with which industrial melanism evolved was never in doubt, questions arose about evidence that bird predation selects against moths that contrast with bark color. Recent results seem to confirm this effect, and many other cases of rapid evolution in response to human-induced environmental change have been reported in recent years. Initially, most of these cases involved evolution of resistance to insecticides by insects and to antibiotics by bacteria. It seemed possible that these cases of



Figure 1. Variation in armor, size, and shape of threespine stickle back. The specimen in the middle is a completely plated (high Eda allele), anadromous (sea run) stickleback, and those around the periphery are low plated (low Eda allele), but otherwise pheno typically diverse, freshwater stickleback from western North

America. All specimens were drawn to be the same size, and the scale bars equal 1 cm. Variable armor traits include the length and number of dorsal spines, expression of the pelvis (including ab sence), and number (including zero) and distribution of lateral plates. (Reprinted with permission from Bell and Foster, 1994)

contemporary evolution might differ from typical microevolution. Insects have large populations and short generation times, both of which favor rapid evolution, and bacteria have even larger populations and shorter generation times. Additionally, selection imposed by human technology might be more severe than selection under natural conditions. However, it is also possible that evolution of resistance in insects and bacteria may not be atypical; it may just be more conspicuous because it has serious consequences for people. Consequently it is quickly noticed and carefully studied. Darwin’s finches are the classic case of adaptive radiation, and they have been closely observed for decades (see chapter I.19). They occur on the Gala´pagos Islands of Ecuador, which are relatively undisturbed. Dramatic evolutionary changes in body and bill size and in bill shape evolved in the medium ground finch, Geospiza fortis, and the cactus finch, G. scandens, in response to climate-induced changes in food availability during a 30-year period. Similarly, field mustard, Brassica rapa, grown from seeds collected after a 5-year drought in California exhibited higher tolerance

to drying and flowered earlier than seeds collected before the drought. Sea-run (anadromous) threespine stickleback fish, Gasterosteus aculeatus, which colonized a lake in Alaska, also evolved rapidly under relatively natural conditions. Freshwater stickleback are phenotypically diverse and differ strikingly from sea-run populations (figure 1). Within a dozen generations after the lake was colonized, several armor, body shape, and feeding traits evolved, and this young population has become indistinguishable from adjacent lake populations. Contemporary evolution is not restricted to life history and anatomical traits. The well-known fruit fly, Drosophila melanogaster, originated in the Old World but has been transported to every continent except Antarctica by humans. It lays its eggs in rotting fruit, where ethanol may reach lethal concentrations. Alcohol dehydrogenase (ADH) is one of the enzymes that detoxifies ethanol. There are two common alleles for the alcohol dehydrogenase gene, Adh-F and Adh-S. Adh-F is less stable at high temperatures and has higher activities at lower temperatures. It gradually increases



in frequency (i.e., clinally) going away from the equator in Australia, Asia, and North America. The functional attributes of enzymes encoded by these enzyme alleles (i.e., allozymes) and other evidence suggest that these clines result from natural selection, and they must have evolved since D. melanogaster was introduced to these continents. Moreover, within the last 20 years, the position of an Adh cline in Australia has shifted southward by 400 km so that high Adh-F allele frequencies now occur farther from the equator than before, as would be expected from global warming. Thus, this recent microevolutionary change can be used to monitor global climate change. Numerous additional examples of contemporary evolution have been described in recent years (see Hendry and Kinnison, 2001), indicating that natural environmental change and colonization of new habitats frequently cause detectable microevolutionary change on a human time scale and that microevolution is often so fast that close monitoring is necessary to catch it in the act. 5. THE UNINTENDED CONSEQUENCES OF HUMAN TECHNOLOGY

In 1962, Rachel Carson’s Silent Spring sounded the alarm that insecticides, which had been in use since the late 1940s to limit crop pests and disease vectors, were also eliminating many desirable, nontarget species. She did not know that insecticides also caused rapid microevolution of resistance in the insect populations that were their targets. Of course, large population size, short generation time, and intense selection by insecticides should cause rapid evolution of resistance. By 1997, more than 500 insect species had evolved resistance to insecticides. Weeds were no different, and herbicide resistance had evolved in more than 200 weed populations by 2001. It has also become painfully clear that pathogenic bacteria, fungi, protozoans, and metazoans routinely evolve resistance to drugs that had previously produced cures. These microevolutionary responses adversely affect human health and agriculture, but they can sometimes be mitigated by using management strategies based on microevolutionary theory. For example, genes from the bacterium Bacillus thuringiensis (Bt) have been inserted into the genome of cotton plants, conferring on them the ability to express Bt toxin, which kills pink bollworm, Pectinophora gossypiella. After 8 years of planting Bt-transgenic cotton, bollworm populations in Arizona still had not evolved resistance to Bt toxin. Failure of P. gossypiella to evolve resistance was apparently achieved by growing small plots of nontransgenic cotton, where non-

resistant bollworms thrived. Because the nonresistant bollworms far outnumbered rare, resistant bollworms that survived in the Bt-transgenic-cotton fields, and they tend to disperse into the Bt-transgenic fields, most bollworms the following season are either nonresistant or hybrids between resistant and nonresistant parents, both of which are killed by Bt toxin. By sacrificing small plots of nonresistant cotton to bollworm infestation, natural selection favoring resistance to Bt toxin has been retarded by gene flow from nonresistant bollworm populations. Experience with the evolution of antibiotic resistance has been far less encouraging. By 1943, penicillin production was under way, and other antibiotics would soon follow. Even as the first antibiotics became available for clinical use, penicillin-resistant bacteria had already appeared, and resistance soon occurred in one bacterial pathogen after another. For example, penicillin could easily cure Staphylococcus aureus (staph) infection in the early 1950s, but by the late 1960s it had become ineffective. Methicillin still worked, but it became ineffective by the 1990s. Most staph infections can still be cured by vancomycin, but resistance to this ‘‘antibiotic of last resort’’ is spreading. New drugs continue to be developed, but this is an expensive arms race with tragic consequences and no end in sight. In retrospect, microevolution of antibiotic resistance in bacteria is not surprising. Genetic variation for antibiotic resistance is common in bacterial populations, and their large size (i.e., Ne) and short generation time both facilitate the appearance of new mutants. Genes for resistance may protect bacteria from multiple antibiotics, and they can be transferred in plasmids between bacterial species. Natural selection for antibiotic resistance has been hastened by the misuse of antibiotics for diseases against which they are ineffective, failure of patients to complete antibiotic treatment, and widespread, chronic, low-dose antibiotic treatment to increase livestock productivity. All of these practices selectively eliminate less resistant bacterial clones, leaving behind more resistant ones to found new bacterial populations. Under favorable conditions, it is also possible to select for reduction in the severity of disease (i.e., virulence). Many pathogens rely on their host’s social interactions to spread to new hosts before it dies or mounts an immune response that eliminates the infection. Consequently, if hosts with the most severe infections are isolated from other susceptible individuals, the most virulent pathogen strains will fail to spread. For example, installation of window screens in the southeastern United States during the first half of the twentieth century prevented mosquitoes from biting people who stayed indoors with the most serious cases of malaria,

Microevolution contributing to evolution of lower virulence in Plasmodium, the malaria pathogen. By separating malaria victims with the most severe symptoms from mosquitoes that spread Plasmodium, the most virulent strains could not spread, and the disease became less serious. Human-induced microevolution may also play a crucial role in the loss of valuable commercial fish populations. Commercial fishing gear selectively captures larger fish, but smaller individuals slip through the net. Fisheries policies are intended to allow young fish to escape and grow to a larger size, at which they both reproduce and become more valuable as food. However, selective fishing for larger fish also favors individuals that stop growing at a smaller size and reproduce earlier in life. Because body size and reproductive schedules are heritable, size-selective fishing should cause evolution of smaller adult body size and earlier reproduction. After fishing is halted, the survivors should have genotypes for smaller body size and early reproduction. Many commercially fished populations never recover numerically, and those that do are often descendants of small individuals from which they inherit small body size. The conclusion that size-selective fishing has caused evolution of smaller body size in commercial fishes has been controversial because the quality of the marine habitats in which declining fish populations live has also deteriorated. Nevertheless, a growing minority advocates the incorporation of microevolutionary principles into fisheries’ management policy. 6. GEOGRAPHIC VARIATION

Variation that has evolved among populations in response to local ecological differences is a common phenomenon and an important source of evolutionary insight. Comparison of mainland or large central populations to populations on islands or peripheral habitat patches (e.g., mountain tops, desert springs) often reveals variation that is associated with environmental differences. Island populations may be isolated from predators, competitors, parasites, and pathogens that occur on the mainland, or they may encounter resources that are unavailable on continents, leading to evolution of unusual traits. The divergent properties of insular populations must be interpreted with care because insular populations are often small (i.e., low Ne) or were bottlenecked in the past, allowing genetic drift to influence their evolution. Ecological character displacement may occur when closely related species that are usually allopatric occur sympatrically. In sympatry, the members of each species that most closely resemble those of the other one may compete poorly with it, and natural selection will tend to eliminate intermediate individuals and favor


evolution of enhanced differences between them. Inference of ecological character displacement has been controversial, but some cases are well supported. Clines, which were mentioned in passing before, are geographic gradients in the frequencies of genotypes or phenotypes or in phenotypic means. They may evolve in an initially hom*ogeneous population that experiences an environmental gradient or even sharp differences in natural selection in different parts of its range. Clines may also form where previously isolated contrasting populations come into secondary contact and hybridize. Populations separated by a cline may eventually merge, or selection against hybrids owing to ecological or genetic differences may cause the populations to retain their differences. Similar clines may occur in multiple species and are recognized as biogeographic rules. For example, in endothermic vertebrates, body size tends to increase (Bergmann’s rule), extremities tend to be shorter (Allen’s rule), and coloration tends to be paler (Gloger’s rule) toward the poles. Similarly, the number of vertebrae increases toward the poles in fishes (Jordan’s rule). Clinal variation among conspecific populations is a conspicuous and ubiquitous manifestation of microevolution. 7. PHYLOGEOGRAPHY

DNA sequence variation is most strongly influenced by genetic drift, and its evolution should be largely independent of phenotypic microevolution. It should carry a strong signal of evolutionary relationships or phylogeny. Heritable phenotypic differences among populations, however, may reflect both phylogeny and local natural selection (adaptation). Although gene flow complicates the analysis, it is possible to reconstruct the phylogeny of conspecific populations, which is called phylogeography, using DNA sequence data to distinguish the effects of phylogeny and adaptation. It is possible that different genes will indicate different relationships among populations of a species because one gene may have been present when one population split into two, and another may have entered one population by gene flow long after the two populations split. Variation of allozymes and restriction fragment length polymorphisms (RFLP) in mitochondrial DNA (mtDNA) were used in phylogeography until the late 1980s, when development of the polymerase chain reaction (PCR) enabled widespread use of DNA sequences from the nuclear genome. Phylogeographic analysis of sockeye salmon, Oncorhynchus nerka, illustrates the value of this approach. Sea-run sockeye salmon are widespread throughout the north Pacific, but rare, isolated lake-resident populations of O. nerka, called kokanee, also exist.



Anadromous sockeye are about twice the size of kokanee but spend only 1 to 3 years at sea before spawning in fresh water. The smaller kokanee remain in fresh water and spawn after 2 to 7 years. The phenotypic similarity of isolated kokanee populations throughout their range suggested that they evolved in one place and spread from there, but their wide distribution suggested that they evolved repeatedly from local anadromous sockeye. Analyses of DNA sequence variation showed decisively that kokanee populations are genetically similar to adjacent migratory sockeye populations. They must have evolved many separate times from sockeye, and the similarity of isolated kokanee populations is a result of repeated (convergent) microevolution of similar adaptations to similar habitats. 8. GENOMICS AND MICROEVOLUTION

Development of large DNA-sequence databases, including whole, sequenced genomes, laboratory methods to inexpensively obtain DNA data, and statistical methods to analyze them have created exciting opportunities to study microevolutionary processes, phylogeography, and the genetics of microevolution. Molecular markers, including RFLP, single-nucleotide polymorphisms (SNPs), and short tandem repeats (microsatellites) can be used to study the number of genes, the relative importance of different genes, their location in the genome, and even which parts of genes underlie phenotypic evolution. In the threespine stickleback, for example, at least four genes on separate chromosomes control variation in the number of lateral armor plates (see figure 1), but the Ectodysplasin (Eda) gene accounts for more than 75% of the variation in plate number. Eda affects development of other vertebrate traits (e.g., teeth, fish scales, mammal hair, and sweat glands), so a change in the structure of the EDA protein would probably have adverse effects on other stickleback traits. As expected, the protein-coding region of Eda does not differ consistently between alleles for high- and low-plate-number phenotypes, implicating altered expression of Eda in evolution of plate number. Insertion of an Eda gene from a mouse into the genome of a low-plated stickleback caused an increase in the number of plates, confirming Eda’s role in plate number evolution. Remarkably, an ancestral Eda allele for low-armored phenotypes originated by mutation more than 10 million years ago, and alleles that have evolved from it have spread across the Pacific, Arctic, and Atlantic oceans from a single point of origin, providing the genetic variation on which natural selection acted to cause evolution of low plate number throughout this huge range. However, different genes with smaller effects

may cause plate number differences in neighboring populations. Genomics has also contributed an entirely new method to study natural selection. If a novel allele enters a population by mutation or gene flow, it may initially be represented by one copy. If this allele experiences strong positive selection, it will quickly be fixed (i.e., reach 100%), and the DNA sequence surrounding it will also be fixed before recombination with DNA surrounding other alleles for the same gene can occur. Until this region accumulates mutations, heterozygosity will tend to be depressed around the positively selected allele. Such depressed heterozygosity is the signature of a recent rapid selective sweep. Studies of the human and other genomes have detected numerous selective sweeps, and extended DNA sequences surrounding genes already believed to have experienced selective sweeps tend to be surrounded by regions of depressed heterozygosity. For example, an allele for the Lct gene that confers the ability in adult humans to digest milk sugar (lactose) is surrounded by a long stretch of DNA with reduced heterozygosity, suggesting a recent selective sweep when humans became consumers of raw milk. 9. PROSPECTS

A wide range of methods continues to be used to develop new insights into the mechanisms for evolution within species. Since the mid-1960s, application of molecular biology to microevolutionary problems has revolutionized the field. Molecular methods and genomic data will continue to be used in phylogeography, and they will increasingly be applied to research on the genetics and development of phenotypes that vary within and among conspecific populations. As the power of molecular methods and number of sequenced genomes increase, new opportunities to use geographic variation to investigate basic problems in genetics and development will appear and create additional tools for research in microevolution. Increasingly, microevolutionary theory and molecular biology will be combined to address problems related to human health, agriculture, and natural resource management. See also chapters I.9 and I.13–I.16. FURTHER READING Avise, John. 2000. Phylogeography: The History and For mation of Species. Cambridge, MA: Harvard University Press. A review of methods to use molecular traits to infer microevolutionary history of conspecific populations. Bell, Michael A., and Susan A. Foster, eds. 1994. The Evo lutionary Biology of the Threespine Stickleback. Oxford:

Microevolution Oxford University Press. A review of microevolutionary phenomena and mechanisms in this microevolutionary model species. Colosimo, P. R., K. E. Housemann, S. Balabhadra, G. Vil larreal, Jr., M. Dickson, J. Grimwood, J. Schmutz, R. Myers, D. Schluter, and D. M. Kingsley. 2005. Wide spread parallel evolution in sticklebacks by repeated fixation of Ectodysplasin alleles. Science 307: 1928 1933. An extraordinary molecular analysis demonstrating that a single mutation created the variation on which natural selection has acted to produce a global pattern of geo graphic variation in a conspicuous skeletal trait. Endler, John A. 1986. Natural Selection in the Wild. Monographs in Population Biology 21. Princeton, NJ: Princeton University Press. An encyclopediac review of natural selection. Endler, John A. 1995. Multiple trait coevolution and envi ronmental gradients in guppies. Trends in Ecology and Evolution 10: 22 29. An overview of the microevolution of multiple phenotypes in guppies. Ewald, Paul W. 1994. Evolution of Infectious Disease. New York: Oxford University Press. A ground breaking book on pathogen microevolution. Grant, Peter R. 1986. Ecology and Evolution of Darwin’s Finches. Princeton, NJ: Princeton University Press. An


account of microevolution and speciation placed within exceptionally thorough biological and environmental contexts. Hendry, A. P., and M. T. Kinnison. 2001. Microevolution: Rate, Pattern, Process. Dordrecht: Kluwer Academic Publishers. (Also published as Genetica, vol. 112 113.) An excellent collection of review papers concerning rates, patterns, mechanisms, and ecological contexts for micro evolution in contemporary populations. Lederberg, Joshua, and Esther M. Lederberg. 1952. Replica plating and indirect selection of bacterial mutants. Journal of Bacteriology 63: 399 406. A classic study showing that mutations are random with respect to adaptation. Majerus, Michael E. N. 1998. Melanism: Evolution in Ac tion. Oxford: Oxford University Press. An excellent re view of the classic case of contemporary microevolution, it exposed some methodological flaws, which, however, do not invalidate this case. Palumbi, Stephen R. 2001. The Evolution Explosion. New York: W. W. Norton & Company. An excellent popular review of microevolution in response to human technology. Wilson, Edward O., and William H. Bossert. 1971. A Primer of Population Biology. Sunderland, MA: Sinauer Associ ates. A concise quantitative treatment of basic population genetics.

I.18 Ecological Speciation: Natural Selection and the Formation of New Species Patrik Nosil and Howard Rundle OUTLINE

1. Ecological speciation: What it is and how to test for it 2. Forms of divergent selection 3. Forms of reproductive isolation 4. Genetic mechanisms linking selection and reproductive isolation 5. Geography of ecological speciation 6. Generality of ecological speciation 7. Remaining questions in the study of ecological speciation Understanding how new species arise is a central goal of evolutionary biology. Recent years have seen renewed interest in the classic idea that adaptive evolution within species and the origin of new species are intimately linked. More specifically, barriers to genetic exchange between populations (termed reproductive isolation) are the hallmark of species, and evolutionary biologists have been asking whether ecologically based divergent natural selection, the process that is responsible for adaptive divergence between populations, may cause such reproductive barriers to evolve. Convincing examples of this process, termed ecological speciation, are accumulating in the literature, and comparative approaches suggest that it may be a widespread phenomenon taxonomically. Attention is now being given to understanding details of the process and uncovering generalities in its operation. Three main components of ecological speciation can be recognized: a source of ecologically based divergent selection, a form of reproductive isolation, and a genetic mechanism linking the two. Current research is focused on understanding these components during the various stages of ecological speciation from initiation to completion.

GLOSSARY ecologically based divergent selection. Selection arising

from environmental differences and/or ecological interactions (e.g., competition) that acts in contrasting directions on two populations (e.g., large body size confers high survival in one environment and low survival in the other) or favors opposite extremes of a trait within a single population (i.e., disruptive selection) linkage disequilibrium. A statistical association between alleles at one locus and alleles at a different locus, the consequence of which is that selection on one locus (e.g., a locus affecting an ecological trait such as color pattern) causes a correlated evolutionary response at the other locus (e.g., a locus affecting mating preference) pleiotropy. Multiple phenotypic effects of a gene (e.g., a gene affecting color pattern also affects mating preferences) postmating isolation. Barriers to gene flow that act after mating (e.g., intermediate trait values of hybrids that make them poor competitors for resources, reducing their fitness) premating isolation. Barriers to gene flow that act before mating (e.g., divergent mate preferences that prevent copulation between individuals from different populations) reproductive isolation. A reduction or lack of genetic exchange (gene flow) between taxa sympatric speciation. A geographic mode of speciation whereby a single population splits into two species in the absence of any geographic separation, often via disruptive selection


The idea that the macroevolutionary phenomenon of speciation is the result of the microevolutionary process of adaptation dates back at least to Charles Darwin. However, it was not until the popularization of the biological species concept in the middle of the last century, whereby speciation was defined as the process by which barriers to genetic exchange evolve between populations, that the study of Darwin’s ‘‘mystery of mysteries,’’ the origin of species, became empirically tractable. The past two decades have witnessed an explosion of speciation research, with much attention being given to understanding the role of divergent selection in speciation. As defined by Dolph Schluter and others, ecological speciation is the process in which barriers to genetic exchange evolve between populations as a result of ecologically based divergent natural selection. Selection is ecological when it arises from differences in the environment or from interactions between populations over resource acquisition. Ecologically based selection can thus arise, for example, from an individual’s quest to obtain food and other nutrients, attract pollinators, or avoid predators. It can also arise from an individual’s interaction with other organisms in its attempt to achieve these goals (e.g., resource competition, predation). Selection is divergent when it acts in contrasting directions in the two populations. Included here is the special case of disruptive selection on a single population, in which selection favors opposite extremes of the same trait. During ecological speciation, populations experience divergent selection between environments or niches and thus differentiate in ecologically important traits. If these traits, or ones that are genetically correlated with them, affect reproductive isolation, then speciation occurs as a consequence. Ecological speciation is distinguished from other models of speciation in which the evolution of reproductive isolation involves processes other than ecologically based divergent selection. These include models in which chance events play a central role, including speciation by polyploidization, hybridization, genetic drift, and population bottlenecks (i.e., drastic reductions in population size). Nonecological speciation also includes models in which selection is involved, but it is nonecological (e.g., sexual conflict, in which selection arises from an evolutionary conflict of interest between the sexes over traits related to reproduction), or it is not divergent between environments. An alternative definition of ecological speciation would restrict it to situations in which the reproductive barriers themselves are ecological in nature, such as


reduced hybrid fitness arising because intermediate hybrid phenotypes cause them to perform poorly in either parental environment (i.e., see the third point below). In contrast, incompatibilities between the parental genomes, expressed when they are brought together in hybrids, is an example of a nonecological barrier. However, when the goal is to understand mechanisms of speciation, it is of interest when both ecological and nonecological forms of reproductive isolation evolve through a specific evolutionary process (e.g., ecologically based divergent selection). Ecological speciation can therefore involve the evolution of any type of reproductive barrier, so long as ecologically based divergent selection is responsible. Ecological speciation can also occur under any spatial arrangement of populations, with population pairs being geographically separated (allopatry), contiguous (parapatry), or in complete contact (sympatry). Under any of these geographic scenarios, if divergent selection drives the evolution of reproductive isolation, then speciation is classified as ecological. Laboratory evolution experiments using Drosophila fruit flies have shown that ecological speciation is feasible: when replicate populations are independently adapted to one of two environments, reproductive isolation tends to arise between populations from different environments, but not between populations evolved in similar environments. Classic examples of such experiments come from the work of Diane M. B. Dodd and of George Kilias and colleagues (figure 1). Convincing examples of ecological speciation in nature are also accumulating, with empirical tests tending to focus on three forms of evidence. First, ecological speciation predicts that the strength of reproductive isolation between pairs of populations will be positively related to the magnitude of their ecological differentiation, independent of any correlation with divergence time (figure 1). This has been shown in Timema walking-stick insects studied by Patrik Nosil, Bernie J. Crespi, and Cristina P. Sandoval, in which pairs of populations adapted to different hostplant species exhibit stronger reproductive isolation than do pairs of populations adapted to the same hostplant species (figure 1). A special case of this scenario, termed parallel speciation, occurs when the same reproductive barriers evolve in independent populations experiencing similar environments. The Drosophila laboratory experiments described above demonstrate the initial stages of parallel speciation: independent populations adapted to one environment were reproductively isolated from populations adapted to the other environment but not from one another. A prime example of parallel speciation in nature comes from freshwater stickleback (Gasterosteus) fishes studied by

Autecology B. Plants

Ec divolog erg ica en l ce

Reproductive isolation

A. Hypothetical

Residual postmating isolation


Genetic distance time

R2 = 0.18 P < 0.001 0

0.25 0.5 0.75 1.0 Ecological (habitat) divergence

D. Walking-sticks 0.8 Total reproductive isolation

Sexual isolation (proportion between-population mating)





0.2 Same Environment



Cryptic postmating isolation Sexual isolation Immigrant inviability Habitat isolation (via host preference


Similar (same hosts) Ecological divergence

Figure 1. Tests for ecological speciation, where the premise is to isolate an association between ecological divergence and levels of reproductive isolation, independent of the amount of time that population pairs have had to diverge from one another via non ecological processes such as genetic drift. (A) A hypothetical sce nario in which reproductive isolation increases with both genetic distance (a proxy for time) and ecological divergence. (B) The pat tern predicted by ecological speciation, in this case a positive as sociation between habitat divergence and residual postmating isolation (the effects of time have been statistically removed) be tween angiosperm species. The data come from a comparative study by Funk and colleagues, and the figure is reprinted with permission of the National Academy of Sciences U.S.A. (C) Evidence for ecological speciation from laboratory evolution studies. Shown

is the proportion of matings occurring between independently evolved lines of Drosophila as a function of the similarity of their environments. Between population mating is less common when populations have been adapted to different environments. Open circles are from work by Dodd, and closed circles are from work by Kilias and colleagues. (D) In Timema cristinae walking stick in sects, multiple forms of reproductive isolation are stronger be tween pairs of populations using different host plant species (i.e., pairs with divergent ecologies) than between similar aged pairs of populations using the same host plant species (i.e., pairs with similar ecologies). The pattern was documented in a series of studies by Nosil, Crespi, and Sandoval, and the figure is reprinted with permission of the American Society for Naturalists.

Dolph Schluter, Howard D. Rundle, and colleagues. These fish come in two main forms, a slender limnetic that feeds primarily on plankton in the open water of a lake and a more robust benthic, which feeds on invertebrates in the shallows. Sympatric limnetic–benthic pairs occur in a number of lakes in western Canada, and molecular genetic evidence suggests that the pairs

have arisen independently (i.e., the present-day phenotypic similarity of limnetics, and of benthics, from separate lakes is the result of parallel evolution and not shared ancestry). Mating trials demonstrate that reproductive isolation between limnetics and benthics has likewise evolved in parallel: premating isolation is strong between limnetics and benthics, even when they

Ecological Speciation A. Divergent selection


Environment #1 Environment #2

Trait value B. Form of reproductive isolation Premating

Postmating Hybrids

C. Genetic mechanism RI gene and selected gene Pleiotropy

RI gene Selected gene

LD with physical linkage

RI gene Selected gene LD without physical linkage


Weak Expected strength of genetic association

Figure 2. A schematic illustration of the three components of eco logical speciation. (A) A form of divergent selection is required, where selection is divergent when it acts in contrasting directions in two populations. (B) Forms of reproductive isolation are numerous and can act either before or after mating (premating and postmating isolation, respectively). Depicted are butterflies from populations adapted to different habitats (dark gray versus white winged indi viduals) and hybrids between these parental forms (light gray). The form of premating isolation depicted is sexual isolation, where indi viduals prefer to mate with individuals that are the same color as themselves. The form of postmating isolation shown is one where hybrids suffer reduced fitness because their intermediate phenotype renders them unfit in either parental environment. Specifically, the hybrids do not match the substrate in either parental environment and suffer increased rates of visual predation as a result of this lack of crypsis. (C) A genetic mechanism is required to transmit selection on genes conferring local adaptation to genes causing reproductive isolation. Reproductive isolation can evolve because the genes under selection are the same as those conferring reproductive isolation (i.e., pleiotropy). Alternatively, reproductive isolation might evolve via the statistical association between genes under selection and those conferring reproductive isolation. This statistical association is termed linkage disequilibrium and is facilitated by proximity of the different genes on the same chromosome (i.e., physical linkage).


are taken from separate lakes, whereas premating isolation is absent within a form, even when they derive from different lakes (e.g., premating isolation is lacking between limnetic forms from different lakes). The repeated evolution of the same barriers to gene flow, in correlation with ecological divergence, is unlikely to occur via nonecological processes (e.g., genetic drift) and thus provides strong comparative evidence for ecological speciation. Second, ecological speciation is facilitated when the traits under divergent selection also cause reproductive isolation pleiotropically; such pleiotropy is therefore predicted to be common in cases of ecological speciation. A clear example comes from the work of Jim Mallet, Chris Jiggins, and colleagues on butterflies in the genus Heliconius. In this group of tropical butterflies, natural selection acts on mimetic coloration to reduce visual predation. Geographic variation in the phenotype of the comimic generates divergent selection among populations to match the local form. Because these color patterns are also used in mate choice, divergence in coloration generates premating isolation (i.e., sexual isolation) as a side effect. Third, ecologically based divergent selection predicts that if hybrids can be formed between the parental populations, their fitness should be reduced for ecological reasons. This normally occurs because hybrid phenotypes are intermediate between the two parental forms, making them ill-suited for various tasks (e.g., acquiring resources, avoiding predation, finding a mate) in either parental environment. This type of hybrid fitness reduction is unlikely to arise via nonecological mechanisms of speciation. Ecologically dependent reductions in growth rate have been shown in limnetic–benthic hybrids, and reduced survival of Heliconius hybrids is likely to arise from their intermediate coloration that fails to mimic either parental form. As the above studies highlight, the case for ecological speciation is compelling: it clearly can, and does, occur. Attention is now being given to understanding the details of the process, including the three main components (a source of divergent selection, a form of reproductive isolation, and a genetic mechanism linking the two; figure 2), to testing for generalities, and to uncovering the geographic context of ecological speciation. 2. FORMS OF DIVERGENT SELECTION

The first component of ecological speciation is a source of divergent selection (figure 2A). Three main sources have been recognized: differences between environments, ecological interactions, and sexual selection.



Differences between Environments

Divergent selection can arise because of differences between populations in their environments, including, for example, habitat structure, climate, and resource availability. As populations adapt to different environments, they may diverge from one another in many ways, evolving to look different, smell different, and behave differently. Such differences will contribute to speciation if, for example, they reduce the likelihood of betweenpopulation mating (perhaps because individuals from different populations no longer recognize one another as potential mates as in the Heliconius butterflies discussed above), or they reduce the fitness of any hybrids that are formed (perhaps, as mentioned earlier, because such hybrids are ill-suited to either parental niche, as in the stickleback example discussed above). It is highly unlikely that any two environments are identical, and it is not surprising, therefore, that environmental differences appear to be a common cause of divergent selection during ecological speciation.

Ecological Interactions

Divergent selection may also arise between populations as a result of their ecological interactions with one another, most notably competition for shared resources. Divergent selection arising from such interactions is frequency dependent because individual fitness depends on the frequency of the various phenotypes within the population. Although interspecific competition appears common in nature and may play a key role in sympatric divergence between taxa, its consequences for the evolution of reproductive isolation are poorly understood. In the threespine sticklebacks discussed earlier, for example, resource competition has been strongly implicated in the morphological divergence of limnetics and benthics (i.e., ecological character displacement), as has adaptation to their different environments. Although the latter form of divergent selection has been implicated in the evolution of reproductive isolation, unambiguous evidence that the former promotes reproductive isolation is lacking. Interbreeding (hybridization), another type of interaction between populations, can also contribute to ecological speciation via a process known as reinforcement. Reinforcement occurs when hybrids have reduced fitness such that selection favors parental individuals that are less likely to hybridize, thereby strengthening premating isolation. Although it features prominently in many models of speciation, reinforcement is difficult to categorize because it can complete a speciation process initiated by any mechanism, eco-

logical or not. However, if hybrid fitness is reduced by ecological means, then reinforcement can be considered a component of ecological speciation. Reinforcement has been implicated in the ecological speciation of limnetic and benthic threespine sticklebacks: postmating isolation is ecological in nature, and premating isolation in sympatry appears to have been strengthened in response. Sexual Selection

Sexual selection has long been hypothesized to be a powerful mechanism of speciation because it involves communication between a signaler and a receiver, thereby creating the potential for rapid coevolutionary diversification of mating signals and preferences that may generate reproductive isolation. Divergent sexual selection arises when mate preferences differ between populations. Such selection is considered a component of ecological speciation when it is initiated by divergent selection between environments. This can occur, for example, if two habitats vary in their signal transmission properties such that different signals are most detectable (i.e., favored by natural selection) within each. Different mating signals and preferences may then evolve in populations occupying either habitat. For example, Manuel Leal and Leo J. Fleishman studied populations of Anolis lizards that occupy different (mesic versus xeric) habitats. They found that light conditions differ between these habitats and that the dewlap spectral traits of the lizards, which are important for social and mating communication, have diverged between populations using different habitats in ways that increase signal detectability within the native habitat, potentially generating premating isolation between these populations. Divergence of mating signals can also occur if sensory or communication systems adapt to their specific environment, even outside of the mating context (e.g., to facilitate resource acquisition or predator avoidance). Several examples of divergence in display traits, sensory systems, or preferences in correlation with environment have now been reported. 3. FORMS OF REPRODUCTIVE ISOLATION

The second component of ecological speciation is the form of reproductive isolation, of which many are possible, and speciation may involve one or more of them. Forms of reproductive isolation are commonly classified according to whether they occur before or after mating (premating and postmating isolation, respectively; figure 2B). The role of these reproductive barriers in speciation was thoroughly reviewed in a

Ecological Speciation recent book on speciation by Jerry Coyne and H. Allen Orr (2004). Premating Isolation

Premating isolation can arise when populations are separated in space (habitat) or time. Habitat isolation occurs when populations exhibit genetically based preferences for separate habitats, reducing the likelihood of between-population encounters and thus of interbreeding. For example, divergent host-plant preferences cause partial reproductive isolation between many herbivorous insect populations that mate on the plant on which they feed. Temporal isolation occurs when populations exhibit divergent developmental schedules such that mating happens at different times in each. A classic empirical example of such forms of reproductive isolation comes from the apple and hawthorne host races of Rhagoletis flies studied by Guy Bush, Jeffery L. Feder, and colleagues. Differences in host-plant preferences and developmental schedules cause substantial reproductive isolation between these host races. Additionally, if individuals immigrating into a foreign habitat are maladapted and die before mating, this immigrant inviability will also act to reduce interbreeding. All these forms of reproductive isolation are inherently ecological and thus are expected to commonly play a role in ecological speciation. Another barrier that acts before mating is sexual isolation, arising from differences between populations in their mating signals and preferences. Sexual isolation is considered by many to be the main component of reproductive isolation between recently evolved taxa. Consistent with this, studies in cichlids and Drosophila have shown that sexual isolation appears necessary for species to coexist in nature. Likewise, the laboratory Drosophila experiments discussed above demonstrate that sexual isolation can evolve relatively rapidly when populations are subjected to divergent natural selection (figure 1). A number of examples from nature also exist in which adaptation to different environments has been implicated in the evolution of sexual isolation, including beetles, walking sticks, butterflies, and stickleback fish. For example, in the stickleback fish discussed previously, adaptation to their different habitats (open water versus shallows) causes divergence in body size, and because mate choice is assortative with respect to size, sexual isolation arises as a by-product. Postmating Isolation

Postmating isolation can arise when hybrid fitness is reduced because of an ecological mismatch between intermediate hybrid phenotypes and the environment,


as was discussed earlier in the evidence for ecological speciation. An example of such ecologically dependent reductions in hybrid fitness stems from work on limnetic–benthic sticklebacks. Hybrids between the limnetic and benthic forms exhibit high fitness in the laboratory. In contrast, the fitness of hybrids in the wild is reduced relative to parental forms. Use of various types of hybrid crosses has shown that this reduction was a direct result of their intermediate phenotype and was not caused by genetic incompatibilities between the two forms (that could arise via any mechanism of speciation). Hybrids can also suffer reduced fitness because their sexual display traits and/or mate preferences reduce their mating success, in effect generating sexual selection against them. This has been shown in hybrid male sticklebacks in work by Steven Vamosi. Postmating isolation can also result from genetic incompatibilities between divergent genomes, caused by negative interactions between genes that differ between populations, when these genes are brought together in hybrids. These incompatibilities reduce the fitness of hybrids and do not depend on an ecological interaction between phenotype and environment. However, it is still possible that such incompatibilities evolve as a byproduct of ecologically based divergent natural selection, for example, if alleles favored by selection within each population are incompatible with one another when brought together in the genome of a hybrid. 4. GENETIC MECHANISMS LINKING SELECTION AND REPRODUCTIVE ISOLATION

The final component of ecological speciation is the genetic mechanism by which selection on ecological traits is transmitted to the genes causing reproductive isolation, thereby driving the evolution of the latter. There are two ways this can occur, distinguished by the relationship between the genes under divergent selection (i.e., those affecting ecological traits) and those causing reproductive isolation (figure 2C). In the first, these genes are the same (e.g., a gene affecting color pattern pleiotropically affects mate preference). In this case, reproductive isolation is said to evolve by direct selection because the alleles responsible for reproductive isolation are themselves under selection, albeit for another reason. In the second, the genes under divergent selection are physically different from those causing reproductive isolation. In this case, reproductive isolation is said to evolve by indirect selection because selection acts on genes causing reproductive isolation only to the extent that they are nonrandomly associated (i.e., in linkage disequilibrium) with the genes directly under selection. When selection acts on genes affecting ecological traits, such nonrandom associations will



cause a correlated evolutionary response in genes conferring reproductive isolation. The nature of these genetic associations is important because it affects the strength of selection transmitted to the genes affecting reproductive isolation. Pleiotropy and Direct Selection

Speciation is facilitated when genes under divergent selection cause reproductive isolation pleiotropically, and there are numerous ways this can occur. These include, for example, habitat isolation that evolves as a direct consequence of selection on genes affecting habitat choice. Selection might also act on ecological traits that incidentally affect mate preferences; the Drosophila lab experiments discussed previously suggest that this is not an unlikely occurrence, and the previously discussed mimetic color patterns in tropical butterflies of the genus Heliconius provide a classic example from nature. In plants, adaptation to different pollinators can cause premating isolation as a side effect. For example, work by Douglas Schemske and Toby Bradshaw has shown that divergent natural selection acts on a flower color gene in Mimulus monkeyflowers via the effects of color on attractiveness to pollinators. In Mimulus lewisii, pink-colored flowers are favored by bumblebees and discriminated against by hummingbirds. In contrast, M. cardinalis has red flowers, which are favored by hummingbirds and discriminated against by bumblebees. Adaptation to different pollinators via divergence in this flower color gene therefore directly affects the probability of cross-pollination (i.e., hybridization), a form of sexual isolation. Other forms of reproductive isolation could also involve pleiotropy. For example, temporal isolation, caused by differences in flowering time could arise as a pleiotropic effect of adaptation to different environments, whereas postmating isolation can arise pleiotropically if alleles favored by selection within each population contribute to genetic incompatibilities in hybrids. Linkage Disequilibrium and Indirect Selection

