Encyclopedia of Animal Behavior, Volumes I-IV [2nd ed.] 0128132515, 9780128132517

Citation preview

About the pagination of this eBook This eBook contains a multi-volume set. To navigate this eBook by page number, you will need to use the volume number and the page number, separated by punctuation or a space. Refer to the Cumulative Index and match the page reference style exactly in the Go box at the bottom of the screen.

ENCYCLOPEDIA OF ANIMAL BEHAVIOR SECOND EDITION

This page intentionally left blank

ENCYCLOPEDIA OF ANIMAL BEHAVIOR SECOND EDITION EDITOR IN CHIEF

Jae Chun Choe Division of EcoScience, Ewha Womans University, Seodaemun-gu Ewhayeodae-gil 52, Seoul, Korea

VOLUME 1 Overview Essays • Historical Overviews • Animal Welfare and Conservation • Cognition • Communication

Amsterdam • Boston • Heidelberg • London • New York • Oxford Paris • San Diego • San Francisco • Singapore • Sydney • Tokyo Academic Press is an imprint of Elsevier

Academic Press is an imprint of Elsevier 125, London Wall, EC2Y, 5AS 525 B Street, Suite 1650, San Diego, CA 92101, United States 50 Hampshire Street, 5th Floor, Cambridge MA 02139, United States The Boulevard, Langford Lane, Kidlington, Oxford OX5 1GB, United Kingdom Copyright Ó 2019 Elsevier Ltd. All rights reserved No part of this publication may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopying, recording, or any information storage and retrieval system, without permission in writing from the publisher. Details on how to seek permission, further information about the Publisher’s permissions policies and our arrangements with organizations such as the Copyright Clearance Center and the Copyright Licensing Agency, can be found at our website: www.elsevier.com/permissions. This book and the individual contributions contained in it are protected under copyright by the Publisher (other than as may be noted herein). Notices Knowledge and best practice in this field are constantly changing. As new research and experience broaden our understanding, changes in research methods, professional practices, or medical treatment may become necessary. Practitioners and researchers may always rely on their own experience and knowledge in evaluating and using any information, methods, compounds, or experiments described herein. In using such information or methods they should be mindful of their own safety and the safety of others, including parties for whom they have a professional responsibility. To the fullest extent of the law, neither the Publisher nor the authors, contributors, or editors, assume any liability for any injury and/or damage to persons or property as a matter of products liability, negligence or otherwise, or from any use or operation of any methods, products, instructions, or ideas contained in the material herein. Library of Congress Cataloging-in-Publication Data A catalog record for this book is available from the Library of Congress British Library Cataloguing-in-Publication Data A catalogue record for this book is available from the British Library ISBN 978-0-12-813251-7

For information on all publications visit our website at http://store.elsevier.com

Publisher: Oliver Walter Acquisition Editor: Sam Crowe Content Project Manager: Michael Nicholls Associate Content Project Manager: Greetal Carolyn Designer: Miles Hitchen

Cover image: Bullet Ant carrying a liquid droplet. Courtesy of Dan Perlman

In Memoriam Charles “Mich” Michener (22 September 1918  1 November 2015) Robert Hinde (26 October 1923  23 December 2016) Peter Marler (24 February 1928  5 July 2014) Richard “Dick” Alexander (18 November 1929  20 August 2018) Koko the Gorilla (4 July 1971  19 June 2018)

This page intentionally left blank

EDITORIAL BOARD

EDITOR IN CHIEF Jae Chun Choe Division of EcoScience, Ewha Womans University, Seodaemun-gu Ewhayeodae-gil 52, Seoul, Korea

SECTION EDITORS OVERVIEW ESSAYS/HISTORICAL OVERVIEWS/SOCIAL BEHAVIOR Jae Chun Choe Division of EcoScience, Ewha Womans University, Seodaemun-gu Ewhayeodae-gil 52, Seoul, Korea

ANIMAL WELFARE AND CONSERVATION Donald Broom Centre for Animal Welfare and Anthrozoology, Department of Veterinary Medicine, University of Cambridge, Cambridge, United Kingdom

COGNITION Tetsuro Matsuzawa Institute for Advanced Study, Kyoto University, Kyoto, Japan

COMMUNICATION Michael Greenfield Department of Ecology and Evolutionary Biology, University of Kansas, Lawrence, Kansas, United States Yikweon Jang Division of EcoScience and Department of Life Sciences, Ewha University, 52, Ewhayeodae-gil, Seodaemun-gu, Seoul, Korea

EVOLUTION/METHODOLOGY Gil Rosenthal Department of Biology, Texas A&M University, College Station, Texas, United States

FORAGING Graham Pyke School of Life Sciences, University of Technology Sydney, Sydney, New South Wales, Australia

vii

viii

Editorial Board

GENES AND BEHAVIOR Ben Oldroyd Behaviour & Genetics of Social Insects Lab, Macleay Building, University of Sydney, Sydney, New South Wales, Australia

HORMONES AND BEHAVIOR Randy J Nelson Department of Neuroscience, Hazel Ruby McQuain Chair for Neurological Research, West Virginia University School of Medicine, Morgantown, West Virginia, United States

HOST-PARASITE INTERACTIONS Kenneth Wilson Lancaster Environment Centre, Lancaster University, Lancaster, United Kingdom

LANDMARK STUDIES Michael Breed Director, Baker Residential Academic Program, University of Colorado, Boulder, Colorado, United States

LEARNING AND TEACHING Martin Giurfa Research Center on Animal Cognition, Center of Integrative Biology, CNRS - University Paul Sabatier - Toulouse III, Toulouse, France

MIGRATION AND ORIENTATION Franz Bairlein Institut fuer Vogelforschung (Institute of Avian Research) “Vogelwarte Helgoland”, An der Vogelwarte 21, Wilhelmshaven, Germany

NEURONS AND SENSES Sabrina Burmeister Department of Biology, University of North Carolina, Chapel Hill, North Carolina, United States

PREDATOR-PREY INTERACTIONS Theodore Stankowich Department of Biological Sciences, California State University, Long Beach, California, United States

REPRODUCTIVE BEHAVIOR Patricia Adair Gowaty Ecology and Evolutionary Biology, UCLA, Institute of Environment and Sustainability, UCLA, Los Angeles, California, United States

CONTRIBUTORS TO VOLUME 1 Elizabeth Adkins-Regan Cornell University, Ithaca, NY, United States

Jerram L Brown University at Albany, Albany, NY, United States

Katarina Almeida-Warren University of Oxford, Oxford, United Kingdom

Joel S Brown University of Illinois at Chicago, Chicago, IL, United States

James R Anderson Kyoto University Graduate School of Letters, Kyoto, Japan F Aureli Veracruz University, Xalapa, Mexico; and Liverpool John Moores University, Liverpool, United Kingdom A Avarguès-Weber CNRS, Université de Toulouse, Toulouse, France; and Centre de Recherches sur la Cognition Animale, Toulouse, France

Henrik Brumm Max Planck Institute for Ornithology, Seewiesen, Germany Gordon M Burghardt University of Tennessee, Knoxville, TN, United States D Shallin Busch National Marine Fisheries Service, Seattle, WA, United States

Melissa Bain University of California School of Veterinary Medicine, Davis, CA, United States

RW Byrne University of St. Andrews, Fife, Scotland, United Kingdom

Christa A Baker Princeton University, Princeton, NJ, United States

J Call Max Planck Institute for Evolutionary Anthropology, Leipzig, Germany

M Bateson Newcastle University, Newcastle upon Tyne, UK Madeleine Beekman The University of Sydney, Sydney, NSW, Australia Sarah N Bevins National Wildlife Research Center, Fort Collins, CO, United States Susan Blackmore University of Plymouth, Plymouth, United Kingdom

Ulrika Candolin University of Helsinki, Helsinki, Finland JF Cantlon Rochester University, Rochester, NC, United States Bruce A Carlson Washington University in St. Louis, St. Louis, MO, United States

EM Brannon Duke University, Durham, NC, United States

Nora V Carlson Max Plank Institute of Ornithology, Radolfzell, Germany

Mark Briffa Plymouth University, Plymouth, United Kingdom

C Sue Carter Indiana University, Bloomington, IN, United States

Donald Maurice Broom University of Cambridge, Cambridge, United Kingdom

Susana Carvalho University of Oxford, Oxford, United Kingdom

ix

x

Contributors to Volume 1

Sergio Castellano University of Turin, Turin, Italy Cinzia Chiandetti University of Trieste, Trieste, Italy

M Giurfa CNRS, Université de Toulouse, Toulouse, France; and Centre de Recherches sur la Cognition Animale, Toulouse, France

Jae C Choe Division of EcoScience, Ewha Womans University, Seoul, Korea

Patricia A Gowaty University of California, Los Angeles, CA, United States; and Smithsonian Tropical Research Institute, Panama, United States

Jakob Christensen-Dalsgaard University of Southern Denmark, Odense M, Denmark

Kirsty E Graham University of York, York, England

Zanna Clay Durham University, Durham, United Kingdom

Temple Grandin Colorado State University, Fort Collins, CO, United States

NS Clayton University of Cambridge, Cambridge, UK Tim Clutton-Brock University of Cambridge, Cambridge, United Kingdom

Linda Greening University Centre Hartpury, Gloucester, Gloucestershire, United Kingdom

Reginald B Cocroft University of Missouri, Columbia, MO, United States

Alison L Greggor San Diego Zoo’s Institute for Conservation Research, Escondido, CA, United States

Murray James Corke University of Cambridge, Cambridge, United Kingdom

Andrea S Griffin University of Newcastle, Callaghan, NSW, Australia

JD Crystal University of Georgia, Athens, GA, USA

Wouter Halfwerk VU University, Amsterdam, The Netherlands

Camille Desjonquères University of Wisconsin-Milwaukee, Milwaukee, WI, United States

RR Hampton Emory University, Atlanta, GA, United States

FBM de Waal Emory University, Atlanta, GA, USA

Mike Hansell University of Glasgow, Glasgow, Scotland, United Kingdom

Donald A Dewsbury University of Florida, Gainesville, FL, United States

Brian Hare Duke University, Durham, NC, United States

A Dickinson University of Cambridge, Cambridge, UK

Benjamin L Hart University of California, Davis, CA, United States

Lee C Drickamer Northern Arizona University, Flagstaff, AZ, United States

Lynette A Hart University of California, Davis, CA, United States

Mark A Elgar University of Melbourne, VIC, Australia ON Fraser University of Vienna, Vienna, Austria Alejandro Frid Central Coast Indigenous Resource Alliance, Vargo Rd, Campbell River, BC, Canada Constantino M García Universidad Nacional Autónoma de México, Coyoacán, México

Lisa S Hayward Northwest Climate Science Center, Seattle, WA, United States Eileen A Hebets University of Nebraska-Lincoln, Lincoln, NE, United States Michael R Heithaus Florida International University, Miami, FL, United States Satoshi Hirata Kyoto University, Sakyo, Kyoto, Japan

Contributors to Volume 1

Catherine Hobaiter University of St Andrews, St Andrews, Scotland Andrew G Horn Dalhousie University, Halifax, Nova Scotia, Canada L Huber Messerli Research Institute, University of Veterinary Medicine Vienna, Medical University of Vienna, University of Vienna, Vienna, Austria Vincent M Janik University of St. Andrews, Fife, Scotland, United Kingdom Keith Jensen Queen Mary University of London, London, United Kingdom Linda J Keeling Swedish University of Agricultural Sciences, Uppsala, Sweden Jennifer L Kelley University of Western Australia, Crawley, WA, Australia Danielle A Klomp University of Porto, Vairão, Portugal Sarah D Kocher Princeton University, Princeton, NJ, United States Paul Koene Wageningen University and Research, Wageningen, The Netherlands Sonja A Kotz Maastricht University, Maastricht, The Netherlands Sarah M Lane Plymouth University, Plymouth, United Kingdom

Randy J Nelson The Ohio State University Wexner Medical Center, Columbus, OH, United States Stephen Nowicki Duke University, Durham, NC, United States Todd H Oakley University of California, Santa Barbara, CA, United States Sara O’Brien Radford University, Radford, VA, United States David Outomuro University of Cincinnati, Cincinnati, OH, United States LA Parr Yerkes National Primate Research Center, Atlanta, GA, USA Katarzyna Pisanski University of Sussex, East Sussex, United Kingdom Graham H Pyke University of Technology, Sydney Ultimo, NSW, Australia; and Macquarie University, NSW, Australia Andrea Ravignani Vrije Universiteit Brussel, Brussels, Belgium David Reby University of Sussex, East Sussex, United Kingdom Josue Reyes-Amaya Autonomous University of the State of Morelos, Cuernavaca, Morelos, Mexico

Marty L Leonard Dalhousie University, Halifax, NS, Canada

Rafael L Rodríguez University of Wisconsin-Milwaukee, Milwaukee, WI, United States

T Matsuzawa Kyoto University, Kyoto, Japan

Holly Root-Gutteridge University of Sussex, East Sussex, United Kingdom

Rowan H McGinley University of Nebraska-Lincoln, Lincoln, NE, United States

Kathryn Lee Gruchalla Russart The Ohio State University Wexner Medical Center, Columbus, OH, United States

R Menzel Freie Universität Berlin, Berlin, Germany

Crickette Sanz Washington University, St. Louis, MO, United States; and Congo Program, Wildlife Conservation Society, Republic of Congo

Nathan I Morehouse University of Cincinnati, Cincinnati, OH, United States Stephanie Musgrave Washington University, St. Louis, MO, United States

Robert M Sapolsky Stanford University, Stanford, CA, United States

xi

xii

Contributors to Volume 1

Laela S Sayigh Hampshire College, Amherst, MA, United States; and Woods Hole Oceanographic Institution, Woods Hole, MA, United States

Kirill Tokarev Hunter College, The City University of New York, New York, NY, United States; and Weill Cornell Medicine, New York, NY, United States

William A Searcy University of Miami, Coral Gables, FL, United States

Nahoko Tokuyama The Graduate University for Advanced Studies, Miuragun, Kanagawa, Japan

AM Seed Max Planck Institute for Evolutionary Anthropology, Leipzig, Germany

Michael Tomasello Max Planck Institute for Evolutionary Anthropology, Leipzig, Germany

Manuel Soler University of Granada, Granada, Spain

Alejandra Valero National Autonomous University of Mexico, Mexico City, Mexico

Kathrin F Stanger-Hall University of Georgia, Athens, GA, United States Ronald R Swaisgood San Diego Zoo’s Institute for Conservation Research, Escondido, CA, United States Michael Taborsky University of Bern, Hinterkappelen, Switzerland Ofer Tchernichovski Hunter College, The City University of New York, New York, NY, United States; and The Graduate Center of The City University of New York, New York, NY, United States Christopher N Templeton Pacific University, Forest Grove, OR, United States

Estefania Velilla VU University, Amsterdam, The Netherlands Stephen H Vessey Bowling Green State University, Bowling Green, OH, United States Elisabet V Wehncke Autonomous University of the State of Morelos, Cuernavaca, Morelos, Mexico Shinya Yamamoto Kyoto University Institute for Advanced Study, Sakyo, Kyoto, Japan Ken Yasukawa Beloit College, Beloit, WI, United States

PREFACE The Ecological Theater, the Evolutionary Play, and the Behavioral Act

“Cats and monkeys, monkeys and catsdall human life is there.” d Henry James

“I had a wonderful teacher about animal behavior - my dog Rusty. He taught me that animals have personalities, minds, and feelings.” d Jane Goodall

Animal behavior is a fatal attraction. A rancher in Texas who herded cattle all day long on his farm comes back home, grabs a beer, sits on the couch, and watches on the television an animal documentary presented by Sir David Attenborough. Many of us have an insatiable appetite for animal images and stories. Ancient drawings on the cave walls speak for our deep-rooted fascination. In fact, I would nominate the study of animal behavior the world’s second-oldest profession. Back in the days of our hunting and gathering past were ‘animal behaviorists’ with watchful eyes on game animals as well as fierce predators. I would argue with confidence that the tribes with such animal behaviorists survived better and reproduced more than those without. The title of the book The Ecological Theater and the Evolutionary Play by the eminent biologist G. Evelyn Hutchinson in 1965 captures the inseparable nature of the two academic disciplinesdEcology and Evolution. To make this scheme more complete, I would add one more component. When the evolutionary play is unfolded in the ecological theater, the actor is behavior. Animals and plants behave to cope with ecological pressures. Many behaviors directly increase an organism’s fitness, that is, they help it survive and reproduce. Thus, behavior is subject to natural selection. Animal behavior is an academic discipline of transdisciplinary nature. Comprehending the concepts, methods, and applied aspects of animal behavior requires understanding of other fields in biology such as evolutionary biology, physiology, endocrinology, neurobiology, psychology, and so on. In accordance with rapid development in such academic fields as neuroscience, genomic research, data science, and robotics, just to name a few, animal behavior is also experiencing a renaissance with exciting new lines of research and applications. Information on animal behavior is scattered everywhere, which makes the retrieval of useful information a formidable task. Authoritative and up-to-date accounts on recent discoveries and concepts will be extremely useful. The readers who would benefit from comprehensive provisioning of necessary information should include, first of all, upper-class undergraduates, graduate students, and researchers in biology, psychology, veterinary science, and medicine (especially, psychiatry and neurology). Professionals, academic and nonacademic, in the fields related to animal behavior, such as conservation biologists, zookeepers, park rangers, animal trainers, veterinarians, researchers and practitioners in pest management, nature photographers, secretarial staffs, and so on, would also enjoy having access to the encyclopedic accounts. The first edition of the Encyclopedia of Animal Behavior was published in 2010 and contained a total of 323 chapters in three volumes. I congratulate Michael Breed and Janice Moore the Editors-in-Chief as well as 21 section editors for their excellent job. Consequently, the work sold all over the world into universities and industry companies. As well as an outstanding performance from the physical print run across the world, the

xiii

xiv

Preface

first edition saw online usage spanning 61 different countries on ScienceDirect. North America showed significantly successful with an additional consistent performance over the years since the first publication in 2010, peaking in 2011, with individual downloads holding at a high level year-on-year. The title also had a strong performance with digital downloads across Europe, with the next top-usage countries being Canada, Australia, Mexico and Taiwan. Over half a million users at premier institutions worldwide have access to the Encyclopedia through the Reference Module in Life Sciences, with subscriptions growing. Recently the field of animal behavior has exploded to keep pace with rapid changes in complementary disciplines. The need to understand the concepts and discoveries in related fields is mutual. Students and researchers in other fields also require a good reliable source for convenient and comprehensive coverage of information in animal behavior. As more people demand easy access to the field of animal behavior, there is a growing need for an encyclopedia that provides readers of diverse educational backgrounds with a complete collection of both basic and applied information of the field on and off the line. We have made some changes from the first edition by adding an abstract and in-text citation with up-to-date references listed up in the References section at the end of each chapter. We retained nearly all the chapters from the first edition with a majority of them updated by the original authors. When the original authors were not available, however, the section editors commissioned new authors. The chapter entries were realigned to certain extent as some of the section headings are combined or modified and some new headings were added to accommodate recent developments in the field. This edition has expanded to be: l l l l l

Comprehensive: Information available in one place provides one-stop service Authoritative: Written by top researchers in the field Up-to-date: Includes the most recent discoveries and concepts Transdisciplinary: Make connections to related fields of research Multimedia-driven: Information disseminated through various media.

I am confident that this new revised and expanded edition will outperform its predecessor. My confidence comes from the collective experience and expertise of the team of section editors I assembled. I dare to call it ‘Dream Team’. Michael Breed, Donald Broom, Graham Pyke, Patty Gowaty, Tetsuro Matsuzawa, Franz Bairlein, Ken Wilson, Benjamin Oldroyd, Randy Nelson, Michael Greenfield, Martin Giurfa, Sabrina Burmeister, Ted Stankowich, Gil Rosenthal, Yikweon Jangdthese are literally Who’s Who in Animal Behavior. I readily admit that I the editor-in-chief is the least qualified of all in the team. Once I formed the team, I had little to navigate. They sailed on their own to their respective beautiful islands. From the bottom of my heart I respect and thank each and every one of them. A work of such enormous scope as this requires equally enormous dedication and cooperation. I am grateful to Priscilla Braglia, the Acquisitions Editor, and Blerina Osmanaj, the former Content Project Manager, who guided me through the formative phase of this project. Mike Nicholls, the Senior Content Project Manager, along with Sam Crowe, the Acquisitions Editor, Rebecca Gelson, the Content Project Manager, and Diana Meadowcroft and Greetal Carolyn, the Associate Content Project Managers, made it possible for this project to come to fruition. Among these I must single out Mike Nicholls, who never failed me in answering all my questions, serious or silly, and did not hesitate to provide me with his expert opinions whenever I needed. Thank you, Mike. Last, but surely not least, I thank all my colleagues who took time off from their hectic schedules to write exemplary articles for those who wish to learn about behaviors of other animals. You are the champions and we should all congratulate ourselves for putting out this tome together. Jae Chun Choe Seoul, Korea July 2018

CONTENTS OF VOLUME 1 Editorial Board

vii

Contributors to Volume 1

ix

Preface

xiii

Overview Essays Animal Architecture Mike Hansell

1

Brood Parasitism Manuel Soler

17

Cooperative Breeding Tim Clutton-Brock

31

Game Theory Ken Yasukawa

45

Hormones and Behavior: Basic Concepts Kathryn Lee Gruchalla Russart and Randy J Nelson

51

Integration of Proximate and Ultimate Causes Stephen H Vessey and Lee C Drickamer

61

Gene and Meme Susan Blackmore

67

Foraging: A Fundamental Activity for All Life Graham H Pyke

75

Welfare Concepts Donald Maurice Broom

80

Historical Overviews Animal Behavior: Antiquity to the Sixteenth Century Lee C Drickamer

84

Animal Behavior: The Seventeenth to the Twentieth Centuries Lee C Drickamer

89

Comparative Animal Behavior – 1920–1973 Gordon M Burghardt

94

xv

xvi

Contents of Volume 1

From Ethology to Behavioral Biology Michael Taborsky

99

Behavioral Ecology and Sociobiology Jerram L Brown and Jae C Choe

103

Psychology of Animals Donald A Dewsbury and Jae C Choe

109

Neurobiology, Endocrinology and Behavior Elizabeth Adkins-Regan and C Sue Carter

116

Future of Animal Behavior: Predicting Trends Lee C Drickamer and Patricia A Gowaty

124

Animal Welfare and Conservation Sentience DM Broom

131

Indicators of Good Welfare Linda J Keeling

134

Stereotypies and Other Abnormal Behavior in Welfare Assessment Linda Greening

141

Indicators of Pain Murray James Corke

147

Slaughter Plants: Behavior and Welfare Assessment Temple Grandin

153

Stress, Health and Social Behavior Robert M Sapolsky

163

Sickness Behavior in Animals: Implications for Health and Wellness Benjamin L Hart and Lynette A Hart

171

Welfare and Interactions Between Humans and Companion Animals Paul Koene

176

Ecology of Fear Joel S Brown

196

Overview of Animal Training: A Welfare Perspective Melissa Bain

203

Punishment Keith Jensen and Michael Tomasello

214

Applications of Animal Behavior to Conservation Ronald R Swaisgood and Alison L Greggor

220

Human Impact, Behavior and Conservation Alejandro Frid and Michael R Heithaus

230

Conservation Behavior and Endocrinology Sara O’Brien, Lisa S Hayward, and D Shallin Busch

242

Noise Pollution and Conservation Henrik Brumm and Andrew G Horn

254

Contents of Volume 1

xvii

Learning and Conservation Andrea S Griffin

260

Habitat Deterioration, Signals and Conservation Ulrika Candolin

266

Parasitism, Host Behavior, and Invasive Species Sarah N Bevins

273

Mating Interference Due to Introduction of Exotic Species Alejandra Valero

279

Seed Dispersal and Conservation Elisabet V Wehncke and Josue Reyes-Amaya

283

Captive Breeding, Conservation and Welfare Jennifer L Kelley and Constantino M García

291

Cognition Animal Arithmetic JF Cantlon and EM Brannon

302

Animal Tool Use Stephanie Musgrave and Crickette Sanz

310

Categories and Concepts: Language-Related Competences in Non-Linguistic Species L Huber

318

Chimpanzee and Bonobo Shinya Yamamoto, Nahoko Tokuyama, Zanna Clay, and Brian Hare

324

Cognitive Development in Chimpanzees T Matsuzawa

335

Collaborative Behavior Satoshi Hirata

343

Conflict Resolution F Aureli and ON Fraser

349

Deception: Competition by Misleading Behavior RW Byrne

354

Emotion and Social Cognition in Primates LA Parr

359

Empathetic Behavior FBM de Waal

366

Gestural Communication in the Great Apes Kirsty E Graham and Catherine Hobaiter

371

Mental Time Travel: Can Animals Recall the Past and Plan for the Future? NS Clayton and A Dickinson

378

Metacognition and Metamemory in Non-Human Animals RR Hampton

383

Non-Elemental Learning in Invertebrates M Giurfa, A Avarguès-Weber, and R Menzel

390

xviii

Contents of Volume 1

Primate Archaeology Susana Carvalho and Katarina Almeida-Warren

397

Problem-Solving in Tool-Using and Non-Tool-Using Animals AM Seed and J Call

408

Rational Choice Behavior: Definitions and Evidence M Bateson

416

Responses to Death James R Anderson

424

Time: What Animals Know JD Crystal

429

Communication: Signal Modality Acoustical Signals – In Air and Water Jakob Christensen-Dalsgaard

438

Bioluminescent Signals Kathrin F Stanger-Hall and Todd H Oakley

449

Chemical Signaling: Air, Water, and on the Substrate Mark A Elgar

462

Electric Signals Christa A Baker and Bruce A Carlson

474

Multimodal Signaling Eileen A Hebets and Rowan H McGinley

487

Visual Signals Using Incident Light Nathan I Morehouse and David Outomuro

500

Vibrational Signals: Sounds Transmitted Through Solids Rafael L Rodríguez and Camille Desjonquères

508

Communication: Context Mating Signals, Including Advertisement and Courtship Sergio Castellano

518

Parent–Offspring Signaling Andrew G Horn and Marty L Leonard

525

Signals in Conflict Resolution: Conventional Signals, Aggression and Territoriality Mark Briffa and Sarah M Lane

531

Communication: Communication in Animal Groups Individual Signatures in Animal Groups: Cetaceans Laela S Sayigh and Vincent M Janik

539

Telling Your Friends Where the Goodies are – Recruitment Signals for Food and Habitat Madeleine Beekman

550

Signals in Insect Social Organization Sarah D Kocher and Reginald B Cocroft

558

Contents of Volume 1

xix

Communication: Analysis of Animal Communication Communication Networks Christopher N Templeton and Nora V Carlson

568

The Use of Phylogenetic Comparative Methods in the Study of Evolution and Visual Signalling Danielle A Klomp

581

Signal Reliability and Intraspecific Deception William A Searcy and Stephen Nowicki

589

Communication: Special Topics Adjustments to Facilitate Communication in Noisy Environments Estefania Velilla and Wouter Halfwerk

598

Culture and Learning: Bird Song Kirill Tokarev and Ofer Tchernichovski

606

Rhythm and Music in Animal Signals Andrea Ravignani, Cinzia Chiandetti, and Sonja A Kotz

615

Vocal Communication Between Humans and Animals Holly Root-Gutteridge, Katarzyna Pisanski, and David Reby

623

This page intentionally left blank

PERMISSION ACKNOWLEDGMENTS The following material is reproduced with kind permission of Oxford University Press Figure 1 Threespine Stickleback Table 1 Chimpanzee and Bonobo www.oup.com The following material is reproduced with kind permission of American Association for the Advancement of Science Figure 3 Singing Behavior in Fishes: Hormones, Neurons, and Evolution Figure 3 Ant, Bee and Wasp Social Evolution Figure 4 Ant, Bee and Wasp Social Evolution Figure 5 Ant, Bee and Wasp Social Evolution Figure 5 Caste in Social Insects: Genetic Influences Over Caste Determination Figure 6 Magnetic Orientation in Migratory Songbirds Figure 3 Sexual Selection and Speciation Figure 1 The Oxytocin System: Single Gene Effects on Social Behavior Across Species Figure 4 Sea Turtles: Navigation and Orientation Figure 1 Vocal–Acoustic Communication in Fishes: Neuroethology Figure 2 Paramecium Behavioral Genetics Figure 3 Paramecium Behavioral Genetics Figure 4 Paramecium Behavioral Genetics Figure 5 Paramecium Behavioral Genetics Figure 4 Group Movement www.aaas.org The following material is reproduced with kind permission of Nature Publishing Group Figure Figure Figure Figure Figure Figure Figure Figure Figure Figure Figure Figure Figure Figure

10 Spotted Hyenas 15 Spotted Hyenas 5 Blue Tit 3 Orientation, Navigation and Homing in Bats 4 Orientation, Navigation and Homing in Bats 5 Orientation, Navigation and Homing in Bats 1 Sexual Selection and Speciation 4 Sexual Selection and Speciation 2 The Oxytocin System: Single Gene Effects on Social Behavior Across Species 3a Collective Intelligence in Social Animals 3 The use of phylogenetic comparative methods in the study of evolution and visual signalling 6 Neuroethology of Sound Localization 4 Flexible Mate Choice 5 Sea Turtles: Navigation and Orientation

i

ii

Permission Acknowledgments

Figure 3 Helpers and Reproductive Behavior in Birds and Mammals Figure 2 Aposematism as a Defence Against Predation Figure 1 Development, Evolution and Behavior Figure 1 Disease Transmission and Network Figure 2 Drosophila Behavior Genetics Figure 6 Chemical Signalling: Air, Water, and on the Substrate Figure 1 Disease Transmission and Networks http://www.nature.com

OVERVIEW ESSAYS Animal Architecture Mike Hansell, University of Glasgow, Glasgow, Scotland, United Kingdom © 2019 Elsevier Ltd. All rights reserved.

Abstract The complexity of animal building behavior varies from genetically constrained to flexible and creative. However, even invertebrates can create complex architecture using apparently stereotyped behavior routines. The anatomy employed for building behavior is jaws and/or paired limbs but these generally show little specialisation. Building materials can be either collected or self-secreted. Unlike collected materials, self-secreted ones such as mucus and silk are highly specialised, having co-evolved with the building behavior. The functions of built structures are, in order of importance, homes, traps and signals. Materials used in home building are varied. However, traps, which take the form of webs or nets, require advanced engineering and therefore require self-secreted materials; only invertebrates and some chordate Appendicularia make these structures. Devices used by vertebrates to obtain food take the form of tools; these are generally simple structures made from collected materials. Builders add complexity to the environment. In doing so they can enhance biological diversity and influence their own evolutionary pathway by inheriting environments that have already been altered by previous generations.

Keywords Aesthetics; Building anatomy; Burrows; Cognition; Extended phenotype; Homes; Inheritance; Manipulation; Mucus; Nests; Signals; Silk; Speciation; Tools; Webs

Introduction One of the tales told of the Renaissance artist Giotto di Bondone (1270–1337) is that when Pope Boniface VIII sent a messenger to him and other artists requiring them to provide a drawing demonstrating their artistic skill, Giotto took a brush and in red paint simply drew a perfect freehand circle. In competition with the other artists’ efforts, Giotto won the commission. Near to where I live there is a water-filled ditch where each winter I can reliably find leaves in which have been cut perfectly circular holes (Fig. 1). What should be my response? Should I publicise the creature that made these perfect leaf holes as a hitherto unrecognised animal genius, or wag my finger at Pope Boniface VIII saying he was lucky to get away with such a poor test of artistic ability? The perfectly circular holes, by the way, are created by the caddis larva Glyphotaelius pellucidus that makes the roof and floor of its portable case out of the cut leaf panels.

Fig. 1 Holes left in leaves after the removal of circular panels that form the roof and floor of the portable case of the caddis larva Glyphotaelius pellucidus. The larva also cuts smaller panels that form the sides of the case. (Photo: Mike Hansell).

Encyclopedia of Animal Behavior, 2nd edition, Volume 1

https://doi.org/10.1016/B978-0-12-809633-8.90728-3

1

2

Overview Essays j Animal Architecture

There is a serious point here in relation to animal architecture and to animal behavior more generally. Relatively lowly animals can produce architecturally complex outcomes with their small brains using apparently simple repertoires of behavior. Fig. 2(a) shows the caterpillar of the moth Cyana sp. (Lithosiini, Arctiinae, Erebidae) just beginning to remove the hairs that have given it protection during the larval phase and use them to construct a pupal shelter. Fig. 2(b) shows the completed shelter. It is a dome composed of a rectangular lattice of long hairs bound together with silk. There is a valve arrangement of the hairs at the anterior and posterior ends. The anterior valve is to facilitate adult emergence, the posterior is the route through which the cast larval skin has been ejected. The pupa is suspended within the dome in a silken hammock, avoiding contact with any surface. How this cocoon provides protection is unknown but, since its construction is a one-off sequence using a fixed reservoir of silk and supply of hairs, the behavior is likely to be quite stereotyped and very much under genetic control. Bowerbirds (Ptilonorhynchidae), however, show that building behavior may also exhibit a high degree of variation, flexibility and innovation. The bower of the male vogelkop, Amblyornis inornatus has a central tower of sticks built around a thin sapling. This ‘maypole’ is, at 2.5 m, about ten times the height of the builder and stands in the centre of a circular arena constructed of darkly stained moss. Outside this arena are assembled distinct piles of a variety of dull-coloured ornaments such as acorns and snail shells that may be several hundred in number. But in this species there are also distinct regional variations so, in another part of its range, males build a tent-like canopy over the maypole. In the cleared forecourt they arrange small piles of ornaments that vary between individuals but are sorted by colour and kind (Fig. 3). The bower forms a courtship stage where, in the presence of the female, the male produces a continuous stream of song that includes precise mimicry of the songs of other birds and human generated mechanical sounds (Frith and Frith, 2004). Animal architecture covers the whole range of behavioral complexity and balance between genetic control and developmental flexibility. It is a microcosm of animal behavior.

Object Manipulation What makes the study of animal architecture a coherent and distinct discipline is that it involves the interaction of behavior with materials. This interaction is a dialogue. Materials constrain and guide building behavior and self-secreted materials co-evolve with the building behavior. In this essay I am looking for themes in the study of animal construction behavior that illustrate this relationship. In doing so I have taken a broad view by looking at all behavior where objects are manipulated in order to achieve a functional

Fig. 2 (a) Caterpillar of the moth Cyana sp. beginning to remove its long body hairs and silk them together to create a protective cocoon. (b) In the completed cocoon of the moth Cyana sp. the hairs are arranged in a rectangular lattice with a valve arrangement at the anterior and posterior ends. (Photo: Vijay Anand Ismavel).

Overview Essays j Animal Architecture

3

Fig. 3 The maypole bower of the vogelkop bowerbird (Amblyornis inornatus) has a stick canopy. In the cleared forecourt numerous ornaments are neatly arranged in piles of different kinds and colours. (Photo: Mamoru Suzuki).

end; that is to say, the biology of object manipulation. To me, this includes tool use and tool making, however, this has traditionally been viewed as occupying a distinct behavioral domain and so its inclusion here needs a word of explanation. Tool related literature is unusually preoccupied with definitions. Beck (1980) defined tools as objects which, when in use, were detached from the substrate and from the animal itself, a definition essentially retained in Shumaker et al. (2011). Beck (1980) acknowledged the connection between tool behavior and building behavior more generally but was concerned that its inclusion would make the subject ‘meaninglessly broad’. Bentley-Condit and Smith (2010) make the same distinction, arguing that it is necessary if ‘. tool use is to remain a useful classification’. None of these publications claims that there is a necessary relationship between tool behavior and causal understanding, however for St Amant and Horton (2008) the distinction between tool behavior and building fixed structures is not pragmatic. They emphasise the goal of the tool user in the dynamic interaction between the freely manipulated tool and the environment, a feature that distinguishes tool use from the construction of fixed objects. Tools have proved difficult to define compactly and comprehensively. For example, Shumaker et al. (2011) include as tool makers caddis larvae that construct portable cases but not ones that attach them to the substrate; Bentley-Condit and Smith (2010) include bowerbirds as tool users on the grounds that bower ornaments are tools but that the fixed structure of the bower is not. Ants of the species Aphaenogaster rudis in placing pieces of soil into liquid food and carrying them back to the nest qualify as tool users. However, ants of the species Formica montana use pieces of soil to absorb the sticky liquid produced by the spittlebug Philaenus spumarius and then use this as a composite building material to create a shelter over honeydew-producing aphids (Henderson et al., 1990). So, is that tool use or building behavior? In Hansell and Ruxton (2008) we propose that tool behavior and the building of fixed structures share so much of their biology that to attempt to define a division between them is a distraction that is unhelpful to research. That is the position I have adopted in this essay.

Materials Sources of Materials All non-human animal builders obtain their materials from only two basic sources: They collect them or the synthesise them. Bearing in mind the number of possible building materials, and their varied origins and properties, this section goes into some detail in examining the interaction between these materials and building behavior. Materials of whatever origin may be processed before use. Leaves and sticks may be cut to particular sizes or shapes, soil and water may be blended to make mud. The wax used by honeybees to make honeycomb is a combination of a secretion from beneath the abdominal sternites of young workers and an oral secretion mixed in through mandibulation. The various pathways to the production of building materials are shown in Fig. 4. Materials will not necessarily be used in a mechanical or structural role (e.g., they may be used as insulation), however, I am going to concentrate on structural materials as that is their most common and important role.

Collected Materials The most frequently used collected building materials are of plant origin although those of mineral and animal origin can also be found. What is true of all of them is that none was designed to perform the function for which the builder will use them; most will therefore need some preparation before use. Mud is an example of a material that needs to be in a particular state during

4

Overview Essays j Animal Architecture

Fig. 4 The origins of building materials and the pathways they take prior to incorporation in the built structure. Hansell, 2005. Reproduced from Animal architecture, Oxford, OUP.

construction but then performs its structural role once it has dried out and become rigid. The house martin, Delichon urbica collects wet mud where it finds it, however, the eumeninae wasp Zeta abdominale creates its own mud by bringing water in its crop to dry soil (Hansell, 2005). The construction of the nest of the Southern masked weaverbird Ploceus velatus (Fig. 5(a) and (b)) also illustrates the need for a material to have different properties during the construction phase and in the completed structure. The male tears long, thin strips of fresh grass leaf. These are flexible but strong in tension because of the long, tough veins that run in parallel along their length. This allows them to be knotted, looped and woven to make what is ultimately a hanging basket. These grass strips then dry and, in the process, shrink to make the structure more rigid. Paper is a material of plant fibres or fragments mixed with a liquid medium so that, during the construction phase, it can be moulded into the desired shape. The liquid may be just water but in some cases includes a self-secreted component. Either way, on drying, the plant fragments become tightly bound together to make a light structure that, in some species, can bear heavy loads in tension. Its potential is fully expressed in the nests of social Vespidae, although forms of it can be seen in the nests of some arboreal ants and termites (Hansell, 2005).

Overview Essays j Animal Architecture

5

Fig. 5 (a) The behavior of the southern masked weaver (Ploceus velatus) in creating the initial nest attachment does not follow a predictable sequence. (b) The ‘ring’ stage of nest construction by the southern masked weaver defines the shape of the nest. The nest building sequence at this stage has now become more predictable. (Photo: Felicity Muth).

Paper as a nest material is found in all three sub-families of the Vespidae: Vespinae, Polistinae and Stenogastrinae. The Polistinae use a wide range of plant pulps (long fibres, woody chips, plant leaf hairs) typically combined with an oral, glandular secretion and, in some cases, mud (Hansell, 2005). The papers of Vespinae typically contain little or no self-secreted material although the origins of the plant materials vary and are probably adapted to the ecological conditions of the nest. A comparison between the nest material of Dolichovespula norwegica and Vespula vulgaris suggests this. Nests of the former are located in relatively exposed sites, the latter in more protected spaces. The nest envelope of D. norwegica is made of long woody fibres and is stronger in tension than that of V. vulgaris, which has shorter fibres, often obtained from rotted woody sources (Hansell, 2005).

Self-Secreted Materials The relationship between self-secreted materials and building behavior is very different from that of collected materials. Selfsecreted materials are specialised for the job they do and are the result of millions of years of co-evolution with the behavior. They can broadly be treated under three headings: wax, mucus and silk. However, all of these are somewhat generic terms and, as we should expect, any individual example is highly adapted to the functional needs of the producer. Waxes are complex mixtures of long-chain carboxylic acids and long-chain esters (Hansell, 2005). They occur in protective structures made by some Hemiptera and more obviously in the nest structures of many social bees. Mucus is a broad term applied to extracellular secretions with generally sticky or slippery properties. The biochemistry of these when used as building materials is generally undescribed, however, an essential ingredient of mammalian mucus is polysaccharide in complex long-chain molecules functioning as lubricants. Some building materials with these properties are known. For example, they are used in the hydrated state to create the bubble nests that protect the eggs of the catfish Hoplosternum littorale, or the foam nests of frogs in the family Leptodactylidae and others (Hansell, 2005). The variety and complexity of architecture possible with mucus as a building material is particularly well shown in the house built by the small planktonic organism Oikopleura labradorensis (Chordata, Appendicularia, Larvacea). This tadpole-like organism lives within a mucus house into which it draws water through a pair of inlets by the beating of its tail. Across these inlets lie mucus nets with a mesh size of 8010 mm2. The water then passes through a pair of complex food filters of mesh 0.30.3 mm2 before being expelled. The ability of the organism to produce such a complex structure so readily depends upon the production of a specific kind of mucus for each part of the structure from glandular areas located on different parts of its head. The secretions from these initially create a mucus helmet covering the head which the animal then expands with rapid back and forth movements. When this enlarges sufficiently, the organism inserts its tail and, with further vigorous beating of it, continues the expansion of the capsule. Then as the capsule enlarges to become a house, the various architectural details appear. This very simple building behavior is made possible through the use of a group of highly specialised yet related building materials. The most important of the self-secreted building materials is silk. It is produced by a wide range of arthropods including the larvae or adults of 20 hexapod orders. It is extruded as fine threads from glands located in a variety of different parts of the body depending upon the species. Silks have evolved independently a number of times and are used to create a variety of structures; consequently the precise composition of each depends upon the function it will perform. In general terms, silk is composed of long unbranched polypeptide chains, the threads of which are deployed to create an astonishing range of structures, either alone or in combination with other materials. In Trichoptera and Lepidoptera it is used to create individual pupal or larval shelters (Fig. 2(b)). A spider can produce more than one kind of silk, each evolved to perform a particular function. The orb webs of spiders such as species of Araneus and Nephila are designed to bring large flying insects to a halt and then hold them fast until they are killed. To do this the spiders produce six different kinds of silk and a silk-like glue. Glycine and alanine are

6

Overview Essays j Animal Architecture

prominent in all of these although the proportions of them and other amino acids differ in each and determine their mechanical properties. Draglines and orb web frame threads are produced from the major ampullate gland (MA). The silk of the minor ampullate gland (MI) is used for the radii, while the flagelliform gland silk forms the core of the spiral capture thread. This combination of materials is what absorbs and dissipates the kinetic energy of a flying insect (Vollrath and Knight, 2001). The MA silk of the orb web builder Araneus diadematus is characterised as having the strength of high tensile steel, however, in another respect it is a very different material. Under tension, this silk is capable of extending 27% before breaking. The steel, on the other hand, is designed to be stiff, extending only 1% before failure. Flagelliform gland silk is less stiff even than MA silk, being capable of an impressive 270% extension before failure. Due to the conformation of the silk molecules, the extension of these silk threads is viscoelastic, giving them 65% hysteresis, so dissipating the majority of the energy of impact of an insect as heat (Gosline et al., 1999). Finally, the secretion of the aggregate gland, which forms sticky droplets on the surface of the capture threads, acts as the glue that prevents the escape of the insect prey.

Materials Predictions In view of the two contrasting origins of building materials, collected or self-secreted, Hansell and Ruxton (2013) and Hansell et al. (2014) tested a series of predictions on how building biology might differ between the two groups. These reveal that the two fundamentally different origins of the materials do not produce such clearly different building behaviors as anticipated. The specialised nature of secreted materials does however lead to more reliable predictions on the biology of secretors themselves. A prediction on the size limits of structures in tension or compression reveals interesting complications. Prediction 1. Standardisation Standardisation of building materials should favour both collectors and synthesisers because it should allow the simplification of construction and contribute to the orderliness and consequent structural integrity of the completed artefact. The most speciose group of animals to make structures almost entirely out of collected materials are the birds. Their nests are often quite speciestypical, implying standardisation (Hansell, 2000). However, the case building of caddis larvae illustrates that, while there are selection rules favouring standardisation, flexibility may also occur. This allows sub-optimal solutions and trade-offs against selection forces such as predation risk, associated with extensive searching for the best materials. Caddis larvae evicted from their cases initially relax selection criteria and built a temporary structure from whatever is locally available, becoming more selective again as their bodies became covered (Hansell, 2005). Caddis larvae have also been shown to alter selection criteria in response to perceived predation risk. Calamoceras marsupus builds cases of tough leaves or sticks when exposed to the odour of a predatory fish but of soft leaves when the odour is of a harmless crayfish. Flexibility in the nature of building materials also turns out to be a feature of some self-secreted materials. Although the synthesis of specialised silk proteins by spiders imposes a substantial degree of standardisation on both the material and its use, there is a significant degree of variability. Regional variation has been noted in the amino acid composition of MA silk of Argiope argentata and Nephila pilipes. In N. pilipes a rich protein diet was found to alter amino acid composition as well as enhance the diameter of the MA-composed frame threads and their stiffness (Tso et al., 2007). Prediction 2. Repertoire size This predicts that the behavior repertoire of material collectors will be more complex than for secretors because they will need to search for the materials and test for their suitability before use whereas secretors will not. Material collectors may also need to transport the materials to the building site and possibly engage in some processing of the materials before use. This prediction proves to be an over-simplification for several reasons. Firstly, as already indicated, self-secreted materials may not be entirely standard. Secondly, there is the problem of getting the structure started that faces all animal builders whatever the materials (Hansell, 2000, 2005). Creating the foundations of a fixed structure requires a builder to be somewhat flexible in the choice of building site in order not to limit its options. Consequently, it must then behave flexibly to accommodate variations in the local topography to create for itself a standard template (Fig. 5(a)). Only then will it be in a position to use simple construction rules to advance and complete the building process (Fig. 5(b)). A third complication is that, for some material secretors as well as collectors, processing of the building material may be necessary before use. The blending of honey-bee scale wax with an oral secretion to create comb wax has already been mentioned. Another example is the creation of capture threads on the webs of cribellate spiders by the combing onto the axial thread a fine ‘mist’ of cribellum silk threads using a special leg structure, the calamistrum. Finally, not all building material secretors avoid an element of collecting and carrying building materials. This takes the form of nest material recycling. For example, honeybees, after the emergence of an adult from a brood cell, may dismantle it and recycle the material to create new comb. The worker bee is essentially incorporating into its behavior repertoire a searching and transporting phase, albeit on a local scale. Prediction 3. Dietary constraint The raw materials of self-secreted materials originate in the diet of the builders. From this we can predict that the evolution of self-secreted materials has been significantly determined by the availability of a suitable raw material already in the ancestral diet. This prediction can be tested by the examination of wax and silk secretions; the former is carbohydrate rich, the latter protein rich (Hansell and Ruxton, 2013).

Overview Essays j Animal Architecture

7

Plant sucking bugs (Hemiptera) generally uphold this dietary constraint hypothesis. Aphids are well known for excreting the excess sugars in their diet as honeydew which attracts ants that, in turn, protect them from predators. A number of species of plant sucking bugs from three different superfamilies (Aphidoidea, Coccoidea, Psylloidea) use the excess sugar in their diets to make protective structures that are known to contain waxes (Hansell, 2005). The most conspicuous use of wax as a building material is by social bees. Here the explanation is more complicated. Adult honey-bees collect sugary nectar in excess of their immediate needs and use the surplus raw material to synthesise wax to make the cells in which the food can be stored as honey. However, mature bee larvae spin silken pupal cocoons and for this they must use some of their larval protein intake from pollen that would otherwise go into growth. Spiders, all being carnivores, have a protein rich diet and, as the dietary constraint hypothesis predicts, are builders with silk. However, an indication that this has been a high cost in the ecology of some spiders is the loss of web making in some lineages (Blackledge et al., 2009). For builders using self-secreted materials other than sap sucking bugs and spiders, the predictions are not so easy. Of particular interest are the Lepidoptera, where larval silk production is almost universal but varies greatly between species in the amount produced. The diet of Lepidoptera is generally herbivorous and in some species such as wood borers, very nutrient poor. However at the other extreme, there are in Hawaii six species of the genus Eupithecia where the caterpillars are ambush predators of other insects. Silk expenditure by lepidopteran larvae in relation to dietary protein is a subject worth investigating. Prediction 4. Castes of secretors In social insects much research has been devoted to social structure and caste formation at the behavioral level. An aspect of this that deserves more attention is the possibility of unequal contributions of self-secreted building materials by group members. A well-known example is that honey bee scale wax secretion begins in an adult worker a few days after emergence and ends a few days later. In this respect it resembles the age-related job specialisation that is known in a variety of social insects. More interesting is the relative contribution of individuals in terms of self-secreted building materials in species that live communally but are capable of independent reproduction. There are 20 families of Lepidoptera in which the caterpillars live communally and, in some of these, there are species that live in collectively-built silk structures. Suggestive evidence of individual specialisation in silk production is found in the caterpillars of Eucheria socialis (Pieridae). In their larval phase there is a strongly male-biased sex ratio. Male larvae also contribute more silk, grow more slowly and have a higher larval mortality than females (Underwood and Shapiro, 1999). The possibility of unequal material contributions like this suggests further predictions. For example, some species of bees and wasps live within the colonies of social species exploiting the behavior of host workers. They are known as inquilines and in some cases have evidently evolved directly from the species they parasitize. It seems possible that a lepidopteran species could similarly evolve to live within a communally built silk structure to which it makes no contribution. Prediction 5. Self-secreted materials are needed for prey capture devices This predicts that prey capture devices will be composed of self-secreted materials because they need precise engineering whereas, for home building, self-secreted materials are often not essential. This prediction is broadly supported by the evidence. Animal-built traps are overwhelmingly of self-secreted materials. The composition of homes, by contrast, is very variable and frequently of collected materials alone. Traps in the form of nets and webs of mucus or silk are relatively common among the invertebrates. These devices, by restraining prey or food particles, have to resist tension stresses. To be effective they also need to be precisely engineered, a feature well illustrated by the mucus filters in the houses of the Chordate Oikopleura and in the specialised silks of spider orb webs. This prediction, although confirmed, cannot be applied to true vertebrates because not one among them builds a trap or capture net, excepting that is, we humans. Homes can be built on the ground or otherwise supported from below and so largely experience stresses of compression. They can be of a wide variety of materials, generally collected materials, and exhibit simple engineering principles. Very large structures in compression such as termite mounds and smaller ones such as bird nests show that these are common among both vertebrates and invertebrates. However, in the birds there are a significant number of species that make hanging nests of collected materials which, naturally, bear loads in tension. Two possible reasons suggest themselves for the lack of a vertebrate self-secreted material to bear loads in tension. One is of evolutionary constraint, i.e., that, for whatever reason vertebrates never evolved the physiology to produce a suitable material, such as silk. However, more likely is that the cost of the structure required to support a load in tension increases more rapidly with the increasing weight of the load than the cost of supporting the load in compression, so imposing an upper economic limit (Hansell et al., 2014). Prediction 6. Size limits on structures in compression and tension For a load supported in tension, its weight will vary in proportion to its volume, i.e., in proportion to the cube of its radius. However, the strength of the supporting cord will vary in proportion to its cross sectional area, i.e., in proportion only to the square of the radius. Therefore, as the weight of a hanging structure increases, so the cost of the cord supporting it will become proportionally greater. Add to this the need for attachments at both ends of the cord, the costs of which are similarly affected, and the relative advantage of supporting the load in compression becomes apparent. This effect makes the prediction that the maximum weight of a hanging bird nest, for example, will be significantly less than the heaviest bird nests in compression. This prediction is upheld. The nest of the hamerkop, Scopus umbretta is a massive pile of sticks that may weigh hundreds of kilos, while the heaviest bird nest in tension is probably the side-attached mud nest of the white-necked rock fowl, Picathartes gymnocephalus that weighs 2.0 kg (Hansell, 2000).

8

Overview Essays j Animal Architecture

The evolutionary constraint hypothesis is more difficult to test, however, data from bird nests suggest that, even if birds had been able to synthesise a material like silk, it would not have allowed them to construct much larger hanging nests. Arthropod silk is in fact a component of the nests of some species belonging to 36% of passerine bird families (Hansell, 2000) but all these species are small. A fairly typical example is the nest of the origma, Origma solitaria which hangs from a suspension composed almost entirely of arthropod silk and weighs only 33 g (Hansell et al., 2014). It seems that the properties of a building material resembling arthropod silk and secreted by birds would be of limited use in their nest building. There is an interesting discrepancy between the maximum size of suspended structures built by vertebrates and invertebrates. The largest hanging bird nest that I have been able to weigh as a dry museum specimen is that of the dusky broadbill Corydon sumatranus, which is made of and supported by a suspension of fine woody plant stems; it weighed 505 g. Even allowing for the fact that this nest was weighed when dry, the suspended nests of social insects can be much heavier. Nests of the wasp Polybia scutellaris may reach 50 cm in diameter and as empty museum specimens, weigh 4650 g. The heaviest of the suspended social insect nests, however, seems likely to be that of the giant honeybee Apis dorsata. It takes the form of a single top-suspended, wax comb about 150 cm across and 100 cm deep. A comb can have around 25,000 cells containing either brood or honey. For protection, the comb is also completely covered with a curtain of 35,000 adult workers; they alone may weigh 4500 g. There seem to be two possible explanations for the very large differences between the maximum weights of suspended structures of vertebrates and invertebrates. The first is the marked difference in construction anatomy between vertebrate and invertebrate builders and the second is the difference in life expectancy of the relevant structures. The vertebrates that build the most intricate suspended structures are the birds. To achieve this they predominantly use their beaks although these are not ideally suited to the task, nor are they specially anatomically adapted for it (Hansell, 2000). Insects on the other hand, have small jaws and, in the case of social insects, large workforces. This makes them well adapted to make very large structures with fine architectural detail. The second possible explanation for this weight difference is that large structures are also investments in the future. A bird nest is typically a structure of quite short life expectancy. Most are built to raise one clutch of eggs to the fledging stage, a working life of only a few weeks (Hansell, 2000). The large nests of social hymenoptera can have very long lives. Although accurate figures are lacking, it has for example been speculated that the age of a large Polybia scutellaris nest might be 30 years.