Indirect selection is less effective than direct selection in the evolution of reproductive isolation. This is because the genetic association between the two sets of genes (i.e., linkage disequilibrium) is not perfect, thereby reducing the strength of selection transmitted to the genes causing reproductive isolation (a phenomenon that has been likened to the slipping of a car’s clutch in that the wheels experience only a fraction of the power provided by the engine). The amount of linkage disequilibrium

that exists can be affected by three factors. The first is the genetic basis of reproductive isolation, of which there is an important distinction between what are termed one-allele and two-allele mechanisms. In a one-allele mechanism, reproductive isolation arises from the fixation of the same allele in both populations (e.g., an allele causing individuals to prefer mates phenotypically similar to themselves, for example, individuals similar in body size). In a two-allele mechanism, different alleles fix in each population (e.g., a preference allele for large individuals in one population and a preference allele for small individuals in the other). This distinction is important because, in a two-allele mechanism, recombination will tend to break down linkage disequilibrium between the genes under divergent selection and those causing reproductive isolation. In contrast, recombination creates no such problem for a one-allele mechanism, and it is therefore a more powerful mechanism of speciation. The prevalence of these genetic mechanisms in nature is unknown. The second factor is physical linkage. The maintenance of linkage disequilibrium is greatly facilitated by the physical linkage of genes on a chromosome because the likelihood of a recombination event declines with decreasing genetic map distance. Chromosomal inversions may thus play a role in ecological speciation by suppressing recombination, thus physically linking large regions of the genome. The third factor is the strength and form of selection. Linkage disequilibrium can be maintained by strong selection that favors specific combinations of genes (i.e., correlational selection), and such selection may be important during sympatric speciation. In general, data examining the relationship between genes under divergent selection and those causing reproductive isolation are sparse. In practice, separating pleiotropy from close physical linkage will be a difficult task, although their effects may ultimately be very similar. Important questions are how common pleiotropy and tight physical linkage are and how often they are of the form that would facilitate ecological speciation. Finally, we note that almost nothing is known about the types of genes involved in ecological speciation. Information on the genetics underlying ecological speciation may improve our mechanistic understanding of its operation in nature, including the type of genes involved and how they cause reproductive isolation. 5. GEOGRAPHY OF ECOLOGICAL SPECIATION

Traditionally, speciation has been classified not by the mechanisms responsible (e.g., ecological speciation) but rather by the geographic context under which it

Ecological Speciation Complete Reproductive isolation

occurs. These include allopatric, parapatric, and sympatric, with the latter being especially controversial and garnering much attention. Ecological speciation, however, can occur under any of these geographic contexts. The divergence of allopatric populations is unimpeded by the constraining effects of gene flow, so reproductive isolation is eventually expected to arise between them from chance events (e.g., genetic drift). Ecologically based divergent selection, however, can greatly accelerate this process and may commonly do so because allopatric pairs of populations often occupy different environments and are therefore subject to divergent selection. Sympatric speciation, in contrast, represents the opposite extreme in which speciation occurs in the absence of any geographic isolation. Strong disruptive selection is therefore required to overcome the hom*ogenizing effects of gene flow, and such selection is expected to be ecological in origin. Ecological speciation is therefore a likely mechanism of sympatric speciation. Parapatric speciation represents an intermediate scenario in which gene flow is reduced but not eliminated by geographic barriers (including distance). Divergent selection is again required to overcome the effects of gene flow, making ecological speciation a likely mechanism. Although attractive, the classification of speciation into these distinct geographic contexts may be overly simplistic and fail to capture the complexity of some speciation events in nature. Ecological speciation, for example, may often occur in stages that involve different geographic contexts (figure 3). The idea is that speciation begins when populations are allopatric, with reproductive isolation accumulating as a by-product of adaptation to their different environments. The second stage is initiated on secondary contact (parapatry or sympatry), with genetic exchange becoming possible at this point. Although the resulting gene flow is generally thought to constrain adaptive divergence and hamper speciation, ecological interactions are added as a source of divergent selection, and reinforcement also becomes possible. The amount of reproductive isolation that evolves during each stage indicates the primary geographic context of speciation, with the classic scenarios of allopatric and sympatric speciation representing the extremes in which, before secondary contact, reproductive isolation was either essentially complete or absent, respectively. Intermediate scenarios may be more common in nature, however, as suggested in the speciation of limnetic and benthic sticklebacks, which appears to have involved some reproductive isolation evolving during both phases. More complex scenarios are also possible, as suggested by recent molecular data from the apple and hawthorn races of Rhagoletis flies,


• Divergent selection between environments • Sexual selection


• Divergent selection between environments • Sexual selection • Ecological interactions • Reinforcement

Contact Allopatric stage

Time Sympatric/ parapatric stage

Figure 3. A scenario for ecological speciation under various geo graphic contexts. Reproductive isolation between two populations is absent at the beginning of the speciation process (at left) and evolves to completion (at right), as indicated by the solid line. The geographic context of speciation is indicated by the position of the dashed vertical line dividing the allopatric stage (left) from the sympatric/parapatric stage (right). This division can occur at any point in time, thus ac commodating a range of geographic contexts and either a one or two stage process. For example, the extreme case of fully sympatric speciation occurs when the allopatric stage is absent (division line coincides with the y axis). The opposite extreme of fully allopatric speciation occurs when reproductive isolation has evolved to com pletion before secondary contact occurs (division line falls at the right hand extreme). Depicted is an intermediate, two stage sce nario in which partial reproductive isolation evolves in allopatry, but reproductive isolation does not evolve to completion until after sec ondary contact has occurred. The ecological causes of divergent selection by which reproductive isolation may evolve are listed within the panel for each stage. The figure is modified from a review of ecological speciation by Rundle and Nosil and reprinted with the permission of the British Ecological Society.

traditionally put forward as a classic case of sympatric speciation. Feder and colleagues (2004) have shown that inversion polymorphisms, containing genetic variation affecting ecologically important diapause traits, trace their origins to allopatric populations in Mexico. 6. GENERALITY OF ECOLOGICAL SPECIATION

Comparative approaches can be used to investigate generalities of ecological speciation, including which, if any, forms of reproductive isolation are common, in what order they tend to arise, and the forms of divergent selection that drove their evolution. Likewise, the taxonomic generality of ecological speciation can also be explored using comparative approaches, as was done in a recent study by Daniel J. Funk, Patrik Nosil, and William B. Etges (2006). Using published data involving more than 500 species pairs of plants, invertebrates, and vertebrates, they found a positive



association between ecological divergence and reproductive isolation for seven of eight groups. By controlling for divergence time using published genetic data, their results suggest that ecological speciation may be taxonomically widespread. 7. REMAINING QUESTIONS IN THE STUDY OF ECOLOGICAL SPECIATION

There is convincing evidence from the laboratory and nature that divergent natural selection can drive the evolution of reproductive isolation. Ecological speciation therefore most certainly occurs. Current research is aimed at determining the ecological sources of divergent selection, the types of reproductive barriers involved, and the genetic mechanisms linking them. Attention is also being given to the generality of the findings associated with these components. There is much work remaining. Insufficient attention has been given to understanding the contribution of ecological interactions and sexual selection to the evolution of reproductive isolation. The relative importance of various forms of reproductive isolation and the likelihood that each evolves via divergent selection are also not well resolved. Direct tests of the genetic link between traits under selection and those conferring reproductive isolation are lacking too. Finally, the factors affecting the degree of progress toward the completion of ecological speciation are poorly understood. We hope that studies using a diversity of taxa and examining a wide range of divergence, from incipient to established species, will shed light on how ecological speciation unfolds from beginning to end.

FURTHER READING Bradshaw, Toby, and Douglas Schemske. 2003. Allele sub stitution at a flower colour locus produces a pollinator shift in monkeyflowers. Nature 426: 176 178. Coyne, Jerry, and H. Allen Orr. 2004. Speciation. Sunder land, MA: Sinauer Associates. Feder, Jeffery, Stewart Berlocher, Joe Roethele, Hattie Dambroski, James Smith, William Perry, Vesna Gavri lovic, Kenneth Filchak, Juan Rull, and Martin Aluja. 2003. Allopatric genetic origins for sympatric host plant shifts and race formation in Rhagoletis. Proceedings of the National Academy of Sciences U.S.A. 100: 10314 10319. Funk, Daniel, Patrik Nosil, and William Etges. 2006. Eco logical divergence exhibits consistently positive associa tions with reproductive isolation across disparate taxa. Proceedings of the National Academy of Sciences U.S.A. 103: 3209 3213. Jiggins, Chris, Russell Naisbit, Rebecca Coe, and James Mallet. 2001. Reproductive isolation caused by colour pattern mimicry. Nature 411: 302 305. Leal, M., and Leo J. Fleishman. 2004. Differences in visual signal design and detectability between allopatric popu lations of Anolis lizards. American Naturalist 163: 26 39. Mayr, Ernst. 1963. Animal Species and Evolution. Cam bridge, MA: Harvard University Press. Nosil, Patrik. 2007. Divergent host plant adaptation and reproductive isolation between ecotypes of Timema cris tinae. American Naturalist 169: 151 162. Rundle, Howard, Laura Nagel, Janette Boughman, and Dolph Schluter. 2000. Natural selection and parallel speciation in sympatric sticklebacks. Science 287: 306 308. Rundle, Howard, and Patrik Nosil. 2005. Ecological specia tion. Ecology Letters 8: 336 352. Schluter, Dolph. 2000. Ecology of Adaptive Radiation. Ox ford: Oxford University Press.

I.19 Adaptive Radiation Rosemary Gillespie OUTLINE

1. Conditions promoting adaptive radiation 2. Are certain taxa more likely to undergo adaptive radiation? 3. Examples of current adaptive radiations 4. Initiation of adaptive radiation 5. Speciation in adaptive radiation 6. Community assembly 7. Molecular basis for adaptive change 8. Testing adaptive radiation Adaptive radiation is generally triggered by the appearance of available niche space, which could result from (1) intrinsic factors or key innovations that allow an organism to exploit a novel resource, and/or (2) extrinsic factors, in which physical ecological space is created as a result of climatic changes or the appearance de novo of islands. There are no general rules as to what taxa are more likely to undergo adaptive radiation, although some lineages may have certain attributes that facilitate adaptive radiation in the appropriate setting. The process of adaptive radiation is described below, as well as some prime examples of the phenomenon. Adaptive radiation is generally initiated by expansion of ecological amplitude of a taxon into newly available ecological space, followed by specialization, the process possibly facilitated through adaptive plasticity. Speciation associated with adaptive radiation may involve one or more of the following: founder events, divergent natural selection, sexual selection, and hybridization. Competition is generally implicated in divergent natural selection and in dictating the communities of species formed during the course of adaptive radiation. Current research is focused on (1) examining the molecular underpinnings of apparently complex morphological and behavioral changes that occur during the course of adaptive radiation, and (2) experimental manipulation of bacteria to assess the conditions under which adaptive radiation occurs.

GLOSSARY adaptive radiation. Rapid diversification of an ancestral

species into several ecologically different species, associated with adaptive morphological, physiological, and/or behavioral divergence attenuation. Decline in number of species represented on islands with distance from a source of colonists divergent natural selection. Selection arising from environmental forces acting differentially on phenotypic traits (morphology, physiology, or behavior) resulting in divergent phenotypes; reproductive isolation may occur as a side effect, either in sympatry or allopatry ecological character displacement. Divergence in ecological traits (which may lead to reproductive isolation as a by-product) caused by competition for shared resources ecological release. Expansion of habitat or use of resources by populations into areas of lower species diversity with reduced interspecific competition ecological speciation. Process by which barriers to gene flow evolve between populations as a result of ecologically based divergent natural selection ecomorph. A group of populations, species, etc., whose appearance is determined by the environment escalation/diversification. Diversification of a herbivore/ parasite in concert with its host in which the adaptations of the host to counter exploitation by the herbivore or parasite build one on each other, and vice versa escape and radiation. Diversification of a herbivore/ parasite in concert with its host in which the host is generally considered to radiate before exploitation and subsequent radiation by the herbivore or parasite, and vice versa founder event. Establishment of a new population with few individuals that contain a small, and hence unrepresentative, portion of the genetic diversity



relative to the original population, potentially leading to speciation key innovation. Any newly acquired structure or property that permits the occupation of a new environment, or performance of a new function, which, in turn, opens a new adaptive zone nonadaptive radiation. Elevated rate of speciation in the absence of noticeable ecological shifts sexual selection. Form of natural selection based on an organism’s ability to mate such that individuals with attributes that allow them greater access to the opposite sex, either through (1) combat with the same sex or (2) attributes that render them more attractive to the opposite sex, mate at higher rates than those that lack these attributes taxon cycle. Temporal sequence of geographic distribution of species from (1) colonizing to (2) differentiating to (3) fragmenting and to (4) specializing Adaptive radiation is the rapid diversification of a lineage into multiple ecologically different species, generally associated with morphological or physiological divergence. The phenomenon can be characterized by four criteria: common ancestry, a phenotype– environment correlation, trait utility, and rapid speciation. The concept of adaptive radiation, and particularly diversification of ecological roles by means of natural selection, has had a long history, beginning with the observations of Charles Darwin on the Gala´pagos Islands, and has played a pivotal role in the development of the Modern Synthesis (see Givnish and Sytsma, 1997, for a detailed history of the concept). 1. CONDITIONS PROMOTING ADAPTIVE RADIATION

The most familiar present-day adaptive radiations are known from isolated archipelagoes or similar islandlike settings (e.g., lakes). However, it is quite likely that much of the diversity of life originated through episodes of adaptive radiation during periods when ecological space became available for diversification. Within this context there are two primary mechanisms through which ecological space can become available: (1) intrinsic changes in the organism often associated with key innovations, and (2) extrinsic effects, including environmental change and colonization of isolated landmasses. The two situations may be linked; for example, an intrinsic change may allow an organism to exploit a new environment. In either case, individuals exploiting the newly available niche space must be isolated to some extent from the remainder of the population to allow for genetic divergence associated with adaptive radiation. If a new habitat becomes available in close proximity to other such habitats (not isolated),

it will be colonized repeatedly by taxa from those habitats, and patterns of species diversity will be governed by ecological processes of immigration and extinction, rather than by evolutionary processes. Intrinsic Factors: Key Innovations

A key innovation is a trait, or a suite of traits, that allows an organism to exploit a novel resource or increase the efficiency with which a resource is used, thereby allowing a species to enter a ‘‘new’’ adaptive zone; the ecological opportunity thus provided may permit diversification. For example, the evolution of C4 photosynthesis, which enhanced rates of carbohydrate synthesis in open environments, likely served as a key innovation preceding radiation of most of the major lineages of grasses. Among recent adaptive radiations, one of the best-known key innovations is that of the pharyngeal jaw apparatus of cichlid fish (figure 1). Although most bony fishes have pharyngeal gill arches modified to process prey, the cichlid pharyngeal jaw is unique in having upper pharyngeal jaw joints, a ‘‘muscular sling,’’ and suturing that functionally fuses the lower pharyngeal jaw. The features of the jaw appear to have allowed cichlids to exploit a diversity of prey types, including large fish and hard-shelled prey that are unavailable to most other aquatic vertebrates. In flowering plants, floral spurs have evolved at least seven times, each time resulting in higher rates of diversification, perhaps the best known radiation being in the columbines (Aquilegia, Ranunculaceae). Interacting species, such as herbivores or parasites and their hosts, may themselves create ecological opportunity and provide some notable examples of key innovations. Here, the host may develop a defense to the herbivore or parasite (e.g., a plant may develop toxicity), but in due course the herbivore or parasite may develop resistance to the defense of the host. Paul Ehrlich and Peter Raven (1964) examined such coevolutionary responses and hypothesized that when plant lineages are temporarily freed from herbivore pressure via the origin of novel defenses, they enter a new adaptive zone in which they can undergo evolutionary radiation. However, if a mutation then arises in a group of insects that allows them to overcome these defenses and feed on the plants, the insect would then be free to diversify on the plants in the absence of competition. The major radiations of herbivorous insects and plants may have arisen through repeated steplike opening of novel adaptive zones that each has presented to the other over evolutionary time. Often referred to as escape and radiation, the host is generally considered to radiate before exploitation and subsequent radiation by the herbivore or parasite. This idea has been supported by

Adaptive Radiation A. a








more recent studies of insect diversification in the context of their host plants, with repeated evolution of angiosperm feeding in phytophagous beetles associated with an increased rate of diversification. There is a consistently greater diversity of beetles among plants in which latex or resin canals have evolved as protection against insect attack. In the same way, adaptive radiation of parasites may occur as a consequence of host switching to a new lineage of hosts. The development of a symbiotic association can also serve as a key innovation, providing a possible avenue through which taxonomic partners can enter into a new set of habitats unavailable to one or both of the symbiotic partners alone. For example, the development of gut endosymbionts and the concomitant ability to digest cellulose in ruminants appear to have led to the radiation of bovids in the African savannas. Likewise, the presence of algal endosymbionts has played a major role in the evolution and diversification of certain clades of Foraminifera.


Extrinsic Factors d



Figure 1. Adaptive radiation in cichlid fish. The pharyngeal jaw morphology, which allows dietary specialization, appears to have served as a key innovation in facilitating adaptive radiation in this group. (A) Biting and sucking species exhibit distinct morphologies. Labeotropheus fuelleborni (top) is a specialized biting species characterized by a short, robust lower jaw and an outer row of closely spaced tricuspid teeth. Metriaclima zebra (bottom) forages with a sucking mode of feeding and has a more elongate jaw and an outer series of larger bicuspid teeth. (B) Cichlids exhibit remark able evolutionary convergence. Similar ecomorphs have evolved repeatedly within different cichlid assemblages. All of the cichlids in the left hand column are from Lake Tanganyika. All of the cichlids in the right hand column are from Lake Malawi and are more closely related to one another than to any species within Lake Tanganyika. Note the similarities among color patterns and trophic morphologies. (From Albertson, R. C., and T. D. Kocher. 2006. Genetic and developmental basis of cichlid trophic diversity. Her edity 97: 211 221, 6800864a.html)

Rates of speciation are frequently accelerated with the physical appearance of new habitat (figure 2). In particular, changes in the temperature or humidity of the environment over various eras have repeatedly resulted in mass extinctions coupled with the opening of ecological space (see also figure 4). For example, the Cretaceous–Paleocene boundary event resulted in numerous extinctions of plants and insects and set the stage for the subsequent adaptive radiation of other groups. Environmental changes appear to form the basis of the Phanerozoic revolutions. Rising temperature and nutrient supplies as a result of submarine volcanism may have triggered later Mesozoic and perhaps early Paleozoic diversification episodes. Similar factors may underlie the iterative radiations of ammonoids through the geological record in which each radiation appears to have originated from a few taxa, which then went on to give rise to a wealth of morphological diversity. The emergence de novo of isolated habitats can be considered a kind of environmental change, as, for example, the formation of islands in the ocean, and many adaptive radiations are associated with such situations. For example, the adaptive radiation of finches in the Gala´pagos Islands appears to have been triggered by the appearance of land through volcanic activity over the last 3 million years. As the number of islands increased, so did the number of finch species. In the same way, adaptive radiations in the Hawaiian Islands are mostly associated with the volcanic activity that resulted in the formation of the current high islands that date back approximately 5 million years. Likewise, the




Autecology Mammals

Ornithischian dinosaurs



Saurischian dinosaurs


Figure 2. Open niches vacated by dinosaur extinctions at the end of the Cretaceous may have created empty ecological space and allowed mammals to radiate into these positions. Likewise, line ages that colonize isolated islands may give rise to adaptive radi ations because the colonists are free from competition with other species. (From Understanding Evolution, http://evolution.berkeley .edu/evolibrary/article/side o 0/adaptiveradiation 01)

evolution and adaptive radiation of the African cichlids appears to have been initiated in Lake Tanganyika approximately 5–6 million years ago when rivers in the area became progressively more swampy, with diversification of fish being initiated when river species became isolated in the deepening lake. In each of these cases, the new habitats were extremely isolated, resulting in infrequent colonization, thus giving the few successful colonists sufficient time to ‘‘explore’’ the ecological space available and diversify into multiple species. 2. ARE CERTAIN TAXA MORE LIKELY TO UNDERGO ADAPTIVE RADIATION?

Whether some species are predisposed to undergo adaptive radiation because of a broad environmental tolerance, generalized feeding patterns, or perhaps some proclivity to develop novel associations has been the subject of much debate. For example, birds have undergone extensive adaptive radiation in the insular Pacific, whereas butterflies have not. This has led some authors to suggest that speciation in butterflies may be constrained by the mechanics of insect–plant coevolution preventing rapid diversification in the insects. However, this argument is not well supported, as other insects with similar coevolutionary ties have undergone some of the most spectacular insular adaptive radiations known. It appears that, given conditions of isolation and time, almost any group of organisms is capable of undergoing adaptive radiation given ecological opportunity that it can exploit. However, certain groups do appear to be predisposed to adaptive radiation. For example, the occurrence of parallel radiations of sister clades of plants, the Hawaiian silverswords and California tarweeds, suggests that this lineage has certain

attributes that facilitate adaptive radiation in the appropriate settings. 3. EXAMPLES OF CURRENT ADAPTIVE RADIATIONS

Adaptive radiation has now been documented in every kingdom of life and in a large number of phyla and in many different circ*mstances. However, the original concepts were developed through studies on islands, as these systems are discrete and amenable for studying the basis of adaptive radiation. The earliest groups to be examined in this context were vertebrates, in particular birds, with perhaps the best-known example being that of the Gala´pagos finches (figure 3) initially described by John Gould, and used by Charles Darwin as a key demonstration of his theory of evolution by natural selection. Currently, there are 13 recognized species in three lineages: ground finches, tree finches, and warbler-like finches, with sympatric species occupying distinct ecological roles. Research by Peter and Rosemary Grant and colleagues has shown that the finches have considerable genetic variation within populations, which is intermittently subject to both natural and sexual selection, with the final community of finches on an island dictated by food resources and interspecific competition for these resources. In the Hawaiian Islands, the endemic honeycreepers, which show even more extraordinary morphological and ecological differentiation than the Gala´pagos finches, comprise 56 species in a single lineage. However, only 22 species of honeycreeper are currently extant, with others known only from historical or fossil collections, making it difficult to develop hypotheses regarding processes underlying the adaptive radiation. Other well-known vertebrate radiations include lizards (anoles) in the Greater Antilles of the Caribbean in which diversification has allowed species to occupy a range of ecological roles, with as many as 11 species occurring sympatrically. Different species live, for example, on twigs, in the grass, or on tree trunks near the ground. Jonathan Losos and colleagues recognize six types of habitat specialists on the basis of morphological measurements (see plate 4). Among frogs, a remarkable example of adaptive radiation has recently been found in Madagascar. Among mammals, the best-known adaptive radiation is also from Madagascar, where lemurs constitute a spectacular diversification of more than 65 species, although at least 15 of these are now extinct. In addition, some striking radiations of small mammals have been documented, including the rodents on the islands of the Philippines and bats in southeast Asia. Additional spectacular examples of adaptive radiation in vertebrates come from lacustrine fish, with the best known being cichlids (mentioned above), which


Adaptive Radiation Geospiza difficilis

iza osp G e dens n sca

Pin aro ino loxia s rna ta

G co eos n i r pi z os a tri s

a i de rth ea Ce livac o


iza Geosp fortis






Cama rhync pallid hus us







Geosp magn iza rostris

hu rhync Cama ates hellob













za pi os osa Ge igin l fu

Ca m a pa rhy r v nc ulu hu s s


Pla cra tyspiz ssir ost a ris

Camarhynchus psittacula

u nch rhy er a Cam paup


Figure 3. Adaptive radiation in Darwin’s finches. Diagram illus trating the morphological and associated ecological diversity among the radiation of Darwin’s finches in the genus Geospiza (Emberezidae). The 14 species evolved from a common ancestor

about 3 million years ago. (From Grant, P. R., and B. R. Grant. 2008. How and Why Species Multiply: The Radiation of Darwin’s Finches. Princeton, NJ: Princeton University Press)

reach their highest diversity in the East African lakes of Victoria (over 400 species), Malawi (300–500 species), and Tanganyika (approximately 200 species). In each of these lakes, the fish exhibit diversity in trophic morphology, including specialist algal scrapers, planktivores, insectivores, piscivores, paedophages, snail crushers, and fin biters. In addition to their trophic diversity, they display a striking array of color patterns, which appear to be involved in courtship and recognition. Other fish radiations include darters of the

Central Highlands of eastern North America and threespine sticklebacks in deglaciated lakes in Canada. The latter in particular have now been heralded as an outstanding example of recent and ongoing adaptive radiation, within which it is possible to study processes involved. No more than two species occur in any one lake, but pairs of species in different lakes appear to have evolved independently of other pairs as a result of parallel bouts of selection for alternate trophic environments.



Among invertebrates, there are now a number of examples that show adaptive radiation associated with historic climatic change (figure 4). One of the best known ongoing radiations in insects is that of the Hawaiian Drosophila flies, in which courtship behavior can be very elaborate; the males of the so-called picture winged species often have ornately patterned wings as well as unusual modifications of the mouthparts and legs. At the same time, clades are characterized according to whether they breed on fungi, leaves, fruit, or bark, suggesting a role of both sexual selection and ecological shifts in allowing diversification of these flies. Hawaiian swordtail crickets have become increasingly recognized over recent years as a striking example of a very rapid island radiation. In common with other crickets, courting males ‘‘sing’’ to attract females by rubbing their forewings together. Each species has a unique song, and females respond preferentially to the song of the same species. Differentiation appears to occur through sexual selection on genetically well-structured populations. Among spiders, several radiations have been described in the Hawaiian Islands, one of the largest being in the genus Tetragnatha, where ecomorphs have arisen independently in much the same way as Caribbean lizards. In other parts of the Pacific, the land snail genus Partula is particularly well known for its radiation on different islands in the South Pacific. Like other land snails, they are highly polymorphic with respect to the color, banding, and chirality of the shell; competition appears to be important in dictating the array of species at a site. The Canary Islands are also well known for adaptive radiations of insects (in particular, beetles and psyllids) and spiders. Here, although most groups show evidence of competition in shaping communities, there is little biogeographic congruence between groups; stochasticity in species arrival patterns plays a prominent role in dictating species composition on any one island. The Hawaiian silversword alliance (Asteraceae) has been considered a prime example of adaptive radiation in plants (plate 3). It consists of 28 species, which exhibit a large diversity of life forms, including trees, shrubs, mat-plants, monocarpic and polycarpic rosette plants, cushion plants, and vines, that occur across a broad environmental spectrum, from rainforests to desert-like settings. Additional plant radiations include columbines (mentioned above) in North America and numerous other island radiations, such as Argyranthemum in the Canary Islands and Psiadia in the Mascarenes, both of which are in the same family as the Hawaiian silverswords (Asteraceae) and have diversified in a parallel fashion. One of the most recently described adaptive radiations is that of the soil bacterium Pseudomonas fluor-

escens, which, over a short period of time, can develop from an isogenic population under conditions of environmental heterogeneity into several somewhat predictable ecomorphs. Accordingly, this system has been hailed as one within which it is possible to conduct experimental studies of adaptive radiation. 4. INITIATION OF ADAPTIVE RADIATION

Although there have been many studies that describe different adaptive radiations, the initiation of the process remains poorly understood. Yet it is possible to recognize some general patterns. Dispersal and Colonization

Because adaptive radiation requires colonization and differentiation in an ecologically available space, the taxa that colonize must necessarily be few. Although there appears to be a substantial random element to colonization, successful colonization of very isolated locations generally requires high dispersal abilities. Accordingly, representation of taxa within biotas in isolated areas will be skewed toward those that can disperse readily. For example, as one ventures farther into the Pacific Ocean from west to east (i.e., toward greater isolation), there is an attenuation in the number of lineages of terrestrial groups that have colonized by over-water dispersal. In less isolated archipelagoes, such as Fiji, the fauna is relatively rich with numerous continental affinities. Farther east, Samoa is less rich at higher taxonomic levels than Fiji but still has many families and orders that are lacking from the native fauna of more remote Polynesian islands. East of Samoa, the number of floral and faunal groups that have been able to reach the remote islands diminishes, and here the few colonists have frequently undergone adaptive radiation, accentuating the unrepresentative nature of the biota. The Hawaiian archipelago (4000 km from the nearest continent, North America; 3200 km from nearest island group) demonstrates this pattern most acutely: among insects, the native fauna is represented by only 50% of insect orders and 15% of known families and exhibits extraordinarily high levels of endemism (95–99% in invertebrates) with numerous cases of adaptive radiation. High dispersal of colonists clearly contrasts with the apparently much restricted ranges and dispersal abilities that are frequently associated with members within an adaptive radiation. In general, there appears to be a dramatic loss of dispersal ability and/or attainment of a more specialized habitat at the outset of adaptive radiation. Indeed, a tendency toward reduced dispersal following colonization of new ecological space


10 Changes Cephaloleia championi Baly 141

Cephaloleia fenestrata Weise 013

Cephaloleia lata Baly 067

Cephaloleia costaricensis Uhmann 023

B. Onset of Paleocene -Eocene global warming ~59 Ma

Bridging of the Isthmus of Panama ~3.5 Ma Panama Arc collides with NW South America ~12.8–7.1 Ma

12 Temperature (°C)

Cumulative number of lineages

K-P boundary event ~65.5 Ma 90


8 4 0

1 Upper Paleocene Cretaceous 60



Oligocene Cenozoic

40 30 20 Millions of years before present


Plio. Plt.


Figure 4. Tempo in the adaptive radiation of Cephaloleia (Coleoptera: Chrysomelidae). (A) Chronogram showing significant diversification rate shifts (each terminal corre sponds to a species) and source(s) of support. Shading indicates the timeframe over which significant diversification rate shifts occurred. (B) Semilogarithmic plot of lineages through time (LTT) for Cephaloleia (LTT plot; heavy upper line) super imposed on a time averaged record of high latitude sea surface tem peratures (lower irregularly de scending line), a proxy for global climate. The constant diversification rate model is rejected if the empir ical LTT curve falls outside the 95% confidence interval generated by simulation (middle ascending lines). Upturns or downturns in the em pirical LTT plot reflect changes in rates of speciation or extinction. (From McKenna, Duane D., and Brian D. Farrell. 2006. Tropical for ests are both evolutionary cradles and museums of leaf beetle diver sity. Proceedings of the National Academy of Sciences U.S.A. 103: 10947 10951)



may be an important factor in allowing diversification to proceed. Ecological Release and Specialization

When a taxon first colonizes a new area, it frequently expands its ecological range, a phenomenon referred to as ecological release. Regular cycles of ecological and distributional change following colonization of islands are well known—the phenomenon was described first in detail for Melanesian ants and subsequently for Caribbean birds. The idea is that widespread, dispersive populations or species (Stage I) give rise to differentiating (Stage II) and then fragmented (Stage III) and ultimately specialized endemic species (Stage IV). This pattern is consistent with the early stages of adaptive radiation. However, once local endemics have formed, there appears to be little evidence to suggest that species become progressively more specialized over time. Phylogenetic analysis of some radiations suggests that specialized species, when they colonize new habitat, may be able to expand their range and accordingly may reinitiate the ‘‘cycle’’ to give rise to other specialist species. Behavioral Plasticity

Although it has been suggested that behavioral and ecological plasticity may impede adaptive diversification as it allows a single taxon to exploit a broad environmental range, recent research suggests otherwise. Mary Jane West-Eberhard (2003) has argued that adaptive plasticity (including behavior) promotes evolutionary diversification, in particular when the environment is variable. Recent theoretical work has supported this idea, and plasticity has now been suggested as playing a role in a number of radiations. For example, in Caribbean anoles, plasticity may allow species to occupy a new habitat in which they otherwise might not be able to survive. Once in these habitats, their behavior may become modified, and, as new mutations arise, selection may act to accentuate the initial, relatively minor, morphological changes. Conclusive demonstration that phenotypic plasticity precedes, and then permits, subsequent evolution has been difficult to obtain because once the ancestral populations have evolved, they may lose the pattern of plasticity present at the start of adaptive differentiation. In the threespine stickleback, however, where ancestral oceanic sticklebacks likely have changed little since colonization and diversification of freshwater species, the pattern of behavioral plasticity in the ancestral species supports the argument that phenotypic plasticity can guide subsequent evolutionary change and facilitate adaptive radiation.


Rapid speciation coupled with phenotypic diversification are key features of adaptive radiation, and accordingly, many studies have examined the basis for genetic diversification. Because adaptive radiation involves ecological shifts, divergent natural selection has been most broadly implicated across a spectrum of lineages, with founder effects and sexual selection also playing a role in certain situations. However, the mechanism through which species form during adaptive radiation is still only very poorly understood, although it appears that geographic barriers to gene flow are generally involved at least in the initiation of speciation. Founder Events

A founder event refers to the establishment of a new population by a few individuals that carry only a small sample of the genetic diversity of the parent population. Many studies have suggested that founder events can play a role in adaptive radiation, as taxa within a radiation often have very small population sizes, with ample opportunity for isolation. Because of the effects of random sampling, a founder event will lead to differences in allele frequencies at some loci as compared to the parent population. However, considerable debate has focused on the nature of genetic changes that occur subsequent to the founder event, during the period of population growth, with some traditional arguments suggesting that founder events trigger rapid species formation, although more recent studies have largely refuted a role for founder events in reproductive isolation. Genetic drift can lead to changes during the bottleneck, but the effect becomes weaker as the population starts to grow. At the same time, a large proportion of alleles are lost during a bottleneck, and few new mutations can occur while the population size is small. The effect of these opposing forces is that the number of beneficial mutations fixed per generation remains largely unchanged by the bottleneck. Nevertheless, selection subsequent to a genetic bottleneck can preserve alleles that are initially rare and that would otherwise tend to be lost as a result of stochastic events in populations of constant size. Divergent Natural Selection

As with other factors considered to play a key role in adaptive radiation, competition plays a dual role. On the one hand, reduced competition is generally associated with the presence of open resource niches, either as organisms move into new habitats or through the acquisition of a key innovation, thereby providing in-

Adaptive Radiation


creased opportunities for diversification. On the other hand, competition, often with ecological character displacement, is frequently implicated in promoting adaptive change between close relatives. However, there are no known direct tests that link the evolution of reproductive isolation to interspecific competition, and the role of competition in the early stages of adaptive radiation remains unclear. Nevertheless, divergent natural selection driven by interspecific competition appears to have shaped current phenotypic differences in many radiations ranging from sticklebacks in the Canadian lakes to lizards on the islands of the Caribbean, finches in the Gala´pagos, and spiders in the Hawaiian Islands. Predation has also been suggested as a possible operative that may work together with (or instead of) competition to allow adaptive differentiation, but this has been difficult to test. However, the specific role of predation in facilitating adaptive radiation has recently been demonstrated in both walking-stick insects and bacteria. Sexual selection (see below) has been shown in some cases to act in concert with divergent natural selection. For example, sexual dimorphism in Caribbean anole lizards allows a species to exploit different niches, thereby serving as an alternative means of ecological diversification.

of a shift in the distribution of mating preferences. In Hawaiian Laupala crickets, closely related species are morphologically similar with no ecologically recognizable features and distinguishable only by the pulse rate of the male courtship song, a secondary sexual trait used in mate attraction. These crickets demonstrate the highest rate of speciation recorded so far in arthropods. Other taxa in which extreme diversification has been attributed in part to sexual selection include haplochromine cichlids, in which sexually dimorphic breeding coloration, with brightly colored males and often dull females, appears to have arisen through female mate choice for male coloration. Disruptive sexual selection on male coloration can result in genetic isolation between fish that exhibit small differences in male coloration and female preference for male coloration. Likewise, diversification in jumping spiders in the sky islands of the western United States appears to be the product of female preference for greater signal complexity or novelty. In all of these radiations, ecological differentiation still occurs, but sexual selection may act somewhat independently and accelerate the rate of differentiation. Mate choice is the primary isolating mechanism, and hybrids between rapidly diverging sibling species are often fully viable and fertile.

Hybridization and Gene Flow


Gene flow among populations in the process of diverging, or hybridization between incipient species, may slow the process of diversification and hom*ogenize populations. However, interspecific hybridization may also be a possible source of additional genetic variation within species: Hybridization increases the size of the gene pool on which selection may act and therefore may be a significant process in the adaptive radiation of some species. Such effects have now been shown in Darwin’s finches, African cichlids, Lake Baikal sculpins, and several lineages of Hawaiian plants.

During the course of adaptive radiation, speciation plays a role similar to that of immigration, although over an extended time period, in adding species to a community. It appears that when a lineage diversifies in a community, it may adapt to multiple different ecological settings, with the development of specific sets of attributes, or ecomorphs, to match a given microhabitat. In an archipelago situation, where similar habitats occur on different islands, it has been found that similar sets of ecomorphs can arise independently within the same lineage through convergent evolution, a phenomenon first demonstrated in anole lizards in the Caribbean (plate 4) and now shown in a wide range of other adaptive radiations including Himalayan birds, Gala´pagos finches, cichlid fish of the African rift lakes, Canadian lake sticklebacks, ranid frogs in Madagascar, spiders in the Hawaiian islands, and snails in the Bonin Islands of Japan. These results point to a model of ecological community assembly that incorporates evolutionary effects of interspecific competition. The central importance of competition in shaping communities has recently been challenged by Stephen Hubbell’s ‘‘neutral theory,’’ which postulates that differences between members of a community of ecologically equivalent species are ‘‘neutral’’ with respect to

Sexual Selection

Sexual selection has been implicated in the diversification of species within some of the most explosive adaptive radiations. In particular, it has been suggested that sexual selection drives species proliferation in Hawaiian Drosophila flies and Laupala crickets. The mechanism for this, as proposed by Kenneth Kaneshiro and colleagues for Drosophila, is that, when a newly founded population is small, female discrimination is relaxed; accordingly, sexual behavior becomes simpler with more intraspecific variability. Divergence of sibling species may then occur during isolation as a result



their success. One outcome of neutral theory has been to prompt investigation of when, and to what extent, ecological equivalence might play a role. In the course of adaptive radiation, neutral processes may govern the identity of taxa initially colonizing a new area, potentially resulting in a transient period with multiple ecological equivalents. Likewise, communities formed through sexual selection may include some ecologically similar species. However, the end product of subsequent ecological speciation appears inevitably to be a set of taxa that are ecologically distinct.

predation, and time in dictating when and how adaptive radiation occurs. These studies, although still in their infancy, have shown many parallels between adaptive diversification of bacterial genotypes over the space of a few weeks and the presumed early stages of adaptive radiation of macroorganisms over millions of years. The challenge is to apply the knowledge gained from these rich bacterial systems to a more general understanding of adaptive radiation. FURTHER READING


That repeated evolution of similar forms has occurred in many species undergoing adaptive radiation has led to considerable research on the molecular underpinnings of such apparently complex changes. With the development of molecular techniques, we now have the opportunity to analyze the genetic architecture of species differences and the role of new mutations versus standing genetic variation in adaptation and divergence in adaptive radiation. For example, in fish, it appears that gene duplication provided a genomic mechanism for adaptive radiation of teleosts, with lineages arising after duplication being much more species-rich than the more basal groups. Studies on the body plating in a radiation of sticklebacks have shown that related alleles are responsible for a given phenotype, indicating that even though a given phenotype (ecomorph) may have multiple independent origins, the same ancestral allele appears to be involved in producing the same phenotypes. In groups in which sexual selection has been implicated in diversification, investigation of rapidly evolving genes for traits associated with sex and reproduction have been the target of research on the process of speciation. For Drosophila, results show differential patterns of evolution of genes expressed in reproductive and nonreproductive tissues, supporting the role of sexual selection as a driving force of genetic change between species. 8. TESTING ADAPTIVE RADIATION

Experimental tests of the processes underlying adaptive radiation are now possible using microbial systems to probe the relative roles of niche space, competition,

Benton, Michael J., and Brent C. Emerson. 2007. How did life become so diverse? The dynamics of diversification according to the fossil record and molecular phylogenet ics. Palaeontology 50: 23 40. Carlquist, Sherwin, Bruce G. Baldwin, and Gerald D. Carr, eds. 2003. Tarweeds and Silverswords: Evolution of the Madiinae (Asteraceae). St. Louis: Missouri Botanical Garden Press. Ehrlich, Paul R., and Peter H. Raven. 1964. Butterflies and plants: A study in coevolution. Evolution 18: 586 608. Gillespie, R. G. and D.A. Clague, eds. 2009. Encyclopedia of Islands. Berkeley: University of California Press. Givnish, T.J.K., and K. J. Sytsma, eds. 1997. Molecular evolution and adaptive radiation. New York: Cambridge University Press. Grant, Peter R., and B. Rosemary Grant. 2007. How and Why Species Multiply: The Radiation of Darwin’s Finches. Princeton, NJ: Princeton University Press. Jessup, C. M., R. Kassen, S. E. Forde, B. Kerr, A. Buckling, P. B. Rainey and B.J.M. Bohannan. 2004. Big questions, small worlds: Microbial model systems in ecology. Trends in Ecology and Evolution 19: 189 197. Losos, Jonathan B. 2001. Evolution: A lizard’s tale. Scientific American 284: 64 69. Ricklefs, Robert E., and Eldredge Bermingham. 2007. The causes of evolutionary radiations in archipelagoes: Pas serine birds in the Lesser Antilles. American Naturalist 169: 285 297. Schluter, Dolph. 2000. The Ecology of Adaptive Radiation. Oxford: Oxford University Press. Seehausen, Ole. 2006. African cichlid fish: A model system in adaptive radiation research. Proceedings of the Royal Society B Biological Sciences 273: 1987 1998. Wagner, Warren L., and Vicki A. Funk, eds. 1995. Hawaiian Biogeography: Evolution in a Hotspot Archipelago. Washington, DC: Smithsonian Institution Press. West Eberhard, Mary Jane. 2003. Developmental Plasticity and Evolution. New York: Oxford University Press.