Functions There are essentially only three functions for built structures. They are: as homes, food gathering devices and as signalling structures. Creating a home or protected living space is the most common of these and also the one with the most diverse taxonomic representation. It is almost universal in birds, common in small mammals and present in a variety of invertebrate phyla, particularly in the Arthropoda. Food gathering devices differ markedly in type between invertebrates and vertebrates, as explained above. Specialised communication structures while having a scattered occurrence across the phyla, are relatively rare. Animals have a variety of media available to them through which they can communicate (visual, auditory, tactile, chemical and electrical) without building any structure to supplement them. The small number of examples of specialised constructed signals may be a reflection of this. However, some traps and homes are known to also communicate information, for example the quality of bird nest building in mate choice (Hansell, 2000), but the scarcity of such examples probably reflects lack of research. By far the most complex, specialised constructed displays are the bowers made by the males of several species of bowerbirds (Ptilonorhynchidae) (Fig. 3). These are discussed in more detail in Section “Control and Organisation of Building Behavior”. The prey capture and feeding devices of vertebrates and invertebrates differ in the materials they use and the foraging principles they employ. With rare exceptions, the feeding devices of invertebrates are webs or nets and are made from self-secreted materials. An example is the mucus net of the marine polychaete Chaetopterus which is deployed across the burrow and captures tiny food particles that are driven into it by the beating of the worm’s fan-like notopodia (Hansell, 2005). Silken filter nets, often of an extremely regular mesh size, are produced by caddis larvae of three different families (Hydropsychidae, Polycentropidae and Philopotamidae). The mesh size of the net of each species is adapted to the rate of water flow and hence of food particle size. In Macronema transversum, a species adapted to slow flowing water and therefore feeding on very small particles, the mesh size is only 253 mm2 (Hansell, 2005). Spiders provide the most numerous examples of silken prey capture devices and, although some take the form of aerial filter nets, the orb web being an example, they are much more varied in design and capture principle than those of caddis larvae. Some use three-dimensional arrays of threads to delay the escape of prey. Others employ one of two types of capture thread; for the cribellate spiders it is the dry so-called ‘hackled’ silk and, for the ecribellate spiders, sticky droplets. Vertebrates, where they employ artefacts to aid prey capture and feeding, use tools. These are made of collected materials and are actively manipulated to be effective. Their function is to allow the user to gain access to prey that is concealed or in otherwise inaccessible cavities. Examples of this are the use of long thin flexible stems by chimpanzees Pan troglodytes to extract termites from their mounds, and the use by New Caledonian crows Corvus moneduloides of hooked tools made from Pandanus leaf, to extract beetle larvae from burrows in wood (Shumaker et al., 2011). Homes are the most common built structures. These vary greatly but all incorporate some form of wall or barrier to protect the builder from physical and biological hazards. Both endothermic and exothermic species build structures that help to regulate environmental temperature. For endotherms such as a bird incubating eggs, the thermal priority is to maximise heat transmission from parent to clutch. A layer of insulating materials lining the nest cup performs this function. For exotherms, the surrounding wall or

Overview Essays j Animal Architecture

9

envelope may be used to trap solar energy. Caterpillars of the small eggar moth Eriogaster lanestris live communally in multi-layered silk tents that maintain the temperature above ambient even in the absence of direct sun, so accelerating larval development. One possible consequence of creating a protective wall is the reduction of oxygen availability to the builder. This has led in some species to architectural devices that promote ventilation. Termite mounds with large populations of insects contained within massive walls have adapted to this problem by evolving two different systems of ventilation, one in which air is circulated within the mound driven by temperature differences and the other by air pressure differences (Hansell, 2005). The architecture of the latter is characterised by distinct apertures projecting from the mound at different heights. Wind passes over the more elevated of these at a greater speed than the lower ones. This creates a pressure difference between them by the Bernoulli effect, drawing air into the lower and out of the more elevated apertures. This ventilation principle is seen in the architecture of the termite mounds Macrotermes subhyalinus, and the large subterranean mounds of the ant Atta vollenweideri. It is also a design feature of the burrow systems of the black-tailed prairie dog Cynomis ludovicianus and, by means of water currents, the burrows of the marine fish Valenciennea longipinnis (Hansell, 2005). Homes can provide protection from the hazards of the biological world in essentially two ways, as a physical deterrent or by camouflage. The deterrent may be simply mechanical, as in the tough envelope of the nest of the social wasp Polybia scutellaris or chemical, as appears to be the case in the ant-deterring material smeared on the nest petiole by species of polistine wasp, for example, of the genus Mischocyttarus (Hansell, 2005). Camouflage is a defensive technique that is, necessarily, more effective on small structures. Terminology is not always consistent in this area but I distinguish three types: concealment, crypsis and masquerade. In the first the home is made invisible simply by being hidden from view, like a bird nest in a thicket. In crypsis, the structure is exposed to view but, by blending into the background, is not noticed. An example of this appears to be the external covering on some small bird nests of pale lichen flakes and/or white spider cocoons (Hansell, 2000). An example of masquerade is the hanging nest of the red-faced spinetail, Cranioleuca erythrops. It has streamers of loose vegetation hanging above and below the egg chamber so that although clearly visible, it resembles a clump of hanging forest debris (Hansell, 2000).

Control and Organisation of Building Behavior Building Anatomy A structure must be built so that its component parts stay together for the duration of its functional life. As animals are limited anatomically to the size of load they can carry, most structures are a combination of a large number of elements. For structures that are well supported from below, gravity may be sufficient to hold these pieces together. However most structures will experience some tension stresses, and this requires the building elements to be secured together in some way. Animals have evolved a diverse range of methods to achieve this (Hansell, 2005). Under the general heading of interlocking and weaving, I identify: Velcro fastening, stitches, pop-rivets, weaving, and entangle. I also identify five additional distinct categories: sticking together, folding and rolling, spinning, modelling and sculpting, where modelling is the shaping of a plastic material such as mud or wood pulp, and sculpting covers burrowing. In view of these varied methods of construction and assembly across the animal kingdom, it is very striking that the range of anatomical features used by all these builders is very narrow and consistent. They are either jaws, legs or both. The use of jaws is unsurprising; they are the anatomy of feeding and this may involve the manipulation of the environment to reveal food or prepare the food for ingestion. However, in virtually all cases where jaws are used in construction, they retain their feeding function. Legs are the organs of locomotion and are usually also equipped with devices at the ends of the limbs with an ability to grip the substrate. The flexibility of limbs and their ability to grip has allowed them to become organs of object manipulation and, in some species, to being specialised for that purpose. In the case of the jaws, the extent of anatomical specialisation for building turns out to be rather small (Hansell, 2005) not simply because they have retained their feeding function but more especially because feeding remains overwhelmingly their primary function. This is well illustrated by the beaks of birds. Use of the beak is essential for nest building but this behavior may only be shown for a few days a year. The shape of a bird beak is a good indicator of feeding specialisation but provides no clue as to the nature of the nest (Hansell, 2005). More detailed study might reveal anatomical specialisations in the beaks of males where they are the sole or major nest building sex. In mammals, a study of the jaws of tent-making bats (Chiroptera) could also be rewarding. A number of tropical species from three families bite through the supporting stems or ribs of large, living leaves that then droop to create tent-like structures which function as diurnal roosts (Choe, 1997). This appears to be a year-round activity and, as mammal teeth are functionally differentiated (e.g., incisors adapted for grooming fur in some mammals), some specialisation of teeth in these bats for biting through tough plant material may have evolved. Specialisation of leg anatomy for building is more obvious. Animals with legs generally possess more than one pair of them. In some caddis fly larvae, the first pair of legs is shortened apparently to facilitate the manipulation of building materials, leaving the two other pairs primarily for walking. Among the vertebrates, object manipulation using paired limbs is generally scarce. Birds, with over 9000 species most of which build nests, are the most conspicuous vertebrate builders, however, there are a similar number of species of fish that prepare some sort of a nest. In neither of these two vertebrate classes is there a free pair of limbs to become specialised for building. Only in the primates do we see forelimbs with organs resembling hands at the ends of them. Only in the great apes do we see hands with sufficient dexterity to make tools, and even then this is not their primary function. In general therefore, anatomical adaptations for building, whether of jaws or limbs are surprisingly modest. There is

10

Overview Essays j Animal Architecture

however one exception: adaptations for burrowing where power is needed to excavate the substrate. This is evident, for example in the much enlarged jaws or forelimbs for burrowing that have evolved in some insects and burrowing mammals (Hansell, 2005).

Cognition and Aesthetics There has been a persistent tendency to regard the building behavior, even of higher vertebrates, as being largely under genetic control (Seed and Byrne, 2010). This is surprising in view of the known learning skills and behavioral flexibility demonstrated in other aspects of bird and mammal behavior. By contrast, tool use exhibited by vertebrates has generally been perceived as evidence of intelligence, due in part to it originally being regarded as a uniquely human attribute. The observations of tool use by chimpanzees in the wild (Goodall, 1964) was therefore an important stimulus to the study of tool behavior of higher vertebrates in order to understand better the evolution of human intelligence. The discovery of tool using birds as well as primates then widened the interest in research on tool users to the study of animal intelligence more generally. The simple structure of animal tools also offered an ideal model system for experimental manipulation. The result has been a flourishing of this area of research and important discoveries, particularly on cognition in primates and birds. Chimpanzees in different parts of their range are now known to use a variety of tools in a number of ways to gain access to foods as varied as termites and nuts. Boesch et al. (2009) observed chimpanzees obtaining honey from three different species of bee, two arboreal, one subterranean. For each species they fashioned and used a set of up to five stick tools, each one with a particular function and used in sequence to locate a nest, enlarge access to it and extract the honey. Among the birds, the most experimentally productive species has been the New Caledonian crow. It can shape and use three kinds of tool: a straight stick, a stick with a hooked end and a strip of the edge of a Pandanus leaf, tapering in steps towards one end and bearing hooks along the outer edge. All are used to extract insect prey concealed in cavities in wood. Evidence of cognitive processes involved on the making and use of these tools are now available from a number of studies (Seed and Byrne, 2010). The hooked stick tool cut from a tree by a New Caledonian crow has three characteristic features: a hook and, at the same end, the bark stripped away and the shaft curved. Presented with artificial tools where the spatial relationship between these features was changed, the crows handled and used them in a way that demonstrated, without trial-and-error learning, that they understood the functional properties of the tool (St Claire and Rutz, 2013). New Caledonian crows have also shown apparent understanding of the value of a tool and its future usefulness. It takes time for them to make these tools, which must then be conveyed to the foraging site. In the wild, New Caledonian crows will ‘safekeep’ the tool between instances of use by holding it under their foot or parking it safely in a tree hole. Klump et al. (2015) showed that in a captive environment, the birds would ‘safekeep’ tools significantly more often when foraging higher off the ground or when foraging for difficult prey. Studies on tool behavior such as these have tended to reinforce the perception that tool behavior contrasted with the essentially genetically determined behavior of bird nest building. However recent evidence has demonstrated that this view is mistaken and that bird nest building should also be regarded as a valuable model system for the study of cognition (Breen et al., 2016). The southern masked weaver Ploceus velatus shows evidence of learning and flexibility in the early stages of nest building (Walsh et al., 2011). Attachment of the first grass strips to the overhead twig suspension (Fig. 5(a)) did not follow a predictable sequence, but became more predictable as the building proceeded into the ‘ring’ stage (Fig. 5(b)). Males also improved their handling of the grass strips as they built more nests. Flexibility in the choice of nest building material in the Zebra finch, Taeniopygia guttata was shown by Muth and Healy (2011). Males that were successful in raising chicks in a nest that they had been obliged to build in a non-preferred nesting material, tended to change preference to this material when next offered a choice. Zebra finch males have also demonstrated observational learning in the choice of colour of nest materials. An inexperienced male, observing another male nest building and then being allowed to build itself, changed its preference to that it saw being used, provided the model bird was familiar to it (Guillette et al., 2016). Zebra finch males have also been shown to have some understanding of the mechanical properties of nest materials. Naïve males presented with stiff or flexible pieces of string increased their preference for the stiff string which they were able to use more effectively (Bailey et al., 2014). Some understanding of the properties of building materials has also been observed in great apes. Casteren et al. (2012) found that orangutans, Pongo sp. used the principle of the green-stick fracture in weaving living branches to make a nest that will hold firmly together. They chose thicker branches with greater rigidity and strength for the main structure and thinner branches for the lining using a green-stick fracture and twist technique. There is also evidence now that the building behavior of invertebrates may not be stereotyped and fully under genetic control. The change in building material preference of a caddis larva in response to a perceived predation threat has already been mentioned. The orb web spider Larinioides sclopetarius, will increase the area of web above the hub if repeatedly offered prey in the upper rather than the lower part of the web (Heiling and Herberstein, 1999). The finding of a correlation across bird species between tool use and the complexity of the cerebellum has suggested that this behavior places special demands on manipulative skills (Iwaniuk et al., 2009). However, Hall et al. (2013) have also found a similar correlation in birds between cerebellum structure and nest complexity. Day et al. (2005) comparing the bower complexity of five bowerbird species have also found a positive relationship between it and cerebellum size. These findings on physical cognition in tool use and nest building raise the question of whether there are specialised central neural systems for different sorts of task or a more dispersed one. Bird and Emery (2009) found that rooks Corvus frugilegus, which do not show tool use in the wild, were able, in captivity, to use a variety of tools even spontaneously modifying them to obtain a food reward. The authors concluded that this demonstrated that tool behavior is a constituent of a ‘domain general’ cognitive

Overview Essays j Animal Architecture

11

capacity rather than a specialised system. Lefebvre (2011) reviewing the overall occurrence of innovation and intelligence of animals in the wild draws a similar conclusion. The bowers built by male bowerbirds are among the most elaborate fixed structures built by any vertebrate and their quality is known to influence female mate choice. However, it is important to recognise that they are only the static setting for a dynamic vocal and visual performance by the male himself. The ‘maypole’ design of the bower and the vocal mimicry of the male vogelkop have already been described. The other basic bower design is the ‘avenue’. That of the male satin bowerbird, P. violaceus is an avenue of twigs in which the female stands while the male displays. It is oriented North-South and at the North end the male places a variety of ornaments, particularly blue ones. He may also paint the inside of the stick avenue with a combination of saliva and berry juice. In the presence of the female, the male exhibits vigorous wing and tail movements, repeatedly picks up ornaments and emits a stream of vocalisations including mimicked calls (Frith and Frith, 2004). Another avenue builder, the great bowerbird, Ptilonorhynchus (Chlamydera) nuchalis performs his display on a platform of white shells and bones that may number hundreds. These are graduated in size with the smallest ones nearest to the end of the stick avenue in which the female stands, creating for her a perspective illusion. It is on this stage that the male performs his display of vigorous movements and complex vocalisations. After experimental removal of all the ornaments from the display area, the male great bowerbirds have restored their graduated arrangement after replacing only ten of them, apparently emphasising its importance (Kelly and Endler, 2017). The energy, detail and individuality seen in male bowerbird’s displays suggest that they could provide information to females on a male’s manipulative skill and inventiveness, not simply his physical fitness. Darwin (1871) in The Descent of Man claimed that bowerbirds had an aesthetic sense and that ‘... the playing passages of bowerbirds are tastefully ornamented with gaily-coloured objects; and this shews that they must receive some kind of pleasure from the sight of such things.’ This claim was largely rejected for about a century until bowerbird courtship displays had been studied in detail and their complexity better understood. This evidence suggests that the possibility of an aesthetic sense in bowerbirds should be taken seriously. Hansell (2007) proposed an ‘art school hypothesis’. This was that male bowerbirds learn to be artists and females learn to be art critics. There is evidence that supports the former. Males take several years before they can successfully attract a female, during which they practice bower building and inspect bowers built by older adults. A weak but suggestive piece of additional evidence is that males show a fastidious attention to detail in the structure and ornamentation of the bower. For females there is much less evidence. They do visit and revisit several males before choosing a mate, but we have no evidence that their judgement of male quality changes over time. Critically, we lack evidence of any special brain function in bowerbirds that can be linked to aesthetic judgement. However, I feel that with the increasing sophistication of brain scanning techniques such as fMRI, evidence to test this may soon become available (Hansell, 2007). Miller (2000) proposes that an aesthetic sense might be selected for where a courtship display contains a great deal of information delivered through a variety of sensory channels. A pleasure centre would be a mechanism for simplifying this information, making comparison of rival males easier for females. Miller (2000) is talking in the context of human evolution but I am suggesting that the same reasoning could be applied to the problem facing a female bowerbird in choosing a mate. However, I believe that the context for evolution of a bowerbird aesthetic sense would have been very different. We humans are a very social species. Our close personal relationships are based on a complex of personality, shared interests and skill attributes. Our sexual partnerships typically develop gradually and may last for decades. These can help to explain the significance in human culture of dance, song and the elaboration of language, from all of which we derive pleasure. Bowerbirds, on the other hand, are not social species. Males display in a dispersed lek and a female, after mating, builds a nest and raises her chicks unaided. A female bowerbird, after exposure to the complex displays of several males, must therefore decide which is to be the father of her offspring after a very brief exposure. A pleasure mechanism such as suggested by Miller (2000) could make this decision manageable but would operate in a different way from our own.

Builders Change the World Species that alter the environment in a way that changes the availability of resources to other species are referred to as ecosystem engineers. However this is a very broad definition that might be seen to include a tree that casts its shadow on the ground and draws nutrients and water from it. A related concept, but one with greater significance to animal architecture is niche construction. This is the concept that, by imposing a new structure on the environment, the builder creates new niches. It predicts that these can be colonised by new species, so increasing diversity. However, the contrary prediction can also be made, i.e., that by creating an enduring home, builders come to dominate the landscape, monopolise resources and so limit diversity (Hansell, 2005). Builders redistribute materials in the environment. Darwin was one of the first to quantify this in his calculation that earthworms through the casts that they excreted, deposited on the soil surface an impressive 4.6 kg per sq m per year. Soft marine sediments provide another habitat where high densities of burrowing organisms can alter the environment. Mud shrimps such as species of Callianassa and Jaxea, are deposit feeders whose burrowing activities produce spoil heaps on the mud surface. The burrowing depth of mud shrimps varies, however, Jaxea nocturna for example displaces sediment from as deep as 90 cm (Hansell, 2005). Pocket gophers Thomomys displace large amounts of soil in their burrowing activity and can occur at sufficient density that their mounds cover 5%–8% of the habitat’s surface area. The endurance of these burrows and their repeated reoccupation has led in some locations to a very even spacing of their burrows that is referred to as ‘mima prairie’. This type of grassland habitat is created not only by burrowing mammals but also, in similar habitats, by termite mounds. Regular spaced mounds with a mean diameter of 28 m in South Africa are apparently formed by the combined activities of a termite species Hodotermes viator and the common molerat Cryptomys hottentotus.

12

Overview Essays j Animal Architecture

Soil disturbance around a rodent burrow entrance can create both patches of bare ground and patches of lush vegetation enriched by faecal material. This vegetation, in turn creates patches of shade with limited vegetation growth (Hansell, 2005). Different plant species are suited to these different conditions, fostering diversity. A marine example of patchiness in an otherwise uniform habitat creating species diversity is the nest mound built by the sand tilefish Malacanthus plumieri. It digs burrows in the sandy substrate just beyond the coral reefs. However, it is the large mound of coral rubble that the fish builds over the burrow, not the burrow itself, that creates habitat opportunities for other species. These mounds have mean dimensions of 1.331.020.22 m3 and attract a wide variety of invertebrates and 32 species of fish (Fig. 6) (Büttner, 1996). There is therefore good evidence that, in terrestrial and marine habitats, building and burrowing behavior produces patchiness that enhances biological diversity. It is not only built structures themselves that can influence the diversity of other species associated with them, but the activities of the resident builders. This is illustrated by the seed storage behavior of a wide variety of burrowing rodents, particularly in the families Cricetidae and Heteromyidae, and of some ants, chiefly in the Myrmicinae. This seed-storing behavior is particularly evident in habitats where seed production is determined by short periods of unpredictable rain. In some North American desert habitats up to six species of seed eating rodents may co-exist. This co-existence is facilitated by the size of the seed collected being correlated with the size of the rodent. However, in less productive habitats, species with similar seed preferences are most likely to be excluded (Hansell, 2005). Species that take advantage of niches created by builders may do so in an opportunistic, facultative way or in an obligate relationship. In the latter case, the specialisation may be parasitic or even symbiotic. Even very simple built structures may create new habitat niches for opportunists. The lepidopteran larva Anacampsis niveopulvella makes leaf rolls as a protective larval environment. Martinsen et al. (2000) found that, experimentally created leaf rolls on trees were colonised by seven orders of insects as well as spiders and mites. These were not species that were otherwise absent from the trees but, when leaf rolls were provided, became much more abundant. The nests of social insects, because of their size, complexity and endurance provide many examples of both facultative and obligate social relationships. The nest-defending power of large numbers of wasps is exploited by as many as 28 species of birds that are known to build their nests close by. This exploitation of wasp nests by the birds is generally facultative, but for the red-cheeked cordon-bleu Uraeginthus bengalus the relationship seems to be essential, although the species of wasp it chooses may vary (Hansell, 2000). Symbiotic relationships are common in the nests of ants. This includes species from 10 orders of insects. In the Coleoptera alone there are 35 families where there are species dependent upon ant nests in some way. Some are beneficial, for example detritus feeders, while others exploit their hosts by soliciting food or preying upon larvae. Animal builders and burrowers can therefore change the environment in ways that have substantial effects on the populations and diversity of other species. In the main, their effects are to create new niches and increase diversity. In some cases however, the dominance and secure position of the builders themselves may exclude competitors.

Evolution Builders, by changing the environment in which they live, alter the selection pressures acting upon them and consequently their own evolutionary path. This section examines this in terms of opportunities for habitat expansion, speciation and the evolution of social life.

Fig. 6 Schematic illustration of the nest mound of the sand tilefish Malacanthus plumieri provides a habitat that attracts various other species. In this illustration: eight species of fish (1, 2, 3, 4, 5, 6, 7, 8), two species of echinoderm (9, 10), one species of mollusc (11), four species of crustacea (12, 13, 14, 15), and one species of polychaete worm (16). Adapted from Büttner, H., 1996. Bulletin of Marine Science 58, 248–260.

Overview Essays j Animal Architecture

13

Social Evolution A possible evolutionary consequence of building a home is that the protection it provides allows the population of the builders to reach the carrying capacity of the environment, reducing the opportunities for individuals to leave the parental home to found a new one. The increase of the number of individuals in the home should then facilitate social evolution (Hansell, 2005). Evidence in support of this prediction is provided by naked molerats, Heterocephalus glaber. They live in extensive burrow systems that have a constant, equitable temperature and have evolved a social organisation similar to that of social hymenoptera, with a queen and workers. This explanation of an evolutionary tendency towards greater social living in home builders can be called the habitat saturation or lack of opportunity hypothesis. However, an alternative hypothesis for the same trend is the cost of nest hypothesis. This is that a reluctance to leave the natal nest is due to the cost of building a new one. The evolution of a fully eusocial shrimp Synalpheus gregalis (that has colony sizes of over 300 and a single reproductive female) has been ascribed to habitat saturation resulting from a lack of suitable sponges in which to live (Duffy, 1996). A social spider example suggests building costs have been a factor. A few dozen species of spiders live socially, generally in collectively built silk structures although none have evolved full eusociality. Agelana consociata feeds on small insect prey and, as social group size goes up, capture rates and individual egg production go down. However, as group size goes up, so individual investment in silk goes down. Riechert (1985) sees this as an important cost saving in a habitat where heavy rains during six months of the year result in web rebuilding two out of every five days.

Speciation Various lines of evidence show a correlation between changes in building behavior and speciation. The nest building of birds is generally fairly species typical but highly variable across the class as a whole. Collias (1997) claimed that this gave birds the ability to construct nests in such a variety of places that it contributed significantly to their speciation and increase in habitat range, although I would add that other aspects of bird biology such as feeding adaptations or courtship behavior should also be regarded as candidates (Hansell, 2000, 2005). Social wasps provide another example suggestive of an association between nest building and speciation. In this case it is the technological innovation of paper as a nest building material for large nests and exploitation of new nest sites that seem likely to have been the stimuli. Bond and Opell (1998) demonstrated that the sticky thread producing clade of spiders (Araneoidea) is significantly more species diverse than the cribellate silk producing clade (Deinopoidea) and that this diversification was at the point of origin of the araneoid group, suggesting a causal link between the evolution of a new capture principle and speciation. There has also been an evolutionary trend in the Araneoidea from larger to smaller size and a reduction in the original orb web design to reduced forms adapted to the capture of special prey types (Hansell, 2005). These spider examples illustrate the two distinct ways in which built structures may change during evolution: changes in technology and changes in design. The innovation of a sticky droplet capture thread is a technological change. Evolutionary changes in which the sticky droplet orb web gave rise to webs with reduced numbers of radii and capture threads are changes of design.

Extended Phenotypes In 1982, Richard Dawkins published The Extended Phenotype, in which he discussed the significance of phenotypes that existed separate from the organisms that created them. This concept has obvious implications for built structures although it could apply to other phenomena such as the manipulation of host behavior by a parasite. It is likely that the species-specific portable case of a caddis larva is largely genetically determined. Dawkins therefore argues that, although the case contains no genes of its own, its success in protecting the builder will affect the allele frequency of genes in the larva for the behavior sequence through which the anatomy of the case is expressed. In this way the structure of the portable case can evolve in response for example to a change in predation pressure, in the same way that other anti-predator behavior might evolve. The situation is, however, more complicated if several individuals together are responsible for the building and they are not genetically identical, as is generally the case in social insect colonies. How can disputes be avoided during the building process and a coherent structure completed? This is where the distinction between evolutionary changes in design and in technology leads to different conclusions. Dawkins (1982) suggests how conflicts over building technology might be resolved by imagining a termite colony in which some individuals gather a dark material and some a light one. The result is a blended building material. This will then be subject to selection pressures which tend to favour, let’s say, the dark material. This will result in an evolutionary trend towards termite mounds with more of the darker material. For the problem of conflicts of design, Dawkins (1982) suggests that it might be possible for the workforce to come to a democratic decision on nest architecture. However, in the enlargement of a termite mound, for example, where the builders are dispersed within the mound, collective decision making would be impossible. If we imagine two different opinions (thresholds) in the workforce for the diameter of a chamber (broad and narrow), what is to prevent some individuals from converting all broad chambers into narrow ones? An interesting insight into how such conflicts might be avoided has come from computer simulations in which virtual ‘workers’ collectively build a nest in an imagined three-dimensional space. These model systems are called ‘lattice swarms’ (Theraulaz and

14

Overview Essays j Animal Architecture

Bonabeau, 1995) and are designed to test whether relatively simple organisms such as social insects could produce complex architecture using only simple behavioral rules. In these models the virtual workers (notionally wasps) are endowed with very limited abilities. They are unable to communicate with one another, have no memory of what they have done, and no spatial awareness beyond their immediate location. They simply move around the available space, adding building material wherever they find an acceptable local configuration. Surprisingly, given appropriate building rules, virtual architecture is created that resembles, in this case, social wasp nest structures. The relevance of this to the problem of conflicts within the workforce in nest design comes from an additional finding from these simulations. This is that only some sets of values given to the builders produce orderly designs resembling the arrangement of wasp nest combs (coordinated algorithms). The majority, however, produce very irregular architecture that is different each time the simulation is run (non-coordinated). Importantly, coordinated algorithms turn out to be very resilient to the addition of random behavioral rules, because rogue workers can rarely find a configuration in the nest structure to which their random rule applies. This suggests that a colony of builders could continue to produce coherent, functional architecture while harbouring individuals that wish to apply different building rules. There is an interesting additional evolutionary point to add to this. It might be possible for colonies producing a particular nest design to accumulate individuals with additional building rules until these conform to a new coordinated algorithm leading to the appearance of a new nest architecture in a single generation (Hansell, 2005).

Environmental Inheritance When structures survive beyond the life of the original builder, the inheritor does not simply inherit the genes for creating a particular extended phenotype but an environment already altered by its ancestors. Some possible consequences of this are explored by Odling-Smee et al. (2003). Imagine a niche-constructing organism that has a pair of alleles at locus E that have the effect of modifying the environment that influences the availability of a resource R that is used by the niche constructor in a way that impacts on the frequency of alleles at locus A. This might be any of a number of possible environmental effects but food availability would be an example. Suppose now that the effect of the alleles at E leads to a change in resource R that is inherited as an ecological change by the next generation, then various predictions follow. For example, if the effect of E on R is small for the next generation, it may take several generations for the level of R to be altered in such a way that it has any impact upon the frequency of alleles at A. The greater the number of previous generations necessary to alter R, the greater the time lag (evolutionary inertia) in its effect on A. This model, has widespread implications for environmental inheritance; for example, in co-evolutionary relationships, where it predicts that selection at the A locus of one organism might be subject to niche constructing influences originating from the E locus of another.

Cultural Inheritance In addition to genetic inheritance and inheritance of altered environments, there is a third avenue of inheritance that influences the evolution of at least some builders. It is cultural inheritance through observational learning. Transmission can be vertical (from one generation to the next) or horizontal (within the same generation). In environments that are changing slowly, genetic evolution is favoured, whereas where the environment is changing rapidly, learning and cultural transmission begin to dominate. For humans, cultural transmission has become the dominant force in changing our behavior. In considering human material cultural evolution, there has been a tendency to emphasise tool use as a particularly important evolutionary driver. There are various reasons for this, but an important one is that stone tools have created a strong bias in the available behavioral evidence. We have a 3 million year old record of stone tool making by our ancestors and it coincides with a rapid increase in brain size and hence presumed cognitive abilities. However, evidence from non-human animals suggests that this is an oversimplification. Tool use and tool making by non-human animals seems to have had little impact on the evolution of their descendants, judging by living examples. In birds for example, tool use has not led to an episode of rapid speciation or extension of habitat range. The woodpecker finch Cactospiza pallida and the New Caledonian crow for example are notably restricted in their range. Building behavior generally does however seem to have had an important evolutionary impact upon for example the ecological dominance of birds and social insects (Hansell and Ruxton, 2008). One possible reason for home building having a greater impact on evolution than tools is that the building of a home is an obligate activity and one that can alter habitats in a substantial way. Tool use on the other hand, is generally concerned with foraging for food and is facultative. It is generally not the only method of food gathering in that species, with some individuals showing it only occasionally or not at all. Incidentally, this suggests that tool use may well have emerged in a number of species and subsequently gone extinct, whereas construction of protective shelters may be more resilient to extinction (Hansell, 2005). So it would be very surprising if it was tool use alone, rather than it together with all kinds of construction behavior that propelled the dramatic evolution of human material culture. Certainly, we are lucky to have a 3 million year record of stone tools but chimpanzees make nests as well as tools, suggesting that similar behavior may have had a long history in human evolution. However, because of the perishable nature of organic building materials, we have no way of knowing. The increase in brain size of our ancestors over the last few million years brought with it enormous scope for us to manipulate our environment. I believe that the making of artefacts, fixed as well as handled, has had an important role in that.

Overview Essays j Animal Architecture

15

Acknowledgments I am very grateful to Graeme Ruxton and Jacqueline Hansell for their careful reading of and valuable comments upon the final draft of this essay.

References Bailey, I.E., Morgan, K.V., Bertin, M., Meddle, S.L., Healy, S.D., 2014. Physical cognition: Birds learn the structural efficiency of nest material. Proceedings of the Royal Society B 281, 20133225. Beck, B.B., 1980. Animal Tool Behavior. Garland, New York. Bentley-Condit, V.K., Smith, E.O., 2010. Animal tool use: Current definitions and an updated comprehensive catalogue. Behaviour 147, 185–221. Bird, C.D., Emery, N.J., 2009. Insightful problem solving and creative tool modification by captive nontool-using rooks. Proceedings of the National Academy of Sciences of the United States of America 106, 10370–10375. Blackledge, T.A., Scharff, N., Coddington, J.A., et al., 2009. Reconstructing web evolution and spider diversification in the molecular era. Proceedings of the National Academy of Sciences of the United States of America 106, 5229–5234. Boesch, C., Head, J., Robbins, M.M., 2009. Complex tool sets for honey extraction among chimpanzees in Loango National Park, Gabon. Journal of Human Evolution 56, 560–569. Bond, J.E., Opell, B.D., 1998. Testing adaptive radiation and key innovation hypotheses in spiders. Evolution 52, 403–414. Breen, A.J., Guillette, L.M., Healy, S.D., 2016. What can nest-building birds teach us? Comparative Cognition and Behavior Reviews 11, 83–102. Büttner, H., 1996. Rubble mounds of sand tilefish Malacanthus plumieri (Bloch 1787) and associated fishes in Colombia. Bulletin of Marine Science 58, 248–260. Casteren, A., van Sellers, W.I., Thorpe, S.K.S., Coward, S., Crompton, R.H.C., 2012. Nest-building orangutans demonstrate engineering know-how to produce safe comfortable beds. Proceedings of the National Academy of Sciences of the United States of America 109, 6873–6877. Choe, J.C., 1997. A new tent roost of Thomas’ fruit-eating bat, Artibeus watsoni (Chiroptera: Phyllostomidae), in Panama. Korean Journal of Biological Sciences 1, 313–316. Collias, N.E., 1997. On the origin and evolution of nest building by passerine birds. The Condor 99, 253–269. Darwin, C., 1871. The Descent of Man, second ed. 1922. John Murray, London. Dawkins, R., 1982. The Extended Phenotype. Freeman, Oxford. Day, L.B., Westcott, D.A., Olster, D.H., 2005. Evolution of bower complexity and cerebellum size in bowerbirds. Brain Behavior and Evolution 66, 62–72. Duffy, J.E., 1996. Eusociality in a coral reef shrimp. Nature 381, 512–514. Frith, C.B., Frith, D.W., 2004. The Bowerbirds. Oxford University Press, Oxford. Goodall, J., 1964. Tool using and aimed throwing in a community of free-living chimpanzees. Nature 201, 1264–1266. Gosline, J.M., Guerette, P.A., Ortlepp, C.S., Savage, K.N., 1999. The mechanical design of spider silks: From fibroin sequence to mechanical function. Journal of Experimental Biology 202, 3295–3303. Guillette, L.M., Scott, A.C.Y., Healy, S.D., 2016. Social learning in nest-building birds: A role for familiarity. Proceedings of the Royal Society B 283. doi:10.1098/rspb. 2015.2685. Hall, Z.J., Street, S.E., Healy, S.D., 2013. The evolution of cerebellum structure correlates with nest complexity. Biology Letters 9. doi:10.1098/rsbl.2013.0687. Hansell, M.H., 2000. Bird Nests and Construction Behaviour. Cambridge University Press, Cambridge. Hansell, M.H., 2005. Animal Architecture. Oxford University Press, Oxford. Hansell, M.H., 2007. Built by Animals. Oxford University Press, Oxford. Hansell, M.H., Ruxton, G.D., 2008. Setting tool use within the context of animal construction behaviour. Trends in Ecology and Evolution 23, 73–78. Hansell, M.H., Ruxton, G.D., 2013. Exploring the dichotomy between animals building using self-secreted materials and using materials collected from the environment. Biological Journal of the Linnean Society 108, 688–701. Hansell, M.H., Ruxton, G.D., Ennos, A.R., 2014. Collected and self-secreted building materials and their contribution to compression and tension structures. Biological Journal of the Linnean Society 112, 625–639. Henderson, G., Hoffman, G.D., Jeanne, R.L., 1990. Predation on Cercropids and material use of the spittle in aphid-tent construction by prairie ants. Psyche 97, 43–53. Heiling, A.M., Herberstein, M.E., 1999. The role of experience in web-building spiders (Araneidae). Animal Cognition 2, 171–177. Iwaniuk, A.N., Lefebvre, L., Wylie, D.R., 2009. The comparative approach and brain-behavior relationships: A tool for understanding tool use. Canadian Journal of Experimental Psychology 63, 150–159. Kelly, L.A., Endler, J.A., 2017. How do great bowerbirds construct perspective illusions. Royal Society Open Science 4 (1), 160661. Klump, B.C., van der Wal, J.E.M., St Claire, J.J.H., Rutz, C., 2015. Context-dependent ‘safekeeping’ of foraging tools in New Caledonian crows. Proceedings of the Royal Society BBiological 282, 20150278. Lefebvre, L., 2011. Taxonomic counts of cognition in the wild. Biology Letters 7, 631–633. Martinsen, G.D., Floate, K.D., Waltz, A.M., Wimp, G.M., Whitham, T.G., 2000. Positive interactions between leafrollers and other arthropods enhance biodiversity on hybrid cottonwoods. Oecologia 123, 82–89. Miller, G., 2000. The Mating Mind. Heinemann, London. Muth, F., Healy, S.D., 2011. The role of adult experience in nest building in the zebra finch Taeniopygia guttata. Animal Behaviour 82, 185–189. Odling-Smee, F.J., Laland, K.N., Feldman, M.W., 2003. Niche Construction: The Neglected Process in Evolution. Princeton University, Princeton. Riechert, S.E., 1985. Why do some spiders co-operate? Agelana consociata, a case study. Florida Entomologist 68, 105–116. Seed, A., Byrne, R., 2010. Animal tool-use. Current Biology 20 (23), R1032–R1039. Shumaker, R.W., Walkup, K.R., Beck, B.B., 2011. Animal Tool Behaviour. The Johns Hopkins University, Baltimore. St Amant, R., Horton, T.E., 2008. Revisiting the definition of tool use. Animal Behaviour 75, 1199–1208. St Claire, J.J.H., Rutz, C., 2013. New Caledonian crows attend to multiple functional properties of complex tools. Philosophical Transactions of the Royal Society B 368 (1630), 20120415. Theraulaz, G., Bonabeau, E., 1995. Modelling the collective building of complex architectures in social insects with lattice swarms. Journal of Theoretical Biology 177, 381–400. Tso, I.-M., Chiang, S.-Y., Blackledge, T.A., 2007. Does the giant wood spider Nephila pilipes respond to prey variation by altering web silk proteins? Ethology 113, 324–333. Underwood, D.L.A., Shapiro, A.M., 1999. Evidence for division of labour in the social caterpillar Eucheria socialis (Lepidoptera: Pieridae). Behavioral Ecology and Sociobiology 46, 228–236. Vollrath, F., Knight, D.P., 2001. Liquid crystalline spinning of spider silk. Nature 410, 541–548. Walsh, P.T., Hansell, M.H., Borello, W.D., Healy, S.D., 2011. Individuality in nest building: Do southern masked weaver (Ploceus velatus) males vary in their nest building behaviour? Behavioural Processes 88, 1–6.

16

Overview Essays j Animal Architecture

Relevant Websites https://en.wikipedia.org/wiki/Bowerbird–Bowerbirds. https://en.wikipedia.org/wiki/Polistinae–Paper wasps, Polistinae. https://en.wikipedia.org/wiki/Spider–Spiders. https://en.wikipedia.org/wiki/Tool_use_by_animals–Tool use by animals. https://en.wikipedia.org/wiki/Ploceidae–Weaverbirds, family Ploceidae.

Brood Parasitism Manuel Soler, University of Granada, Granada, Spain © 2019 Elsevier Ltd. All rights reserved.

Glossary Altricial Avian species in which chicks are raised at the nest with high costs. Coevolution Reciprocal evolution among interacting species. Coevolutionary arms race Coevolutionary model characterized by the continuous escalation in defenses and counter-defenses in both brood parasites and their hosts. Misimprinting costs To imprint on the parasite egg when parasitism occurs during the first breeding attempt. Nest-mate killers Brood parasite species in which chicks eliminate all host offspring. Non-evictors Brood parasite species in which chicks share the nest with host nestlings Phenotypic plasticity Mechanism that allows animals to adjust their behavior in response to changes in environmental conditions. Precocial Avian species in which chicks are able to move and forage for themselves soon after hatching Recognition costs When hosts reject their own oddly colored eggs. Rejection costs When hosts damage their own eggs while trying to eject the parasitic egg. Resistance Host defense based on prevent infection or limit parasite reproduction Tolerance Host defense based on acceptance of parasitism while minimizing the harm caused by the parasite.

Abstract Brood parasitism is a breeding strategy in which the brood parasite female evades all parental care by laying its eggs in host nests of the same (conspecific brood parasitism, CBP) or different species (interspecific brood parasitism, IBP) relying on unrelated foster parents to care for their parasitic offspring. Both CBP and IBP are widespread throughout the animal kingdom providing an extraordinary diversity of brood parasite – host interactions. Costs imposed by brood parasites on hosts selects for the evolution of host defenses, which in turn select for counter-defenses in the brood parasite, triggering improved host defenses, further parasitic adaptations, and so on, giving rise to a coevolutionary arms race. The relationships between brood parasites and their hosts have been studied in most detail in birds, where the subject of brood parasitism have provoked an explosion of compelling discoveries revealing the refinement and complexity of the strategies evolved by brood parasites and their hosts. Current evidence shows that adaptations and counter-adaptations may occur at any stage of the breeding cycle and that each defense and counter-defense will likely influence the evolution of subsequent defenses and counter-defenses, leading to different long-term trajectories and outcomes of coevolutionary interactions between brood parasites and their hosts.

Keywords Brood parasitism; Coevolution; Coevolutionary arms race; Coevolutionary cycles; Conspecific brood parasitism (CBP); Host defenses; Interspecific brood parasitism (IBP); Lack of rejection; Ongoing interactions; Social parasitism; Species interactions; Successful resistance

Introduction Parental care involves any trait of the parents that increases the survival prospects of their offspring often at a cost of their own (Royle et al., 2012). This care involves a long list of traits that range from attendance of eggs only until hatching, to complete care including the provision of food, warmth, protection, and opportunities to learn the skills needed for survival or reproduction (Glazier, 2002). Parental care involves highly costly activities that affect an individual’s fitness, not only because the activities are both time- and energy-consuming, but also because they imply greater predation risk. These costs of parenting could be partially or completely avoided by exploiting any type of offspring care provided by other individuals. This possibility is known as parental-care parasitism (Roldán and Soler, 2011), in which the parasite leaves its offspring in a position appropriate for it to take advantage of hosts’ parental care. Such tactics can be used both within and between species, and are enormously variable depending on the type of interaction. They can include, for example, nest usurpation, theft of stored food,

Encyclopedia of Animal Behavior, 2nd edition, Volume 1

https://doi.org/10.1016/B978-0-12-809633-8.90777-5

17

18

Overview Essays j Brood Parasitism

exploitation of worker labor in eusocial animals, or completely leaving parasitic offspring to be fed and defended by hosts (Roldán and Soler, 2011). In brood parasitism the parasitic females procure care for their offspring by leaving them in the nests of the same species (conspecific brood parasitism, CBP) or a different one (interspecific brood parasitism, IBP) and rely on these unrelated foster parents to care for their parasitic young (Davies, 2000; Roldán and Soler, 2011). By fooling hosts into raising their parasitic young, brood parasites impose fitness costs on hosts given that, at the very least, hosts are forced to divert care and resources away from their own young. However, sometimes parasites evict all host offspring or outcompete them for food, triggering the highest costs for hosts and thereby reducing host reproductive success to zero (Davies, 2000; Kilner and Langmore, 2011). Costs imposed by brood parasites on hosts impels selection for the evolution of host defenses, which in turn drives selection for the brood parasite to evolve counter-defenses, triggering improved host defenses, further parasitic adaptations, and so on. This process, called an “arms race”, usually occurs within a coevolutionary process that implies reciprocal evolutionary change in both the brood parasite and the host species (Kilner and Langmore, 2011; Soler, 2014).

Taxonomic Distribution of Brood Parasitism Brood parasitism is widespread throughout the animal kingdom. Placing the eggs facultatively under the care of conspecific (conspecific brood parasitism, CBP) can be assumed to exist in most animal groups that exhibit parental care. It has been reported mainly in lace bugs and other heteroptera, burying beetles, treehoppers, bees, wasps, and birds (Tallamy, 2005; Yom-Tov and Geffen, 2017). In birds, the group in which this reproductive strategy has been studied in most detail, CBP has been reported in 256 species, being more prominent in precocial than in altricial bird species (Yom-Tov and Geffen, 2017). On the other hand, obligate IBP has been found in spiders, butterflies, bees, wasps, fish, frogs, salamanders, mammals, and especially in ants and birds (Kilner and Langmore, 2011; Roldán and Soler, 2011). Among insects, IBP is especially frequent in social species such as wasps, bees, bumblebees, and ants (especially in eusocial ants with more than 200 species described), and therefore it is frequently referred to as “social parasitism” (Cervo, 2006; Buschinger, 2009; Kilner and Langmore, 2011). In birds, again the best studied group, obligate interspecific brood parasitism is confined to only four out of 40 orders in birds. A total of 109 species of 27 genera are known to be obligate intraspecific brood parasites (Mann, 2017) and, in this case, contrary to what occurs in CBP, IBP is more frequent in altricial than in precocial birds. It has been reported in five families belonging to four orders: Anseriformes (family Anatidae), Cuculiformes (three lineages in the family Cuculidae), Piciformes (family Indicatoridae) and Passeriformes (families Viduidae and Icteridae) (Mann, 2017).

Diversity in Brood Parasite – Host Interactions Two types of brood parasitism exist, one in which parasitic females rear their own offspring but, in addition, place some of their eggs in the care of other females (facultative brood parasitism) and another type in which parasitic females do not provide parental care at all (obligate brood parasitism). Facultative brood parasitic females lay their parasitic eggs in the nests of only conspecific females (CBP), only heterospecific females (IBP) or both, but CBP is the most common form of facultative brood parasitism. Brood parasites display an extraordinary variety of strategies for tricking hosts. The main challenges to brood parasites are for parasitic females to enter host nests or colonies and for parasitic females and/or offspring to gain acceptance and care by host foster parents.

In Insects Sometimes, the parasitic female manages to enter the host nest or colony by avoiding confrontation with the host female, for instance entering the nest when the host female is absent. However, when the hosts are social insects this entry is usually achieved by preventing recognition via chemical camouflage – that is, the parasitic females coat themselves with the odor of a captured host worker or by biosynthesizing the appropriate odor signature themselves (Davies et al., 1989; Kilner and Langmore, 2011). Many other strategies used by parasitic females to enter host nests have been described. For instance, such adaptations as a thickened cuticle, can protect against host stings, a chemical repellent or deterrent of host workers, an appeasement substance that diminishes host aggressiveness, or killing (or expelling) all host individuals present during the parasitic invasion (Cervo, 2006; Kilner and Langmore, 2011).

In Ants Most of the information about brood parasitism in insects comes from ants, the best studied group. Brood parasitic ants exhibit many strategies that can vary greatly among species. For instance, parasitism may be temporary or permanent, the parasitic female can coexist with the host female or kill her, brood parasitic species can have a worker caste or not, or they can steal workers from host colonies (slave-making ants) or not. Parasitic ants are highly specialized as they usually parasitize just one (more rarely up to three

Overview Essays j Brood Parasitism

19

in some slave-making ants) very close phylogenetic relative host species. Four different kinds of brood parasitic ants can be considered (Buschinger, 2009): (1) Guest ants: This is a low-cost form of brood parasitism, also known as “xenobiosis”, in which brood parasitic females use food collected by the host workers and take advantage of the nest built by the host. These “guests” care for their brood themselves, but parasitize food and shelter. (2) Temporary parasitism: In this type, host parental care is parasitized during the foundation of the parasitic colony. After entering the host colony, the parasitic female kills the host queen and manages host workers to forage and care for her and their offspring, which gradually replace host workers. Thus, when all host workers die, the colony becomes a society exclusively of the parasitic species. (3) Permanent parasitism: In this type of brood parasitism the parasitic queen produces only sexual offspring, and thus the worker caste does not exist. This implies that the parasitic queens, throughout their lives, depend upon the host workers. Most frequently these queens do not kill the host queen, but rather the two coexist in the same colony and for this reason this relationship is frequently termed “inquilinism”. In this case, since host workers are subsequently replaced, the parasitic queen can reproduce for a long period. But in some permanent parasitic species, queens kill host queens and therefore can produce sexual offspring only for the short period during which host workers are available. (4) Slave-making ants: This type of brood parasitism, also known as “dulosis”, is a kind of permanent parasitism in which a worker parasitic caste is adapted to steal workers (slaves) from host colonies. The parasitic queen, once in the host colony, kills the host queen and monopolizes the care provided by host workers for her own offspring, which are highly specialized workers equipped with morphological adaptations for effective fighting, such as enormous mandibles or powerful stings. However, they are not able to perform the normal worker activities of non-parasitic species such as foraging or feeding larvae, and they are even unable to eat by themselves. Therefore, both the parasitic queen and workers depend completely on slave workers stolen from host colonies. When the number of host workers diminishes, the slave-maker workers attack host colonies, fight against the defenders, killing most of them, and carry worker pupae to their nest. When pupae emerge they work to the benefit of the parasitic colony rearing more slave-maker workers, who make additional raids several times a year on host colonies to steal new slaves.

In Birds As in insects, laying parasitic eggs by both CBP and IBP requires access to the host nest, which as in insects is achieved mainly by avoiding detection by their hosts. The general rule is that egg laying occurs by stealth; brood parasite females behave secretively during egg laying, usually when hosts are less likely to be at their nests and is extremely rapid. However, sometimes, efficient vigilance or nest guarding by hosts force parasitic females to lay their eggs facing the high risk of being attacked by the incubating host female (Reboreda et al., 2017). Obligate avian brood parasites also exhibit different strategies that can vary among species. The main difference with respect to insect brood parasites is that, while the latter usually are highly specific (i.e., parasitizing one or very few host species), the former include both specific and generalist brood parasitic species, even in the same genus. An extreme example is the genus Molothrus (i.e., cowbirds), which includes five brood parasitic species; two of these species parasitize more than 200 species while one parasitizes just one closely related species (Davies, 2000). Another possibility is for the parasite to be a generalist at the species level but specialist at the female level. This is the case of the most famous brood parasite, the common cuckoo (Cuculus canorus). This species uses about 15 passerine species as regular hosts, and, in each case, the cuckoo females lay eggs that mimic the color of the host species. This host-specific egg mimicry is maintained because there are female-specific races (termed gentes, singular gens) specialized in parasitizing only a single host species whose eggs they mimic. Although females can mate with any male, these races remain distinct because genes determining cuckoo egg type are located on the female-specific W sex chromosome (Gibbs et al., 2000). Regarding brood parasitic nestlings, two different strategies can be differentiated: sharing the nest with host nestlings or eliminating all host offspring. The former (“non-evictors”) would imply to share feedings with host nestlings, while the latter (“nestmate killers”) would allow brood parasitic nestlings to monopolize all feedings provided by the foster parents. Then, being raised as the sole inhabitants of the foster parents’ nest would be highly beneficial for the parasitic nestling. Thus, the second strategy would be expected to be the more frequent, and indeed this is the case. On the one hand, nestlings of many brood parasitic species, just after hatching, eliminate all their host nestmates either by evicting them from the nest or by killing them by using specialized bill structures such as the bill hooks present in honeyguides. On the other hand, nestlings of species that share the nest with host nestlings have different adaptations that make the parasitic nestlings more efficient than host nestlings in getting food, and this usually provokes the starvation of host nestlings. However, this is not always the case, and some host nestlings are able to outcompete parasitic nestlings for food mainly when host nestlings are larger than parasitic nestlings (Soler, 2017a).

Evolution of Brood Parasitism Very little is known about the evolution of brood parasitism in insects. This breeding strategy is widespread throughout many taxa of insects, many brood parasite species are known and, surely, many others remain to be discovered. Considering that in this group

20

Overview Essays j Brood Parasitism

of parasitic species and their hosts are frequently close phylogenetic relatives, it could be stated that brood parasitism has evolved independently many times in insects, nearly as many as brood parasitic species exist. Even when parasitizing very close phylogenetic relatives, the evolution of obligate brood parasitism needs to be accompanied by profound changes in both ecology and life history. These changes are even more remarkable when brood parasites and their hosts are not close relatives, as occurs in the caterpillar of the cuckoo butterfly (Maculinea rebeli) when parasitizing Myrmica ants. Caterpillars synthesize mimetic chemicals that make them recognized as ant larvae by foraging Myrmica workers, which transport them into their nests and feed them (Akino et al., 1999). The origin and evolution of parasitic strategies in insect brood parasites remains a fascinating and challenging matter of research that has often been neglected. The evolution of brood parasitism in birds and its phylogenetic history has received much more attention, although there is no clear consensus on how brood parasitism has evolved. On current evidence, the most likely sequence of steps is: (1) ancestral species with parental care, (2) facultative brood parasitism, (3) conspecific brood parasitism and (4) obligate brood parasitism. Point (3) considering CBP as a stepping stone to obligate interspecific brood parasitism was broadly accepted in the past for mainly two reasons: first, it seems logical that conspecific brood parasitic females, which successfully increase their fitness, will be selected for laying their eggs in the nests of other species, mainly when CBP is too frequent in a population; and second, most facultative interspecific brood parasites are also conspecific brood parasites, and facultative interspecific parasitism is reported in the same families as obligate interspecific brood parasitism (Feeney et al., 2014; Lyon and Eadie, 2017). However, recent studies have shown that CBP is not a necessary precursor for the evolution of IBP given that phylogenetic analyses have shown that, contrary to previous expectations, the occurrence of CBP and IBP do not closely correlate, signifying that parental ancestral care state can lead to the brood parasitism state without the need for CBP acting as a precursor (Krüger and Pauli, 2017). Several factors have been suggested as potentially favoring the evolution of IBP: cooperative breeding, nest predation, competition for nest sites, nest takeover, and parasitism of closely related species. However, convincing support has not been found for any of them. Obligate IBP in birds is known to have evolved independently seven times, once in precocial and six in altricial birds. The only evolutionary origin known in precocial birds occurred recently (10 million years), but in cowbirds occurred rather recently (P), which produces the dilemma. Each suspect must then choose to defect or to cooperate. What should the suspects do? Axelrod and Hamilton (1981) answered this question for an iterated (repeated interactions) prisoner’s dilemma in two ways: by soliciting strategies and then playing the 14 that were submitted in a round-robin computer tournament, and by determining mathematically whether one strategy is an ESS. One strategy, ‘tit for tat’ (TFT), which was submitted by Anatol Rapoport, won the tournament and was shown to be an ESS when the probability of interacting with the same player on the next move of the game was high enough. An animal (or suspect) playing TFT cooperates when first meeting an opponent, and subsequently does whatever the opponent does. The three characteristics that make TFT a winning strategy are as follows: it is nice initially, it retaliates, and it forgives immediately. Since the original use of the prisoner’s dilemma as a model for the evolution of reciprocity (one route to cooperation), many modifications have been developed, including changing the number of players, number of strategies, relatedness of players, and degree of stochasticity (e.g., Dugatkin et al., 1992; Hammerstein and Selten, 1994; McNamara et al., 2004; Doebeli and Hauert, 2005). One possible example of reciprocation by TFT is predator inspection by guppies (Dugatkin, 1991). When guppies and other fish first encounter a potential predator, individuals often approach it, perhaps to gather information about the identity and motivation of the predator. It is very likely that the payoff for inspecting in a group is greater than the payoff if no fish inspects, so R>P. In addition, although having no inspectors is dangerous, it is more dangerous to be the lone inspector, so P>S. Thus, guppies engaging in predator inspection seem to experience a prisoner’s dilemma. A reciprocal strategy such as TFT would ensure that the advantages of inspection exceed those of keeping a safe distance. Guppies are capable of recognizing and remembering the inspection behavior of partners and may employ a conditional approach strategy in which a fish swims toward a predator (inspect) on the first move of a game and subsequently only moves forward if the other fish swims beside it. Inspectors thus appear to be nice (starts to inspect), retaliatory (ceases inspecting if partner stops inspecting), and forgiving (resumes inspecting if partner resumes inspecting).

Table 3

Cooperate Defect

Payoff matrix for the prisoner’s dilemma game. Fitness payoffs accrue to the strategies on the left when each plays the strategies at the top Cooperate

Defect

R¼3 T¼5

S¼0 P¼1

Overview Essays j Game Theory

49

Honest Communication Communication occurs when the behavior or some other cue of one animal affects the behavior of another (Searcy and Nowicki, 2005). Signals and displays are traits that function specifically for communication and have evolved by natural selection for that function. They are thus products of signaler/receiver coevolution, but communication is not necessarily cooperative because the interests of signaler and receiver are not identical. In many cases signalers and receivers have conflicting interests, so that communication has the potential to be manipulative (Searcy and Nowicki, 2005). In such situations, game theory can provide powerful insights into signal evolution and mechanisms that maintain signal honesty or reliability. Many game theory models have been developed to study interactions in which communication occurs, such as between parents and offspring (Godfray and Johnstone, 2000), contestants for important resources (Hammerstein and Riechert, 1988), prey and predators (Brown et al., 1999), and males and females (Johnstone et al., 1996). One question they have in common is, How is honest communication possible when signalers can benefit from deceiving receivers? Maynard Smith (1974) considered this question when he wondered why accurate information is ever transferred during contests. He viewed the coevolution of sender and receiver as analogous to an arms race, with senders attempting to manipulate receivers and receivers attempting the resist the manipulation of senders. But because empirical studies showed clearly that animals communicate information honestly, he concluded that lying may be impossible or may be punished in some cases. Subsequent studies showed that displays that are physically constrained (e.g., large frogs produce deep croaks, but small ones cannot) are honest. After considerable controversy, game theory models showed that signals that confer a handicap on the signaler are also honest (the handicap principle; Grafen, 1990). Signals can be costly to signalers because they are energetically expensive to produce, or they are disadvantageous in some other way such as interfering with locomotion or the immune system. They can also be costly because signals attract unwanted eavesdroppers such as predators or receivers that attack (punish) dishonest signalers (Searcy and Nowicki, 2005). Honest signaling models indicate that reliable information transfer is common even if deceit occurs occasionally and that the form of the communication is influenced by selection for signal efficacy and for signal reliability because unreliable signals are not evolutionarily stable (Johnstone, 1998).

Game Theory and Animal Behavior Many researchers consider the application of the concept of inclusive fitness (Hamilton, 1964) and of game theory to animal behavior to be two of the most important theoretical developments since the modern synthesis of evolution and genetics. In many respects, game theory has changed the thinking of those who study aggression, sequential assessment, honest communication, cooperation, and personality in animals. The fundamental principles of game theory that the behavior of one animal affects the fitness of others and that these effects must be understood when explaining the evolution of behavior, and the concept of an ESS, have become fundamental principles of animal behavior in particular and of biology more generally. The concept of ESS has also invaded psychology, political science, and even mathematics itself. Despite the ability of game theory models to address many aspects of the social behavior of animals, there have been relatively few empirical tests of game theory models, especially in comparison with other classes of models such as optimal foraging (e.g., Sih et al., 2015, but compare Beekman and Jordan, 2017). In addition to the usual objections to theoretical approaches to biology, game theory models seem to face some particular difficulties, including that they are unnecessary or that they are irrelevant because they ignore the underlying genetic structure and constraints. On the other hand, game theory models have yielded a rigorous evolutionary understanding of social behavior that is otherwise difficult to explain, such as settling contests conventionally, communicating honestly, cooperation in the face of the temptation to cheat, and the maintenance of behavioral polymorphisms. Recent models even account for gang foraging (Dall and Wright, 2009), animal personalities (Sih et al., 2015), the evolution of learning (Fawcett et al., 2013) and how strategies and payoff matrices originate and evolve (van der Post et al., 2015). It is likely that empirical testing will become more common as game theory models make more realistic assumptions and more explicit predictions, which will make these models more accessible to empiricists. If you are interested in learning how to develop game theory models, try Gamebug (see “Relevant Websites section”), a teaching and learning resource.

References Axelrod, R., 1984. The Evolution of Cooperation. Basic Books, New York, NY. Axelrod, R., Hamilton, W.D., 1981. The evolution of cooperation. Science 211, 1390–1396. Barnes, D.J., Chu, D., 2010. Introduction of Modeling for Biosciences. Springer, London, UK. Beekman, M., Jordan, L.A., 2017. Does the field of animal personality provide any new insights for behavioral ecology? Behavioral Ecology 28, 617–623. Borgia, G., 1986. Sexual selection in bowerbirds. Scientific American 254, 70–79. Brown, J.S., Laundré, J.W., Gurung, M., 1999. The ecology of fear: Optimal foraging, game theory, and trophic interactors. Journal of Mammalogy 80, 385–399. Dall, S.R.X., Wright, J., 2009. Rich pickings near large communal roosts favour ‘gang’ foraging by juvenile common ravens, Corvus corax. PLOS ONE 4, e4530. Available at: https:// doi.org/10.1371/journal.pone.0004530. Doebeli, M., Hauert, C., 2005. Models of cooperation based on the prisoner’s dilemma and the snowdrift game. Ecology Letters 8, 748–766.

50

Overview Essays j Game Theory

Dugatkin, L.A., 1991. Dynamics of the TIT for TAT strategy during predator inspection in the guppy (Poecilia retinulata). Behavioral Ecology and Sociobiology 29, 127–132. Dugatkin, L.A., Mesterton-Gibbons, M., Houston, A.I., 1992. Beyond the prisoner’s dilemma: Toward models to discriminate among mechanisms of cooperation in nature. Trends in Ecology and Evolution 6, 202–205. Dugatkin, L.A., Reeve, H.K. (Eds.), 1998. Game Theory and Animal Behavior. Oxford University Press, New York, NY. Fawcett, T.W., Hamblin, S., Giraldeau, L.-A., 2013. Exposing the behavioral gambit: The evolution of learning and decision rules. Behavioral Ecology 23, 2–11. Godfray, H.C.J., Johnstone, R.A., 2000. Begging and bleating: The evolution of parent-offspring signalling. Philosophical Transactions of the Royal Society of London B 355, 1581–1591. Grafen, A., 1979. The hawk-dove game played between relatives. Animal Behaviour 27, 905–907. Grafen, A., 1990. Biological signals as handicaps. Journal of Theoretical Biology 144, 517–546. Hamilton, W.D., 1964. The genetical evolution of social behaviour. I and II. Journal of Theoretical Biology 7, 1–52. Hammerstein, P., Riechert, S.E., 1988. Payoffs and strategies in territorial contests: ESS analyses of two ecotypes in the spider, Agelenopsis aperta. Evolutionary Ecology 2, 115–138. Hammerstein, P., Selten, R., 1994. Game theory and evolutionary biology. In: Aumann, R.J., Hart, S. (Eds.), Handbook of Game Theory, vol. 2. Elsevier Science, Amsterdam, NE. Houston, A., McNamara, J., 1988. Fighting for food: A dynamic version of the hawk-dove game. Evolutionary Ecology 2, 51–64. Johnstone, R.A., 1998. Game theory and communication. In: Dugatkin, L.A., Reeve, H.K. (Eds.), Game Theory and Animal Behavior. Oxford University Press, New York, NY. Johnstone, R.A., Reynolds, J.D., Deutsch, J.C., 1996. Mutual mate choice and sex differences in choosiness. Evolution 50, 1382–1391. Lin, C.C., Segel, L.A., 1988. Mathematics Applied to Deterministic Problems in the Natural Sciences. Society for Industrial and Applied Mathematics, Philadelphia, PA. Maynard Smith, J., 1974. The theory of games and the evolution of animal conflicts. Journal of Theoretical Biology 47, 209–221. Maynard Smith, J., 1976. Evolution and the theory of games. American Scientist 64, 41–45. Maynard Smith, J., 1982. Evolution and the Theory of Games. Cambridge University Press, Cambridge, UK. Maynard Smith, J., Parker, G.A., 1976. The logic of asymmetric contests. Animal Behaviour 24, 159–175. Maynard Smith, J., Price, G.R., 1973. The logic of animal conflict. Nature 246, 15–18. McNamara, J.M., Barta, Z., Houston, A.I., 2004. Variation in behaviour promotes cooperation in the prisoner’s dilemma game. Nature 428, 745–748. Nash, J., 1951. Non-cooperative games. Annals of Mathematics 54, 286–295. Pruett-Jones, S., Pruett-Jones, M., 1994. Sexual competition and courtship disruptions: Why do male bowerbirds destroy each other’s bowers? Animal Behaviour 47, 607–620. Searcy, W.A., Nowicki, S., 2005. The Evolution of Animal Communication: Reliability and Deception in Signaling Systems. Princeton University Press, Princeton, NJ. Sih, A., Mathot, K.J., Moirón, M., et al., 2015. Animal personality and state-behaviour feedbacks: A review and guide for empiricists. Trends in Ecology and Evolution 30, 50–60. van der Post, D.J., Verbrugge, R., Hemelrijk, C.K., 2015. The evolution of different forms of sociality: Behavioral mechanisms and eco-evolutionary feedback. PLOS ONE 10, e0117027. https://doi.org/10.1371/journal.pone.0117027. von Neumann, J., Morgenstern, O., 1944. Theory of Games and Economic Behavior. Princeton University Press, Princeton, NJ.

Relevant Websites http://www.crafoordprize.se/press/arkivpressreleases/thecrafoordprize1999.5.32d4db7210df50fec2d800018201.html. –Craaford Prize. http://ess.nbb.cornell.edu/gbgmanual.html. –GameBug.

Hormones and Behavior: Basic Conceptsq Kathryn Lee Gruchalla Russart and Randy J Nelson, The Ohio State University Wexner Medical Center, Columbus, OH, United States © 2019 Elsevier Ltd. All rights reserved.

Abstract Behavioral endocrinology is the scientific study of the interaction between hormones and behavior. The goal of this article is to provide a précis to the field by describing the principles of behavioral endocrinology, the basics of behavioral and hormonal analyses, as well as the sorts of approaches used in this subdiscipline of animal behavior.

Keywords Behavioral endocrinology techniques; Endocrine glands; Hormone evolution; Hormone receptors; Hormones; Hypothalamus; Nervous system; Neuroendocrinology; Sexual behavior; Steroids

Introduction Behavioral endocrinology is the scientific study of the interaction between hormones and behavior. This interaction is bidirectional: hormones can affect behavior, and behavior can feedback to influence hormone concentrations. Hormones are chemical messengers released from endocrine glands that influence the nervous system to regulate the physiology and behavior of individuals. Over evolutionary time, hormones regulating physiological processes have been co-opted to influence behaviors linked to these processes. For example, hormones associated with gamete maturation such as estrogens are now broadly associated with the regulation of female sexual behaviors. Such dual hormonal actions ensure that mating behavior occurs when animals have mature gametes available for fertilization. Generally speaking, hormones change gene expression or cellular function, and affect behavior by increasing the likelihood that specific behaviors occur in the presence of precise stimuli. Hormones achieve this by affecting individuals’ sensory systems, central integrators, and/or peripheral effectors. To gain a full understanding of hormone–behavior interactions, it is important to monitor hormone values, as well as receptor interactions in the brain. Because certain chemicals in the environment can mimic natural hormones, these chemicals can have pro-found effects on the behavior of humans and other animals.

Behavioral Endocrinology Techniques A number of methods are used to gather the evidence needed to establish hormone–behavior relationships. Much of the recent progress in behavioral endocrinology has resulted from technical advances in the tools that allow us to detect, measure, and probe the functions of hormones and their receptors. These techniques, with a brief description, are listed in Table 1. Several of these techniques are the result of advanced research in behavioral endocrinology, including the time-honored ablation-replacement techniques, bioassays, as well as modern assays that utilize the concept of competitive binding of antibodies that include radioimmunoassay (RIA), enzyme-linked immunosorbent assay (ELISA; enzyme-linked immunoassay (EIA)), autoradiography, and immunocytochemistry. Other techniques commonly used in behavioral endocrinology include neural stimulation and single-unit recording, techniques that activate or block hormone receptors with drugs, and gene arrays and genetic manipulations including interfering with RNA and use of viral gene vectors to deliver novel genes directly into the brain. Because hormones must inter-act with specific receptors to evoke a response, many of these techniques are used to influence or measure hormone secretion, hormone binding, or the physiological and behavioral effects that ensue after hormones bind to their respective receptors.

Hormones Hormones are organic chemical messengers produced and released by specialized glands called ‘endocrine glands.’ Endocrine is derived from the Greek root words endon, meaning ‘within,’ and krinein, meaning ‘to release,’ whereas the term hormone is based on the Greek word hormon, meaning ‘to excite.’ Hormones are released from these glands into the bloodstream (or the tissue fluid system in invertebrates), where they act on target organs (or tissues) generally at some distance from their origin. Hormones

q

Change History: October 2017. Randy J Nelson made changes to the text and references. This is an update of R.J. Nelson, Hormones and Behavior: Basic Concepts, Editor(s): Michael D. Breed, Janice Moore, Encyclopedia of Animal Behavior, Academic Press, 2010, Pages 97-105.

Encyclopedia of Animal Behavior, 2nd edition, Volume 1

https://doi.org/10.1016/B978-0-12-809633-8.20697-3

51

52 Table 1

Overview Essays j Hormones and Behavior: Basic Concepts Common techniques in behavioral endocrinology

Ablation (removal or extirpation) of the suspected source of a hormone to determine its function is a classic technique in endocrinology. Suspected brain regions that may regulate the behavior in question can also be ablated. Typically, four steps are required: (1) a gland that is suspected to be the source of a hormone affecting a behavior is surgically removed; (2) the effects of removal are observed; (3) the hormone is replaced, by reimplanting the removed gland, by injecting a homogenate or extract from the gland, or by injecting a purified hormone; and (4) a determination is made whether the observed consequences of ablation have been reversed by the replacement therapy. Radioimmunoassay (RIA) is based on the principle of competitive binding of an antibody to its antigen. An antibody produced in response to any antigen, in this case a hormone, has a binding site that is specific for that antigen. Antigen molecules can be ‘labeled’ with radioactivity, and an antibody cannot discriminate between an antigen that has been radiolabeled (or ‘hot’) and a normal, nonradioactive (‘cold’) antigen. A standard curve is produced with several tubes, each containing the same measured amount of antibody, the same measured amount of radiolabeled hormone, and different amounts of cold purified hormone of known concentrations. The radiolabeled hormone and unlabeled hormone compete for binding sites on the antibody, so the more cold hormone that is present in the tube, the less hot hormone will bind to the antibody. The quantity of bound hormone can be determined by precipitating the antibody and measuring the associated radioactivity resulting from the radiolabeled hormone that remains bound. The unknown concentration of hormone in a sample can then be determined by subjecting it to the same procedure and comparing the results with the standard curve. Enzymoimmunoassay (EIA), as RIA, works on the principle of competitive binding of an antibody to its antigen. EIAs do not require radioactive tags; instead, the antibody is tagged with a compound that changes optical density (color) in response to binding with antigen. Other than home pregnancy tests, most EIAs are developed to provide quantitative information. A standard curve is generated so that different known amounts of the hormone in question provide a gradient of color that can be read on a spectrometer. The unknown sample is then added, and the amount of hormone is interpolated by the standard curves. A similar technique is called ‘enzyme-linked immunosorbent assay’ (ELISA). Immunocytochemistry (ICC) techniques use antibodies to determine the location of a hormone in cells. Antibodies linked to marker molecules, such as those of a fluorescent dye, are usually introduced into dissected tissue from an animal, where they bind with the hormone or neurotransmitter of interest. For example, if a thin slice of brain tissue is immersed in a solution of antibodies to a protein hormone linked to a fluorescent dye, and the tissue is then examined under a fluorescent microscope, concentrated spots of fluorescence will appear, indicating where the hormone is located. Autoradiography is typically used to determine hormonal uptake and indicate receptor locations. Radiolabeled hormone is injected into an individual or into dissected tissue. Suspected target tissues are sliced into several very thin sections; adjacent sections are then subjected to different treatments. One section of the suspected target tissue is stained in the usual way to highlight various cellular structures. The next section is placed in contact with photographic film or emulsion for some period of time, and the emission of radiation from the radiolabeled hormone develops an image on the film. The areas of high radioactivity on the film can then be compared with the stained section to determine how the areas of highest hormone concentration correlate with cellular structures. This technique has been very useful in determining the sites of hormone action in nervous tissue, and consequently has increased our understanding of hormone–behavior interactions. Blot tests use electrophoresis to determine in which cells specific DNA, RNA, or proteins are located. Homogenized tissue of interest is placed on a nitrocellulose filter, which is subjected to electrophoresis that involves application of an electric current through a matrix or gel that results in a gradient of molecules separating out along the current on the basis of size (smaller molecules move farther than larger molecules during a set time period). The filter is then incubated with a labeled substance that can act as a tracer for the protein or nucleic acid of interest: radiolabeled complementary deoxyribonucleic acid (cDNA) for a nucleic acid assay, or an antibody that has been radiolabeled or linked to an enzyme for a protein assay. If radiolabeling is used, the filter is then put over film to locate and measure radioactivity. In enzyme-linked protein assays, the filter is incubated with chromogenic chemicals, and standard curves reflecting different spectral densities are generated. Southern blots assay DNA; Northern blots assay RNA, whereas Western blots test for proteins. In situ hybridization is used to identify cells or tissues in which mRNA molecules for a specific protein (e.g., a peptide hormone) are produced. The tissue is fixed, mounted on slides, and either dipped into emulsion or placed over film and developed with photographic chemicals. Typically, the tissue is also counterstained to identify specific cellular structures. A radiolabeled cDNA probe is introduced into the tissue. If the mRNA of interest is present in the tissue, the cDNA will form a tight association (i.e., hybridize) with it. The tightly bound cDNA, and hence the messenger RNA (mRNA), will appear as dark spots. This technique can be used to determine whether a particular substance is produced in a specific tissue. Pharmacological techniques are used to identify hormones and neurotransmitters involved in specific behaviors. Some specific chemical agents can act to stimulate or inhibit endocrine function by affecting hormonal release; these agents are called ‘general agonists’ and ‘antagonists,’ respectively. Other drugs act directly on receptors, either enhancing or negating the effects of the hormone under study; these drugs are referred to as ‘receptor agonists’ and ‘antagonists,’ respectively. Anterograde and retrograde tract tracing is used to follow the complex neural pathways involved in controlling behaviors beyond a single locus. Visible tracer molecules or molecular tracers are injected into a specific brain region of interest where they are absorbed by the neuron. In anterograde tract tracing, the dyes are absorbed by the cell body and transported down the axon to their terminals. Retrograde tract tracing is similar, except in this case the injected molecules are absorbed by axonal terminals and transported to the cell body. The axonal projections and cell bodies can then be visualized using immunohistochemistry. Some anterograde tracers can cross the synaptic cleft, allowing for the tracing of neural networks. Brain imaging techniques reveal brain activation during behaviors. Paired with endocrine manipulations or monitoring, imaging can provide important information about hormone–behavior interactions. Positron emission tomography (PET) scanning permits detailed measurements of real-time functioning of specific brain regions of people who are conscious and alert. PET gives a dynamic representation of the brain at work. Computer-assisted tomography (CT) scanner shoots fine beams of X-rays into the brain from several directions. The emitted information is fed into a computer that constructs a composite picture of the anatomical details within a ‘slice’ through the brain of the person. Magnetic resonance imaging (MRI) does much the same thing, but uses nonionizing radiation formed by the excitation of protons by radiofrequency energy in the presence of large magnetic fields. Functional MRI (fMRI) uses a very high spatial (1 mm) and temporal resolution to detect changes in brain activity during specific tasks or conditions. When neurons become more active, they use more energy, and require additional blood flow to deliver glucose and oxygen. The fMRI scanner detects this change in cerebral blood flow by detecting changes in the ratio of oxyhemoglobin and deoxyhemoglobin.

Overview Essays j Hormones and Behavior: Basic Concepts Table 1

53

Common techniques in behavioral endocrinologydcont'd

Gene manipulations. In behavioral endocrinology research, common genetic manipulations include the insertion (transgenic or knockin) or removal (knockout) of the genetic instructions encoding a hormone or the receptor for a hormone. In knockout mice, behavioral performance can then be compared among wild-type (þ/þ), heterozygous (þ/), and homozygous (/) mice, in which the gene product is produced normally, produced at reduced levels, or completely missing, respectively. Inducible knockouts when specific genes are inactivated in adulthood promise to become important tools in behavioral endocrinology. An alternative approach involved gene silencing via RNA interference (RNAi), which is used to deplete protein products made in cells. Gene arrays can be used to determine relative gene expression during the onset of a behavior, or a change in developmental state, or among individuals that vary in the frequency of a given behavior or hormonal state. Essentially, a miniscule spot of nucleic acid of known sequence is attached to a glass slide (or occasionally nylon matrix) in a precise location often by high speed robotics. This identified, attached nucleic acid is called ‘the probe,’ whereas the sample nucleic acid is the target. The identification of the target is revealed by hybridization, the process by which the nucleotides link to their base pair. RNA sequencing technologies can be used similarly to gene arrays, except they can determine absolute gene expression instead of relative gene expression. RNA is isolated from a specific tissue and sequenced using high-throughput technology. Gene expression is then calculated based on the depth of sequencing of each known gene. Expression of each gene is then compared between treatment groups to determine all of the differentially expressed genes. -Omics informally refers to the field of study that aims to characterize and quantify entire genomes, proteomes, epigenomes, transcriptomes, metabolomes, etc. Omics studies use bioinformatics and computational methods to more fully understand signaling pathways. Omics can also be used to compare tissues from different treatments. Electrical recording is used to record neuronal activity before and after exposure to hormones. A small electrode is placed near an individual neuron or a group of neurons. Electrical brain stimulation is used to stimulate specific neurons in culture or in a live animal. A microelectrode is positioned precisely in the brain, and the neuron or neural network is stimulated by a weak electric current. Correlations between neural activity and behavior can be made using this technique. Optogenetics is an improved technique for electrical brain stimulation that allows precise control over the neurons being stimulated. Light-sensitive ion channels are genetically inserted and expressed in specific neurons, and the activity of these neurons can then be controlled with light and monitored in real time. This technique allows a more precise understanding of the neuronal regulation of behavior. DREADDS (Designer Receptors Exclusively Activated by Designer Drugs) is used to transiently activate specific brain regions using synthetically derived receptors and ligands. Genetic techniques are used to insert the receptors into specific neurons, which can then be activated by applying exogenous ligands. DREADDS allow for repeatable activation or inhibition of specific neurons, and can be used to study the hormonal activity or behaviors resulting from specific neuronal manipulations.

coordinate the physiology and behavior of an animal by regulating, integrating, and controlling its bodily function. Hormones are similar in function to other chemical mediators including neurotransmitters and cytokines. Indeed, the division of chemical mediators into categories mainly reflects the need by researchers to organize biological systems into endocrine, nervous, and immune systems, rather than real functional differences among these chemical signals. Hormones often function locally as neurotransmitters and also interact with neurotransmitters and cytokines to influence behavior. Hormones can be grouped into four classes: (1) peptides or proteins, (2) steroids, (3) monoamines, and (4) lipid-based hormones. Generally, only one class of hormone is produced by a single endocrine gland, but there are some notable exceptions. It is important and useful to discriminate among the four types of hormones because they differ in several important characteristics, including their mode of release, how they move through the blood, the location of their target tissue receptors, and the manner by which the interaction of the hormone with its receptor results in a biological response. The major vertebrate hormones and their primary biological actions are listed in Table 2. Although exceptions always exist, the endocrine system has several general features: (1) endocrine glands are ductless, (2) endocrine glands have a rich blood supply, (3) hormones, the products of endocrine glands, are secreted into the bloodstream, (4) hormones can travel in the blood to virtually every cell in the body and can thus potentially interact with any cell that has appropriate receptors, and (5) hormone receptors are rather specific binding sites, embedded in the cell membrane or located elsewhere in the cell that interact with a particular hormone or class of hormones. As mentioned, the products of endocrine glands are secreted directly into the blood, whereas other glands, called ‘exocrine glands,’ have ducts into which their products are secreted (e.g., salivary, sweat, and mammary glands). Some glands have both endocrine and exocrine structures (e.g., the pancreas). Recently, the definition of an endocrine gland also had to be reconsidered. For example, adipose tissue produces the hormone, leptin, and the stomach produces a hormone called ‘ghrelin.’ Probably the most active endocrine organ, and the one that produces the most diverse types of hormones, is the brain. As single cells evolved into multicellular organisms, chemical communication within and between cells, as well as between individuals and populations, developed. The endocrine system evolved to become a key component of this complicated intra- and intercellular communication system, although other systems of chemical mediation exist. For example, chemical mediation of intracellular events is called ‘intracrine mediation.’ Some intracrine mediators may have changed their function over the course of evolution and now serve as hormones or pheromones. Autocrine cells secrete products that may feed back to affect processes in the cells that originally produced them. For example, many steroid–hormone-producing cells possess receptors for their own secreted products. Chemical mediators released by one cell that induce a biological response in nearby cells are called ‘paracrine agents’; nerve cells are well-known paracrine cells. In several cases, a single hormone (especially pep-tides) can have autocrine, paracrine, or endocrine functions. For example, leptin stimulates expression of itself and its receptor. Generally, leptin is produced in adipose tissue and it functions as

54

Overview Essays j Hormones and Behavior: Basic Concepts

Table 2

Vertebrate hormones

Glands/hormone Adrenal glands Mineralocorticoids Aldosterone 11-Deoxycorticosterone Glucocorticoids

Abbreviation

Source

Major biological action

DOC

Zona glomerulosa of adrenal cortex Zona glomerulosa of adrenal cortex

Sodium retention in kidney Sodium retention in kidney Increases carbohydrate metabolism; antistress hormone Increased carbohydrate metabolism; antistress hormone Weak androgen; primary secretory product of fetal adrenal cortex

Cortisol (hydrocortisone)

F

Corticosterone

B

Zona fasciculata and z. reticularis of adrenal cortex Zona fasiculata and z. reticularis of adrenal cortex

Dehydroepiandro-sterone Ovaries Estradiol Estriol Estrone

DHEA

Zona reticularis of adrenal cortex

Progesterone Testes Androstenedione Dihydrotestosterone

P

Corpora lutea, placenta

DHT

Leydig cells Seminiferous tubules and prostate

T

Leydig cells

Male sex characters Male secondary sex characters Spermatogenesis; male secondary sex characters

Abbreviation

Source

Major biological action

PAI-1

Adipocytes Adipocytes Adipocytes

Regulation of energy balance Modulates endothelial adhesion molecules Regulation of vascular hemostasis

Adrenal medulla Adrenal medulla

Analgesic actions in CNS Analgesic actions in CNS

Testosterone Peptide and protein hormones Hormone Adipose tissue Leptin (Ob protein) Adiponectin Plasminogen activator inhibitor-1 Adrenal glands Met-enkephalin Leu-enkephalin Gut

Follicles Follicles Follicles

Bombesin Cholecystokinin (pancreozymin) Gastric inhibitory polypeptide Gastrin Gastrin-releasing peptide Ghrelin Glucogon-like peptide-1 Motilin

Neurons and endo-crine cells of gut CCK GIP

Duodenum and CNS Duodenum G-cells of midpyloric glands in stomach antrum GI tract Stomach mucosa/GI tract L cells of intestine Duodenum, pineal gland

Uterine and other female tissue development Uterine and mammary tissue development Uterine and mammary tissue development Uterine development; mammary gland development; maintenance of pregnancy

Hypothermic hormone; increases gastrin secretion Stimulates gallbladder contraction and bile flow; affects memory, eating behavior Inhibits gastric acid secretion

Vasoactive intestinal polypeptide Peptide YY Heart Atrial naturetic factor Hypothalamus Agouti-related protein Arg-vasotocin

VIP PPY

GI tract, hypothalamus GI tract

Increases secretion of gastric acid and pepsin Stimulates gastrin secretion Regulation of energy balance Regulates insulin secretion Alters motility of GI tract Stimulates pancreatic acinar cells to release bicarbonate and water Increases secretion of water and electrolytes from pancreas and gut; acts as neurotransmitter in autonomic nervous system Regulation of energy balance/food intake

ANF

Atrial myocytes

Regulation of urinary sodium excretion

AGRP AVT

Corticotropin-releasing hormone

CRH

Gonadotropin-releasing hormone

GnRH

Arcuate nucleus Hypothalamus and pineal gland Paraventricular nuclei, anterior periventricular nuclei Preoptic area; anterior hypothalamus; suprachiasmatic

Regulation of energy balance Regulates reproductive organs Stimulates release of ACTH and b-endorphin from anterior pituitary Stimulates release of FSH and LH from anterior pituitary

GRP GLP-1

Secretin

Duodenum

Overview Essays j Hormones and Behavior: Basic Concepts Table 2

55

Vertebrate hormonesdcont'd

Glands/hormone

Abbreviation

Source

Major biological action

Gonadotropin-inhibiting hormone

GnIH

Inhibits release of LH (in birds)

Kisspeptin

KISS

Luteinizing hormone-releasing hormone

LHRH

Species-dependent loci Arcuate and anteroventral periventricular nuclei Nuclei; medial basal hypothalamus (rodents and primates); arcuate nuclei (primates)

GHRH

Anterior periventricular nuclei Medial basal hypothala mus; arcuate nuclei

MIF (DA)

Arcuate nuclei

MRF NPY

Paraventricular nuclei Arcuate nuclei Hypothalamus; intestinal mucosa Lateral hypothalamic area Arcuate nuclei

Somatostatin (growth hormone-inhibiting hormone) Somatocrinin (growth hormone-releasing hormone) Melanotropin-release inhibitory factor (Dopamine) Melanotropin-releasing factor Neuropeptide Y Neurotensin Orexin A and B Prolactin-inhibitory factor (Dopamine)

PIF (DA)

Prolactin-releasing hormone

Paraventricular nuclei

Substance P

SP

Hypothalamus, CNS, intestine

Thyrotropin-releasing hormone Urocortin Liver Somatomedins

TRH

Paraventricular nuclei Lateral hypothalamus

Inhibits release of GH and TSH from anterior pituitary inhibits release of insulin and glucagon from pancreas Stimulates release of GH from anterior pituitary Inhibits the release of MSH (no evidence of this peptide in humans) Stimulates the release of MSH from anterior pituitary (no evidence of this peptide in humans) Regulation of energy balance May act as a neurohormone Regulation of energy balance/food intake Inhibits PRL secretion Stimulates release of PRL from anterior pituitary Transmits pain; increases smooth muscle contractions of GI tract Stimulates release of TSH and PRL from anterior pituitary CRH-related peptide

Angiotensinogen Ovaries

Liver, blood

Cartilage sulfation, somatic cell growth Precursor of angiotensins, which affect blood pressure

Relaxin Inhibin (folliculostatin) Gonadotropin surge-attenuating factor Activin Pancreas Glucagon

Corpora lutea Follicles

Permits relaxation of various ligaments during parturition Inhibits FSH secretion

Follicles Sertoli cells

Control of LH secretion during menstruation Stimulates FSH secretion

a-cells

Glycogenolysis in liver Glucose uptake from blood; glycogen storage in liver Inhibits insulin and glucagon secretion Effects on gut in pharmacological doses

Insulin Somatostatin Pancreatic polypeptide Pituitary

Liver, kidney

Critical for normal puberty

GnSAF

PP

b-cells d-cells Peripheral cells of pancreatic islets

Adrenocorticotropic hormone Vasopressin (antidiuretic hormone) b-endorphin

ACTH ADH or AVP

Anterior pituitary Posterior pituitary Intermediate lobe of pituitary

Follicle-stimulating hormone Growth hormone Lipotropin

FSH GH LPH

Anterior pituitary Anterior pituitary Anterior pituitary

Luteinizing hormone

LH

Anterior pituitary

Melanocyte-stimulating hormone

MSH

Anterior pituitary

Stimulates synthesis and release of glucocorticoids Increases water reabsorption in kidney Analgesic actions Stimulates development of ovarian follicles and secretion of estrogens; stimulates spermatogenesis Mediates somatic cell growth Fat mobilization; precursor of opioids Stimulates Leydig cell development and testosterone production in males; stimulates corpora lutea development and production of progesterone in females Affects memory; affects skin color in amphibians (Continued)

56

Overview Essays j Hormones and Behavior: Basic Concepts

Table 2

Vertebrate hormonesdcont'd

Glands/hormone

Abbreviation

Source

Major biological action

PRL

Anterior pituitary

Stimulates milk letdown and uterine contractions during birth Many actions relating to reproduction, water balance, etc.

TSH

Anterior pituitary

Stimulates thyroid hormone secretion

CG

Placenta

LH-like functions; maintains progesterone production during pregnancy

CS (PL)

Placenta

Acts like PRL and GH

MIH

Fetal Sertoli cells of testes Seminiferous tubules (and ovaries) Sertoli cells

Mediates regression of Müllerian duct system Inhibits FSH secretion Stimulates FSH secretion

CT

C-cells of thyroid

Parathyroid hormone Thyroxine (tetraiodothyronine) Triiodothyronine Parathyroid-related peptide Thymus Thymosin Thymostatin Monoamine hormones Adrenal glands Hormone

PTH T4 T3 PTHrP

Parathyroid gland Follicular cells Follicular cells Parathyroid gland (and other tissues)

Lowers serum Ca2þ levels Stimulates bone resorption; increases serum Ca2þ levels Increases oxidation rates in tissue Increases oxidation rates in tissue Regulation of bone/skin development

Thymocytes Thymocytes

Proliferation/differentiation of lymphocytes Proliferation/differentiation of lymphocytes

Abbreviation

Source

Epinephrine (adrenaline) Norepinephrine (noradrenaline) Central nervous system Dopamine

EP NE

Adrenal medulla (and CNS) Adrenal medulla (and CNS)

Major biological action Glycogenolysis in liver; increases blood pressure Increases blood pressure

DA

Arcuate nuclei of hypothalamus

Serotonin Pineal gland Melatonin Lipid-based hormones (eicosanoids) Hormone Leukotrienes

5-HT

CNS (also pineal)

Pineal gland

Affects reproductive functions

Abbreviation LT PGE1 and PGE2 PGF1a and PGF2a PGA2 PGI2 TX2

Source Lung

Major biological action Long-acting bronchoconstrictors

Variety of cells

Stimulates cAMP Active in dissolution of corpus luteum and in ovulation Hypotensive effects Increased second messenger formation Increased second messenger formation

Oxytocin Prolactin Thyroid-stimulating hormone (thyrotropin) Placenta Chorionic gonadotropin Chorionic somatomammotropin (placental lactogen) Testes Müllerian inhibitory hormone Inhibin (folliculostatin) Activin Thyroid/parathyroid Calcitonin

Prostaglandins E1 and E2 Prostaglandins F1a and F2a Prostaglandin A2 Prostacyclin I Thromboxane A2

Posterior pituitary

Variety of cells Kidney Variety of cells Variety of cells

Inhibits prolactin release (and other actions) Stimulates release of GH, TSH, ACTH; inhibits release of LH

Source: Reproduced from Nelson, R.J., Kriegsfeld, 2016. An Introduction to Behavioral Endocrinology. Sunderland, MA: Sinauer Associates.

a hormone when released into the blood by regulating energy balance at the level of the hypothalamus. However, leptin is also produced in the anterior pituitary gland where it diffuses locally to influence thyroid-stimulating hormone (TSH) secretion (paracrine) (Chen et al., 2006). Many chemical mediators display similar diversity in function. Some hormones are water-soluble proteins or small peptides that are stored in the endocrine cell in secretory granules, or vesicles. In response to a specific stimulus for secretion, the secretory vesicle fuses its membrane with the cellular membrane, an opening develops, and the hormones diffuse into the extracellular space via a process called ‘exocytosis.’ The expelled hormones then enter the blood system from the extracellular space. Other hormones, such as steroid hormones, are lipid soluble (i.e., fat soluble), and because they can move easily through the cell’s membrane, they are not stored in the endocrine cells. Instead, a signal to an endocrine gland to produce steroid hormones essentially serves as a signal to release them into the blood as soon as produced by the cellular machinery.

Overview Essays j Hormones and Behavior: Basic Concepts

57

Hormone receptors, that are either embedded in the cell membrane or located elsewhere within the cell, inter-act with a particular hormone or class of hormones. Receptor proteins bind to hormones with high affinity and generally high specificity. As a result of the high affinity of hormone receptors, hormones can be very potent in their effects, despite their very dilute concentrations in the blood (e.g., 1 ng ml1 of blood). However, when blood concentrations of a hormone are high, binding with receptors that are specific for other related hormones can occur in sufficient numbers to cause a biological response (i.e., cross-reaction). Also, many hormones are structurally similar so that antibodies designed to attach to one hormone may cross react with other similarly shaped molecules (e.g., growth hormone and prolactin, the sex steroids, and the glycoproteins, viz., luteinizing hormone, folliclestimulating hormone, and thyroid-stimulating hormone). But generally, hormones can directly influence only cells that have specific receptors for that particular hormone and serve as target cells. The interaction of a hormone with its receptor begins a series of cellular events that either eventually leads to activation of enzymatic pathway or to a genomic response wherein the hormone acts directly or indirectly to activate genes that regulate protein synthesis. The newly synthesized proteins may activate or deactivate other genes, causing yet another cascade of cellular events see section ‘Steroid Hormones’ below. When sufficient receptors are unavailable because of a clinical condition, or because previous high concentrations of a hormone have occupied all the available receptors and new ones have yet to be made, a biological response may not be sustained (see later). Such a reduction in the numbers of receptors may lead to a so-called endocrine deficiency despite normal or even supernormal levels of circulating hormones. For example, a deficiency of androgen receptors can prevent the development of male traits despite normal circulating testosterone concentrations (Brinkman et al., 1996). Conversely, elevated receptor numbers may produce clinical manifestations of endocrine excess despite a normal blood concentration of the hormone. Thus, in order to understand hormone– behavior interactions, it is sometimes necessary to characterize target tissue sensitivity (i.e., the number and type of receptors possessed by the tissue in question) in addition to measuring hormone concentrations.

Protein Hormones Most vertebrate hormones are proteins. Protein hormones that comprise only a few amino acids in length are called ‘peptide hormones,’ whereas larger ones are called ‘protein’ or ‘polypeptide hormones.’ Protein and peptide hormones include insulin, the glucagons, the neurohormones of the hypothalamus, the tropic hormones of the anterior pituitary, inhibin, calcitonin, parathyroid hormone, the gastrointestinal (gut) hormones, ghrelin, leptin, adiponectin, and the posterior pituitary hormones. Protein and peptide hormones can be stored in endocrine cells and are released into the circulatory system by means of exocytosis. Protein and peptide hormones are soluble in blood, and therefore, do not require a carrier protein to travel to their target cells, as do steroid hormones. However, protein and peptide hormones may bind with other plasma proteins, which slow their metabolism by peptidases in the blood. Hormones are removed from the blood via degradation or excretion. The metabolism of a hormone is reported in terms of its biological half-life, which is the amount of time required to remove half of the hormone (radioactively tagged) from the blood. Generally, larger protein hormones have longer half-lives than smaller peptide hormones (e.g., growth hormone has 200 amino acids and a biological half-life of 20–30 min; thyroid-releasing hormone has three amino acids and a biological half-life of blue pathway). The attachment is probably influences a caregiver in the same environment (c). Reproduced from Purewal, R., Christley, R., Kordas, K., et al., 2017. Companion animals and child/adolescent development: A systematic review of the evidence. International Journal of Environmental Research and Public Health 14.

188

Animal Welfare and Conservation j Welfare and Interactions Between Humans and Companion Animals

Fig. 6 A hypothetical model based on reviewing literature of human (child)-companion animal (dog) interaction influences the physical and mental health of children. Reproduced from Gadomski, A.M., Scribani, M.B., Krupa, N., et al., 2015. Pet dogs and children’s health: Opportunities for chronic disease prevention? Preventing Chronic Disease 12, E205.

misunderstanding cycle (Fig. 4; bottom) shows the situation in which both participants do not interact in a successful as if they do not understand each other’s communication and no functional relation is reached. As described, this cycle may result with continuation of the mismatch in conflict and/or abnormal behavior in both participant (Rehn and Keeling, 2016). Individual differences between human caretaker and companion animal lead to a large variety of possible (mis)understanding situations and the authors of the model suggest focusing in future on such differences and not on the average companion animal. Another model of pet-child interaction shows how pets might influence play, caretaking, companionship, social interaction and health (Gadomski et al., 2015, 2017). Pet-child play increases activity and exploration promoting responsibility, social interactions and self-esteem and might reduce weight and increase health of the child. In summary, several models of HCAI in relation to individual welfare are found. They provide a framework of further study in human – companion animal interactions with species other than dogs. Especially in case the modelled interactions imply situations of negative welfare, more research should be done to resolve doubts concerning the pet suitability of the specific individuals and species kept.

Population Frameworks Interactions between humans and companion animals differ between type of human, species or breed of companion animal and the environment given to them. Concerning the positive and the negative processes and outcomes of the interactions the general scheme is as is depicted for the one health process approach (Fig. 1; Treadwell, 2008). Interactions between human-animal-environment deliver Social Network transmission of positive aspects of health and welfare responsibility. Cats, dogs, horses, rodents may develop overall as just pets or even pests and modelling the negative, but also the positive interactions shows how to handle changes in the balance on population level. Based on the Global One Health (GOH) approach the interdependence of human welfare with animal welfare and the interaction environment are described (Fig. 7). For that purpose, the center of the flow chart of Treadwell should be for example changed in “Welfare benefits and costs”. Environmental change, adaptation and resilience can be depicted best by the original description of the global and specific effects of changing environments on the welfare of farm animals (Fraser et al., 1997). The idea is applied here as the change in environment of human and companion animals and its consequences on behavior and welfare of both humans and animals (Fig. 8). It can be seen for instance as the change from the wild environment at the domestic one, the change from mostly outdoor living to mostly indoor living, or he change when companion animals become feral again and start to interact with humans in that new environment. Use figure for a visualization of changes in environment (Milieu). The description of Fraser the total of adaptations

Animal Welfare and Conservation j Welfare and Interactions Between Humans and Companion Animals

189

Fig. 7 Illustration of the domains and interfaces relevant for human companion animal interaction. Reproduced from Treadwell, T., 2008. Convergence of forces behind emerging and reemerging zoonoses, and future trends in zoonoses. Presented at the Institute of Medicine/National Research Council Workshop on Sustainable Global Capacity for Surveillance and Response to Emerging Zoonoses, Washington, DC, 25–26 June.

an animals (or here and human-companion animal interaction) possesses nine (circle A) and the challenges of that interaction in changed circumstances (Circle B) (Fig. 8). The care that human should give to companion animals should preferably dependent on the needs of the animal species and the environment in which it is kept (Swart and Keulartz, 2011). In case of companion animals in many cases is in a non-natural environment that according the hypothetical model (Fig. 9) deserves more specific than non-specific care (care for the environment). As probably more than 75% of the global dog population is free-ranging in a partly natural environment also non-specific care is important for maintaining a healthy dog population. Many wild species are also kept as companion animals and should in good cases kept with as much attention for specific as well as nonspecific care from the caretaker. As most dogs (75%) in the world are feral or stray animals in the UK no measurable population of such dogs is present. Most stray, feral, abandoned and ceded dogs are rehomed or reunited to new or old owners (Aegerter et al., 2017). The two main populations are owned and controlled (private owners) and the unowned and controlled (welfare organizations) groups (Fig. 10). The model can be used to understand balance between owned and unowned populations for instance in case no welfare organizations are present. A large increase in stray animals may be the consequence, dependent on the human-dog interaction of local authorities or the human population. In case stray dogs are fed a huge stray population can develop, in case stray dogs are not fed or cared for there might be even no population of stray dogs at all (Fig. 11). The structure of the cat population in the UK is simpler than the dog population, although differences between cats confined within homes or free-roaming cats exists with flexible exchange to the stray or even the feral population that avoid humans (Texel, the Netherlands), Veterinary consequences are described (Aegerter et al., 2017). These may lead health policies concerning pets described for instance in a pet policy model (Rock et al., 2015). Human and companion animal health and welfare are on municipality level cared for by (1) Preventing threats and nuisance from pets, (2) providing veterinary services for pets, (3) procuring pets ethically, (4) meeting emotional and physical needs of pets and (5) licensing and identifying pets. These five activities (Fig. 12) are associated with local governments’ jurisdiction over pets (Rock et al., 2015). By having a transparent system pet and human health and welfare issues can be retraced and cured by mediating complaints, fees, punishments to benefit the described human – companion animal interaction(HCAI) and ameliorate health and welfare of both pets and humans (Fig. 7).

190

Animal Welfare and Conservation j Welfare and Interactions Between Humans and Companion Animals

Fig. 8 Conceptual model of Fraser applied to human-CA interactions includes three categories of adaptations that fit challenges in the environment (3), adaptations that are useless (1), and adaptation that should be there to counteract challenges, but are absent (2).

Fig. 9 Model describing specific (species) and non-specific (milieu) care for as a function of their environment. Six classes of human-animal relation are presented based on dependence and on adaptivity to humans, i.e. (I) non-adaptive wild animals, (II) adaptive wild animals, (III) animals in human environmentsr, (IV) feral animals, (V) wild animals in zoos, and (VI) pet animals and livestock. See Swart, J.A.A., Keulartz, J., 2011. Wild animals in our backyard. A contextual approach to the intrinsic value of animals. Acta Biotheoretica 59, 185–200 for details.

Areas Where More Research is Needed Methodology In many cases the methodology behind research in the influence of pets on human health and welfare is not conclusive, limited and sometimes even wrong, for instance by interpreting correlational relations and outcomes as causal ones. By comparing pet and nonpet owners many differences between them in measures as gender, age, race/ethnicity, living arrangements, and income (Saunders et al., 2017). Pet owner characteristics are associated with better mental and physical health outcomes with implications for the proper interpretation of a large number of studies. Based on this comparison statistical advice could be given and statistical correction of bias in studies could be applied. Using fear and intimidation to manage pets is pervasive and results in conditioned responses that perpetuate fearful responses and fear-aggression. Additional research is needed to better understand the physiological changes that underlie acute and chronic stress responses when forceful methods are used to manage pets. These findings may then be compared to management techniques that establish positive conditioned responses and enhance welfare. In addition, the differential behavioral response of pets to familiar and unfamiliar individuals, both conspecific and heterospecific, warrants further research (Waiblinger et al., 2006; Hosey, 2008). Discovering key factors using the combined knowledge of veterinarians and animal welfare researchers probably provides a step forward (Dawson et al., 2016).

Animal Welfare and Conservation j Welfare and Interactions Between Humans and Companion Animals

191

Fig. 10 Changes in environment in dogs in Britain, i.e. with input of the dog population in the UK. Reproduced from Aegerter, J., Fouracre, D., Smith, G.C., 2017. A first estimate of the structure and density of the populations of pet cats and dogs across Great Britain. PLOS ONE 12.

Fig. 11 Changes in environment in cats in Britain. Reproduced from Aegerter, J., Fouracre, D., Smith, G.C., 2017. A first estimate of the structure and density of the populations of pet cats and dogs across Great Britain. PLOS ONE 12.

Pioneering studies correlate behavior signs with pain, canine cardiac disease, and feline interstitial cystitis. This approach is needed for other common diseases and conditions, particularly since owners often present ill pets to veterinarians on the basis of behavioral signs. For example, behavioral concomitants of endocrine diseases such as diabetes, hypothyroidism in dogs, and hyperthyroidism in cats will improve our understanding of behavioral correlates of disease processes.

Behavioral Biology or Ethology There is a lack of data on behavior and welfare of pets. Literature shows few follow-up studies in veterinary medicine other than an evaluation of the health condition and treatment, and few consider animal welfare based on behavior assessments. In addition to veterinary knowledge, more ethology or behavioral knowledge has to be used and should be made available to assess the welfare status of companion animals. Especially communicative signals of animal species can be recorded to enhance our knowledge of positive and negative expressions or behaviors of emotion and/or affect. In addition, validation in its several forms (construct, internal, external) must have much more attention than before. A while may in the end limited or even single behavioral elements

192

Animal Welfare and Conservation j Welfare and Interactions Between Humans and Companion Animals

Fig. 12 The presented Human-Companion animal interaction policy framework emphasizes the interaction between welfare of humans and companion animals and their interdependence.

may be recorded, a broader perspective using many behaviors or behavioral catalogues should give insight in the welfare status of an animal. In some areas ethology/behavioral biology is needed (Christiansen and Forkman, 2007):

• • • • • •

The analysis, validation and use of behavior that reveals information about subjective experience in pet animals. Ethological methods are the most suitable for this purpose. Development of behavioral tests to measure preference, motivation, need or want of pet animals. Especially short tests are needed that can be applied in veterinary practice for individual animals of many species. Integration or correlation of veterinary, physiological and behavioral tests to assess the welfare state of the animal. Collection of data during practice and of course in experiments can reveal hidden patterns (using big data) that support decision about welfare in practice. There is a huge need to identify welfare assessment tools for all companion and pet animals, especially since there is a tendency to keep more exotic pets compared to traditional ones (Schuppli et al., 2014). In fact, most knowledge is about cats and dog and a little on rabbits, while welfare assessment tools are needed for all species kept. Investigate the suitability of all species as pets, i.e. mammals, birds, reptiles and fish. Decide on the suitability of a species and design protocols to secure the welfare of that species. Investigate normal, adaptive, unwanted and abnormal behavior in pet animals and determine the consequences for keeping the individual animal and/or keeping the species.

Continue to Develop Welfare Quality Approaches The principles of good feeding, good housing, good health and appropriate behavior could be applied to all pet species kept. More effort should be put in development of such protocols for mammal, bird and reptile species that are kept as pets. Such protocols should be developed for groups of animals and individual animals for application in a large variety of circumstances and practices. The protocols should be easy to apply. The Welfare Quality approach for on-farm welfare assessments of cattle, pigs and chicken (Blokhuis et al., 2010) provides an example of an approach towards welfare assessment that could be used on group level (see dog and cat shelter welfare) and preferable should be developed for welfare assessment on individual level. For example, a dog that growls may be interpreted as dominant by its owner in spite of the fact that its visual communication signals indicate fear. The dual problems of anthropomorphism and misinterpretation of communication signals bias the owner as reporter with regard to the animal’s welfare and may result in the pet being treated by the owner in a manner considered conducive but actually detrimental to its welfare, failing to meet its species-typical behavioral needs. In conclusion, more research and ideas need to be developed in the area of using owner information (proxy assessment) and concerning the subjective experience of the pet animal. Using behavioral methods to assess the welfare state of an animal may replace the use of Quality of Life (QoL) or more consistent use will develop in a renewal of the QoL concept.

Animal Welfare and Conservation j Welfare and Interactions Between Humans and Companion Animals

193

Fig. 13 Framework to assess pet suitability using bibliographic information of animal species and best professional judgment of animal scientists (AS). The AS1-team selects relevant oneliners from literature, the AS2-team assesses from the oneliners the strengths’ of needs and risks, and the AS3-team assesses from those assessed strengths combined the final pet suitability of 90 mammal species. Reproduced from Koene, P., De Mol, R.M., Ipema, B., 2016b. Behavioral ecology of captive species: Using bibliographic information to assess pet suitability of mammal species. Frontiers in Veterinary Science 3.

Suitability and Welfare in Animal Species and Breeds In the Netherlands apart from dogs and cats (4.1 M individuals) thousands of other species are kept (30 M individuals), including hundreds of mammal species. Based on welfare considerations and welfare assessments the keeping and the suitability of many pet species should be reconsidered and maybe less species should be kept (Schuppli and Fraser, 2000; Engebretson, 2006; Soulsbury et al., 2009; Schuppli et al., 2014; Koene et al., 2016). On the other hand, of the species kept we need much more information about their behavior and welfare assessment is needed. Of cats and dogs several protocols and assessments of welfare are published, most for shelters and rescue centres. There are only few protocols and assessments available for other species, while many are needed for veterinary practice, rescue centres and shelters. Only for rabbits (Schepers et al., 2009) and reptiles (Warwick et al., 2013) protocols are available and welfare in parrots attracts some attention (Speer, 2014). How could we assess the suitability of an animal species as pet or companion animal? Several frameworks are available ranging from using characteristics of domestication and exaptation (Hale, 1962; Koene, 2014), the suitability framework (Schuppli and Fraser, 2000; Schuppli et al., 2014), the Emode approach (Warwick et al., 2014) to the bibliographical and behavioral needs approach (Koene et al., 2016). In this last approach knowledge and information about behavior, welfare, health and HCAI (HAR) from the animal in the wild and in captivity are combined and assess by several team of animal scientists (Fig. 13). In addition, stakeholders could make their assessment of suitability after adding bibliographic information and assessing the information in the database. The database can additionally be filled with measured animal behavior and animal welfare observations. The outcome generated by this system are a rank order of Odd ratio assessments of the risk on good welfare divided by the risk on bad welfare in captive conditions of the analysed species. The system was in the end not adopted in the Netherlands because of stakeholders pressure and the demand on more stakeholder influence on the final pet suitability assessment. The above flowchart depicts the assessment process of bibliographic information to arrive at a general assessment of pet suitability of a species. The final stage of such a flow chart is reached when all parameters relevant for the calculation of pet suitability are in fact measurable and measured making assessments by experts superfluous (Jones et al., 2009). Until all animal species are assessed for their pet suitability however, the suggested model might have a function in guaranteeing welfare in pet animal species.

See also: Animal Welfare and Conservation: Indicators of Good Welfare; Sentience. Methodology: Assessment of Welfare and Needs. Overview Essays: Welfare Concepts.