II Population Ecology H. Charles J. Godfray Understanding what determines the average abundance of species, why their numbers fluctuate, and how they interact with each other is a major part of modern ecology often united under the term population ecology. Of course, the boundaries of population ecology are ill-defined and porous: on the one hand the field grades into physiological ecology—how individuals interact with the environment—and on the other hand into community ecology—the study of large assemblages of species. Population ecology is part of the larger subject of population biology that encompasses both the evolutionary and the ecological processes affecting populations. The human race has always been concerned with the abundance and fluctuations of the plants and animals that share its environment, not least because they provide its food. But the modern study of populations begins with Thomas Malthus (1798), who, in his Essay on the Principle of Population, realized that if birth and death rates remain constant with the former greater than the latter, then population size will grow geometrically until some extrinsic factor comes into play. The conclusions that Malthus, an upper-class English vicar, drew from his insights were of the importance of doing something about the ‘‘irresponsibly fecund lower orders’’ (as well as the need to attend to other ‘‘problems’’ such as ‘‘liberal women’’ and the French!). Fortunately, Malthus is not remembered as a politician, but his writing hugely influenced the first generation of biologists to think about animal populations, and in particular Charles Darwin, who realized that geometric population growth implied massive mortality and hence a huge advantage to any heritable trait that helped individuals in the struggle for survival. Today we use the Malthusian parameter, the population’s rate of geometric growth assuming demographic parameters remain the same, as an index of the state of the population. A closely related parameter, the growth rate of a rare mutation, is intimately connected to notions of evolutionary fitness. Calculating population growth rates (population projection) is quite

straightforward for some species, for example, those with discrete generations. It can be much more complicated when there are overlapping generations and where the population is composed of individuals of different classes (differing in age, size, or other variable), issues discussed in chapter II.1. But demographic rates do not remain the same forever, and in particular, as population densities increase, birthrates go down or death rates go up. It is these density-dependent effects that are critical in determining the typical range of abundance of different organisms, as discussed in chapter II.2. Densitydependent effects may increase smoothly as population size gets larger but may also be much more capricious, cutting in only above a threshold, the latter itself possibly varying from year to year. The chief factor determining observed population densities at any particular time is often a density-independent process such as the weather, and the densities of some populations may fluctuate in a random way for many generations before they become large enough for densitydependent processes to come into play. However, no population can be regulated, that is, persist indefinitely within certain bounds, without density dependence occurring. Where density-dependent processes act instantaneously and increase gently with population size, the outcome of population regulation will be a stable equilibrium (though in nature random perturbations will mean that an absolutely constant population density is unlikely to be observed). But if there is a time lag between population increase and the impact of density dependence, or if density dependence is very strong, then overcompensation may occur, and the population will show cycles. As was first realized by ecological theoreticians, particularly by Robert May, in the 1970s, stronger density dependence and larger time lags may lead to population fluctuations that are chaotic—purely deterministic yet impossible to predict in detail. Hastings (chapter II.3) explores these issues and discusses recent findings about how deterministic


Population Ecology

population dynamics and random environmental effects may produce a complex array of fascinating dynamics. Animal and plant populations do not exist in one place but occur in a normally complex spatial landscape. The first generation of ecological studies tended to ignore this spatial component, but their considerations are now central to much theoretical and practical ecology and are discussed in several of the chapters in this section. One particularly fruitful line of inquiry is discussed in greater depth by Hanski (chapter II.4). Consider a species that inhabits a constellation of habitat patches: some may be empty, and others may be occupied by subpopulations with substantially independent dynamics, which may become extinct or send out colonizers to form new populations. This is a metapopulation, and the theoretical study of their dynamics, coupled with some superb long-term field studies, in particular by Hanski himself, has greatly enriched the subject and proved very important in applied ecology, especially in conservation biology. Turning from single populations to pairs and small collections, a number of authors explore the different ways in which species may interact. This is important in its own right but also as the building blocks from which communities and ecosystems are composed. Populations can flourish only if they have resources for growth and reproduction, and interspecific resource competition is a potent force that is thought to structure many communities. We have two entries on competition: chapter II.5 focuses on plants and chapter II.6 on animals. Although there are commonalities, the facts that plants are (literally) rooted to the ground and compete for a relatively small range of essential quantities (space, light, water, nutrients) has meant that plant competition ecology has developed rather differently from its zoological cousin. Competition is sometimes referred to as a ‘‘ ’’ interaction because each species involved suffers from the presence of the other. In a ‘‘þ ’’ interaction, one species gains at the expense of the other. We have four chapters on such interactions. Chapter II.7 discusses predator–prey dynamics including issues such as the circ*mstances under which predator–prey cycles may occur and what processes may tame the intrinsic tendency for instability in such interactions. Such concerns are also explored in chapter II.8, which considers parasitoid–host dynamics. Parasitoids are insects, typically wasps and flies, with a life history somewhat intermediary between predators and parasites. They

are numerically abundant and important mortality factors affecting a wide variety of insect hosts and are also very important as biological agents. Chiefly because of this last reason their population dynamics has been closely studied by ecologists. Both the dynamics of true parasites and the dynamics of pathogens are discussed in chapter II.9. Although the importance of disease for humans and their livestock and crops has long been realized, it is only relatively recently that the importance of disease in natural ecosystems has been appreciated, a shift in focus that has been greatly helped by the molecular biology revolution that has furnished new tools for studying microbial pathogens. Interestingly, mathematical tools and techniques developed by ecologists are now used routinely by epidemiologists studying human, farm animal, and crop pathogens. The final ‘‘þ ’’ interaction chapter is by Morris (chapter II.10) on plant–herbivore interactions. The major difference here is that plants are frequently able to tolerate varying degrees of herbivory in a way that, say, an individual zebra cannot tolerate predation by a lion. Morris explores the consequences of these differences and asks the degree to which herbivory may influence plant population dynamics. Mutualism and symbiosis, ‘‘þ þ’’ interactions, are rather the Cinderella of this type of population ecology and have received far less attention than competition and consumer–resource interactions. In chapter II.11, Bronstein discusses why ecologists may have underestimated their importance, drawing examples from some wonderful model systems, in particular involving plants and their insect pollinators. Many partners in mutualisms and symbioses are microorganisms, and we also include a chapter discussing some of the special issues concerning microbial population ecology. Chapter II.12 explores how all the types of ecological processes that affect larger organisms also operate at this smaller scale but shows how factors such as the size of microbial populations and their particular growth dynamics require special treatment. It also illustrates how microbial systems, with their very short generation times, provide fabulous models systems for exploring processes that affect all living organisms. Finally, Thompson (chapter II.13) explores how ecological interactions such as competition, predation, and mutualism result in evolutionary pressures that are experienced by both the interacting species. The resultant coevolution has shaped the morphology, biochemistry and behavior of many, perhaps most, organisms on Earth, including ourselves.

II.1 Age-Structured and Stage-Structured Population Dynamics Mark Rees and Stephen P. Ellner



1. Age-structured models: Life tables and the Leslie matrix 2. Stage-structured matrix models 3. Integral projection models 4. Continuous-time models with age structure 5. Applications and extensions 6. Coda

age structure. Distribution of ages in a population matrix. A rectangular array of symbols, which could

When all individuals in a population are identical, we can characterize the population just by counting the number of individuals. However, the individuals within many animal and plant populations differ in important ways that influence their current and future prospects of survival and reproduction. For example, larger individuals typically have greater chances of survival, produce more and sometimes larger offspring, and often have slower growth rates. In such cases, characterizing the population structure—the numbers of individuals of each different type—is critical for understanding how the population will change through time. In this chapter, we examine some of the main types of models used for describing and forecasting the dynamics of structured populations. Age-structured models in discrete time, appropriate for populations in seasonal environments, were developed centuries ago by the great mathematician Leonhard Euler (1707–1783). These are considered first, before moving to models where individuals are characterized by their stage in the life cycle (e.g., seed versus flowering plant, larva versus adult). Next we look at how to incorporate differences among individuals that vary continuously, such as size. Having explored discrete-time models, we briefly turn to continuoustime models and then present some applications and extensions.

The simplest age-structured models assume that each individual’s chance of survival and reproduction depends only on its age; there are no effects of population density. The standard model counts only females (assuming no shortage of mates) and assumes that all births occur in a single birth pulse immediately before the population is censused (a so-called postbreeding census). The population dynamics is then summarized by the following equations:

represent numbers, variables, or functions


n0 (t þ 1) ¼ f0 n0 (t) þ f1 n1 (t) þ f2 n2 (t) þ ¼

A X a

na (t þ 1) ¼ pa

fa na (t)

0 1 na

1 (t),


where na(t) is the number of individuals of age a at time t, fa, and pa are, respectively, the average fecundity and the probability of survival to age a þ 1 of age a individuals, and A is the maximum possible age (or the maximum age at which reproduction occurs, if postreproductives are omitted from the population count). Because births occur just before the next census, fa ¼ pa ma þ 1 , where ma þ 1 is the number of offspring produced by an age a þ 1 female.


Population Ecology

Another way of formulating the model is to assume a prebreeding census, so the population is censused immediately before the birth pulse. This has two important consequences: (1) all individuals are at least age 1, and (2) in this case fa ¼ p0 ma , so fecundity depends on the number of offspring produced now, ma, and the chance that they survive to be censused at age 1, p0. The simple age-structured model can be written as a matrix, commonly known as a Leslie matrix after British ecologist P. H. Leslie. Expressing equation 1 in matrix form simply means putting the fs and ps in the right places: 3 2 f0 n0 (t þ 1) 6 n1 (t þ 1) 7 6 p0 6 7 6 6 7 60 .. 6 7¼6 . 6 7 6 .. 4 5 4 . nA (t þ 1) 0 2

f1 0 p1 .. .

f2 0 0 .. .



32 3 fA n0 (t) 6 7 07 76 n1 (t) 7 7 6 . 0 76 . 7 7 .. 76 . 7 5 4 5 . nA (t) 0


n(t þ 1) ¼ Ln(t),


where L is the matrix is equation 2. When L is a matrix with n columns and n(t) a column vector of length n, then Ln(t) is a vector whose ith element is [Ln(t)]i ¼

n X j

Lij n(t)j ,



where Lij is the number in the ith row and jth column of L. Matrix multiplication expresses equation 1 as a single operation; it also means that the tools of linear algebra can be used to study how the population varies through time. Now that we have formulated the model, how does it behave? To answer this question we need to solve equation 3. Starting with some initial age distribution, n(0), we find:

n0 (t þ 1) ¼

A X a

n(2) ¼ Ln(1) ¼ L[Ln(0)] ¼ L2 n(0), and so on; so the general solution is n(t) ¼ Lt n(0):


It is difficult to intuit what Lt is doing, but some insight is gained by solving the model numerically. For an example, setting

fa la n0 (t a):


Assuming that n0 grows exponentially at some rate l, substituting n0 (t) ¼ clt into equation 7 and simplifying gives the famous Euler-Lotka equation, 1¼

A X a


(a þ 1)

la fa ¼

A þ1 X a



la ma :



This equation shows how the long-term population growth rate l is determined by the age-specific survival and mortality. Critically, when l > 1 the population increases, and when l < 1 it decreases. Consequently, l is of great importance in applied contexts: for control of pest species we would like to make l < 1, whereas for species of conservation interest we would like to ensure that l > 1. As l gets larger, the right-hand side of equation 8 gets smaller; using this fact and substituting l ¼ 1 into equation 8, we find that l > 1 if and only if R0 > 1, where R0 ¼

A X a

n(1) ¼ Ln(0)


and iterating the model (figure 1) shows that after some initial fluctuations, the total population size grows at a constant rate, and the proportion of individuals in each age class becomes constant. This suggests that, rather remarkably, an age-structured population will behave very much like a simple unstructured population undergoing exponential growth, in which the numbers at time t are given by n(t) ¼ n(0)lt . This fact allows us to derive one of the most important equations in agestructured dynamics. We know that na (t) ¼ la n0 (t a), where la is the probability an individual survives to age a (l0 ¼ 1, la ¼ p0 p1 p2 . . . pa 1 ). Substituting this into the first line of equation 1 gives

(2) or, more compactly,

3 5 05 0

0 3 L ¼ 4 0:2 0 0 0:5

la f a :

R0 is the average number of offspring produced by a newborn female over her lifetime, given by summing the chance of surviving to each age times the number of offspring produced at that age. Thus, the population will increase (l > 1) in the long run only if each female more than replaces herself. Understanding the model further requires some results from matrix algebra. Eigenvalues turn out to be key quantities and are defined as follows: l is an eigenvalue of L if there is a nonzero vector w such that

Population Dynamics B. 40 Number of age 0 individuals (log scale)

Total population size (log scale)












2 1













D. 0.8



2.0 N(t+1)/N(t)

Proportion age 0


0.4 0.2

1.5 1.0

0.5 0








20 Time

Figure 1. Numerical solution of the Leslie matrix model (equation 6) assuming n(0) ¼ (1, 1, 1). The panels show time series plots of (A) total population size N(t), (B) the number of newborn individuals, (C)

the proportion of individuals of each possible age (black ¼ 0, dark gray ¼ 1, and light gray ¼ 2), and (D) the population growth rate N (tþ1)/ N (t) .

Lw ¼ lw, and w is called the corresponding eigenvector. If there are A þ 1 distinct eigenvalues, then the corresponding eigenvectors are linearly independent, and so any population vector can be expressed as n(0) ¼ c0 w0 þ c1 w1 þ c2 w2 þ þ cA wA . Then we can rewrite equation 3 as

So as t becomes large, the solution will be determined by the largest-magnitude li, termed the dominant eigenvalue and its eigenvector; hence,

n(1) ¼ Ln(0) ¼ L(c0 w0 þ c1 w1 þ c2 w2 þ þ cA wA ) ¼ c0 Lw0 þ c1 Lw1 þ c2 Lw2 þ þ cA LwA ¼ c0 l0 w0 þ c1 l1 w1 þ c2 l2 w2 þ þ cA lA wA , so moving one year forward corresponds to multiplying the coefficients ci by the corresponding li. Thus, the model solution (equation 3) can be written as n(t) ¼

A X i

ci lti wi :


n(t)&c1 l1 t w1 ,


where & means approximately with a relative error decreasing to zero as t becomes large. This explains the numerical results (figure 1) that after an initial period of transients, the population grows at a constant rate, given by l1, and the proportion of individuals in each age class becomes constant and is proportional to w1. For this reason w1 is called the stable age distribution. The existence of a dominant eigenvalue is guaranteed so long as L is power positive: some power Lm has all entries greater than zero. L will be power positive providing fA > 0 and any two consecutive fs are posi2 tive, or more generally if all the entries of LA þ 1 are positive. For a nonnegative L that is power positive, the Perron-Frobenius Theorem implies that L has a


Population Ecology

unique dominant eigenvalue that is real, positive, and strictly larger in magnitude than any other eigenvalue, guaranteeing convergence to the stable age distribution and stable growth rate l1. When the matrix lacks power positivity, we can get more exotic behavior. For example, consider a population where all the reproduction is concentrated in the last age class, such as 2 3 0 0 11 L ¼ 4 0:2 0 0 5: (11) 0 0:5 0 In this case, the age structure continually cycles with a cycle length of 3, and population size never settles into growing at a constant rate (figure 2). These population waves were initially explored by Harro Bernardelli in relation to oscillations in the age structure of the Burmese population. For a matrix like that in equation 11, with all reproduction in the final age class, each individual of age a at time t gives rise, after m time steps (where m is the number of age classes) to R0 age-a individuals at time t þ m, and R0 is the product of the nonzero matrix entries. Consequently, any initial age structure gives rise to a cycle of age structures that repeats indefinitely with period A, and the long-term 1=m population growth rate is R0 . 2. STAGE-STRUCTURED MATRIX MODELS

In a stage-structured model, individuals are divided into discrete categories conventionally called ‘‘stages’’ or ‘‘stage classes.’’ These sometimes represent discrete stages in the life history, say eggs, larvae, pupae, and adults of an insect species, but very commonly stages are categories imposed on a continuously varying trait such as size. For example, a plant population might be characterized by small, medium, and large individuals, and all between-stage transitions may be possible as a result of growth and shrinkage. Despite this, many of the ideas developed for age-structured populations carry over. In place of the Leslie matrix, reproduction and the movements of surviving individuals between stages are governed by a population projection or Lefkovitch matrix, M. The dynamics are then given by n(t þ 1) ¼ Mn(t):


The Perron-Frobenius Theorem still applies provided that M is power positive, so the long-term growth rate is given by the dominant eigenvalue, l1, of M, and the stable stage distribution by the corresponding eigenvector, w1.

To give a concrete example, here is the (slightly rounded) projection matrix used by Katriona Shea and David Kelly (1998) to explore the dynamics of Carduus nutans, an invasive thistle:


SB SB 0:04 M ¼ S 0:19 M 0 L 0

S M L 8:25 179:41 503:14 1:09 22:18 62:18 0:01 0 0 0:01 0:02 0

# :


SB is the number of seeds in the seedbank, and S, M, L refer to thistle rosettes that are small, medium, and large in size. The matrix has the following simple interpretation: each column gives the expected contribution of a particular stage to each of the other stages. So the first column says that 4% of the seeds in the seedbank will stay there, and 19% will become small rosettes; the second column says that each small rosette will give rise to 8.25 seeds in the seedbank, 1.09 small rosettes, and a small number of medium and large rosettes, and so on. Constructing the matrix M for a real population requires selecting appropriate stages. If the life cycle is divided into discrete stages, this is straightforward. Otherwise things become more complicated, as it is necessary to (1) decide on the appropriate measure of individual state and (2) set the boundaries between stages. Practical issues of data collection and the ability to predict an individual’s fate may determine how to measure an individual’s state. Typically a single variable is used (e.g., longest leaf length or rosette diameter as a measure of plant size), but more complex classifications, say by age and size, are also possible. Setting boundaries may be problematic. Ideally there should be many categories, so all individuals within a category really behave in a similar way, as the model assumes. However, the more categories there are, the fewer observations there are on each category, so estimates of the elements of M become less reliable. Integral projection models, discussed in the next section, provide an elegant way around these problems. An enormous amount of work has been done analyzing the properties of projection matrices and using those properties to study real populations, much of it summarized in the landmark monograph by Hal Caswell (Caswell, 2001; first edition 1989). For example, we can use elasticity analysis to explore how fractional changes in matrix elements affect the long-term population growth rate l1. Specifically, defining eij ¼

fractional change in l1 @l1 =l1 ¼ , fractional change in mij @mij =mij


Population Dynamics B. 50 25

Mean total population size (log scale)

Total population size (log scale)



















D. 4

1.0 0.8

3 N(t+1)/N(t)

Proportion age i


0.6 0.4

2 1


0 0



30 Time





30 Time



Figure 2. Numerical solution of the Leslie matrix model (equation 11) assuming n(0) ¼ (1, 1, 1) in each panel; we have a time series plot of (A) total population size N(t), (B) total population size averaged over the

cycle, (C) the proportion of individuals in each age class (black ¼ 0, dark gray ¼ 1, and light gray ¼ 2), and (D) the population growth rate N(tþ1)/ N(t).

it can be shown that

Matrix models can be generalized in many ways, such as by adding density dependence and/or stochastic variation from one time step to the next. Exploring all these would require an entire large book, which, fortunately, Caswell (2001) has already written.

eij ¼

mij vi wj , l1 vw


where vw is the dot-product (vw ¼ v1 w1 þ v2 w2 þ vn wn ), and v is the left eigenvector of M (vM ¼ lM). For the thistle matrix (equation 13), the elasticities are


SB e¼ S M L

SB 0:004 0:247 0 0

S 0:198 0:308 0:044 0:075


M L 0:017 0:031 0:025 0:045 : 0 0 0:001 0


These results suggest that the transitions SB ! S, S ! SB, and S ! S are critical for population growth, and therefore, management strategies should focus on reducing these transitions. The unintuitive prediction that it will be far more effective to concentrate on small plants rather than large ones is made apparent only by computing the elasticities.


Plant and many other types of organisms do not just come in small, medium, and large sizes. For example, consider Platte thistle, with individual size measured by the root crown diameter (figure 3). If individuals are divided into three size classes (indicated by the vertical lines in figure 3), then from the fitted curves, it is clear that some categories contain very different individuals; for example, individuals in the ‘‘large’’ category have probabilities of flowering that vary systematically from &0.2 to over 0.8. A matrix projection model with three size classes ignores these differences and treats all ‘‘large’’ individuals as identical. To avoid this problem, in 2000 Michael R. Easterling, Stephen P. Ellner, and Philip M. Dixon proposed


Population Ecology B. 3.0

2.0 1.5 1.0 0.5 0

0.6 0.4 0.2 0

0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 Root crown diameter year t mm (log scale) C.

1.0 0.8


Probability of survival

Root crown diameter year t +1 mm (log scale)


0.5 1.0 1.5 2.0 2.5 3.0 3.5 Root crown diameter mm (log scale)

0.5 1.0 1.5 2.0 2.5 3.0 3.5 Root crown diameter mm (log scale)



800 Seed production

0.8 0.6 0.4

600 400 200


0 0

0.5 1.0 1.5 2.0 2.5 3.0 3.5 Root crown diameter mm (log scale)

Figure 3. Size structured demographic rates for Platte thistle, Cirsium canescens. (A) Growth (as characterized by plant size in successive years), (B) survival , (C) the probability of flowering, and (D) seed production all vary continuously with size and can be de scribed by simple regression models. (Redrawn from Rose et al.,

2005) In panels B and C, the data were divided into 20 equal sized categories, and the plotted points are fractions within each cate gory, but the logistic regression models (plotted as curves) were fitted to the binary values (e.g., flowering or not flowering) for each individual.

the integral projection model (IPM) in which individuals are characterized by a continuous variable x such as size. The state of the population given by n(x,t), such thatRthe number of individuals with sizes between a and b b is a n(x, t)dx. Instead of the matrix M, the IPM has a projection kernel K(y,x), so that

matrix model, and so the results described above carry over. Constructing the projection kernel K(y,x) is straightforward using the regressions shown in figure 3. For an individual of size x to become size y, it must (1) grow from x to y, (2) survive, and (3) not flower (flowering is fatal in monocarpic plants like Platte thistle). These probabilities are calculated from the fitted relationships in figures 3A, 3B, and 3C, respectively. The use of regression models to construct the projection kernel brings some advantages: (1) accepted statistical approaches can be used for selecting an appropriate regression model; and (2) additional variables characterizing individuals’ states can be included by adding explanatory variables rather than having to select a single best state variable. For example, in some thistles the probability of flowering depends on both an

n(y, t þ 1) ¼


K(y, x)n(x, t)dx,


where s and S are the minimum and maximum possible sizes. The integration is the continuous version of equation 4, adding up all the contributions to size y at time t þ 1 by individuals of size x at time t. Providing some technical conditions are met (see Ellner and Rees, 2006, for details), the IPM behaves essentially like a

Population Dynamics individual’s size and age and is often described by a logistic regression such as logit pf (a, x) ¼ exp (b0 þ bs x þ ba a). Extending a size-structured model to include age-dependent flowering therefore requires the estimation of a single additional parameter rather than estimation of many age- and size-class-specific flowering probabilities in the analogous matrix model. 4. CONTINUOUS-TIME MODELS WITH AGE STRUCTURE

The simplest starting point is the continuous-time analog of the Leslie matrix, in which vital rates depend only on individual age a, ignoring effects of population density and environmental factors. The state of the population (as usual counting only females) is then characterized by n(a,t), so that as in IPMs Z


n(s, t)ds ¼ Number of individuals of age a to b:


(16) The dynamics of n(a,t) are generated by the age-specific per capita birthrate b(a) and death rate m(a). To be age a at time t, an individual must have been age a dt at time t dt and have not died; that is, n(a, t) ¼ n(a dt, t dt)[1 m(a dt)dt]: Rearranging and letting dt ! 0, we obtain the McKendrick–von Foerster equation @n @n þ ¼ m(a)n(a, t), @t @a


which describes the dynamics of n(a,t) for a > 0. The boundary condition, describing the birth of new individuals, is Z1 n(0, t) ¼ b(a)n(a, t)da:


As with the previous models we expect exponential solutions, so n(a, t)&Cn*(a)ert . By arguments analogous to those for the discrete-time age-structured model, the long-term instantaneous growth rate r can be shown to satisfy the continuous Euler-Lotka equation Z1 1¼ e 0





where l(a) is the chance of surviving to age a given by Z a l(a) ¼ exp m(s)ds : o

The fate of the population then depends on r, with increasing populations having r > 0 and decreasing ones r < 0. So not surprisingly r, like l, plays a key role in population management and life-history theory. Just as in the discrete case, we can compute expected lifetime fecundity R0 ¼



so r > 0 if and only if R0 > 1, as expected. Age has the unforgiving property that one year on, you will always be one year older, and this property was exploited when studying age-structured models. In a size-structured model, we must specify how size changes over time. It is then possible, by looking at the flows of individuals into and out of some small size range, to derive the dynamics of the size structure. If individuals grow deterministically, growth can be described by an equation dm=dt ¼ g(m). This leads to the McKendrick–von Foerster equation for size-structured dynamics (although it was almost surely known to Euler as the equation for passive particles carried by a moving fluid): @n @(gn) þ ¼ m(m)n(m, t): @t @m


Specifying appropriate boundary conditions is less straightforward for equation 20 than for equation 17. If we assume that individuals are size m0 at birth, then equation 20 applies for m > m0 , and the boundary condition is n(m0 , t) ¼

1 g(m0 )

Z1 b(m)n(m, t)dm;



the prefactor before the integral is needed to convert the birthrate (individuals per unit time) into the resulting contribution to the size distribution (individuals per increment of size). Much as for age-structured models, the size-structured model has a long-term size structure and population growth rate, which can be derived using the (nonlinear) relationship between age and size entailed by the deterministic growth pattern. The basic model can be extended in many ways, such as allowing a random component to growth (including a


Population Ecology

chance of shrinkage), variable offspring size, reproduction by fission, and characterization of individuals by multiple measures of size (e.g., lipid and nonlipid body mass) or by size and age. Extending the basic models to include density dependence and interspecific interactions is difficult, and indeed, William S. C. Gurney and Roger Nisbet (1998) described continuous-time models in which individuals are distinguishable by age and size as ‘‘a traditional source of mathematical headaches.’’ To make some progress, Gurney, Nisbet, and John Lawton (1983) suggested grouping individuals into stages and assuming constant vital rates within each stage. However, unlike the stage-structured matrix model, individuals within a stage may have different states. For example, even if all juveniles have the same growth rate, younger juveniles may be smaller and less likely to mature soon into adults. To see what this means, consider Gurney, Nisbet, and Lawton’s model for laboratory blowfly populations based on the classic experimental studies of A. J. Nicholson. The model assumes two stages, Juvenile and Adult. 1. Ages 0 to t are Juveniles with a constant per capita mortality rate m(a, t) ¼ mJ and birthrate b(a, t) ¼ 0. 2. Ages t and above are Adults with a constant per capita mortality rate m(a, t) ¼ mA and per capita birthrate b(a, t) ¼ qe cNA (t) , where NA (t) is adult population size, so the birthrate is density dependent. These assumptions imply a set of differential equations describing the dynamics, dNJ =dt ¼ RJ (t) RA (t) mJ NJ dNA =dt ¼ RA (t) mA NA ,


where RJ (t) and RA (t) are the recruitment rates into the Juvenile and Adult stages. By definition, RJ (t) ¼ qNA (t)e cNA (t) . Because the Juvenile stage lasts exactly t time units, RA (t) equals the recruitment into the Juvenile stage t t time units ago times the survival through the Juvenile stage SJ ¼ e tmJ ; hence, RA (t) ¼ SJ RJ (t t). Substituting RJ (t) into RA (t) gives a single equation for the dynamics dNA =dt ¼ SJ NA (t t)e

cNA (t


mA NA :


The key simplifying assumption in this model is that all Juveniles have the same demographic rates. Juveniles do differ in their state though: some are nearly

mature and will soon become adults, whereas others are recently born and will not become mature for some time. Although the final model involves only the total numbers in each class, its structure reflects the fact that newborns all wait t time units before maturing into Adults. The presence of the time delay t is the price for allowing individuals within a stage to differ in state. In this model, stages correspond exactly to a range of ages, but similar models can be constructed in which stages correspond to a range of sizes, or in which there is no exact correspondence of stages to age or size; rather, each individual within a stage has a probability (potentially depending on age, size, etc.) of making the transition to another stage. These models may require state variables in addition to the population counts for each stage to track the within-stage state dynamics and its consequences.


Here we discuss some applications of the ideas presented in the previous sections: a stochastic densitydependent Leslie matrix model for an Asiatic wild ass; stage-structured models used to understand the dynamics of laboratory populations; and use of integral projection models to make evolutionary predictions of life-history strategies.

Climate Variability and Persistence of Asiatic Wild Ass

David Saltz, Daniel I. Rubinstein, and Gary C. White (2006) developed a Leslie matrix-type model for the population of an Asiatic wild ass (Equus hemionus) reintroduced into the Makhtesh Ramon Nature Reserve in Israel, based on long-term monitoring (1985– 1999). Their main goal was to explore possible effects of increased rainfall variability on the population’s risk of extinction because increased variability is predicted under some global climate change scenarios even in areas where no changes in mean rainfall are predicted. Their model includes a number of important extensions to the basic Leslie matrix model described above. Saltz et al. used their data to fit a model predicting an adult (age 3) female’s chance of successful reproduction as a function of total rainfall in the current and previous years. Their model also included a negative effect of adult female density, but reproductive success was unrelated to age. Age-specific survivorship was based on a published survivorship curve for zebra, with additional mortality of 30% or higher during drought years (rainfall < 40 mm) based on data for other ungulates.

Population Dynamics

leading to competition. To apply the stage-structured model, Gurney et al. (1983) estimated the stage-specific mortalities, fecundities, and durations:

Probability of extinction

1.0 Global change “future” conditions No change Global change “past” conditions


0.6 0.4


0 0

0.1 0.2 0.3 Reduction in survival (%)



Figure 4. Probabilities of extinction of an Asiatic wild ass population under various climate change scenarios.

Because the model was explicitly linked to variation in rainfall, model simulations could be based on the 41 years of rainfall data for the study area. In particular, simulations to assess extinction risk over a 100-year time period were run by bootstrapping from either the first 20 years of rainfall data (when variance was lower), the second 21 years (when variance was higher), or the complete data set. They also incorporated demographic stochasticity: for example, rather than having 30% of adults die in a drought year, they did a simulated ‘‘coin toss’’ to determine whether each individual lived or died. The strong effect of drought years on survival and reproduction produced a strong impact of rainfall variability on population persistence. At the low-end estimate of drought-induced mortality (30%), the increase in variance between the first and second halves of the rainfall data produced a more then fivefold increase in the probability of population extinction within 100 years (figure 4). Nicholson’s Blowflies: Continuous-Time Stage Structured Models and Density-Dependent Leslie Matrices

Gurney et al. (1983) used Nicholson’s data to estimate the parameters for the model in equation 23. Nicholson conducted a series of long-term experiments, using sheep blowflies, designed to explore the effects of resource limitation at different life stages. The blowfly has four distinct life stages—eggs, larvae, pupae, and adults—but feeds only in the larval and adult stages. In the experiments considered by Gurney et al., larvae were given unlimited resources, whereas the adults received protein (in the form of ground liver) at a fixed rate,

Using the duration of each stage and the stagespecific mortality to estimate the egg-to-adult delay time as t&15:6 days and egg-to-adult survival SJ &0:91. Estimating the egg-production rate by combining data on egg production versus food supply with the assumption that food is divided evenly among adults, to get b(NA )&8:5e NA =600 . Using the rate of decline in adult population when no recruitment is occurring to estimate the adult mortality rate mA &0:27=day.

With these estimates, the model produces sustained cycles with a period of about 37 days (compared to an average observed period of about 38 days), and adult population varying between a minimum of 150 and a maximum of 5400 (compared to observed minima and maxima of 270 120 and 7500 500; figure 5). This is remarkable given that no model parameters were adjusted to fit the experimental time series, and perhaps even more remarkably the model solutions exhibit the ‘‘double peak’’ that usually occurred in the data. The population cycles occur because egg production is overcompensating, and the period of the cycles is determined by t; analysis of the model suggests the period will be in the range (2t, 4t), in good agreement with the numerical solution. To simulate the model without the difficulties of solving delay-differential equations, it can be expressed as an age-structured model, similar to equation 2. Because the maturation time is 15.6 days, it is convenient to use time and age increments of 0.1 days. The instantaneous mortality and fecundity rates in the continuous-time model can be converted into rates per 0.1 days. For example, if juveniles in the discrete-time model become mature adults when they exit the 15.6day-old age class, the juvenile survival probability in 1=157 the discrete-time model is SJ per time increment. Then 157 age classes are needed for the juveniles, but only one for the adults, giving a Leslie matrix that is large (of size 158158) and density dependent but straightforward to implement on a computer. Integral Projection Models for Plants: Linking Evolution and Ecology

Population dynamics and evolution are intimately linked because the fate of new genetic mutants depends on their ability to spread in a population. Coupling evolutionary ideas with demographic models for the


Population Ecology A.

B. 6000 5000 Adults, eggs/day

Figure 5. (A) Experimental time series of blowfly adult (gray) and egg (black) dynamics (from Nicholson, 1954) and (B) predicted dynamics from the simple stage structured model (equation 23).

Adults, eggs/day





4000 3000 2000 1000

0 0

50 100 150 200 250 300 350 Time (days)

growth of the mutant subpopulation thus allows predictions of how natural selection shapes individual behaviors and life histories. The key evolutionary idea is John Maynard Smith’s concept of an evolutionary stable strategy (ESS): a strategy that cannot be displaced by a rare mutant if it has become fixed in a population. In a population at demographic equilibrium (l ¼ 1), the established strategy cannot be displaced if l is less than 1 for a rare mutant strategy with some other strategy. In this way, ESSs in real populations can be identified. As an illustration, an integral projection model for Oenothera biennis (evening primrose) can be used to predict the size dependence of flowering probability (figure 3C). This relationship is determined by balancing the benefits of flowering at a large size (increased seed production, figure 3D) against the mortality costs of growing large. Using published data, Rees and Rose (2002) produced a fully parameterized IPM for this species. The probability of flowering was size dependent and described by a logistic regression, logitpf (x) ¼ b0 þ bs x, where x is log rosette diameter,

50 100 150 200 250 300 350 Time (days)

b0 and bs and the fitted intercept and slope. Making b0 smaller reduces the probability of flowering for all sizes and so increases the mean size at flowering. In Oenothera, density dependence acts only at the recruitment stage, so the ESS is characterized by maximizing R0, the total reproductive output of an individual that survives through the recruitment stage (Mylius and Diekmann, 1995). Numerical evaluation of R0 as a function of b0 shows that estimated value is very close to the predicted ESS (figure 6). This example illustrates how structured population models, coupled with evolutionary ideas, provide a general framework for understanding the evolution of organisms’ life cycles subject to trade-offs and constraints, a vast subject known as life-history theory. 6. CODA

All populations are structured: by age, size, genotype, social status, and so on. Structured population models have arguably become the core theoretical framework for population ecology, and a modern course on


B. 1.0











Figure 6. Relationships between b0 and (A) R0 and (B) l for the Oenothera IPM. The ESS is marked with a dot, and the estimated b0 is indicated by the vertical line.

0 -40




0 -40




Population Dynamics population ecology would be in large part a course on structured population modeling. The scope of theory and applications vastly exceeds the space available here. Read on!