194

Animal Welfare and Conservation j Welfare and Interactions Between Humans and Companion Animals

References Aegerter, J., Fouracre, D., Smith, G.C., 2017. A first estimate of the structure and density of the populations of pet cats and dogs across Great Britain. PLOS ONE 12. Amiot, C., Bastian, B., Martens, P., 2016. People and companion animals: It takes two to tango. BioScience 66, 552–560. Anderson, D.C., 2007. Assessing the Human-Animal Bond: A Compendium of Actual Measures. Purdue University Press, West Lafayette, Indiana, US. Arhant, C., Wogritsch, R., Troxler, J., 2015. Assessment of behavior and physical condition of shelter cats as animal-based indicators of welfare. Journal of Veterinary BehaviorClinical Applications and Research 10, 399–406. Barnard, S., Pedernera, C., Candeloro, L., et al., 2016. Development of a new welfare assessment protocol for practical application in long-term dog shelters. Veterinary Record 178, 18–U44. Belshaw, Z., Asher, L., Harvey, N.D., Dean, R.S., 2015. Quality of life assessment in domestic dogs: An evidence-based rapid review. Veterinary Journal 206, 203–212. Blokhuis, H.J., Veissier, I., Miele, M., Jones, B., 2010. The Welfare Quality (R) project and beyond: Safeguarding farm animal well-being. Acta Agriculturae Scandinavica Section A – Animal Science 60, 129–140. Bokkers, E.A.M., De Vries, M., Antonissen, I.C.M.A., De Boer, I.J.M., 2012. Inter- and intra-observer reliability of experienced and inexperienced observers for the qualitative behaviour assessment in dairy cattle. Animal Welfare 21, 307–318. Bradbury, A.G., Dickens, G.J.E., 2016. Appropriate handling of pet rabbits: A literature review. Journal of Small Animal Practice 57, 503–509. Christiansen, S.B., Forkman, B., 2007. Assessment of animal welfare in a veterinary context – A call for ethologists. Applied Animal Behaviour Science 106, 203–220. Dawkins, M.S., 2004. Using behaviour to assess animal welfare. Animal Welfare 13, S3–S7. Dawkins, M.S., 2008. The science of animal suffering. Ethology 114, 937–945. Dawson, L.C., Dewey, C.E., Stone, E.A., Guerin, M.T., Niel, L., 2016. A survey of animal welfare experts and practicing veterinarians to identify and explore key factors thought to influence canine and feline welfare in relation to veterinary care. Animal Welfare 25, 125–134. Duijvesteijn, N., Benard, M., Reimert, I., Camerlink, I., 2014. Same pig, different conclusions: Stakeholders differ in qualitative behaviour assessment. Journal of Agricultural & Environmental Ethics 27, 1019–1047. Engebretson, M., 2006. The welfare and suitability of parrots as companion animals: A review. Animal Welfare 15, 263–276. Farm Animal Welfare Council, 2012. Five freedoms [Online]. Available at: http://webarchive.nationalarchives.gov.uk/20121007104210/, http://www.fawc.org.uk/freedoms.htm (accessed 17.02.18). Fraser, D., 1999. Animal ethics and animal welfare science: bridging the two cultures. Applied Animal Behaviour Science 65, 171–189. Fraser, D., Weary, D.M., Pajor, E.A., Milligan, B.N., 1997. A scientific conception of animal welfare that reflects ethical concerns. Animal Welfare 6, 187–205. Gadomski, A.M., Scribani, M.B., Krupa, N., et al., 2015. Pet dogs and children’s health: Opportunities for chronic disease prevention? Preventing Chronic Disease 12, E205. Gadomski, A.M., Scribani, M.B., Krupa, N., Jenkins, P., 2017. Pet dogs and child physical activity: The role of child-dog attachment. Pediatric Obesity 12, E37–E40. Gartner, M.C., 2015. Pet personality: A review. Personality and Individual Differences 75, 102–113. Hale, E.B., 1962. Domestication and the evolution of behaviour. In: Hafez, E.S.E. (Ed.), The Behaviour of Domestic Animals, second ed. Baillière, Tindall & Cassell, London, pp. 22–42. Hammerschmidt, J., Molento, C.F.M., 2015. Protocol for expert report on animal welfare in case of companion animal cruelty suspicion. 2015 51, 15. Herwijnen, I.R.V., Van Der Borg, J.A.M., Naguib, M., Beerda, B., 2018. The existence of parenting styles in the owner-dog relationship. PLOS ONE 13, e0193471. Hewson, C.J., Hiby, E.F., Bradshaw, J.W.S., 2007. Assessing quality of life in companion and kennelled dogs: a critical review. Animal Welfare 16, 89–95. Hosey, G., 2008. A preliminary model of human-animal relationships in the zoo. Applied Animal Behaviour Science 109, 105–127. Jones, K., Bielby, J., Cardillo, M., Fritz, S.A., O’dell, J., et al., 2009. PanTHERIA: A species-level database of life history, ecology, and geography of extant and recently extinct mammals. Ecology 90, 2648. Kessler, M.R., Turner, D.C., 1997. Stress and adaptation of cats (felis silvestris catus) housed singly, in pairs and in groups in boarding catteries. Animal Welfare 6, 243–254. Koene, P., De Mol, R., Ipema, B., 2016. Behavioral ecology of captive species: Using bibliographic information to assess pet suitability of mammal species. Frontiers in Veterinary Science 3. Koene, P., 2014. Pet exaptation index indicates potential adaptation of mammal species for living in human environments. In: Estevez, I., Manteca, X., Marin, R.H., Averós, X. (Eds.) Proceedings of the 48th Congress of the International Society for Applied Ethology (ISAE 2014), Wageningen Academic Publishers, The Netherlands. Ladewig, J., 2005. Of mice and men: Improved welfare through clinical ethology. Applied Animal Behaviour Science 92, 183–192. Lavan, R.P., 2013. Development and validation of a survey for quality of life assessment by owners of healthy dogs. Veterinary Journal 197, 578–582. Lesimple, C., Hausberger, M., 2014. How accurate are we at assessing others’ well-being? The example of welfare assessment in horses. Frontiers in Psychology 5. Litchfield, C.A., Quinton, G., Tindle, H., et al., 2017. The ‘Feline Five’: An exploration of personality in pet cats (Felis catus). PLOS ONE 12. Loukaki, K., Koukoutsakis, P., Kostomitsopoulos, N., 2010. Animal welfare issues on the use of rabbits in an animal assisted therapy program for children. Journal of the Hellenic Veterinary Medical Society 61, 220–225. Mason, G., Burn, C.C., Dallaire, J.A., et al., 2013. Plastic animals in cages: behavioural flexibility and responses to captivity. Animal Behaviour 85, 1113–1126. Mason, G.J., 2010. Species differences in responses to captivity: stress, welfare and the comparative method. Trends in Ecology & Evolution 25, 713–721. Mullan, S., 2015. Assessment of quality of life in veterinary practice: Developing tools for companion animal carers and veterinarians. Veterinary Medicine: Research and Reports 6, 203–210. Mullan, S., Main, D., 2007. Preliminary evaluation of a quality-of-life screening programme for pet dogs. Journal of Small Animal Practice 48, 314–322. Nagasawa, M., Mitsui, S., En, S., et al., 2015. Oxytocin-gaze positive loop and the coevolution of human-dog bonds. Science 348, 333–336. Nagasawa, M., Okabe, S., Mogi, K., Kikusui, T., 2012. Oxytocin and mutual communication in mother-infant bonding. Frontiers in Human Neuroscience 6. Nagasawa, M., Ogawa, M., Mogi, K., Kikusui, T., 2017. Intranasal oxytocin treatment increases eye-gaze behavior toward the owner in ancient Japanese dog breeds. Frontiers in Psychology 8. Odendaal, J.S.J., Meintjes, R.A., 2003. Neurophysiological correlates of affiliative behaviour between humans and dogs. Veterinary Journal 165, 296–301. Potter, A., Mills, D.S., 2015. Domestic cats (Felis silvestris catus) do not show signs of secure attachment to their owners. PLOS ONE 10. Rault, J.-L., 2015. Pets in the digital age: Live, robot, or virtual? Frontiers in Veterinary Science 2. Rayment, D.J., De Groef, B., Peters, R.A., Marston, L.C., 2015. Applied personality assessment in domestic dogs: Limitations and caveats. Applied Animal Behaviour Science 163, 1–18. Rehn, T., Keeling, L.J., 2016. Measuring dog-owner relationships: Crossing boundaries between animal behaviour and human psychology. Applied Animal Behaviour Science 183, 1–9. Rock, M.J., Adams, C.L., Degeling, C., Massolo, A., Mccormack, G.R., 2015. Policies on pets for healthy cities: A conceptual framework. Health Promotion International 30, 976–986. Sánchez-Villagra, M.R., Geiger, M., Schneider, R.A., 2016. The taming of the neural crest: A developmental perspective on the origins of morphological covariation in domesticated mammals. Royal Society Open Science 3. Saunders, J., Parast, L., Babey, S.H., Miles, J.V., 2017. Exploring the differences between pet and non-pet owners: Implications for human animal interaction research and policy. PLOS ONE 12. Schepers, F., Koene, P., Beerda, B., 2009. Welfare assessment in pet rabbits. Animal Welfare 18, 477–485.

Animal Welfare and Conservation j Welfare and Interactions Between Humans and Companion Animals

195

Schuppli, C.A., Fraser, D., 2000. A framework for assessing the suitability of different species as companion animals. Animal Welfare 9, 359–372. Schuppli, C.A., Fraser, D., Bacon, H.J., 2014. Welfare of non-traditional pets. Revue Scientifique Et Technique-Office International Des Epizooties 33, 221–231. Schuurmans, L., Enders-Slegers, M.J., Verheggen, T., Schols, J., 2016. Animal-assisted interventions in Dutch nursing homes: A survey. Journal of the American Medical Directors Association 17, 647–653. Soulsbury, C.D., Iossa, G., Kennell, S., Harris, S., 2009. The welfare and suitability of primates kept as pets. Journal of Applied Animal Welfare Science 12, 1–20. Speer, B., 2014. Normal and abnormal parrot behavior. Journal of Exotic Pet Medicine 23, 230–233. Stoeckel, L.E., Palley, L.S., Gollub, R.L., Niemi, S.M., Evins, A.E., 2014. Patterns of brain activation when mothers view their own child and dog: An fMRI study. PLOS ONE 9. Swart, J.A.A., Keulartz, J., 2011. Wild animals in our backyard. A contextual approach to the intrinsic value of animals. Acta Biotheoretica 59, 185–200. Treadwell, T., 2008. Convergence of forces behind emerging and reemerging zoonoses, and future trends in zoonoses. Presented at the Institute of Medicine/National Research Council Workshop on Sustainable Global Capacity for Surveillance and Response to Emerging Zoonoses, Washington, DC, 25–26 June. Trut, L.N., 1999. Early canid domestication: The farm-fox experiment. American Scientist 87, 160–169. Twardek, W.M., Peiman, K.S., Gallagher, A.J., Cooke, S.J., 2017. Fido, Fluffy, and wildlife conservation: The environmental consequences of domesticated animals. Environmental Reviews 25, 381–395. Waiblinger, S., Boivin, X., Pedersen, V., et al., 2006. Assessing the human-animal relationship in farmed species: A critical review. Applied Animal Behaviour Science 101, 185–242. Walsh, F., 2009a. Human-animal bonds I: The relational significance of companion animals. Family Process 48, 462–480. Walsh, F., 2009b. Human-animal bonds II: The role of pets in family systems and family therapy. Family Process 48, 481–499. Warwick, C., Arena, P., Lindley, S., Jessop, M., Steedman, C., 2013. Assessing reptile welfare using behavioural criteria. In Practice 35, 123–131. Warwick, C., Steedman, C., Jessop, M., Toland, E., Lindley, S., 2014. Assigning degrees of ease or difficulty for pet animal maintenance: The EMODE system concept. Journal of Agricultural & Environmental Ethics 27, 87–101. Wemelsfelder, F., Hunter, T.E.A., Mendl, M.T., Lawrence, A.B., 2001. Assessing the ‘whole animal’: A free choice profiling approach. Animal Behaviour 62, 209–220. Wilkins, A.S., Wrangham, R.W., Fitch, W.T., 2014. The “Domestication Syndrome” in mammals: A unified explanation based on neural crest cell behavior and genetics. Genetics 197, 795–808. Wolfensohn, S., Honess, P., 2007. Laboratory animal, pet animal, farm animal, wild animal: Which gets the best deal? Animal Welfare 16, 117–123.

Ecology of Fearq Joel S Brown, University of Illinois at Chicago, Chicago, IL, United States © 2019 Elsevier Ltd. All rights reserved.

Abstract Animals use vigilance, time allocation, and apprehension as fear responses for reducing the likelihood of being killed by their predators. As prey trade off food for safety, they alter predator–prey dynamics, the habitat selection of both prey and predators, and the number of coexisting prey species. Similarly, via predator facilitation, different predator species may coexist by exploiting the weaknesses of different predator-specific fear responses. Finally, prey fear responses may be vital for maintaining ecosystem integrity, preventing overgrazing, and conserving biodiversity. All these population, community, and ecosystem consequences of prey fear responses to predation constitute the ‘Ecology of Fear.’

Keywords Ecology of fear; Giving-up density; Landscape of fear; Mechanism of coexistence; Nonlethal effects; Predation; Predator facilitation; Predator–prey dynamics; Trait-mediated effects; Trophic cascade; Vigilance

Introduction A tourist’s video from South Africa posted on YouTube records an amazing sequence of events. A group of lionesses attempt the foolhardy. They attack a group of African buffalo, separating out a calf and dragging it into the waterhole, presumably to drown it and to fend off any counterattack from the buffalo. The buffalo retreat, but a crocodile unexpectedly grabs the calf. The lionesses have no choice but to drag the hapless calf back out of the water. This emboldens the buffalo which return, drive off the lionesses, and rescue the apparently mauled but otherwise healthy calf. This choreography of buffalo, lions, and crocodile illustrates the nexus between antipredator behaviors and their ecological consequences for births and deaths. The mother buffalo almost lost her calf, the crocodile almost got a meal, and the lionesses in losing their meal deprived a crocodile and inadvertently saved the calf. The ecology of predator–prey interactions involves how predators kill prey and prey feed predators. The behavioral ecology of predator and prey studies stealth and vigilance, and the topics of many of the preceding articles. The Ecology of Fear is the study of how the fear and anti-predatory behaviors of prey and the stealth and hunting behaviors of predators influence predator–prey interactions, population dynamics, species coexistence, and evolutionary dynamics. Long before I applied this term to the consequences of predator–prey behaviors on the ecology of predator–prey systems, others had recognized the strong connections. To name but a few: Michael Rosenzweig noted how fear responses of prey could radically change the dynamics and stability properties of predator–prey interactions (Rosenzweig and MacArthur, 1963). Steve Lima points out how a single goshawk swooping over a flock of flamingos could directly kill one, but it could set off a stampede that might kill several more (Lima and Dill, 1990). Oswald Schmitz showed how the nonlethal effects of predators could be as or more important than the lethal effects. He and his lab enclosed grasshoppers with spiders. In some treatments, the spiders were potentially lethal, whereas in others, the spiders had been rendered harmless; their chelae glued together. Relative to control populations, the number of grasshoppers declined equally whether the spiders were lethal or simply scary. Even without direct mortality from spiders, the grasshoppers experienced reduced fitness as they abandoned perceived risky habitats and forewent other feeding opportunities for the sake of fear (Schmitz et al., 1997). As the ecology of fear, we shall see how predation risk represents an additional activity cost for the prey. These fear responses then influence the population sizes and dynamical stability of predator–prey interactions. Fear responses also structure the spatial and the temporal landscapes of their prey, thus influencing and determining habitat suitability for both predator and prey. Fear responses can also create behavioral cascades up and down food chains. Fear responses may influence the length of food chains, and the coexistence of multiple prey species and predator species. Risk of injury from hunting creates fear responses from the predators. Finally, fear responses may be the unit of conservation, may be necessary for ecosystem health, and may provide behavioral indicators for the status of the prey and the whereabouts of their predators (Stephens et al., 2007).

Predation Risk as an Activity Cost Predators, as a direct effect, increase the mortality rate of prey by killing them. Similarly, the consumption of prey by a predator enhances her survivorship and fecundity. The lethality of predators directly affects the per capita growth rates of the prey and

q

Change History: June 2018. Donald Brown added reference citations to the text. This is an update of J.S. Brown, Ecology of Fear, Editor(s): Michael D. Breed, Janice Moore, Encyclopedia of Animal Behavior, Academic Press, 2010, Pages 581-587.

196

Encyclopedia of Animal Behavior, 2nd edition, Volume 1

https://doi.org/10.1016/B978-0-12-809633-8.20870-4

Animal Welfare and Conservation j Ecology of Fear

197

predators. But, these direct effects become modified as soon as the prey exhibit flexible fear responses. The prey do this by either allocating time away from risky times and places, or using vigilance behaviors to reduce risk while engaged in some risky activity such as foraging. In both cases, the prey may be forgoing other fitness-enhancing opportunities to avoid being killed by the predator. The fear responses reduce the direct effect of the predator on the prey. For this reason, the effects of fear on the lethal effects of predators have often been referred to as ‘trait-mediated effects,’‘behavioral indirect effects,’‘nonlinear effects,’ or even ‘higher-order interactions.’ Regardless, fear can best be thought of as the prey viewing predation risk as an activity cost. How big should this predation cost be? As an adaptation shaped by natural selection, the prey’s cost of predation should integrate three components: predation risk (m), survivor’s fitness (F), and the marginal fitness value of the activity (vF/ve). For a foraging animal, vF/ve describes how much its fitness prospects increase with an incremental increase in energy gained. For a foraging animal, the predation cost of the activity would be: mF/(vF/ve). In other words, an animal’s cost of predation can double either by doubling the level of predation risk, doubling what the animal has to lose from being killed in terms of future reproduction and survivorship, or halving the marginal value of energy to enhancing the animal’s fitness. This cost of predation forms the basis for the prey’s fear responses – the greater the cost of predation, the greater will be the prey’s fear responses. Interestingly, this means that hungry animals (high vF/ve) or animals in a poor state (low F), should be easier to catch. The foraging (or activity) cost of predation forms the basis for antipredatory behaviors that influence predator–prey interactions, population dynamics, and species coexistence.

The Paradox of Fierce Carnivores Consider a textbook predator–prey model. In the state space of predator density versus prey density, the prey’s isocline is hump shaped, and the predator’s is a vertical line (Fig. 1). The prey’s (or predator’s) isocline shows all combinations of prey and predator numbers such that the prey’s (or predator’s) growth rate is zero. The prey’s isocline rises at low prey numbers as the prey experience safety in numbers, but then it reaches a peak and declines as the negative fitness consequences of competition among the prey outweighs benefits of reduced predation risk. The predator’s isocline is independent of the number of predators and only depends on prey abundance. In its simplest version, the predators only interact with each other indirectly through the consumption of prey – they do not interfere with each other. Hence, the predators have a subsistence abundance of prey that they require to just get by. Now, the paradox of such a predator–prey system. If the predators are very inefficient and require a high subsistence abundance of prey, then their isocline intersects the prey’s to the right of the hump (see Fig. 1). This yields a nice stable equilibrium, although there may be damped oscillations as the prey and predator dynamics approach this equilibrium. But, if the environment was to degrade and support fewer prey, then the predator’s face extinction should the prey population numbers become too low. More efficient predators do not face this problem. They only require a low abundance of prey for subsistence and their isocline intersects the prey’s to the left of the hump. But, such an intersection produces nonequilibrium dynamics with permanent limit cycles or fluctuations that may result in the extinction of the prey, predator, or both. Inefficient predators may be extinction prone from extrinsic perturbations to the environment, while efficient predators may create intrinsic instabilities in the dynamics resulting in extinction. Prey fear responses may rescue the paradox and promote the stability of predator–prey system with fierce carnivores. Fierceness is not really a quality of the predator. Squirrels prey on acorns, but acorns do not scream, or get up and run away – hence squirrels are not fierce predators. Fierceness emerges because the prey can perceive and respond to the predators – the capacity of the prey to fear and respond makes a predator fierce. Let’s revisit the isoclines and imagine a fierce predator. When the predators are very scarce, the prey should have a low predation cost of foraging and should show little vigilance or fear responses. To a predator, they are easier to catch – such a predator may be quite efficient and require only a low density of prey to subsist. At low predator numbers, the predator’s isocline starts as an efficient predator. But, as predator numbers increase, so does the prey’s predation cost of foraging, and the prey become harder to catch. With more predators, the predators now have a higher subsistence level of prey. The predator’s isocline is not vertical but has positive slope (Fig. 1). Because of the prey’s fear responses, the predators are extremely efficient on unwary prey when predators are scarce (resistant to extrinsic perturbations), and the predators are extremely inefficient on highly wary prey when predator’s are abundant (resistant to intrinsic instabilities in population dynamics). Fear responses may be a critical component in predator–prey dynamics, and the stability and persistence of predator–prey systems with fierce carnivores. Mule deer and mountain lions, and wolves and elk in Yellowstone provide examples of such systems (Altendorf et al., 2001).

Landscapes of Fear, and Habitat Quality for Prey and Predators In thinking about the wolves and elk of Yellowstone, John Laundre developed the concept of the landscape of fear as an important component of habitat suitability in addition to landscapes of productivity, vegetation cover, food availability and/or physical properties of the landscape (Laundre et al., 2001). One can imagine a vegetation map, or a topographic map of elevation, or a mapping of the soil characteristics. The landscape of fear is a topographic map where the ‘elevation’ lines represent lines of equal predation cost (Fig. 2). Creating this map from direct observations of risk would be nigh impossible. Typically, this mapping of fear is measured by setting out a grid of depletable food patches and measuring the foragers’ giving-up densities. Spatial statistics can then extrapolate these into lines of equal giving-up density. Or, if the giving-up densities are converted into quitting harvest rates, the lines of equal foraging cost can be presented in units of energy per unit time. Either way, changes in the ‘elevation’ of this map represent changes in the predation cost of foraging.

198

Animal Welfare and Conservation j Ecology of Fear

Fig. 1 How prey fear responses can stabilize predator–prey interactions. The graphs show various configurations of predator–prey dynamics as shown using isoclines diagrams. The hump-shaped curve is the prey’s isocline showing all combinations of prey and predator densities such that the prey have zero growth rate (prey growth rate is positive below this isocline and negative above). The predator isoclines (all combinations of prey and predator densities such that the predator has zero population growth rate) represent three different scenarios and outcomes. The vertical predator isocline in (a) shows an inefficient predator that needs a high density of prey to subsist. A drop in prey number below this line would result in the predator’s extinction. The vertical predator isocline in (b) shows a highly efficient predator (low density of prey required to subsist). Because the isocline intersects the prey’s in a region of positive slope, the equilibrium point is unstable yielding nonequilibrium dynamics that could result in the extinction of the predator. The predator isocline in (c) that curves gently upward shows a predator whose prey exhibit appropriate fear responses. At low predator densities, the prey can be fearless and hence easy to catch (the predator is highly efficient); at higher predator densities, the increasingly fearful prey become harder to catch (the predators become highly inefficient). The net effect of this positively sloped isocline is a stable equilibrium and a predator population that is well buffered from declines in prey abundances.

The landscape of fear can be compared with other mappings of productivity, vegetation structure, and other physical features to assess how these features may relate to predation risk (Brown et al., 1999; Brown and Kotler, 2004). A perspective on both the prey’s and predator’s habitat quality emerges by comparing the landscape of fear with the animal’s overall level of feeding activity. We can simplify this into a 22 matrix where columns represent ‘Low’ versus ‘High’ activity levels, and rows represent ‘Low’ versus ‘High’ predation cost. Various combinations represent core habitats, refuge habitats, unsuitable habitats, and valuable but risky habitats. This last habitat provides the greatest opportunities for the predator – high prey activity under conditions of high risk. Such habitats may be core to the predators (Table 1). Spatial variation in predation risk provides opportunities for habitat selection. If prey move freely between habitats and they do not interfere with each other, theory predicts an ideal free distribution where prey distribute themselves between habitats so that each habitat provides the same fitness reward (expected per capita growth rate to the individual). If habitats simply vary in productivity, then the distribution of individuals between habitats should match productivities (resource matching). If habitats also vary in the risk of predation, then individuals should distribute themselves in a manner that roughly matches the ratio of reward to risk (‘m/f rule’ where m is predation risk and f is feeding rate). If the predators are also free to move so as to equalize their fitness opportunities, then habitat selection becomes a predator–prey foraging game (Hugie and Dill, 1994). One outcome has the predators matching the resources available to the prey and the prey distributed so as to equalize the predators’ opportunities. Additional distributions emerge when the predators interfere with each

Animal Welfare and Conservation j Ecology of Fear

199

Fig. 2 A landscape of fear for a colony of eight cape ground squirrels (Xerus inauris) was measured in Augrabies Falls National Park, South Africa. The lines of equal quitting harvest rate (converted from giving-up densities in experimental food patches) have units of kilojoules per minute. Differences between lines show how much the predation cost of foraging changes in space. The extrapolation of this map was made from an 88 grid of food patches (blue open circles) spaced at 10-m intervals (LON refers to 10-m increments from West to East; and LAT indicates 10-m increments from South to North). In this desert grassland, the squirrels’ perception of safety increased with the proximity of burrows (shown as red open circles).

Table 1

How the prey’s landscape of fear (as measured by giving-up densities, the amount of food left behind by prey in food patches), and the prey’s activity patterns can indicate the Prey’s Refuge, Unsuitable, and Core habitats, and the Predator’s Core habitat

Feeding activity Low High

Giving-up density Low

High

Refuge: safe and poor in opportunities Core: safe and rich in opportunities

Unsuitable: risky with low opportunities Predator’s core: risky and rich in opportunities

other (this promotes a more even distribution of predators), when the prey can respond with vigilance behaviors (this tends to even out the distribution of prey), and when the prey exhibit safety in numbers (this may cause prey to herd up in one habitat or the other). The interplay between resource opportunities, predation risk, and available fear responses strongly influences the distribution and abundance of prey, and subsequently that of the predators via foraging games that may involve three or more trophic levels.

Behavioral Trophic Cascades Why is the world green? All those leaves, blades of grass, and small plants represent food for herbivorous insects, vertebrates, and many other taxa. Furthermore, overbrowsing and overgrazing do occur. In fact, the efficient consumption of phytoplankton and periphyton produces the clean, clear oligotrophic lakes that we humans prefer over pond-scummy eutrophic ones. Three hypotheses include the idea of a ‘green desert.’ Herbivores may eat most of the nutritious plant material, leaving behind the unfavorable bits. Or, plants are highly defended with silicates to grind down herbivore mouthparts, secondary compounds meant to poison and deter, and/or structural defenses such as spines, or bark meant to harm or discourage the herbivores. Such defenses represent the plant’s non-cognitive fear responses to their own predation cost of nutrient foraging. Finally, predators may consume herbivores that otherwise would have eaten plants. This represents a classic trophic cascade where the predators indirectly benefit the plants by consuming their herbivorous predators – from the plant’s perspective, ‘the enemy of my enemy is my friend.’ All three of these hypotheses may have analogs in the ecology of fear. It has been suggested, but not empirically verified, that plants deter herbivory by producing tissue of lower nutritional value than otherwise would be optimal. Foragers have higher giving-up densities on foods of low quality or high handling times. Structural and chemical defenses raise the forager’s costs or lower its harvest rate – both will raise giving-up densities. Finally, predators can create a behavioral trophic cascade. The presence of the predators raises the herbivore’s predation cost of foraging, and the herbivores in response forage less intensively on the plants. A behavioral cascade can be reversed very quickly (Abramsky et al., 1985). Desert spiny mice (Acomys cahirinus) consume just a few desert snails. But when numerous rock refugia were added to their home ranges, the spiny mice, now much more protected from their predators, consumed vast numbers of snails in a matter of days. Many cases of overgrazing or overharvesting may have less to do with too many herbivores, and much to do with herbivores that are not fearful.

200

Animal Welfare and Conservation j Ecology of Fear

Fear and Mechanisms of Species Coexistence A mechanism of coexistence requires environmental variability and tradeoffs among the organisms exploiting this variability. If variability takes the form of habitat heterogeneity, then one species may be better at competing in one habitat while the other may have the competitive edge in another. Fear responses of prey to their predators likely create some coexistence mechanisms and promote biodiversity. The mechanism of coexistence generally involves some sort of spatial or temporal variability in predation risk, and the tradeoff among the different prey species is either quantitative (one prey is better at avoiding predators, while the other is better at garnering food) or qualitative (one prey feels safer under one set of conditions, while the other feels safer under another). Differences in antipredator strategies between night-time and daylight provide a stark example of qualitative tradeoffs. Daylight plays into the sense of vision for both prey and predators. Darkness favors hearing and smell. While many species may be active both day and night, most have a strong preference for either daylight, darkness, or as crepuscular animals, the twilight that separates them. While thermoregulatory and water regulation may play large roles in determining temporal patterns of activity, predation risk and the latitude for fear responses likely loom large for most organisms choosing their times of day or night. Zooplankton exhibit a regular pattern of daily migration up and down in the water columns of temperate ponds and lakes – darker depths provide safety during the day from visually orienting fish, the upper water column provides food and relief from predatory invertebrates. Coral reefs experience an almost complete change from the ‘day-shift’ to the ‘night-shift’ community. The colorful reef fishes seem to gain a visual edge over their predators by day with barracudas patrolling just off the reef, and other predators attempting to wait in ambush among the corals. But, by night, they retreat completely to refugia among the coral; presumably, the reduced visibility prevents these reef fishes from distinguishing between a harmless competitor and a predator. Meanwhile, the sea urchins have been hunkered down all day – to come out during the day invites attacks from trigger fish (they can literally blow the urchins onto their backs exposing the urchin’s vulnerable underbellies) and the like. But, come night, the triggerfish retreat, and urchins can more safely patrol and graze throughout the reefs and even the adjacent sandflats. Coral reef fish of the night offer bizarre morphologies of spines, poisons, and behaviors of puffing up. The separation of communities by night and day is not just a matter of predation pruning each species into its temporal niche; rather, these organisms exhibit strong coadaptations of behaviors and morphology that allow them to avoid periods of danger and seek periods of relative safety. Fear responses strongly stabilize the coexistence and dynamics of these temporal communities, as each ‘shift’ avoids intruding on the others. Space provides much the same opportunity for mechanisms of coexistence based on tradeoffs in food and safety. Among Heteromyid rodents, there are situations where kangaroo rats (Dipodomys sp.) seem to have the competitive edge in the risky, open microhabitats, while pocket mice (Chaetodipus and Perognathus) enjoy an edge in the safer bush microhabitat; although, as we shall see, snakes versus owls can wonderfully complicate this simpler scenario. Rosenzweig tested for these effects by augmenting cover on some plots in the desert and removing shrub cover from others. Via population dynamics, we might expect over weeks or months a slow steady shift in numbers as differential mortality and fecundity favored kangaroo rats on the coverless plots and favored pocket mice on the cover-augmented plots. Not so. Within a day or two, the kangaroo rats quickly abandoned the augmented plots, and the pocket mice quickly abandoned the open plots. Fear accelerated the dynamics and likely drove the pocket mice away, but another feedback was likely at work in cover plots. Less fearful pocket mice likely depressed seed resources (remember the behavioral cascade) to the point where kangaroo rats could not profitably feed; they needed to go elsewhere. Fear as a driver of coexistence can operate in unexpected but marvelous ways. My favorites are fox squirrels and gray squirrels of the Midwest United States. Fox squirrels are orange in color, slightly larger, and have a temperament that seems to focus on ‘managing’ their predators, that they will mob or harass. Gray squirrels, on the other hand, are most of the time gray (with melanistic ones making up most of the rest), slightly smaller, and seem to have a temperament focused on managing other squirrels and competition for food. They are smaller, but they are the interference dominant over fox squirrels. Fox squirrels predominate on the riskier and less productive wood margins or savannas, while gray squirrels predominate in the safer deep woods. The temperament, body size, and even coat color (a bit of apomatism?) of fox squirrels may all be antipredator adaptations allowing a competitive edge in riskier habitats. Habitat heterogeneity in fear is necessary for their coexistence. But, the presence of squirrels alone is probably not responsible for the large predator community (hawks, coyotes, foxes) that exists in these habitats and creates this fear! Squirrels form a relatively small component of these predators’ diets, and if there were only squirrels around, there probably would be many fewer predators. Rather, voles, chipmunks, cottontail rabbits, and white-footed mice likely feed the predators that create the mechanism of coexistence for the two squirrel species. In the absence of these more accessible prey, there would be many fewer predators, and gray squirrels would likely outcompete fox squirrels in this much safer world. We see a strong interaction of lethal and nonlethal effects in creating this coexistence.

Predator Facilitation Coexistence among predators may be facilitated if predators can exploit the prey’s predator-specific fear responses. Predator facilitation likely is widespread and may promote differences in predator-hunting tactics (sit-and-wait vs. active pursuit), as well the coexistence of diverse predator taxa. As a direct effect, horned vipers and barn owls would be labeled as ‘competitors’ as both seek to kill gerbils inhabiting the sand dunes of Middle Eastern deserts. But, predator facilitation is at work and snakes likely make it easier for owls to kill gerbils and vice versa. Appropriately, gerbils head for the cover of shrubs in response to owls. Snakes

Animal Welfare and Conservation j Ecology of Fear

201

exploit this fear tactic by lying in ambush under shrubs – fear of owls may reinforce the snake’s behavior. The presence of snakes under shrubs then drives gerbils into the waiting talons of the owls. The lionesses hunting the buffalo calf may have been predator facilitation gone awry. The presence of crocodiles may make the approach of the buffalo to the waterhole slow and deliberate. This may increase the chances for the lions to conduct an ambush of their own. Furthermore, wariness for lions may increase the crocodile’s chances of capturing an antelope. Split attention, predatorspecific fear responses, and the overall penalty of multitasking may enhance both prey and predator diversity through predator facilitation.

What are the Predators Afraid of? The ecology of fear influences predators in at least two ways. First, the predators themselves may be at risk from each other or from other predator species. Second, the predators may be at risk of injury or even death from capturing their prey. Intraguild predation, or additionally, intraguild harassment recognizes that predators, while hunting for the same prey may harass, injure, or interfere with each other. A predator such as a leopard in Chitwan National Park, Nepal, or a red fox in the Midwestern United States loses twice when they face an even larger predator such as tigers or coyotes, respectively. Not only do the tigers and coyotes frighten and compete for the same prey as the leopard and fox, but the leopard and fox become less effective predators as they must vigilantly watch their backs while hunting. The reappearance of coyotes at the Morton Arboretum, Lisle, Illinois, was marked by the rapid disappearance of red foxes. It is unlikely that the coyotes killed many if any foxes, but the foxes now found their former habitat unprofitable in the face of these twin penalties. Intraguild predation adds a predation-like cost of foraging to mesopredators caught in the middle. Mesopredator release refers to the increase in numbers and impact of midsize carnivores when a larger predator has been extirpated from the system. Like overgrazing by herbivores, much of this release may stem from the mesopredator’s ability to forage so much more efficiently once freed from its predation cost of foraging (Abrams, 2010). Large carnivores may actually be quite fragile. In the absence of any assistance from a social group, a large carnivore may face death if it suffers even a moderate injury that temporarily reduces its capacity to hunt. A swollen paw or a sprained muscle may render a cheetah incapable of capturing its prey – starvation may be imminent. This suggests that large carnivores may need to hold back in hopes of reducing the risk of injury. Perhaps they even forgo opportunities to make a risky kill. Interestingly, this scaling of boldness or derring-do allows predators to tradeoff risk of injury with likelihood of prey capture. A cautious predator risks little injury but has a concomitantly lower chance of successfully capturing its prey. Derring-do and risky prey likely give predators additional degrees of freedom to manage their hunting tactics. A hungry or downand-out-on-its-luck mountain lion can cope by being more daring, or by going after riskier prey such as porcupines. Lions and tigers in the extreme may turn to man eating. Similarly, when prey are abundant or the predators are in fine shape, such predators can be more cautious or stick with less risky prey. In going after the young buffalo, were the lionesses exploiting a unique opportunity or exhibiting a degree of desperation? Regardless, the ecology of fear, by permitting predators to ramp up their daring or seek riskier prey, likely stabilizes their population dynamics and forestalls starvation during lean periods. Furthermore, predators may experience a tradeoff between their prey’s catchability and risk of injury. Such is the case for the mountain lion facing deer (hard to catch, little risk of injury) and porcupines (easy to catch, yet high risk of injury). These prey present the predator with a tradeoff between catchability and risk. Such prey may provide a mechanism of coexistence for a predator that is more agile yet injury prone and one that is less agile but brawnier (e.g., cheetah vs. leopard). A tantalizing example may be the mountain lion (agile) and jaguar (brawny) of North and South America preying upon deer and peccaries, respectively. The ranges and diets of these two cat species overlap considerably, but not entirely. The traditional range of the mountain lion corresponds almost exactly to that of deer and guanacos and other South American llama-like antelope – these may be hard to catch but relatively risk-free prey. The original range of the jaguar overlaps almost exactly with that of the collared and white-lipped peccary – easier to catch but more likely to cause injury.

Fear, Behavioral Indicators, and Conservation The games of fear and stealth that go on between predator and prey greatly enhance the behavioral sophistication exhibited in nature. With the elimination of predators, such fear behaviors may decline or become less intense. The prey may become behaviorally less sophisticated and more focused on the business of feeding – recall the behavioral cascades. In domestic animals, we see an evolved reduction or elimination of many fear responses. Goats and sheep likely outforage native antelopes simply because they are less fearful, less distracted from feeding, and less vigilant – we have bred them that way. Retaining the full suite of fear behaviors in prey may be a conservation goal in itself. The absence of predators can reverse the behavioral cascade and result in herbivores being far more destructive on their own resources. The return of wolves to Yellowstone has become iconic. Frightened elk forage less intensively, particularly on riverine willows. The return of willows has returned meanders to the meandering streams, provided more food for moose, nesting and foraging sites for several bird species, and renewed opportunities for beavers. The initiation of these effects could be seen in the rapid and dramatic reappearance of fear behaviors by elk to wolves (Ripple and Beschta, 2004; Manning et al., 2009). Fear behaviors and their role in behavioral cascades may be critical to maintaining or restoring natural areas.

202

Animal Welfare and Conservation j Ecology of Fear

Fearless prey may provide valuable indicators of current habitat quality. Starving animals or animals with greatly reduced feeding prospects will perceive a high marginal value of food (vF/ve) and a reduced sense of future fitness opportunities (F). Both these changes in response to a degrading environment will lower the predation cost of foraging. Such animals will forage their resources more intensively and exhibit fewer fear responses to their predators (such animals will also be easier to catch!). Noting a drop in fear responses even in the face of continued predation threats may provide managers with a valuable indicator of a prey population with a degrading habitat. Fearless prey may also indicate the absence of predators. Mahesh Gurung and Som Ale used the vigilance behaviors, herd sizes, and habitat selection of blue sheep and Himalayan tahr, respectively, to determine the presence or absence of snow leopards from stretches of the Annapurna and Everest ranges of Nepal. They could also use more subtle patterns of fear responses to determine the actual whereabouts of snow leopards in valleys frequented by these cats (Jackson et al., 1996). Fear behaviors can provide an indicator of the long-term presence or absence of predators, and indicators of the imminent presence of a predator.

See also: Animal Welfare and Conservation: Indicators of Pain; Noise Pollution and Conservation; Sickness Behavior in Animals: Implications for Health and Wellness; Slaughter Plants: Behavior and Welfare Assessment. Overview Essays: Welfare Concepts.

References Abrams, P.A., 2010. Implications of flexible foraging for interspecific interactions: Lessons from simple models. Functional Ecology 24, 7–17. Abramsky, Z., Rosenzweig, M.L., Brand, S., 1985. Habitat selection of Israel desert rodents: Comparison of a traditional and new method of analysis. Oikos 45, 79–88. Altendorf, K.B., Laundre, J.W., Lopez Gonzalez, C.A., Brown, J.S., 2001. Assessing effects of predation risk on foraging behavior of mule deer. Journal of Mammalogy 82, 430–439. Brown, J.S., Kotler, B.P., 2004. Hazardous-duty pay and the foraging cost of predation. Ecology Letters 7, 999–1014. Brown, J.S., Laundre, J.W., Gurung, M., 1999. The ecology of fear: Optimal foraging, game theory, and trophic interactions. Journal of Mammalogy 80, 385–399. Hugie, D.M., Dill, L.M., 1994. Fish and game: A game theoretic approach to habitat selection by predators and prey. Journal of Fish Biology 45 (Suppl. A), S151–S169. Jackson, R.M., Ahlborn, G.G., Gurung, M., Ale, S., 1996. Reducing livestock depredation losses in the Nepalese Himalaya. Proceedings of the Seventeenth Vertebrate Pest Conference 1996, 241–247. Laundre, J.W., Hernandez, L., Altendorf, K.B., 2001. Wolves, elk, and bison: Reestablishing the ‘landscape of fear’ in Yellowstone National Park, USA. Canadian Journal of Zoology 79, 1401–1409. Lima, S.L., Dill, L.M., 1990. Behavioral decisions made under the risk of predation – A review and prospectus. Canadian Journal of Zoology 68, 619–640. Manning, A.D., Gordon, I.J., Ripple, W.J., 2009. Restoring landscapes of fear with wolves in the Scottish Highlands. Biological Conservation 142, 2314–2321. Ripple, W.J., Beschta, R.L., 2004. Wolves and the ecology of fear: Can predation risk structure ecosystems. Bioscience 54, 755–766. Rosenzweig, M.L., MacArthur, R.H., 1963. Graphical representation and stability of predator–prey interaction. American Naturalist 97, 209–223. Schmitz, O.J., Beckerman, A.P., Obrien, K.M., 1997. Behaviorally mediated trophic cascades: Effects of predation risk on food web interactions. Ecology 78, 1388–1399. Stephens, D.W., Brown, J.S., Ydenberg, R. (Eds.), 2007. Foraging: Behavior and Ecology. University of Chicago Press, Chicago, IL.

Overview of Animal Training: A Welfare Perspectiveq Melissa Bain, University of California School of Veterinary Medicine, Davis, CA, United States © 2019 Elsevier Ltd. All rights reserved.

Abstract Animal behavior and dog training utilize all quadrants of learning theory, including classical and operant conditioning. Our relationship with dogs has changed over time, as has our understanding of their behavior and our methods of training, which vary from humane and dog-friendly to coercive and based in punishment. This article will address the background of dog training, as well as the aspects related to animal welfare.

Keywords Aggression; Behavior; Classical; Conditioning; Dog; Operant; Punishment; Reinforcement; Training; Welfare

Introduction The field of study of animal behavior is a burgeoning one with a long history, beginning when our ancestors began to hunt animals. They had to identify behavioral patterns that would increase their success in capturing their prey, while allowing them to be safe in the hunt, and this persists today with our use of animals in work. Another important part of our history is when our ancestors learned how to domesticate animals. Through selection over years, humans have been able to select for traits that are useful to people, such as herding, guarding, and hunting. We have also learned over time how to train animals so that they are more likely to perform the behaviors we need them to. Evidence behind some ideologies and methodologies have changed how we approach what we do.

Ethological Views of Animal Behavior Ethology focuses on the study of animals’ behavioral patterns in its environment from a biological perspective. Ethologists study what is termed proximate and ultimate causes of behavior: the “how” and the “why” of behavioral responses. It helps us understand whether a behavior is innate or learned; however, it is not to say that experience also plays a role in what an animal has learned. The behaviors generally associated with feeding, reproduction, territorial defense, and social interaction occur between and among species. Much of the focus is on instinctual or innate behavior patterns (modal or fixed action patterns) that are not a product of learning, and are highly stereotyped motor responses across all individuals in a species. An example is the pouncing of the fox onto a rodent; this behavior is nearly identical across all within that species when encountering the stimulus of the prey. Ethologists argue that animals of different species behave differently because they act within a different set of rules. These rules are partly determined by the animal’s physiological makeup: the form determines the function. One cannot expect a mouse to herd a flock of sheep or a cat to burrow underground. One of the first mentions of dog behavior in modern literature is in Konrad Lorenz’s book, Man Meets Dog. His discussion of how to interact with dogs was to act like a pack leader, including shaking a dog by the scruff of its neck. Another statement in the book describes hitting a dog with a carcass of its prey, so that he does not repeat the offense of killing it. Despite this, it is interesting that he states it is wrong to consider punishment as more effective than the use of rewards when training.

Ethological Views Related to Dog-Human Communication and Relationships Researchers extrapolated hierarchal structure of wolf packs to that of companion dogs amongst each other, as well as with humans. David Mech popularized the term “alpha” regarding interactions in wolf packs (Mech, 1970). He presented the human family as equals with dogs, as if they are in a wolf pack, and owners were to act more “wolf-like”. This led to the popularization of the human construct of dominant gestures on behalf of dog trainers and owners, such as alpha-rolls and scruff shakes. He has since retracted this notion, moving toward classifying wolf pack interactions like that of human family members, where adults and leaders guide the behavior of others in a group (Mech, 2008). Some animals continue to get priority access to valued resources, and dominance-

q

Change History: December 2017. MJ Bain updated the text, references, and relevant websites to this entire article, and added all new figures and tables. This is an update of L.I. Haug and A. Florsheim, Training of Animals, In Encyclopedia of Animal Behavior, edited by Michael D. Breed and Janice Moore, Academic Press, Oxford, 2010, Pages 439-445.

Encyclopedia of Animal Behavior, 2nd edition, Volume 1

https://doi.org/10.1016/B978-0-12-809633-8.20850-9

203

204

Animal Welfare and Conservation j Overview of Animal Training: A Welfare Perspective

submission signaling occurs in these situations. However, our understanding of these interactions has shifted toward cooperative existence. Despite evidence to the contrary, some continue to view dogs continued as part of the human pack. In this paradigm, humans are to be “dominant” over their dog. Unfortunately, the use of these dominance constructs during training remains a persistent theme (Todd, 2018). This philosophy has been a central approach in training programs for many species including dogs, horses, cattle, and elephants. In order to “make” an animal perform the desired behavior, the handler/trainer must establish physical “dominance” over the animal, typically by some form of confrontational interaction, often utilizing a tool. Under the guise of training, handlers give dogs leash corrections with training/choke or prong collars, chase horses with a rope or whip, and subdue elephants with an ankus (hook) (Figs. 1–3). Viewed under the umbrella of operant and respondent learning, these approaches have little merit. A bird bites the hand of someone reaching for it because biting makes the hand go away, not because the bird is dominant. By manipulating the consequences of biting, and of stepping up willingly, the bird’s behavior can be altered with equal efficacy regardless of what we believe the bird’s underlying motivation, it most likely being fearful. Another example is horses that crowd close to handlers during training. They are also often labeled dominant but there is another explanation. “Natural horsemanship” training techniques often use round-penning and lunging exercises to encourage compliance and perceived submission from the horse. The horse is compelled to move away from the trainer by “pressuring”, which is threatening the horse by waving a rope or whip at the animal. A considerable number of horses will kick out at the trainer as they move away, especially during the initial stages of the training. For safety, the trainer typically uses the whip or rope only in a threatening manner when the horse is several feet away, outside the range of the its kick. So when the horse is close to the trainer, the trainer does not generally threaten the horse, but when the horse is several feet away, the trainer feels safe enough to swing the

Fig. 1

Choke collar.

Fig. 2

Prong collar.

Fig. 3

Ankus hook.

Animal Welfare and Conservation j Overview of Animal Training: A Welfare Perspective

205

rope at the horse. In this situation, crowding is not about dominance; it is a simple case of negative reinforcement. The horse learns that staying close to the handler prevents the occurrence of aversive stimuli. Essentially, this training practice can teach the horse to crowd the handler. Behaviorists, on the other hand, attempt to avoid the pitfalls of labels and constructs by focusing on objective, quantifiable measures of behavior. Well-respected organizations have developed position statements on the improper use of the concept of dominance in dog training (The Pet Professional Guild; American Veterinary Society of Animal Behavior, 2014a,b). By objectively observing behavior, behaviorists avoided the trap of trying to assume the animal’s underlying thoughts, feelings, and motivations. Freed from the distraction of worry about an animal’s potential nefarious intentions, a trainer was able to focus on the actual behavior and the consequences driving it, thereby effectively designing a program to manipulate the behavior–consequence contingencies to shape desired behaviors. Despite these advantages, behaviorism has its limits. Animals are not blank slates. As ethologists have demonstrated, instinctive behavior is part of the animal’s repertoire and it can influence operantly learned behaviors. In Breland and Breland’s “The Misbehavior of Organisms”, they describe unrewarded, inappropriate behavior during training (Breland and Breland, 1961). For example, after successful conditioning, raccoons that had learned to pick up a coin and put it in a piggy bank would start to rub the coin between their paws for seconds to minutes at a time, delaying their reward. This “misbehavior” of the raccoons was actually an expression of this food manipulation behavior. The Brelands attempted to use hunger as a motivator thinking that the hungrier the raccoons were, the faster they would seek food reward. In fact, the hungrier they were, the more they exhibited the coin rubbing behavior and further postponed food delivery. While behaviorism has contributed a great deal to our understanding of animal behavior, and has had a significant influence on modern training methodologies, it cannot stand alone in practice, as the Brelands discovered. Instinctual patterns matter. The perspective highlighted respectively by ethologists and behaviorists does not, and should not, argue the existence or importance of the other. Each paradigm has strengths and weaknesses in terms of devising training programs. Further research has demonstrated the differences between wolves and domesticated dogs, in that dogs are more communicative with people, demonstrating the outcomes of domestication. Dogs display more communicative signals and less aggressive and avoidance behaviors in relation to people, and are able to apparently more intuitively follow our direction, compared to similarly-raised wolves (Gácsi et al., 2005; Riedel et al., 2008; Virányi et al., 2008; Udell et al., 2010; Udell and Wynne, 2010). It is via these, and other, studies that we can safely say that domesticated dogs have forged a different relationship with us. More recent studies have shown that dogs and horses can read a person’s facial expressions and respond to pictures of such facial expressions (Albuquerque et al., 2018). The effect of domestication on communication with people is seen in other species. Research with foxes demonstrate that there is some evidence that foxes that were bred for particular tamer behaviors seem to better understand human guidance, shown via pointing toward an object (Hare et al., 2005). The same has been seen in horses, in which horses will orient toward a person’s gaze of attention, perhaps more likely if the person is familiar (Proops and McComb, 2010; Krueger et al., 2011). Certainly some of this is a result of socialization with a particular person. Unlike the studies of dogs and wolves, currently there is no research comparing domesticated horses to wild equids.

Learning “Theory” Animal training is, simply, the manipulation of behavior. Behavior is not the tool with which the animal is trained, but rather the measure of the training procedure: if the animal’s behavior changes, then learning has occurred. There exists an argument that species-specific differences necessitate devising unique training approaches for that particular animal, as though the principles of learning differ between species. Similarly, some trainers believe that we need to assess the animal’s temperament in order to know which techniques to use on a particular animal. Labels such as friendly, nervous, shy, dominant, spooky, etc. may serve to enhance communication between professionals if these are clearly defined; however, there is much misinterpretation of these labels. Additionally these labels themselves do not help us define the actual behaviors that the animal is exhibiting or that we wish to train. Most animal professionals recognize that species show a repertoire of innate behaviors that serve certain functions for the animal. These behaviors are influenced and shaped further by an animal’s experiences. As the Brelands noted, these behaviors can at times interfere with training goals. Nevertheless, these innate behaviors also can be used to enhance the success of training goals. Horse trainers utilizing negative reinforcement for “natural horsemanship” have done this with the manipulation of a horse’s instinct to run when threatened. Escape behaviors (running away) are easy to condition in an animal that is already predisposed to running away. The weakness in the approach is that too few trainers understand the principles of learning. They rely solely on ethological concepts to explain behavior, e.g. the horse runs because it is a prey species and/or is acting submissive, rather than understanding that avoidance behavior is conditioned via negative reinforcement.

Classical Conditioning/Respondent Learning Classical conditioning centers on automatic stimulus – response (S – R) patterns and stimulus – stimulus (S – S) associations. Classical conditioning, also termed respondent learning, is best illustrated by Ivan Pavlov’s famous salivation experiments with dogs.

206

Animal Welfare and Conservation j Overview of Animal Training: A Welfare Perspective

Dogs would begin to salivate (an unconditioned response – UR) at the sight of food (an unconditioned stimulus – US). Pavlov then began ringing a bell (a neutral stimulus – NS) prior to presenting food. After a number of trials of this pairing, the dogs began to salivate (the conditioned response (CR)) at the sound of the bell (now a conditioned stimulus (CS)). Respondent learning is also seen with internal physiologic processes such as hormonal changes. For example, teat stimulation (US) by milking machines elicits oxytocin release (UR) in dairy cattle in milking parlors (previously an NS). Over time, oxytocin release (CR) and milk letdown occur as the cattle begin to approach the milking parlor (CS). Respondent learning is far more complicated than just a simple pairing of two unrelated stimuli. Various studies since the 1960s have demonstrated that not all stimuli are equally associable and that the differences in associability of the CS–US pairings are key in how well respondent learning occurs. In a laboratory setting, researchers studied the effect on male rats of pairing the scent of wintergreen (NS) with sexually receptive female rats (US). Male rats exposed to female rats that are emitting pheromones (US) associated with sexual receptivity experience a reflexive rise in hormones in the bloodstream (UR), which indicates sexual arousal in the male rat. When the rats were exposed to the wintergreen scent (CS) alone, the rise in the measured hormones of the male rats equaled the rise in hormones of the male rats exposed to the sexually receptive female rats (Graham and Desjardins, 1980). Classical conditioning is realistically utilized hand-in-hand with operant conditioning to change an animal’s emotional response, but instead of changing an animal’s behaviors, it is utilized to change an animal’s underlying emotional state. The most frequent situation involves fear-related responses and behaviors that are significant enough to interfere with an operant training paradigm. While fearful behavior is normal in all species, animals exhibiting fear in captive or domestic settings may pose a variety of problems, including difficulty maintaining health and husbandry, destructive escape behavior, or aggression. Trainers often choose to work through the respondent behavior problem first to help clear the way for operant learning. There are well-established procedures to reduce respondent fear via systematic desensitization, counterconditioning, and flooding.