FURTHER READING Caswell, Hal. 2001, Matrix Population Models: Construction, Analysis and Interpretation. Sunderland, MA: Sinauer As sociates (1st edition 1989). This volume is the classic and comprehensive monograph on matrix models for struc tured populations, clear, authoritative, and amusing. This book and the ones below by Metz and Diekmann, Tulja purkar, and Tuljapurkar and Caswell are essential reading for the structured population modeler. Ellner, Stephen P., and Mark Rees. 2006. Integral projection models for species with complex demography. American Naturalist 167: 410 428. This article generalized the size structured IPM introduced by Easterling, Ellner, and Dixon by allowing individuals to be cross classified by several traits. The article includes model construction, elasticity analysis, stable distribution theory for density independent models, evolutionary optimality criteria, and local stability analysis for density dependent models. Gurney, William S. C., and Roger M. Nisbet. 1998. Ecolo gical Dynamics. Oxford: Oxford University Press. This volume is a very readable text covering a wide range of structured population models, with many case studies that illustrate the concepts and processes involved in con structing structured models. Gurney, W.S.C., R. M. Nisbet, and John. H. Lawton. 1983. The systematic formulation of tractable single species population models incorporating age structure. Journal of Animal Ecology 52: 479 495. This paper introduced con tinuous time models with discrete stage structure and showed how they could explain qualitative differences be tween the dynamics of two laboratory insect populations. Metz, Johannes A. J., and Odo Diekmann, eds. 1986. The Dynamics of Physiologically Structured Populations. Ber lin: Springer. This is the volume that moved continuous time size structured models and their relatives into the mainstream of population modeling, with a mix of mathematical theory and applications. Often mathemati cally challenging; the article by de Roos in Tuljapurkar and


Caswell (1997) provides a ‘‘gentle introduction’’ that may be useful to read first. Murdoch, William M., Cheryl J. Briggs, and Roger M. Nis bet. 2003. Consumer Resource Dynamics. Princeton, NJ: Princeton University Press. This definitive volume presents systematic development and real world applications of stage structured models in the style of Gurney Nisbet Lawton for predator prey and host parasitoid population interactions, summarizing the fruits from two decades of focused effort. If we could all work like this, ecology would be much the better for it. Mylius, Sido D., and Odo Diekmann. 1995. On evolutionarily stable life histories, optimization and the need to be specific about density dependence. Oikos 74: 218 224. This rep resents an elegant theoretical paper analyzing the properties that characterize ESSs in structured populations. Tuljapurkar, Shripad. 1990. Population Dynamics in Vari able Environments. New York: Springer. Although sadly now out of print, this volume is an essential reference on stochastic matrix models by the author of many funda mental articles on this topic. Tuljapurkar, S., and H. Caswell, eds. 1997. Structured Population Models in Marine, Terrestrial, and Freshwater Systems. New York: Chapman & Hall. This book provi des a wealth of applications, and some accessible reviews of basic theory, derived from a summer course on struc tured population models at Cornell University.

Applications discussed in this chapter Rees, M., and K. E. Rose. 2002. Evolution of flowering strategies in Oenothera glazioviana: An integral projec tion model approach. Proceedings of the Royal Society 269: 1509 1515. Rose, K. E., S. M. Louda, and M. Rees. 2005. Demographic and evolutionary impacts of native and invasive insect herbivores: A case study with Platte thistle, Cirsium ca nescens. Ecology 86: 453 465. Saltz, David, Daniel I. Rubenstein, and Gary C. White. 2006. The impact of increased environmental stochasticity due to climate change on the dynamics of Asiatic wild ass. Conservation Biology 20: 1402 1409. Shea, K., and D. Kelly. 1998. Estimating biocontrol agent impact with matrix models: Carduus nutans in New Zealand. Ecological Applications 8: 824 832.

II.2 Density Dependence and Single-Species Population Dynamics Anthony R. Ives


1. Three questions about the dynamics of single species 2. Density dependence 3. Endogenous population variability 4. Exogenous population variability 5. Returning to the three questions In ecology, population dynamics refers to how populations of a species change through time. The study of single-species population dynamics encompasses three general questions: (1) What explains the average abundance of a population? (2) What explains the fluctuations in abundance of a population through time? and (3) How do average abundances and fluctuations in abundance vary among populations in different geographic locations? Any of these questions can be asked of any population of any species, yet some populations pose particularly interesting challenges for one or more of the questions. Thus, ecologists often focus on populations that are remarkably large (pests) or small (endangered species), that have dramatic fluctuations through time, or that vary markedly from one location to another.

GLOSSARY dependence. Density-dependent population growth occurs when the per capita population growth rate changes as the population density changes. Because it changes with population density, density-dependent growth is not exponential. dynamics. The dynamics of a population consists of the changes through time in the population size or a related measure such as density. endogenous variability. Endogenous population variability is driven by density-dependent factors that involve interactions among individuals in the system density

specified by a researcher. The system could consist of a single population or populations of interacting species. exogenous variability. Exogenous population variability is driven by factors outside the system that are not themselves influenced by population fluctuations within the system. Examples include not only environmental factors such as weather but also the abundances of other species if the dynamics of these species is not affected by the focal species within the system. exponential population growth and decline. When the per capita population growth rate remains constant, the population experiences exponential growth or decline. Exponential population growth can also occur when the per capita population growth rate varies through time provided its average remains constant. intrinsic rate of increase. The intrinsic rate of increase is the maximum per capita population growth rate for a population with a stable age structure (i.e., the proportions of the population in different age groups remain the same). The intrinsic rate of increase is often achieved when the population is at low density. per capita population growth rate. The per capita population growth rate is the rate at which a population changes per individual in the population. It is often expressed as the natural logarithm of the ratio of population densities at consecutive sample times, log e x(t þ 1)=x(t). population. A population is a group of individuals of the same species occupying a specified geographic area over a specified period of time. The area may be ecologically relevant (an island) or irrelevant (political districts), and the boundaries may be porous, with individuals immigrating to and emigrating from the population.

Density Dependence stability. Stability is defined in many ways in ecology.

In models of population dynamics, stability is generally used in two ways. First, when there is no environmental stochasticity, stability describes how populations change when they are around points or cycles. A stable point, for example, is one in which, if the population density is initially near the point, it will move generally closer to the point through time. Second, when there is environmental stochasticity, a more stable system is one in which population variability is small for a given level of environmental variability in the per capita population growth rate. There are additional ways that stability can be defined in model and real systems, which necessitate care in using the word stability. stochasticity. Stochasticity is random (unpredictable) variability that is described by a probability distribution giving the mean, variance, and other properties of the random process. 1. THREE QUESTIONS ABOUT THE DYNAMICS OF SINGLE SPECIES

The three broad questions about single-species population dynamics boil down to: What explains the abundance of a population and changes in abundance through time and space? As an example of the first question, we could ask why, unfortunately, there are more mosquitoes than moose in Wisconsin. The answer might seem simple; moose are so much bigger than mosquitoes, they simply occupy more space and need more food. The question becomes more difficult, however, when asking why the roughly 60 species of mos-


quitoes in Wisconsin differ hugely in abundance. One species, Aedes vexans, is many times more common than most other species. Is this because A. vexans is more flexible in its breeding requirements and capable of breeding in more diverse habitats than other species? Is it because the females have more catholic tastes for the hosts that unwillingly give up a blood meal that the females convert to eggs? Or is it because A. vexans somehow is more adept at avoiding the many predators that turn mosquitoes into lunch? This set of questions poses a real challenge to ecologists, and a challenge of possible practical importance. Along with being common, A. vexans females also include humans in their range of suitable hosts, and if we understood why it is so common, we might also be able to change this situation. The second question is best illustrated with a figure showing two example populations (figure 1). The first, the moose population on Isle Royale in Lake Superior, fluctuates over the 45 years of data, showing a peak of 2500 individuals followed by a crash in the late 1990s to 500 individuals. This fivefold variation, however, is small in comparison to the 500,000-fold variation in the abundance of midges in Lake Myvatn, Iceland. The root causes of the fluctuations of both populations are the same: a combination of depleted food resources (balsam fir trees for moose, algae for midges) and predation (by wolves on moose and a variety of species on midges). Despite having the same general causes, however, why are the fluctuations in midges so much more dramatic? Finally, an example of the third question is posed by the pattern of population dynamics shown by many small rodent species such as voles. Many populations at high latitudes show strong fluctuations, often exceeding

B. 2.5

1,000 Population abundance (thousands)

Population abundance (thousands)








0 1960


1980 Year



Figure 1. Population abundances of (A) moose on Isle Royale ( and (B) the midge Tanytarsus graci



1990 Year



lentus in Lake Myvatn, Iceland (Ives et al., 2008). Note the log10 scale in B.


Population Ecology

two orders of magnitude, whereas populations of the same species fluctuate much less dramatically at lower latitudes. A tempting explanation for this pattern is simply that populations at higher latitudes have to contend with a more severe climate, in particular hard winters, and the severe climate drives greater population fluctuations. This answer cannot be the sole explanation, however, because the fluctuations in rodent populations at high latitudes do not match the fluctuations in weather conditions. In fact, some peaks in rodent abundances occur in winter instead of summer. Ecologists suspect that predators are involved in the high population fluctuations of rodents at high latitudes and that the importance of these predators for some reason diminishes for populations closer to the equator. This chapter is about single-species population dynamics, so a discussion of what is a single-species population is in order. In ecology, what is meant by a population is often given by the context of the discussion or study. In the examples above, the moose population on Isle Royale is clearly delineated; it is the number of individuals on the island, and Lake Superior gives a clear, ecologically relevant boundary through which moose do not easily pass (despite being excellent swimmers). In other cases, the boundaries might be clear but ecologically irrelevant. For example, the population of moose in Wisconsin is delineated by political, not ecological boundaries. For issues of population management, political boundaries make sense, but not so for ecological concerns. Furthermore, the boundaries are porous, with moose obliviously crossing from Michigan to Wisconsin and back again. Nonetheless, the moose population of Wisconsin can still be delineated and counted, making it clear what the population is. Obviously, single-species populations consist of a single species, although no species lives on an island unto itself. Population dynamics of any species will be affected by other species—species that it consumes, species that consume it, species that compete, and species that might in some way help (such as pollinators helping plants). Although ecologists recognize the importance of interactions among species, very often studies are conducted on a single focal species, with other species considered only to the extent that they affect the dynamics of the focal species. In this way, other species are treated somewhat like the weather. Often, the reason for focusing on a single species is simply pragmatic. If there is a single species that is of applied or academic interest, it is necessary to limit the ecological extent of the investigation to what can sensibly be studied with the resources available. The other topic of this chapter is density dependence, and explaining this requires an entire section.


The easiest way to explain density dependence is to consider when it is absent. In particular, consider European rabbits. Rabbits have remarkable reproductive proclivity. At 3 months of age, the females start to breed, and a single pair can produce up to 40 offspring per year. The consequence of this breeding ability was seen when Thomas Austin introduced 24 rabbits for sport hunting onto his property in the state of Victoria, Australia, on Christmas Day, 1859. Within 10 years, the descendants of these rabbits reached a population numbering in the millions that spread throughout much of eastern Australia. By the early twentieth century, the population plateaued at about 200 million. The population growth of rabbits can described by a simple mathematical formula, x(t þ 1) ¼ er x(t), where x(t) is the population size of rabbits in year t, and r is a biological parameter called the intrinsic rate of increase, which gives the maximum rate at which the rabbit population can increase. Studies have shown that r for invading European rabbits is roughly 2.5 year–1, which when plugged into this equation means the population can increase by a factor of e2:5 ¼ 12 each year. This reproductive potential is impressive, the more so when the equation above is used to predict the growth of the rabbit population. Starting with 24 rabbits, assuming they and their descendants maintain the intrinsic rate of increase of 2.5 year–1, in 5 years the population would be over 250,000, and in 10 years it would be almost 100 billion (1011). Following this exponential growth a little longer, in about 21 years the mass of rabbits would exceed that of the Earth (61024 kg). This did not happen. The reason is that rabbits experienced density-dependent reductions in their per capita population growth rate. The per capita population growth rate is the natural logarithm of the number of individuals in a population at some time t þ 1 divided by the number at time t, loge x(t þ 1)= x(t), where time is measured in units appropriate for the species (years for rabbits, minutes for bacteria, etc.). The per capita population growth rate integrates both reproduction and survival, and when the per capita population growth rate is density dependent, the birth and/or death rates change with density. For rabbits in Australia, when densities became very high, birthrates declined and death rates increased as rabbits suffered food shortages. Like rabbits, all populations cannot maintain their intrinsic rate of increase indefinitely. Eventually, densities will become high enough that birthrates decline and/

Density Dependence or death rates increase. Eventually, the per capita population growth rate will drop to zero. Populations that persist for long periods of time must have negative per capita population growth rates at high densities that stop unbounded increases and positive per capita population growth rates at low densities to stop population extinction. In fact, in the long run, a population must have an average per capita population growth rate of zero. Although all persisting populations must have density-dependent population dynamics, the factors leading to density dependence are often multiple, complex, and not easily identified and understood. For example, the exotic dynamics shown by midges in Lake Myvatn (figure 1B), with fluctuations of over five orders of magnitude, involves density dependence that causes populations to crash from very high densities. But what explains the timing of the crashes, why do the midges crash for several generations in a row, and what saves the population at low density so that the species does not become extinct? Detailed answers to these questions about midges, and similar questions for other species, are often extremely hard to answer. Much of the study of population ecology revolves around explaining the factors causing density dependence and the consequences they have for population dynamics. 3. ENDOGENOUS POPULATION VARIABILITY

Density dependence not only bounds a population above and below but also sets the character of the population dynamics. To illustrate this, it is easiest to use another mathematical model that includes a density-dependent per capita population growth rate, specifically x(t þ 1) ¼ er(1



where, as before, r is the intrinsic rate of increase, and K is often called the carrying capacity. Here, the per capita population growth rate, r(1 x(t)=K), decreases as the population size x(t) increases, reaching zero when x(t) ¼ K. Thus, K for the rabbits in Australia would be around 200 million. Because the per capita population growth rate is positive when x(t) is less than K and negative when x(t) is greater than K, it seems reasonable to expect that the population will eventually settle close to K. Although this is the case sometimes, it is not always so. This is because density dependence itself can generate population variability. Before we proceed, a disclaimer is needed. Simple models such as the one above are extremely helpful in understanding basic ecological phenomena, such as the possible consequences of density dependence. Nonetheless, they are not very realistic and do not


necessarily do a good job describing the dynamics of any real species. But it is in fact their unrealistic simplicity that makes these models didactically valuable; the point is not that real populations act exactly like the model but instead that the general types of phenomena shown by the model may in fact have counterparts in real systems. Figure 2 illustrates the population dynamics generated by the simple model with density-dependent per capita population growth rates for three values of the intrinsic rate of increase, r. When r is low (r ¼ 0:1), the population can increase only slowly, so a graph of x(t þ 1) versus x(t) is nearly a straight line (figure 2A). Nonetheless, the line curves down slightly, showing that the population is increasing when x(t) is below the carrying capacity K and decreasing when x(t) is above K. When r is intermediate (r ¼ 1), there is a higher per capita population growth rate when densities are low, yet the per capita population growth rate declines more rapidly so that x(t þ 1) ¼ x(t) again when x(t) ¼ K (figure 2B). Finally, when r is high (r ¼ 2:2), the population increases very rapidly from low densities and declines rapidly from high densities (figure 2C). In fact, it increases so rapidly that the relationship between x(t þ 1) and x(t) is strongly hump-shaped; when the population density starts at some intermediate value, say K/2, the resulting population at the next time step is higher than K, whereas if the population starts at high values, say 2K, the population subsequently crashes to very low levels. Plots of the model populations over time reflect these patterns. When r is low, the population rises slowly toward the carrying capacity K (figure 2D), whereas when r is intermediate, the population attains K rapidly (figure 2E). However, when r is high, the population experiences perpetual booms and busts; when populations are initially low, they bounce to high densities in the next time step and then drop down again in the time step after that. In this case, the formerly stable carrying capacity becomes unstable; even if the population started very close to K, it would exhibit cycles of increasing amplitude until it settled down to the perpetual cycle. Thus, although the carrying capacity K is unstable, the cycle is stable. Such stable boom-and-bust cycles are rarely seen so starkly in real populations, but this in no way diminishes the lesson from the simple model that density dependence can itself generate population fluctuations. There is no environmental variability in the model, so the only factor creating these cycles is endogenous. Do purely endogenously driven population fluctuations occur in nature? Certainly, although not necessarily as clearly and simply as in the model. In laboratory systems, sustained fluctuations have been

Population Ecology


D. 2K

r = 0.1

x (t)

x (t+1)

A. 2K


B. 2K




0 0




C. 2K



x (t)




K x (t)




20 Time, t


I. 2K



0 0


0 0

F. 2K

r = 2.2




0 0

0 H. 2K

E. 2K


x (t)

x (t+1)


0 0

x (t+1)

G. 2K

20 Time, t


Figure 2. Hypothetical population dynamics generated by the model x (tþ1)¼er [1 x(t)/Kþe(t)] for values of the intrinsic rate of in crease r ¼ 0.1, 1, and 2.2. Panels A C plot the population abundance at time step tþ1 against the abundance in the previous time step t. The gray line gives the one to one line; therefore, x (tþ1) crosses the dashed line at K, because at K, x (tþ1)¼ x (t). Panels D F give tra

jectories of population abundance, x(t), through time t when there is no environmental stochasticity, e (t) ¼ 0. Panels G I give population trajectories with environmental stochasticity starting from K; each

created for flies and beetles that involve either populations that reach high enough densities to consume all their food, or cannibalism in which high densities of adults consume large numbers of juveniles. In nature, sustained cycles consistent with a single-species population model have been observed for fish species in which a few, large adult individuals dominate the population by consuming most of the juveniles. This domination is punctuated by bursts of juvenile success when the current large adults senesce and die, and the burst of juvenile recruitment establishes the next dominant cohort of adults. These types of endogenous fluctuations involve species that have distinct stage structures, a topic discussed in chapter II.1. Even more complex possibilities occur when there are strong interactions among two or more species, another topic described in chapter II.3.


trajectory is subjected to the same sequence of environmental fluctu ations given by values of e(t).

In the models so far, the only source of population fluctuations has been density dependence, but real populations are buffeted by purely environmental forces. The consequences of these exogenous sources of population variability on population dynamics depend on density dependence. To illustrate the importance of density dependence, the population model can be modified to include environmental stochasticity— variability that is unpredictable. Specifically, let the per capita population growth rate be r(1 x(t)=K) þ e(t), where e(t) is selected at random from a normal probability distribution. Thus, the per capita population growth rate includes density dependence and environmental stochasticity. Figure 2G–I illustrates examples of population trajectories generated by the new stochastic model.

Density Dependence Even though the environmental stochasticity [the values of e(t)] is the same for all examples, the impact it has on population fluctuations is markedly different. These differences can be understood by comparing populations with and without environmental stochasticity (figure 2D–F). When the intrinsic rate of increase, r, is low, population densities are brought very slowly toward the carrying capacity K (figure 2D). Therefore, in the presence of environmental stochasticity (figure 2G), the weak endogenous force of density dependence does little to counteract the environmental fluctuations that buffet populations away from their initial abundance at K. In contrast, for intermediate r (figure 2E), population densities away from K are returned rapidly, so in the stochastic case, the environmentally driven fluctuations are more rapidly damped out (figure 2H). Finally, for the high value of r there are sustained fluctuations driven by endogenous processes alone (figure 2F). Environmental stochasticity adds to this variability, creating cycles that tend to have higher booms and lower busts (figure 2I). This simple model with environmental stochasticity illustrates the dual nature of density dependence. Density dependence can reduce the magnitude of population fluctuations if it acts to bring the population rapidly to some stable point (K in the simple model). Conversely, density dependence can itself generate endogenous population variability by driving a point unstable. Thus, understanding population fluctuations requires understanding the interactions between density-dependent per capita population growth rates and exogenous stochasticity. 5. RETURNING TO THE THREE QUESTIONS

The three questions concerning single-species population dynamics—what explains the abundance of a population, and how its abundance changes through time and space—intimately involve density dependence. Density dependence sets the average population abundance because the average abundance is where the average per capita population growth rate is zero. Density dependence also determines the characteristics of the fluctuations in population abundance through time because it can generate endogenous population fluctuations. Furthermore, density dependence determines the impact of environmental stochasticity on population dynamics, either damping out environmental stochasticity rapidly or allowing it to produce large fluctuations in population abundance. Finally, when populations in different geographic locations differ in either average abundance or the characteristics of their dynamics, then density dependence somehow differs among populations.


Although density dependence is fundamental to population dynamics, determining how per capita population growth rates depend on density for a specific species is often very difficult. To measure how per capita population growth rates depend on density, it is necessary to observe populations at both very low and very high abundances. Some populations fluctuate sufficiently violently that they naturally occur at both very low and very high abundances, but other populations do not. Experiments designed to perturb population abundances are often the best way to measure density dependence, although for many species, such manipulative experiments are impossible or unethical. Another difficulty is that species dynamics are generally affected by those of other species, such as species that are eaten by the focal species, compete with the focal species, or eat the focal species. For species with strong and close interactions, the dynamics of one cannot be separated from those of the other species. Despite these difficulties, we know a lot about the dynamics and density dependence of a large number of plant and animal species. This is not a random selection of species. Some species were studied because they have practical significance, either being pests that we want to eliminate or endangered species that we want to protect. Other well-studied species were selected because they are particularly easy to study, at least relative to related species in other areas. For example, the study of moose on Isle Royale has the advantage that the population is well defined and isolated from factors (such as deer and elk populations) that could complicate its dynamics. A final set of well-studied species consists of those whose dynamics is sufficiently dramatic to beg an explanation that an ecologist cannot resist, for example, the hugely fluctuating midge population in Lake Myvatn. To understand the population dynamics of any species requires long-term study into the many facets of the species ecology that affect its per capita population growth rate.

FURTHER READING Gotelli, Nicholas J. 2001. A Primer of Ecology, 3rd ed. Sunderland, MA: Sinauer Associates. Gurney, W.S.C., and R. M. Nisbet. 1998. Ecological Dy namics. New York: Oxford University Press. Ives, Anthony R. 1998. Population ecology. In S. I. Dodson, T.F.H. Allen, S. R. Carpenter, A. R. Ives, R. L. Jeanne, J. F. Kitchell, N. E. Langston, and M. G. Turner, eds. Ecology. New York: Oxford University Press, 235 314. Turchin, Peter. 2003. Complex Population Dynamics: A Theoretical/Empirical Synthesis. Princeton, NJ: Princeton University Press.

II.3 Biological Chaos and Complex Dynamics Alan Hastings OUTLINE

1. Fluctuations in populations 2. Brief guide to dynamic systems 3. Chaotic dynamics in models in ecology and population biology 4. Search for chaos in data 5. Resolution as noisy clockwork 6. Other complex dynamics The cause of fluctuations in ecological populations has long been the subject of study, with the goal of understanding the relative importance of exogenous versus endogenous forces in explaining observed dynamics. The discovery of the likelihood of chaotic dynamics in simple discrete-time models that could be used to describe singlespecies population dynamics spurred much research focused on understanding chaos and its importance and likelihood in ecological systems. To understand the importance of chaos, we consider the role of fluctuations in ecological systems, the generation of chaotic dynamics in models, and the determination of chaos from time series. This naturally leads to more general questions on the role of complex dynamics in ecology and to a more synthetic view of the causes of observed fluctuations.

GLOSSARY asymptotically stable solution. A solution that is ap-

proached by all nearby solutions is asymptotically stable. This is also known as an attractor. chaos. Chaos is a property of an attractor in a dynamic system that can be roughly characterized as sensitive dependence on initial conditions and can be detected by the presence of a positive Lyapunov exponent. cycle. A cycle is a solution that repeats at regular intervals. equilibrium. An equilibrium of a model is a solution that does not change in time. Lyapunov exponent. A Lyapunov exponent represents the exponential rate of divergence (if positive) or

convergence (if negative) of (two) solutions started on or near an attractor.


A key observation that is central in ecology is that populations fluctuate in time. These fluctuations can exhibit some regularity or can be irregular. Periodical cicadas emerge with great regularity, whereas outbreaks of other insects such as locusts are both dramatic and irregular. The cause of fluctuations in populations in ecology has been a central question in ecology for many years. Early in the history of ecology, Volterra and Lotka focused on the regular oscillations produced by interactions between predator and prey in their models. Shortly thereafter, Gause attempted to reproduce these oscillations in laboratory systems using microorganisms and found that sustained oscillations were difficult to reproduce. In the simple laboratory systems, either the predator ate up all the prey and then starved or the predator could not find enough food and starved with only the prey surviving. This set up a problem that remains until today, namely, what allows predator and prey to coexist. Also, many of the mechanisms that might allow coexistence of species might lead to more complex dynamics, and more often coexisting species fluctuate. In any examination of natural populations, fluctuations in numbers have been found to be the almost universal outcome. These fluctuations could range from relatively regular cycles, such as those observed in small mammal populations, or more dramatic changes, such as outbreaks of insect populations. A classic debate in ecology has focused on the causes of these fluctuations. One potential source of fluctuations could be external influences, such as changes in weather or climate. These exogenous forces could be responsible for changes in the dynamics of populations, producing cycles that were either regular or irregular. Another cause of changes in the numbers of populations would

Chaos and Complex Dynamics be endogenous forces within the population that would lead to cycles or more complex dynamics. Enlarging the question to look at fluctuations not just in time but also over space was thought of as one way to decide. Population fluctuations that are synchronous over space would either have to be caused by exogenous forces that were synchronous over space (the Moran effect) or endogenous forces such as dispersal that would synchronize populations. One particular kind of fluctuating population that we will return to later that deserves special attention is the incidence of disease. In particular, the numbers of individuals with various childhood diseases (measles, mumps, rubella, and others) in the prevaccination era have been intensely studied. These particular fluctuations have played a special role for several reasons. The data are of much higher quality, with more, and more accurate, observations than for the numbers of many organisms. Another important aspect is that the underlying interactions producing the dynamics of diseases are relatively simple and relatively easily described. Finally, because data are known for multiple diseases in multiple locations, deeper understanding is possible. 2. BRIEF GUIDE TO DYNAMIC SYSTEMS

The idea I now consider is that the primary cause of the fluctuations in ecological systems is interactions within and between populations rather than primarily external influences. The role of internal dynamics should be carefully examined in model systems using ideas from the mathematical theory of dynamics. This theory has undergone long development and can be traced back to attempts by physical scientists to understand the motion of the planets and other systems. A brief review of the theory of dynamic systems can elucidate many implicit assumptions ecologists make when using models framed in differential or difference equations or partial difference equations in attempts to understand what drives dynamics of ecological systems. First, dynamic systems are systems that incorporate changes through time. In many engineering or physical examples, the systems, initial conditions, and parameters controlling the systems are well defined. However, in ecology, this is rarely the case. Ecological systems are so complex that only rough descriptions are often possible, the measurements of population sizes are notoriously difficult, and parameters such as birth and death rates are known only imprecisely. Thus, the way to understand the behavior of ecological systems is not to look for exact solutions of fully specified dynamic systems through either numerical or analytic means. An understanding of what is known as the


qualitative behavior is much more appropriate and important. Fortunately for ecologists, advances in understanding dynamic systems have focused primarily on the qualitative behavior of these systems. One focus is on long-term or asymptotic behavior of dynamic systems. Several definitions are needed. The simplest long-term dynamic is an equilibrium, or a solution that remains constant in time. The next most complex dynamic behavior is a cycle, or a solution that repeats regularly in time, so x(t þ T) ¼ x(t) for some cycle period T for all times t. Before we turn to other more complex asymptotic behaviors, we need a definition of stability. I use heuristic, rather than mathematically rigorous, definitions. An asymptotic solution to a dynamic system is stable if it is approached from all nearby initial conditions. Such a solution is also known as an attractor. Attractors can be as simple as equilibrium points but can also be more complex. Cyclic attractors are also known as limit cycles. Quasicyclic attractors oscillate but with two (superimposed) incommensurate periods so the solutions do not exactly repeat. Attractors can also be chaotic. A chaotic attractor is a solution that is still stable in the sense that it is approached by nearby solutions. A chaotic attractor has the property that even solutions that start on the chaotic attractor diverge from each other at an exponential rate, so all solutions have very sensitive dependence on initial conditions. The understanding and appreciation of the concept of chaotic dynamics can be traced back to work by Lorenz over 40 years ago and further back to work of Poincare´ and Andronov and others more than a century ago. The importance of chaotic dynamics is that it challenges notions of predictability. However, note that even solutions that exhibit chaotic dynamics are predictable over short time scales, even if prediction is not possible over long time scales. 3. CHAOTIC DYNAMICS IN MODELS IN ECOLOGY AND POPULATION BIOLOGY

Within ecology, the first appreciation of chaotic dynamics arose in studies of single-species models with overcompensatory density dependence and nonoverlapping generations. The key idea of overcompensatory density dependence refers to the fact that not only does the per capita production of individuals in the next generation decline as the number of individuals in the current generation goes up, but additionally, the total population size of the next generation eventually declines as the number of individuals in the current generation is increased. These models take the general form


Population Ecology N(t þ 1) ¼ N(t)f [N(t)],

where N(t) is the population size of a single species at time t, and f[N(t)] is the mean number of individuals in the next generation left by an individual in the current generation. The dependence of the function f on the population size N(t) is used to describe the action of density-dependent factors in determining the population size of the following generation. Thus, if f[N(t)] is a declining function of the population size, the model exhibits density dependence. And if N(t)f[N(t)] eventually declines, then the system is said to incorporate overcompensatory density dependence. Classic density-dependent terms used in fisheries can be incorporated this way, with the Ricker function exhibiting overcompensatory density dependence, and the Beverton-Holt form not exhibiting this property. It is the functional forms that include overcompensatory density dependence that can lead to complex dynamics and chaos. One important observation is that chaotic dynamics is essentially a generic property of discrete-time models with strong enough overcompensatory density dependence, so it makes sense to study simple models. Some of the first investigations of chaotic dynamics used an idealized form to describe the population dynamics, namely, x(t þ 1) ¼ r x(t)[1 x(t)], where x(t) is a scaled (between 0 and 1) measure of population size at time t, and r (between 0 and 4) is a measure both of the growth rate when rare and degree of overcompensation in the density dependence. This seemingly simple model, the quadratic or logistic population model, was extensively investigated by Robert May (1974b) and others in the 1970s and since. A description of the dependence of the dynamic behavior of the quadratic model as a function of the parameter r provides insight into the possible aspects of chaotic dynamics in populations, and similar behavior is found in other models. For small values of r, the population cannot survive. For larger values of r, the population inevitably approaches a stable equilibrium. As the parameter r is increased, the asymptotic behavior of the model is a two-point cycle with the population alternating between two values. This perioddoubling behavior continues as r increases, with asymptotic behavior of four-point cycles, then eightpoint cycles, then 16, and so on. The ranges of values of the parameter r over which the cycles of period 2n occur get smaller and smaller as n gets bigger. A limit is reached at a critical value of r, beyond which the dynamics is much more complex, and chaotic solutions

are found. The presence of this period-doubling sequence is one of several ‘‘routes to chaos.’’ Although an emphasis on the study of the discretetime model for a single species seemed to imply that chaos was strictly a property of ecological systems with nonoverlapping generations, this is not the case. In continuous-time models, there need to be at least three interacting quantities (e.g., species) for the system to exhibit chaotic dynamics. Since the discovery of chaos in discrete-time models, studies have emphasized that chaotic dynamics is likely to be found in simple models of food webs (or food chains) with three or more species with nonlinearities as would arise from functional responses used to describe predation. Essentially, chaos seems to arise when dynamics is characterized by interactions among oscillating systems with different periods as is the case with predator–prey systems. Even systems with enough competitive relationships can produce chaos. Chaos and other forms of complex dynamics also arise very naturally in descriptions of ecological systems that include different population levels at different spatial locations. Chaotic behavior can arise in models as simple as those describing two coupled predator–prey oscillators. One additional class of models that can exhibit chaotic dynamics comes from epidemiology. Among the simplest epidemiological models are those phrased in terms of susceptible, infectives, and removed individuals, the SIR model first discussed by Kermack and McKendrick in the 1920s. Because this is a continuoustime model that can be reduced to two quantities, it cannot produce chaos. However, if the contact rate describing the transition from susceptible to infective varies seasonally (periodically), the model does have chaotic solutions. 4. SEARCH FOR CHAOS IN DATA

The behavior of the simplest population models that could exhibit chaotic dynamics, the single-species discrete-time models, seemed to provide a possible way of explaining irregular fluctuations in natural populations. However, the possibility of chaotic dynamics is not the same as the existence of chaotic dynamics. Thus, great efforts have been made to uncover evidence for chaotic dynamics in ecological systems, both natural and in the laboratory. One procedure for uncovering chaotic dynamics begins with the collection of a time series of population abundances or disease incidence or similar data. At first, the possibility of uncovering chaotic dynamics would seem to be doomed because typical time series of natural populations in ecology or population biology

Chaos and Complex Dynamics focus on only a single species or the incidence of a single disease. Here a powerful idea that shows how to study a system with many dimensions (e.g., many species) from a single dimension (time series for a single species) comes to the rescue. The idea is based on a powerful mathematical argument (though the conditions that justify the procedure are almost never checked) that says that the full dynamic behavior of a system can be understood and replicated by a reconstruction procedure beginning with a single time series. Instead of focusing on the time series itself, say x(t), consider representing the data with lags and plotting x(t) versus x(t T) and x(t 2T). The actual number of lags chosen (which may often be more than two) and the length of the lag (T) can be critical, and unfortunately there is no well-established procedure for their choice. However, by using this procedure, one can clearly focus on the search for chaos. At this point, there is still the problem of looking for a signature of chaos, such as a positive Lyapunov exponent. For data sets much larger than any found in ecology, there would be direct procedures for estimating this Lyapunov exponent. In ecology, given the limited data sets, an approach based on choosing a model and fitting this model to the data and then looking at properties of the best-fitting model must be used. There are relatively standard approaches for estimating a Lyapunov exponent from a model. There are, however, two different kinds of approaches for making models fit, based either on choosing a very general functional form, which is purely phenomenological, or on choosing from among a set of much more mechanistic descriptions of the biological processes. The former approach has the advantage of flexibility while perhaps ignoring important biological constraints. The latter approach has the advantage of biological realism but can be difficult to apply to data from a single time series of a more complex system. The latter approach also depends critically on choosing a good set of candidate models that can exhibit an appropriate range of dynamic behavior. The general approach of using time series to look for chaos has been applied to a variety of data sets, including childhood diseases (Finkenstadt and Grenfell, 2000) and laboratory systems such as flour beetles (Costantino et al., 1997). The evidence for chaos in childhood diseases such as measles is not completely clear-cut, but there at least seems to be a very strong possibility of chaotic dynamics. For the flour beetle system studied by Costantino and collaborators (1997), the evidence for chaos in this highly controlled system is clearer. The calculations of the Lyapunov exponent for a model that fits the system well yield a positive Lyapunov exponent, one hallmark of chaos. Moreover, both the model (clearly) and ex-


perimental results (somewhat less clearly) also exhibit the period-doubling behavior that is one of the signatures of chaos. For systems outside the laboratory other than childhood diseases, the evidence for chaotic dynamics is much weaker. This may be because natural systems are not chaotic, but part of this may be a result of the difficulty of obtaining high-quality data. One point that clearly is important is that stochastic forces must play a role, as environmental fluctuations and demographic heterogeneity inevitably influence all populations. Stochasticity clearly plays an important role. 5. RESOLUTION AS NOISY CLOCKWORK

To some extent, so far we have focused on the causes of observed fluctuations in population levels as a dichotomy: either endogenous or exogenous. It is clear that stochastic forces must be important for the dynamics of natural populations, and it is equally clear that there are strong interactions within and between species affecting population dynamics. This sets up what is essentially a false dichotomy between two forces. Instead, it is much more realistic to consider the interplay between endogenous and exogenous forces. This idea that the dynamics of natural populations must depend on both stochastic and deterministic forces has been referred to as ‘‘noisy clockwork’’ (Bjornstad and Grenfell, 2001), although the idea has a longer history. One important aspect is that the stochastic aspects of population dynamics cannot be thought of as small perturbations of a deterministic population trajectory. Instead, the complex endogenous aspects of population dynamics and the exogenous forces are inexorably intertwined. These two aspects together produce the complex population trajectories we observe. One can study populations from this point of view and obtain new insights. One can define a Lyapunov exponent for a stochastic system in terms of expectations and therefore sensibly ask whether chaotic dynamics exist in natural systems. However, difficulties of limited data still make detection of chaos a challenge. 6. OTHER COMPLEX DYNAMICS

There is one other way that the emphasis on chaotic dynamics in ecology may have led investigators away from important ecological behavior. Although it is possible to sensibly define chaotic behavior on shorter time scales, much of the study of chaotic behavior, especially in model systems, has emphasized asymptotic behavior. However, many ecological systems may best be understood by studying transient dynamics rather than asymptotic behavior.


Population Ecology

The modern approaches to understanding dynamic systems and new statistical approaches for understanding time series that have been used in the study of chaos in ecology can also be used to shed light on other dynamic behavior. Transient dynamics can be studied using ideas from dynamic systems. There are mathematical tools for understanding spatial systems and, in particular, systems of coupled oscillators (e.g., predator–prey systems and epidemiological systems) and synchrony, even in the presence of stochasticity. New and novel statistical approaches will also likely prove useful in understanding the forces producing observed population fluctuations. Model-based frequentist approaches and Bayesian methods that truly incorporate different kinds of stochasticity will both contribute to a deeper understanding of population dynamics. FURTHER READING Bjornstad, O. N., and B. T. Grenfell. 2001. Noisy clockwork: Time series analysis of population fluctuations in animals. Science 293: 638 643.

Costantino, R. F., R. A. Desharnais, J. M. Cushing, and B. Dennis. 1997. Chaotic dynamics in an insect population. Science 275: 389 391. Finkenstadt, B. F., and B. T. Grenfell. 2000. Time series modelling of childhood diseases: A dynamical systems approach. Journal of the Royal Statistical Society Series C Applied Statistics 49: 187 205. Hastings, A. 2004. Transients: The key to long term ecolog ical understanding? Trends in Ecology and Evolution 19: 39 45. Hastings, A., C. Hom, S. Ellner, P. Turchin, and H.C.J. Godfray. 1993. Chaos in ecology: Is mother nature a strange attractor? Annual Reviews of Ecology and Sys tematics 24: 1 33. Huisman, J., and F. J. Weissing. 2001. Fundamental unpre dictability in multispecies competition. American Nat uralist 157: 488 494. May, R. M. 1974a. Stability and Complexity in Model Eco systems. Princeton, NJ: Princeton University Press. May, R. M. 1974b. Biological populations with nonover lapping generations: Stable points, stable cycles, and chaos. Science 186: 645 647.

II.4 Metapopulations and Spatial Population Processes Ilkka Hanski OUTLINE

1. 2. 3. 4. 5. 6.

Metapopulation patterns and processes Long-term viability of metapopulations Metapopulations in changing environments Evolution in metapopulations Spatial dynamics in nonpatchy environments Metapopulations, spatial population processes, and conservation 7. Conclusion Most landscapes are complex mosaics of many kinds of habitat. From the viewpoint of a particular species, only some habitat types, often called ‘‘suitable habitat,’’ provide the necessary resources for population growth. The remaining landscape, often called the (landscape) matrix, can only be traversed by dispersing individuals. Often the suitable habitat occurs in discrete patches, an example of which is a woodland in the midst of cultivated fields—for forest species, the woodland is like an island in the sea. The woodland may be occupied by a local population of a forest species, but many such patches are likely to be temporarily unoccupied because the population became extinct in the past and a new one has not yet been established. At the landscape level, woodlands and other comparable habitat patches comprise networks in which local populations living in individual patches are connected to each other by dispersing individuals. A set of local populations inhabiting a patch network is called a metapopulation. In other cases, the habitat does not consist of discrete patches, but even then, habitat quality is likely to vary from one place to another. Habitat heterogeneity tends to be reflected in a more or less fragmented population structure, and such spatially structured populations may be called metapopulations. Metapopulation biology addresses the ecological, genetic, and evolutionary processes that occur in metapopulations. For instance, in a highly fragmented landscape, all local populations may be so small that they all have a high risk of

extinction, yet the metapopulation may persist if new local populations are established by dispersing individuals fast enough to compensate for extinctions. Metapopulation structure and the extinction–colonization dynamics may greatly influence the maintenance of genetic diversity and the course of evolutionary changes. Metapopulation processes play a role in the dynamics of most species because most landscapes are spatially more or less heterogeneous, and many comprise networks of discrete habitat patches. Human land use tends to increase fragmentation of natural habitats, and hence, metapopulation processes are particularly consequential in many human-dominated landscapes.

GLOSSARY connectivity. An individual habitat patch in a patch

network and a local population in a metapopulation are linked to other local populations, if any exist, via dispersal of individuals. Connectivity measures the expected rate of dispersal to a particular patch or population from the surrounding populations. dispersal. Movement of individuals among local populations in a metapopulation is dispersal. Migration is often used as a synonym of dispersal. extinction–colonization dynamics. Local populations in a metapopulation may go extinct for many reasons, especially when the populations are small. New local populations may become established in currently unoccupied habitat patches. Local extinction and recolonization are called turnover events. extinction debt. If the environment becomes less favorable for the persistence of metapopulations through, e.g., habitat loss and fragmentation, species’ metapopulations start to decline. For some metapopulations, the new environment may be below the extinction threshold. Extinction debt is defined as the number of species for which the


Population Ecology

extinction threshold is not met and that are therefore predicted to become extinct but have not yet had time to become extinct. extinction threshold. Classic metapopulation may persist in a habitat patch network in spite of local extinctions if the rate of recolonization is sufficiently high. Long-term persistence is less likely the smaller the habitat patches are (leads to high extinction rate) and the lower their connectivity (leads to low recolonization rate). Below extinction threshold, recolonizations do not occur fast enough to compensate for local extinctions, and the entire metapopulation becomes extinct even if some habitat patches exist in the landscape. local population. This is an assemblage of individuals sharing common environment, competing for the same resources, and reproducing with each other. In a fragmented landscape, a local population typically inhabits a discrete habitat patch. metapopulation. A classic metapopulation is an assemblage of local populations living in a network of habitat patches. More generally, spatially structured populations at landscape scales are often called metapopulations. metapopulation capacity. This is a measure of the size of the habitat patch network that takes into account the total amount of habitat as well as the influence of fragmentation on metapopulation viability. network of habitat patches. In a fragmented landscape, habitat occurs in discrete patches, each one of which may be occupied by a local population, and which together compose a network that may be occupied by a metapopulation. source and sink populations and habitats. A local population that has negative expected growth rate, and that therefore would go extinct without immigration, is called a sink population, and the respective habitat is a sink habitat. A population that has sufficiently high growth rate when small to persist even without immigration is called a source population, and the respective habitat is source habitat. 1. METAPOPULATION PATTERNS AND PROCESSES Different Kinds of Metapopulations

Many classifications of metapopulations have been proposed, and they serve a purpose in facilitating communication, but it should be recognized that in reality there exists a continuum of spatial population structures rather than discrete types. The following terms are often used.

Classic metapopulations consist of many small or relatively small local populations in patch networks. Small local populations have high or relatively high risk of extinction, and hence, long-term persistence can occur only at the metapopulation level, in a balance between local extinctions and recolonizations (further discussed under the Levins Model below). Mainland–island metapopulations include one or more populations that are so large and live in sufficiently big expanses of habitat that they have a negligible risk of extinction. These populations, called mainland populations, are stable sources of dispersers to other populations in smaller habitat patches (island populations). The MacArthur-Wilson model of island biogeography is an extension of the mainland–island metapopulation model to a community of many independent species. Source–sink metapopulations include local populations that inhabit low-quality habitat patches and would therefore have negative growth rate in the absence of immigration (sink populations), and local populations inhabiting high-quality patches in which the respective populations have positive growth rates (source populations; further discussed under Source and Sink Populations below). Nonequilibrium metapopulations are similar to classic metapopulations, but there is no stochastic balance between extinctions and recolonizations, typically because the environment has recently changed and the extinction rate has increased, the recolonization rate has decreased, or both (this is further discussed under Transient Dynamics and Extinction Debt below). Dispersal and Population Turnover

Three ecological processes are fundamental to metapopulation dynamics: dispersal, colonization of currently unoccupied habitat patches, and local extinction. Dispersal has several components: emigration, departure of individuals from their current population; movement through the landscape; and immigration, arrival at new populations or at empty habitat patches. All three components depend on the traits of the species and on the characteristics of the habitat and the landscape, but they may also depend on the state of the population, for instance, on population density. Dispersal may influence local population dynamics. In the case of very small populations, a high rate of emigration may reduce population size and thereby increase the risk of extinction. Conversely, immigration may enhance population size sufficiently to reduce the risk of extinction. Immigration to a currently unoccupied habitat patch is particularly significant in potentially leading to the establishment of a new local

Metapopulations A.


B. 6 3 2



Log area

Log area


0 -1 -2 -3 -4


-5 -4 -4

-6 -3


-1 0 Log isolation





2 3 Connectivity, S



Figure 1. Two examples of the common metapopulation pattern of increasing incidence (probability) of habitat patch occupancy with increasing patch area and connectivity. Black circles represent oc cupied, gray circles unoccupied, habitat patches at the time of sampling. (A) Mainland island metapopulation of the shrew Sorex cinereus on islands in North America. Isolation, which increases with decreasing connectivity, is here measured by distance to the main land. The lines indicate the combinations of area and isolation for which the predicted incidence of occupancy is greater than 0.1, 0.5,

and 0.9, respectively. (From Hanski, I. 1993. Dynamics of small mammals on islands. Ecography 16: 372 375) (B) Classic meta population of the silver spotted skipper butterfly (Hesperia comma) on dry meadows in southern England. The line indicates the combi nations of area and connectivity above which the predicted incidence of occupancy is greater than 0.5. (From Hanski, I. 1994. A practical model of metapopulation dynamics. Journal of Animal Ecology 63: 151 162)

population. From the genetic viewpoint, two extreme forms of immigration and gene flow have been distinguished, the ‘‘migrant pool’’ model, in which the dispersers are drawn randomly from the metapopulation, and the ‘‘propagule pool’’ model, in which all immigrants to a patch originate from the same source population. The latter is likely to reduce genetic variation in the metapopulation. Colonization of a currently unoccupied habitat patch is more likely the greater the numbers of immigrants, which is measured by the connectivity of the patch (see next section). All populations have smaller or greater risks of extinction from demographic and environmental stochasticities (see Stochasticity in Metapopulation Dynamics below) and other causes. Typically, the smaller the population the greater the risk of extinction. Assuming that a local population in patch i has a constant risk of going extinct, Ei, and that patch i, if unoccupied, has a constant probability of becoming recolonized, Ci, the state of patch i, whether occupied or not, is determined by a stochastic process (Markov chain) with the stationary (time-invariant) probability of occupancy given by

Ji is often called the incidence of occupancy. This formula helps explain the common metapopulation patterns of increasing probability of patch occupancy with patch area (which typically decreases Ei and hence increases Ji) and with decreasing isolation (which typically increases Ci and hence increases Ji). Figure 1 gives two examples.

Ji ¼

Ci : Ci þ Ei



Local populations and habitat patches in a patch network are linked via dispersal. Connectivity measures the strength of this coupling from the viewpoint of a particular patch or local population. Connectivity is best defined as the expected number of individuals arriving per unit time at the focal patch. Connectivity increases with the number of populations (sources of dispersers) in the neighborhood of the focal patch; with decreasing distances to the source populations (making successful dispersal more likely); and with increasing sizes of the source populations (larger populations send out more dispersers). A measure of connectivity for patch i that takes all these factors into account may be defined as Si ¼ Ai im Sj6 i exp ( adij )Jj Aj em :



Population Ecology Table 1. Four types of stochasticity affecting metapopulation dynamics

Type of stochasticity Demographic Environmental Extinction colonization Regional

Entity affected

Correlation among entities

Individuals in local populations Individuals in local populations Populations in metapopulations Populations in metapopulations

No Yes No Yes

Here, Aj is the area and Jj is the incidence of occupancy of patch j, dij is the distance between patches i and j, 1/a is species-specific average dispersal distance, and zim and zem describe the scaling of immigration and emigration with patch area. This formula assumes that the sizes of source populations are proportional to the respective patch areas; if information on the actual population sizes Nj is available, the surrogate Jj Aj in equation 2 may be replaced by Nj. Stochasticity in Metapopulation Dynamics

Metapopulation dynamics are influenced by four kinds of stochasticity (types of random events; table 1): demographic and environmental stochasticity affect each local population, and extinction–colonization and regional stochasticity affect the entire metapopulation. Both local dynamics and metapopulation dynamics are inherently stochastic because births and deaths in local populations are random events (leading to demographic stochasticity) and so are population extinctions and recolonizations in a metapopulation (extinction–colonization stochasticity). Environmental stochasticity refers to correlated temporal variation in birth and death rates among individuals in local populations, whereas regional stochasticity refers to correlated extinction and colonization events in metapopulations. Metapopulations are typically affected by regional stochasticity because the processes generating environmental stochasticity, including temporally varying weather conditions, are typically spatially correlated. It can be demonstrated mathematically that all metapopulations with population turnover caused by extinctions and colonizations will eventually become extinct. It is a certainty that, given enough time, a sufficiently long run of extinctions will arise by ‘‘bad luck’’ and extirpate the metapopulation. However, time to extinction can be very long for large metapopulations inhabiting large patch networks (see the Levins Model below), and the metapopulation settles for a long time to a stochastic quasiequilibrium, in which there is variation but no systematic change in the number of local populations.

Source and Sink Populations

Populations may occur in low-quality sink habitats if there is sufficient dispersal from other populations living in high-quality source habitats. Therefore, the presence of a species in a particular habitat patch does not suffice to demonstrate that the habitat is of sufficient quality to support a viable population. Conversely, a local population may be absent from a habitat patch that is perfectly suitable for population growth when a local population happened to become extinct for reasons unrelated to habitat quality. In a temporally varying environment, sink populations may, counterintuitively, enhance metapopulation persistence. This may happen when source populations exhibit large fluctuations leading to a high risk of extinction. The habitat patches supporting such sources may become recolonized by dispersal from sinks, assuming this happens before the sink populations have declined to extinction. In general, dispersal among local populations that fluctuate relatively independently of each other (weak regional stochasticity) enhances the metapopulation growth rate. This happens because when a population has increased in size in a good year, and the offspring are spread among many independently fluctuating populations, subsequent bad years will not hit them all simultaneously. It can be shown that this spreading-of-risk effect of dispersal may be so substantial that it allows a metapopulation consisting of sink populations only to persist without any sources. 2. LONG-TERM VIABILITY OF METAPOPULATIONS

Mathematical models are used to describe, analyze, and predict the dynamics of metapopulations living in fragmented landscapes. A wide range of models can be constructed differing in the forms of stochasticity they include (see Stochasticity in Metapopulation Dynamics above), in whether change in metapopulation size occurs continuously or in discrete time intervals, in how many local populations the metapopulation consists of, in how the structure of the landscape is represented, and so forth. A minimal metapopulation includes two local populations connected by dispersal. At the other


Levins Model

The Levins model has special significance for metapopulation ecology, as it was with this model that the American biologist Richard Levins introduced the metapopulation concept in 1969. The Levins model captures the essence of the classic metapopulation concept—that a species may persist in a balance between stochastic local extinctions and recolonization of currently unoccupied patches. For mathematical convenience, the model assumes an infinitely large network of identical patches, which have two possible states, occupied or empty. The state of the entire metapopulation can be described by the fraction of currently occupied patches, p, which varies between 0 and 1. If each local population has the same risk of extinction, and each population contributes equally to the rate of recolonization, the rate of change in the size of the metapopulation is given by dp ¼ cp(1 p) ep, dt


where c and e are colonization and extinction rate parameters. This model ignores stochasticity, but it is a good approximation of the corresponding stochastic model for a large metapopulation inhabiting a large patch network. The Levins model is structurally identical with the logistic model of population growth, which can be seen by rewriting equation 3 as dp p ¼ (c e)p 1 : (4) dt 1 e=c c e gives the rate of metapopulation growth when it is small, and 1 e=c is the equilibrium metapopulation size (‘‘carrying capacity’’). The ratio c/e defines the basic reproductive number R0 in the Levins model. A species can increase in a patch network from low occupancy if R0 ¼ c=e > 1. This condition defines the extinction threshold in metapopulation dynamics. In reality, in a finite patch network, a metapopulation may become extinct because of stochasticity even if the threshold condition c=e > 1 is satisfied. When a diffusion approximation is used to analyze the stochastic Levins model, the mean time to extinction T can be calculated as a function of the number of habitat patches n and p*, the size of the metapopulation at quasiequilibrium. Figure 2 shows the number of patches that the network must have to make T at least 100


Number of patches n

extreme, assuming infinitely many habitat patches leads to a particularly simple description of the classic metapopulation, which is discussed next.





0.4 0.6 Occupancy state p*



Figure 2. The number of habitat patches n needed to make the mean time to metapopulation extinction T at least 100 times longer than the mean time to local extinction. The dots show the exact result based on the stochastic logistic model, and the line is based on the following diffusion approximation:

T =

e − (n −1) p* . n p*2 (1 − p*) n −1


(From Ovaskainen and Hanski, 2003)

times as long as the expected lifetime of a single local population (given by the inverse of the extinction rate). For metapopulations with large p*, a modest network of n ¼ 10 patches is sufficient to allow longterm persistence, but for rare species (say p* < 0:2), a large network of n > 100 is needed for long-term persistence. The stochastic Levins model includes extinction– colonization stochasticity but no regional stochasticity, which leads to correlated extinctions and colonizations. In the presence of regional stochasticity, the mean time to metapopulation extinction does not increase exponentially with increasing n as in figure 2 but as a power function of n, the power decreasing with increasing correlation in extinction and recolonization rates, reducing long-term viability. This result is analogous to the well-known effects of demographic and environmental stochasticities on the lifetime of single populations. Spatially Realistic Metapopulation Models

There is no description of landscape structure in the Levins model; hence, it is not possible to investigate with this model the consequences of habitat loss and fragmentation. Real metapopulations live in patch networks with a finite number of patches; the patches are of varying size and quality, and different patches have different connectivities, which affect the rates of immigration and recolonization. These considerations

Population Ecology

have been incorporated into spatially realistic metapopulation models. The key idea is to model the effects of habitat patch area, quality, and connectivity on the processes of local extinction and recolonization. Generally, the extinction risk decreases with increasing patch area because large patches tend to have large populations with a small risk of extinction, and the colonization rate increases with connectivity to existing populations. The theory provides a measure to describe the capacity of an entire patch network to support a metapopulation, denoted by lM and called the metapopulation capacity of the fragmented landscape. Mathematically, lM is the leading eigenvalue of a ‘‘landscape’’ matrix, which is constructed with assumptions about how habitat patch areas and connectivities influence extinctions and recolonizations. The size of the metapopulation at equilibrium is given by pl * ¼ 1 e=(clM ),


which is similar to the equilibrium in the Levins model, but with the difference that metapopulation equilibrium now depends on metapopulation capacity, and metapopulation size pl is measured by a weighted average of patch occupancy probabilities. The threshold condition for metapopulation persistence is given by lM > e=c:


In words, metapopulation capacity has to exceed a threshold value, which is set by the extinction proneness (e) and colonization capacity (c) of the species, for longterm persistence. To compute lM for a particular landscape, one needs to know the range of dispersal of the focal species, which sets the spatial scale for calculating connectivity (parameter a in equation 2), and the areas and spatial locations of the habitat patches. The metapopulation capacity can be used to rank different fragmented landscapes in terms of their capacity to support a viable metapopulation: the larger the value of lM, the better the landscape. Figure 3 gives an example in which metapopulation capacity explains well the size of butterfly metapopulations in dissimilar patch networks.


A fundamental question about metapopulation dynamics concerns long-term viability, which has great significance for the conservation of biodiversity in fragmented landscapes (see Metapopulations, Spatial Population Processes, and Conservation below). A

1.0 0.8 0.6 p*


0.4 0.2 0 -3.0






Figure 3. Metapopulation size of the Glanville fritillary butterfly (Melitaea cinxia) as a function of the metapopulation capacity lM in 25 habitat patch networks. The vertical axis shows the size of the metapopulation based on a survey of habitat patch occupancy in 1 year. The empirical data have been fitted by a spatially realistic model (continuous line; the broken lines give the model fit to the second smallest and the second largest positive values). The result provides a clear cut example of the extinction threshold. (From Hanski and Ovaskainen, 2000)

patch network will not support a viable metapopulation unless the extinction threshold is exceeded, and even if it is, a metapopulation may become extinct for stochastic reasons (see Levins Model above). Long-term viability is further reduced by environmental change. Ephemeral Habitat Patches

Innumerable species of fungi, plants, and animals live in ephemeral habitats such as decaying wood. A dead tree trunk may be viewed as a habitat patch for local populations of such organisms. The trunk is not permanent, largely because of the action of the organisms themselves, and local populations necessarily become extinct at some point. Parasites living in a host individual can be similarly considered as comprising a local population, which necessarily becomes extinct when the host individual dies. This example reflects fundamental similarities between metapopulation biology and epidemiology. Regular disappearance of habitat patches increases extinctions, but the metapopulation may still persist in a stochastic quasiequilibrium. Equation 1 may be extended to include patch extinction and the appearance of new patches: Ji ¼

Ci [Ci (1 Ci Ei )age ] ; Ci þ Ei


Metapopulations where age is the age of patch i. Following its appearance, a new patch is initially unoccupied, Ji ¼ 0 when age ¼ 0. When the patch becomes older, the incidence of occupancy approaches the equilibrium [Ci =(Ci þ Ei )] given by equation 1 and determined by the extinction–colonization dynamics of the species. The precise trajectory is given by equation 7, where the term in square brackets declines from Ci when age ¼ 0 to zero as age becomes large and when equation 1 is recovered. Transient Dynamics and Extinction Debt

Human land use often causes the loss and fragmentation of the habitat for many other species. Following the change in landscape structure, especially if the change is abrupt, it takes some time before the metapopulation has reached the new quasiequilibrium, which may be metapopulation extinction. Considering a community of species, the term extinction debt refers to situations in which, following habitat loss and fragmentation, the threshold condition is not met for some species, but these species have not yet become extinct because they respond relatively slowly to environmental change. More precisely, the extinction debt is the number of extant species that are predicted to become extinct, sooner or later, because the threshold condition for long-term persistence is not satisfied for them following habitat loss and fragmentation. How long does it take before the metapopulation has reached the new quasiequilibrium following a change in the environment? The length of the transient period is longer when the change in landscape structure is greater, when the rates of extinction and recolonization are lower, and when the new quasiequilibrium following environmental change is located close to the extinction threshold. The latter result has important implications for conservation. Species that have become endangered as a result of recent changes in landscape structure are located, by definition, close to their extinction threshold, and hence, the transient period in their response to environmental change is predicted to be long. This means that we are likely to underestimate the level of threat to endangered species because many of them do not occur presently at quasiequilibrium with respect to the current landscape structure but are slowly declining because of past habitat loss and fragmentation. On the positive side, long transient time in metapopulation dynamics following environmental change gives us humans more time to do something to reverse the trend.



The hierarchical structure of metapopulations, from individuals to local populations to the entire metapopulation, has implications for evolutionary dynamics. In addition to natural selection occurring within local populations, different selection pressures may influence the fitness of individuals during dispersal and at colonization. Individuals that disperse from their natal population and succeed in establishing new local populations are likely to comprise a nonrandom group of all individuals in the metapopulation. Particular phenotypes and genotypes may persist in the metapopulation because of their superior performance in dispersal and colonization even if they would be selected against within local populations. This is often called the metapopulation effect. The most obvious example relates to emigration rate and dispersal capacity. The most dispersive individuals are selected against locally because their local reproductive success is reduced by early departure. However, these individuals may find a favorable habitat elsewhere, which increases their fitness in the metapopulation. Which particular phenotypes and genotypes are favored in particular metapopulations depends on many factors. Local competition for resources and competition with relatives for mating opportunities selects for more dispersal, and so does temporal variation in fitness among populations, but mortality during dispersal selects against dispersal. Because of many opposing selection pressures, habitat fragmentation may select for more or less dispersive individuals depending on particular circ*mstances. 5. SPATIAL DYNAMICS IN NONPATCHY ENVIRONMENTS

The classic population concept and corresponding population models assume that all individuals interact equally with all other individuals in the population. This is called panmictic population structure. On the other extreme is the classic metapopulation concept, which assumes a set of dynamically independent local populations, within which individual interactions take place. Many real populations have intermediate spatial population structures: individuals do not occur in discrete local populations, but the population is more continuous across a large area, yet ecological interactions and dispersal are more or less localized; hence, the population is not panmictic at a large spatial scale. In this case, what matters most is the local density experienced by individuals rather than the overall density of individuals in the large population as a whole.


Population Ecology

Population dynamics across a landscape is determined by the opposing forces of localized interactions, which tend to increase differences in local dynamics and dispersal among neighboring population units, which reduces their dynamic independence. In single panmictic populations, strongly nonlinear dynamics may lead to population cycles and other complex dynamics. In large populations distributed across a landscape, nonlinear dynamics and dispersal may generate complex spatially structured dynamics. For instance, the dynamics of measles in human populations may exhibit traveling waves, in which epidemics initiated in large core populations (cities) lead, with some time lag, to epidemics in the surrounding smaller communities (see chapter II.9). Comparable complex spatiotemporal patterns of population density have been detected in some animal populations.

capacity lM replaces the fraction of suitable patches h in the Levins model, and the threshold condition for metapopulation persistence is given by lM > e=c. lM takes into account not only the amount of habitat in the landscape but also how the remaining habitat is distributed among the individual habitat patches and how the spatial configuration of habitat influences extinction and recolonization rates and hence metapopulation viability. The metapopulation capacity can be computed for multiple landscapes, and their relative capacities to support viable metapopulations can be compared: the greater the value of lM, the more favorable the landscape is for the particular species (figure 3). Reserve Selection

Loss and fragmentation of natural habitats are the most important reasons for the current catastrophically high rate of loss of biodiversity on Earth. The amount of habitat matters because long-term viability of populations and metapopulations depends, among other things, on the environmental carrying capacity, which is typically positively related to the total amount of habitat. Additionally, the spatial configuration of habitat may influence viability because most species have limited dispersal range, and hence, not all habitat in a highly fragmented landscape is readily accessible, giving rise to the extinction threshold (see Levins Model above).

Setting aside a sufficient amount of habitat as reserves is essential for conservation of biodiversity. Reserve selection should be made in such a manner that a given amount of resources for conservation makes a maximal contribution toward maintaining biodiversity. In the past, making the optimal selection of reserves out of a larger number of potential reserves was typically based on their current species richness and composition, without any consideration for the long-term viability of the species in the reserves. More appropriately, we should ask which selection of reserves maintains the largest number of species to the future, taking into account the temporal and spatial dynamics of the species and the predicted changes in climate and land use. Metapopulation models can be incorporated into analyses that aim at providing answers to such questions.

Consequences of Habitat Loss and Fragmentation


Metapopulation models have been used to address the population dynamic consequences of habitat loss and fragmentation. In the Levins model, where there is no description of landscape structure, habitat loss has been modeled by assuming that a fraction 1 h of the patches becomes unsuitable for reproduction, while fraction h remains suitable. Such habitat loss reduces the colonization rate to cp(h p). The species persists in a patch network if h exceeds the threshold value e/c. At equilibrium, the fraction of suitable but unoccupied patches (h p*) is constant and equals the amount of habitat at the extinction threshold (h ¼ e=c). The spatially realistic metapopulation model (see above) combines the metapopulation perspective of the Levins model with a description of the spatial distribution of habitat in a fragmented landscape. In the model described by equation 5, the metapopulation

Metapopulations are assemblages of local populations inhabiting networks of habitat patches in fragmented landscapes. The local populations have more or less independent dynamics because of their isolation, but complete independence is prevented by large-scale similarity in environmental conditions and by dispersal, which occurs at a spatial scale characteristic for each species. Metapopulation models are used to describe, analyze, and predict the dynamics of metapopulations. Important questions include the conditions under which metapopulations may persist in particular patch networks and for how long, how landscape structure influences metapopulation persistence, and the response of metapopulations to changing landscape structure. Metapopulation dynamics in highly fragmented landscapes involves an extinction threshold, a critical amount, and spatial configuration of habitat


Metapopulations that is necessary for long-term persistence of the metapopulation. FURTHER READING Hanski, I. 1999. Metapopulation Ecology. Oxford: Oxford University Press. Hanski, I., and O. E. Gaggiotti, eds. 2004. Ecology, Genetics, and Evolution of Metapopulations. Amsterdam: Elsevier. Hanski, I., and O. Ovaskainen. 2000. The metapopulation capacity of a fragmented landscape. Nature 404: 755 758. Hanski, I., and O. Ovaskainen. 2002. Extinction debt at extinction threshold. Conservation Biology 16: 666 673.


Ovaskainen, O., and I. Hanski. 2002. Transient dynamics in metapopulation response to perturbation. Theoretical Population Biology 61: 285 295. Ovaskainen, O., and I. Hanski. 2003. Extinction threshold in metapopulation models. Annales Zoologici Fennici 40: 81 97. Tilman, D., R. M. May, C. L. Lehman, and M. A. Nowak. 1994. Habitat destruction and the extinction debt. Nature 371: 65 66. Verheyen, K., M. Vellend, H. Van Calster, G. Peterken, and M. Hermy. 2004. Metapopulation dynamics in changing landscapes: A new spatially realistic model for forest plants. Ecology 85: 3302 3312.

II.5 Competition and Coexistence in Plant Communities Ray Dybzinski and David Tilman OUTLINE

1. Introduction to competition and stable coexistence 2. Competition for nutrients 3. Competition for light 4. Competition and temperature 5. From models to reality: Future challenges Numerous species commonly compose natural plant assemblages from the poles to the equator, and a wealth of classic ecological experiments have demonstrated that these often compete strongly with one another for resources such as nutrients or light. Theoretical ecologists have demonstrated that competing species are only expected to stably coexist (i.e., coexist in the long run) when each is protected from local extinction by density- or frequency-dependent processes that benefit it when rare, or equivalently, when the net negative effects of intraspecific (within-species) competition exceed those of interspecific (between-species) competition. Competition for nutrients, such as nitrogen and phosphorus, is ‘‘size symmetric’’ because smaller and larger individuals are potentially equal competitors on a per-biomass basis. In contrast, shoot competition for light is ‘‘size asymmetric’’ because taller individuals are advantaged irrespective of biomass. We describe the different modeling approaches that this difference requires. However, in either case, stable coexistence requires trade-offs such that species that are better competitors for one limiting nutrient or in one light environment are necessarily worse competitors for a second limiting nutrient or in a second light environment. As one important example of how these models might account for the stable coexistence of numerous species across landscapes, we consider the effects of habitat heterogeneity in mean growing season temperature coupled with trade-offs in performance among species at different mean temperatures. Finally, given our

theoretical understanding, we close with a discussion of the current challenges to and opportunities for advancing our empirical understanding of competition and coexistence in the real world.

GLOSSARY coexistence. The indefinite persistence of two or more

species. The empirically relevant sort of coexistence is termed ‘‘stable coexistence’’ in which species will continue to persist in the face of perturbations in their abundances. It is important to note that species that co-occur may or may not be stably coexisting; it is possible that one or more species are on their way to local extinction at a time scale that might appear slow to a casual observer. competition. Most broadly, an interaction between individuals in which neither benefits. Here, we are considering exploitation competition for limiting resources in which the resource consumed or intercepted by one individual is no longer available to the second individual, thereby decreasing its fitness. exclusion. A condition in which a species is driven to local extinction as a consequence of a competitive interaction. founder control. A condition in which the dominant species in a competitive interaction is the species that is initially most abundant. interspecific competition. Competition among individuals of different species. intraspecific competition. Competition among individuals of the same species. invader/invasion. In the context of theoretical ecology, an invader is a species introduced at arbitrarily small abundance to a habitat of a resident species at equilibrium. The question is asked: will the invader increase in abundance? Note that this use of the term is different from the sense in which an ‘‘invader’’

Competition and Coexistence: Plants may be a foreign species with negative ecological consequences. local extinction. A condition in which a species is no longer present within a defined habitat area. Local extinction is very different from the common use of the term ‘‘extinction,’’ in which a species is no longer present anywhere. resource. Broadly, something that may be consumed by one individual such that it is no longer available to another organism. Relevant resources for plants are nutrients, such as nitrogen, phosphorus, potassium, and micronutrients, along with light, water, and carbon dioxide. 1. INTRODUCTION TO COMPETITION AND STABLE COEXISTENCE

Plant species, be they trees in tropical or temperate forests, forbs in prairies or tundra, or algae in lakes or oceans, frequently compete with other plant species for various limiting resources, such as nitrate, phosphate, and light (Harper, 1977). Plants are also impacted by interactions such as mutualism, predation, herbivory, and disease. Each of these interactions has the potential to influence both the types of habitats in which species occur and their abundances in those habitats. Here we focus on mechanisms of competition for nutrients and light and their influence on coexistence and competitive exclusion among competing plant species. As developed in the pioneering work of Lotka and Volterra, competition between two species, or interspecific competition, is defined as an interaction in which an increase in the abundance of one species causes a decrease in the growth and abundance of the other species and vice versa. By removing neighboring plants from around a target plant and documenting an increase in the target plant’s growth rate, numerous experimental studies have shown that interspecific competition can be a strong force in plant communities. Competition between two individuals of the same species, intraspecific competition, is similarly defined and is almost certainly as strong as or stronger than interspecific competition. Competition occurs because all vascular plants require light, water, and 20 or so mineral nutrients, including N, P, K, Ca, Mg, and various trace metals to survive and grow (e.g., Harper, 1977). As plants grow and consume these resources in a particular habitat, one or more resources become limiting because their rates of supply are insufficient to meet the demand from consumption. The outcome of the resultant resource competition may be exclusion or coexistence and depends on the dynamics of resource supply, on the resource requirements of the various species, and


on the dependence of these factors on the physical attributes of the habitat, including its temperature, humidity, soil pH, slope, and aspect. The outcome of competition may also depend on initial species abundances. If each species maintains dominance whenever it is initially dominant, competition is said to be ‘‘founder controlled.’’ Species will coexist if each species can increase when it is rare and other species are at equilibrial abundances. This generally requires that intraspecific competition be stronger than interspecific competition. The mere co-occurrence of two or more competing species does not necessarily mean that they are coexisting; some of these species may be on their way to local extinction, albeit too slowly to be detected by casual observation. Ecological theory provides a way to formally distinguish stable coexistence from competitive exclusion at any time scale. First, if a competition model has a multispecies equilibrium point at which competing species have abundances greater than zero, the stability of this point can be assessed mathematically. If species abundances return to the equilibrium point after abundances are perturbed away from that point, this indicates that coexistence is stable. However, it is also possible for two or more competing species to persist indefinitely even when they do not have a stable multispecies equilibrium point. In these cases, one way to test for stable coexistence is to determine if there is ‘‘mutual invasibility’’ of species within the model (note that ‘‘invasion’’ in this chapter has no connotation of processes related to problematic nonnative species). In the simple case of two species, A and B, one asks if the geometric mean growth rate of a relatively small number of As in a monoculture of Bs is positive. If so, species A is said to be able to invade species B. If B can also invade A, then each species increases when rare and is protected from local extinction. The two species stably persist (coexist), even though their abundances are not constant. If A can invade B but B cannot invade A, then species A will exclude species B. If neither species can invade an equilibrial population of the other, then there is founder control; the initial dominant will maintain dominance. 2. COMPETITION FOR NUTRIENTS

Although it is possible for many nutrients to limit growth, it is easiest to begin by considering a single consumer species and a single limiting nutrient (hereafter referred to more generically as a ‘‘resource’’). A resource is considered limiting if increases/decreases in its supply lead to increases/decreases in the growth rate of the species (Tilman, 1982). To understand

Population Ecology

1 dB ¼ f [R] m: B dt


A. f [R] Per capita growth rate, f [R], and loss rate, m

dynamics, we must consider the factors that control both the rates of growth and loss of a species and the rates of supply and consumption of the resource. Because many plants have indeterminate growth and must increase in mass by two to six or more orders of magnitude when growing from a seed to an adult plant, it is customary to consider the dynamics of plant biomass rather than the dynamics of the number of individuals (as often done for animal populations). Thus, the specific rate of growth or loss of plant biomass is represented by dB=dt 1=B, where B is biomass per unit area (for terrestrial plants) or per unit volume (for marine and freshwater planktonic algae). The specific growth rate of a species is often a monotonically increasing but asymptotically saturating function, f[R], of the concentration or level of the limiting resource, R (figure 1A). Plant species experience a variety of sources of loss, including mortality and tissue loss from senescence, damage, herbivory, and disease. These losses, m, can be expressed in terms of density-independent changes in plant biomass, giving equation 1:

loss > growth


growth > loss

Resource concentration, R B.

As a plant population grows, it will consume the limiting resource. To see this, consider the dynamics of the resource:

fA [R] fB [R] mA mB


R*B Resource concentration, R



where S is the supply rate of all forms of resource R, a is the rate of conversion of currently plant-unavailable forms of this resource (such as, for example, soil organic compounds containing nitrogen) into a plantavailable form (such as nitrate), and Q is the quota of this resource needed to make a unit of new biomass (e.g., the ratio of nitrogen in plant tissues to biomass). Equation 2 thus states that the rate of change in available resource, R, depends on its rates of supply, a(S R), and consumption, QBf[R]. In combination, equations 1 and 2 state that, at equilibrium, a species growing by itself will reach a biomass at which growth and loss balance each other, at which resource supply and consumption balance each other, and thus, at which dB=dt ¼ dR=dt ¼ 0. This occurs when the balance between supply and consumption has reduced the level of the resource down to R* (figure 1A; Tilman, 1982). R* is the resource requirement of this species in the sense that it is the amount of resource at which resource-dependent growth balances all sources of loss (figure 1A). If the

A Biomass (A and B) or resource concentration (R)

dR ¼ a(S R) QBf [R], dt


Per capita growth rate, f [R], and loss rate, m





Time Figure 1. Details of resource consumption and competition for a single nutrient. (A) A single plant species growing on a single resource will reach equilibrium at resource concentration R*, where per capita growth, f [R], exactly balances per capita losses, m. (B) Depending on their physiology, morphology, and ecology, different species (e.g., Species A and Species B) are expected to have different resource dependent growth rates and loss rates. As a consequence, different species are expected to have different R* values for a given resource. (C) When species compete for a single resource, the species with the lowest R* value is expected to win. The species with the lowest R* value reduces equilibrial resource concentrations to a level where the loss rates of its competitors exceed their growth rates, and they are driven locally extinct.

Competition and Coexistence: Plants level of the resource were greater than R*, the abundance of the species would increase (dB=dt > 0) because growth would be greater than loss. Greater plant abundance would ultimately reduce the level of the resource to R*. If the level of the resource were less than R*, the abundance of the species would decrease (dB=dt < 0) because loss would be greater than growth. Less plant abundance would decrease resource consumption and ultimately increase the level of the resource to R*. When two or more species compete for a single limiting soil resource, the long-term outcome of their competition, should an equilibrium be reached, is that the species with the lowest R* value (as determined by its physiology, morphology, and environmental conditions and their combined effects on f[R] and m; figure 1B) would win and exclude all other competitors (figure 1C). This occurs because the species with the lowest R* value would continue to increase in abundance until it had reduced the resource down to its R*. At this resource level, the growth rates of all other species would be less than their loss rates, causing them to decline in abundance until locally extinct. Thus, if there is only one limiting resource, coexistence is not possible at equilibrium; only a single species will persist. There are two important points to be made about this result. First, note that the model is predictive in the sense that a species trait measured in monoculture (R* or physiological or morphological traits, such as root mass and fine root diameter, correlated with R* in a given environment) allows one to predict the outcome of competition. This is much different from phenomenological models, such as the Lotka-Volterra competition equations, in which the competition coefficients that might allow one to predict the outcome of competition are measurable only after a competition experiment has been performed (thereby making a prediction of the outcome unnecessary!). Second, note that the model’s prediction that one species should displace all is at odds with the coexistence of numerous competing species in almost all natural ecosystems. This discrepancy suggests that one or more of the simplifying assumptions used in the model are violated in nature. This would not be surprising because the model implicitly assumes that there is only one limiting resource; that sites within a habitat do not differ in physical factors that may also limit growth, such as temperature, soil pH, and humidity; that plant abundances are not limited by dispersal; that interspecific interactions go to equilibrium; and that organisms on other trophic levels that might exert density- or frequency-dependent loss rates, such as


herbivores, seed predators, pathogens, and mutualists, are unimportant. More complex and realistic theories have shown that addition of any one or more of these complexities of nature can allow many species to coexist, but only if species have appropriate trade-offs, as discussed below. Competition for Two Limiting Nutrients

Few ecosystems have but a single limiting nutrient. Nutrient addition experiments in freshwater and marine ecosystems have shown that two or more nutrients often limit phytoplankton, especially nitrate, phosphate, silicate, and/or iron. The most commonly limiting nutrients in most terrestrial ecosystems are soil nitrogen and phosphorus. Other nutrients, especially potassium, calcium, or trace metals, can be limiting depending on the age and origin of the soil. Although not a nutrient per se, water is also limiting in many terrestrial ecosystems. In all but very nutrient poor or dry terrestrial ecosystems and in many freshwater and marine ecosystems, light also limits at least some plant species. We address the special case of light limitation in a separate section below. Although a single number, R*, can summarize the competitive ability of a plant for a single limiting resource, resource-dependent growth isoclines are needed to do so for two or more limiting resources. A resourcedependent growth isocline defines the concentrations of all limiting resources for which a species has a given growth rate. The isocline defining the growth rate that is equal to the loss rate of a given species (i.e., dB=dt ¼ 0), which is called the nullcline, is directly analogous to R* in its ability to predict the outcome of competition for two or more limiting resources. Most commonly, nullclines are plotted on graphs in which the axes represent the possible concentrations of limiting resources (figure 2A). When resource concentrations fall between the origin and a species’ nullcline, the species will experience greater loss than growth. When resource concentrations fall beyond the nullcline relative to the origin, the species will experience greater growth than loss (figure 2A). The shape of the nullcline reflects the type of resources that are limiting. Most plant resources are nutritionally essential; individual plants must have N, P, K, etc., and they cannot eliminate their need for one of these elements by substituting increased amounts of a different element. Nutrients tend to be perfectly essential for morphologically simple plants, such as singlecelled phytoplankton species. This means that the nullclines of algae tend to have right-angle corners (figure 2A), with each branch of the isocline representing the R* of the species for that resource. However, because higher

Population Ecology

190 A.

dA = 0 dt


growth > loss


loss > growth

R* 1




dA = 0 dt

R1 C. dB = 0 dt


dA = 0 dt

R A,2*

R B,2*

R A,1*

R B,1*

plants can vary their morphology and the physiology of different tissue types, they have some ability to substitute one resource for the other when both are approximately equally limiting. Thus, they have interactive-essential resources: their nullclines tend to have rounded corners (figure 2B). The difference between perfectly essential and interactive-essential resources has no major qualitative effects on the outcomes of interspecific resource competition described below. Two species can stably coexist at equilibrium in a hom*ogeneous habitat only if their nullclines cross, which occurs only if the species have an interspecific trade-off. For instance, consider species A and B of figure 2C. Species A has a lower requirement (i.e., a lower R* value) for R1 than species B, but species B has a lower requirement for R2 than species A. The twospecies equilibrium point at which the nullclines cross shows the levels to which these two species would reduce R1 and R2 if they were to stably coexist at equilibrium. Although nullclines must cross for two species to coexist at equilibrium, that alone is not sufficient to ensure coexistence. The outcome of competition is also determined by their rates of consumption of the resources. Assuming that each species forages optimally for these two resources, such that no resources are consumed that do not increase growth (i.e., no ‘‘luxury consumption,’’ a reasonable assumption given that resource consumption requires energy), each species will consume resources in the ratio defined by the slope of the line drawn from the corner of its nullcline to the origin. Hence, consumption may be represented by vectors, cA and cB, that take those slopes (figure 3A). Consider the habitat represented by the supply point (S1, S2) in figure 3A. Such habitats have intermediate rates of supply of the two limiting resources. At equilibrium, the consumption by the two competitors would reduce resources down to the two-species equilibrium point. Indeed, coexistence would occur for any habitats with supply points within the area bounded by the shaded triangle that extend the slopes of the consumption vectors of the two species away from the


Figure 2. Details of resource consumption for two nutrients. (A) A nullcline defines the concentration of two resources for which a plant species will be in equilibrium, i.e., where dA/dt ¼ 0 and per capita growth is exactly balanced by per capita loss. The nullcline shown depicts the response of a species to perfectly essential re sources, such as nitrogen and phosphorus, where no amount of one resource can compensate for a deficit of the other resource. (B) A bent nullcline depicts the response of a species to interactive essen tial resources, where plants can somewhat shift investment toward physiological, morphological, or ecological structures that capture the most limiting resource. For example, a plant may invest less pho tosynthate in mycorrhizal associations and more photosynthate in

the construction of fine roots when nitrogen is more limiting than phosphorus. Characterizing resources as interactive essential in stead of perfectly essential does not change the qualitative outcome of competition. (C) The intersection of the nullclines of two species is necessary, but not sufficient, for their coexistence. Their intersection implies that one species has a lower R* value for one resource and that the other species has a lower R* value for the other resource. If coexistence does occur (see figure 3), equilibrial resource concentra tions will be fixed at the intersection point. When the nullclines do not intersect (not shown), the species with lower R* values for both resources will always win in competition, as in the case for a single resource.