Desensitization and Counterconditioning Systematic desensitization is a process in which a conditioned emotional response (CER) such as fear is extinguished by exposing the animal in a graduated manner to the fear-eliciting stimuli (Fig. 2). When creating a program for systematic desensitization, a stimulus hierarchy is created which ranges from a level that elicits no discernable response to a level that elicits an extreme response. The animal is exposed to the first step in this hierarchy, where it shows a very mild response, such as noticing that the trigger is present, until the mild response is no longer being displayed. Once this occurs, the animal is exposed to the next step on the stimulus hierarchy. This process continues until the animal no longer shows a fear response to the stimulus even at full intensity. Systematic desensitization’s effectiveness is often improved when it is paired with a process called counterconditioning. The animal’s initial CER is replaced with an alternative, competing response by pairing it with an eliciting stimulus that will trigger an opposing emotional or physiologic response. For example, a show cat may need to be bathed and dried prior to a show. If the cat is afraid of the sound of the blow dryer, the sound of the dryer can be paired with a favored food, which elicits pleasure. It’s even more beneficial if the cat never gets this favorite food unless it is going through desensitization and counterconditioning. Counterconditioning occurs only if the new eliciting stimulus triggers a response powerful enough to supersede the original CER. If the cat is extremely afraid of the sound of the dryer, it is very likely that it will not eat in the presence of the blow dryer. Hence, it is often advantageous to pair this counterconditioning with systematic desensitization. By minimizing the intensity of the original CS, the new eliciting stimulus is likely to be salient enough to overcome it.

Flooding Flooding is an exposure-based technique in which an animal is exposed to a fear-evoking stimulus at full intensity until the animal’s fear of that stimulus is extinguished (Fig. 4). The animal is prevented from escaping the stimulus, typically by some form of restraint, hence the term flooding. A dog that is afraid of the sound of fireworks would be exposed constantly to the sound until the fireworks no longer elicited a fearful response. Similarly, a draft horse that is afraid of having a harness on would be placed in the harness until it no longer showed a fearful response (Fig. 5).

Fig. 4

An example of systematic desensitization and counterconditioning.

Animal Welfare and Conservation j Overview of Animal Training: A Welfare Perspective

Fig. 5

207

An example of flooding.

While a favored technique in the past, the use of flooding has largely fallen out of favor with educated behavior professionals because of the problems that can arise. If the session is aborted prior to the abatement of the animal’s fear, the process can actually exacerbate the animal’s fearful response. Additionally, during the session, there is a very real risk of injury to the animal and any human participants if the animal shows severe panic behavior or aggression. Flooding may also lead to a condition known as learned helplessness. The animal learns that its behavior has no effect on its environment; this results in decreased response to the stimulus even when the opportunity to escape the stimulus is returned. Learned helplessness can induce a global suppression on responding such that the animal also fails to respond to other previously learned cues, and has been associated with wider reaching detrimental side effects including physiologic problems.

Operant Conditioning In operant learning, the behavior performed is considered voluntary rather than reflexive, and the behavior is controlled by its consequence. Consequences can be either reinforcing or punishing. Reinforcement increases the future probability of the behavior and punishment decreases the future probability of the behavior. The effectiveness of both reinforcers and punishers is dependent on how closely the behavior and the consequence are paired in time and on the salience of the reinforcer or punisher to the animal.

Reinforcement vs. Punishment Reinforcement and punishment are further classified as positive or negative. These are mathematical terms and indicate whether something is added (positive, þ) or subtracted (negative, ) from the environment within which the behavior is performed (Table 1). With positive reinforcement, a stimulus is added to the system, which results in the behavior being more likely to occur again. For example, if a dog sits, it is given a treat. Negative reinforcement describes a situation where a typically aversive stimulus is removed from the environment when the animal performs the target behavior. A rider applies her spurs to a horse’s sides until the horse begins to move, at which point she ceases the spurring. When a behavior is positively punished, a stimulus is added to the system: if a rat is shocked every time it presses a lever, the rat is less likely to press the lever in the future. Negative punishment refers the removal of a stimulus (including the opportunity for reinforcement) making a behavior less likely to occur. If a sea lion exhibits an aggressive response during a training session, the trainer may step away and ignore the animal for a short period of time. This removes the trainer’s attention as well as the animal’s opportunity to earn further rewards. The use of a no-reward marker can be viewed as a form of negative punishment. Actions and words such as “oops” or “too bad” can signal to the dog they have performed the incorrect behavior. While we focus on rewarding the appropriate behavior, unless an Table 1

Quadrants of operant conditioning Add something

Behavior happens more frequently

Behavior happens less frequently

Remove something

Positive reinforcement

Negative reinforcement

• Give food reward when dog sits • Give food to dog if it is not barking • Run after dog if it steals a shoe

• Remove pressure of hand pushing down on hind end when dog sits

• Owner holds mouth, and lets go of mouth when quiet • Pinches dog’s lips, and when he drops the shoe,

Positive punishment

releases pressure Negative punishment

• Yell at dog for not sitting • Smack dog on nose for barking • Hit dog for stealing shoe

• Remove treat if dog does not sit • Walk away when dog begins to bark • Leave the room when the dog steals the shoe

208

Animal Welfare and Conservation j Overview of Animal Training: A Welfare Perspective

animal is able to avoid all mistakes, learning via immediate feedback with a non-reward marker may lead to more effective learning (Kornell and Vaughn, 2016; Vaughn et al., 2017). Reinforcers and punishers are categorized as primary or secondary. A primary reinforcer or punisher reinforces or punishes innately and does not depend on an association with other stimuli. Examples of primary reinforcers include food, water, reproduction, and relief from environmental stress. Examples of primary punishers include pain and fear-inducing stimuli. Secondary, or conditioned, reinforcers and punishers are dependent on their previous association with primary reinforcers or punishers. For example, the reinforcing effect of a light or sound is dependent upon how frequently that light has been paired with a primary reinforcer such as food. The word ‘no’ acquires punishing properties by pairing it with other unpleasant stimuli hitting the dog or applying a collar correction. Another example is when an owner has historically squirted a dog with a squirt bottle; when she brings out the squirt bottle, the dog anticipates a potential squirt in response to its behavior at the time. Secondary reinforcers are acquired by respondent conditioning (S – S pairings); therefore, their continued strength as reinforcers is dependent on consistent contiguity with a primary reinforcer. In his experiments with pigeons, Herrnstein evaluated the use of secondary reinforcers by examining the relative rate at which pigeons would peck a disk to obtain a secondary reinforce (Herrnstein, 1964). Four pigeons were trained to peck at either of two response keys. Pecking at either key occasionally produced a secondary reinforcer. Then, in the presence of the secondary reinforcer, further pecking occasionally produced the primary reinforcer, food. He determined that the rate at which each pigeon pecked to obtain a secondary reinforcer equaled the relative rate that the secondary reinforcer was paired with a primary reinforcer. As they decreased the number of times the secondary reinforcer was paired with the primary reinforcer, the pigeon worked less to obtain the secondary reinforcer. Secondary reinforcers, such as clickers and whistles, have become very popular in modern animal training, as discussed below.

Reinforcement Schedules Reinforcement schedules influence the variability and persistence of behavior over time. There are several types of reinforcement schedules and each has a distinctive effect on the target behavior. The simplest schedule is termed continuous reinforcement where the target behavior is reinforced each and every time it occurs. Continuous reinforcement leads to the most rapid learning of new behavior, and for this reason, it is the most appropriate choice when teaching new behaviors to animals. Intermittent schedules include duration/interval schedules and ratio schedules. Each of these can be applied on a fixed or variable schedule. Variable schedules produce the most enduring and persistent behavior and are most applicable to animal training outside the laboratory. Duration schedules are most appropriate for long-duration behaviors such as teaching a dog to heel or stay. The dog is reinforced after a variable period of time for successfully staying in position. Another example is a predator waiting for the appearance of prey; the predator is reinforced at variable intervals by the appearance of potential prey. Ratio schedules can be fixed or variable and depend on the number of behaviors offered. Reinforcement occurs after a specified number of correct behaviors are emitted. A bear fishing for salmon in a stream might have to dunk his nose in the water multiple times before he is reinforced by successfully catching a salmon. This is an example of a variable ratio – the number of times the bear has to dunk his head before catching a fish varies.

Conditioning New Behaviors New behaviors can be developed in three general ways: capturing, prompting/luring, and shaping. Each technique has strengths and weaknesses and the best approach will vary with the animal and the specific target behavior.

Capturing Capturing is the process of reinforcing the goal behavior when the behavior is spontaneously offered by the animal. The trainer simply observes the animal, and when the animal displays the behavior, reinforcement is provided. Capturing can work well for behaviors that are a frequent part of the animal’s normal behavioral repertoire and when the behavior occurs in a relatively invariant form. For instance, during breeding season, Atlantic Harbor Seals frequently slap the water with their flippers. Skilled trainers can quickly capture this behavior and put the behavior on cue as this behavior does not vary frequently in the form it is offered. Capturing is not an optimal technique to use if the behavior occurs infrequently, as the resulting reinforcement rate would be so low that the animal would become frustrated or the training process would take too long.

Luring Luring and prompting are two techniques using various stimuli to trigger the appearance of the target behavior or an approximation of the target behavior. Luring, typically using food or a toy, is a popular method for teaching behaviors to dogs. A treat is used to manipulate the dog’s body into a position, by holding the food above its nose, which maximizes the likelihood that the dog will sit.

Animal Welfare and Conservation j Overview of Animal Training: A Welfare Perspective

209

Once the dog sits, it is immediately given the food as a reinforcement. Horse owners often use food to try to lure horses into trailers or stalls. Placing honey or molasses on a horse’s bit encourages the horse to open its mouth to accept the bit.

Prompting Prompts are stimuli used to trigger a desired behavior. For example, a trainer can prompt a dog to turn its head to one side by making a noise off in the desired direction and reinforcing the animal when the head turn occurs. A cat owner can prompt the cat to wipe at its face by placing a small piece of tape on the cat’s nose. Once the behavior is triggered by the lure or the prompt, the trainer then reinforces the behavior. Over time, the lure or prompt is gradually faded, and the behavior is transferred to a different cue (discriminative stimulus) so the behavior can be elicited without the lure or prompt. A distinct advantage of luring and prompting is that behavior can be generated quickly, often within minutes. This is also a relatively humane and compulsion-free technique; however, one disadvantage of prompting is that the animal may habituate to the prompt stimulus. For example, after a few repetitions, the dog may no longer orient to the sound used to prompt a head turn. The major disadvantage of using lures and prompts is the process of fading the stimulus. Lures are generally easier to fade than prompts, but in both cases, the process can take time. Another criticism of luring is that the learning process is trainer-driven, rather than animal-driven. Animals trained only by luring may offer little spontaneous behavior, or behavioral experimentation, during training sessions. This can become problematic if the trainer wishes to develop a behavior that is outside the animal’s typical behavioral repertoire. This problem can be avoided by using the last method, shaping, because behavioral experimentation and variability are essential to this technique.

Shaping Shaping, also called differential reinforcement of successive approximations, is a process where a behavior is developed by initially reinforcing any behavior that remotely resembles the target behavior. Over subsequent training sessions, the final target behavior is shaped by progressively reinforcing behaviors that more closely resemble the target behavior. A bird can be trained to elevate its wings for examination by first reinforcing the bird for any tiny lift of the wing away from the body. When this behavior is being offered with relative reliability, reinforcement is withheld for this criteria and the bird is reinforced only if it lifts its wing away from the body and begins to slightly extend the carpal joint. Wing lifting is progressively shaped by only reinforcing greater extensions of the wing until the bird is lifting the wing completely away from the body and extending the wing out to its full span. The smaller the approximations, the more seamless the learning process will be. When shaping criteria are carefully planned, the animal often offers a new level of behavior before the trainer changes reinforcement criteria. Shaping is most effective when done with a bridge, or secondary reinforcer. This is typically a unique sound such as a click or whistle that has been previously paired with a primary reinforcer such as food. The target behavior is ‘marked’ with the bridge signal, and then a primary reinforcement is delivered to the animal. The bridge signal increases the accuracy of the reinforcement process particularly for rapid behaviors or situations where the animal is not close to the trainer, thereby making timely delivery of a primary reinforcer difficult. Results of a study showed that it may be less challenging for a novice owner to teach their dog a cue via clicker training when compared to using just food (Feng et al., 2018). As seen with Herrnstein’s pigeons, the effectiveness of a secondary reinforcer relies on its consistent association with a primary reinforcer. If a bridge is used in the absence of a primary reinforcer, it loses its effectiveness as a secondary reinforcer and marking stimulus. Shaping produces the most learner participation as the process is entirely learner driven: the trainer does not prompt the animal in any way, but rather reinforces behaviors that the animal offers spontaneously. Shaping does require the trainer to have excellent observation skills and knowledge of the evolution of the behavior being shaped. As with other training methods, shaping can be frustrating to the animal if the reinforcement rate is too low or the trainer’s criteria are unclear. Shaping can be combined successfully with prompts in some circumstances. For example, hoof care is an important part of equine husbandry, so it is essential that the horse reliably lift its feet when cued to do so. A shaping procedure combined with a prompt would begin with the trainer gently squeezing the tendons of the horse’s leg. When the horse shifts its weight even the slightest degree to the off foreleg, the trainer reinforces the horse by removing the prompt and offering food reinforcement (a combination of negative and positive reinforcement). Over trials, the horse is reinforced only for greater weight shifts, then lifting the foot slightly, etc. Some new research has looked into the efficacy of social learning, also known as “Do As I Do” (Fugazza and Miklósi, 2015). While both utilize positive reinforcement to reward the wanted behavior, there is some evidence that social learning may be more effective.

Manipulating Antecedents to Change Behavior Antecedents are learned signals for the behavior-consequence contingency to follow. Opening a food bin may become the antecedent for a horse to kick the stall walls because the kicking has been reinforced with the delivery of food. The strength of the antecedent to cue a particular behavior is related to the strength of the reinforcer that follows the behavior. There are three ways to

210

Animal Welfare and Conservation j Overview of Animal Training: A Welfare Perspective

manipulate antecedents to alter behavior patterns: adding/removing a cue; changing the setting events; and strengthening/weakening the motivation for the behavior, termed establishing operations. Antecedents can be changed or removed to alter the behavior/consequence pairing and prevent the elicitation of the problematic behavior. For example, some dogs will bark and run to the door when they hear the doorbell in anticipation of people entering the home. If the doorbell is turned off or guests enter the home without ringing the doorbell, the dog may be less likely to bark. The context, conditions, or situational influences that affect behavior are often some of the easiest antecedents to change. Horses that weave when placed in a stall can be moved to a pasture with other horses. Cats that urinate outside of the box only when the box is dirty can have their boxes cleaned more regularly. If a cat bites when it is petted while on someone’s lap, the owner can refrain from letting the cat on laps or from petting the cat once it is there. Motivating or establishing operations are antecedents that temporarily alter the effectiveness of consequences. For example, certain toys may be highly motivating to a dog if it rarely has access to them but less so if the dog has unlimited access to the toys all the time. A horse may more readily go back into its stall and rest quietly once it has had a long workout and time at pasture with social interaction with other horses. Similarly, food reinforcers are more effective if the animal is hungry, so training will be most effective if the training session is schedule just before the animal’s next mealtime.

Manipulating Consequences to Change Behavior While antecedent behavior-change strategies are preventative solutions, consequence changes rely on learning. Manipulating consequences necessitates that the trainer determines what reinforces a particular behavior. Once the reinforcement is identified, the trainer can withhold or eliminate the reinforcement. When the association between the behavior and its consequence is severed, the number of times the behavior is offered decreases and eventually disappears or returns to baseline. This process is known as extinction. Extinction can be problematic as a strategy to change behavior. Current studies on extinction indicate that extinction does not destroy the original learning but instead generates new learning that is very dependent on context. A change of context after extinction can cause a return of the initial CR, termed the renewal effect. Some behaviors may be maintained by internal reinforcement processes (e.g., physiologic changes); therefore, the reinforcement cannot be withdrawn easily. Other behaviors may have external reinforcements that are very difficult or even impossible to control. For example, many dogs receive substantial reinforcement for chasing small and birds. Because the dog must be let out into the yard for elimination each day, the dog is likely to continue to receive some level of reinforcement for this behavior even if the dog is kept on leash. Extinction can be a very slow process and the problematic behavior may recover over time. Cessation of the behavior is often preceded by an extinction burst, a sharp increase in the frequency and intensity of the problematic behavior. This can be a very frustrating process for those working with animals. Many individuals inadvertently reinforce the animal during this burst, thereby creating an even stronger response. Extinction can also result in frustration-related behaviors, such as displacement or attentionseeking behaviors. Another behavior change strategy is differential reinforcement of alternative behaviors. The alternative behavior selected should be one that replaces the function of, and generally incompatible with, the problem behavior. The new behavior should receive more reinforcement than the problem behavior. If a chimpanzee charges the feeding door to receive its food, the chimpanzee would be trained to sit away from the door. The animal would be highly reinforced for this behavior such that he receives more reinforcement for sitting away from the door than he does for charging at the door when it opens. The Matching Law states that organisms will apportion their behavior in accordance with how much reinforcement each behavior receives. If a pig roots out high value reinforcements in a specific area of the yard, the pig is significantly less likely to root in other parts of the yard where it receives less reinforcement. The alternative behavior should ideally be something the animal already knows how to perform. It will be easier to replace a problem behavior with a behavior that the animal already knows well, rather than trying to teach the animal a brand new response. If a dog is highly proficient with sitting on cue, but is just learning lie down, then sit is the more appropriate choice as an alternative behavior for lunging at people on walks.

Welfare Concerns in Dog Training Definitions of welfare vary; however, a commonly utilized framework is the Five Freedoms. These include the freedom from fear and distress, freedom from pain and suffering, and the freedom to express normal behaviors. They are useful in framing the effects of our treatment of animals. However, an understood limitation is that they do not capture the entire complexity of current knowledge related to animal welfare. The general public has little understanding of learning theory, and can be swayed by incorrectly and deceivingly-used words like dog-friendly, humane, balanced, and reward-based. Popular dog training books often have erroneous information (Browne et al., 2017). Despite our close relationship with dogs, most people report utilizing some sort of positive punishment when training their pet dog (Hiby et al., 2004; Arhant et al., 2010; Rooney and Cowan, 2011). Many well-respected organizations have developed position statements promoting humane training techniques, and highlighting the concerns around utilizing aversive techniques (American Veterinary Society of Animal Behavior, 2014a,b; The Pet Professional Guild, 2016, 2017).

Animal Welfare and Conservation j Overview of Animal Training: A Welfare Perspective

211

Training Methodologies For years, animals have been trained utilizing primarily positive punishment and negative reinforcement. These methods do have some adaptive value, as animals can learn to avoid something that is potentially dangerous. Often these triggers can be taught to be avoided through conditioning a response to a previously neutral stimulus. One such example is the utilization of electronic, or shock, collars. Most collars on the market have distinct levels of electronic stimulation, or shock, along with a tone that precedes the shock. As time progresses, dogs begin to associate the tone with a resultant shock; therefore, that tone now takes on the characteristics of the shock. Unfortunately the owners that select trainers utilizing coercive techniques usually do not understand the scientific principles behind operant and classical conditioning, including punishment. They may find it difficult to stop a trainer performing a certain training technique, such as helicoptering (swinging a dog in a circle by its leash and collar), because they just paid a professional dog trainer to help them with their dog’s problem behavior. A lot of these trainers come highly recommended, either from their friends or from flashy websites touting their training prowess. However, since trainers are currently not required to obtain a dog training license in any state, there is no governmental oversight of their training techniques, abilities, or credentials. There have been a sizable number of published research papers identifying deleterious outcomes when using positive punishment and/or negative reinforcement in training an animal. The methods utilized in these studies varied, from owner surveys, to direct observations of behavior, to interventional studies. Over the past number of years, there has been both a growing body of research on the negative effects from using coercive training methods, including increased aggression and anxiety (Hiby et al., 2004; Blackwell et al., 2008; Herron et al., 2009; Arhant et al., 2010; Casey et al., 2013, 2014; Cooper et al., 2014) One example is a study of 3897 dog owners, where researchers found that the use of positive punishment and negative reinforcement was related to an increased risk for aggression toward people, when compared to the use of positive reinforcement and negative punishment (Blackwell et al., 2008). Problem behaviors are commonly identified as a risk for either euthanasia or relinquishment of a pet dog, often related to the type of training utilized (Kwan and Bain, 2013). Owners of dogs visiting a veterinary behaviorist were more likely to select euthanasia of their dog if they had used punishment-based training methods or previously consulted with a non-veterinary behaviorist or trainer (Siracusa et al., 2017). Another study, a meta-analysis of appropriately-designed studies, was performed on 13 research papers on reasons for relinquishment and/or euthanasia of dogs. Behavioral problems were identified as a reason for relinquishment in eight of nine studies, and as a reason for euthanasia in all five studies (Lambert et al., 2015). As an example of an observational study, with observations of 33 military dog-handler dyads performing standardized exercises, dogs that received more aversive stimuli (either positive punishment or negative reinforcement) were more distracted and showed worse performance compared to dogs that received less of the aversive stimuli. Additionally, the dogs receiving more aversive training showed a lower body posture after receiving an aversive stimulus, which can be interpreted as fearful behavior. And dogs receiving more coercive training were no more successful in their performance (Haverbeke et al., 2008). Additional research has shown a correlation between the use of shock collars and poorer training outcomes and stress-related signs (Schilder and van der Borg, 2004; Blackwell et al., 2012; Arnott et al., 2014). The European Society of Veterinary Clinical Ethology evaluated the research to help them shape their position statement against the use of shock collars in dog training (Masson et al., 2018). Not only have they come up with a position statement against the use of shock collars, some countries, including Austria, Germany, Denmark, Norway, Wales, and Scotland, have banned the use of these tools. While owners and trainers often incorrectly focus on training a behavior more quickly, evidence has shown that dogs learn no faster via punishment-based methods. In a study of owners of 326 dogs, it was revealed that the highest obedience scores were reported by owners who used only reward-based training, followed by those who used a combination of reward and punishmentbased methods, and lastly by those who used only punishment-based methods (Hiby et al., 2004). Few interventional studies have been done evaluating training methods in dogs. In one study of handlers utilizing shock collars, laboratory-bred beagles were divided into three groups: one group received a shock precisely when they grabbed a prey dummy; another received a shock if they failed to respond to a recall while hunting the prey dummy; and the third group received arbitrary and unpredictable shocks (Schalke et al., 2007). The second and third group had roughly a doubling and tripling, respectively, of absolute and relative salivary cortisol levels, whereas the first group had an increase of 22% and 31% of absolute and relative levels. This led the researchers to conclude that, when the dogs were able to clearly associate the shock with their action, such as touching the prey dummy, and consequently were able to predict and control the stressor, they did not show considerable or persistent stress indicators. However, when they were not able to predict whether or not a shock was forthcoming, there was an apparent increase in stress. Compared to the utilization of aversive training, the utilization of positive-reinforcement methods has been correlated to better trained dog (Rooney and Cowan, 2011). It also has been correlated to less stress-related behaviors (Deldalle and Gaunet, 2014).

Summary Animal behavior and training, especially as seen through the lens of dog training, can be fraught with potential concerns. By being educated in the science surrounding ethology and training one can avoid some of the common hazards that can negatively affect an animal’s welfare.

212

Animal Welfare and Conservation j Overview of Animal Training: A Welfare Perspective

See also: Animal Welfare and Conservation: Punishment; Stress, Health and Social Behavior. Landmark Studies: Frustration in Hens. Methodology: Assessment of Welfare and Needs. Overview Essays: Welfare Concepts.

References Albuquerque, N., et al., 2018. Mouth-licking by dogs as a response to emotional stimuli. Behavioural Processes 146, 42–45. American Veterinary Society of Animal Behavior, 2014a. Dominance position statement. Retrieved May 20, 2018, from https://avsab.org/wp-content/uploads/2018/03/Dominance_ Position_Statement_download-10-3-14.pdf. American Veterinary Society of Animal Behavior, 2014b. Punishment position statement. Retrieved May 20, 2018, from https://avsab.org/wp-content/uploads/2018/03/Punishment_Position_Statement-download_-_10-6-14.pdf. Arhant, C., et al., 2010. Behaviour of smaller and larger dogs: Effects of training methods, inconsistency of owner behaviour and level of engagement in activities with the dog. Applied Animal Behaviour Science 123 (3), 131–142. Arnott, E.R., et al., 2014. Environmental factors associated with success rates of Australian stock herding dogs. PLOS ONE 9 (8), e104457. Blackwell, E.J., et al., 2008. The relationship between training methods and the occurrence of behavior problems, as reported by owners, in a population of domestic dogs. Journal of Veterinary Behavior: Clinical Applications and Research 3 (5), 207–217. Blackwell, E.J., et al., 2012. The use of electronic collars for training domestic dogs: Estimated prevalence, reasons and risk factors for use, and owner perceived success as compared to other training methods. BMC Veterinary Research 8 (1), 93. Breland, K., Breland, M., 1961. The misbehavior of organisms. American Psychologist 16 (11), 681–684. Browne, C.M., et al., 2017. Examination of the accuracy and applicability of information in popular books on dog training. Society & Animals 25 (5), 411–435. Casey, R., et al., 2013. Inter-dog aggression in a UK owner survey: Prevalence, co-occurrence in different contexts and risk factors. Veterinary Record 172 (5), 127, 127. Casey, R.A., et al., 2014. Human directed aggression in domestic dogs (Canis familiaris): Occurrence in different contexts and risk factors. Applied Animal Behaviour Science 152, 52–63. Cooper, J.J., et al., 2014. The welfare consequences and efficacy of training pet dogs with remote electronic training collars in comparison to reward based training. PLOS ONE 9 (9), e102722. Deldalle, S., Gaunet, F., 2014. Effects of 2 training methods on stress-related behaviors of the dog (Canis familiaris) and on the dog–owner relationship. Journal of Veterinary Behavior: Clinical Applications and Research 9 (2), 58–65. Feng, L.C., et al., 2018. Is clicker training (clicker þ food) better than food-only training for novice companion dogs and their owners? Applied Animal Behaviour Science (in press, available online) Available at: https://doi.org/10.1016/j.applanim.2018.04.015. Fugazza, C., Miklósi, Á., 2015. Social learning in dog training: The effectiveness of the Do as I do method compared to shaping/clicker training. Applied Animal Behaviour Science 171, 146–151. Gácsi, M., et al., 2005. Species-specific differences and similarities in the behavior of hand-raised dog and wolf pups in social situations with humans. Developmental Psychobiology 47 (2), 111–122. Graham, J., Desjardins, C., 1980. Classical conditioning: Induction of luteinizing hormone and testosterone secretion in anticipation of sexual activity. Science 210 (4473), 1039–1041. Hare, B., et al., 2005. Social cognitive evolution in captive foxes is a correlated by-product of experimental domestication. Current Biology 15 (3), 226–230. Haverbeke, A., et al., 2008. Training methods of military dog handlers and their effects on the team‫׳‬s performances. Applied Animal Behaviour Science 113 (1), 110–122. Herrnstein, R.J., 1964. Secondary reinforcement and rate of primary reinforcement. Journal of the Experimental Analysis of Behavior 7 (1), 27–36. Herron, M.E., et al., 2009. Survey of the use and outcome of confrontational and non-confrontational training methods in client-owned dogs showing undesired behaviors. Applied Animal Behaviour Science 117 (1–2), 47–54. Hiby, E.F., et al., 2004. Dog training methods: Their use, effectiveness and interaction with behaviour and welfare. Animal Welfare 13, 63–69. Kornell, N., Vaughn, K.E., 2016. How retrieval attempts affect learning: A review and synthesis. In: Ross, B.H. (Ed.), Psychology of Learning and Motivation, 65. Elsevier, pp. 183–215. Krueger, K., et al., 2011. Horses (Equus caballus) use human local enhancement cues and adjust to human attention. Animal Cognition 14 (2), 187–201. Kwan, J.Y., Bain, M.J., 2013. Owner attachment and problem behaviors related to relinquishment and training techniques of dogs. Journal of Applied Animal Welfare Science 16 (2), 168–183. Lambert, K., et al., 2015. A systematic review and meta-analysis of the proportion of dogs surrendered for dog-related and owner-related reasons. Preventive Veterinary Medicine 118 (1), 148–160. Masson, S., et al., 2018. Electronic training devices: Discussion on the pros and cons of their use in dogs as a basis for the position statement of the European Society of Veterinary Clinical Ethology. Journal of Veterinary Behavior: Clinical Applications and Research 25, 71–75. Mech, D.L., 1970. The Wolf: The Ecology and Behavior of an Endangered Species. The Natural History Press, Garden City, New York. Mech, D.L., 2008. Whatever happened to the term Alpha Wolf? International Wolf 18 (4), 4–8. Proops, L., McComb, K., 2010. Attributing attention: The use of human-given cues by domestic horses (Equus caballus). Animal Cognition 13 (2), 197–205. Riedel, J., et al., 2008. The early ontogeny of human–dog communication. Animal Behaviour 75 (3), 1003–1014. Rooney, N.J., Cowan, S., 2011. Training methods and owner–dog interactions: Links with dog behaviour and learning ability. Applied Animal Behaviour Science 132 (3), 169–177. Schalke, E., et al., 2007. Clinical signs caused by the use of electric training collars on dogs in everyday life situations. Applied Animal Behaviour Science 105 (4), 369–380. Schilder, M.B.H., van der Borg, J.A.M., 2004. Training dogs with help of the shock collar: Short and long term behavioural effects. Applied Animal Behaviour Science 85 (3–4), 319–334. Siracusa, C., et al., 2017. Dog- and owner-related risk factors for consideration of euthanasia or rehoming before a referral behavior consultation, and for euthanizing or rehoming the dog after the consultation. Journal of Veterinary Behavior: Clinical Applications and Research 22, 46–56. The Pet Professional Guild. Dominance theory in animal training. Retrieved May 20, 2018, from https://www.petprofessionalguild.com/DominanceTheoryPositionStatement. The Pet Professional Guild, 2016. Position statement on the use of shock in animal training. Retrieved May 20, 2018, from https://www.petprofessionalguild.com/ shockcollars. The Pet Professional Guild, 2017. Position statement on the use of choke and prong collars. Retrieved May 20, 2018, from https://www.petprofessionalguild.com/ chokeandprongcollarpositionstatement. Todd, Z., 2018. Barriers to the adoption of humane dog training methods. Journal of Veterinary Behavior: Clinical Applications and Research 25, 28–34. Udell, M.A., et al., 2010. What did domestication do to dogs? A new account of dogs‫ ׳‬sensitivity to human actions. Biological Reviews 85 (2), 327–345. Udell, M.A., Wynne, C.D., 2010. Ontogeny and phylogeny: Both are essential to human-sensitive behaviour in the genus Canis. Animal Behaviour 79, e9–e14.

Animal Welfare and Conservation j Overview of Animal Training: A Welfare Perspective

213

Vaughn, K.E., et al., 2017. Retrieval attempts enhance learning regardless of time spent trying to retrieve. Memory 25 (3), 298–316. Virányi, Z., et al., 2008. Comprehension of human pointing gestures in young human-reared wolves (Canis lupus) and dogs (Canis familiaris). Animal Cognition 11 (3), 373–387.

Relevant Websites www.acaw.org–American College of Animal Welfare. www.dacvb.org–American College of Veterinary Behaviorists. www.avsab.org–American Veterinary Society of Animal Behavior. www.animalbehavior.org–Animal Behavior Society. http://indoorpet.osu.edu–Indoor Pet Initiative.

Punishmentq Keith Jensen, Queen Mary University of London, London, United Kingdom Michael Tomasello, Max Planck Institute for Evolutionary Anthropology, Leipzig, Germany © 2019 Elsevier Ltd. All rights reserved.

Abstract Animals can use punishment as a means to change the behavior of others. Punishment can be done for selfish ends with no regard for how the target of the act is affected. On the other extreme, it can benefit others in a society and be motivated by its effects on others. Altruistic punishment, third-party punishment, and norm enforcement are special cases of punishment that can maintain cooperation, and these may not have analogs in animals other than humans. More socially sophisticated forms of punishment will require more flexible and complex cognitive processes. Of particular interest are social (other-regarding) preferences, since these may have allowed the evolution of the large-scale non-kin cooperation seen only in humans. However, little is known about the cognitive mechanisms underlying punishment in other animals.

Keywords Altruistic punishment; Cooperation; Negative reciprocity; Norm enforcement; Other-regarding preferences; Punishment; Social preferences; Spite; Third-party punishment

What is Punishment? In everyday parlance, punishment occurs when one individual performs an act with the goal of discouraging another individual from engaging in a particular behavior. This may involve doing something aversive to the offender or withholding something of value. For example, owners might punish their dogs for chewing on the furniture by swatting them on the rump, and parents might punish their toddlers for drawing on the walls by withholding dessert. This commonsense view of punishment involves intent – the punisher has the goal of modifying the behavior of another individual. Importantly, it involves a change in behavior in the future; it is not simply reactive or malicious. This future orientation is consistent with the biological view of punishment in which an individual reacts to harmful behavior by reducing the fitness of the instigator and thereby decreases the likelihood of future harm. Inflicting harm on another at a personal fitness cost with no direct fitness benefit (biological spite) is not likely evolve, except under very restrictive conditions leading to indirect fitness benefits (kin selection). However, by being reciprocal – inflicting a harm for a harm done – the punisher gains net fitness benefits in the future. Biologists are agnostic about the cognitive mechanisms of the punisher; intentions are not required. There are two schools of thought on punishment in psychology. For learning theorists, punishment is any stimulus that causes a decrease in an organism’s behavior. In this vein, a rose thorn is a punisher because it decreases the likelihood that the pricked individual will pick roses in the future. Roses, of course, do not have goals and are cognitively uninteresting, nor does the flower picker need any understanding of intentions. Social psychologists, on the other hand, restrict punishment to cases where impartial, outside observers mete out corrections on the basis of normative principles. This requires sophisticated cognitive abilities, as well as social norms and rational, rather than emotional, motives. The social psychological view not only restricts punishment to humans, but also it does not encompass all – nor arguably much of – human punishment in the everyday sense of the word. As for other animals, it is likely that their punishment lies along a continuum between simple reflexive responses to harmful events and intentional, impartial, norm enforcement.

Punishment Shapes the Social Environment Punishment is a powerful means by which an individual can shape its social environment. At the simplest level, punishment achieves selfish benefits. Punishment is distinguished from mere aggression and avoidance by the delay in benefits. Aggression and avoidance, though potentially costly, provide immediate benefits. For instance, a large male that sexually harasses a female leaves behind offspring, and subordinate animals begging from dominant individuals will receive scraps of food. From an evolutionary perspective, the delay in benefits involved in punishment results in both the punisher and the target suffering costs, making the behavior – at least at the time that it is performed – a form of biological spite. For example, male hamadryas baboons (Papio hamadryas) will threaten or attack females that stray from the harem, and dominant male chimpanzees (Pan troglodytes) will attack rivals as well as supporters of rivals. Ultimately, for punishment to evolve, it must eventually benefit the actor, either directly or

q

Change History: June 2018. Donald Broom inserted the references supplied by the authors. This is an update of K. Jensen and M. Tomasello, Punishment, In Reference Module in Life Sciences, Elsevier, 2017.

214

Encyclopedia of Animal Behavior, 2nd edition, Volume 1

https://doi.org/10.1016/B978-0-12-809633-8.01093-1

Animal Welfare and Conservation j Punishment

215

indirectly (i.e., through kin). As with positive reciprocity (often called reciprocal altruism), negative reciprocity is costly at the time that it is performed but yields benefits in the future. Because of delayed benefits, punishment is ‘return benefits spite.’ The functions of punishment in animal societies include achieving and maintaining social dominance, deterring cheaters and parasites, as well as disciplining and coercing offspring and sexual partners. In all these cases, the punisher receives delayed, direct fitness benefits from its actions.

Punishment and Cooperation Retaliation against personally harmful behaviors (negative reciprocity) is a particularly interesting form of punishment (Sigmund, 2007). While the punisher still benefits from its actions, other individuals in the group might also benefit from changes in the target’s behavior. For instance, a juvenile that is attacked when it tries to take food from one adult might learn to not take food from others. The role of punishment in maintaining cooperation is of great interest, because the benefits to the punisher are not so transparent and the consequences are far reaching (Boyd and Richerson, 1992). The reasons why animals cooperate – particularly when they make costly sacrifices for the benefit of others – is one of the longest-standing and most important questions in evolutionary biology. Individuals risk being exploited by group members: free-riders reap the benefits of cooperation without sharing the costs and can cause the collapse of cooperation. Punishment can be an important force for the maintenance of cooperation (Clutton-Brock and Parker, 1995). In animal societies, one context in which punishment would appear important for maintaining cooperation is in cooperative breeders. In cooperative breeders, breeding individuals will coerce their offspring into forfeiting reproduction to aid in the care of the dependent offspring. In some cases such as social insects, the nonreproductive workers will punish egg-laying cheats (biologists refer to this as policing since the punishers do not benefit directly). Cooperative breeding is a fascinating, but limited form of cooperation; punishers either directly benefit or provide benefits to kin (especially true in the social insects), and helpers remain because the costs of being exploited are less than the risks of leaving. Surprisingly, despite temptations to free-ride, examples of punishment of helpers that fail to work are rare. Punishment in cooperative breeders functions as a coercive strategy. Outside of cooperative breeding, examples of punishment to maintain cooperation in non-humans are few, and these are equivocal. In one field experiment designed to test for tit-for-tat -city, African lions (Panthera leo) did not punish laggards that failed to reciprocate the potentially hazardous behavior of approaching a perceived threat. In chimpanzees, there was one observation in captivity of a male attacking a group member who failed to provide support in a conflict, the inference being that the attacker punished the recipient for uncooperative behavior. Alternatively, in a systematic observational study of reciprocity and aggression in chimpanzees, failure to reciprocate grooming or support did not lead to punishment. In the only experiment designed specifically to probe punishment of non-cooperative behavior in chimpanzees, captive individuals had their food stolen from them by a conspecific. They then punished the theft by collapsing the table on which the food sat, thereby preventing the thief from eating. Chimpanzees were punitive in that they were negatively reciprocal (vengeful). It is not clear that they were punishing a non-cooperative behavior (theft) for the purpose of maintaining a cooperative relationship, however (de Waal and Luttrell, 1988). In fact, theft increased over time while punishment declined, suggesting that punishment failed to deter non-cooperative behavior.

Altruistic Punishment A special form of punishment has been described recently in the experimental economics literature. Altruistic punishment is the punishment of free-riders without a return in benefits for the punisher. Because the punisher does not directly benefit from an increase in cooperation, altruistic punishment has been claimed to pose a potential challenge to natural selection at the individual level. Theoretical models and experimental economic studies conducted on humans have shown that in the absence of punishment, the level of cooperation in groups declines with repeated interactions. In fact, the standard prediction in economic theory is that in one-shot encounters, pairs or groups of players should not cooperate at all: free-riders fare better in a population of cooperators by receiving the benefits of cooperation without the costs. As a result, cooperation is driven to extinction. However, if individuals are allowed to punish others – even if this imposes an additional cost on the punishers – cooperation can be maintained as a stable strategy. Punishment is more effective in maintaining cooperation than is direct (positive) reciprocity because the individual costs of punishment decline as the number of free-riders declines, whereas the cost of cooperation rises as the number of cooperators increases. As well, the threat of punishment can be a sufficient deterrent to cheating. The nature of altruistic punishment remains contentious. The reason for the debate is that the experiments that elicit altruistic punishment are artificial, and there are some questions whether the participants truly play the games as if they are anonymous and one-shot. However, studies outside of economics, as well as examples of people intervening on behalf of others, lend credence to the notion of altruistic punishment. While there have been many studies of altruistic punishment in humans, the sole experimental attempt to find something like altruistic punishment in animals is an adaptation of the widely used economic experiment, the ultimatum game. The ultimatum game involves a division of a resource between two players. The first player proposes a division which the second player can accept. But if the second player rejects it, both get nothing; the responder pays a cost to punish the proposer for his offer. This experiment has been run hundreds of times in many human cultures, and contrary to economic models of rational self-interest, people routinely pay this punishment cost. In a reduced form ultimatum game played by chimpanzees, the apes behaved like selfinterested maximizers, conforming to the predictions of standard economic theory (Jensen et al., 2007a). Proposers did not choose equal divisions, and responders did not reject any nonzero offer; that is, chimpanzees did not pay a cost to prevent a conspecific

216

Animal Welfare and Conservation j Punishment

from getting more of a resource. This finding is consistent with the punishment experiment described earlier in which they did not react especially to the outcome of another individual eating as long as it was not stolen.

Third-Party Punishment Clear evidence for altruistic punishment would require impartial, third-party punishment, where the punisher has no direct stake in the matter. (Previous examples were personal, or second-party, punishment.) As stated earlier, third-party intervention is an essential condition for punishment according to social psychologists. Humans do engage in third-party punishment, for instance, by policing all kinds of social norms that are violated – from smoking in restaurants to harming children. Some primates will intervene in disputes and some social insects will destroy eggs not laid by the queen. Biologists use the term policing for this kind of behavior, though it is fundamentally different from the human case. The examples of policing in animal societies are typically executed by dominant individuals who have their own interests at stake, and there is no evidence for punishment by disinterested individuals following the disruption created by a conflict in a group. Experimental evidence for third-party punishment in animals is needed to help resolve the question of altruistic punishment in non-humans.

Norm Enforcement A special case of third-party punishment is norm enforcement. In this case, the simple violation of social norms, such as a teenager wearing torn jeans and a T-shirt to a wedding, will evoke punishment from others, such as expulsion from the wedding or disparaging gossip. The ‘altruism’ refers to the fact that the punisher’s behavior benefits not the offender, but others in the group, for example by maintaining certain behavioral traditions as markers of group identity (Fehr and Gächter, 2002). There may be indirect benefits to the punisher, including a reputational benefit for being an altruistic punisher (attracting more cooperators and deterring cheats). But both third-party punishment and norm enforcement may also benefit the punisher indirectly by benefiting the group, as described earlier. Punishment can maintain any behavior – such as conformity to styles of dress – and not just cooperation, though the selective pressures favoring cooperation are more obvious than for wedding attire (Fehr and Fischbacher, 2003) There have been a few claims for norm enforcement, or punishment of rule violation, in animals other than humans. One example was suggested in rhesus macaques (Macaca mulatta): higher ranking individuals aggressed against lower ranking conspecifics that failed to give food calls when the latter found food. The interpretation was that non-callers were punished for deception (Hauser, 1992). However, a plausible alternative was provided in a study on white-faced capuchins (Cebus capucinus): individuals called to signal ownership of food resources and those that failed to call were more likely to enter into conflict with higher ranking individuals, since ownership was not clear. An observation in captive chimpanzees saw two individuals that failed to come into the enclosure at night in a timely manner (and hence delayed the feeding of the group); they were attacked the next morning by the others when reunited. And an example from the field reported a violent attack by eight individuals on single male who had failed to conform to a social rule such as not exhibiting species-typical submissive behavior. It is not clear, however, that these acts of aggression were really cases of punishment to enforce specific behaviors or were cases of redirected aggression, dominance behavior, or something else. Experiments modeled on such observations will be needed before we can attribute rule enforcement to nonhuman animals.

Social Preferences and Punitive Motives Given the role of punishment in societies, it is important to know why individuals punish others, namely their intentions. As an example for why intentions are important, consider the difference between first-degree murder (intentional killing) and involuntary manslaughter (unintentional killing). From the victim’s perspective, the outcome is the same, but psychologically (and legally), these are very different acts. Of particular importance for punishment are social preferences, namely whether the punisher has the intention or motivation of causing harm in other individuals (Silk, 2008).

Self-Regarding Preferences Self-regarding preferences are nonsocial – the goals involved are purely personal with no regard for the consequences for others. Any effects on other individuals are byproducts. With regard to punishment, any change in future behavior that subsequently benefits the punisher is fortuitous. The goal of the individual is only that the offender immediately refrain from harming it, or that it remove itself from the harmful situation. Effects on the target of punishment – and on others that may be affected (e.g., through a decrease in uncooperativeness) – do not motivate the punishers choices. In other words, the individual is indifferent to the consequences of its actions apart from immediate, personal outcomes. Much aggression and harm avoidance may be seen in this light and would not be seen as punishment in the commonsense use of the term: the aggressor’s motivation is not that the offender learn to refrain from doing something in the long term, but only that it stop it now. However, in terms of delayed costs and benefits for both the punisher and the target, unintended punishment may still have the same effects as intended punishment. The cognitive processes required for self-regarding behaviors such as aggression and avoidance are minimal. In social insects, for example, an aggressive response is triggered by a biochemical cue. It is a fixed, evolved response and is cognitively not much more

Animal Welfare and Conservation j Punishment

217

interesting than a rose thorn. The cognitively relevant cases are those that are intentional and flexible. Retaliatory punishment is an aggressive response to an aversive stimulus. Shunning need only be explained by avoidance. However, the behaviors in many species are not inflexible. For instance, a low-ranking baboon will not behave aggressively to a dominant individual, or even the kin of the dominant. It may have learned this lesson through past experience (i.e., losing violent conflicts). The motivations for punishment are self-regarding – the punishing animal need only have as its goal the reduction of annoyances in its social environment. Consideration for how this affects the target of the punishment is not required. The simplest mechanism for dealing with non-cooperators would be to stop cooperating after being exploited; however, indiscriminate shunning punishes cooperators as well and would cause cooperation to fall apart. Cognitive mechanisms that would allow targeted avoidance of non-cooperators should include individual recognition and perhaps a memory for specific events (episodic-like memory). Emotions such as anger are likely important in mediating punishment. In the punishment experiment with chimpanzees described earlier, when chimpanzees exhibited aggressive displays and tantrums – indicators of anger – they were more likely to collapse the food table. While likely important for motivating immediate reactions to undesirable events, it would be valuable to find evidence for planning punishment after the eliciting event (e.g., cold-blooded revenge). At present, there appears to be no evidence for delayed retributive punishment in non-human animals.

Social (Other-Regarding) Preferences Other-regarding preferences are social. These are behaviors that are motivated by a concern for the welfare of others. Even though the outcome of punishment may benefit the punisher, the motivation to punish need not be self-regarding; the goal might be the effect it has on the behavior or the psychological state of the target, with any personal benefits arising as unintended byproducts. In addition to having the goal of decreasing harmful behavior directed at the self, the punishing individual may have as goal a change in the non-cooperative behavior of the target, or may have the goal of maintaining a cooperative social environment. However, selfish benefits make it hard to rule out self-regarding motives, and it is difficult to establish other-regarding punitive motives. Furthermore, any individual action can arise from multiple motivations. For example, someone might scold a youth for crossing the street against a red light so as to protect him from potential harm, as well as to uphold the norms of society, to set an example for children who are present, to impress upon one’s peers that one is an upstanding member of society, to relieve the moral outrage he feels, to experience pleasure in causing embarrassment in the youth, and to be the first to cross the street when the light is green so as to get the best seat on the waiting tram. The important point is not to elucidate the complex suite of motivations for every action, but to determine whether certain motives even exist.

Antisocial preferences Punishment can be motivated by a concern for the negative welfare (suffering) of others. This is not the same thing as self-regarding punishment, which is neutral to the consequences of others; it has as its primary goal that the target suffer. Antisocial (negative other-regarding) preferences such as spite and schadenfreude (pleasure in the misfortunes of others) can motivate punishment. For example, humans (at least males) experience pleasure in seeing an individual who previously cheated them receive a painful stimulus, yet will exhibit empathy for a cooperator in pain. An important aspect of human cooperation may be a sensitivity to fairness. Aversion to inequity can motivate people to correct unfairness. People are ultracompetitive in that they compare their gains and losses to the gains and losses of others. They are sensitive to – and reciprocate strongly against – personally unfair outcomes as well as unfair intentions (West et al., 2007). When faced with personal unfairness in experiments such as the ultimatum game, people report feelings of anger and they show appropriate facial expressions and physiological responses such as an increased heart rate. This ‘wounded pride’ can motivate punishment in the absence of anticipated rewards such as future reciprocity or reputation (Gächter and Hermmann, 2009). Spite, in the typical, psychological sense, has the suffering of the target as the ultimate goal and is not a means to an end. Altruistic punishment may therefore be a byproduct of spiteful (antisocial) punishment; the punisher inflicts harm on another individual for the sake of causing harm, rather than out of altruistic motives for others. There is considerable debate about whether other animals have a sensitivity to fairness, particularly disadvantageous inequity aversion. Several experiments have suggested that some non-human primates and even dogs (Canis familiaris) are sensitive to inequity. However, several other experiments dispute these findings. In none of these studies can the animals inflict harm on others in response to inequity; they can only react to the experimenter. Two studies in chimpanzees have attempted to address antisocial punishment directly. In the punishment study described earlier, chimpanzees did not react spitefully by ‘punishing’ unfair outcomes when an experimenter pulled their food away and gave it to a conspecific (Jensen et al., 2007b). Nor, in the miniultimatum game did responders react in any way to unfair outcomes. Taken together, these studies suggest that harming others for the sake of causing others to experience loss might be unique to humans. Counter-intuitively, such antisocial preferences might lead to altruistic outcomes; altruistic punishment is not necessarily motivated by altruism, and instead may be driven by spite.

Prosocial preferences Punishment need not be self-regarding nor negatively other-regarding. One might be motivated by the positive outcomes for others. Empathy is the standard for prosocial preferences. Many parents, for instance, will tell their children that they are punishing them ‘for their own good’ and that it hurts them more than it hurts the child. The punishment is intended to produce delayed benefits for the target through the imposition of immediate costs; any benefits or costs to the punisher are unintended. Parental discipline could

218

Animal Welfare and Conservation j Punishment

qualify as being prosocially motivated. Examples of this are intentional teaching (discipline) and reform. For example, humans will punish their child if it engages in a potentially risky activity to teach it to not do this; they will reform it for having done something that is harmful to others to prevent that from happening in the future. The majority of examples of discipline in animals are inconclusive, however, since the parents punish behavior that is personally aversive; the punishment therefore provides immediate, direct benefits to the parents. One potential exception is a field observation in chimpanzees of a mother aggressing against her infant that was about to eat poisonous leaves, though single observations are always inconclusive. It is not yet clear what motivates altruistic, third-party or normative punishment in humans. It is possible that people have altruistic motives, that they have as a goal the modifying of the behavior of defectors so that others will benefit (Gardner and West, 2004). They can be motivated by advantageous inequity aversion and empathy. It is also possible that they are indifferent to the effects of the punishment on both themselves and the target, but are motivated to achieve a particular social effect, such as a sense of justice. Punishment is used as a means to an end, and a punitive sentiment is a motive to see a non-cooperator harmed for a cooperative end (and not an end in itself as in the case of spite). An altruistic punitive sentiment is a motive to harm the target for the target’s own good, or for the benefit of others. Altruistic punitive motives can also be for the benefit of individuals other than the target, such as preventing free-riders from exploiting others in the future. The important point is that the benefit of the target or of other individuals is the final goal of altruistic punishment, and that it is not a self-serving means to an end.

Other Cognitive Mechanisms As alluded to earlier, emotions are likely important components for punishment. Anger, in particular, appears to be a feature of all forms of punishment in humans and is likely to be prevalent in other animals. The circumstances that motivate anger are likely to vary; moralistic outrage, for instance, will only apply to altruistic, normative, or third-party punishment, since the emotion is in response to rule violations and not personal offenses. Other cognitive features that might be important for punishment in humans – and potentially their closest living relatives – are perspective-taking and intention reading (Seymour et al., 2007). Chimpanzees, but not capuchin monkeys, have been shown to take the visual perspective of conspecifics in a competitive situation. Such visual perspective-taking allows them to know what another individual sees, has seen, and presumably knows. Chimpanzees, as well as capuchin monkeys, also recognize intent in others and are less likely to remain engaged in an interaction with a human who is unwilling to give them food than one who attempts to give them food but is unable to do so (Gros-Louis, 2004). However, there is as yet no evidence for an understanding of beliefs and desires (theory of mind) in non-human animals. Secondary (social) emotions such as satisfaction at revenge. pride in seeing justice done, guilt in failing to punish, and forgiveness when the punishment has been effective may also be important for punishment, at least in humans.

Conclusion Punishment is an important way in which organisms can control or manipulate their social environments. Most often, it provides direct benefits to the punisher in that the offender stops engaging in the harmful behavior – ideally both at the moment and in the long term. In some cases, the direct benefits are great enough that the punisher will suffer some costs to bring it about. In addition, by punishing free-riders, others can benefit, and this can stabilize cooperation. Altruistic punishment, third-party punishment, and norm enforcement are special forms of punishment that may allow for the development of cooperation of unrelated individuals on a large scale seen only in humans. The cognitive mechanisms of particular importance to punishment are other-regarding preferences. Cooperative outcomes can result from purely self-regarding motives. The evolution of complex, stratified societies are likely to require other-regarding preferences. Antisocial preferences such as spite and prosocial preferences such as empathy can facilitate cooperation by being especially sensitive to cheaters and other norm violators.

See also: Animal Welfare and Conservation: Learning and Conservation; Overview of Animal Training: A Welfare Perspective; Stress, Health and Social Behavior. Methodology: Assessment of Welfare and Needs. Overview Essays: Welfare Concepts.

References Boyd, R., Richerson, P., 1992. Punishment allows the evolution of cooperation (or anything else) in sizable groups. Ethology and Sociobiology 13, 171–195. Clutton-Brock, T.H., Parker, G.A., 1995. Punishment in animal societies. Nature 373, 209–216. de Waal, F.B.M., Luttrell, L.M., 1988. Mechanisms of social reciprocity in three primate species: Symmetrical relationship characteristics or cognition? Ethology and Sociobiology 9, 101–118. Fehr, E., Fischbacher, U., 2003. The nature of human altruism. Nature 425, 785–791. Fehr, E., Gächter, S., 2002. Altruistic punishment in humans. Nature 415, 137–140. Gächter, S., Hermmann, B., 2009. Reciprocity, culture and human cooperation: Previous insights and a new cross-cultural experiment. Philosophical Transactions of the Royal Society B 364, 791–806. Gardner, A., West, S.A., 2004. Cooperation and punishment, especially in humans. American Naturalist 164, 753–764.

Animal Welfare and Conservation j Punishment

219

Gros-Louis, J., 2004. The function of food-associated calls in white-faced capuchin monkeys, Cebus capucinus, from the perspective of the signaller. Animal Behaviour 67, 431–440. Hauser, M.D., 1992. Costs of deception: Cheaters are punished in rhesus monkeys (Macaca mulatta). Proceedings of the National Academy of Sciences of the United States of America 89, 12137–12139. Jensen, K., Call, J., Tomasello, M., 2007a. Chimpanzees are rational maximizers in an ultimatum game. Science 318, 107–109. Jensen, K., Call, J., Tomasello, M., 2007b. Chimpanzees are vengeful but not spiteful. Proceedings of the National Academy of Sciences of the United States of America 104, 13046–13050. Seymour, B., Singer, T., Dolan, R., 2007. The neurobiology of punishment. Nature Reviews Neuroscience 8, 300–311. Sigmund, K., 2007. Punish or perish? Retaliation and collaboration among humans. Trends in Ecology & Evolution 22, 593–600. Silk, J., 2008. Social preferences in primates. In: Glimcher, P.W., Camerer, C.F., Fehr, E., Poldrack, R.A. (Eds.), Neuroeconomics: Decision Making and the Brain. Elsevier, London, UK, pp. 269–284. West, S.A., Griffin, A.S., Gardner, A., 2007. Social semantics: Altruism, cooperation, mutualism, strong reciprocity and group selection. Journal of Evolutionary Biology 20, 415–432.