Competition and Coexistence: Plants A. coexist S2





S1 R1 B. S2





cA S1 R1 C. S2

A wins




two-species equilibrium point (figure 3). This is because it is possible to create a linear combination of cA and cB that is exactly equal in magnitude and opposite in sign to the rate of resource replenishment, represented by the vector U, such that resource supply is exactly balanced by consumption (compare figures 3A and 3B) and dR=dt ¼ 0. In habitats that have low rates of supply of R1 relative to R2, such as for the habitat depicted in figure 3C, both species are most limited by R1, and species A, which is the better competitor for R1 because it has a lower R1*, would competitively displace species B at equilibrium. Notice that because their slopes are fixed by assumption of optimal foraging, there is no linear combination of cA and cB in figure 3C that could exactly balance U, as in figures 3A and 3B. Similarly, species B would win for habitats that have low relative rates of supply of R2. If species have trade-offs in their abilities to compete for two or more limiting resources, these trade-offs would mean that each species would have habitats, defined by supply rate of these resources, in which it would persist in competition with any other species. This is illustrated for four species (C, D, E, and F) competing for two limiting resources in figure 4A. Note that, with two limiting resources, only two species can stably coexist, at equilibrium, in any given habitat (i.e., any specific S1, S2) but that all four species can persist together in a landscape that has spatial heterogeneity in the supply rates of these resources as shown. Even more species could coexist with a fixed amount of heterogeneity if new species invaded with intermediate requirements for R1 and R2, an example of which is shown using a dashed nullcline in figure 4A. Alternatively, more species could coexist if the habitat encompassed even greater heterogeneity (not shown). The concept of trade-offs can easily be expanded to three or more resources. 3. COMPETITION FOR LIGHT

cA S1 R1 Figure 3. The outcome of competition for two nutrients. For all three panels: Species A’s nullcline is solid and Species B’s nullcline is dashed. The vector cA represents Species A’s rate of consumption vector. By assumption of optimal foraging, its slope is fixed by the ratio of resource concentrations that limit Species A, i.e., the slope between the origin and the bend in Species A’s nullcline. Its magni tude is proportional to Species A’s equilibrial biomass. The vector cB is analogous for Species B. The vector U represents the rate of re source replenishment. The point (S1, S2) represents the supply concentration of the two resources in a given habitat. At equilibrium,

The model of competition described above applies to competition for nutrients but not to competition for light (Tilman, 1988). To see why, first realize that the U will point toward (S1, S2) and will be equal and opposite to the vector sum of the consumption vectors cA and cB. Because their slopes are fixed, coexistence will occur at equilibrium only when it is possible to add cA and cB from the nullcline intersection point such that their sum is equal and opposite to U. This will occur for all (S1, S2) inside the medium gray triangle (panels A and B). For all (S1, S2) inside the lightest gray area (panel C), Species A will win (cB ¼ 0). For all (S1, S2) inside the darkest gray area (not shown), Species B will win (cA ¼ 0). For all (S1, S2) inside the white area (not shown), both species will become locally extinct.

Population Ecology



A. A

















H S1 R1

Temperature-dependent R*




A wins

B wins

C C wins

concentration to which a plant reduces a nutrient through consumption (R) is the same concentration at which it must subsequently extract the nutrient (R). In other words, the nutrient concentration it ‘‘gets’’ is the same as the nutrient concentration it ‘‘creates.’’ This is fundamentally different from competition for light: a tall tree intercepts full sunlight and reduces the intensity of light below its crown, such that trees below it receive less than full sunlight. Thus, the light level that the tall tree ‘‘gets’’ is distinct from the light level it ‘‘creates.’’ Weiner (1990) described differences in competitive interactions in terms of symmetry and asymmetry. ‘‘Size symmetric’’ competition occurs when larger individuals have no advantage over smaller individuals on a per-biomass basis, i.e., when the resource consumption of an individual is proportional to the biomass of its resource-capturing structures, such as roots for soil nutrients. ‘‘Size asymmetric’’ competition occurs when larger individuals take a disproportionate share of the resource, as is expected when taller plants preempt the light of shorter plants below them. Indeed, empirical studies have generally supported the idea that competition for soil resources is size symmetric, whereas competition for light is size asymmetric. Given size asymmetry, it is challenging to create a realistic mathematical model of light competition that is as simple and readily interpretable as the nutrient model described above.

Mean growing season temperature

The Perfect Plasticity Approximation Temperature-dependent Z*

C. C B A

A wins

B wins

C wins

Recently, S. W. Pacala and co-workers discovered an approach that allowed them to create a simple yet mechanistic and predictive analytical model of plant competition for light. Although their model is formulated explicitly for single-canopy forests, it is extendable to other plant communities. The model treats trees as morphologically plastic, as being able to ‘‘lean’’ toward gaps by concentrating their growth in areas of higher light, where the photosynthetic return on structural investment is greater. To a first approximation, termed the ‘‘perfect plasticity approximation’’ (PPA), the result of each tree’s individual plasticity and

Mean growing season temperature Figure 4. The effect of heterogeneity in resource supply points and temperature on coexistence. (A) With two limiting resources, at most two species can stably coexist at a single resource supply point (S1, S2). However, if supply points shift across a landscape (as depicted by the circle and the range of S1 and S2), many species can coexist, provided a trade off causes a lowered R * value for one resource to necessarily lead to an increased R * value for the other resource. Light gray regions depict those for which supply points will cause a single species to dominate. Dark gray regions depict those for which supply points will cause two species to stably coexist. Gray letters represent

species nullclines. Black letters represent the supply point zones in which a species will persist. The dashed nullcline shows an example of a new species that would be able to invade and coexist in this landscape. (B and C) Hypothetical dependence of R * and Z * values on mean growing season temperature. Ecophysiological constraints suggest that those characteristics that will lead to low R * values or relatively high Z * values at one temperature should prevent a species from having a low R * value or relatively high Z * value at a different temperature. Thus, heterogeneity in growing season temperature across a landscape should lead to the coexistence of multiple species.

Competition and Coexistence: Plants phototropism is canopy closure via tessellation, where each crown fills as much space as it can without growing into adjacent crowns. Moreover, under the PPA, the height above the ground at which each canopy crown touches an adjacent canopy crown tends toward a constant value across a forest canopy with a given set of species at a given stage of development. Thus, forest trees may be classed as either in the canopy or in the understory and consequently subject to either full sunlight or the mean light remaining after transmission through the canopy. This has the dual advantage of simplifying the mathematics of a light competition model and yet making it more realistic (Adams et al., 2007). Consider a forest monoculture of species A at equilibrium. The forest consists of reproductive canopy trees and nonreproductive understory trees. The canopy trees reduce light to a level, LA, in the understory. Because the forest is assumed to be in equilibrium, each canopy tree exactly replaces itself over a lifetime of reproductive effort, on average. Understory trees grow at rate GD,A[L] and die at rate mD,A[L]. Canopy trees grow at rate GL,A, die at rate mL,A, and reproduce at rate FA per unit area of crown. Crown area, p(aAD)2, and height, HA(D)^bA, scale allometrically with trunk diameter, D. The equilibrial ˆ A*. canopy height is Z Invasion of species A’s monoculture by a small number of species B individuals will be successful only if, on average, each invader produces more than one successful (i.e., reproductive) offspring. By assumption of the Adams et al. (2007) model, individuals reproduce only once they have reached the canopy. Thus, if some of the invaders die before reaching the canopy, successful invasion will depend on whether those individuals that do reach the canopy can produce enough successful offspring to make up for their loss. Resident species A can influence this process only by affecting the fraction, z, of invaders that survive to reach the canopy:


mD, B [LA ] GD, B [LA ]

^ 1=bB ZA HB



Adams et al. (2007) point out that invasion will be less likely if LA is smaller (by decreasing the invader’s understory growth rate, GD,B[LA], and increasing ˆ A* is greater (by inits mortality rate, mD,B[LA]) or if Z creasing the time required to reach the canopy and thus the likelihood of mortality before reaching the canopy). If the crown transmissivities of the two species are equivalent, i.e., LA ¼ LB , then Adams et al. (2007) demonstrate that light competition among trees is analogous to competition for a single soil resource,


in the sense that one species is always expected to win, and the winner may be determined by examining the traits of the species in equilibrial monoculture. Specifically, the winner will be the tallest ˆ i*, apspecies, with the canopy height of species i, Z proximated by: ˆ i &Hi Z

#!bi ! " 2pa2i Fi G2L, i GD, i [Li ] ln : mD, i [Li ] m3L, i


Foresters have long recognized that forest succession, which may be viewed as prolonged competitive exclusion, generally proceeds from shade-intolerant species to shade-tolerant species. Purves et al. (2008) showed that shade-intolerant species tend to have low Zˆ*, that shade-tolerant species tend to have high ˆ *, and that succession in the forests near the Great Z Lakes predictably proceeded from species with low Zˆ* to species with high Zˆ*. This suggests that, relative to the importance of other factors that affect light competition during succession, the crown transmissivities of different tree species may be effectively equivalent. Clearly, numerous tree species stably coexist in most forests, an observation that is at odds with the prediction that the single species with the highest Zˆ* should exclude all competitors. Again, this suggests that some of the model’s assumptions, which largely overlap with the assumptions of the simple model of competition for a single soil resource described in the previous section, are false. However, before introducing exogenous factors into the model to explain diversity, Adams et al. (2007) note that at least some coexistence may be explained if we relax the assumption that the transmissivities of different tree species are equal, i.e., let LA 6¼ LB . Adams et al. (2007) showed that it is possible for two species to coexist when competing for light if each species performs less well in the understory light environment that its canopy creates. Specifically, two species will coexist via competition for light when the species that casts more shade performs relatively better under less shade and the species that casts less shade performs relatively better under more shade. Each species creates the conditions that disfavor it. As discussed earlier, species are expected to stably coexist in resource-limited conditions when intraspecific competition is stronger than interspecific competition. Further theoretical and experimental work is needed to determine how likely this mechanism of coexistence might be relative to mechanisms that involve having both light and one or more additional factors limit plant growth.


Population Ecology


The rates of photosynthesis, respiration, and other physiological processes of all plants are dependent on temperature during the growing season. Temperatures below freezing are a separate but equally important limiting factor for plants because a host of chemical and mechanical adaptations are needed for cells to avoid the damage that freezing can cause. Moreover, the magnitude of such adaptations increases for temperatures further below freezing. Here we consider just the effects of the mean temperature of the growing season on plant growth rates. The net effect of the dependence of plant physiology on temperature is that each plant species should have a mean growing season temperature at which it performs best. For habitats in which a single nutrient is limiting, the R* of each species would be a function of mean growing season temperature, with each species having its lowest R* value at its optimal temperature (figure 4B). For habitats in which light is limiting, equilibrial ˆ *, will vary with temperature because canopy height, Z of the dependence of growth, mortality, and fecundity rates on temperature. Habitats have spatial variation in mean growing season temperature, such as along elevational gradients or among slopes with different aspects (north- versus south-facing slopes), and temporal variation, such as year-to-year or decade-to-decade differences. Such variation can allow numerous species to coexist. For example, the temperature-dependent R* values of figure 4B define the temperature ranges at which each species would be the superior competitor (figure 4B). Species A would dominate the coolest sites/times, species B would dominate warmer sites/times, and species C would dominate the warmest sites/times. Strictly interpreted, only a single species would persist in a given site, but all three species would stably coexist on the larger landscape that encompassed the temperature variation ilˆ * values of diflustrated in figure 4B. Similarly, the Z ferent species are expected to be maximal at a temperature for which they are well adapted, relative to other species at that temperature. Thus, across a landscape with variation in mean growing season temperature, the relative position of Zˆ* values would shift such that species A might win at lower temperatures, whereas species B might win at higher temperatures (figure 4C). This coexistence occurs because plants, like all organisms, are constrained by trade-offs. No species has the lowest R* value or highest Zˆ* value for all temperatures. Rather, each species has a range of temperatures at which it is the superior competitor and a range for which it is inferior. Just such a trade-off is

expected based on the temperature dependence of both the activation energy and the denaturation of enzymes. If such a trade-off were an unavoidable aspect of plant life, there could be many other species, each with its own optimal temperature, that could persist between species A and B (figure 4B,C). Indeed, there would be no simple limit to the number of competing species that could coexist on a single limiting resource, be it nutrients or light, in a landscape with spatial variation in growing season mean temperature. 5. FROM MODELS TO REALITY: FUTURE CHALLENGES

The models described above provide several mechanisms, each of which, by itself, can maintain a potentially infinite number of stably coexisting species. The ecological literature is replete with other coexistence mechanisms that incorporate effects of space; disturbance and other nonequilibrium dynamics; or natural enemies on coexistence (see Further Reading). Given the appropriate trade-offs between species, most of these mechanisms can also maintain a potentially infinite number of stably coexisting species. The current and future challenge for ecology is determining which of these possible mechanisms actually maintain plant species diversity in nature. This is difficult for at least two reasons. First, feedbacks between plants and their abiotic and biotic environment and between soil biota and their abiotic and biotic environment make it difficult to isolate or manipulate only those aspects of the community or environment that theory suggests should impact some observable characteristic. For instance, it is difficult to remove specialized insect herbivores to test for their importance in diversity maintenance without also removing other insects, some of which may be mutualists. The second difficulty lies in the long time scales over which many mechanisms are expected to operate, either because the dynamics are slow or because the mechanism relies on periodic but rare events. These defy detection in short-term field studies. As has happened in geology, substantial progress might be made if theoretical ecologists apply themselves to articulating short-term signatures of long-term mechanisms that field ecologists could then look for or test for. However, this appears more challenging in ecology than in geology because many of the most obvious signatures (e.g., particular rank abundance relationships) may be generated by multiple mechanisms. Alternatively, ecologists might turn to long-term data. For example, Purves et al. (2008) used long-term Forest Inventory and Analysis data (available from the U.S. Department of Agriculture) to parameterize and then test a variant

Competition and Coexistence: Plants of the light competition model described above and showed that the model largely predicted changes in total basal area, species composition, and the size structure of forest stands over 100 years of secondary succession. Unfortunately, such relevant long-term data are rare in ecology. Despite these difficulties, many studies have demonstrated that particular mechanisms of coexistence may operate in particular habitats (Fargione and Tilman, 2002), typically by altering some aspect of the habitat and observing the predicted response. Such studies provide important insights into the ways particular mechanisms of coexistence operate and, to the extent that publications are not biased against negative results, provide data for broader conclusions via metaanalysis of multiple studies across a range of habitats and conditions. Although it might be intellectually gratifying to learn that just one mechanism maintains diversity throughout all of the varied biomes of the world, it seems much more likely that multiple mechanisms operate in any given habitat and that the relative importance of different mechanisms shifts across biomes and environmental gradients. In all cases, though, theory suggests that interspecific trade-offs will be both a root cause of coexistence of competing plant species and a useful signature of the mechanisms responsible for coexistence. FURTHER READING Adams, T. P., D. W. Purves, and S. W. Pacala. 2007. Un derstanding height structured competition in forests: Is there an R* for light? Proceedings of the Royal Society B: Biological Sciences 274: 3039 3047. Chesson, P. 2000. Mechanisms of maintenance of species diversity. Annual Review of Ecology and Systematics 31: 343 366. This review differentiates between the ‘‘equalizing’’ and ‘‘stabilizing’’ components of coexistence mechanisms. Chesson’s distinction has received consider able attention in the recent literature, especially in the debate between niche and neutral processes. See, for in stance: Adler et al. 2007. A niche for neutrality. Ecology Letters 10: 95 104. Fargione, J., and D. Tilman. 2002. Competition and coexis tence in terrestrial plants. In U. Sommer and B. Worm,


eds. Competition and Coexistence. Berlin: Springer Verlag, 165 206. This review has a good discussion of the limited number of studies that have successfully identified mechanisms of coexistence operating in natural systems. Harper, J. L. 1977. Population Biology of Plants. London: Academic Press. This volume is a classic synthesis of early work on plant competition. Huisman, J., and F. J. Weissing. 1999. Biodiversity of plankton by species oscillations and chaos. Nature 402: 407 410. This article is an elegant account of how the number of coexisting species may greatly exceed the number of limiting resources. The authors use a variant of the soil resource model presented above and find that the dynamics of competition lead to chaos and, as a result, species coexistence. Miller, T. E., J. H. Burns, P. Munguia, E. L. Walters, M. M. Kneitel, P. M. Richards, N. Mouquet, and H. L. Buckley. 2005. A critical review of twenty years’ use of the resource ratio theory. The American Naturalist 165: 439 448. See also a reply by Wilson et al. and rejoinder by Miller et al. in 169: 700 708. Purves, Drew W., Jeremy Lichstein, Nikolay Strigul, and Stephen W. Pacala. 2008. Predicting and understanding forest dynamics using a simple tractable model. Proceed ings of the National Academy of Sciences 105: 17018 17022. This parameterized version of the light competi tion model presented above predicts the 100 year out come of secondary succession in the forests of the Great Lakes states. Tilman, D. 1982. Resource Competition and Community Structure. Princeton, NJ: Princeton University Press. This volume presents a detailed treatment of the soil resource competition model and its implications. Tilman, D. 1988. Plant Strategies and the Dynamics and Structure of Plant Communities. Princeton, NJ: Princeton University Press. This volume presents a detailed treat ment of a simulation model of plant allocation to leaves, roots, etc. and its effect on competition and coexistence. Tilman, D., and S. Pacala. 1993. The maintenance of species richness in plant communities. In R. E. Ricklefs and D. Schluter, eds. Species Diversity in Ecological Commu nities. Chicago: The University of Chicago Press, 13 25. This is a review of coexistence mechanisms. Weiner, J. 1990. Asymmetric competition in plant popula tions. Trends in Ecology and Evolution 5: 360 364. This article presents a clear presentation of the difference be tween ‘‘symmetric’’ competition for soil resources and ‘‘asymmetric’’ competition for light.

II.6 Competition and Coexistence in Animal Communities Priyanga Amarasekare OUTLINE

1. 2. 3. 4. 5. 6.

Introduction Basic principles of competitive coexistence Coexistence under a single limiting factor Coexistence under multiple limiting factors Multiple coexistence mechanisms Summary and conclusions

Competition is the most ubiquitous of species interactions. It occurs any time a resource that is essential to growth and reproduction (e.g., food, shelter, nesting sites) occurs in short supply. The acquisition of the resource by one individual simultaneously deprives others of access to it, and this deprivation has a negative effect on both the fitness of individuals and the per capita growth rates of populations. Competition is thus an interaction that has mutually negative effects on its participants. Coexistence results when populations of several species that utilize the same limiting resources manage to persist within the same locality. This chapter focuses on mechanisms that allow competitive coexistence in animal communities. Animals have two characteristics that determine the kinds of resources they can use and the mechanisms by which they can tolerate or avoid competition for these resources. First, animals are heterotrophs and have to ingest other organisms to obtain the energy required for growth and reproduction; competition thus involves biotic resources. Second, most animals are mobile and hence able to avoid or reduce competitive effects through dispersal.

GLOSSARY density dependence. Dependence of the per capita

growth rate on the abundance or density of the organism in question. exploitative competition. Individuals have indirect negative effects on other individuals by acquiring a resource and thus depriving others of access to it.

functional response. The relationship between per

capita resource consumption and resource abundance. interference competition. Individuals have direct negative effects on other individuals by preventing access to the resource via aggressive behaviors such as territoriality, larval competition, overgrowth, or undercutting. per capita growth rate. Per-individual rate of increase as a result of reproduction, mortality, emigration, and immigration. stable coexistence. Competing species maintain positive abundances in the long term and are able to recover from perturbations that cause them to deviate from their long-term or steady-state abundances. 1. INTRODUCTION

A thorough understanding of the mechanisms of coexistence requires a thorough understanding of the mechanism of competition. Because animals rely on biotic resources which themselves grow and reproduce, the appropriate theoretical framework is one in which the resource dynamics are considered explicitly. Tilman’s resource competition theory, although motivated by plant competition, provides such a framework for animal communities as well. When two or more species are limited by the same resource, the species that can maintain a positive per capita growth rate at the lowest resource level will exclude all other species. This is called the R* rule in exploitative competition (Tilman, 1982). Coexistence mechanisms are the processes that counteract the R* rule. They do so by increasing the strength of intraspecific competition relative to that of interspecific competition. The exact means by which this is achieved is obvious in some cases and quite subtle in others. There are several basic principles that underlie

Competition and Coexistence: Animals all coexistence mechanisms, and a clear grasp of these principles is necessary to understand the more subtle coexistence mechanisms. 2. BASIC PRINCIPLES OF COMPETITIVE COEXISTENCE

First, stable coexistence requires species to exhibit ecological differences. These differences are typically thought of as the species’ niches. Following Chesson (2000), a species’ niche has four dimensions: resources, natural enemies, space, and time. Species could differ in terms of (1) which resources or natural enemies they are limited by, (2) when they use the resource or encounter the natural enemy, or (3) where they use the resource or encounter the natural enemy. Niche differences are essential to coexistence because they allow species to depress their own per capita growth rates more than they do the growth rates of their competitors (Chesson, 2000). Elucidating exactly how this occurs is often difficult, but such an understanding is vital for a mechanistic understanding of coexistence. A useful starting point is the idea of a negative feedback loop. Such a loop can cause a species’ per capita growth rate to decrease when the population size is large and to increase when the population size is small. Negative feedback processes arise naturally when individuals compete with conspecifics (other individuals of the same species) for a limiting resource. Coexistence requires that this self-limiting negative feedback be stronger than the negative effect that the species has on the per capita growth rate of another species it competes with; i.e., when the focal species’ population size is large, effects of resource limitation should affect its survival and reproduction more than it affects the survival and reproduction of its competitor. In what follows, I discuss the various ways in which this can be achieved. I begin with the case of species competing for a single limiting factor, which is the most restrictive case for coexistence. I then discuss cases where species compete for multiple limiting factors. 3. COEXISTENCE UNDER A SINGLE LIMITING FACTOR

The salient point to keep in mind is that coexistence requires mechanisms that counteract the R* rule. When two or more species compete for a single limiting resource, this can occur via two basic classes of mechanisms. The first class of mechanisms enables coexistence via density-dependent negative feedback processes that operate within local communities. These mechanisms can operate in the absence of spatial or


temporal variation in the environment. The second class of mechanisms enables coexistence by allowing organisms to avoid or minimize interspecific competition in space or time. These mechanisms rely on spatial or temporal variation in the abiotic environment. I next discuss these two classes of mechanisms. Mechanisms That Are Independent of Environmental Variation: Intraspecific Interference

Intraspecific interference mechanisms are inherently density-dependent phenomena that have little or no effect on a species’ per capita growth rate when it is rare and a strong negative effect when it is abundant. Such mechanisms therefore enable species to depress their per capita growth rates more than they would the per capita growth rates of other species they compete with. Intraspecific interference mechanisms are widespread in animal communities. In species where mates and/or nest sites are in short supply, territoriality limits per capita reproductive success when population sizes are large. Overgrowth and undercutting in sessile marine organisms have a similar self-limiting effect. In insect parasitoids, direct interference via aggression during oviposition or superparasitism (several females oviposit within the same host, leading to larval competition via direct combat or physiological suppression) reduces the per capita reproductive success of individual female parasitoids and decreases the per capita population growth rate at high density. A similar self-limiting effect can arise even in the absence of direct interference: per capita oviposition success declines with increasing parasitoid density because females keep rediscovering already parasitized hosts. This phenomenon, termed pseudointerference, can allow coexistence if parasitoid species that compete for a common host species have aggregated distributions, or attacks, that are independent of one another (Murdoch et al. 2003). This allows higher encounter rates with conspecifics than with heterospecifics, thus leading to stronger intraspecific than interspecific competition. Mechanisms Dependent on Environmental Variation

When species compete for a single limiting factor, coexistence can occur via mechanisms that depend on environmental variation. Temporal variation enables coexistence by allowing species to differ in terms of when they use a limiting resource, and spatial variation enables coexistence by allowing species to differ in terms of where they use a limiting resource. In both cases, coexistence results because intraspecific competition is concentrated, in time or space, relative to interspecific


Population Ecology

competition. Below I discuss specific examples of spatial and temporal coexistence mechanisms.

Coexistence Mechanisms Driven by Temporal Variation

Nonlinear Competitive Responses Most animal species exhibit nonlinear functional responses. The most common of these is the Type II functional response where the per capita consumption rate saturates at high resource abundances because of long handling times in predators or egg limitation in insect parasitoids. Type II functional responses cause the species’ per capita growth rates to depend on resource abundance in a nonlinear manner. If the resource abundance fluctuates over time, and species that compete for the resource differ in the degree of nonlinearity in their functional responses, stable coexistence is possible if the species with the more nonlinear response is more disadvantaged, when abundant, by resource fluctuations than the species with the less nonlinear response (Armstrong and McGehee, 1980). Large resource fluctuations depress the per capita growth rate of the species with the more nonlinear response. This in turn reduces competition on the species with the less nonlinear response and allows it to invade a community where the former species is resident. The species with the more nonlinear functional response, which has the lower R*, is better at exploiting the resource when resource abundance is lower, and the species with the less nonlinear functional response is better at exploiting the resource when resource abundance is higher. Thus, temporal fluctuations in the resource allow coexistence via temporal resource partitioning. Resource fluctuations can arise via abiotic environmental variation (e.g., seasonal variation in temperature and/or humidity) or via the resource’s interaction with the consumer species with the more nonlinear functional response (Armstrong and McGehee, 1980). Temporal Storage Effect A second type of temporal coexistence mechanism, termed the temporal storage effect (Chesson, 2000), occurs when competing species differ in their responses to abiotic environmental variation. For instance, a species’ developmental period or resource consumption rate may vary depending on seasonal variation in temperature or humidity. Such speciesspecific differences, termed the environmental response (Chesson, 2000), modify the nature of com-

petitive interactions between species. For example, two competing species that differ in their temperature sensitivity may be active at different times of the year, and the resulting reduction in temporal overlap will reduce the intensity of interspecific competition between them. This effect is quantified as the covariance between the environmental response and competition. When species-specific responses to environmental variation modify competition, intraspecific competition is the strongest when a species is favored by the environment, and interspecific competition is the strongest when a species’ competitors are favored by the environment. Thus, the covariance between the environmental response and competition is negative (or zero) when a species is rare and positive when it is abundant. This relationship can be understood as follows. When a species is favored by the environment, it can respond with a high per capita population growth rate (i.e., positive or strong environmental response) and reach a high abundance. When abundant, the species experiences mostly intraspecific competition because the environment is unfavorable to the other species it competes with. Thus, a strong environmental response increases intraspecific competition, resulting in a positive covariance between the environmental response and the strength of competition. In contrast, when the species is not favored by the environment, its per capita population growth rate is low (weak environmental response), and it will remain rare. It will experience mostly interspecific competition (because the environment is now favorable to the species’ competitors), but because it is rare, the strength of competition it experiences will be low. Thus, a weak environmental response reduces interspecific competition, resulting in a zero or negative covariance between the environmental response and the strength of competition. Stable coexistence via the temporal storage effect requires a third ingredient, buffered population growth (Chesson, 2000). Buffered population growth occurs when the decrease in population growth when the abiotic environment is unfavorable is offset by an increase in population growth when the abiotic environment is favorable. This follows from the mathematical phenomenon called Jensen’s inequality, i.e., when the per capita growth rate is a nonlinear function of a life history trait or a vital rate that is subject to environmental variation, the growth rate averaged over the range of environmental variation experienced by the species is different from the average of the growth rates it experiences at different points of the environmental gradient (Ruel and Ayres, 1999). Buffered population growth allows species to tide

Competition and Coexistence: Animals over unfavorable periods resulting from strong interspecific competition and/or abiotic conditions that are not conducive to growth and reproduction. Life history traits that enable buffered population growth in animals include resting eggs in zooplankton species (e.g., Daphnia), high adult longevity in species in which competition occurs at the juvenile stage (e.g., coral reef fish), and dormancy and diapause (e.g., desert rodents) that allow species to be inactive during harsh environmental conditions. Thus, species-specific responses to the abiotic environment ensure that species experience mostly intraspecific competition when they are favored by the environment, and buffered population growth ensures that species experience minimal interspecific competition when they are not favored by the environment. The overall outcome is an increase in the strength of intraspecific competition relative to interspecific competition, and this ensures stable coexistence. The difference between nonlinear competition and a temporal storage effect is that, in the former, the competing species differ in the nonlinearity of their responses to a limiting resource that fluctuates over time (where fluctuations are a response to abiotic environmental variation or a result of the resource’s interaction with a consumer species), whereas in the latter, the species differ in their responses to an abiotic environmental factor that fluctuates over time, which alters the species’ responses to the limiting factor. In the first case, the differences in species’ responses affect competition directly because they respond to the limiting factor itself, whereas in the latter, differences in species’ responses affect competition indirectly because they respond to an environmental factor whose variation determines when species engage in competition. The most obvious biological example of temporal storage effect comes from temporal refuges. Such refuges are common in invertebrates, particularly insects, whose developmental periods, life history traits, and vital rates are all strongly influenced by temperature variation. Such temporal refuges typically arise from seasonal variation in the environment, with species differing in their tolerance of harsh environmental conditions. In insects in particular, species-specific differences are typically manifested as differences in activity periods. This situation is well documented in pest–enemy systems where multiple natural enemy species attack the same pest species. For instance, the leaf-feeding beetle, Galerucella calmariensis, is a successful biological control agent of the invasive plant pest purple loosestrife (Lythrum salicaria), but predation by the omnivorous mirid bug (Plagiognathus politis) disrupts


control; however, the beetle emerges earlier in the season than the bug, and hence has a predation-free window that allows it to establish and inflict severe damage on its host plant. The parasitoids of the olive scale (Parlatoria oleae) Aphytis maculicornis and Cocophagoides utilis engage in intraguild predation and also exhibit differential temperature sensitivity: A. maculicornis is intolerant of warmer temperatures, whereas C. utilis is intolerant of colder temperatures; C. utilis thus has a temporal refuge from intraguild predation during the warm summer months, which promotes coexistence as well as complementary pest control. Coexistence Mechanisms Driven by Spatial Variation

Spatial Storage Effect A spatial storage effect can arise if competing species differ in their responses to spatial variation in the abiotic environment (Chesson, 2000). This can occur if spatial variation changes the species’ R* values such that a species that is the superior competitor in one locality may be an inferior competitor in another locality. When species differ in their responses to spatial variation in the abiotic environment, they grow to high abundances in the habitat patches most favorable to their growth and reproduction and experience strong intraspecific competition. The advantage of favorable environmental conditions is counteracted by strong intraspecific competition, leading to a negative covariance between the environmental response and strength of competition. In contrast, species tend to be rare in habitat patches unfavorable to their growth and reproduction, and they experience mostly interspecific competition (because the habitat patches unfavorable to the focal species tend to be favorable to its competitors). However, the species’ rarity minimizes the impact of interspecific competition, leading to a weak or negative covariance between strength of competition and the quality of the habitat patch. In a spatially varying environment, the fact that per capita growth rates will depend on the habitat patch quality ensures buffered population growth. For instance, the reduction in the per capita growth rate in unfavorable patches (because of low habitat quality and strong interspecific competition) is compensated for by an increase in the per capita growth rate, when the species is rare, in favorable patches. Together, the environmental response (and the resulting covariance between competition strength and habitat quality) and buffered population growth ensure that intraspecific competition is stronger than interspecific competition, and this promotes stable coexistence.


Population Ecology Dispersal-Mediated Coexistence: Source–Sink Dynamics

A spatial storage effect automatically implies the existence of sources and sinks, localities where a species, because of its lower R*, can escape competitive exclusion (sources) as opposed to localities where a species, because of its higher R*, is subject to competitive exclusion (sinks). In such situations, dispersal from source habitats can prevent the exclusion of inferior competitors from sink habitats. Competitive exclusion is prevented because dispersal creates a negative density-dependent effect that increases the strength of intraspecific competition relative to interspecific competition (Amarasekare, 2003). Such dispersal-mediated coexistence is, however, not robust to increases in the dispersal rate. High rates of dispersal eliminate the spatial variation in competitive rankings such that the R* rule operates at the metacommunity scale, and the species that has the lowest R* when averaged across all habitat patches excludes the other species. Moreover, if emigration rates are sufficiently large, source communities themselves can experience negative growth rates, resulting in region-wide exclusion of the species. A key point to appreciate is that coexistence via source–sink dynamics is contingent on there being a mechanism within the source communities that ensures the persistence or coexistence of a given species; i.e., the species should exhibit a self-limiting negative feedback loop in localities that are favorable to its growth and reproduction. The Role of Trade-offs in Coexistence

The role of trade-offs (energetic and other constraints that prevent species from simultaneously doing well on all aspects of growth and reproduction) in coexistence has spawned many a misconception. When two or more species compete for a single limiting resource, trade-offs cannot allow coexistence unless they involve a density- or frequency-dependent negative feedback mechanism. Trade-offs that lack such negative feedback mechanisms can reduce the differences between species in their R* values but cannot provide the kind of negative feedback processes that are necessary for stable coexistence (Chesson, 2000). For instance, a species that is efficient at acquiring resources may do so at the cost of higher density-independent mortality and hence have a higher R* than it would otherwise, but this trade-off will not allow coexistence because it will still competitively exclude (or be excluded by) a less efficient species subject to lower density-independent mortality that has a higher (lower) R*. Trade-offs that involve negative feedback mechanisms arise naturally when species compete for multiple limiting factors.


When species compete for multiple limiting factors, trade-offs in resource use can allow coexistence by increasing intraspecific competition relative to interspecific competition. This occurs because strong dependence of a species on a particular resource (which also leads to resource depletion) creates a negative feedback loop that leads to self-limitation (Chesson, 2000). For instance, a species will enjoy a high per capita growth rate when it is rare and the resource is abundant but suffer a low or negative per capita growth rate when it is abundant and the resource is rare. Thus, if species 1 has a negative feedback loop with resource A and species 2 has a similar feedback loop with resource B, they can coexist even if species 2 does utilize resource A because their overlap in the use of resource A is less that it would be had they not been utilizing two resources. Each species exhibits a trade-off in resource use (i.e., it is a superior exploiter of one resource but at the cost of being an inferior exploiter of another resource), and as long as such trade-offs cause species to limit themselves more than they would their competitors, stable coexistence is possible. Trade-offs between Competitive Ability and Susceptibility to Natural Enemies

Most animals are limited by resources and natural enemies alike. Coexistence is possible if species that share common resources and natural enemies exhibit a trade-off between resource exploitation ability (as determined by their R*s) and susceptibility to a natural enemy (as determined by an equivalent P* rule, i.e., the prey species with the greatest resistance/tolerance to a natural enemy will exclude all other prey species). For instance, two competing prey species can coexist if one is more strongly limited by the resource and the other is more strongly limited by the natural enemy. Coexistence becomes possible because each species has a negative feedback loop with the resource or natural enemy that enables it to depress its own per capita growth rates more than it depresses the per capita growth rates of the other species. 5. MULTIPLE COEXISTENCE MECHANISMS

Species coexistence in most animal communities is likely to result from multiple mechanisms. For instance, parasitoid species that attack the same host species often engage in interference mechanisms, but

Competition and Coexistence: Animals they also exhibit temporal refuges resulting from differential sensitivities to abiotic environmental variation. The key issue in elucidating the operation of multiple mechanisms is the potential for interactions between different types of coexistence mechanisms. Such interactions often involve nonadditive effects and lead to outcomes that are often unexpected and counterintuitive. Thus, it is crucial that empirical investigations of coexistence mechanisms be guided by a theoretical framework that can both distinguish between different coexistence mechanisms and predict the outcomes of interactions between mechanisms. 6. SUMMARY AND CONCLUSIONS

Coexistence in animal communities can occur via a variety of mechanisms, all of which share the property that negative feedback loops between a given species and the resource (or natural enemy) that it is most limited by leads to strong self-limitation, which overwhelms the effects of interspecific competition. The key issue in elucidating coexistence mechanisms in natural communities is to understand the mechanisms by which such self-limitation occurs. It is equally important to be able to distinguish between different types of mechanisms and to understand the nature of interactions between mechanisms that operate simultaneously. The axes provided here, competition for single versus multiple limiting factors and coexistence mechanisms that are independent of versus dependent on abiotic environmental variation, provide the basis for developing comparative predictions that can guide empirical investigations of multiple coexistence mechanisms in animal communities.