Applications of Animal Behavior to Conservationq Ronald R Swaisgood and Alison L Greggor, San Diego Zoo’s Institute for Conservation Research, Escondido, CA, United States © 2019 Elsevier Ltd. All rights reserved.

Abstract Animal behavior can influence conservation outcomes, and can be used as a tool for diagnosing anthropogenic impacts and managing species’ recovery. Researchers from disparate backgrounds in animal behavior, most notably behavioral ecology and applied ethology, are using their research to contribute to conservation efforts, including reserve design, human disturbance, and reintroduction programs. The potential for animal behavior to contribute to conservation is growing in subdisciplines ranging from social behavior to animal learning, but progress in utilizing behavioral research will rely on increasing access to evidence of its effectiveness in comparison to traditional methods.

Keywords Antipredator behavior; Captive breeding; Conservation; Dispersal; Effective population size; Habitat selection; Human disturbance; Learning; Mate choice; Reintroduction; Social processes; Spatial ecology; Temperament

Introduction Conservation stems from a human value system that seeks to maintain the diversity of life on earth and ensure the ecological integrity of our natural heritage. People may be motivated to conserve nature by utilitarian values, such as recognition of the important services to humanity that a functioning ecosystem provides, or by a deep and abiding philosophy that there is intrinsic value to other forms of life. The biophilia hypothesis – championed by the founding father of biodiversity and sociobiology, E.O. Wilson – posits that humans are predisposed to an emotional attachment to nature that motivates them toward environmental stewardship. It may be argued that animal behaviorists, who spend long hours observing animals, have a more-than-average dose of biophilia, making their late arrival on the conservation stage surprising. Although the modern academic discipline of conservation biology has its roots in wildlife management that goes back generations, it was not founded until the 1980s. Conservation behavior – as the application of behavioral research to conservation is sometimes called – traces its formal, academic beginnings to the waning moments of the last millennium. With almost two decades of concerted effort, much of the heady promise of this nascent field has yet to come to fruition. However, conservation behaviorists are attempting to reinvent the way that behavioral research is applied to conservation efforts. Influential books and papers continue to highlight the many implications that behavior – particularly behavioral ecology – has for conservation. The strong theoretical framework afforded by behavioral ecology provides the basis for a hypothetico-deductive approach to conservation science. Integrating the four levels of explanation in animal behavior – causation, development, adaptive utility and evolutionary history – across larger ecological scales – population, community, ecosystem, landscape – holds the most promise for the successful application of behavioral research to conservation. The challenge for behavioral scientists has been to move from the implications phase of conservation behavior to more active applications that solve real-world conservation problems. Several subdisciplines within animal behavior have contributed to the emergence of conservation behavior, but to date their influence is not spread evenly. Many behavioral ecology topics appear promising for conservation application – including dispersal and habitat selection, foraging patterns, mate choice, and behavioral responses to habitat fragmentation and anthropogenic disturbance. However, efforts thus far have focused more on behaviors such as dispersal and foraging than on antipredator, social or learning behaviors (Berger-Tal et al., 2015). Moreover, behavior is still widely underutilized, even in areas of conservation where it should be highly applicable. For example, although greater than 30% of papers on animal reintroductions published between 1990 and 2005 reported difficulties that stem from behavior, only 5% actually mention behavior (Berger-Tal et al., 2015). Despite the historic disconnect between conservationists and behavioral ecologists, routes have been identified for improving the influence and usability of behavioral knowledge; for example, in fostering better communication with conservation managers, conducting research with defined management applications, gaining a greater understanding of wildlife management decisions and by offering easily accessible evidence. To better explain the scope, challenges and opportunities for conservation behavior, we first briefly outline the range of conservation problems where behavior could be of use and then expand on the role of different areas of animal behavior in solving those problems.

q

Change History: October 2017. Ronald Swaisgood and Alison Greggor updated all the sections and added new figures for 2 and 4. This is an update of R. Swaisgood, Conservation and Animal Behavior, In Encyclopedia of Animal Behavior, edited by Michael D. Breed and Janice Moore, Academic Press, Oxford, 2010, Pages 359-365.

220

Encyclopedia of Animal Behavior, 2nd edition, Volume 1

https://doi.org/10.1016/B978-0-12-809633-8.90027-X

Animal Welfare and Conservation j Applications of Animal Behavior to Conservation

221

Taking Stock of the Problem: What are Some of the Challenges Facing Animal Conservation? Today’s natural world faces an onslaught of anthropogenic processes that threaten the functioning of ecosystems, leading to historically high rates of species loss and defaunation (Dirzo et al., 2014). Habitat degradation and destruction are the single largest culprits, and the resulting fragmentation of habitat supporting wildlife has multiplicative rather than additive effects on biodiversity loss. Widespread urbanization and intensive agricultural practices eliminate many species, whereas other human activities degrade much of the remaining natural areas. Chemical pollutants may affect survival and reproduction, and thus population recruitment rates. Noise pollution may likewise disturb mating and parenting behavior or lead to chronic stress, with its attendant consequences for immunocompetence, fertility, and allocation of energetic resources away from other demands of survival. A single aircraft overflight has been known to cause the immediate loss of most chicks in a white pelican nesting colony. Sadly, even our love for nature in the form of ecotourism may be harming animals (Geffroy et al., 2015). A simple walk on the beach disturbs shorebirds, diverting them away from foraging, mating or parenting behavior. Multiplied by thousands of beachgoers, the cumulative effects on survival and reproduction can mean chronic sublethal effects and population decline. Pets, acting as predators along edges of natural communities, can have reverberating effects on community dynamics deeper in nature reserves. Studies have shown that even on-leash dogs can reduce diversity along a surprisingly wide trail margin. Humans also often aid transport and colonization of invasive nonnative species. Species with the right suite of behavioral and life-history characteristics sometimes get a foothold in their new environment and – without the ecological controls in their place of origin – undergo rapid population expansion. Over-run with exotics, the results for native competitors or prey species can be devastating. Take the case of the mountain yellow-legged frog (Rana muscosa) Fig. 1 in California. The introduction of brown trout and the larger and more aggressive bullfrog have contributed to the decline of this endangered species. This high-mountain frog has no coevolutionary history with fish predators and its tadpoles are vulnerable to predation by nonnative trout. Populations have recovered rapidly when trout have been experimentally removed (Vredenburg, 2004). With the many changes to habitats that humans make, animals’ evolutionarily advantageous behaviors may no longer lead to the beneficial outcomes they have evolved to expect. Prior habitat and cue preferences can lead animals to make suboptimal dispersal, settlement, or foraging decisions, even when suitable novel habitat or resources may exist nearby. These types of poor decisions lead animals into what has been deemed ‘ecological traps’ and ‘perceptual errors’, which can greatly reduce fitness and survival (Gilroy and Sutherland, 2007). Overlaid on top of these long-known threats to animal populations is the unpredictable impact of anthropogenic climate change, which will exacerbate other impacts. Climate change may precipitate range shifts in many species, but fragmented habitat may prevent movement, calling upon even greater human intervention, such as widespread translocations of animals to help track shifting habitats (Hällfors et al., 2017). While there has been progress in creating models that predict climactic changes, models designed to predict at-risk species and populations require an understanding of animals’ behavioral flexibility in their range or migratory patterns (Keith and Bull, 2017).

How Does Conservation Address Species Loss? To combat and reverse human-caused changes to the environment, conservation efforts engage in three types of action: (1) Determining what level of species diversity or ecosystems should be conserved, (2) identifying and quantifying the threats to said units of diversity or ecosystems, (3) and mitigating those threats (Primack, 2014; Berger-Tal and Saltz, 2016).

Fig. 1 One of the estimated 90%), it was possible to identify the most prototypical movement combinations for each expression category. Table 1 illustrates some of these prototypical expression categories. Of interest for homology is whether any of these expressions resemble human emotional expressions, so the latter are also plotted along with their specific movement configurations. Many of the matching human expressions are not identical to those described by Paul Ekman as basic facial emotions (anger, fear, happiness, sadness, surprise, and disgust), but represent some variation of these. Sadness may be most similar as both young human infants and chimpanzees have pout faces used in very similar situations. Happiness, conveyed best by laughter, is thought to be homologous to the chimpanzee play face, but in humans, the upper teeth are fully exposed while the chimpanzee play face conspicuously covers the upper teeth. Note that chimpanzee play faces also have a distinct laughter-type vocalization. Smiling, which can also convey happiness, shares most in common with the baredteeth face and based on the earlier discussion, at least some versions of this facial expression are best characterized as appeasing/ reassuring in both chimpanzees and humans. Anger shares some features in common with the chimpanzee bulging-lip face, although chimpanzees do not show the brow knitting that is a characteristic of human anger. Anger can also be conveyed by screaming, as in strong protests or rage, and chimpanzee screams are largely expressions of protest as opposed to real terror or pain. Although these are only semiquantitative comparisons, they provide a clear visual representation of how similar many of the facial expressions are between chimpanzees and humans, enabling informed assumptions about their underlying emotional content.

Communication and Cognition Categorization of Facial Expressions Also important for social cognition is how different species perceive and categorize social signals, like facial expressions. The signals must be easily distinguished from one another or they would fail to have any predictive function. Seemingly at odds with this

Cognition j Emotion and Social Cognition in Primates

363

prediction is the well-established finding that primate, particularly chimpanzee, facial expressions are highly graded, both in terms of their form, the exact movements involved in the display, and their overall intensity. Some researchers have even noted that chimpanzees have some expressions that appear to be blends, showing features in common with several different types and that these expressions can be used in many different contexts. Explaining how chimpanzees perceive and process these highly variable signals would be difficult based on cue-response contingency learning. By invoking the concept of behavioral/cognitive flexibility, however, regularities in the use of these specific movement patterns can come to communicate general tendencies or motivations that can then be parsed in to more detailed meaning by incorporating additional factors, such as actor, environment, and antecedent conditions. Thus, the concept of behavioral/cognitive flexibility nicely fits with existing data on emotional communication. At this stage, perceptual discrimination studies in chimpanzees have only utilized the most prototypical examples of facial expressions, like those shown in Table 1. However, it is very plausible that graded signals can provide meaning to group members mentioned earlier and beyond that related to motivational intensity. Small changes in expressions or blends between expressions may alter meaning in completely new and, as yet, unstudied ways. Parr et al. (1998) were the first to examine facial expression discrimination in chimpanzees using a computerized, joysticktesting paradigm. According to this procedure, subjects first must orient to a sample stimulus, for example, a facial expression, by contacting it with a joystick-controlled cursor. Then, two comparison images appear on the screen, one showing another example of the same expression type as the sample and the other showing a neutral face (see Figure 2). In these experiments, all individuals were different, so subjects could not match based on individual identity. Of the five expression categories presented – bared-teeth display, pant hoot, scream, relaxed lip face, and the play face – subjects learned to discriminate all but the relaxed-lip and the neutral face. This was interesting as these two faces are emotionally-neutral, but the relaxed-lip contains the distinctive droopy lower lip. In an effort to understand the role of unique and distinctive features in expression discrimination, such as mouth position, visibility of the teeth, staring eyes, etc., a follow-up experiment was performed in which each expression category was paired with every other, so every pair-wise combination of expression types was represented. Performance discriminating these dyads was then correlated with the number of features shared between the two expression types, such as mouth open, teeth visible, etc. If performance

Figure 2 An example of the matching-to-sample task used to study expression categorization in chimpanzees. The image on top is the sample, the one to match. The correct choice is the matching expression (lower left).

364

Cognition j Emotion and Social Cognition in Primates

was based on the detection of specific features, like a droopy lip, then expression dyads that shared these features would result in poorer performance than expression dyads that had no overlap in features. This hypothesis was supported for only some expression types – the bared-teeth display, pant-hoot, and relaxed-lip face – but not others, such as the play face or scream. Thus, it was concluded that the features of some expressions are more salient than others. After the development of the ChimpFACS, this experiment was replicated using a standardized set of chimpanzee facial expressions generated using a three-dimensional chimpanzee model (see Figure 3). In this way, all facial expressions of the same category contained identical movements and each was standardized to its peak intensity. The pair-wise matching task was then replicated and the pattern of errors made by subjects was analyzed using a multidimensional scaling analysis. This creates a graph of how similar or different each expression is to each other. Those that cluster together are perceived as highly similar, while more dispersed expressions are perceived as more distinct. Then, the dimensions of the plot can be interpreted based on the features of each expression. One dimension appeared to reveal the degree of vertical mouth opening or closing and the other was the degree of horizontal mouth opening with lips either being retracted (as in the bared-teeth display) or puckered as in the pout. In a slight modification of the task, subjects were required to match each expression prototype by selecting only one of its individual component movements, thus asking what individual movement, if any, was the most representative of each expression prototype. For example, the bared-teeth display is comprised of three movements, one raising the upper lip (action unit (AU) 10), one lowering the lower lip (AU16) and one retracting the lips back (AU12). These were combined into three trials, pairing the bared-teeth prototype with each combination of these three movements (see Figure 4). Subjects showed a clear preference for which movement was most representative of each prototype and, moreover, this single movement explained most of the errors found in the previous cluster plot. Most expressions were relatively easy to discriminate and little confusion occurred, but when subjects did have trouble it was not because the two expressions in the pair shared the most number of features in common, as was the earlier hypothesis, it was that the two expressions shared their most salient movement in common. It is perhaps these salient movements that are involved in a general pattern recognition heuristic and provides chimpanzees with an ability to categorize facial expressions into basic groups, despite the degree of blending or grading, the meaning of which can be elaborated on by other factors, like context.

Figure 3

An example of standardized, prototypical chimpanzee facial expressions created using 3D animation software.

Figure 4 The Poser chimpanzee model, showing the bared-teeth display prototype and each individual component movement it contains. Reproduced from Parr LA, Waller BM, and Heintz M (2008) Facial expression categorization by chimpanzees using standardized stimuli. Emotion 8: 216– 231, with permission from $ $ $.

Cognition j Emotion and Social Cognition in Primates

365

Conclusion: The Role of Emotion in Social Cognition This article has argued that social cognition is best understood through detailed and objective measurements of observable behavior, such as facial expressions, with an appreciation for the dynamic environment in which these signals take place. By comparing how these emotional signals look and function in related species, we can begin to understand their evolutionary origins, thus better understanding their overall function and meaning. What has been learned is that the primate social environment is extremely complex and most behavioral interactions do not lend themselves to interpretation based on learned or innate cueresponse contingencies, nor do they warrant rich interpretations based on unobservable mental states. Instead, it is suggested that primates interpret their social interactions using more flexible behavioral heuristics. Primates have the ability to recognize specific behavioral cues, such as faces and facial expressions, performed in a variety of social contexts, providing them with an understanding of the basic motivational states of others. This is then filtered through an assessment of other factors, such as social context and previous experience, giving more detailed information about the overall emotional meaning of the signal, and what the sender is likely to do. As a result, the cognitive and emotional state of others can be interpreted and maybe even predicted, not from mind reading, but from the perception of the others’ behaviors as it occurs dynamically within a specific environment. The dynamic nature of social communication creates an even more complicated scenario, as these interactions occur rapidly and can change course frequently, necessitating behavioral adjustments based on ongoing processing. Facial expressions do not appear in a simple sequence: behavior a, then expression b, leads to behavior c, interpreted as expression b functions to change behavior from a to c. Rather, facial expressions are embedded within dynamic social interactions that takes the form of multiple feedback loops, dependent on the reaction of the social partner and changing in ongoing needs. To understand the real essence of social cognition, one must study these interactions within the distributed cognitive system in which they occur, where behaviors do not occur in linear cause and effect sequences, but rather communication emerges within a structure of a dynamic and everchanging society.

See also: Cognition: Cognitive Development in Chimpanzees. Communication: Visual Signals Using Incident Light.

Further Reading Andrew, R.J., 1963. The origin and evolution of the calls and facial expressions of the primates. Behaviour 20, 1–109. Barrett, L., Henzi, P., Rendall, D., 2007. Social brains, simple minds: Does social complexity really require cognitive complexity? Philosophical Transactions of the Royal Society B 362, 561–575. Burrows, A.M., 2008. The facial expression musculature in primates and its evolutionary significance. BioEssays 30, 212–225. Dunbar, R.I.M., 1988. Primate Social Systems. Chapman & Hall, London. Parr, L.A., Cohen, M., de Waal, F.B.M., 2005. The influence of social context on the use of blended and graded facial displays in chimpanzees (Pan troglodytes). International Journal of Primatology 26, 73–103. Parr, L.A., Hopkins, W.D., de Waal, F.B.M., 1998. The perception of facial expressions in chimpanzees (Pan troglodytes). Evolution of Communication 2, 1–23. Parr, L.A., Waller, B.M., Heintz, M., 2008. Facial expression categorization by chimpanzees using standardized stimuli. Emotion 8, 216–231. Parr, L.A., Waller, B.M., Vick, S.J., Bard, K.A., 2007. Classifying chimpanzee facial expressions using muscle action. Emotion 7, 172–181. Povinelli, D.J., Vonk, J., 2003. Chimpanzee minds: Suspiciously human? Trends in Cognitive Sciences 7, 157–160. Preston, S., de Waal, F.B.M., 2002. Empathy: Its ultimate and proximate bases. Behavioral and Brain Sciences 25, 1–72. Preuschoft, S., 1992. Laughter’ and ‘smile’ in Barbary macaques (Macaca sylvanus). Ethology 91, 220–236. Tomasello, M., Call, J., Hare, B., 2003. Chimpanzees understand psychological states-the question is which ones and to what extent. Trends in Cognitive Science 7, 153–156. van Hooff, J.A.R.A.M., 1967. The facial displays of the Catarrhine monkeys and apes. In: Morris, D. (Ed.), Primate Ethology. Aldine, Chicago, IL, pp. 7–68. van Hooff, J.A.R.A.M., 1972. A comparative approach to the phylogeny of laughter and smiling. In: Hinde, R.A. (Ed.), Nonverbal Communication. Cambridge University Press, Cambridge, pp. 209–241. Waller, B., Dunbar, R.I.M., 2005. Differential behavioural effects of silent bared teeth display and relaxed open mouth display in chimpanzees (Pan troglodytes). Ethology 111, 129–142.

Relevant Websites http://userwww.service.emory.edu/lparr/index.html. – Dr. Lisa Parr’s website. www.chimpfacs.com. – The ChimpFACS website. http://www.do2learn.com/games/facialexpressions/face.htm. – Do2Learn-Facial expressions.

Empathetic Behavior FBM de Waal, Emory University, Atlanta, GA, USA © 2010 Elsevier Ltd. All rights reserved.

Abstract State matching and emotional contagion probably evolved with parental care, which required appropriate responses to offspring distress, after which it generalized to other social domains. Built around this core evolved higher empathy levels, including responses geared to the other’s specific situation, thus increasing the effectiveness of support, care, and reassurance. Inasmuch as some of these higher levels require perspective-taking, they are more typical of large-brained species such as apes, dolphins, and elephants. Empathy in general, however, characterizes all mammals, and probably also birds. It is proposed here as a proximate mechanism of directed altruism and prosocial behavior.

Keywords Altruism; Cooperation; Emotions; Empathy; Reciprocity; Self-awareness; Theory-of-mind

Definitions of empathy commonly emphasize two aspects, which is the sharing of emotions and the adoption of another’s perspective. The latter, cognitive aspect remains controversial for many species, but the first, emotional aspect is hard to deny, and was already recognized by Darwin in The Descent of Man: “. many animals certainly sympathize with each other’s distress or danger.” Empathy allows the organism to quickly relate to the states of others, which is essential for the regulation of social interactions, coordinated activity, and cooperation toward shared goals. Even though perspective-taking is often critical, it is a secondary development. This is even true for our own species, as Hoffman noted: “humans must be equipped biologically to function effectively in many social situations without undue reliance on cognitive processes.” The selection pressure to evolve rapid emotional connectedness likely started in the context of parental care. Signaling their state through smiling and crying, human infants urge their caregiver to come into action, and equivalent mechanisms operate in other animals in which reproduction relies on feeding, cleaning, and warming the young. Offspring signals are not just responded to but induce an agitated state, suggestive of parental distress at the perception of offspring distress. Avian and mammalian parents alert to and affected by their offspring’s signals must have out-reproduced those that remained indifferent. Once the empathic capacity existed, it could be applied outside the rearing context and play a role in the wider network of social relationships. The fact that mammals retain distress vocalizations into adulthood hints at the continued survival value of careinducing signals. For example, primates often lick and clean the wounds of conspecifics, which is so critical to healing that migrating adult male macaques have been observed to temporarily return to their native group, where they are more likely to receive this service. One of the first experimental studies of animal empathy was Church’s (1959) study entitled ‘Emotional Reactions of Rats to the Pain of Others.’ Having trained rats to obtain food by pressing a lever, Church found that if a rat pressing the lever perceived that another rat in a neighboring cage receives a shock from an electrified cage floor, the first rat would interrupt its activity. Why should this rat not continue to acquire food? The larger issue is whether rats that stopped pressing the lever were concerned about their companions or just fearful that something aversive might also happen to themselves. Church’s work inspired a brief flurry of research during the 1960s that investigated concepts such as ‘empathy,’ ‘sympathy,’ and ‘altruism’ in animals. This included studies of monkeys, which showed a much more dramatic empathy response than rats. Monkeys will for many days refuse to pull a chain that delivers food to them if doing so delivers an electric shock to a companion. In order to avoid accusations of anthropomorphism, however, authors often placed the topic of their research in quotation marks, and their studies went largely ignored in ensuing years. In the meantime, human empathy became a respectable research topic. In the 1970s began studies of empathy in young children, in the 1980s in human adults, and finally, in the 1990s, neuroimaging of humans watching others in pain, distress, or with a disgusted facial expression. Mirror neurons are commonly invoked to explain human empathy, but despite the fact that these neurons were discovered not in humans, but in macaques, animal empathy research has lagged. Half a century after Church’s study, however, there is a revival of interest in animal empathy and a basic mechanism common to humans and other animals has been proposed. Accordingly, seeing another in a given situation or display certain emotions reactivates neural representations of when the subject was itself in similar situations or experienced similar emotions, which in turn generates a bodily state resembling that of the object of attention. Thus, seeing another individual’s pain may lead the observer to share the bodily and neural experience. These reactions are so automatic and instantaneous that it is not unusual for humans to shout ‘ouch!’ while watching a child scraping its knee. The Perception–Action Mechanism (PAM) of Preston and de Waal manifests itself very early in human life, such as when newborns cry contagiously, and is increasingly suggested for other animals. Examples range from involuntary facial mimicry during play in orangutans, yawn contagion in both primates and canines, to heart-rate increases in geese while watching their mated partner in a fight. Physiological continuity between the ways humans and chimpanzees process emotional stimuli is suggested

366

Encyclopedia of Animal Behavior, 2nd edition, Volume 1

https://doi.org/10.1016/B978-0-12-813251-7.00105-X

Cognition j Empathetic Behavior

367

by the apes’ drop in skin temperature while watching aversive images as well as human-like lateralized changes in brain temperature.

Emotional Contagion in Mice Langford and colleagues put pairs of mice through a so-called ‘writhing test.’ In each trial, two mice were placed in two transparent Plexiglas tubes such that they were able to see one another. Either one or both mice were injected with diluted acetic acid, known to cause a mild stomach-ache. Mice respond to this treatment with characteristic writhing movements. The researchers found that an injected mouse would show more writhing if its partner was writhing, too, than it would if its partner had not been injected. Significantly, this applied only to mouse pairs that were cage mates. Male (but not female) mice showed an interesting additional phenomenon while witnessing another male in pain: its own pain sensitivity actually dropped. This counter-empathic reaction occurred only in male pairs that did not know each other, which are probably also the pairs with the greatest degree of rivalry. Was that rivalry suppressing their reaction, or did they actually feel less empathy for a strange rival? Finally, Langford and colleagues exposed pairs of mice to different sources of pain – the acetic acid as before and a radiant heat source. Mice observing a cage mate writhing because of the acid injection withdrew more quickly from the heat source. In other words, their reactions could not be attributed to mere motor imitation, but involved emotional contagion, because seeing a companion react to pain caused sensitization to pain in general.

Consolation Behavior A well-studied primate response to others’ distress is so-called consolation behavior, that is, friendly, reassuring contact by an uninvolved bystander toward a distressed party, such as the loser of a fight (Figure 1). Similar behavior in children is typically attributed to sympathetic concern. That consolation serves to alleviate distress is suggested by the finding that such contact is directed more at recipients of aggression than at aggressors, and more at recipients of intense than mild aggression (Figure 1). Other studies have confirmed consolation behavior in chimpanzees, gorillas, and bonobos as well as canines and corvids. However, when de Waal and Aureli set out to apply the same observation protocol to detect consolation in monkeys, they failed to find any, as did others. The consolation gap between monkeys and the Hominidae (i.e., humans and apes) extends even to the one situation where one would most expect consolation to occur: macaque mothers fail to comfort their own offspring after the receipt of aggression. Content analysis of hundreds of reports confirms that reassurance of distressed parties is typical of apes yet uncommon in monkeys.

Figure 1 Consolation behavior is common in chimpanzees, and functions to reassure distressed parties. A juvenile puts an arm around a screaming adult male who has just been defeated in a fight with a rival. Photograph by Frans de Waal.

368

Cognition j Empathetic Behavior

After initial failures to find effects of consolation among chimpanzees, a comprehensive analysis demonstrated that the rate of self-directed behavior (i.e., self-scratching and self-grooming) is elevated following aggressive conflicts, but drops significantly as soon as individuals receive consolation. Self-directed behavior serves as an index of anxiety, hence these observations suggest that consolation effectively counters anxiety induced by agonistic conflict. The early primate literature contains many striking qualitative accounts of consolation among apes. In 1925, Yerkes reported how his bonobo was so concerned about his sickly chimpanzee companion, Panzee, that the scientific establishment might not accept his claims: “If I were to tell of his altruistic and obviously sympathetic behavior towards Panzee I should be suspected of idealizing an ape.” Ladygina-Kohts noticed similar empathic tendencies in her young chimpanzee, Joni, which she raised at the beginning of the previous century. Kohts, who analyzed Joni’s behavior in the minutest detail, discovered that the only way to get it off the roof of her house after an escape (much better than holding out a reward) was to appeal to his sympathy: ‘If I pretend to be crying, Joni immediately stops his plays or any other activities, quickly runs over to me . he tenderly takes my chin in his palm, lightly touches my face with his finger, as though trying to understand what is happening.’

Empathic Perspective-Taking In 1996, I suggested that apart from emotional connectedness, apes have an appreciation of the other’s situation. Psychologists usually speak of empathy only if it involves such understanding combined with the adoption of the other’s point of view. Thus, one of the oldest and best-known definitions stresses the ‘changing places in fancy’ with the sufferer. The prevailing view of empathy as a cognitive affair dependent on imagination and mental state attribution explains the occasional skepticism about non-human empathy. But perspective-taking by itself is, of course, hardly empathy: it is so only in conjunction with emotional engagement. Menze was the first to investigate whether chimpanzees understand what others know, setting the stage for studies of nonhuman theory-of-mind and perspective-taking. After several ups and downs in the evidence, current consensus seems to be that apes, but perhaps not monkeys, show perspective-taking both in their spontaneous social behavior and under controlled experimental conditions. One manifestation of empathic perspective-taking is so-called ‘targeted helping,’ which is help fine-tuned to another’s specific situation. For an individual to move from emotional sensitivity toward an explicit other-orientation requires a shift in perspective. The emotional state induced in oneself by the other now needs to be attributed to the other instead of the self. A heightened selfidentity allows a subject to relate to the object’s emotional state without losing sight of the actual source of this state. The required self-representation is hard to establish independently, but one common avenue is to gauge reactions to a mirror. The coemergence hypothesis predicts that mirror self-recognition (MSR) and advanced expressions of empathy appear together in both ontogeny and phylogeny. Ontogenetically, there exists compelling evidence for the coemergence hypothesis. Bischof-Köhler found that the relation between MSR and the development of complex forms of empathy holds even after controlling for age. Gallup was the first to propose phylogenetic coemergence, a prediction empirically supported by the contrast between monkeys and apes, with compelling evidence for both MSR, consolation, and targeted helping only in the apes. Apart from the great apes, the animals for which we have the most striking accounts of consolation and targeted helping are cetaceans and elephants, which is why the coemergence hypothesis predicts MSR in these taxa. The mark test, in which an individual uses a mirror to locate a mark on itself that it cannot see without a mirror, has now tentatively confirmed this prediction for dolphins and elephants. MSR is believed to be absent in all other mammals. The coemergence hypothesis may have its own neural correlate, that is, the presence of Von Economo Neurons, or VEN cells. These special neurons have been found only in the brains of Hominoids (not other primates), cetaceans, and elephants. In the future, we may be able to address the self–other distinction more directly through neural investigation. In humans, the right inferior parietal cortex at the temporoparietal junction helps distinguish between self- and other-produced actions, and this distinction may be critical to full-blown empathy.

Altruism and Prosocial Behavior The common claim that humans are the only truly altruistic species, since animals are driven by return-benefits, assumes that animals not only engage in reciprocal exchange, but do so with a full appreciation of how this will ultimately benefit themselves. However, return-benefits generally remain beyond the cognitive horizon of animals, that is, occur too distantly in time to be linked to the original act. Since animals cannot be motivated by future events that they cannot predict, intentionally selfish altruism involves highly speculative (and as yet unproven) assumptions. When animals alert others to an outside threat, work together for immediate self-reward, or vocally attract others to discovered food, biologists may speak of altruism or cooperation, but such behavior is unlikely to be motivated by empathy with the beneficiary. One category, however, termed directed altruism (i.e., altruistic behavior aimed at others in need, pain, or distress), is traditionally explained as a product of empathy in humans, and may share the same proximate causation in animals.

Cognition j Empathetic Behavior

369

Apart from evidence for consolation behavior (seen earlier), there exists a rich literature on primate support in aggressive contexts, costly cooperation, and food-sharing. Primates engage in the latter even when separated by bars, hence while protected from physical pressure to share. Emotional activation is often visible in the facial expressions and vocalizations of both altruists and beneficiaries. Empathy is a perfect candidate mechanism for directed altruism, since it provides a unitary explanation for a wide variety of situations in which assistance is dispensed according to need. Perhaps confusingly, the mechanism is relatively autonomous in both animals and humans. Thus, empathy often reaches beyond its original evolutionary context, such as when people send money to complete strangers, primates bestow care on unrelated orphans, or a bonobo saves an injured bird. Experimentation on risky altruism is ethically problematic, yet there are increasingly experiments on low-cost altruism, also known as ‘other-regarding preferences.’ A typical paradigm is to offer one member of a pair the option to either secure food for itself or food for both itself and a companion. Recent experiments in my lab have shown well-developed prosocial tendencies in monkeys, such as marmosets and capuchins. In these studies, the role of reward could be ruled out in that the prosocial choices were made even if they delivered no extra rewards compared with selfish choices, and one study ruled out punishment in that the least vulnerable parties (i.e., dominants) turned out to be the most prosocial. These results are explained more easily by empathy with another’s situation than cost/benefit calculations. With regard to chimpanzees, the same tendencies have proved harder to establish until a series of experiments by Warneken and colleagues yielded an outcome in line with the overwhelming observational evidence for spontaneous aiding behavior in this species.

Empathy as Evolved Proximate Mechanism The PAM model predicts that the more similar or familiar subject and object, the more their neural representations will agree, and hence the more accurate their state-matching. Generally, the empathic response is amplified by similarity, familiarity, social closeness, and positive experience with the other (see Table 1 in Preston and de Waal’s 2002 paper). In studies in humans, subjects empathize with a confederate’s pleasure or distress if they perceive the relationship as cooperative, yet show a counter-empathic response – also known as Schadenfreude – if they perceive the relationship as competitive. These effects of previous experience have been confirmed by fMRI research: seeing the pain of a cooperative confederate activates pain-related brain areas, but seeing the pain of an unfair confederate activates reward-related brain areas, at least in men. Relationship effects are also known for rodents, in which emotional contagion is measurable between cage mates, but not strangers. In monkeys, empathic responses to another’s fear or pain are enhanced by familiarity between subject and object, and prosocial tendencies in capuchin monkeys vary with social closeness, being strongest between partners that spend most time together. This means that empathy and prosocial tendencies are biased the way evolutionary theory would predict. Empathy is (1) activated in relation to those with which one has a close or positive relationship and (2) suppressed, or even turned into its opposite, in relation to strangers and defectors. The latter, retaliatory aspect is well-documented in chimpanzees, which show both reciprocation of favors within positive relationships and a tendency to square accounts with those that have acted against them. A common way in which mutually beneficial exchanges are achieved is through investment in long-term bonds to which both parties contribute. This reciprocity mechanism is commonplace in non-human primates, and has been suggested for human relations as well. Individual interests may be served by partnerships (e.g., marriages, friendships) that create a long-lasting communal ‘fitness interdependence’ that flows from mutual empathy. Within close human relationships, partners do not necessarily keep careful track of which did what for which, as also indicated for reciprocal exchange among chimpanzees.

Conclusion Empathy covers all the ways in which one individual’s emotional state affects another’s, with simple mechanisms at its core and more complex mechanisms and perspective-taking abilities as its outer layers. Because of this layered nature of the capacities involved, we speak of the Russian Doll Model, in which higher cognitive levels of empathy build upon a firm, hard-wired basis, such as the PAM (Figure 2). The claim is not that PAM by itself explains sympathetic concern or perspective-taking, but that it underpins these cognitively more advanced forms of empathy, and serves to motivate behavioral outcomes. Empathy may provide the main motivation that makes individuals that have exchanged benefits in the past to continue doing so in the future. Instead of the cognitively demanding assumption of cost–benefit calculations and learned expectations to explain such behavior, the assumption here is one of empathy-based altruism mediated by bonding and emotional sensitivity. It is summarized in the following conclusions: 1. Evolutionary parsimony assumes similar motivations to underlie directed altruism in humans and other animals. 2. Consistent with kin selection and reciprocal altruism theory, empathy is biased toward close, familiar individuals and previous cooperators, and biased against defectors. 3. Empathy, broadly defined, may characterize all mammals and birds.

370

Cognition j Empathetic Behavior

Figure 2 The Russian Doll model of empathy and imitation. Empathy induces a similar emotional state in the subject as the object, with at its core the perception-action mechanism (PAM). The doll’s outer layers, such as sympathetic concern and perspective-taking (which includes understanding the reasons for another’s emotions), build upon this hard-wired socio-affective basis. Even though the doll’s outer layers depend on prefrontal functioning and an increasing self-other distinction, they remain fundamentally linked to its core.

See also: Cognition: Conflict Resolution; Emotion and Social Cognition in Primates.

Further Reading Bischof-köhler, D., 1988. Uber den Zusammenhang von Empathie und der Fäshigkeit sich im Spiegel zu erkennen. Schw. Z. Psychol 47, 47–159. de Waal, F.B.M., 1996. Good Natured: The Origins of Right and Wrong in Humans and Other Animals. Harvard University Press, Cambridge, MA. de Waal, F.B.M., 2008. Putting the altruism back into altruism: The evolution of empathy. Annual Review of Psychology 59, 279–300. Decety, J., Jackson, P.L., 2006. A social-neuroscience perspective on empathy. Current Directions in Psychological Science 15, 54–58. Gallup, G.G., 1982. Slef-awareness and the emergence of mind in primates. American Journal of Primatology 2, 237–248. Langford, D.J., Crager, S.E., Shehzad, Z., et al., 2006. Social modulation of pain as evidence for empathy in mice. Science 312, 1967–1970. Menzel, E.W., 1974. A group of young chimpanzees in a one-acre field. In: Schrier, A.M., Stollnitz, F. (Eds.), Behavior of Non-Human Primates, Vol. 5. Academic Press, New York, pp. 83–153. Preston, S.D., de Waal, F.B.M., 2002. Empathy: Its ultimate and proximate bases. Behavioral and Brain Sciences 25, 1–72.

Gestural Communication in the Great Apes Kirsty E Graham, University of York, York, England Catherine Hobaiter, University of St Andrews, St Andrews, Scotland © 2019 Elsevier Ltd. All rights reserved.

Abstract Language provides us with the most powerful social tool that any species has evolved. With it we can articulate any idea that comes to mind: From chaos theory to a knock-knock joke. And yet we have very little idea of how or why this remarkable faculty evolved. Other species’ communication permits rich information exchange; but humans do more than broadcast information, whether spoken or signed, with language we communicate particular goals to specific partners. Language goes beyond information; it has meaning. When it was demonstrated that great apes employ their large repertoires of gestures to communicate in this language-like way, it revolutionised our understanding of non-human communication. Here we describe the gestural communication of great apes, with a particular focus on its language-like characteristics including intentional use, meaning, and flexibility.

Keywords Communication; Evolution of language; Flexibility; Gesture; Goal-directed; Great ape; Intentional; Meaning; Multimodal; Openness; Repertoire; Structure; Syntax

What is Gestural Communication? All apes – human and non-human – use gestures. Gestures are mechanically ineffective limb and body movements that are used to communicate (Gómez, 1994). Human gesturing is a universal behaviour. People point, wave, stick out their lips, shake their heads, and present certain fingers...(!). They give someone a gentle nudge to indicate the direction they‫׳‬d like them to move to make space on the couch (gesture), rather than shoving them out of the way (action). Humans use gestures alongside speech; the co-occurrence of speech and gesture is so strong that many speakers continue to use gestures while talking on the phone, when their audience is unable to see them (Bavelas et al., 2008). In a series of studies, participants were asked to describe ‘Greebles’, digital 3D objects, but were not explicitly told to use gestures (Hoetjes et al., 2015). The participants used spoken words to describe the objects, but they also accompanied their speech with gestures – regardless of whether or not their audience could see them. There is an undeniable link between gesture, language, and cognition. Gesture may be used as a tool to help us to think through tasks, like when you are looking for scissors and walk around the room making a ‘scissor gesture’ with your fingers. Gestures are linked to numeracy ability in young children, with evidence that some aspects of number knowledge are evidence through gesture before they are shown in speech (Gunderson et al., 2015). In childhood, gesture develops alongside language acquisition. Young children reach to request for things, and children as young as 11- to 12-months point as a way of referring to objects (Carpenter et al., 1998; Tomasello et al., 2007). As adults we incorporate gesture in many ways: deictic gestures refer to objects, for example by pointing your lips or index finger towards something (Wilkins, 2003; Butterworth, 1998); iconic gestures convey meanings by resembling the referent, for example by making a scissor movement with your index and middle finger; and conventionalised gestures are arbitrary body movements to which we attribute meanings. Conventionalised gestures vary between linguistic groups (Kendon, 1997) with different gestures used for the same meanings, for example: greeting (hand waving, bowing), approving (thumbs up, head nodding, clapping), disapproving (thumbs down, head shake), or swearing (middle finger, middle and index finger, chin flicking); and the same gestures used for different meanings, for example: the middle and index finger in a V-sign, the vertical extension of the palm, or thumbs up (each of which might be supportive or obscene, depending on your interpretation). Nevertheless, despite its obvious stand-alone importance, human gesture typically accompanies language. Whether language is spoken or signed, co-speech and co-sign gesture regularly accompanies the linguistic content conveyed in words and signs. The combination of different modalities (forms) of communication is not unique to humans – all apes combine different types of signals, and can express themselves using vocal, gestural, postural, and facial signals. In non-human apes, from our current knowledge about each modality, it is in their gestures that we most clearly see the expression of language-like meanings.

Great Ape Gestures The use of gesture in non-human great ape (hereafter ‘great ape’) communication was recognized in the earliest studies of both captive (Witmer, 1909; Ladygina-Kohts, 1935, 2002; Kellogg and Kellogg, 1933) and wild apes (Schaller, 1963; Goodall, 1968; Plooij, 1978). A series of early studies examined the evolution and acquisition of human language by trying to teach it to apes who were raised ‘as nearly as possible like a human child’ (Hayes, 1951). While the development of these enculturated apes showed many parallels to that of human children, they learned little more than a few unconvincing ‫׳‬words‫( ׳‬Bryan, 1963; Hayes and Hayes,

Encyclopedia of Animal Behavior, 2nd edition, Volume 1

https://doi.org/10.1016/B978-0-12-809633-8.90158-4

371

372

Cognition j Gestural Communication in the Great Apes

1951). Chimpanzee vocal production is apparently not malleable enough for human speech (Dunn and Smaers, 2018). However, subsequent studies had far greater success when switching modality and teaching apes sign language. After 21 months, chimpanzee Washoe acquired over 30 signs; after 30 months, gorilla Koko acquired around 100 signs; and in total, chimpanzee Nim acquired around 125 signs (Gardner and Gardner, 1969; Patterson, 1978; Terrace et al., 1979). However, the field fell apart in an acrimonious debate over potential methodological flaws (Terrace et al., 1979; Umiker-Sebeok and Sebeok, 1981). A similar line of research, which trained apes on symbol keyboards, continued to explore apes’ potential for structured – syntax-like – combinations of symbols with some success (Savage-Rumbaugh and Lewin, 1994; Savage-Rumbaugh et al., 1993; Savage-Rumbaugh et al., 1986). The significant life-long harm created by separating infant apes from their mothers and social groups, and raising them to be ‫׳‬human‫ ׳‬means that there are no new enculturation studies (Leavens et al., 2017). Even if the ethical issues of such studies could be addressed (and they cannot), assessing the cognitive capacity for language in great apes through enculturation in the human world raises methodological questions too. While extraordinary individuals like Washoe, Koko, and Nim showed extraordinary capacities under extraordinary conditions, they were just that: extra-ordinary. Human language is adapted to the communication needs of the human species; if a non-human ape ‫׳‬fails‫ ׳‬to acquire specific aspects of language expression or comprehension, our first question ought to be: why should they? When considering the evolution of language and language-like communication, we may be better served by investigating the patterns of similarity and difference among apes’ own natural systems of communication. At around the same time as the ape language studies in the 1960s and 70s, Goodall’s seminal field studies of wild chimpanzees were revealing incredible descriptions of their communication and cognition (Goodall, 1986). Her observation of an adult male chimpanzee making and using a tool to dip for ants led Louis Leakey to write to her: ‘Now we must redefine tool, redefine Man, or accept chimpanzees as human’ (quoted, Goodall, 1998). Since then we humans have strenuously resisted the third option (accepting chimpanzees as human, or perhaps more correctly accepting humans as a third species of chimpanzee (Diamond, 1991)), in favour of either redefining tool, or, more commonly, redefining Man. But Goodall’s work established the scientific potential of careful observation of natural behaviour in wild ape populations, and the study of great ape language-capacities shifted to examining those that exist within their own systems of communication.

Gestures and Other Signals All apes employ a diverse range of signals. The bright pink oestrus swelling of a female chimpanzee or bonobo contains obvious visual (and probably tactile and olfactory) information regarding her proximity to ovulation (Reichert et al., 2002; Deschner et al., 2003). However, like the warning colouration on a caterpillar, this information is fixed, it is a feature of the signaller‫׳‬s body, visible to all, – she doesn’t choose to display it just to certain males and not to others. But apes can also produce large repertoires of more flexible signals including vocalizations, facial expressions, bodily postures, and gestures. Facial expressions and body postures may also be relatively fixed – they are hard to inhibit (Tanner and Byrne, 1993) and changes in how they are displayed could be attributable to differing levels of arousal (Waller et al., 2015). Vocalizations appear to be more flexible. There is evidence for a range of audience effects, including ‘exaggeration’ of distress screams when the audience contains an individual who might intervene on your behalf. Or the deft manipulation of the social consequences of increasing (Schel et al., 2013a) or decreasing (Hauser and Wrangham, 1987) the production of food calls, depending on the audience and the probable consequences of letting them know that there is food nearby. But, in research to date, it is the large repertoires of great ape gestures that show the most dramatic range and expression of flexibility. Gestures are selected and adjusted depending on the signaller’s goal, and whether or not the audience can see them (e.g. Tanner and Byrne, 1996; Pika et al., 2003, 2005; Tomasello and Call, 2007; Genty et al., 2009; Hobaiter and Byrne, 2011a; McCarthy et al., 2012) and the level of understanding or misunderstanding of previous communications (Cartmill and Byrne, 2007). An individual gesture can be used to achieve different goals, or the same goal can be achieved by different gestures (Plooij, 1978; Call and Tomasello, 2007; Graham et al., 2018). Crucially, many of these features are shared with in human language.

Gestures and Intentional Communication Humans also produce a range of signals, including fixed involuntary calls or movements (a yelp when you pick up something that’s too hot, or a flinch when something startles you). Some of these, although involuntary, are susceptible to audience effects – we’re more likely to smile when someone is smiling at us (e.g., Provine, 1992; Wild et al., 2003), or to gesture more expansively when someone can see us (Bavelas et al., 2008). But language is different. We choose whether or not to produce a word or sentence based not only on whether or not someone is there; but on who they are and what they do or don’t know. With language we do more than broadcast information, we intend to communicate a goal or other information to a partner, who we recognise as having their own behaviour, goals, and knowledge. Like language, great ape gestural communication is produced intentionally. That is to say, gestures are not produced as a fixed response to a stimulus (which we would term zero-order intentional production; see Dennett, 1983), but with the aim of altering at least the behaviour (first order intentional - I get you to bring me some water), and perhaps the mental state (second order intentional - I get you to bring me some water because you know I‫׳‬m thirsty) of the recipient (Dennett, 1983; Sebeok, 1996). Grice (1969) believed that everyday language use can operate on at least fourth order intentionality, “thinking about knowing about believing

Cognition j Gestural Communication in the Great Apes

373

about wanting” (I believe that Sarah knows that I didn‫׳‬t buy her a birthday card because I know that she believes birthday cards are a waste of trees, and I don‫׳‬t believe that Sarah thinks that it‫׳‬s because I just forgot), although many exchanges probably take place at a lower order (thank goodness! see Grice, 1969; Millikan, 1984). To date, non-human animal signals that meet the criteria for firstorder-intentional use remain relatively rare. These include a specific alarm call in chimpanzees (Schel et al., 2013b; Crockford et al., 2015, 2017), a potentially referential hunting gesture in grouper fish (Vail et al., 2012), and the only non-human system of intentional signals: the large repertoires of great ape gestures (for example: Bard, 1992; Bard and Vauclair, 1984; Bard et al., 2014; Cartmill and Byrne, 2007, 2010; Fröhlich et al., 2016a; Genty et al., 2009; Graham et al., 2017, 2018; Hobaiter and Byrne, 2011a,b, 2012, 2014; Leavens et al., 1996; Leavens and Hopkins, 1998; Liebal et al., 2006; McCarthy et al., 2012; Perlman et al., 2012; Pika et al., 2003, 2005; Plooij, 1978; Pollick and deWaal, 2007; Tanner and Byrne, 1993, 1999; Tomasello et al., 1985; Tomasello et al., 1989; Roberts et al., 2012a). Over the past 30-years, the rigorous study of gesture in captive great apes has allowed substantial progress in our understanding of the intentional and flexible nature of ape gestures (Bard, 1992; Bard et al., 2014, 2017; Cartmill and Byrne, 2007, 2010; Genty et al, V1.36., 2009; Leavens et al., 1996; Leavens and Hopkins, 1998; Liebal et al,., 2006; McCarthy, Jensvold and Fouts, 2012; Pika et al., 2003, 2005; Pollick and deWaal, 2007; Tanner and Byrne, 1993, 1999; Tomasello et al., 1985, 1989). More recently, with developments in data collection, systematic research on wild populations has become possible (Fröhlich et al., 2016a,b, 2017; Graham et al., 2017, 2018; Hobaiter and Byrne, 2011a,b, 2012, 2014; Hobaiter et al., 2014; Hobaiter et al., 2017; Roberts et al., 2012a,b) allowing scientists to explore ape gesture under the natural conditions that elicit the full expression of the apes’ communicative capacity (Hobaiter and Byrne, 2011a; Seyfarth and Cheney, 2017). Assessing intentional use of a signal is not straightforward (Townsend et al., 2017). We can’t simply ask a non-human species (or, for that matter, a non-verbal person or pre-verbal child) ‘what are you thinking?’ In response to this problem researchers have developed a set of criteria to assess first-order intentionality (Bates et al., 1975; Call and Tomasello, 2007; Cartmill and Byrne, 2007; Gómez, 1994; Hobaiter and Byrne, 2011a,b; Liebal et al., 2004). First, the signal must be directed towards a target audience, showing that the signaller is communicating with a specific individual and not just making random body movements on their own. A mechanically ineffective movement, like raising an arm over your head, is not communicative unless directed towards another individual. Second, the signaller must monitor the visual attentional state of the recipient, i.e., make eye contact or look in their direction (Bates et al., 1975; Gómez, 1994), and adjust gestures accordingly - using a silent visual gesture, such as a wave, only makes sense if the recipient is facing the signaller and can see the movement (Liebal et al., 2004). In human developmental studies, eye contact is a sign of ostension – the signaller wants the recipient to be aware of their communicative attempts (Bates et al., 1975). Like the first criterion, this demonstrates that the signaller is targeting a specific individual and is making the recipient aware of their communication, either by coming into their visual field, by making contact with them, or by using an audible gesture that does not require visual attention to already be present. Third, the signaller must wait for a response Fourth, the signaller should persist or elaborate if their goal is not met (Cartmill and Byrne, 2007; Hobaiter and Byrne, 2011b; Roberts et al., 2012b). These last two criteria show that the signaller aims to alter the behaviour of the recipient in a specific way, waits to receive that reaction, and will continue to communicate unless the reaction they want to happen, happens. Intentional communication is goal-directed communication, and so criteria demonstrating this goal-directedness are imperative. To be included in analysis in most studies, gestures must meet one or more of these intentionality criteria (Fig. 1).

Fig. 1 Great apes have large repertoires of gestures that exploit different modalities of information. All gestures are visible, but some – for example, Big Loud Scratch and Leaf Clip are audible, while others – for example, Bite, Grab, Mouth Stroke, or Tapping other, include physical contact with the recipient. Great apes target their use of gestures to a recipient’s visual attention, employing audible and contact gestures when the recipient is unable to see the signaller.

374

Cognition j Gestural Communication in the Great Apes

The criterion that signallers should persist or elaborate when they are unsuccessful, has been elegantly shown in captive orangutans’ discrimination of different types of recipient misunderstanding (Cartmill and Byrne, 2007). If the gesture that an orang-utan subject used to request a food item was partially successful (the orang-utan received a small piece of the banana but not the whole thing), then they persisted in using that same gesture type. If the gesture was unsuccessful (they didn‫׳‬t get any banana at all); however, they elaborated by using a different gesture type (Cartmill and Byrne, 2007). This use of further gestures after an ‫׳‬incorrect‫׳‬ response clearly demonstrated that the orang-utans were directing their gestures towards a goal. More strikingly, it showed that orang-utans distinguished among different reactions in their target audience and adjusted their gesturing accordingly. If the audience apparently misunderstood their basic aims (to get high value food), they used a completely different gesture type (elaboration); if the audience apparently half-understood (they gave only some of the high value food when the signaller wanted all of it), they continued to use the same gesture type (repetition).

Meaning and Openness in Gesture In language, one word can be used to convey multiple meanings (in English, for example: bank, bark, nail, jam, pool, mine, season, novel, hatch, cast, to name just a few), and the same meaning can be expressed with multiple words (again in English: cheerful, merry, joyful, buoyant, etc.). This openness, or means-ends dissociation, is relatively rare in non-human communication. Vocalizations are typically associated with specific contexts, of which ‫׳‬predation; or predator responses have been extensively studied. In response to detecting a predator, vocalizations encode highly specific information dependent on social relationships, predator type, and proximity to the predator (primate species: Arnold and Zuberbühler, 2006; Cheney and Seyfarth, 1990; Marler, 1961; Ouattara et al., 2009; Schel et al., 2013a; Seyfarth and Cheney, 2003; non-primate species: Manser, 2001; Manser et al., 2002; Templeton et al., 2005; Wilson and Evans, 2012). In contrast ape gestures are regularly employed across varied behavioural contexts, a gesture such as ‫׳‬arm raise‫ ׳‬might be deployed in playing, grooming, feeding, or sexual contexts (Genty et al., 2009; Hobaiter and Byrne, 2011a; Liebal et al., 2006; Pika et al., 2003, 2005; Plooij, 1978; Pollick and de Waal, 2007; Tomasello et al., 1985; Tomasello et al., 1989). Plooij (1978) first reported that the regular combination of gesture types and the use of the same gesture forms in different contexts suggested ‘openness, which is one of the most characteristic design features of human language’. Gestures are used by an ape signaller to achieve a particular goal with a specific recipient. This means that we can ask not only about the information a recipient may be able to decode from the behaviour, but what the signaller ‘means’ to communicate. However, this language-like use of meaning in communication is tricky to explore: we have no direct access to a signaller’s internal state of mind. Most adults have had enough misunderstandings to know that it can be hard to understand what another human means to say, even when we are using the same language. Early explorations of gestural communication employed behavioural context as a proxy for meaning – for example if a gesture was seen during play it would be classified as ‫׳‬meaning‫ ׳‬play, or feeding, or affiliation (e.g. Liebal et al., 2006; Pika et al., 2003, 2005; Tomasello et al., 1985, 1989). With context we can apply established ethograms to categorise on the basis of externally observable behaviour. However, if the signaller employs a gesture that means ‘come here’ or ‘stop that’, these gestures could be employed to convey the same meaning in different behavioural contexts. Inferring meaning from context can therefore give a false impression of ‘flexibility’ in the meaning of a gesture. Later work explored recipient responses as a means of inferring signaller‫׳‬s intention (e.g., Roberts et al., 2012a), but these responses may also include refusal, misunderstanding, or even the start of a negotiation, and so may not reflect the signaller’s desired response. More recently, researchers have employed the apparently satisfactory outcome– the recipient response that stopped the signaller from continuing to signal (Cartmill and Byrne, 2010; Genty et al., 2009; Graham et al.2018; Hobaiter and Byrne, 2014). In any one instance of communication there remains the possibility that there was a miscommunication, after which the signaller gives up. However, across hundreds of cases from different signallers where there is a consistent pattern of association between a gesture and particular outcomes, these indicate the signaller’s intended meaning. Using this approach, gestural communication retained its flexibility. Species’ repertoires are large – containing 80 or more distinct gesture types (Byrne et al., 2017; Genty et al., 2009; Graham et al., 2017; Hobaiter and Byrne, 2011a, 2014; Liebal et al., 2006; Pika et al., 2003, 2005; Tomasello et al., 1985, 1989; Roberts et al., 2012a). However, the number of goals described for ape gestures are relatively small: just 19 in chimpanzees, suggesting substantial redundancy in the system (Hobaiter and Byrne, 2014). The meanings associated with specific gestures appear to be consistent across signallers (Graham et al., 2018; Hobaiter and Byrne, 2014) and even across species, with similar chimpanzee and bonobo gestures being employed to achieve similar goals (Graham et al., 2018). However, the redundancy in the system is not evenly distributed – some gestures express single meanings, and others express a more diverse range. Negations, such as ‘stop that’ or ‘move away’ seem to be associated with a particularly wide range of gesture types, and it is likely that there is further nuance in their usage that remains to be described.