I want to acknowledge Peter Chesson, whose ideas have shaped my thinking about coexistence and whose work I have drawn heavily from in writing this chapter. FURTHER READING Amarasekare, P. 2003. Competitive coexistence in spatially structured environments: A synthesis. Ecology Letters 6: 1109 1122. This article synthesizes recent theory on spatial coexistence mechanisms and develops comparative predictions that can be used to distinguish between dif ferent mechanisms. Armstrong, R. A., and R. McGehee. 1980. Competitive ex clusion. American Naturalist 115: 151 170. A classic paper that provides a rigorous mathematical analysis of coexistence via nonlinear competitive responses. Chesson, P. 2000. Mechanisms of maintenance of species diversity. Annual Review of Ecology and Systematics 31: 343 366. An important paper that is essential reading for those who want a rigorous and comprehensive under standing of coexistence mechanisms. Murdoch, W. W., C. J. Briggs, and R. M. Nisbet. 2003. Consumer Resource Dynamics. Princeton, NJ: Princeton University Press. The definitive work on consumer resource dynamics, with a comprehensive analysis of co existence mechanisms in parasitoids and predators. Ruel, J., and M. Ayres. 1999. Jensen’s inequality predicts effects of environmental variation. Trends in Ecology and Evolution 14: 361 366. A discussion of Jensen’s in equality that illustrates how the mathematical concept applies to ecological phenomena. Tilman, D. 1982. Resource Competition and Community Structure. Princeton, NJ: Princeton University Press. A comprehensive treatment of resource competition theory, based on the R* rule, that has provided the foundation for much of the later theory on competitive coexistence.

II.7 Predator–Prey Interactions Robert F. Denno and Danny Lewis OUTLINE

1. Evidence that predators reduce prey populations 2. Reciprocal density effects and predator–prey cycles 3. Mathematical models of predator–prey interactions 4. Factors stabilizing predator–prey interactions and promoting their persistence 5. Predation in complex food webs 6. Predation, biodiversity, and biological control 7. Evolutionary interactions between predators and prey 8. Epilogue In natural food webs, consumers fall victim to other consumers such as predators, parasitoids, parasites, or pathogens. Predators kill and consume all or parts of their prey and do so either before or after their catch has reproduced. A lynx stalking, attacking, and consuming a snowshoe hare is an example from the vertebrate world. Spiders snaring moths in their webs, assassin bugs lancing caterpillars with their beaks, and starfish ravaging mussel beds in rocky intertidal habitats are all instances of invertebrate predation. By contrast, parasitoids such as small wasps and flies usually attack only the immature stages of their arthropod hosts, thus killing them before they reproduce. Parasites live on (e.g., fleas and lice) or in (e.g., tape worms) host tissues, often reducing the fitness of their host but not killing it. Pathogens (e.g., viruses, bacteria, and fungi) induce disease and either weaken or ultimately kill their hosts. Although this chapter focuses on predators, there are many similarities among predator–prey, host– parasitoid, and host–pathogen interactions.

GLOSSARY food web. Network of feeding relationships among

organisms in a community functional response. The relationship between prey

density and the number of prey consumed by an individual predator

intraguild predation. A predation event in which one

member of the feeding guild preys on another member of the same guild (predators consuming predators) keystone species. A species that has a disproportionate effect on its environment relative to its abundance mesopredator. A predator that is fed on by another predator, usually a top carnivore numerical response. The relationship between the number of predators in an area and prey density omnivory. Feeding at more than one trophic level such as occurs when a predator consumes herbivores as well as other predators top carnivore. A predator at the top of the food chain feeding on organisms at lower trophic levels (e.g., mesopredators and herbivores) trophic cascade. Reciprocal predator–prey effects that alter the abundance, biomass, or productivity of a community across more than one trophic link in the food web (e.g., removing predators enhances herbivore density, which in turn diminishes plant biomass) Unlike many consumers, predators are often generalized in their feeding habits, consuming a diversity of prey species that can even represent different trophic groups. For instance, coyotes (top carnivores) feed on other predators such as foxes (mesopredators), both of which consume rabbits (herbivores) and opossums (omnivores). When predators consume other predator species, the act is called intraguild predation, whereas cannibalism occurs when predators consume members of their own species. Wolf spiders (Lycosa and Pardosa), for example, are notoriously cannibalistic, consuming smaller individuals in the population and even their own offspring. Although many predators are generalists, feeding on a diversity of prey species, there are some very specialized feeders. Desert horned lizards (Phrynosoma platyrhinos) are ant specialists, and ground beetles in the genus Scaphinotus feed selectively on mollusks and have a long head and mandibles adapted for reaching deep into snail shells.

Predator–Prey Interactions Predation can have widespread ecological, evolutionary, and economic effects on biological communities in both natural and managed habitats. Predation, for instance, can be a powerful evolutionary force with natural selection favoring more effective predators and less vulnerable prey. In an ecological sense, predators can dramatically affect the abundance and distribution of their prey populations. Moreover, the diverse feeding habits of predators form linkages that are responsible for the flow of energy through food webs, thus affecting food-web dynamics. Predators can act as keystone species, preventing superior competitors from dominating the community and promoting biodiversity at lower trophic levels. In contrast, the invasion of native ecosystems by exotic predators often has very negative effects on resident prey species. On a more positive note, invertebrate predators have been used as effective control agents of agricultural pests, increasing crop yields without the adverse consequences of pesticides. Thus, in both theoretical and applied contexts, it is imperative to understand the process of predation and its complex effects on species interactions, foodweb dynamics, and biodiversity. In the remainder of this article, we explore critical elements of predator– prey interactions, namely how predators and prey influence each other’s population size and dynamics, what factors stabilize predator–prey interactions and promote their persistence, how predation promotes complex species interactions and stabilizes food webs, and how predators and prey have reciprocally influenced each other’s evolution. 1. EVIDENCE THAT PREDATORS REDUCE PREY POPULATIONS

Excluding or adding predators to natural prey populations provides support that predators indeed can reduce populations of their hosts, very significantly in some cases. A classic example involves the mule deer herd on the Kaibab Plateau on the north rim of the Grand Canyon in Arizona. Before 1905, the deer herd numbered about 4000 individuals, but it erupted more than 10-fold over the course of the next 20 years when a bounty resulted in the demise of native deer predators such as wolves, coyotes, and cougars. At a much larger spatial scale in eastern North America, white-tailed deer populations have erupted following the extinction of top predators. These examples and others from the vertebrate world suggest that predators impose natural controls on prey populations and diminish the chances for so-called prey release. The biological control of crop pests following the release of natural enemies provides further evidence that predators suppress prey populations. With the


accidental introduction of Cottony cushion scale (Icerya purchasi) from Australia, the California citrus industry became seriously threatened by this severe insect pest. In the late 1800s, a predaceous ladybug beetle (Rodolia cardinalis) was collected in Australia and subsequently released into California citrus groves. Shortly after the release of this efficient predator, it completely controlled the scale insect and saved the citrus industry from financial ruin (Caltagirone and Doutt, 1989). Since this classic case, the encouragement or release of arthropod predators has frequently resulted in reduced pest populations (Symondson et al., 2002). Manipulative experiments also show that invertebrate predators impose controls on prey populations in natural habitats. For example, herbivorous planthoppers (Prokelisia marginata) and their wolf spider predators (Pardosa littoralis) co-occur on the intertidal marshes of North America (Do¨bel and Denno, 1994). When spiders are removed from habitat patches, planthopper populations erupt to very high levels. If spiders are removed but are then added back into habitat patches at natural densities, planthopper populations remain suppressed. The question arises as to how predators reduce prey populations. In planthopper– spider systems, spiders can reduce prey directly by consuming them (consumptive effect), or they can indirectly affect prey populations via nonconsumptive effects (Cronin et al., 2004). For example, the mere presence of spiders promotes the local dispersal of planthoppers. Similarly, when grasshoppers are exposed to nonlethal (i.e., defanged) spiders, they undergo a feeding shift from grasses to poor-quality forbs, where they avoid spiders but incur increased mortality from starvation (Schmitz et al., 1997). Notably, the mortality arising from this antipredator behavior rivals that seen when grasshoppers are killed directly by spiders with their fangs intact. In fact, evidence is building from many systems that predators adversely affect prey populations via both consumptive and nonconsumptive effects. It should be evident that predators often inflict high mortality on prey populations and that in the absence of predation prey populations often erupt. There are cases, however, in which predator removal does not result in increased prey density. Often, such cases involve compensatory mortality whereby the mortality inflicted by predation is replaced by mortality from another limiting factor such as food shortage. In the sections that follow, we consider how predators and prey interact to affect each other’s long-term population dynamics, specifically address the role of predation in population cycles, and explore factors that promote the persistence of predator–prey interactions in nature.


Population Ecology 150 Snowshoe hare Canadian lynx

Figure 1. Fluctuations in lynx and snowshoe hare populations based on the number of pelts purchased by the Hudson Bay Company between 1845 and 1930 (data from NERC, 1999).

Number (thousands)

125 100 75 50 25 0 1850



Population cycles occur in a diverse array of animals ranging from arctic mammals to forest insect pests. Traditionally, predation is one factor thought to induce such cycles (Gilg et al., 2003). Historic support for the view that a coupled predator–prey interaction can drive population cycles came from an analysis of about 100 years of fur-trapping records by the Hudson Bay Company in boreal Canada. An analysis of the number of lynx and snowshoe hare pelts showed spectacular cyclicity with peaks and valleys of abundance occurring at roughly 10-year intervals (figure 1). When hares were numerous, lynx increased in numbers, reducing the hare population, which in turn caused a decline in the lynx population. With predation relaxed, the hare population recovered, and the cycle began anew. It should be noted, however, that there is controversy over the singular role of predation in driving population cycles in boreal mammals.


The first ecologists to model the cyclic dynamics of predator–prey interactions were Alfred Lotka (1925) and Vito Volterra (1926), who independently derived the ‘‘predator–prey equations’’ (Lotka-Volterra equations), a pair of differential equations describing the coupled dynamics of a single specialized predator and one prey species. Both ecologists based their models on observations of reciprocal predator–prey cycles in nature. Volterra’s ideas were motivated by watching the rise of fish populations in response to decreased fishing pressure during World War I, whereas Lotka was in-





spired by observing parasitoid–moth cycles. The LotkaVolterra equations demonstrate the inherent propensity for predator–prey populations to oscillate, in what is called ‘‘neutral stability’’ (Begon et al., 1996). For the prey or host population, the rate of population change through time (dH/dt) is represented by the equation: dH ¼ rh H aHP, dt where H is prey density, rh is the rate of increase of the prey population (birthrate), a is a constant that measures the prey’s vulnerability and predator’s searching ability, and P is predator density. Thus, exponential growth of the prey population (rhH) is countered by deaths from predation (aHP). Change in the predator population through time (dP/dt) is shown by: dP ¼ caHP dp P, dt where c is a constant, namely the rate that prey are converted to predator offspring, and dp is the rate of decrease in the predator population (death rate). The death rate of the predator population ( dp P) is offset by the rate at which predators kill prey and convert them to offspring (caHP). The two equations provide a periodic solution in that predator and prey populations oscillate in reciprocal fashion through time (figure 2A). When the dynamics of predator and prey populations resulting from the Lotka-Volterra equations is plotted in two-phase space (predator density versus prey density), a neutral limit cycle results whereby both predator and prey populations cycle perpetually in time (figure 2B).

Predator–Prey Interactions B.

Prey Predator

Predator density

Predator and prey densities




Seeing that simple models could generate predator– prey oscillations prompted numerous researchers to duplicate such persistent cycles under simple laboratory conditions. However, these attempts often failed. A representative example involved the predaceous ciliate protozoan Didinium nasutum and its prey, another ciliate, Paramecium caudatum. Five Paramecium were placed in laboratory cultures, and after 2 days three predators were added. Initially, prey populations exploded in the absence of predators, but with the addition of predators, Paramecium populations were quickly driven to extinction. Moreover, in the absence of prey, predators subsequently perished (figure 3A). These unexpected results, and those from many other laboratory attempts, raised the question of why predator–prey cycles could not be easily reproduced in the laboratory, why such simple systems were inherently unstable, and why the predator–prey interaction did not persist. Simply stated, more is needed to understand why prey is not driven to extinction at high predator densities and why predators persist when focal prey are rare. As the following sections demonstrate, ecologists have since identified multiple factors missing from the Lotka-Volterra model that introduce realism into predator–prey interactions and lend accuracy in predicting real-world dynamics. As a result, more recent models incorporate biological features such as saturating predator functional responses (inability of predators to capture all available prey when prey are abundant), nonlinear reproductive responses, predator interference, refuges, spatial processes such as immigration, alternative prey, and multiple trophic levels (Canham et al., 2003; Grimm and Railsback, 2005).

Prey density

Figure 2. (A) Oscillating pred ator and prey populations and (B) a neutrally stable predator prey limit cycle generated by the Lotka Volterra equations.


Although the Lotka-Volterra equations generate coupled predator–prey oscillations, they are inherently oversimplified. It is unrealistic to expect predators and prey to cycle as predicted (figure 2). For instance, several assumptions of the early predator–prey models are not met by real organisms. The models assume exponential growth of prey in the absence of predation and exponential decline of the predator population in the absence of prey. Prey, however, are often resource limited, and their population growth can be slowed independent of predation. As shown in the discussion of functional responses below, predators are rate limited in their ability to capture and process prey, which constrains their ability to suppress prey populations at high densities. Also, predators often interfere with one another at high predator densities, further relaxing predation pressure on prey. There is also the unrealistic model expectation that predators and prey respond instantaneously to changes in each other’s densities. For microorganisms with high reproductive rates, this expectation is not far fetched. However, for larger predators, their reproductive response to increased prey density is lagged, providing the opportunity for prey to escape predator control, ultimately leading to an unstable dynamic. There are a multitude of other reasons why simple models inadequately predict predator–prey dynamics and do not capture the complexity of predator–prey interactions in nature. Foremost is that predator– prey interactions do not take place in closed systems in the absence of spatial processes such as emigration and

Population Ecology A.

60 P. caudatum


Number of individuals

Number of individuals


80 60 40

D. nasutum

20 0 2

3 4 Days



Predator Prey

20 10









10 0

0 20

30 Days






8 10 12 14 16 18 Days

Low level immigration of predators and prey into the system pro moted the cycling and persistence of predator and prey populations. (From Gause, 1934)

B. Population density of predator (T. occidentalis)

Population density of prey (E. sexmaculatus)






D. nasutum



1400 1200

Population density of prey (E. sexmaculatus)


Figure 3. Interaction between predator (Didinium nasutum) and prey (Paramecium caudatum) in laboratory microcosms (A) without im migration and (B) with immigration (the addition of a single individual of the predator and prey once every 3 days as indicated by arrows).


P. caudatum


0 0




Predator Prey












Population density of predator (T. occidentalis)


0 10


30 Days




Figure 4. In a simple structured habitat without refuges (A), pre dacious mites drive herbivorous mites to low densities, leading to the starvation and ultimate extinction of predatory mites. In a

complex structured habitat (B), herbaceous mites find refuge from predation, and the predator prey oscillation persists until food (oranges) quality for the prey deteriorates (Huffaker, 1958).

immigration. At low prey densities, predators often disperse to areas of higher prey density, thus relaxing predation on the local prey population rather than driving it to extinction. Moreover, immigration from neighboring patches can rescue declining local populations. Even in very simple lab settings, immigration can encourage the persistence and cycling of predators and prey. Returning to the Didinium–Paramecium system, the addition of a single individual predator and prey every third day of the experiment resulted in a persistent predator–prey cycle (figure 3B). In addition to spatial processes, complex habitat structure and the refuge it provides for prey from predation also lend persistence to predator–prey interactions. A classic example involves interactions between the citrus-feeding mite Eotetranychus sexmaculatus and

its predatory mite Typhlodromus occidentalis (Huffaker, 1958). The population dynamics of the mites was compared between two experimental habitats: a simple habitat consisting of a monoculture of oranges arranged on trays and a complex-structured habitat where oranges were interspersed among rubber balls and little posts from which prey could disperse. In the simple universe, predaceous mites easily dispersed throughout the habitat, prey were driven to a threateningly low density, and the predator then became extinct (figure 4A). In the complex habitat, prey dispersed and found refuges from predation, and three complete predator– prey oscillations resulted before the food quality of oranges deteriorated and the system collapsed (figure 4B). This study highlights the importance of refuges in promoting the coexistence of predators and prey, but it also

Predator–Prey Interactions

C = rN

C. Type II functional response

D. Type II fractional consumption

C = rN / (1 + rhN) Fraction of prey consumed

Predator’s per capita consumption rate

B. Type I fractional consumption

Fraction of prey consumed

Predator’s per capita consumption rate

A. Type I functional response

E. Type III functional response

F. Type III fractional consumption

C = rN2 / (1 + rhN2) Fraction of prey consumed

Predator’s per capita consumption rate

emphasizes that other factors such as food quality bear on the persistence of the interaction. Prey species also escape predation as a result of constraints on the ability of predators to catch and handle prey (functional response) and increase their population size (numerical response) as prey densities rise (Holling, 1959, 1965). In his component analysis of predation, Holling described three types of functional response (figure 5). For predators exhibiting a Type I functional response, the consumption rate of a single individual is limited only by prey density (figure 5A). Thus, over a wide range of densities, per capita consumption and prey density are linearly related. Many filter feeders (e.g., rotifers and sponges) that consume suspended zooplankton exhibit this response. For predators showing a Type I response, the proportion of prey captured of the total number offered remains constant and independent of prey density (figure 5B). Most invertebrate predators (e.g., hunting spiders, preying mantises, ladybug beetles) exhibit a Type II functional response, in which consumption rate levels off with increasing prey density to an upper plateau (saturating response) set by handling time (the time required to subdue and consume each prey item) and satiation (figure 5C). At high prey densities, most of a predator’s time is spent handling captured prey, and little time is spent searching for additional prey. Notably, for predators with a Type II response, the fraction of prey captured of the available total decreases with increasing prey density (figure 5D). As prey density increases, such predators are less able to reduce prey population growth, thus providing prey with an ever-growing opportunity to escape from predation. Many vertebrate predators (e.g., birds and mammals) and some invertebrate predators show a sigmoidal or Type III functional response (figure 5E). For such predators, consumption rate responds slowly to increases in prey density when prey is scarce. At somewhat higher prey densities, consumption rate rises rapidly, and at very high prey densities, consumption rate saturates and is limited by handling time as in a Type II response. The rapid rise in consumption rate at intermediate prey densities occurs because predators learn to discover and capture prey with increased efficiency or they simply increase their searching rate as they encounter more prey. In some cases, predators respond to volatile chemicals emitted by their prey and thereafter rapidly increase and direct their searching rate accordingly. Polyphagous predators often switch to alternative prey when the density of their preferred prey species falls below a certain threshold. Cases of prey switching can transform a Type II response into a Type III be-


Prey density

Prey density

Figure 5. Consumption rate of predators when offered prey at in creasing densities. Per capita consumption rates are shown for predators exhibiting Type I, II, and III functional responses (A, C, and E, respectively), as is the fractional consumption rate (proportion of prey taken of the total number offered) for each functional re sponse (B, D, and F). Associated equations describe the specific functional response: C ¼ per capita consumption rate, N ¼ prey density, r ¼ ‘‘risk of discovery’’ (a measure of prey vulnerability), and h ¼ handling time.

cause the consumption rate of focal prey is relaxed at low prey densities. Regardless of the mechanism (learning, increased search rate, or prey switching), predators exhibiting a Type III response are thought to stabilize predator–prey interactions. Stability is imposed because the fraction of prey consumed by a predator is low at low densities, preventing predators from driving prey to extinction (figure 5F). Yet, with an increase in prey density, the fraction of prey consumed increases (is density dependent), thus reducing the opportunity for prey to escape predator control. Only at very high densities are predators satiated such that the fraction of prey consumed decreases and the prey population escapes.


Population Ecology

So far, we have considered only the consumption rate of an individual predator under conditions of increasing prey density. To gain a complete picture of how predators might control prey populations, we also need to know how many predators are present in the population and how they respond to increasing prey densities. Most predators exhibit a numerical response by becoming more abundant as the density of their preferred prey increases. Two independent mechanisms underlie this pattern. First, predators often aggregate in areas where prey abound, a response that results from a short-term change in the spatial distribution of predators. For instance, the local density of the wolf spider Pardosa littoralis can be dramatically enhanced over a 3-day period when prey is experimentally added to its habitat (Do¨bel and Denno, 1994). Second, if prey density remains high for an extended period of time, predator populations often build as a consequence of increased reproduction. The density of Arctic foxes, for instance, increases greatly during peak lemming years because of elevated breeding success (Gilg et al., 2003). Thus, predator aggregation and enhanced reproduction can both account for the numerical response of a predator to increased prey density. Like functional responses, numerical responses level off at intermediate prey densities because continued increases in prey densities do not result in a higher predator density because of reproductive limitations or interference among conspecific predators. Nonetheless, strong aggregative responses are often thought to stabilize predator–prey interactions because predation is relaxed in vacated habitats where prey density is low, whereas predation is increased in colonized habitats where prey is more abundant. Combining a predator’s functional and numerical responses into its total response predicts a predator’s overall response to increased prey density and thus its overall impact on the prey population. A predator’s total response (number of prey consumed/unit area) can be calculated by multiplying its functional response (number of prey consumed/predator) by its numerical response (number of predators/unit area). Because both functional and numerical responses level off at intermediate prey densities, further increases in prey density result in an increasingly smaller proportion of the prey population that is killed by predators. Thus, the opportunity for escape exists at high prey densities, and one defensive strategy is for prey to satiate the predator population by emerging synchronously at very high densities. Periodical cicadas (Magicicada) appear to employ this strategy in that the proportion of mortality attributable to predation is drastically reduced during times of peak emergence when cicadas reach incredibly high densities.

Other life history traits of prey, such as dispersal capability, stage class, and body size may also provide escape from predation and thus contribute to the persistence of predator–prey interactions. Regarding dispersal, a highly mobile lifestyle appears to promote escape from predator control. Planthopper species, for example, vary tremendously in their dispersal ability with both highly mobile and extremely sedentary species represented. In a survey of species, invertebrate predators inflicted significantly more mortality on immobile species than on their migratory counterparts (Denno and Peterson, 2000). Invulnerable stage classes also provide a refuge from predation. The true bug Tytthus attacks only the eggs of planthoppers; thus, once planthopper eggs have hatched, emerging nymphs are immune to predation from this specialist predator. Also, because the act of predation requires overpowering victims, predators usually profit by attacking smaller or weaker individuals in the prey population. Size-selective predation has been observed across a wide range of vertebrate (snakes, fish, birds, and mammals) and invertebrate predators (insects, spiders, starfish). Importantly, because predators often focus their attacks on smaller prey, larger prey obtain a partial refuge from predation. 5. PREDATION IN COMPLEX FOOD WEBS

So far, our focus has been on interactions between a single predator and prey species and how inherent limitations imposed by a predator’s functional and numerical responses and size-selective predation can offer prey a partial escape from predation. However, predators and prey are nested into food webs and do not occur in isolation from other players. In fact, refuges for prey exist because of other species in the system. We have already seen how the presence of alternative prey species can relax predation on focal prey when its density drops to low levels. Moreover, mesopredators are also subject to predation themselves from top carnivores. Thus, interactions among predators can result in intraguild predation, which often relaxes predation on shared prey. A good example involves heteropterans bugs (Zelus and Nabis) and lacewing larvae (Chrysoperla), all of which prey on cotton aphids (Rosenheim et al., 1993). In the absence of heteropterans predators, lacewing predation drives the aphid population to a low level. When heteropterans are added to the system, they focus their attack on the more vulnerable lacewings, aphids experience a partial refuge from predation, and aphid populations rebound. Thus, the presence of multiple predator and prey species in the community can alter the interaction

Predator–Prey Interactions between a specific predator–prey pair and often lends stability to any specific interaction. By now it should be clear that many factors influence the dynamics of a specific predator–prey interaction. Even in the simplest of systems, it is difficult to draw solely on the focal pair of players to explain each other’s population fluctuations. For instance, the 4-year population cycles of lemmings in Greenland are driven by a 1-year delay in the numerical response of stoat and stabilized by density-dependent predation imposed by arctic foxes, snowy owls, and skuas (Gilg et al., 2003). Moreover, recent assessments of snowshoe hare population dynamics (figure 1), including a large-scale predator exclusion and food enhancement experiment, suggest that hare population cycles result from interactions among three trophic levels (Krebs et al., 1995, 2001). When lynx and other predators were experimentally excluded, hare populations doubled. Hare populations tripled when plant biomass was increased via fertilization. Strikingly, hare populations increased 11-fold when predators were excluded and plant resources were enhanced. This finding highlights the view that predators and prey can indeed affect each other’s abundance, but the dynamic of the interaction is complex and can not be studied in isolation from other factors. Visiting an invertebrate system further underscores why understanding predator–prey interactions requires a multitrophic approach (Finke and Denno, 2004, 2006). Spartina cordgrass is the only host plant for Prokelisia planthoppers, which in turn are consumed by the mesopredator Tytthus and the intraguild predator Pardosa. In this intertidal system, there is considerable variation in leaf litter as a result of elevational differences in tidal flushing and decomposition. In litter-rich habitats, Pardosa spiders abound and readily aggregate in areas of elevated planthopper density. In these structurally complex habitats, Tytthus finds refuge from Pardosa predation, the predator complex effectively suppresses planthoppers, and cordgrass flourishes. By contrast, in litter-poor habitats, spiders are less abundant, Tytthus experiences intraguild predation, and overall predation on planthoppers is relaxed, leading to planthopper outbreak and reduced plant biomass. Thus, both vegetation structure and the predator assemblage interact in complex ways to influence the strength of the spider– planthopper interaction and the probability for a trophic cascade, namely the extent to which predator effects cascade to affect herbivore suppression and plant biomass. This example and many others further emphasize that alternative population equilibria exist for prey and that release from predation is dependent on spatial refuges and the composition of other players in the system.


Although there is evidence that cycling does occur in some simple predator–prey systems in the boreal north, coupled cycling is not often characteristic of predator– prey dynamics in more complex food webs. Here, plant-mediated effects, alternative prey, intraguild predation, and refuges collectively dampen predator– prey cycles. Moreover, such multitrophic interactions are the rule and are thought to lend stability to food webs, making them more resistant to environmental disturbance and invasion by other species (fa*gan, 1997). 6. PREDATION, BIODIVERSITY, AND BIOLOGICAL CONTROL

Predators can act as keystone species influencing the species composition and biodiversity of the prey community. A classic case involves starfish, which graze mussels and barnacles in intertidal habitats, precluding them from dominating the community, allowing other invertebrate species to persist, and enhancing the overall diversity of the benthic community (Paine, 1974). From a conservation perspective, however, exotic predators that invade natural habitats can have very negative effects on resident prey species, effects that can cascade throughout the food web. When the brown treesnake was accidentally introduced to Guam, its population erupted in the absence of native predators, ultimately leading to the widespread extirpation of many native vertebrate species including birds, mammals, and lizards. Similarly, rainbow trout have been purposefully introduced throughout the world, often with devastating effects on native stream communities. In parts of New Zealand, trout incursion resulted in a trophic cascade, whereby this efficient predator reduced populations of native invertebrates that graze on benthic algae, which in turn promoted dramatic increases in algal biomass. Another alarming consequence of our rapidly changing world is the loss of biodiversity as a result of habitat disturbance, fragmentation, and loss. In particular, consumers at higher trophic levels such as predators are at risk of extinction. In coastal California, for example, urbanization and habitat fragmentation have promoted the disappearance of coyote, the historic top carnivore in this sage–scrub habitat (Crooks and Soule´, 1999). Its disappearance has fostered increased numbers of smaller mesopredators (e.g., foxes and skunks), which in turn are contributing to the extinction of scrub-breeding birds. A practical extension of the consequences of multiplepredator interactions is whether single or multiple predators are more effective in suppressing agricultural pests. The effect of increased predator diversity on

Population Ecology


Prey suppression

Synergism (facilitation)


Antagonism (interference, IGP)

Number of predator species Figure 6. Relationships between the number of predator species in the system and prey suppression. Synergism results from facilita tion, additivity arises from predator complementarity, and antago nism occurs from intraguild predation (IGP) or interference.

biological control, and ecosystem function at large, depends on how predator species interact and complement each other (figure 6). We have already seen how increasing predator diversity by adding an intraguild predator to the enemy complex can relax predation on shared herbivore prey. However, not all predator– predator interactions are antagonistic. In stream systems, predators interact synergistically, whereby stonefly predators drive mayflies from under stones making them more susceptible to fish attack (Soluk and Collins, 1988). Thus, predators that interact synergistically can enhance prey suppression beyond additive expectations. Likewise, if predators complement each other by attacking prey at different developmental stages or during different times of the year, increasing predator diversity can enhance prey suppression. The key to elucidating the relationship between predator diversity and ecosystem function or manipulating the composition of the predator assemblage for more effective biological control rests on the nature of predator–predator interactions in the system. The issue remains open in biological control because there is system-specific evidence that increasing predator diversity can either increase or decrease pest suppression. 7. EVOLUTIONARY INTERACTIONS BETWEEN PREDATORS AND PREY

There is little doubt that predators have exerted selection on prey that has resulted in evolutionary

change. For instance, prey species have evolved a wide range of defenses in response to selection from predation. Such defenses can be categorized as primary, secondary, or tertiary depending on when in the predation sequence (detection, capture, or handling) they operate. Primary defenses (e.g., crypsis and reduced activity when predators forage) operate before prey is detected by a predator. Secondary defenses operate to deter capture after prey is detected by the predator. Examples include escape mechanisms (aphids dropping from leaves in the presence of a foraging ladybug), startle behaviors (moths displaying wings with eye spots to frighten away birds), and evasive behaviors (moths detecting the sonar of bats and initiating strategic flight-avoidance tactics). Tertiary defenses interrupt predation after capture and during the handling phase. Such defenses include mechanisms that deter, repel, or even kill the predator directly (contact toxins, venoms, or morphological structures such as spines). The consequences of some tertiary defenses for predators can be severe. The neurotoxin injected by the death-stalker scorpion (Leiurus quinquestriatus) can cause rapid paralysis and death to an attacking small mammal. Clearly, predation has promoted a wide array of prey defenses, and the question arises as to whether there have been counteradaptations in predators. Have predators and prey engaged in an ‘‘evolutionary arms race’’ such that reciprocal selection has promoted a continuing escalation of predator offense and prey defense? Some evidence is consistent with this hypothesis. For instance, the drilling abilities of predaceous gastropods and the shell thickness of their bivalve prey have increased through geologic time (Vermeij, 1994). Similarly, marine snails have become more heavily armored, while the correlated response in predaceous crabs has been the evolution of larger claws for crushing the better-defended snails. In both of these instances, predators may have evolved greater weaponry in reciprocal response to improved prey defense (coevolution hypothesis), or predator armaments may have evolved in response to other predators or competitors (escalation scenario). Overall, however, evidence suggests that reciprocal selection on predators may be weaker than that on prey, thus precluding a classic evolutionary arms race (Brodie and Brodie, 1999). In part, the asymmetry arises because many predators are generalist feeders, and selection imposed by any one of its prey options is likely small. In general, coevolution between exploiter and exploited is unlikely when the intimacy of the interaction is low. Moreover, selection on predators from effective primary and secondary prey defenses is probably weak. For instance,

Predator–Prey Interactions if a predator fails to detect cryptic prey or catch a stealthy individual, it simply searches for another, without penalty. The exception occurs when predators interact with dangerous prey, prey that possess tertiary defenses such as toxins that can kill the attacker. In such cases, predators experience strong selection from prey and are expected to evolve either innate avoidance behavior or traits that allow them to exploit dangerous prey. Such a case of coevolution has likely occurred between the toxic newt Taricha granulosa and its garter snake predator Thamnophis sirtalis. The skin of the newt contains one of the most potent neurotoxins known, one that kills all other predators outright. Across populations, geographic variation in the level of newt toxin covaries with levels of resistance in the snake. Thus, garter snakes are evolving in response to newt defenses, and the ‘‘arms race’’ is apparently taking place (Brodie and Brodie, 1999).


Predation is a central process in community ecology because it is responsible for energy flow among multiple trophic levels. Moreover, the many reticulate linkages resulting from predation across multiple trophic levels (omnivory) are an important stabilizing force in food-web dynamics. Predators, however, by virtue of their precarious position at the apex of the food chain, are often at greater risk of extinction when natural systems are disturbed, often with dire consequences for the diversity and functioning of the community at large. Also, predators are important members of natural-enemy complexes that can provide effective pest control in agroecosystems. From the perspective of the consequences of predation, however, it is crucial to realize that the objectives of conservation biology and biological control seek very different ends. Conservation biologists seek to retain trophic diversity, preserve trophic linkages (e.g., intraguild predation) that stabilize food-web interactions, and reduce the probability that predator effects will cascade to affect plant productivity. Reconstructing stable food-web dynamics and ecosystem function are particularly important in habitat restoration projects. By contrast, biological control aims to induce trophic cascades, whereby antagonistic interactions among predators are minimized, pests are collectively suppressed by the enemy complex, and crop yield is enhanced. This contrast in objectives provides an ideal impetus for exploring the complex role of predation in community dynamics from both applied and theoretical perspectives.


FURTHER READING Begon, M., J. A. Harper, and C. R. Townsend. 1996. Ecol ogy, 3rd ed. London: Blackwell Science. Brodie, E. D. III, and E. D. Brodie, Jr. 1999. Predator prey arms races. BioScience 49: 557 568. Caltagirone, L. E., and R. L. Doutt. 1989. The history of the vedalia beetle importation to California and its impact on the development of biological control. Annual Review of Entomology 34: 1 16. Canham, Charles D., Jonathan J. Cole, and William K. Lauenroth. 2003. Models in Ecosystem Science. Prince ton, NJ: Princeton University Press. Cronin, J. T., K. J. Haynes, and F. Dillemuth. 2004. Spider effects on planthopper mortality, dispersal and spatial population dynamics. Ecology 85: 2134 2143. Crooks, K. R., and M. E. Soule´. 1999. Mesopredator release and avian extinctions in a fragmented habitat. Nature 400: 563 566. Denno, R. F., and M. A. Peterson. 2000. Caught between the devil and the deep blue sea, mobile planthoppers elude natural enemies and deteriorating host plants. American Entomologist 46: 95 109. Do¨bel, H. G., and R. F. Denno. 1994. Predator planthopper interactions. In Planthoppers: Their Ecology and Man agement. R. F. Denno and T. J. Perfect, eds. New York: Chapman & Hall, 325 399. fa*gan, W. F. 1997. Omnivory as a stabilizing feature of natural communities. American Naturalist 150: 554 567. Finke, D. L., and R. F. Denno. 2004. Predator diversity dampens trophic cascades. Nature 429: 407 410. Finke, D. L., and R. F. Denno. 2006. Spatial refuge from intraguild predation: Implications for prey suppression and trophic cascades. Oecologia 149: 265 275. Gause, G. F. 1934. The Struggle for Existence. Baltimore: Williams & Wilkins. Reprinted 1964, New York: Hafner. Gilg, O., I. Hanski, and B. Sittler. 2003. Cyclic dynamics in a simple vertebrate predator prey community. Science 302: 866 868. Grimm, V., and S. F. Railsback. 2005. Individual based Modeling and Ecology. Princeton, NJ: Princeton Univer sity Press. Holling, C. S. 1959. The components of predation as revealed by a study of small mammal predation of the European pine sawfly. Canadian Entomologist 91: 293 320. Holling, C. S. 1965. The functional response of predators to prey density and its role in mimicry and population reg ulation. Memoirs of the Entomological Society of Canada 45: 5 60. Huffaker, C. B. 1958. Experimental studies on predation: Dispersion factors and predator prey oscillations. Hil gardia 27: 343 383. Krebs, C. J., R. Boonstra, S. Boutin, and A.R.E. Sinclair. 2001. What drives the 10 yr cycle of snowshoe hares? BioScience 51: 25 36. Krebs, C. J., S. Boutin, R. Boonstra, A.R.E. Sinclair, J.N.M. Smith, M.R.T. Dale, K. Martin, and R. Tarkington. 1995. Impact of food and predation on the snowshoe hare cycle. Science 269: 1112 1115.


Population Ecology

NERC Centre for Population Biology, Imperial College. 1999. The Global Population Dynamics Database. http:// Paine, R. T. 1974. Intertidal community structure: Experi mental studies of the relationship between a dominant competitor and its principal predator. Oecologia 15: 93 120. Rosenheim, J. A., L. R. Wilhoit, and C. A. Armer. 1993. Influence of intraguild predation among generalist insect predators on the suppression of an herbivore population. Oecologia 96: 439 449. Schmitz, O. J., A. P. Beckerman, and K. M. O’Brien. 1997. Behaviorally mediated trophic cascades: Effects of pre

dation risk on food web interactions. Ecology 78: 1388 1399. Soluk, D. A., and N. C. Collins. 1988. Balancing risks? Re sponses and nonresponses of mayfly larvae to fish and stonefly predators. Oecologia 77: 370 374. Symondson, W.O.C., K. D. Sunderland, and M. H. Green stone. 2002. Can generalist predators be effective bio control agents? Annual Review of Entomology 47: 561 594. Vermeij, G. J. 1994. The evolutionary interaction among species: Selection, escalation, and coevolution. Annual Review of Ecology and Systematics 25: 219 236.

II.8 Host–Parasitoid Interactions Cheryl J. Briggs OUTLINE

1. Parasitoid terminology and taxonomy 2. Parasitoids and behavioral ecology 3. Parasitoids in theory: The quest for persistence and stability 4. Heterogeneity in risk of parasitism 5. Large-scale spatial dynamics 6. Stage structure in systems with overlapping generations 7. Parasitoids in biological control: Case study of the successful control of California red scale Parasitoid–host interactions have been popular topics of study in the areas of population biology and behavioral ecology because they represent potentially simple, tightly coupled interactions in which the oviposition behavior of the adult female parasitoid searching for hosts translates directly into fecundity and therefore fitness. The Nicholson-Bailey model, which predicts that the interaction between a single host and a single parasitoid species, in its simplest form, will result in the extinction of one or both species, spawned several decades of research into uncovering the mechanisms that allow real host–parasitoid interactions to persist. Work in this area has focused on the potential stabilizing effects of heterogeneity across the host population in the risk of parasitism, large-scale spatial processes, and stage-structured interactions.