Gesture: The Next Steps Describing gesture types and the meanings for which they are used were crucial first steps, but there remain substantial areas to explore, including revisiting central questions such as ‘what is a gesture?’ To date, most studies have employed relatively arbitrary categories – the way we lump or split ape movements into gesture ‘types’ can result in dramatic differences in the repertoires described (Cartmill and Byrne, 2007; Genty et al, V1.36., 2009; Hobaiter and Byrne, 2011a; Leavens and Hopkins, 1998; Leavens

Cognition j Gestural Communication in the Great Apes

375

et al., 2010; Liebal et al.2006; Pika et al., 2005; Pollick and deWaal, 2007; Tanner and Byrne, 1996; Tomasello et al., 1985; Roberts et al., 2012a,b). Are a ‘hand shake’ and an ‘arm shake’ distinct gestures; or is it the movement ‘shake‫ ׳‬that’s key, irrespective of the limb used? One approach to addressing this question is to use the apes’ own behaviour to discriminate relevant gesture types. These analyses can be achieve with the large data sets of ape gestural behaviour that are now available to researchers. So, in the example above, if the ‫׳‬hand shake‫ ׳‬and ‫׳‬arm shake‫ ׳‬are both used to consistently mean the same thing they can be lumped together, or if there are differences in meaning they can be split (Hobaiter and Byrne, 2017). In a similar fashion, we can explore the use of intentional signals, without necessarily drawing a distinction between what we label as a gesture as opposed to a vocalization. While these distinctions can be useful for some analyses, there are examples that blur the boundaries – non-voiced sounds made with the mouth such as teeth-clacking or blowing ‫׳‬raspberries‫ ׳‬are both under voluntary control and used in social interactions (Leavens et al., 2004; Watts, 2016); and orang-utans modify the sound of a voiced call by using their hands (Hardus et al., 2009).

Gesture in Context While substantial effort has gone into describing gestural repertoires and the meanings for which gestures are used, these studies have focused on the common features of gestural communication as a system. But that system of communication functions in a messy ‘real world’ in which each communication takes place within in a specific social and ecological context (Smith, 1965). Recipients, your relationship with them, their state of attention, behaviour, knowledge, and goals all vary. Dense rainforests restrict lines of sight to just a few meters. This environment limits which gestures you can effectively use to communicate with your intended recipient and – at the same time – allows you to gesture to specific partners without the risk of eavesdropping (Hobaiter and Byrne, 2012; Hobaiter et al., 2017). So while the repertoires of gesture types available to a signaller appear to be share by all members of the species (e.g., Byrne et al., 2017; Genty et al., 2009; Hobaiter and Byrne, 2011a), the development of the use of gestures and their expression by individuals in day-to-day communication is highly flexible (Bard, 1992; Bard et al., 2014, 2017; Fröhlich et al., 2016a,b; Halina et al., 2013; Hobaiter and Byrne, 2011b; Liebal et al., 2006; Perlman et al., 2012; Pika et al., 2003, 2005; Plooij, 1978; Schneider et al., 2012, 2017; Tanner and Byrne, 1993, 1999; Tomasello et al., 1985, 1989). Apes vary their use of gesture depending on both the current recipient, and their history of social interactions with that individual and previous partners (Bard et al., 2014, 2017; Fröhlich et al., 2016b, 2017; Genty et al., 2015; Schneider et al., 2012; Roberts and Roberts, 2016), and the expression of a gesture type can be adjusted to the specific physical and social context of the communication (Bard et al., 2017; Perlman et al., 2012). Like language, gestural communication in great apes combines biologically inherited species-typical features together with substantial flexibility in how these are used in day-to-day communication. However, to date, the majority of our knowledge comes from just a handful of ape social groups, and work in the wild has been heavily biased towards the study of chimpanzees. In human language terms, this would be the equivalent of studying just a handful of communities in just one or two cultures. Just as psychologists today recognise the problems in extrapolating human behaviour from a few W.E.I.R.D cultures (Western, Educated, Industrialised, Rich, Democratic) we must avoid the same trap studying the behaviour of other apes. To fully explore the flexibility and variation of great ape gesture it is vital to expand our understanding across communities and species. Indeed doing so is becoming an urgent priority. With primate populations across the world in steep decline (Estrada et al., 2017), and all great ape species currently classed as Endangered, the loss of any single group is far more than a percentage decline in the population size. It is the loss of individuals and of their group’s culture. As a result, collaborating across research groups, sites, species, and methodologies is about more than adding new pieces to the pile. It’s only when we are able to put those pieces together that we will be able to see the full picture – a road map from which we can explore the mechanisms behind the acquisition and use of gestural communication.

References Arnold, K., Zuberbühler, K., 2006. Language evolution: Sematic combinations in primate calls. Nature 441, 303. Bard, K.A., 1992. Intentional behavior and intentional communication in young free-ranging orang-utans. Child Development 63, 1186–1197. Bard, K.A., Vauclair, J., 1984. The communicative context of object manipulation in ape and human adult-infant pairs. Journal of Human Evolution 13, 181–190. Bard, K.A., Dunbar, S., Maguire-Herring, V., et al., 2014. Gestures and social-emotional communicative development in chimpanzee infants. American Journal of Primatology 76, 14–29. Bard, K.A., Maguire-Herring, V., Tomonaga, M., Matsuzawa, T., 2017. The gesture ‘Touch’: Does meaning-making develop in chimpanzees’ use of a very flexible gesture? Animal Cognition 1–16. https://doi.org/10.1007/s10071-017-1136-0. Bates, E., Camaioni, L., Volterra, V., 1975. The acquisition of performatives prior to speech. Merrill-Palmer Quarterly of Behavior and Development 21, 205–226. Bavelas, J., Gerwing, J., Sutton, C., Prevost, D., 2008. Gesturing on the telephone: Independent effects of dialogue and visibility. Journal of Memory and Language 58, 495–520. Bryan, A.L., 1963. The essential basis for human culture. Current Anthropology 4, 297–306. Butterworth, G., 1998. Origins of joint visual attention in infancy. Monographs of the Society for Research in Child Development 63, 144–166. Byrne, R.W., Cartmill, E., Genty, E., et al., 2017. Great ape gestures: Intentional communication with a rich set of innate signals. Animal Cognition 20, 755–769. Call, J., Tomasello, M., 2007. The gestural repertoire of chimpanzees (Pan troglodytes). In: Call, J., Tomasello, M. (Eds.), The Gestural Communication of Apes and Monkeys. Lawrence Erlbaum Associates, London, UK, pp. 17–40. Carpenter, M., Nagell, K., Tomasello, M., Butterworth, G., Moore, C., 1998. Social cognition, joint attention, and communicative competence from 9 to 15 months of age. Monographs of the Society for Research in Child Development 63 (4) (i-174).

376

Cognition j Gestural Communication in the Great Apes

Cartmill, E., Byrne, R.W., 2007. Orangutans modify their gestural signaling according to their audience’s comprehension. Current Biology 17, 1345–1348. Cartmill, E., Byrne, R.W., 2010. Semantics of primate gestures: Intentional meanings of orangutan gestures. Animal Cognition 13, 793–804. Cheney, D.L., Seyfarth, R.M., 1990. The representation of social relations by monkeys. Cognition 37, 167–196. Crockford, C., Wittig, R.M., Zuberbühler, K., 2015. An intentional vocalization draws others’ attention: A playback experiment with wild chimpanzees. Animal Cognition 18, 581–591. Crockford, C., Wittig, R.M., Zuberbühler, K., 2017. Vocalizing in chimpanzees is influence by social-cognitive processes. Science Advances 3, e1701742. Dennett, D.C., 1983. Intentional systems in cognitive ethology. The Brain and Behavioural Sciences 6, 343–390. Deschner, T., Heistermann, M., Hodges, K., Boesch, C., 2003. Timing and probability of ovulation in relation to sex skin swelling in wild West African chimpanzees, Pan troglodytes verus. Animal Behaviour 66, 551–560. Diamond, J., 1991. The Third Chimpanzee: The Evolution and Future of the Human Animal. Hutchison Radius, London. Dunn, J.C, Smaers, J.B, 2018DunnandSmaers. Neural Correlates of Vocal Repertoire in Primates. Frontiers in Neuroscience. https://doi.org/10.3389/fnins.2018.00534. Estrada, A., Garber, P.A., Rylands, A.B., et al., 2017. Impending extinction crisis of the world’s primates: Why primates matter. Science Advances 3 (1), e1600946. Fröhlich, M., Wittig, R.M., Pika, S., 2016a. Play-solicitation gestures in chimpanzees in the wild: Flexible adjustment to social circumstances and individual matrices. Royal Society Open Science 3, 160278. Fröhlich, M., Kuchenbuch, P., Müller, G., et al., 2016b. Unpeeling the layers of language: Bonobos and chimpanzees engage in cooperative turn-taking sequences. Scientific Reports 6, 25887. Fröhlich, M., Müller, G., Zeiträg, C., Wittig, R.M., Pika, S., 2017. Gestural development of chimpanzees in the wild: The impact of interactional experience. Animal Behaviour 134 (2017), 271–282. https://doi.org/10.1016/j.anbehav.2016.12.018. Gardner, R.A., Gardner, B.T., 1969. Teaching sign language to a chimpanzee. Science 165, 664–672. Genty, E., Breuer, T., Hobaiter, C., Byrne, R.W., 2009. Gestural communication of the gorilla (Gorilla gorilla): Repertoire, intentionality and possible origins. Animal Cognition 12, 527–546. Genty, E., Neumann, C., Zuberbühler, K., 2015. Bonobos modify communication signals according to recipient familiarity. Scientific Reports 5, 16442. Gómez, J.C., 1994. Mutual awareness in primate communication: A Gricean approach. In: Parker, S.T., Mitchell, R.W., Boccia, M.L. (Eds.), Self-Awareness in Animals and Humans. Cambridge University Press, Cambridge, pp. 61–80. Goodall, J., 1968. The behaviour of free-living chimpanzees in the Gombe Stream Reserve. Animal Behaviour Monographs 1, 161–311 (IN12). Goodall, J., 1986. The Chimpanzees of Gombe: Patterns of Behavior. The Belknap Press of Harvard University Press, Cambridge, MA. Goodall, J., 1998. Essays on science and society: Learning from the chimpanzees: A message humans can understand. Science 282, 2184–2185. Graham, K.E., Furuichi, T., Byrne, R.W., 2017. The gestural repertoire of the wild bonobo (Pan paniscus): A mutually understood communication system. Animal Cognition 20, 171–177. Graham, K.E., Hobaiter, C., Ounsley, J., Furuichi, T., Byrne, R.W., 2018. Bonobo and chimpanzee gestures overlap extensively in meaning. PLOS Biology 16, e2004825. Grice, H.P., 1969. Utterer’s meaning and intention. The Philosophical Review 78, 147–177. Gunderson, E.A., Spaepen, E., Gibson, D., Goldin-Meadow, S., Levine, S.C., 2015. Gesture as a window onto children’s number knowledge. Cognition 144, 14–28. Halina, M., Rossano, F., Tomasello, M., 2013. The ontogenetic ritualization of bonobo gestures. Animal Cognition 16, 653–666. Hardus, M.E., Lameira, A.R., Van Schaik, C.P., Wich, S.A., 2009. Tool use in wild orang-utans modifies sound production: A functionally deceptive innovation? Proceedings of the Royal Society of London B: Biological Sciences 276 (1673). https://doi.org/10.1098/rspb20091027. Hauser, M.D., Wrangham, R.W., 1987. Manipulation of food calls in captive chimpanzees. Folia Primatologica 48, 207–210. Hayes, C., 1951. The Ape in our House. Harper, Oxford, England. Hayes, K.J., Hayes, C., 1951. The intellectual development of a home-raised chimpanzee. Proceedings of the American Philosophical Society 95, 105–109. Hobaiter, C., Byrne, R.W., 2011a. The gestural repertoire of the wild chimpanzee. Animal Cognition 14, 745–767. Hobaiter, C., Byrne, R.W., 2011b. Serial gesturing by wild chimpanzees: Its nature and function for communication. Animal Cognition 14, 827–838. Hobaiter, C., Byrne, R.W., 2012. Laterality in the gestural communication of wild chimpanzees. Annals of the New York Academy of Sciences 1288, 9–16. Hobaiter, C., Byrne, R.W., 2014. The meanings of chimpanzee gestures. Current Biology 24, 1596–1600. Hobaiter, C., Byrne, R.W., 2017. What is a gesture? A meaning based approach to defining gestural repertoires. Neuroscience and Biobehavioural Reviews 82, 3–12. Hobaiter, C., Leavens, D.A., Byrne, R.W., 2014. Deictic gesturing in wild chimpanzees (Pan troglodytes)? Some possible cases. Journal of Comparative Psychology 128 (1), 82–87. https://doi.org/10.1037/a0033757. Hobaiter, C., Byrne, R.W., Zuberbühler, K., 2017. Wild chimpanzees’ use of single and combined vocal and gestural signals. Behavioural Ecology and Sociobiology 71, 96. Hoetjes, M., Krahmer, E., Swerts, M., 2015. On what happens in gesture when communication is unsuccessful. Speech Communication 72, 160–175. Kellogg, W.N., Kellogg, L.A., 1933. The Ape and the Child: A Study of Environmental Influence Upon Early Behavior. Whittlesey House. Kendon, A., 1997. Gesture. Annual Review of Anthropology 26, 109–128. Ladygina-Kohts, N.N., 1935/2002/. Infant Chimpanzee and Human Child: A Classic 1935 Comparative Study of Ape Emotions and Intelligence, translated by Borris Wekker. Oxford University Press, Oxford, UK. Leavens, D.A., Hopkins, W.D., Bard, K.A., 1996. Indexical and referential pointing in chimpanzees (Pan troglodytes). Journal of Comparative Psychology 110, 346–353. Leavens, D.A., Hopkins, W.D., 1998. Intentional communication by chimpanzees: A cross-sectional study of the use of referential gestures. Developmental Psychology 34, 813–822. Leavens, D.A., Hostetter, A.B., Wesley, M.J., Hopkins, W.D., 2004. Tactical use of unimodal and bimodal communication in chimpanzees, Pan troglodytes. Animal Behaviour 67, 467–476. Leavens, D.A., Russell, J.L., Hopkins, W.D., 2010. Multimodal communication by captive chimpanzees (Pan troglodytes). Animal Cognition 13, 33–44. Leavens, D.A., Bard, K.A., Hopkins, W.D., 2017. The mismeasure of ape social cognition. Animal Cognition 1–18. https://doi.org/10.1007/s10071-017-1119-1. Liebal, K., Call, J., Tomasello, M., 2004. The use of gesture sequences in chimpanzees. American Journal of Primatology 64, 377–396. Liebal, K., Pika, S., Tomasello, M., 2006. Gestural communication of orang-utans (Pongo pygmaeus). Gesture 6, 1–36. Manser, M.B., 2001. The acoustic structure of suricates’ alarm calls varies with predator type and the level of response urgency. Proceedings of the Royal Society of London B: Biological Sciences 268, 2315–2324. Manser, M.B., Seyfarth, R.M., Cheney, D.L., 2002. Suricate alarm calls signal predator class and urgency. Trends in Cognitive Sciences 6, 55–57. Marler, P., 1961. The logical analysis of animal communication. Journal of Theoretical Biology 1, 295–317s. McCarthy, M.S., Jensvold, M.-L.A., Fouts, D.H., 2012. Use of gesture sequences in captive chimpanzee (Pan troglodytes) play. Animal Cognition 16 (3), 471–481. https://doi.org/ 10.1007s/10071-012-0587-6. Millikan, R.G., 1984. Language, Thought, and Other Biological Categories: New Foundations for Realism. MIT Press. Ouattara, K., Lemasson, A., Zuberbühler, K., 2009. Campbell’s monkeys use affixation to alter call meaning. PLOS ONE 4, e7808. Patterson, F.G., 1978. The gestures of a gorilla: Language acquisition in another pongid. Brain and Language 5, 72–97. Perlman, M., Tanner, J.E., King, B.J., 2012. A mother gorilla’s variable use of touch to guide her infant. In: Pika, S., Liebal, K. (Eds.), Developments in Primate Gesture Research. John Benjamins Publishing, Amsterdam, p. 55. Pika, S., Liebal, K., Tomasello, M., 2003. Gestural communication in young gorillas (Gorilla gorilla): Gestural repertoire, learning and use. American Journal of Primatology 60, 95–111.

Cognition j Gestural Communication in the Great Apes

377

Pika, S., Liebal, K., Tomasello, M., 2005. Gestural communication in subadult bonobos (Pan paniscus): Repertoire and use. American Journal of Primatology 65, 39–61. Plooij, F.X., 1978. Some basic traits of human language in wild chimpanzees. In: Locke, A. (Ed.), Action, Gesture and Symbol: The Emergence of Language. Academic Press, London, UK. Pollick, A.S., de Waal, F.B.M., 2007. Ape gestures and language evolution. Proceedings of the National Academy of Sciences USA 104, 8184–8189. Provine, R.R., 1992. Contagious laughter: Laughter is a sufficient stimulus for laughs and smiles. Bulletin of the Psychonomic Society 30, 1–4. Reichert, K.E., Heistermann, M., Hodges, K.J., Boesch, C., Hohmann, G., 2002. What females tell males about their reproductive status: Are morphological and behavioural cues reliable signals of ovulation in bonobos (Pan paniscus)? Ethology 108, 583–600. Roberts, A.I., Vick, S.-J., Buchanan-Smith, H., 2012a. Usage and comprehension of manual gestures in wild chimpanzees. Animal Behaviour 84, 459–470. Roberts, A.I., Vick, S.-J., Buchanan-Smith, H., 2012b. Communicative intentions in wild chimpanzees: Persistence and elaboration in gestural signalling. Animal Cognition 16 (2). https://doi.org/10.1007/s10071-012-0563-1. Roberts, A.I., Roberts, S.G.B., 2016. Wild chimpanzees modify modality of gestures according to the strength of social bonds and personal network size. Scientific Reports 6, 33864. Savage-Rumbaugh, E.S., Lewin, R., 1994. Kanzi: The Ape at the Brink of the Human Mind. Wiley. Savage-Rumbaugh, E.S., Murphy, J., Sevcik, R.A., et al., 1993. Language comprehension in ape and child. Monographs of the Society for Research in Child Development 58 (3–4), 1–222 (i-252). Savage-Rumbaugh, S., McDonald, K., Sevcik, R.A., Hopkins, W.D., Rubert, E., 1986. Spontaneous symbol acquisition and communicative use by pygmy chimpanzees (Pan paniscus). Journal of Experimental Psychology: General 115, 211. Schaller, G.E., 1963. The Mountain Gorilla: Ecology and Behaviour. Chicago Press, Oxford, England. Schel, A.M., Townsend, S., Machanda, Z., Zuberbühler, K., Slocombe, K., 2013a. Chimpanzee food calls are directed at specific individuals. Animal Behaviour 86, 955–965. Schel, A.M., Townsend, S., Machanda, Z., Zuberbühler, K., Slocombe, K., 2013b. Chimpanzee alarm call production meets key criteria for intentionality. PLOS ONE 8, e76674. Schneider, C., Call, J., Liebal, K., 2012. What role do mothers play in the gestural acquisition of bonobos (Pan paniscus) and chimpanzees (Pan troglodytes)? International Journal of Primatology 33, 246–262. Schneider, C., Call, J., Liebal, K., 2017. ‘Giving’ and ‘responding’ difference in gestural communication between nonhuman great ape mothers and infants. Developmental Psychobiology 59, 303–313. Sebeok, T.A., 1996. Signs: An Introduction to Semiotics. University of Toronto Press, Toronto. Seyfarth, R.M., Cheney, D.L., 2003. Signalers and receivers in animal communication. Annual Review of Psychology 54, 145–173. Seyfarth, R.M., Cheney, D.L., 2017. The origin of meaning in animal signals. Animal Behaviour 124, 339–346. Smith, W.J., 1965. Message, meaning, and context in ethology. The American Naturalist 99, 405–409. Tanner, J.E., Byrne, R.W., 1996. Representation of action through iconic gesture in a captive lowland gorilla. Current Anthropology 37, 162–173. Tanner, J.E., Byrne, R.W., 1999. Spontaneous gestural communication in captive lowland gorillas. In: Parker, S.T., Mitchell, R.W., Miles, L.H. (Eds.), The Mentalities of Gorillas and Orang-Utans in Comparative Perspective. Cambridge University Press, Cambridge, pp. 211–240. Templeton, C.N., Greene, E., Davis, K., 2005. Allometry of alarm calls: Black-capped chickadees encode information about predator size. Science 308, 1934–1937. Terrace, H.S., Petitto, L.A., Sanders, R.J., Bever, T.G., 1979. Can an ape create a sentence? Science 206, 891–902. Tomasello, M., Call, J., 2007. Introduction: Intentional communication in nonhuman primates. In: Call, J., Tomasello, M. (Eds.), The Gestural Communication of Apes and Monkeys. Lawrence Erlbaum Associates, London, pp. 1–15. Tomasello, M., George, B.L., Kruger, A.C., Jeffrey, M., Evans, A., 1985. The development of gestural communication in young chimpanzees. Journal of Human Evolution 14, 175–186. Tomasello, M., Gust, D., Frost, G.T., 1989. A longitudinal investigation of gestural communication in young chimpanzees. Primates 30, 35–50. Tomasello, M., Carpenter, M., Liszkowski, U., 2007. A new look at infant pointing. Child Development 78, 705–722. Townsend, S.W., Koski, S.E., Byrne, R.W., et al., 2017. Exorcising Grice’s ghost: An empirical approach to studying intentional communication in animals. Biological Reviews 92, 1427–1433. Umiker-Sebeok, J., Sebeok, T.A., 1981. Clever Hans and smart simians: The self-fulfilling prophecy and kindred methodological pitfalls. Anthropes 76, 89–165. Vail, A.L., Manica, A., Bshary, R., 2012. Referential gestures in fish collaborative hunting. Nature Communications 4, 1765. Waller, B.M., Caeiro, C.C., Davila-Ross, M., 2015. Orangutans modify facial displays depending on recipient attention. PeerJ 3, e827. Watts, D.P., 2016. Production of grooming-associated sounds by chimpanzees (Pan troglodytes) at Ngogo: Variation, social learning, and possible functions. Primates 57, 61–72. Wild, B., Erb, M., Eyb, M., Bartels, M., Grodd, W., 2003. Whey are smiles contagious? An fMRI study of the interaction between perceptions of facial affect and facial movements. Psychiatry Research: Neuroimaging 123, 17–36. Wilkins, D., 2003. Why pointing with the index finger is not a universal (in sociocultural and semiotic terms). In: Kita, S. (Ed.), Pointing: Where Language, Culture, and Cognition Meet. Lawrence Erlbaum Associates, Mahwah, NJ, pp. 171–215. Wilson, D.R., Evans, C.S., 2012. Fowl communicate the size, speed, and proximity of avian stimuli through graded structure in referential alarm calls. Animal Behaviour 83, 535–544. Witmer, L., 1909. A monkey with a mind. Psychological Clinic 3 (7), 179–205.

Relevant Website http://greatapedictionary.ac.uk/video-resources/gesture-videos/–The Great Ape Dictionary.

Mental Time Travel: Can Animals Recall the Past and Plan for the Future? NS Clayton and A Dickinson, University of Cambridge, Cambridge, UK © 2010 Elsevier Ltd. All rights reserved.

Abstract According to the mental time travel hypothesis, only humans can mentally dissociate themselves from the present, traveling backward in time to recollect specific past events about what happened where and when (episodic memory) and traveling forward in time to anticipate future needs (future planning). A series of studies of the mnemonic capabilities of food-caching western scrub-jays question this assumption. In terms of the retrospective component of episodic memory, these birds remember the ‘what, where, and when’ of specific past caching episodes; they keep track of how long ago they cached different types of perishable foods that decay at different rates, and also remember whether another individual was present at the time of caching, and if so, which bird was watching when. Recent work demonstrates that the jays also make provision for a future need, caching more food in places in which they will not be given breakfast the next morning than in places where they will be receive breakfast the next morning even though there is plenty of food available to them at the time when they cache the food. Taken together these results challenge the mental time travel hypothesis by showing that some elements of both retrospective and prospective mental time travel appear not to be uniquely human.

Keywords Autonoesis; Chimpanzees; Chronesthesia; Corvids; Episodic-like memory; Episodic memory; Future planning; Mental time travel; Western scrub-jays

Introduction In an influential paper that was published in 1997, Suddendorf and Corballis argued that we humans are unique among the animal kingdom in being able to mentally dissociate ourselves from the present. To do so, we travel backwards and forwards in the mind’s eye to remember and reexperience specific events that happened in the past (episodic memory) and to anticipate and preexperience future scenarios (future Planning). Although physical time travel remains a fictional conception, mental time Travel is something we do for a living, and the fact that we spend so much of our time thinking about the past and the future led to Mark Twain’s witty remark that “my life has been filled with many tragedies, most of which never occurred.” Mental time travel then has two components: a retrospective one in the form of episodic memory and a prospective one in the form of future planning. In formulating their mental time travel hypothesis, Suddendorf and Corballis were the first to suggest that episodic memory and future planning are intimately linked and can be viewed as two sides of the same coin so to speak. In fact, their proposal consisted of two claims. In addition to integrating the retrospective and prospective components of mental time travel, they also argued that such abilities were unique to humans and reflected a striking cognitive dichotomy between ourselves and other animals. The latter idea was not new, however, but rather an extension of what others have argued makes episodic memory special. Indeed in his seminal studies of human memory, Tulving coined the term episodic memory in 1972 to refer to the recollection of specific personal happenings, a form of memory that he claimed was uniquely human and fundamentally distinct from semantic memory, the ability to acquire general factual knowledge about the world, which he argued we share with most, if not all, animals. Ever since he made this remember–know distinction, most cognitive psychologists and neuroscientists have assumed that episodic memory is special because of the experiential nature of these memories, namely that our episodic reminiscences are accompanied by a subjective awareness of currently reexperiencing an event that happened in the past, as opposed to just knowing that it happened. Of course we also have many instances of knowledge acquisition in which we do not remember the episode in which we acquired that information. For example, although most of us know when and where we were born, we do not remember the birth itself nor the episode in which we were told when our birthday is, and therefore such memories are classified as semantic as opposed to episodic. Episodic and semantic memory, then, are thought to be marked by two separate states of awareness; episodic remembering requires an awareness of reliving the past events in the mind’s eye and of mentally traveling back in one’s own mind’s eye to do so, whereas semantic knowing only involves an awareness of the acquired information without any need to travel mentally back in time to personally reexperience the past event. It is for this reason that in later writings, Tulving has argued that one of the cardinal features of episodic memory is that it operates in ‘subjective time,’ and he refers to the awareness of such subjective time as chronesthesia. Language-based reports of episodic recall suggest that the retrieved experiences are not only explicitly located in the past but are also accompanied by the conscious experience of one’s recollections, feeling that one is the author of the memory, or of traveling back not in any mind’s eye but in my mind’s eye, what Tulving called autonoetic consciousness. In other words, Tulving and others argue that episodic memory differs from semantic memory not only in being oriented to the past, but specifically in the past of

378

Encyclopedia of Animal Behavior, 2nd edition, Volume 1

https://doi.org/10.1016/B978-0-12-813251-7.00094-8

Cognition j Mental Time Travel: Can Animals Recall the Past and Plan for the Future?

379

the owner of that memory. So while some semantic knowledge, such as the birth date example described earlier, does involve a datable occurrence, these memories are fundamentally distinct from episodic memories because they do not require any mental time travel. As William James so aptly wrote “Memory requires more than the mere dating of a fact in the past. It must be dated in my past.” From a biological perspective, the characterization of episodic memory in terms of these two phenomenological properties of consciousness, namely autonoesis and chronesthesia, presents major problems for two reasons. The first is that positing a subjective state of awareness is difficult to integrate with evolutionary processes of natural and sexual selection, which operates on behavioral attributes such as reproductive success and survival rather than on mental states. The second is that this definition makes it impossible to test in nonverbal animals, in the absence any agreed behavioral markers of non-Linguistic consciousness. Adopting an ethological approach to comparative cognition necessitates two requirements. The first is that the memory needs to be characterized in terms of behaviorally defined properties as opposed to purely phenomenological ones, such as the types of information encoded. Indeed, we shall argue that the ability to remember what happened, where and how long ago is a critical behavioral criterion for episodic memory. The second requirement is the identification of an ethological context in which these memories would confer a selective advantage. Note that by doing so, we transform this debate about the human uniqueness of mental time travel into an empirical evaluation in non-Linguistic animals as opposed to restricting it to the realms of philosophical personal ponderings. But before doing so, let us return to the two claims made by the mental time travel hypothesis: (1) future planning and episodic memory are subserved by a common process, mental time travel, and (2) this process is uniquely human. We shall evaluate each of these claims in turn, and argue that there is good evidence to support the first claim, but that considerably more controversy surrounds the second component of Suddendorf and Corballis’ thesis. Evidence to support the first claim comes from a number of sources. First, studies of brain activity while engaged in either memory retrieval or future-oriented tasks identify a specific core network of regions in the brain of healthy human adults that support both episodic recollection and future planning. Moreover, there are patients such as DB and KC, who show specific impairments in episodic but not semantic memory, and these patients have similar deficits in episodic but not semantic forethought. Finally, studies of cognitive development in young children suggest that episodic memory and future planning both emerge at about the same age, and are not properly developed until children reach the age of about four.

Is Mental Time Travel Unique to Humans? Regarding the second claim about the uniqueness of episodic memory and future planning, if we are to adopt an ethological approach of the form we outlined earlier, then the question becomes one of asking where in the natural world these two processes might intersect, in which species, and under what conditions. One classic candidate is the food-caching behavior of corvids, members of the crow family that include jays, magpies, and ravens as well as the crows. These large-brained, long-lived, and highly social birds hide food caches for future consumption, and rely on memory to recover their caches of hidden food at a later date, typically weeks if not months into the future. So clearly food-caching is a behavior that is oriented toward future needs. Indeed, the act of hiding food is without obvious immediate benefit and yields its return only when the bird comes to recover the caches it made. Given that these birds are dependent on finding a significant number of these caches for survival in the wild, it seems likely that the selection pressure for an excellent memory for the caches would have been particularly strong, especially as they cache year round. These birds also cache reliably in the laboratory, providing both ethological validity and experimental control. At issue, however, is whether or not these birds episodically remember the past and plan for the future. For these reasons, we shall now turn our attention to assessing the evidence as to whether or not these food-caching corvids can remember the past and plan for the future.

Episodic Memory As we stated earlier, language-based reports of episodic recall in humans suggest that the retrieved experiences are not only explicitly located in the past but are also accompanied by the conscious experience of one’s recollections. From a comparative perspective, the problem with this definition, however, is that in the absence of agreed non-Linguistic markers of consciousness, it is not clear how one could ever test whether animals are capable of episodic recollection. For how would one assess whether or not an animal can experience an awareness of the passing of time and of reexperiencing one’s own memories while retrieving information about a specific past event.

Behavioral criteria for episodic memory This dilemma can be resolved to some degree, however, by using Tulving’s original definition of episodic memory, in which he identified episodic recall as the retrieval of information about ‘where’ a unique event occurred, ‘what’ happened during the episode, and ‘when’ it took place. The advantage of using this definition is that the simultaneous retrieval and integration of information about these three features of a single, unique experience may be demonstrated behaviorally in animals. Clayton and Dickinson termed this ability ‘episodic-like memory’ rather than episodic memory because we have no way of knowing whether or not this form of remembering is accompanied by the autonoetic and chronesthetic consciousness that accompanies human episodic

380

Cognition j Mental Time Travel: Can Animals Recall the Past and Plan for the Future?

recollections. Indeed, we have argued that the ability to remember the ‘what-where-and-when’ of unique past episodes is the hallmark of episodic memory that can be tested in animals.

Empirical tests of episodic-like memory We focus our analysis on one particular species of food-caching corvid, the western scrub-jay, capitalizing on one feature of their ecology, namely, the fact that these birds cache perishable foods, such as worms, as well as nondegradable nuts, and as they do not eat rotten items, recovering perishable food is only valuable as long as the food is still fresh. In a classic experiment published in 1998, we tested whether the jays could remember the ‘what, where, and when’ of specific caching events. Although the birds had no cue predicting whether or not the worms had perished other than the passage of time that had elapsed between the time of caching and the time at which the birds could recover the caches they had hidden previously, the birds rapidly learned that highly preferred worms were fresh and still delicious when recovered 4 h after caching, whereas after 124 h, the worms had decayed and tasted unpleasant. Consequently, the birds avoided the wax worm caches after the longer retention interval and instead recovered exclusively peanuts, which never perish. Following experience with caching and recovering worms and peanuts after the short and long intervals, probe tests, in which the food was removed prior to recovery, showed that they relied on memory to do so rather than cues emanating directly form the food. Subsequent tests revealed that the jays could remember which perishable foods they have hidden where and how long ago, and irrespective of whether the foods decayed or ripened. Since the initial studies, a number of other laboratories have also turned their attention to the question of whether or not animals have episodic-like memory. Using paradigms analogous to those employed with the jays, there is now good evidence that rats, mice, and magpies can remember the what-where-and-when and what-where-and-which of past events.

Forethought If forethought, at least in the form of episodic future thinking, falls under the general umbrella of mental time travel and is the reason for why episodic memory evolved in the first place as we suggested in the introduction, then we should expect to find a concomitant development of episodic memory and episodic future thinking. So if one accepts the evidence that the scrub-jays can episodically recall the past, at least in terms of the behavioral criteria, then these birds should also be capable of planning for the future. The topic is of course a controversial one, and indeed there is much debate about whether non-human animals are capable of forethought (see, e.g., the arguments of Suddendorf and Corballis, and the responses from my laboratory). For how does one test whether the jays’ caching decisions are controlled by future planning?

Behavioral criteria for future planning The first distinction that one must draw is between prospectively oriented behavior and future planning. Several anticipatory activities, including migration, hibernation, nest building, and food-caching, are clearly conducted for a future benefit as opposed to a current one, but they would not constitute a case of future planning unless one could demonstrate the flexibility underlying cognitive control, and thereby rule out simpler accounts in terms of behavior triggered by seasonal cues or previous reinforcement of the anticipatory act. So the first issue to address is whether the caching behavior of the jays is sensitive to its consequences. To do so, once again we capitalized on the fact that the jays love to eat and cache fresh worms but that they do not eat them once they have degraded. We used a variant of the Clayton and Dickinson (1998) caching paradigm in which the jays were given fresh worms and nuts to cache before recovering them 2 days later. In contrast to the original experiments on episodic-like memory, in which the state of the worm caches varied with the retention interval, in the future-planning experiment, the worms were always degraded at recovery in order to investigate their choice of what to cache, as opposed to where to search at recovery. The objective of this experiment was to assess whether or not the birds could learn that even though the worms were fresh at the time of caching there was no point in caching them because they would always be degraded and therefore unpalatable at the time of recovery. The jays rapidly learned to stop caching the worms, even though they continued to eat the fresh worms at the time of caching, thereby demonstrating that caching is indeed selective to its consequences in the sense that the jays could learn what not to cache.

The Bischof–Ko¨hler hypothesis Suddendorf and Corballis have also argued that a critical feature of future planning is that the subject can take action in the present for a future motivational need, independent of the current motivation. Indeed, they argued that mental time travel provided a profound challenge to the motivational system in requiring the subject to suppress thoughts about one’s current motivational state in order to allow one to imagine future needs, and to dissociate them from current desires. To illustrate this distinction between current and future motivational states, consider the following example. A current desire for a croissant at breakfast may lead to an early morning trip to the local baker. Of course it will take some time to reach the market, and therefore the croissant will not be eaten now but in a few minutes time. But although the croissant will be eaten at a future time as opposed to the present, this behavior would not fulfill the Bischof–Köhler criterion because the action is governed by one’s current motivational state. By contrast, going to the baker’s shop in order to ensure that there are croissants for tomorrow’s Sunday brunch would be an example of the future planning envisaged by the Bischof–Köhler hypothesis because this action would be performed for a future motivational need, independent of one’s current needs.

Cognition j Mental Time Travel: Can Animals Recall the Past and Plan for the Future?

381

This hypothesis was inspired by a comparative perspective, from reviewing the evidence for human and non-human primate cognition, and indeed it has led to a number of tests of whether animals can dissociate current from future motivational needs. In one study to address this issue, Naqshbandi and Roberts gave squirrel monkeys the opportunity to choose between eating four dates and eating just one date. Eating dates makes monkeys thirsty, but rather than asking the monkeys to chose between water and the dates, the experimenters manipulated the delay between the choice (one vs. four dates) and receiving water such that the monkeys received water after a shorter delay if they had chosen the one date rather than the four dates. The monkeys gradually reversed their natural preference for four dates, suggesting that they were anticipating their future thirst. However, because the monkeys received repeated trials in which they learnt the consequences of their choices, one can give a simple associative explanation in terms of reinforcement of the anticipatory act by avoidance of the induction of thirst. More convincing evidence for a dissociation of current and future motivational states comes from a study by Correia, Dickinson, and Clayton on the food-caching scrub-jays. Like many other animals, when sated of one type of food, these birds prefer to eat and cache another type of food. Correia and colleagues capitalized on this specific satiety effect to test whether the birds would choose to cache the food they want now or the food they think they will want when they come to recover their caches in the future. In the critical group, the birds were sated on one of two foods that were both then made available for caching. Then, immediately prior to the recovery of these caches, they were sated on the other food. Consequently, the food that was valuable at recovery was the one that was less valuable at the time of caching. At the beginning of the experiment, the birds cached the food they desired at the time, but then rapidly switched to storing preferentially the food that was valuable at the time of recovery rather than the one they wanted to eat at the time of caching, suggesting that the jays can plan future actions on the basis of what they anticipate they will desire in the future as opposed to what they need now. So this study supports the notion that jays can dissociate future from current motivational needs, and therefore provides direct evidence to challenge the Bischof–Köhler hypothesis (for further discussion see our recent review in Animal Behaviour). For the skeptic, however, this kind of task need not require prospective mental time travel because the scrub-jay does not need to imagine a future situation. Suppose that the act of recovering a particular food recalls the episode of caching that food. If the bird is hungry for that particular food, then recovering it will be rewarding and therefore this could directly reinforce the act of caching the food through the memory of doing so. The point is that such memory-mediated reinforcement does not require the bird to envisage future motivational states.

Tulving’s spoon test Tulving has argued that it is possible to test whether animals are capable of such episodic future thinking, and devised what he calls the ‘spoon test,’ which he argues is a ‘future-based test of autonoetic consciousness that does not rely on and need not be expressed through language.’ The test is based on an Estonian children’s story tale, in which a young girl dreams about going to a birthday party. In the dream, all of her friends are eating a delicious chocolate mousse, which is her favorite pudding, but alas she cannot because she does not have a spoon with her, and no one is allowed to eat the pudding unless they have their own spoon. As soon as she gets home she finds a spoon in the kitchen, carry it up to her bedroom and hides – or caches – it under her pillow, in preparation for future birthday parties and even dreams of future birthday parties for that matter. The point then is to use past experience to take action now for an imagined future event. To pass the spoon test, an animal must act analogously to the little girl carrying her own spoon to a new party, a spoon that has been obtained in another place and at another time. Is there any evidence that animals and young children can pass this spoon test? Although some animals, notably primates and corvids (namely the scrub-jays we discussed earlier), have been shown to take actions now based on their future consequences, most of these studies have not shown that an action can be selected with reference to future motivational states independent of current needs as discussed in the previous section. Mulcahy and Call were the first to devise a spoon test for animals. In their study a variety of species of non-human apes were first taught to use a tool to obtain a food reward that would otherwise have been out of reach, before being given the opportunity to select a tool from the experimental room, which they could carry into the sleeping room for use the following morning. Although most of the subjects did choose the correct tool on some trials, the individual patterns of success for each subject was not consistent across subsequent trials, as one would expect if they had a true understanding of the task. Furthermore, the apes received a number of training trials, so reinforcement of the anticipatory act cannot be ruled out. A more convincing case of planning was provided by Osvath and Osvath. In a recent series of experiments, these authors demonstrated that when selecting a tool for use in the future, chimpanzees and orangutans can override immediate drives in favor of future needs. One of the most striking examples of the spoon test in animals comes from recent studies of the food-caching scrub-jays. In the laboratory, work by Raby and colleagues showed that our jays can spontaneously plan for tomorrow’s breakfast without reference to their current motivational state. The birds were given the opportunity to learn that they received either no food or a particular type of food, for breakfast in one compartment, while receiving a different type of food for breakfast in an alternative compartment. Having been confined to each compartment at breakfast time for an equal number of times, the birds were unexpectedly given the opportunity to cache food in both compartments one evening, at a time when there was plenty of food for them to eat and therefore no reason for them to be hungry. Given that the birds did not know which compartment they would find themselves in at breakfast tomorrow and on the assumption that they prefer a variety of foods for breakfast, we predicted that if they could plan for the future, then they should cache a particular food in the compartment in which they had not previously had it for breakfast. This the birds did, suggesting that they could anticipate their future desires at breakfast time tomorrow when they would be hungry. Importantly, because the birds had not been given the opportunity to cache during training, we can in this experiment

382

Cognition j Mental Time Travel: Can Animals Recall the Past and Plan for the Future?

rule out an explanation in terms of mediated reinforcement of the anticipatory act. These findings led Shettleworth to argue that “two requirements for genuine future planning are that the behavior involved should be a novel action or combination of actions and that it should be appropriate to a motivational state other than the one the animal is in at that moment . Raby et al. describe the first observations that unambiguously fulfill both requirements.” Although it seems clear that the scrub-jays and chimpanzees do pass the spoon test, at issue, however, is whether or not these tasks truly tap episodic future thinking. Indeed, we have argued that in the absence of language, there is no way of knowing whether the jays’ ability to plan for future breakfasts reflects episodic future thinking, in which the jay projects itself into tomorrow morning’s situation, or semantic future thinking, in which the jays acts prospectively but without personal mental time travel into the future. In the latter case, all that the subject has to do is to work out what has to be done to ensure that the implement is to hand, be it a spoon, some other tool, or a food-cache. In no sense does this task require the subject to imagine or project one’s self into possible future episodes or scenarios. As Raby et al. have argued, however, what these studies do demonstrate is the capacity of animals to plan for a future motivational state that stretches over a timescale of at least tomorrow, thereby challenging the assumption that this ability to anticipate and act for future needs evolved only in the hominid lineage.

Concluding Remarks The mental time travel hypothesis of Suddendorf and Corballis makes two claims. We have argued that the first claim that episodic memory and future planning are intimately linked and subserved by the same common process of mental time travel has good support. However, we challenge the second claim about human uniqueness. Indeed, we have argued that at least some animals, notably a few primates and corvids, are capable of recollecting the past and planning for the future. In the case of the scrub-jays, the functional account of caching appears to be reflected in the psychological processes underlying this behavior; by fulfilling the behavioral criteria we have outlined, they therefore show at least some elements of episodic memory and forethought. It also serves as a superb illustration of the integration of the retrospective and prospective components of mental time travel for there is no benefit to the animal of hiding food at the time of caching. The benefit occurs when recovering the caches at a future time, and to do so effectively, the jays must rely on their episodic-like memories of past caching events to know where to search for their hidden stashes of food.

See also: Cognition: Time: What Animals Know.

Further Reading Clayton, N.S., Bussey, T.J., Dickinson, A., 2003. Can animals recall the past and plan for the future? Nature Reviews Neuroscience 4, 685–691. Clayton, N.S., Correia, S.P.C., Raby, C.R., Alexis, D.M., Emery, N.J., Dickinson, A., 2008. In defense of animal foresight. Animal Behaviour 76, e1–e3. Clayton, N.S., Dickinson, A., 1998. Episodic-like memory during cache recovery by scrub jays. Nature 395, 272–274. Correia, S.P.C., Alexis, D.M., Dickinson, A., Clayton, N.S., 2007. Western scrub-jays (Aphelocoma californica) anticipate future needs independently of their current motivational state. Current Biology 17, 856–861. James, W., 1890. The Principles of Psychology. Holy, New York. Mulcahy, N.J., Call, J., 2006. Apes save tools for future use. Science 312, 1038–1040. Naqshbandi, M., Roberts, W.A., 2006. Anticipation of future events in squirrel monkeys (Saimiri sciureus) and rats (Rattus norvegicus): Tests of the Bischof–Kohler hypothesis. Journal of Comparative Psychology 120, 345–357. Osvath, M., Osvath, H., 2008. Chimpanzee (Pan troglodytes) and orang-utan (Pongo abelii) forethought: Self-control and pre-experience in the face of future tool use. Animal Cognition 11, 661–674. Raby, C.R., Alexis, D.M., Dickinson, A., Clayton, N.S., 2007. Planning for the future by western scrub-jays. Nature 445, 919–921.

Relevant Websites http://www.psychol.cam.ac.uk/ccl/. – Department of Experimental Psychology: Research. http://www.neuroscience.cam.ac.uk/directory/profile.php?nsclayton. – Professor Nicky Clayton: Cambridge Neuroscience. http://www.youtube.com/watch?v¼y_MnwNyX0Ds. – Bird Tango. http://www.sciencemag.org/cgi/content/full/315/5815/1074.

Metacognition and Metamemory in Non-Human Animalsq RR Hampton, Emory University, Atlanta, GA, United States © 2017 Elsevier Inc. All rights reserved.

Abstract Metacognition, or thinking about thinking, allows one to monitor and control cognitive processing. Metacognition is inferred when subjects accurately evaluate (either prospectively or retrospectively) performance in a cognitive task, such as a perceptual discrimination or a memory test. Non-human animals including dolphins, rats, monkeys, and apes have been reported to show metacognition. Efforts to identify and discriminate among the variety of cognitive mechanisms that may underlie metacognition are ongoing. Some mechanisms for metacognitive performance invoke associative learning about publicly observable stimuli. Other accounts posit introspective cognitive mechanisms by which subjects have privileged access to subjective mental states. Current evidence suggests the existence of both types of mechanisms.

Keywords Awareness; Cognitive control; Confidence; Consciousness; Declarative; Explicit; Introspection; Memory; Memory monitoring; Metacognition; Metamemory; Perception; Self-awareness; Self-control; Self-regulation; Uncertainty

Introduction Metacognition generally means thinking about thinking (Flavell, 1979). Metacognition can allow one to monitor and adaptively control cognitive processing. For example, a human student might improve her grade by dedicating more of her study effort to the longest textbook chapters and the most difficult topics on an upcoming exam. She might restudy the definitions of terms she is less familiar with or finds that she forgot after a single study session (Tu et al., 2015). During the exam, she might skip questions whose answers she is unsure of, returning to them only after first answering questions about which she is confident. In each case, our student has monitored the difficulty faced in learning or performing and has controlled his/her behavior appropriately. While most interest in metacognition is focused on such monitoring and control of one‫׳‬s own cognitive processes, metacognition can also refer to a general knowledge about how cognition works (Flavell, 1979). For example, metacognitive knowledge refers to a variety of information characterizing cognition in general, such as knowing that forgetting happens over time, that one has to attend carefully to follow complex directions, and that some people are better at math than others. This article deals with metacognition as monitoring and controlling one‫׳‬s own cognitive functioning rather than knowing more generally how cognition works.

Approaching the Study of Metacognition in Non-Human Animals Metacognition in humans is often associated with consciousness and complex cognition (Nelson, 1996). These associations raise concerns about the feasibility of studying metacognition in non-human species. But metacognition can be operationalized and studied with objectively observable behavior as will be described in this article (Hampton 2001, 2009). Studies of metacognition in non-human animals have focused on the ability of subjects to monitor and control their own cognitive states. In order to objectively determine whether such monitoring and control occurs, experiments have been designed with three critical features. First, the experimenter defines a primary behavior that can be scored for accuracy or efficiency such as performance in a test of matching-tosample (MTS; this is a memory test in which subjects are required to select a recently experienced stimulus from among a set of distracter stimuli). Next, the experimenter defines a secondary behavior that can be used to infer monitoring or control of the cognition underlying the primary behavior, such as the subjects avoiding difficult tests, or seeking additional information when they do not know the correct response to make. Finally, the experimental design must allow for an explicit assessment of whether the primary and secondary behaviors are correlated. For example, were the tests that the subjects avoided indeed ones on which they were likely to respond incorrectly? This correlation can be assessed most powerfully when the subjects‫ ׳‬state of knowledge is experimentally manipulated and can therefore be confidently known. If subjects avoid memory tests for which they have never been shown the correct answer while taking tests for which the answer was recently presented, this would be consistent with metacognition.

q

Change History: March 2016. R.R. Hampton updated the text and added new citations for the references.

Encyclopedia of Animal Behavior, 2nd edition, Volume 1

https://doi.org/10.1016/B978-0-12-809633-8.01089-X

383

384

Cognition j Metacognition and Metamemory in Non-Human Animals

Studies of Metacognition in Non-Human Animals Non-human animals have demonstrated metacognition in a variety of experiments with the features described earlier. These experiments can be classified according to whether they required metacognition about perception or about memory. Monkeys, dolphins, pigeons, and rats have been shown to either decline difficult trials or make accurate post-trial confidence judgments in perceptual tests. Apes and monkeys have similarly performed in ways consistent with metacognition on memory tests, while pigeons are generally reported not to do so. It should be emphasized, however, that while species differences in metacognition would clearly be of interest, there is currently insufficient data available to reach any firm comparative conclusions. In the following section, a few representative types of test of non-human metacogntion are described.

Avoiding Difficult Perceptual Tests The first published study of metacognition in a non-human species was conducted by David Smith and his colleagues and described the performance of a bottle-nosed dolphin (Tursiops truncatus) in an auditory psychophysical task (Smith et al., 1995). The dolphin was required to discriminate between tones of 2100 Hz and tones of any lower frequency (ranging from 1200–2099 Hz). It was initially trained to make this discrimination (the primary behavior) by responding to a left paddle following 2100 Hz tones and to a right paddle for any lower frequency tone. As expected, the dolphin‫׳‬s accuracy decreased as the tested frequency approached 2100 Hz (the dolphin was likely to respond to the left paddle when the frequency was close to 2100 Hz, treating these tones as if they were 2100 Hz tones). After the dolphin had acquired this primary discrimination, a third paddle was introduced that allowed the dolphin to decline a given discrimination trial (the secondary behavior) in favor of an easy discrimination (a 1200 Hz tone). With these contingencies in place, the dolphin could maximize the rate of reward by performing the primary discrimination (choosing the left or right paddle) when the discrimination was easy, while selecting the third paddle when the discrimination was difficult. The dolphin‫׳‬s behavior generally conformed to these contingencies. It was unlikely to use the third paddle following low frequencies (the easiest trials) and was increasingly likely to use this “decline test” paddle following frequencies near 2100 Hz (the most difficult trials). Later work by Smith and his colleagues showed that monkeys behaved the same way in an analogous psychophysical test in which the density of pixels in a visual display substituted for tones (Smith et al., 1997). Humans given a nearly identical test showed patterns of behavior very similar to those shown by the monkeys. It is interesting to note that the humans reported that they used the “decline test” response only when they felt uncertain.

Confidence Judgments Following Tests A retrospective gambling paradigm was developed by Herb Terrace and his colleagues to assess the ability of monkeys to accurately judge how likely their choices on trials they had just completed were to be correct (Kornell et al., 2007). In this paradigm, monkeys rated their “confidence” by wagering either a large or small number of video tokens on the accuracy of each test trial immediately after they completed it. The video tokens were secondary reinforcers that were periodically “cashed out” for actual food when a sufficient number had accumulated. Critically, monkeys placed their wager after answering, but before receiving feedback about their accuracy. In this paradigm, metacognition predicts large wagers following easy tests (ie, when monkeys are confident of their answer) and small wagers following difficult tests (ie, when monkeys would be unsure of their answer). This in indeed how the monkeys performed in tests on which they were required to discriminate line lengths. These results suggest that they knew whether they had responded correctly despite the lack of feedback prior to placing their bet. Monkeys trained to make these confidence judgments immediately generalized the ability from perceptual tests to memory tests, showing that performance was not restricted to a specific set of test stimuli or even a particular cognitive domain (Kornell et al., 2007). Such generalization suggests the existence of an introspective cue that is relevant to both perceptual and memory tests.

Avoiding Difficult Memory Tests When subjects are presented with lists of items to remember (such as the list of salad dressings available with your order at a restaurant), it is typical for items early and late in the list to be remembered better than items in the middle of the list. Such serial position effects have been a staple of memory research in humans and non-humans. Work with monkeys took advantage of this predictable pattern of memory performance to assess whether monkeys showed metacognition for memory (Smith et al., 1998). Monkeys saw a list of four consecutive random dot polygon figures and their memory for individual polygons from the list was probed using a yes–no recognition test. Monkeys showed the expected serial position effect; their memory was better for the first and last items than for the middle items. Monkeys were then presented with a decline test response, concurrently with a probe polygon that may or may not have been from the studied list and a “not there” response used to indicate that the polygon was not from the studied list. The monkeys declined tests of the middle list items more often than tests of the first and last list items, thus showing that use of the metacognitive response again correlated with accuracy in the primary memory test.