GLOSSARY classical biological control. The purposeful release of

natural enemies of a pest (often from the pest’s area of origin) with the hope that the enemy will both suppress the density of a pest species and also persist to suppress future outbreaks of the pest. oviposition/ovipositor. The act of laying an egg on or in a host/the specialized structure that many adult female parasitoids have that allows them to lay an egg on or in a host.

parasitoid. Parasitoids are insects in which the adult

female lays one or more eggs on, in, or near the body of another insect (the host), and the resulting parasitoid offspring use the host for food as they develop, killing the host in the process. population regulation. In the history of ecology, this has been a surprisingly difficult term to define; the tendency of a population to persist within bounds. pseudointerference. A form of temporal density dependence in which the parasitoid efficiency decreases at high parasitoid densities because an increasing fraction of parasitoid attacks are wasted on alreadyparasitized hosts. stability. A population equilibrium is stable if the population returns to the equilibrium following a perturbation. 1. PARASITOID TERMINOLOGY AND TAXONOMY

Parasitoids are insects the adult female of which lays one or more eggs on, in, or near the body of another insect (the host), and the resulting parasitoid offspring use the host for food as they develop, killing the host in the process. Some authors have used the term parasitoid more generally to describe parasitic species that spend the majority of their life in close association with a single host individual, ultimately resulting in the death of that host; however, the term has been used mainly in reference to insects with this type of life history. Parasitoids are distinguished from parasites in that parasitoids kill their host in the process of completing their life cycle. They differ from predators because they require only a single host to complete their development. The majority of insect parasitoids are Hymenoptera (wasps) or Diptera (flies), but the parasitoid life history is also present in the Coleoptera (beetles) and occasionally in representatives of other orders of insects. Most parasitoids attack the juvenile stages of their host, and the parasitoid literature is filled with specialized terminology to describe their mode of attack.


Population Ecology

Parasitoids are often characterized by the stage of host that is attacked, e.g., in egg parasitoids, the adult female parasitoid lays her egg in the egg stage of the host, and in larval parasitoids, it is the larval host stage that is initially attacked. Some parasitoids (termed idiobionts) immediately kill or permanently paralyze their host at the time of attack, whereas others (koinobionts) permit their host to continue to feed, grow, and develop for some time before it is killed, allowing the developing parasitoid offspring to gain more resources from the host beyond those present at the time of oviposition. In solitary parasitoids, a single egg is laid on a host, whereas in gregarious parasitoids, a few to several hundred eggs can be laid on the same host. In some species, the female parasitoid gains all of the protein needed to produce all of her eggs from the host on which she developed (proovigenic parasitoids), whereas in others, the female continues to develop eggs during her adult life (synovigenic parasitoids), and the adult female can feed on additional hosts to gain the necessary nutrients. When host feeding, the adult female parasitoid can pierce the host with her ovipositor and then turn around and consume the host fluids, rather than laying an egg. Host feeding generally results in the death of the host. 2. PARASITOIDS AND BEHAVIORAL ECOLOGY

Parasitoid–host interactions have been highly influential in the development of ecological theory pertaining to population dynamics, population regulation, and species interactions, in part because, at least conceptually, the parasitoid–host interaction represents a simple, tightly coupled consumer–resource interaction in which each host attacked can lead to one or a clutch of new parasitoids. Because of this direct link between the oviposition behavior by searching female parasitoids and their fecundity, and therefore fitness, parasitoids are also the subjects of intensive study in behavioral ecology. Parasitoids display a staggering array of interesting behaviors. Parasitoids have been shown to assess the quality and size of the hosts that they encounter and modify their oviposition strategy accordingly. Larger hosts contain more resources for the developing parasitoids and therefore can produce larger, or more, parasitoid offspring. In many cases, parasitoids have been shown to lay larger clutches of eggs in larger or higher-quality hosts. Hymenopterous parasitoids are haplodiploid, such that females result from fertilized eggs and males result from unfertilized eggs. Therefore, the adult female can determine the sex of her offspring through fertilization. Female parasitoids (which produce costly eggs) generally benefit more than males from being large, and as such, many para-

sitoid species have been shown to lay female eggs in large hosts and male eggs in small hosts. In synovigenic parasitoid species, females tend to host-feed on smaller hosts and oviposit in larger, higher-quality, hosts. In many cases, parasitoids have been shown to be able to discriminate between unparasitized and parasitized hosts, and they can distinguish between hosts that have been parasitized by members of their own species (conspecifics) versus other parasitoid species (heterospecifics). Superparasitism describes the situation in which the adult female parasitoid lays additional eggs on a host individual that has already been parasitized by a conspecific, and in multiparasitism, the adult female parasitoid oviposits on a host that has been parasitized by a heterospecific competitor parasitoid. In either case, the adult parasitoid may kill any eggs present on the host (ovicide) by probing them with her ovipositor, or the resulting parasitoid larvae may compete within the host resulting in the death of one or both larvae. A further type of interaction occurs when a heterospecific parasitoid larva that is already present is itself used as a host. Some parasitoids are facultative hyperparasitoids in that they can attack either unparasitized hosts or hosts that have already been parasitized by individuals of another species, whereas other species are obligate hyperparasitoids in that they attack only parasitized hosts. Autoparasitism (heteronomous hyperparasitism) is one of the more unusual life history strategies exhibited by parasitoids in which female eggs are laid on an unparasitized host while male eggs are laid on juvenile parasitoids, either of the same species or another species. In addition to being of interest to behavioral and evolutionary ecologists, all of these oviposition behaviors have the potential to influence the dynamics of the host–parasitoid interaction. For example, superparasitism can lead to density-dependent juvenile parasitoid survival and can have a potentially stabilizing effect. Facultative hyperparasitism has the same dynamic effect as intraguild predation (in which competing predator species also eat each other) and can therefore affect the outcome of competition between parasitoid species. 3. PARASITOIDS IN THEORY: THE QUEST FOR PERSISTENCE AND STABILITY

The Nicholson-Bailey model lies at the heart of most of the theory of host–parasitoid population dynamics. This model considers the host–parasitoid interaction in its simplest form: the population of a single host species is attacked by a single parasitoid that searches for hosts at random within the closed population. The hosts and parasitoids have synchronized, nonoverlapping

Host–Parasitoid Interactions generations, making the model most relevant for temperate insect systems with a single generation per year. The host population is not regulated by any factor other than the parasitoid (i.e., no other sources of density dependence). The model predicts the density of hosts (N) and parasitoids (P) in year t þ 1 based on their densities in year t: Hosts : Parasitoids :

Nt þ 1 ¼ R Nt f (Nt , Pt ) Pt þ 1 ¼ cNt [1 f (Nt , Pt )]

where R is the net reproductive rate of hosts in the absence of the parasitoids (the host population would increase by a factor of R each year if no parasitoids were present). The function f(Nt, Pt) is the fraction of hosts that avoid being parasitized in year t. The Nicholson-Bailey model assumes that this function does not depend on the density of hosts but decreases exponentially as the density of parasitoids in year t increases: f (Nt , Pt ) ¼ f (Pt ) ¼ exp( a Pt ). This function is the zero term of a Poisson distribution and can be thought of as the result of parasitoids searching randomly for hosts with the average number of encouters (aPt) increasing linearly with parasitoid density. The constant, a, can be interpreted as the fraction of the total area searched by each parasitoid during year t. If f(Nt, Pt) is the fraction of hosts that escape parasitism, then the remaining fraction [1 f (Nt , Pt )] is the fraction of hosts in year t that are parasitized. Each of these parasitized hosts results in c new parasitoids in the following year. As long as R is greater than 1, this system always has a positive equilibrium with both host and parasitoid present. The equilibrium, however, is always unstable. For all parameter combinations, the host and parasitoid populations oscillate, with the magnitude of the oscillations rapidly increasing in amplitude through time until one or both populations become extinct. The parasitoid population lags behind the host, and the only two potential outcomes are extinction of the parasitoids followed by geometric growth of the host and extinction of the host followed by extinction of the parasitoid. In this simple model, persistence of the host–parasitoid interaction is not possible. The instability occurs because the parasitoids overexploit the host population, causing them to become rare, which in turn leads to a crash in parasitoid numbers, allowing the host population to recover; because of the time lags in the system, the successive cycles of overexploitation and recovery increase progressively in amplitude until one party becomes extinct. The inherent instability and lack of persistence of the Nicholson-Bailey model fly in the face of the fact


that parasitoid–host interactions do exist in nature and are therefore apparently persistent. Numerous examples of persistent and sometimes stable dynamics have been observed in real laboratory and field host– parasitoid interactions. In many cases, parasitoid species that have been released to control pest insect species have resulted in successful classical biological control, which relies on persistence of the parasitoid with its host. This disconnect between the predictions of the Nicholson-Bailey model and the observed persistence of real host–parasitoid systems spawned decades of research into uncovering the mechanisms that could potentially regulate and stabilize host–parasitoid interactions. The Nicholson-Bailey model makes very simple assumptions about how the host–parasitoid interaction works. Much of the theoretical research in this area has investigated the effects of violating these assumptions to determine their effects on the persistence and stability of the host–parasitoid interaction. In most cases, the theoretical developments have been inspired by empirical studies, either the results of laboratory experiments or observations on the life history features of specific species. Key assumptions of the Nicholson-Bailey model are:

No density dependence in the host reproduction or survival Random searching by parasitoids in a well-mixed host population Constant parasitoid search rate that does not depend on parasitoid or host density Efficiency at converting parasitized hosts to new searching parasitoids does not depend on host or parasitoid density No density dependence in parasitoid survival Synchronized, nonoverlapping generations Closed populations of host and parasitoid One host/one parasitoid: the host is attacked by only a single parasitoid species, and the parasitoid species specializes on a single host

Models have been developed to determine the effects on persistence and stability of all of these biological mechanisms. Three types of potentially stabilizing mechanisms are highlighted here: heterogeneity in risk of parasitism, large-scale spatial dynamics (e.g., metapopulation dynamics), and stage structure in systems with overlapping generations. 4. HETEROGENEITY IN RISK OF PARASITISM

The Nicholson-Bailey model assumes that all hosts in the population at any given time are at equal risk of


Population Ecology

being attacked by parasitoids and that the parasitoids search at random within the host population. Much attention has been directed to the different ways that these assumptions can be violated and to the effects of heterogeneity in risk of parasitism. Robert May showed that allowing nonrandom attack by the parasitoids can drastically alter the predictions of the Nicholson-Bailey model. He did this in a phenomenological way by replacing the function that describes the fraction of hosts that escape parasitism, f(Pt), in the Nicholson-Bailey model with the zero term of a negative-binomial distribution [f (Pt ) ¼ (1 þ aPt =k) k , rather than the zero term of a Poisson distribution as in the original model]. The negative-binomial model assumes that there is heterogeneous risk of parasitoid attack across the host population, according to the clumping parameter, k. For high values of k, the negative-binomial model approaches the case of hom*ogeneous parasitoid attack, as in the original model. As k is reduced, the distribution of risk across the host population becomes more skewed, with high risk of parasitism concentrated in a smaller fraction of the population. This could come about, for example, if certain hosts are more exposed to parasitoids (e.g., near the plant surface) or are on plants that attract more parasitoids. Adding heterogeneous attack to the Nicholson-Bailey model dramatically alters the dynamics. With small values of k, hosts and parasitoids are not only able to persist, they persist at a stable equilibrium. The condition for stability is k < 1. However, as k gets smaller, and the risk of parasitism gets concentrated on a smaller fraction of the host population, a larger fraction of the host population escapes parasitism, and the host density at equilibrium gets larger. Therefore, as with many factors that can stabilize host–parasitoid dynamics, increasing stability also leads to increasing host equilibrium density, potentially presenting biological control practitioners with a dilemma. The negative-binomial model is a phenomenological description of heterogeneity of risk of parasitism across the host population, which could represent a number of different biological mechanisms, including heterogeneity in host susceptibility caused by behavioral or physiological differences among individuals in the population. Some individuals in the population may be better able to fight off parasitoid attack because of their behavioral traits or physiological condition. Alternatively, as a result of differences in phenology, some individuals in the population might overlap temporally with the adult searching parasitoids for a longer period of the year than others. The far greatest amount of attention in this area, however, has been devoted to the potential for spatial heterogeneity in parasitoid search

behavior and parasitoid aggregation to stabilize host– parasitoid interactions. Michael Hassell and colleagues have explored a range of different variants of a discrete-time model in which each generation the host population is distributed across a number of patches (e.g., host plants), and at the start of each generation the parasitoids distribute themselves across these host patches according to some distribution. The parasitoid distribution can either be independent of host density or aggregated such that parasitoids are concentrated in areas of high host density. At the start of each generation, the hosts and parasitoids from all of the patches mix and redistribute themselves. Hassell and colleagues proposed a general rule, the ‘‘CV2 > 1 Rule, ’’ which states that the host– parasitoid equilibrium will be stable if the coefficient of variation squared of the density of searching parasitoids around each host is greater than 1. This rule is a reasonable approximation for the stability criterion for a range of different models of this type. Both host density-dependent and host density-independent heterogeneity in parasitoid attack can be stabilizing. The stabilizing effect of spatial heterogeneity, however, results not directly from the spatial distribution of parasitoid attack but from how this translates into temporal density dependence. In these models, the parasitoid attack is concentrated on the hosts at high risk (i.e., those in patches with high parasitoid density), whereas those at low risk (low parasitoid density) may escape parasitism. This leads to a form of temporal density dependence, termed pseudointerference, in which the parasitoid efficiency decreases at high parasitoid densities because an increasing fraction of parasitoid attacks are wasted on already-parasitized hosts. In this type of discrete-time model, the parasitoids are assumed to effectively choose a patch at the start of the generation and not redistribute as the density of hosts in the high-risk patches are depleted. A general finding of subsequent models appears to be that spatial heterogeneity that maintains this heterogeneity of risk to the hosts across the generation is stabilizing, but the stabilizing effect is lost if redistribution of parasitoids hom*ogenizes the risk. 5. LARGE-SCALE SPATIAL DYNAMICS

An alternative to persistence of host–parasitoid interactions through local processes is the possibility envisioned by A. J. Nicholson in the 1930s that the dynamics of hosts and parasitoids in any single location might be unstable and characterized by frequent extinctions, but the collection of host and parasitoid populations across a larger (metapopulation) scale

Host–Parasitoid Interactions might persist if the local populations are loosely connected by dispersal. If each of the local populations has an unstable equilibrium, then the key to long-term persistence of a host–parasitoid metapopulation (or metacommunity) is that the dynamics on patches in different parts of the environment remain asynchronous to some degree, such that not all populations of hosts are becoming extinct or reaching outbreak densities at the same time. There must be some intermediate level of dispersal of hosts and parasitoids between the patches: too much dispersal and the patches will become synchronized; too little and they will revert to isolated, nonpersistent populations. Over the last two decades, there has been an explosion of models describing spatial host–parasitoid interactions. The structure of these models varies greatly, including patch-occupancy models that characterize each patch only as being occupied or unoccupied by hosts and/or parasitoids, patch models in which the within-patch models are described explicitly, grid or lattice models in which each cell of the grid is either an individual or a population, continuous-time reaction-diffusion models that follow the density of the populations distributed across continuous space, interacting particle models in which discrete individuals bump into each other in continuous space, etc. In general, this vast array of models has shown that spatial host–parasitoid interactions can frequently persist for longer than their nonspatial counterparts, and in many cases, they can persist indefinitely. In some cases, long-term persistence results from only ‘‘statistical stabilization,’’ in which the variability observed in the sum of a number of unregulated population trajectories will be less than the variability of individual trajectories (e.g., diversifying your stock portfolio). In other cases, linking together populations that would be unstable in isolation can actually stabilize the dynamics at both the local population and metapopulation levels (for this to occur, there must be some mechanism that maintains asynchrony between the local populations). In models in which space is modeled explicitly, persistence in space is sometimes accompanied by various types of spatial pattern formation, where parts of the environment have high densities of hosts and/or parasitoids and others have low densities. The spatial patterning can either be static in space or can move through the environment (e.g., spiral waves). Theoretical studies of spatial host–parasitoid interactions far outnumber empirical studies. Work by John Maron and Susan Harrison on western tussock moths on lupines in California provides one potential example of spatial pattern formation caused by a parasitoid. Patches of lupine with high densities of tussock moth are surrounded by apparently habitable lupine


plants with low moth densities, and these high-density patches can remain in place for many years. Moths experience lower parasitism rates within the patches than in the areas immediately surrounding the patches. Maron and Harrison hypothesized that the parasitoids, which have higher dispersal ability than their hosts, maintain the spatial patterning by spilling out of the high-density patches, causing a ‘‘halo of death’’ around the patch. In another study, Jens Roland and Philip Taylor showed that the degree of spatial fragmentation, such as that caused by deforestation, can alter the ability of parasitoids to find and potentially control their host populations. For three of four of the parasitoid species attacking a single species of forest tent caterpillar, the parasitism rate decreased as the degree of fragmentation increased. But for the smallest of the parasitoid species, the pattern was reversed, and the highest parasitism rates were achieved in highly fragmented forests. 6. STAGE STRUCTURE IN SYSTEMS WITH OVERLAPPING GENERATIONS

The Nicholson-Bailey model and its descendants assume discrete host and parasitoid generations, as in many temperate insect systems in which there is a single generation of each species per year. In many tropical, subtropical, and Mediterranean (e.g., California) climates, insects can have many generations per year with all life stages present at any time, and because all life stages co-occur, there is the potential for interesting stage-dependent interactions that can affect the dynamics of the host–parasitoid interaction. Continuous-time host–parasitoid models with stage structure have been written to describe this type of situation. In this type of model, frequently written as delayed differential equations, the life cycle of each species is divided into a number of discrete stages (e.g., eggs, larvae, pupae, adults) that can have different demographic rates. In the case of the parasitoid, only the adult female stage searches for hosts, and in the case of the host, generally only a subset of the juvenile stages is attacked by the parasitoids (e.g., the larval stage). The dynamic effects of a range of stage-dependent parasitoid oviposition behaviors have also been investigated through this type of model. Stage-structured models can produce new types of interesting dynamics (e.g., generation cycles) caused by time lags and interactions between stages, and details of the stage-structured interaction can affect the persistence and stability of the host–parasitoid interaction. Because many parasitoids attack only juvenile stages of their host, one key finding that is likely to be relevant to a number of real systems is that a relatively long-lived


Population Ecology

adult stage that is invulnerable to parasitism can have a strong stabilizing effect on the host–parasitoid equilibrium. The long-lived adult stages act as a life-history refuge for the host when parasitoid populations are high and would otherwise overexploit the host. Additionally, the dynamics of a system in which the parasitoid has a relatively short development time tends to be more stable than one in which the parasitoid has a long development time (the yearly time lag between parasitism of the host and the production of new parasitoids inherent in the Nicholson-Bailey model is central to the instability of its dynamics). The reason for this is that the parasitoid then acts more as a direct rather than delayed density-dependent cause of mortality, and the reduction of the time lag tends to prevent the cycles of overexploitation, crash, and recovery. In many species, parasitoids produce female offspring from attacks on older (larger) juvenile hosts but continue to attack and kill younger (smaller) hosts, either to produce male offspring or for host feeding. The attacks on younger host stages lead to a form of delayed feedback in the parasitoid recruitment rate (because attacks on the young stages now will result in fewer female parasitoid-producing larger hosts later). Murdoch and co-workers found that stage-dependent parasitoid attacks can have a stabilizing effect but also can lead to a new type of instability (delayed-feedback cycles) if this effect is too strong.

interaction is not only persistent but remarkably stable, with host and parasitoid densities fluctuating within very narrow bounds through time. Murdoch and coworkers tested and rejected many of the potential stabilizing mechanisms through field observations and experiments, including parasitoid aggregation, a refuge from parasitism, and large-scale spatial dynamics. Stability in this case appears to arise from details of the stage structure of the interaction. Red scale has a longlived adult stage that is invulnerable to attack by the parasitoid. A. melinus has a short development time relative to that of the host, allowing it to respond rapidly to increases in host abundance. Additionally, in situations in which the parasitoid population reaches high densities (e.g., following high densities of the host), high levels of reattack (both parasitism and host feeding) on already parasitized hosts occur, leading to density dependence in the juvenile parasitoid survival. This system is not only an excellent example of topdown control of an insect by a natural enemy, it is also a classic example of competitive displacement. A. melinus was one of many parasitoid species introduced. Following its introduction, A. melinus rapidly displaced the earlier-released and less-effective parasitoids by suppressing the host abundance to lower levels than that achieved by the earlier parasitoids. FURTHER READING


Parasitoids have been commonly used as in classical biological control in which natural enemies are introduced (often from the pest’s area of origin) with the hope that the enemy will both suppress the density of a pest species and also persist to quell future outbreaks of the pest. Although most biological control efforts fail, either because the natural enemy does not become established or because it is not successful at suppressing the target pest population, there are many examples of phenomenal successes. These successes are dramatic illustrations of the ability of parasitoids to exert topdown control on their host populations and to persist with their host, sometimes in very stable interactions. One of the most detailed efforts to determine the factors that allow persistence of a host–parasitoid system is the work of William Murdoch and colleagues in attempting to understand the successful biological control of California red scale, Aonidiella aurantii, by a parasitoid. California red scale is a pest of citrus that is controlled at a small fraction of its potential density by the action of the parasitic wasp, Aphytis melinus. The

Briggs, Cheryl J., and Martha F. Hoopes. 2004. Stabilizing effects in spatial parasitoid host and predator prey models: A review. Theoretical Population Biology 65: 299 315. Summarizes different ways that parasitoid host interactions can be stabilized by spatial processes. Godfray, H. Charles J. 1994. Parasitoids: Behavioral and Evolutionary Ecology. Princeton, NJ: Princeton Uni versity Press. Summary of parasitoid natural history, re productive strategies, and the relevant evolutionary and ecological theory. Hassell, Michael P. 1978. The Dynamics of Arthropod Predator Prey Systems. Princeton, NJ: Princeton Univer sity Press. The classic summary of early host parasitoid models. Hassell, Michael P. 2000. The Spatial and Temporal Dy namics of Host Parasitoid Interactions. Oxford: Oxford University Press. An update on more recent work, in cluding host density dependent and host density independent parasitoid aggregation, and patch models. Concentrates on discrete time models. Hassell, Michael P., and Steven W. Pacala. 1990. Hetero geneity and the dynamics of host parasitoid interactions. Philosophical Transactions of the Royal Society: Biologi cal Sciences 330: 203 220. More on the CV2 Rule. Hawkins, Bradford A., and Howard V. Cornell, eds. 1999. Theoretical Approaches to Biological Control. Cam bridge, UK: Cambridge University Press. Edited volume

Host–Parasitoid Interactions with more information on parasitoids in biological con trol situations. Mills, Nicholas J., and Wayne M. Getz. 1996. Modeling the biological control of insect pests: A review of host parasitoid models. Ecological Modeling 92: 121 143. Murdoch, William W., Cheryl J. Briggs, and Roger M. Nis bet. 2003. Consumer Resource Dynamics. Princeton, NJ: Princeton University Press. Discusses more continuous time models than Hassell’s books and has a heavy em phasis on stage structured interactions. Murdoch, William W., Cheryl J. Briggs, and Susan Swar brick. 2006. Biological control: Lessons from a study of California red scale. Population Ecology 48: 297 305. Recent summary of the efforts to understand the factors


leading to stability in the California red scale biological control example. Nicholson, A. J. 1933. The balance of animal populations. Journal of Animal Ecology 2: 131 178. Foresees most of the major developments in population ecology of the twentieth century. Tayor, Andrew D. 1993. Heterogeneity in host parasitoid interactions: ‘‘aggregation of risk’’ and the ‘‘CV 2 > 1 Rule:’’ Trends in Ecology and Evolution 8: 400 405. A good description of how heterogeneity of risk is stabiliz ing in discrete time models through leading to pseudo interference and temporal density dependence in the par asitoid recruitment rate.

II.9 Ecological Epidemiology Michael Begon OUTLINE

1. Parasites, pathogens, and other definitions 2. The importance of ecological epidemiology 3. The dynamics of parasites within populations: Transmission 4. The population dynamics of infection 5. Parasites and the dynamics of hosts 6. Shared parasites—zoonoses Strictly speaking, epidemiology is the study of the dynamics of disease in a population of humans. In ecology, however, the term takes on a slightly different meaning. Ecologists tend to expand the usage to cover populations of any species, animal or plant, but they then restrict it to infectious diseases (as opposed to, say, cancers or heart disease). Studies of human epidemiology usually treat the host (human) population as fixed in size and focus on the dynamics of disease within this population. What distinguishes ‘‘ecological’’ epidemiology is an acknowledgment that the dynamics of the parasite and the host populations may interact. Hence, we are interested in the dynamics of parasites in host populations, that may themselves vary substantially in size, and also in the effects of the parasites on the dynamics of the hosts.

GLOSSARY basic reproductive number. Usually denoted R0, for

microparasites, the average number of new infections that would arise from a single infectious host introduced into a population of susceptible hosts; for macroparasites, the average number of established, reproductively mature offspring produced by a mature parasite throughout its life in a population of uninfected hosts critical population size. The population size of susceptible hosts for which R0 ¼ 1, where R0 is the basic reproductive number, and which must therefore be exceeded if an infection is to spread in a population

density-dependent transmission. Parasite transmission

in which the rate of contact between susceptible hosts and the source of new infections increases with host density frequency-dependent transmission. Parasite transmission in which the rate of contact between susceptible hosts and the source of new infections is independent of host density herd immunity. Where a population contains too few susceptible hosts (either because of natural infection or immunization) for infection to be able to establish and spread within a population macroparasite. A parasite that grows but does not multiply in its host, producing infective stages that are released to infect new hosts; the macroparasites of animals mostly live on the body or in the body cavities (e.g., the gut); in plants, they are generally intercellular microparasite. A small, often intracellular parasite that multiplies directly within its host transmission threshold. The condition R0 ¼ 1, where R0 is the basic reproductive number, which must be crossed if an infection is to spread in a population vector. An organism carrying parasites from one host individual to another, within which there may or may not be parasite multiplication zoonosis. An infection that occurs naturally and can be sustained in a wildlife species but can also infect and cause disease in humans 1. PARASITES, PATHOGENS, AND OTHER DEFINITIONS

A parasite is an organism that obtains its nutrients from one or a very few host individuals, normally causing harm but not causing death immediately. This distinguishes parasites from predators, which kill and consume many prey in their lifetime, and from grazers, which take small parts from many different prey. If a parasite infection gives rise to symptoms that are clearly harmful, the host is said to have a

Ecological Epidemiology disease. Pathogen, then, is a term that may be applied to any parasite that gives rise to a disease (i.e., is pathogenic). The language used by plant pathologists and animal parasitologists is often very different, but for the ecologist, these differences are less striking than the resemblances. One distinction that is useful is that between microparasites and macroparasites. Microparasites are small, often intracellular, and they multiply directly within their host where they are often extremely numerous. Hence, it is usually impossible to count the number of microparasites in a host: ecologists normally study the number of infected hosts in a population. Examples include bacteria and viruses (e.g., the typhoid bacterium and the yellow net viruses of beet and tomato), protozoa infecting animals (e.g., the Plasmodium species that cause malaria), and some of the simpler fungi that infect plants. Macroparasites grow but do not multiply in their host. They produce infective stages that are released to infect new hosts. The macroparasites of animals mostly live on the body or in the body cavities (e.g., the gut) of their hosts. In plants, they are generally intercellular. It is often possible to count or at least to estimate the numbers of macroparasites in or on a host. Hence, ecologists study the numbers of parasites as well as the numbers of infected hosts. Examples include parasitic helminths such as the intestinal nematodes and tapeworms of humans, the fleas and ticks that are parasitic in their own right but also transmit many microparasites between their hosts, and plant macroparasites such as the higher fungi that give rise to the mildews, rusts, and smuts. Cutting across the distinction between micro- and macroparasites, parasites can also be subdivided into those that are transmitted directly from host to host and those that require a vector or intermediate host for transmission, i.e., are either simply carried from host to host by another species (aphids carrying viruses from plant to plant) or need to parasitize a succession of two (or more) host species to complete their life cycle (both mosquitoes and humans being parasitized by the malaria Plasmodium). 2. THE IMPORTANCE OF ECOLOGICAL EPIDEMIOLOGY

Parasites are an important group of organisms in the most direct sense. Millions of people are killed each year by various types of infection, and many millions more are debilitated or deformed. When the effects of parasites on domesticated animals and crops are added to this, the cost in terms of human misery and economic loss becomes incalculable. Parasites are also important numerically. A free-living organism that does not harbor


several parasitic individuals of a number of species is a rarity. Thus, ecological epidemiology is important from an entirely practical point of view. If we wish to control the diseases that have afflicted us and our domesticated species historically—malaria, tuberculosis—and those that have emerged recently or threaten us—HIV-AIDS, SARS, avian influenza—then we must seek to understand their dynamics. But it is also the case that a major question in ecology that not only remains unanswered but has only recently been seriously addressed is: To what extent are animal and plant populations and communities in general affected by parasitism and disease? Ecologists have long been concerned with the effects of food resources, competitors, and predators on their focal species; only recently have parasites and pathogens been afforded similar attention. 3. THE DYNAMICS OF PARASITES WITHIN POPULATIONS: TRANSMISSION

Transmission dynamics, in a very real sense, is the driving force behind the overall population dynamics of pathogens. Different species of parasite are of course transmitted in different ways between hosts, the most fundamental distinction being between parasites that are transmitted directly (either through close contact between hosts or via an environmental reservoir to which infectious hosts have contributed) and those that require a vector or intermediate host for transmission. Irrespective of these distinctions, the rate of production of new infections in a population depends on the per capita transmission rate (the rate of transmission per susceptible host ‘‘target’’) and also on the number of susceptible hosts there are. That per capita transmission rate depends on the infectiousness of the parasite, the susceptibility of the host, and so on, but it also depends on the contact rate between susceptible hosts and whatever it is that carries the infection. For directly transmitted parasites, we deal with the contact rate between infected hosts and susceptible (uninfected) hosts; for hosts infected by long-lived infective agents, it is the contact rate between these and susceptible hosts; with vector-transmitted parasites it is the contact rate between host and vector. But what determines this contact rate? Essentially, two factors are determinative: the contact rate between a susceptible individual and all other hosts, and the proportion of these that are actually infectious. For the first of these, ecologists have tended to make one of two simplifying assumptions: either that this contact rate increases in direct proportion to the density of the population (density-dependent transmission) or that it is utterly independent of population density


Population Ecology

(frequency-dependent transmission). The former imagines individuals bumping into one another at random: the more crowded they become, the more contacts they make. The latter, by contrast, assumes that the number of contacts an individual makes is a fixed aspect of its behavior. Frequency-dependent transmission has therefore conventionally been assumed for sexually transmitted diseases—the frequency of sexual contacts is independent of population density—but it is increasingly recognized that many social contacts, territory defense for instance, may come into the same category. It has also become increasingly apparent that real contact patterns usually conform to neither of these simplifying assumptions exactly, but they nonetheless represent two valuable benchmarks through which real data sets can be understood. There has also often been an assumption that the ‘‘infectious proportion’’ can be calculated from, and also applies throughout, the whole host population. In reality, however, transmission typically occurs locally, between adjacent individuals. Thus, there are likely to be hot spots of infection in a population, where the infected proportion is high, and corresponding cool zones. Transmission, therefore, often gives rise to spatial waves of infection passing through a population rather than simply an overall, global rise. 4. THE POPULATION DYNAMICS OF INFECTION

We begin by looking at the dynamics of disease within host populations without considering any possible effects on the total abundance of hosts. We then take the more ‘‘ecological’’ approach of considering the effects of parasites on host abundance in a manner much more akin to conventional predator–prey dynamics (see chapter II.7). The Basic Reproductive Number and the Transmission Threshold

In the study of the dynamics of parasites, there are a number of particularly key concepts. The first is the basic reproductive number, usually denoted R0. For microparasites, this is the average number of new infections that would arise from a single infectious host introduced into a population of susceptible hosts. For macroparasites, it is the average number of established, reproductively mature offspring produced by a mature parasite throughout its life in a population of uninfected hosts. The transmission threshold, which must be crossed if an infection is to spread, is then given by the condition R0 ¼ 1. An infection will eventually die out for R0 < 1 (each present infection or parasite leads to

fewer than one infection or parasite in the future), but an infection will spread for R0 > 1. Insights into the dynamics of infection can be gained by considering the various determinants of the basic reproductive number. We do this in some detail for directly transmitted microparasites with density-dependent transmission (see above) and then deal more briefly with related issues for other parasites. Directly Transmitted Microparasites and the Critical Population Size

For such microparasites, R0 can be said to increase (1) with the average period of time over which an infected host remains infectious, L; (2) with the number of susceptible individuals in the host population, S, because greater numbers offer more opportunities for transmission; and (3) with the transmission coefficient, b, the strength or force of transmission. Thus, overall: R0 ¼ SbL:


Note immediately that by this definition, the greater the number of susceptible hosts, the higher the basic reproductive number of the infection. But in particular, the transmission threshold can now be expressed in terms of a critical population size, ST, where, because R0 ¼ 1 at that threshold: ST ¼ 1=bL:


In populations with numbers of susceptibles less than this, the infection will die out (R0 < 1). With numbers greater than this, the infection will spread (R0 > 1). These simple considerations allow us to make sense of some very basic patterns in the dynamics of infection. Consider first the kinds of population in which we might expect to find different sorts of infection. If microparasites are highly infectious (large bs), or give rise to long periods of infectiousness (large Ls), then they will have relatively high R0 values even in small populations and will therefore be able to persist there (ST is small). Conversely, if parasites are of low infectivity or have short periods of infectiousness, they will have relatively small R0 values and will be able to persist only in large populations. Many protozoan infections of vertebrates, and also some viruses such as herpes, are persistent within individual hosts (large L), often because the immune response to them is either ineffective or short lived. A number of plant diseases, too, such as club-root, have very long periods of infectiousness. In each case, the critical population size is therefore small, explaining why the diseases can and do survive endemically even in small host populations.

Ecological Epidemiology On the other hand, the immune responses to many other human viral and bacterial infections are powerful enough to ensure that they are only very transient in individual hosts (small L), and they often induce lasting immunity. Thus, for example, a disease such as measles has a critical population size of around 300,000 individuals and is unlikely to have been of great importance until quite recently in human biology. However, it has generated major epidemics in the growing cities of the industrialized world in the eighteenth and nineteenth centuries, and in the growing concentrations of population in the developing world in the twentieth century. The Epidemic Curve

The value of R0 itself is also related to the nature of the epidemic curve of an infection. This is the time series of new cases following the introduction of the parasite into a population of hosts. Assuming there are sufficient susceptible hosts present for the parasite to invade (i.e., the critical population size, ST, is exceeded), the initial growth of the epidemic will be rapid as the parasite sweeps through the population of susceptibles. But as these susceptibles either die or recover to immunity, their number, S, will decline, and so too, therefore, will R0. Hence, the rate of appearance of new cases will slow down and then decline. And if S falls below ST and stays there, the infection will disappear—the epidemic will have ended. Not surprisingly, the higher the initial value of R0, the more rapid will be the rise in the epidemic curve. But this will also lead to the more rapid removal of susceptibles from the population and hence to an earlier end to the epidemic: higher values of R0 tend to give rise to shorter, sharper epidemic curves. Also, whether the infection disappears altogether (i.e., the epidemic simply ends) depends very largely on the rate at which new susceptibles either move into or are born into the population because this determines how long the population remains below ST. If this rate is too low, then the epidemic will indeed simply end. But a sufficiently rapid input of new susceptibles should prolong the epidemic or even allow the infection to establish endemically in the population after the initial epidemic has passed. Cycles of Infection

This leads us naturally to consider the longer-term patterns in the dynamics of different types of endemic infection. As described above, the immunity induced by many bacterial and viral infections reduces S, which reduces R0, which therefore tends to lead to a decline in the incidence of the infection itself. However, in due


course, and before the infection disappears altogether from the population, there is likely to be an influx of new susceptibles into the population, a subsequent increase in S and R0, and so on. There is thus a marked tendency with such infections to generate a sequence from many susceptibles (R0 high), to high incidence, to few susceptibles (R0 low), to low incidence, to many susceptibles, etc., just as in any other predator–prey cycle. This undoubtedly underlies the observed cyclic incidence of many human diseases, with the differing lengths of cycle reflecting the differing characteristics of the diseases: measles with peaks every 1 or 2 years, whooping cough every 3 to 4 years, and so on. By contrast, infections that do not induce an effective immune response tend to be longer lasting within individual hosts, but they also tend not to give rise to the same sort of fluctuations in S and R0. Thus, for example, protozoan infections tend to be much less variable (less cyclic) in their prevalence. Immunization Programs

Recognizing the importance of critical population sizes also throws light on immunization programs in which susceptible hosts are rendered nonsusceptible without ever becoming diseased (showing clinical symptoms), usually through exposure to a killed or attenuated pathogen. The direct effects here are obvious: the immunized individual is protected. But by reducing the number of susceptibles, such programs also have the indirect effect of reducing R0. Indeed, seen in these terms, the fundamental aim of an immunization program is clear: to hold the number of susceptibles below ST so that R0 remains less than 1. To do so is said to provide ‘‘herd immunity.’’ In fact, a simple manipulation of equation 2 gives rise to a formula for the critical proportion of the population, pc, that needs to be immunized in order to provide herd immunity (reducing R0 to a maximum of 1, at most). This reiterates the point that in order to eradicate a disease, it is not necessary to immunize the whole population—just a proportion sufficient to bring R0 below 1. Moreover, this proportion will be higher the greater the ‘‘natural’’ basic reproductive number of the disease (without immunization). It is striking, then, that smallpox, the only known disease where in practice immunization seems to have led to eradication, has unusually low values of R0 (and hence pc). Frequency-Dependent Transmission

Suppose, however, that transmission is frequency dependent. Then there is no longer the same dependence for spread on the number of susceptibles, and hence, no


Population Ecology

threshold population size. Such infections can therefore persist even in extremely small populations, where, to a first approximation, the chances of sexual contact, say, for an infected host are the same as in large populations. Vector-Borne Infections

For microparasites that are spread from one host to another by a vector, the life-cycle characteristics of both host and vector enter into the calculation of R0. In particular, the transmission threshold (R0 ¼ 1) is dependent on a ratio of vector:host numbers. For a disease to establish itself and spread, that ratio must exceed a critical level; hence, disease control measures are usually aimed directly at reducing the numbers of vectors and are aimed only indirectly at the parasite. Many virus diseases of crops, and vector-transmitted diseases of humans and their livestock (malaria, onchocerciasis, etc.), are therefore controlled by insecticides rather than chemicals directly targeting the parasite. Directly Transmitted Macroparasites

The effective reproductive number of a directly transmitted macroparasite (no intermediate host) is directly related to the length of its reproductive period within the host (i.e., again, to L) and to its rate of reproduction (rate of production of infective stages). Most directly transmitted helminths have an enormous reproductive capability. For instance, the female of the human hookworm Necator produces roughly 15,000 eggs per worm per day. The critical threshold densities for these parasites are therefore very low, and they occur and persist endemically in lowdensity human populations, such as hunter–gatherer communities. Indirectly Transmitted Macroparasites

Finally, for macroparasites with intermediate hosts, the threshold for the spread of infection depends