Cognition j Metacognition and Metamemory in Non-Human Animals

385

Seeking Information When Ignorant Metacognition is shown when subjects collect additional information when ignorant and act without expending the effort to seek information when already informed. The first tests of this capacity were conducted with human children, chimpanzees (Pan troglodytes), and orangutans (Pongo pygmaeus; Call and Carpenter, 2001). A modified version of this same test was subsequently used with rhesus monkeys and capuchin monkeys (Fig. 1; Hampton et al., 2004). Subjects were presented with a set of opaque tubes in which food was hidden. Subjects either witnessed the baiting (seen trials) or did not (unseen trials), and therefore were either informed or ignorant about the food‫׳‬s location on each trial. At test, subjects could select a single tube and collect the reward, if they were correct. This test is an interesting assessment of metacognition because the subjects could bend over and look down the length of the tubes to locate the food before choosing (see Fig. 1). Subjects demonstrate metacognition by collecting information when ignorant on unseen trials, and choosing immediately when informed on seen trials. Human children, chimpanzees, orangutans, and rhesus monkeys clearly showed this pattern of behavior, while the case for capuchin monkeys was less clear (some capuchins made this differentiation under at least some conditions). Pigeons tested in related conditions in which they were given an opportunity to study before taking memory tests did not learn to do so, and instead proceeded to the tests without the information needed to succeed (Roberts et al., 2009).

Avoiding Upcoming Tests A few studies have required subjects to make a metacognitive judgment before seeing the actual test. In one study (Hampton, 2001), monkeys were initially trained to match to sample, and then the delay between the study and test phases was gradually lengthened until monkeys performed at an intermediate level between chance and perfection. A metacognitive response was then introduced at the end of the delay interval that allowed monkeys to accept the memory test and receive a favored reward if correct, or decline the memory test and receive a guaranteed, but less desirable, reward. On other trials, only the option to take the memory test was offered at the end of the delay (Fig. 2). Monkeys were more accurate on trials on which they accepted the test than on trials on which they were required to take the test, demonstrating that they accepted tests when memory was relatively good and declined tests when memory was relatively poor. Use of the decline test response generalized to conditions in which memory was directly manipulated either by providing no sample to remember (monkeys overwhelming declined subsequent memory tests) or by increasing the delay interval (monkeys were more likely to decline tests after long than after short delay intervals). Rats were similarly shown to avoid an upcoming auditory duration classification when the signal to be classified was of ambiguous duration (Foote and Crystal, 2007). Studies in which pigeons could avoid upcoming memory tests did not find metacognitive performance (Sutton and Shettleworth, 2008).

Interpreting Metacognitive Performance The performances of some non-human animals in the tests described earlier clearly meet the criteria for metacognition. Subjects adaptively took easy tests and declined difficult tests. Animals judged past performance correctly, sought more information

Fig. 1 Left, a rhesus monkey, ignorant of the food‫׳‬s location (unseen trial), makes the effort to bend down and collect more information by looking through the ends of the opaque tubes before making a choice. Right, an informed monkey makes a choice without going to the effort of confirming the location of the food (seen trial). Such selective information seeking suggests that the monkey knows when he knows, and only seeks more information as needed.

386

Cognition j Metacognition and Metamemory in Non-Human Animals

Fig. 2 Metacognition about memory, or metamemory, in monkeys. Each panel depicts what monkeys saw on a touch-sensitive computer monitor at different stages in a trial.

when needed, and predicted accuracy even before seeing the test. But behaving in a metacognitive way does not by itself specify what particular mechanism underlies the performance (Hampton, 2009). Metacognition in humans is often associated with the conscious awareness of one‫׳‬s own cognitive states and is therefore presumed to reflect private monitoring of those states. But the evidence presented in this article proves neither that metacognitive performance is based on private monitoring of mental states, nor that if it were, those states would need to be conscious states. In the following section, some approaches to explaining metacognitive performance are described. It is likely that no one explanation is sufficient to account for all metacognition; rather, there is a diversity of ways in which metacognition can come about.

Private Versus Public Stimuli for Metacognition It is useful to distinguish between private and public mechanisms for metacognition (Hampton, 2009). Private mechanisms are those by which cognitive control is contingent on the privileged access the subject has to their own cognitive states. In the case of public mechanisms, adaptive cognitive control is based upon the use of publicly available information, such as the perceivable difficulty of a problem or the subject‫׳‬s reinforcement history with particular stimuli. Contrast the following two situations requiring a metacognitive judgment: (1) a colleague asks whether you remember the title of B. F. Skinner‫׳‬s first book and (2) a friend asks whether you can answer a question his 6-year-old has about psychology. In the first case, you would surely check the contents of your memory and determine whether you can retrieve a memory of the book title. Your metacognitive judgment would therefore depend on your success or failure at privately retrieving the relevant explicit memory, a cognitive state to which you, as the one doing the remembering, have privileged access. In the second case, your friend has not even asked you to retrieve a specific memory. If you are an expert in Psychology, you might feel confident (probably correctly) that you can answer the question of a 6-year-old. However, your confidence would not depend on a private evaluation of your memory. Instead, your confidence would depend on your history of expertise, your past ability to answer such questions, and your assessment of the intellectual capacity of 6-year-olds – all publicly available information. It is significant that, in the second case, your friend‫׳‬s judgment about your ability to answer correctly would be about as accurate as your own. This would not be true if you were introspectively accessing a specific explicit memory, in which case you as the

Cognition j Metacognition and Metamemory in Non-Human Animals

387

introspecting individual would have a distinct advantage over others in accurately estimating your knowledge. Thus, the observation of adaptive cognitive control should not be uncritically equated with private mechanisms. In the following section, several mechanisms for adaptive cognitive control are proposed that do not require access to private mental states.

Classes of Stimuli Sufficient for Metacognitive Control Many cases of metacognition may be adequately accounted for by public mechanisms. Because we cannot obtain from non-humans the verbal reports that constitute part of the evidence for private introspective metacognition in humans, we can only infer private metacognition in non-humans by excluding the likely public mechanisms. Below, four classes of mechanisms for metacognition are described. This list is representative rather than exhaustive (see Basile et al., 2014).

Environmental Cue Associations Some stimuli are more difficult to discriminate or remember than others, and some test conditions are more challenging than are others. Stimuli that are close together on a continuum are more difficult to discriminate than are those that are far apart. Highly similar images are difficult to identify in MTS tests. Memory tests after long delays are more difficult than those following short delays. Stimulus magnitude, image similarity, and delay interval are all types of publicly available information that indicate the difficulty of a particular test trial. Subjects performing tests with such stimuli might use the identity, magnitude, similarity, delay, or other publicly available information as a discriminative cue for declining tests or rating confidence. For example, if subjects have experienced low rates of reward with stimuli in a specific magnitude range, they could learn to avoid tests with all stimuli in that range. The probability that such can account for performance in a given paradigm is best assessed by generalization tests which determine whether or not performance is maintained across changes in the particular stimuli used and specific conditions of testing. If performance immediately generalizes to new test conditions or new stimuli, it is safe to conclude that metacognitive responding was not controlled by stimuli that were changed for the generalization test.

Behavioral Cue Associations This account of metacognitive behavior is similar to environmental cue associations, with the exception that the discriminative stimuli controlling use of the metacognitive response are systematically generated by the subject in a way that correlates with accuracy in the primary task. For example, the subject may vacillate when it does not know the correct response on a given test. This vacillation itself does not necessarily represent metacognition by the subject that it does not know the answer, but can rather be an unmediated result of not knowing how to respond. It is common to see this sort of vacillation in monkeys taking MTS tests, for example, in which they look back and forth between the choice stimuli before choosing. It is also well known that response latency is often longer for incorrect than correct responses. Because vacillation and response latency correlate with accuracy, subjects could use these self-generated cues as discriminative stimuli for the metacognitive response, for example, by declining tests on which they experience a relatively long response latency. One way to assess whether behavioral cue associations account for metacognitive performance is to require subjects to make the secondary metacognitive judgment before they have seen the relevant primary test, and therefore before the test could have elicited vacillation or similar behavioral responses, as was done in some of the studies described earlier.

Response Competition In most reports of metacognition in non-human animals, subjects are confronted with the primary discrimination problem or memory test and the secondary metacognitive response option simultaneously. Because subjects can only make one response (eg, a primary test response or a secondary decline test response), simultaneous presentation puts these two behaviors in direct competition. As indicated earlier, animals are often slower to respond on error trials than on correct trials. On error trials with no prepotent primary test response, the probability that the subject will make the secondary metacognitive decline test response is greater, simply because no other competing response occurs immediately. On correct trials, when the inclination to make a primary test response is strong, it may dominate the tendency to decline the test or collect more information before responding (see Le Pelley, 2012 for related arguments). In all of the studies described earlier, the evidence for metacognition is that difficult primary test trials are declined or delayed (while more information is collected). Higher probabilities of the metacognitive response on difficult trials may therefore result from competition between primary choice responses and secondary metacognitive responses. As an example of how different behaviors can compete, consider a rat that has good knowledge of the location of food on a maze. Such a rat is likely to go directly to the baited locations and is consequently unlikely to explore other locations or engage in other behavior. Response competition can be ruled out as an account for metacognitive responding by presenting the secondary metacognitive response option either before or after the primary test, so that the two types of response do not compete directly.

388

Cognition j Metacognition and Metamemory in Non-Human Animals

Introspection Metacognition could also be mediated by private introspective assessment of the subject‫׳‬s mental states. While introspection (ie, the contemplation or perception of ones own mental states) might not necessarily require consciousness, it is closely allied with consciousness in humans. By the introspection account, the discriminative stimulus controlling a metacognitive response (eg, declining to take a test) is the private experience of uncertainty or the weakness of memory. In the case of uncertainty, subjects are suggested to experience conscious (at least in humans) “feelings of uncertainty” that differ from the experience of objective stimuli. In the case of memory, subjects are proposed to assess the strength of their memory. The assessment of memory might be accomplished through several mechanisms that vary in sophistication from detecting whether a memory is present (while knowing nothing of the content of the memory) to attempting to retrieve the relevant memory and determining the success of that effort. Subjects use the decline response or other metacognitive response when memory is determined to be absent or weak. The important difference between this account and the preceding three is that the use of the metacognitive response is based on privileged introspective access to the subject‫׳‬s cognitive states, rather than on publicly available information or response competition. Due to the private nature of introspection, the conclusion that it accounts for metacognitive performance in non-humans can probably be reached only by ruling out other accounts (Hampton, 2009; Basile et al., 2014). The most direct evidence that metacognition in nonhumans sometimes involves introspection comes from a recent study showing that increasing cognitive load selective impairs metacognitive judgments, while leaving primary judgments intact (Smith et al., 2013).

Inferring Consciousness While humans often describe metacognition as accompanied by conscious experience, it is difficult or impossible to specify the causal role that consciousness per se plays in metacognition. But the study of nonverbal species highlights the fact that the functional properties of cognitive systems, but not the phenomenological experiences associated with them, can be determined from behavioral experiments. Functional descriptions of cognitive systems can be applied equally well to human and nonhuman animals. In contrast, description of cognitive systems in terms of subjective experience and various conscious states creates a rift between the study of human and non-human cognition. Whether or not metacognition in other animals is associated with subjective conscious states like those experienced by humans, the growing literature on non-human metacognition demonstrates that the processes underlying metacognition can be effectively studied in non-human species. Metacognitive performance can be achieved through a variety of mechanisms, some of which may be entirely consistent with traditional views of non-human cognition and others that might call for a re-evaluation of the richness of comparative cognition (Smith et al., 2014).

See also: Cognition: Mental Time Travel: Can Animals Recall the Past and Plan for the Future?; Time: What Animals Know.

References Basile, B.M., Schroeder, G.R., Brown, E.K., Templer, V.L., Hampton., R.R., 2014. Evaluation of seven hypotheses for metamemory performance in rhesus monkeys. Journal of Experimental Psychology: General 144, 85–102. Call, J., Carpenter, M., 2001. Do apes and children know what they have seen? Animal Cognition 4, 207–220. Flavell, J.H., 1979. Metacognition and cognitive monitoring: A new area of cognitive-developmental inquiry. American Psychologist 34, 906–911. Foote, A.L., Crystal, J.D., 2007. Metacognition in the rat. Current Biology 17 (6), 551–555. Hampton, R.R., 2001. Rhesus monkeys know when they remember. Proceedings of the National Academy of Sciences of the United States of America 98 (9), 5359–5362. Hamtpon, R.R., 2009. Multiple demonstrations of metacognition in nonhumans: Converging evidence or multiple mechanisms? Comparative Cognition & Behavior Reviews 4, 17– 28. Available at: http://psyc.queensu.ca/ccbr/index.html. Hampton, R.R., Zivin, A., Murray, E.A., 2004. Rhesus monkeys (Macaca mulatta) discriminate between knowing and not knowing and collect information as needed before acting. Animal Cognition 7, 239–254. Kornell, N., Son, L.K., Terrace, H.S., 2007. Transfer of metacognitive skills and hint seeking in monkeys. Psychological Science 18 (1), 64–71. Le Pelley, M.E., 2012. Metacognitive monkeys or associative animals? Simple reinforcement learning explains uncertainty in nonhuman animals. Journal of Experimental Psychology: Learning Memory, and Cognition 38 (3), 686–708. Nelson, T.O., 1996. Consciousness and metacognition. American Psychologist 51 (2), 102–116. Roberts, W.A., Feeney, M.C., McMillan, N., et al., 2009. Do pigeons (Columba livia) study for a test? Journal of Experimental Psychology: Animal Behavior Processes 35 (2), 129–142. Smith, J.D., Couchman, J.J., Beran, M.J., 2014. Animal metacognition: A tale of two comparative psychologies. Journal of Comparative Psychology 128 (2), 115–131. Smith, J.D., Coutinho, M.V.C., Church, B.A., Beran, M.J., 2013. Executive-attentional uncertainty responses by rhesus macaques (Macaca mulatta). Journal of Experimental Psychology-General 142 (2), 458–475. Smith, J.D., Schull, J., Strote, J., et al., 1995. The uncertain response in the bottle-nosed-dolphin (Tursiops truncatus). Journal of Experimental Psychology: General 124 (4), 391–408. Smith, J.D., Shields, W.E., Schull, J., Washburn, D.A., 1997. The uncertain response in humans and animals. Cognition 62 (1), 75–97.

Cognition j Metacognition and Metamemory in Non-Human Animals

389

Smith, J.D., Shields, W.E., Washburn, D.A., 1998. Memory monitoring by animals and humans. Journal of Experimental Psychology: General 127 (3), 227–250. Sutton, J.E., Shettleworth, S.J., 2008. Memory without awareness: Pigeons do not show metamemory in delayed matching to sample. Journal of Experimental Psychology: Animal Behavior Processes 34 (2), 266–282. Tu, H.W., Pani, A.A., Hampton, R.R., 2015. Rhesus monkeys (Macaca mulatta) adaptively adjust information seeking in response to information accumulated. Journal of Comparative Psychology 129 (4), 347–355.

Further Reading Hampton, R.R., 2005. Can Rhesus monkeys discriminate between remembering and forgetting? In: Terrace, H.S., Metcalfe, J. (Eds.), The Missing Link in Cognition: Origins of SelfReflective Consciousness. Oxford University Press, New York, NY, pp. 272–295. Kornell, N., 2009. Metacognition in humans and animals. Current Directions in Psychological Science 18 (1), 11–15. Smith, J.D., Shields, W.E., Washburn, D.A., 2003. The comparative psychology of uncertainty monitoring and metacognition. Behavioral and Brain Sciences 26, 317–374. Smith, J.D., Washburn, D.A., 2005. Uncertainty monitoring and metacognition by animals. Current Directions in Psychological Science 14 (1), 19–24.

Relevant Websites http://www.psychology.emory.edu/lcpc/bailout.high.html– Video of a monkey performing a metamemory test.

Non-Elemental Learning in Invertebrates M Giurfa and A Avargue`s-Weber, CNRS, Université de Toulouse, Toulouse, France; and Centre de Recherches sur la Cognition Animale, Toulouse, France R Menzel, Freie Universität Berlin, Berlin, Germany © 2010 Elsevier Ltd. All rights reserved.

Abstract Invertebrates are historically considered as good models for the study of elemental forms of associative learning such as pavlovian or operant conditioning. Here, we review and discuss studies showing that besides the capacity to master simple learning forms, invertebrates also learn to solve nonelemental associative problems, in which the relationship between events or stimuli is ambiguous or nonlinear. We focus in particular on contextual learning in which a stimulus may or not be predictive of a certain outcome depending on the particular environment, and on rule learning in which animals have to exhibit a positive transfer of learning toward new stimuli that have never been encountered before on the basis of an ‘if then’ principle. Invertebrates constitute attractive models for cognitive neurosciences because of the facility they offer to access the neural mechanisms underlying these complex types of learning. Indeed neural networks mediating associative learning in these models are well known, and neurogenetic tools available in Drosophila open new possibilities for dissecting the basis of complex learning forms.

Keywords Cognition; Contextual learning; Invertebrate; Learning; Nonelemental learning; Rule learning

Introduction Cognitive science provides a fresh look at animal behavior, and its merge with neuroscience overcomes the conceptual limitations of traditional experimental psychology and ethology. Despite the multitude of approaches in cognitive neuroscience and the respective attempts to define these approaches, a general definition for the term ‘cognition’ remains elusive probably because the key terms are understood differently depending on the conceptual traditions to which the scientists relate themselves, the behaviors in question, and the considered complexity of the neural substrates underlying them. A key term is ‘representation,’ the understanding that the brain is actively involved in perceiving the world and creating motor patterns by recruiting memories, expecting outcomes, and making decisions between neural instantiations of behavioral options. Gaining information by learning and by storing it in multiple forms of memory, as a fundamental form of representation, is an essential and most likely a basic property of any neural system of some complexity. Here, we shall focus on nonelemental forms of associative learning, that is, on learning forms in which simple, unambiguous links between specific events in an animal’s environment cannot account for experience-dependent changes in behavior, and which require operations on remote and recent memories. In this respect, nonelemental associative learning transcends elemental forms of associative learning, in which animals learn univocal connections between specific events in their environment. In particular, we shall ask whether animals with small brains like molluscs and insects are capable of performing nonelemental associative learning.

Elemental Forms of Associative Learning in Invertebrates Associative learning allows extracting the logical structure of the world by evaluating the sequential order of events. Two major forms of associative learning are usually recognized: in classical conditioning, animals learn to associate an originally neutral stimulus (conditioned stimulus (CS)) with a biologically relevant stimulus (unconditioned stimulus (US)); in operant Conditioning, they learn to associate their own behavior with a reinforcer and relate this connection to the context conditions of the environment. In their most simple version, both learning forms rely on the establishment of associative links connecting two (or more) specific and unambiguous events in the animal’s world. For instance, in absolute classical conditioning (Aþ), a direct link between an event (A) and reinforcement (þ) is learned, while in differential classical conditioning (Aþ vs. B), simple, unambiguous links between A and reinforcement and between B and the absence of reinforcement are simultaneously learned. Multiple cases of these simple learning forms have been described for invertebrates. For instance, in the honeybee Apis mellifera, olfactory conditioning of the proboscis extension response (PER) has been repeatedly used for the study of elemental classical conditioning and its neural substrates. Individually harnessed hungry bees that do not respond to an odor presentation with an extension of their proboscis do so when their antennae are stimulated with sucrose solution (the US). If the odor (the CS) is forward paired with sugar, the bees learn an association between odor and sugar reward so that they exhibit conditioned PER to future presentations of the odor alone (Figure 1). An example of elemental operant conditioning is provided by the aquatic mollusc Lymnaea stagnalis, which can be trained to suppress the opening of its pneumostome, a small respiratory orifice, when the animal surfaces and

390

Encyclopedia of Animal Behavior, 2nd edition, Volume 1

https://doi.org/10.1016/B978-0-12-813251-7.00095-X

Cognition j Non-Elemental Learning in Invertebrates

391

Figure 1 Olfactory conditioning of the proboscis extension reflex. (a) An individual bee is immobilized in a metal tube so that only the antennae and mouth parts (the proboscis) are free to move. The bee is set in front of an odorant stimulation setup which is controlled by a computer and which sends a constant flow of clean air to the bee. The air flow can be rerouted through cartridges presenting chemicals used for olfactory stimulation (conditioned stimuli or CS). A toothpick soaked in sucrose solution (unconditioned stimulus or US) is delivered to the antennae and the proboscis. In this appetitive classical conditioning, the bee learns to associate odorants (CS) and sucrose solution (US). (b) The proboscis extension reflex of the honeybee. Bees exhibit this reflex when their antennae are touched with sucrose solution (US). After successful conditioning, bees extend the proboscis to the odorant (CS) which predicts the US.

attempts to breathe. This is achieved by an aversive and repeated mechanical stimulation of the pneumostome, which determines that the mollusc learns to reduce its attempts to open the pneumostome as training progresses. In both examples, the neural networks mediating associative learning are relatively simple and well studied, thus underlining the advantages of invertebrates as model systems for the understanding the neural mechanisms of simple forms of learning.

Nonelemental Forms of Associative Learning in Invertebrates In the higher-order forms of learning on which we focus here, simple links connecting specific events are generally not useful because ambiguity characterizes the events under consideration. For instance, in the discrimination termed negative patterning discrimination, an animal has to learn to differentiate a nonreinforced binary compound AB from its reinforced elements (Aþ, Bþ). This situation is particularly challenging as each element A and B appears as often reinforced as nonreinforced. Relying on elemental links between A (or B) and reinforcement (or absence of reinforcement) is useless to solve this problem. Another example of nonelemental learning is the so-called biconditional discrimination where the subject learns to respond to the compounds AB and CD and not to the compounds AC and BD (ABþ, CDþ, AC, BD). As in negative patterning, each element, A, B, C, and D appears reinforced as often as nonreinforced so that it is impossible to rely only on the associative strength of a single element to solve the task. In both examples, animals have to suppress linear processing of compounds and learn that a compound is an entity different from its components. A second form of nonelemental learning is contextual learning, in which animals learn to produce adaptive responses that can be linked to a specific context. They learn that, given a certain stimulus or condition, a particular response is appropriate whereas, given a different stimulus or condition, the same response is no longer appropriate. This form of learning, usually referred to as conditional learning or occasion setting, cannot be viewed as elemental learning because a given stimulus may or not be predictive of a certain outcome, depending on the particular environment. A third form of nonelemental rule is provided by rule learning in which animals respond to novel stimuli that they have never encountered before or can generate novel responses that are adaptive given the context in which they are produced. In doing this, animals exhibit a positive transfer of learning, a capacity that cannot be referred to as an elemental learning because the responses are aimed at stimuli that do not predict a specific outcome per se based on the animals’ past experience. One of the first works adopting a nonelemental learning perspective in invertebrates was performed on lobsters. These animals normally exhibit exploratory behavior when placed in an aquarium. They can be aversively conditioned to stop searching by pairing an olfactory stimulus delivered in water with a mechanosensory disturbance produced by the experimenter. Lobsters were trained in this way with an olfactory compound AX reinforced by the aversive mechanosensory stimulation (AXþ). Conditioning was either absolute (AXþ) or differential, when a second compound AY (AXþ vs. AY) was used. After absolute conditioning, lobsters inhibited their search behavior when presented with AX as expected, but searched when presented with A, X, or with a novel odor Y. This result is consistent with learning the compound AX as an entity different from its components A and X, as proposed by the configural theory (Pearce, 1994). After differential conditioning, lobsters again inhibited their searching behavior when presented with AX but not with AY. Interestingly, they also inhibited search when presented with the element X but not with the element Y. A was not useful as it was common to the reinforced and the nonreinforced compounds AXþ and AY, respectively. In this case, lobsters seemed to have learned the compounds AX and AY in elemental terms, thus being able to fully generalize their respective responses to X and Y. This work shows that depending on the conditioning protocol, lobsters treat and learn an olfactory compound differently so that either elemental or nonelemental associations with the negative mechanosensory reinforcer are built.

392

Cognition j Non-Elemental Learning in Invertebrates

In honeybees, several studies have addressed the issue of elemental versus nonelemental learning, using visual conditioning of free-flying animals or olfactory PER conditioning of harnessed animals. In the first protocol, bees flying between the hive and a feeding site are trained to discriminate different kinds of visual targets (colors, shapes, motion cues, etc.) at the food source. Correct choices are rewarded with a drop of sucrose solution. In the second protocol, described earlier, harnessed bees learn a Pavlovian association between odor and the sucrose reward. In both experimental protocols, bees were shown to solve a biconditional discrimination (ABþ, CDþ, AC, BD). In the visual modality, free-flying bees had to discriminate complex patterns that were arranged to fulfill the principles of this discrimination problem. In the olfactory modality, olfactory compounds were used and bees learned to respond appropriately to each compound, independently of the ambiguity inherent to the components. Bees also proved to be able to solve a negative patterning discrimination (Aþ, Bþ, AB) in the olfactory domain. It was shown that in situations in which ambiguity is created at the level of the odorants integrating a compound, olfactory processing is consistent with the unique cue theory, a form of processing in which animals detect to some extent the presence of the components in the compound but in which they also assign a unique identity to the compound (the unique cue), resulting from the interaction between its components.

Neural Bases of Nonelemental Learning in Invertebrates The interest in nonelemental olfactory learning protocols in insects relates to the possibility of correlating the behavior with the plasticity of the underlying neural circuits. The olfactory circuit is relatively well known. In the case of honeybees (Figure 2), peripheral processing of odor molecules occurs at 60 000 olfactory receptor neurons (ORNs) and in 160 glomeruli of the antennal lobe (AL). ORNs and glomeruli in the AL have broad, overlapping and combinatorial responses to a range of odors. Processed olfactory information is conveyed by 800 projection neurons (PNs) to higher-order brain centers (mushroom bodies (MBs) or lateral protocerebrum). MBs are particularly interesting from the perspective of nonelemental learning since they receive segregated information of different sensory modalities (visual, olfactory, mechanosensory) and provide multimodal output that reflects the integration of information between modalities at the level of the neurons that constitute them, the Kenyon cells and the mushroom body output neurons. In honeybees, bilateral olfactory input to both antennae is required to solve a negative patterning discrimination. Given that the olfactory circuit remains practically unconnected between hemispheres until the MBs, this result suggests that the reading of a unique cue, arising from odorant interactions within the mixture, occurs upstream the ALs, that is, at the level of the MBs. Mushroom body-ablated honeybees were used to determine whether these structures are necessary to solve nonelemental olfactory discriminations. Bees were conditioned in a side-specific discrimination so that when odorants were delivered to one antenna, the contingency was Aþ versus B, while it was reversed (A vs. Bþ) when they were delivered to other antenna. Bees without lesions could solve this nonelemental problem (each odor is as often rewarded as nonrewarded), while bees with unilateral lesions of the MBs were impaired in this problem solving but not in elemental discriminations. It was therefore proposed that MBs are required for solving nonelemental discriminations. It thus appears that at least lobsters and honeybees are capable of nonelemental learning in the strict sense and that in insects, MBs are involved in such kind of problem solving. Such forms of learning are highly dependent on the way in which animals are

Figure 2 The basic organization of the honeybee olfactory system. (a) Frontal view of the brain with the main olfactory centers; (b) Threedimensional reconstruction of the olfactory circuit based on confocal microscopy; AL: antennal lobe; LH: lateral horn; MB: mushroom body; m-ACT: medial antenno-cerebral tract; l-ACT: lateral antenno-cerebral tract; mCa: medial calyx; lCa: lateral calyx; alpha and beta: alpha and beta lobes of the mushroom body. Courtesy of Wolfgang Roessler.

Cognition j Non-Elemental Learning in Invertebrates

393

trained, the number of trials, and on the similarity between elements in a compound. Further research should ask whether other invertebrates particularly Drosophila solve nonlinear discrimination problems. Neurogenetic tools available in this insect could be a most useful tool for identifying in a more precise way the neuronal circuits involved in nonlinear discriminations.

Contextual Learning in Invertebrates Contextual learning can be subsumed in the so-called occasion setting problem. In this problem, a given stimulus, the occasion setter, informs the animal about the outcome of its choice (for instance, given stimulus C, the occasion setter, the animal has to choose A and not B because the former but not the latter is rewarded). This basic form of conditional learning admits of different variants, depending on the number of occasion setters and discriminations involved, which have received different names. For instance, another form of occasion setting involving two occasion setters is the so-called transwitching problem. In this problem, an animal is trained differentially with two stimuli, A and B, and with two different occasion setters C1 and C2. When C1 is available, stimulus A is reinforced while stimulus B is not (Aþ vs. B), while it is the opposite (A vs. Bþ) with C2. This problem does not admit lineal solutions as each element (A, B) and each occasion setter (C1, C2) appear equally as often connected with reinforcement as with absence of reinforcement. Focusing on A or B alone does not allow solving the problem. Animals have, therefore, to learn that C1 and C2 define the valid contingency. In the mollusc Aplysia californica, exposure to two different contexts (a smooth, round bowl containing lemon-flavored seawater and a rectangular chamber with a ridged surface containing unscented seawater that was gently vibrated by an aerator located in one corner) and experiencing a series of moderate electric shocks (US) in one of these two contexts lead to the establishment of an association between the context and the shock. The context alone elicited a defensive reaction which was exclusive for the reinforced context. Crickets Gryllus bimaculatus and cockroaches Periplaneta americana also exhibit contextual learning as they solve a typical version of the transwitching problem. Both crickets and cockroaches associate one odorant with water reward (appetitive US) and another odorant with saline solution (aversive US) under illumination, and learn the reversed contingency in the dark. Thus, the visual context affected the learning performance only when crickets were requested to use it to disambiguate the meaning of stimuli and to predict the nature of reinforcement. Bumblebees Bombus terrestris have also been trained in a transwitching problem to choose a 45 grating and to avoid a 135 grating to reach a feeder, and to do the opposite to reach their nest. They also learn that an annular or a radial disc must be chosen, depending on the disc’s association with a 45 or a 135 grating either at the feeder or at the nest entrance: in one context (the nest), access was allowed by the combinations 45 þ radial disc and 135 þ annular disc, but not by the combinations 45 þ annular disc and 135 þ radial disc; at the feeder, the opposite was true. In both cases, the potentially competing visuomotor associations were insulated from each other because they were set in different contexts. Comparable behavior was found in honeybees, where distinct odors or times of the day were the occasion setters for a given flight vector or rewarded color. Further examples for contextual learning could be provided but they would be redundant for the main conclusion of this section: invertebrates are capable of different forms of conditional learning. Despite this cumulative body of evidences, the nature of the associations underlying this kind of learning and the neural substrates underlying this form of learning remain unclear. Studies of decision making in the fruit fly Drosophila melanogaster indicate that MBs are of fundamental importance for this behavior. In this case, an individual fly suspended at a torque meter from a copper wire glued to its thorax beats its wings when hanging in the middle of a cylindrical arena displaying a visual panorama with identifiable landmarks (Figure 3). An unpleasant heat-beam is focused on the fly’s thorax and switched on whenever the insect fly toward a given landmark on the cylinder. The fly controls the reinforcer delivery as its flight maneuvers determine the on/off switching of the heat beam if the appropriate flight directions (i.e., landmarks) are chosen. In studies of decision-making in Drosophila, flies were conditioned to choose one of two flight paths in response to color and shape cues; after the training, they were tested with contradictory cues. Normal flies made a discrete choice that switched from one alternative to another as the relative salience of color and shape cues gradually changed, but this ability was greatly diminished in mutant flies with miniature MBs or with hydroxyurea ablation of MBs. Although this protocol does not provide a formalized nonlinear discrimination problem such as those presented earlier (e.g., negative patterning), it has the merit of moving from the traditional elemental learning protocols applied so far in Drosophila to a more sophisticated problem in which the cognitive richness of fly behavior could be revealed and related to the MBs. Furthermore, it was shown that salience-dependent choice behavior consists of early and late phases; the former requires activation of the dopaminergic system and MBs, whereas the latter is independent of these activities. Immunohistological analysis showed that MBs are densely innervated by dopaminergic axons, thus suggesting that the circuit from the dopamine system to MBs is crucial for choice behavior in Drosophila.

Positive Transfer in Rule Learning by Invertebrates Nonelemental associative learning also underlies problem solving in which animals respond to novel stimuli that they have never encountered before or can generate novel responses that are adaptive given the context in which they are produced. In doing this, the

394

Cognition j Non-Elemental Learning in Invertebrates

Figure 3 The flight simulator used for visual conditioning of a tethered fruit fly. A Drosophila is flying stationarily in a cylindrical arena homogeneously illuminated from behind. The fly’s tendency to perform left or right turns (yaw torque) is measured continuously and fed into the computer. In closed-loop, the computer controls arena rotation. On the screen four ‘landmarks,’ two Ts and two inverted Ts, are displayed in order to provide a referential frame for flight direction choice. The illumination of the arena can be changed using color filters. A heat beam focused on the fly’s thorax is used as an aversive reinforcer. The reinforcer is switched on whenever the fly flies towards a prohibitive direction. The fly controls therefore reinforcer delivery by means of its flight direction so that operant conditioning mediates the performance observed. However, classical associations between landmarks and reinforcer (or its absence) can also be established in this protocol. Courtesy of B. Brembs.

animals exhibit a positive transfer of learning, a capacity that cannot be referred to as an elemental learning because the responses are aimed at stimuli that do not predict a specific outcome per se based on the animals’ past experience. A typical example of rule learning is the acquisition of the sameness or difference principle. These rules are demonstrated through the protocols of delayed matching to sample (DMTS) and delayed nonmatching to sample (DNMTS), respectively. In DMTS, animals are presented with a sample and then with a set of stimuli, one of which is identical to the sample and which is reinforced. Since the sample is regularly changed, animals must learn the sameness rule, that is, ‘always choose what is shown to you (the sample), independent of what else is shown to you.’ In DNMTS, the animal has to learn the opposite. Honeybees foraging in a Y-maze learn both rules. Bees were trained in a DMTS problem in which they were presented with a changing nonrewarded sample (i.e., one of two different color disks or one of two different black-and-white gratings, vertical or horizontal) at the entrance of a maze (Figure 4). The bees were rewarded only if they chose the stimulus identical to the sample once within the maze. Bees trained with colors and presented in transfer tests with black-and-white gratings that they have not experienced before solved the problem and chose the grating identical to the sample at the entrance of the maze. Similarly, bees trained with the gratings and tested with colors in transfer tests also solved the problem and chose the novel color corresponding to that of the sample grating at the maze entrance. Transfer was not limited to different kinds of modalities (pattern vs. color) within the visual domain, but could also operate between drastically different domains such as olfaction and vision. Furthermore, bees also mastered a DNMTS task, thus showing that they also learn a rule of difference between stimuli. These results document that bees learn rules relating stimuli in their environment. The capacity of honeybees to solve a DMTS task has recently been verified and studied with respect to the working memory underlying it. It was found that the working memory for the sample underlying the solving of DMTS is around 5 s and thus coincides with the duration of other visual and olfactory short-term memories characterized in simpler forms of associative learning in honeybees (Menzel, 1999). Moreover, bees trained in a DMTS task can learn to pay attention to one of two different samples presented successively in a flight tunnel (either to the first or to the second) and can transfer the learning of this sequence weight to novel samples. The neural basis of rule extraction has not been addressed yet in invertebrates. The potentials offered by Drosophila with respect to molecular genetics and by the bee with respect to the recording of neural correlates will certainly be used in the near future to establish closer links to the neural substrates.

Conclusion Here we focused on a particular basic cognitive faculty that relates to the ability of animals to process sequences of associative connections such that structures of interrelatedness are derived which are not housed in the elemental associations. In some cases, rules are learned and applied across sensory modalities, in others temporal relations are acquired. Learning under natural conditions will be much richer than implied here because bees, for example, are known to navigate along novel routes according to the expected outcome of the navigational choices, and Drosophila decides between flight goals by integrating multiple stimulus

Cognition j Non-Elemental Learning in Invertebrates

395

Figure 4 Rule learning in honeybees. Honeybees trained to collect sugar solution in a Y-maze (a) on a series of different patterns (b) learn a rule of sameness. Learning and transfer performance of bees in a delayed matching-to-sample task in which they were trained to colors (Experiment 1) or to black-and-white, vertical and horizontal gratings (Experiment 2). (c, d) Transfer tests with novel stimuli. (c) In Experiment 1, bees trained on the colors were tested on the gratings. (d) In Experiment 2, bees trained on the gratings were tested on the colors. In both cases bees chose the novel stimuli corresponding to the sample although they had no experience with such test stimuli. n denotes number of choices evaluated. Reproduced from Giurfa M, Zhang SW, Jenett A, Menzel R, and Srinivasan M (2001) The concepts of sameness and difference in an insect. Nature 410: 930–933.

conditions. These and the examples discussed here require brain functions best conceptualized as representations, since the relations established during learning cannot reside in basic cellular modules of associative connections as they were so successfully studied in invertebrates. Rather they must be represented in network properties composed of multiple cellular association modules which incorporate new information into already stored information by some self-organization process, retrieve appropriate information from remote stores, and allow decisions to be made according to the current conditions, the internal status of the animal, and the evaluated expected outcomes. Hints for memory processing during both memory storage and retrieval come from multiple observations. For example, consolidation of earlier forms of memory into later and stable forms changes the content of the memory and is accompanied by transfer between structures, for example, between the gamma lobe and the alpha/beta lobe neurons in the mushroom body of Drosophila. Memory retrieval initiates processes described as reconsolidation, and decisions between simultaneously activated memories are being made without access to stimuli according to the expected outcome. In this respect, memory processing during storage and retrieval in invertebrates resembles basic features described for mammals and humans, and it is conceivable that analog network processes may be responsible despite the large differences in the structure and functional organization between for example, insect and mammalian brains. How are we to discover these processes? A fundamental requirement for any experimental approach is that the working of the neural nets is monitored at the level of multiple but single neurons under conditions in which the animal learns, retrieves, and processes memory. Ideally, these neurons should be identifiable anatomically, aiming to establish a close relationship between structure and function. These strict requirements are not met by any animal although recent advances in optical and electrical recordings from neurons in the Drosophila and the bee brain come close. Two streams of new developments have to meet in an attempt to take advantage of invertebrates as models for a cognitive neuroscience approach, a conceptual shift in addressing the phenomena of learning and memory, and a major methodological advance. Methodological advances are already on the verge. Calcium and voltage sensitive dyes as well as light driven dyes for controlling neural excitation can be expressed in defined neurons of the Drosophila brain, while recordings from multiple neurons in the bee brain can be performed for several days when the animals learn and perform. A more important achievement will be the conceptual

396

Cognition j Non-Elemental Learning in Invertebrates

shift, which relates to the necessity to include invertebrates into the cognitive view of behavior. It is the combination of stereotypical and highly flexible behavior of invertebrates that makes them such attractive study objects for a cognitive approach. Evidence presented and discussed in this article aims at promoting this cognitive framework to understand invertebrate behavior.

See also: Cognition: Categories and Concepts: Language-Related Competences in Non-Linguistic Species; Metacognition and Metamemory in Non-Human Animals. Genes and Behavior: Drosophila: Behavior Genetics. Learning and Teaching: Insect Social Learning. Neurons and Senses: Crabs and Their Visual World; Invertebrate Vision; Nematode Learning and Memory: Neuroethology; Taste: Invertebrates.

Further Reading Davis, R., 2005. Olfactory memory formation in Drosophila: From molecular to systems neuroscience. Annual Review of Neuroscience 28, 275–302. Giurfa, M., 2007. Behavioral and neural analysis of associative learning in the honeybee: A taste from the magic well. Journal of Comparative Physiology A 193, 801–824. Giurfa, M., Zhang, S.W., Jenett, A., Menzel, R., Srinivasan, M., 2001. The concepts of sameness and difference in an insect. Nature 410, 930–933. Greenspan, R.J., van Swinderen, B., 2004. Cognitive consonance: Complex brain functions in the fruit fly and its relatives. Trends in Neurosciences 27, 707–711. Guo, J., Guo, A., 2005. Crossmodal interactions between olfactory and visual learning in Drosophila. Science 309, 307–310. Heisenberg, M., 2003. Mushroom body memoir: From maps to models. Nature Reviews Neuroscience 4, 266–275. Lachnit, H., Giurfa, M., Menzel, R., 2004. Odor processing in honeybees: Is the whole equal to, more than, or different from the sum of its parts? In: Slater, P.J.G. (Ed.), Advances in the Study of Behavior, vol.34 Elsevier, San Diego, CA, pp. 241–264. Livermore, A., Hutson, M., Ngo, V., Hadjisimos, R., Derby, C.D., 1997. Elemental and configural learning and the perception of odorant mixtures by the spiny lobster Panulirus argus. Physiology & Behavior 62, 169–174. Menzel, R., 1999. Memory dynamics in the honeybee. Journal of Comparative Physiology A 185, 323–340. Menzel, R., Brembs, B., Giurfa, M., 2006. Cognition in invertebrates. In: Strausfeld, N.J., Bullock, T.H. (Eds.), The Evolution of Nervous Systems. Vol II: Evolution of Nervous Systems in Invertebrates. Elsevier Life Sciences, London, pp. 403–422. North, G., Greenspan, R., 2007. In: Invertebrate Neurobiology. CSHL Press, New York, NY, p. 665. Pearce, J.M., 1994. Similarity and discrimination: A selective review and a connectionist model. Psychological Review 101, 587–607. Swinderen, B.V., 2005. The remote roots of consciousness in fruit-fly selective attention? BioEssays 27, 321–330. Zhang, K., Guo, J.Z., Peng, Y., Xi, W., Guo, A., 2007. Dopamine-mushroom body circuit regulates saliency-based decision-making in Drosophila. Science 316, 1901–1914. Zhang, S.W., Srinivasan, M.V., 2004. Exploration of cognitive capacity in honeybees. In: Prete, F.R. (Ed.), Complex Worlds from Simpler Nervous Systems. MIT Press, Cambridge, pp. 41–74.

Primate Archaeology Susana Carvalho and Katarina Almeida-Warren, University of Oxford, Oxford, United Kingdom © 2019 Elsevier Ltd. All rights reserved.

Glossary Assemblage In archaeological terms, it refers to a collection of archaeological material and artefacts (e.g., tools, ceramics, coins etc.) that have accumulated in one location over time. Chaîne opératoire In archaeological terms, it refers to a method employed to reconstruct the making of a tool, that can include the complete operational sequence of a technological activity, from raw material selection, to object modification, utilization, and, eventually, discard. Clinometer Or inclinometer, is an instrument used to measure vertical angles (tilt), between two objects. GIS Geographic Information System. It is a digital framework or program, specialised in collecting, visualizing, managing, and processing geographic and spatial information. GNSS Global Navigation Satellite System. It is an umbrella term for satellite navigation systems that provide autonomous geospatial positioning at a global scale. Examples of GNSS systems include: GPS, Galileo, and GLONASS. Hominin Also Hominini is a taxonomic group that refers to our species, Homo sapiens, and all our ancestors (e.g., H. habilis, H. erectus, H. neanderthalensis). H. sapiens is the only Hominin species alive today. In-situ In archaeological terms, it refers to archaeological remains that are found in their original location (i.e., where they were last used), without any evidence of displacement by taphonomic processes. Material culture Represents the physical evidence (e.g., objects, resources, buildings, spaces) that define the culture of a population. Micro-wear Defines microscopic abrasion, wear, polishing, left on the surface of artefacts, such as tools, as a result of their use in particular activities. Morphometric Refers to the quantitative analysis of the dimensions and shape of an object and its features. Plumb bob Or plummet, is a weight with a pointed tip which is suspended from a string and is used as a vertical reference line. Use-wear Defines the surface modification on an object or artefact, as a result of its use in a particular activity.

Abstract Primate Archaeology is a novel field of research that has had transformed our knowledge about the origins and evolution of technology. It expanded the focus of traditional archaeological studies from humans to all primates that are tool-users. Research in this area focuses on present and past processes of tool-making, use and site formation in extant and extinct primates, such as chimpanzees and monkeys. The discipline is flourishing and is at the forefront of key scientific developments in the fields of primatology, archaeology and palaeoanthropology. In this article, we provide an historical overview of the discipline, describe main methodological tools, highlight key case studies and discuss implications of recent findings and future research directions.

Keywords Behavioral evolution; Capuchins; Chimpanzees; Cognition; Extractive foraging; Human evolution; Macaques; Non-human artefacts; Nut-cracking; Pliocene archaeology; Primate archaeology; Primate heritage; Technological Origins; Termitefishing; Tool-making; Tool-use

Introduction Primate Archaeology is a recently created, and burgeoning, field of research at the inter-section of Primatology and Archaeology (Carvalho et al., 2008, 2009; Haslam et al., 2009, 2017). The sub-discipline focuses on the study of the technological behavior of extant (and extinct) non-human primates and its material record. Research in this field typically combines archaeological methods with ethological approaches by studying the material evidence of technology in tandem with the behavioral evidence, whether directly or indirectly observed, associated with tool-use and tool-making in the Primate order. The need to establish a new discipline that considers non-human primate tools, and behavior, as critical to further understand how humans developed such complex technology (Carvalho et al., 2008; Haslam et al., 2009) emerged as a natural consequence of milestone, first time developments during the last decade: archaeologists went to West Africa to study chimpanzee stone tools (e.g., Joulian, 1994, 1996; Mercader et al., 2002; Mercader et al., 2007), and primatologists applied archaeological methods to infer

Encyclopedia of Animal Behavior, 2nd edition, Volume 1

https://doi.org/10.1016/B978-0-12-809633-8.90156-0

397

398

Cognition j Primate Archaeology

invisible behavior from chimpanzee tools (see Hernandez-Aguilar et al., 2007; Carvalho et al., 2007). Primate Archaeology provided a novel framework to investigate the origins of technological behavior using a cross-species, evolutionary approach (Haslam et al., 2009, 2017). The growing research operating within this paradigm has recently brought researchers to suggest that the variation we see today in the tool-use of the primate order indicates that: (1) Tool-use in humans and chimpanzees may be a shared, primitive trait, thus the Last Common Ancestor between humans and chimpanzees (ca. 7 million years ago) was likely a tool-user; (2) Tool-use in the more distant related primates, like monkeys, may be better explained by convergence, i.e., parallel evolution (see also Campbell and Carvalho, 2017 for an extensive review on tool-use and manufacture in the Last Common Ancestor of Pan and Homo). The foundation of Primate Archaeology has had wide scientific implications, leading to important re-interpretations of past archaeological records (see Proffitt et al., 2016; Roche, 2016). It has opened the way to the new “Pliocene Archaeology” and the acknowledgement that archaeological assemblages pre-dating 2.6 million years should exist in Africa (Carvalho and McGrew, 2012; Harmand et al., 2015). It is illuminating variables that can explain the loss of technological traditions (Luncz et al., 2017b), and is challenging the old premise that hominins must have manufactured all of the artefacts found in the archaeological record. This new paradigm suggests that in the past, as is the case today, a variety of primate species generated material culture. Primate archaeology also encapsulates ethoarchaeology, a branch of archaeology which studies, in effect, living archaeology and the relationships between observed behaviors and the active archaeological signatures they produce on a daily basis (e.g., McGrew et al., 2007). The study of this living archaeology, enables an in-depth understanding of contemporary primate technology and the associated cognitive, social and cultural mechanisms. Thus, in using the same methodological framework of traditional archaeology, Primate Archaeology is producing a wealth of information that can be directly translated into the archaeological record of our earliest ancestors and is becoming a vital asset to Paleolithic archaeologists for understanding the functional and cognitive contexts of early hominin technology, as well as reconstructing patterns of resource exploitation and landscape use associated to the earliest archaeological assemblages (Toth et al., 2006; Toth and Schick, 2009). It can help identify the evolutionary origins of hominin tool-use and provided tools for identifying currently archaeologically invisible behaviors (Panger et al., 2002; Carvalho et al., 2009). Non-human primate tool-use itself, has been known for over 60 years, when Jane Goodall witnessed, for the first time, chimpanzee termite-fishing at the wild chimpanzee field site now known as The Gombe Stream National Park. She provided the first scientific account of tool-use in a species other than our own (Goodall, 1963). This ground-breaking discovery defied the longstanding anthropocentric view that technology was a uniquely human trait (Carvalho and McGrew, 2012; Carvalho and Beardmore-Herd, in press). However, despite decades of primatological studies focusing on primate material culture (McGrew, 2002, 2004), and the awareness that tool-using primates could be useful models for hominin technological evolution (Wynn and McGrew, 1989; Sakura and Matsuzawa, 1991; Marchant and McGrew, 2005), archaeology continued to be synonymous of ‘human’ until 2002, when the first excavation of a non-human archaeological site was published, in a groundbreaking study in Science (Mercader et al., 2002). The excavations took place at three wild chimpanzee nut-cracking sites in the Taï National Park, Cote d’Ivoire, and yielded several buried tree roots used as anvils with associated hammerstone fragments and nutshell debris. The earliest remains dated back to 4,300 years ago, revealing that, much like our hominin ancestors, chimpanzees too are leaving behind durable archaeological records (Mercader et al., 2007). It was this excavation that would open the way for the establishment of a field of non-human Primate Archaeology. A re-assessment of the original Taï chimpanzee assemblage is being published as we write this piece (Proffitt et al., in press). In 2008, a team of researchers published a milestone study combining ethological and archaeological methods, including direct and indirect records of tool-use at several stone tool sites by the wild chimpanzees of Bossou, Guinea (Carvalho et al., 2008). It was described soon after as ‘so far the most comprehensive study bringing archaeological methodology to the study of primate tools, applying the theoretical concept of operational sequences to chimpanzee nut cracking and highlighting variability in tool function and regional diversity in tool use.’ (Haslam et al., 2009: 344). Interdisciplinary training in archaeology and primatology was then highlighted as a pre-requisite for the field of Primate Archaeology to emerge. A follow-up conference aiming to bridge these fields (Palaeoanthropology meets Primatology, Ling et al., 2009) led to a review ‘manifesto’ urging the establishment of this field of research (Haslam et al., 2009). Today, primate technology encapsulates a range of different tool-using behaviors from Western chimpanzee (Pan troglodytes verus) nut-cracking with stone or wood hammers and anvils, to the use of stones by Burmese long-tailed Macaque (Macaca fascicularis aurea) to extract sea-snails and other marine invertebrates from their hard shells (Sugiyama and Koman, 1979; Malaivijitnond et al., 2007). It also includes the Brazilian bearded capuchin (Sapajus libidinosus) stone-on-stone percussion for extracting minerals, chimpanzee and orangutan (Pongo sp.) honey dipping with sticks, and nest-building in several ape species (Sept, 1992; Stewart et al., 2011; van Schaik et al., 2003; Sanz and Morgan, 2009; Hernandez-Aguilar, 2009; Proffitt et al., 2016). But, how can archaeological theory and methods enhance our knowledge about primate technology? For example, an excavation of a primate tool site, such as a chimpanzee nut-cracking assemblage, can help determine the age of the behavior at that location (Mercader et al., 2007; Haslam et al., 2016a,b). It can also reveal whether the behavior has remained consistent or has changed over time, and can help identify the timing of technological innovations or loss of them (Haslam et al., 2017; Luncz et al., 2017b). From an active assemblage, it is possible, when analyzed within the broader ecological context, to assess whether a primate population is targeting specific raw materials that they use for tools (Carvalho et al., 2008; Almeida-Warren et al., 2017). Analysis of the mechanical properties of the tools used can also provide information about the knowledge and expertise acquired by individuals in order to be successful and proficient tool-users (Visalberghi et al., 2009). Lastly, comparison with assemblages from other populations can

Cognition j Primate Archaeology

399

help determine whether there is technological conformity or variation of the behavior and which variables may be determining the variation (Carvalho et al., 2008; Luncz et al., 2018), which in turn, when combined with genetic evidence, can provide insight into past dispersal patterns, networks of information transmission, and the evolutionary origins of technology throughout the primate lineage. Perhaps expectedly, primate archaeology, has largely focused on the primate species that use stone tools, particularly stone tools used for extractive foraging via percussive activities. Studies on stone tool-use by chimpanzees (Pan troglodytes), capuchins (Sapajus libidinosus), and long-tailed macaques (Macaca fascicularis aurea) predominate the literature between 2008 and 2018, with a behavioral focus on nut-cracking and shellfish pounding. Lithic technology comprises the vast majority of the early archaeological records (Hovers and Braun, 2009; Shea, 2016) - together with fossil bones, these are the most durable materials that can last for millions of years. Because archaeologists have had to infer past behavior mostly from lithic artifacts, they have well-established methodologies for its study. However, re-constructions of early hominin evolution continue seriously hampered by this limited evidence and the low number of early archaeological sites (Hovers and Braun, 2009; McPherron et al., 2010; De la Torre, 2011; Harmand et al., 2015). Thus, the opportunity to focus on invisible records, i.e., tools made of perishable materials that are unknown to the oldest archaeological records, is another important area of studying extant primate technology. Making use of this ‘living archaeology’, the field has been expanding towards the study of ephemeral technologies, such as nest building, ant-dipping and termite fishing (Stewart et al., 2011; Almeida-Warren et al., 2017; Pascual-Garrido et al., 2012). While the lifespan of perishable tools is relatively short compared to stone tools, there are many archaeological methods that are applicable, such as the selection, re-use and spatial distribution of nesting locations, as well as the selection and transport of raw materials for insect predation. Plant-based technology is the most common and, arguably, the most diverse and complex manifestation of primate tool-use (Whiten et al., 1999; Wynn et al., 2011). Thus, such studies hold the key for obtaining a comprehensive understanding of primate technology as a whole (Fig. 1). At a broader, evolutionary level, the study of perishable technology through an archaeological lense is providing a framework for reconstructing archaeologically invisible aspects of hominin behavior. It is offering information on selection and use of plant-based technology and on landscape use and foraging patterns based on the location and exploitation of resources. This will be crucial to contextualize the currently known archaeological sites and to re-think ranging patterns and locations of potential unknown sites (Almeida-Warren et al., 2017; Stewart et al., 2011; Panger et al., 2002; Pascual-Garrido et al., 2012). Not all primate species use tools in the wild, however. In fact, tool use in the animal kingdom is rare (Beck, 1980). Thus, primate archaeological research has largely focused on the most studied primate of all, the chimpanzee (Pan troglodytes sp.), or in other versatile tool-users: bearded capuchins (Sapajus libidinosus), and Burmese long-tailed macaques (Macaca fascicularis aurea). Chimpanzee tool-use is unique in that it is a generalized feature of all studied populations, includes tool-use that is not for extractive foraging (e.g., hygiene, comfort, see Whiten et al., 1999), and encompasses lengthy tool-kit repertoires of more than 20 different tooltypes in a single community (Carvalho and McGrew, 2012; McGrew, 2004, 2010; Campbell and Carvalho, 2017). Nevertheless, the fact that tool use is not widespread in the primate order, but occurs in apes, New World and Old World monkeys poses in itself and interesting evolutionary and behavioral conundrum.

Fig. 1 Chimpanzee plant tool-use. Top left, chimpanzee termite fishing tools from the Issa valley (Tanzania); bottom left, map of raw material sources at a chimpanzee termite-fishing site in the Issa valley (Tanzania); right, chimpanzee termite fishing.

400

Cognition j Primate Archaeology

Further research operating within the framework of primate archaeology will no doubt begin to tap into the foundations and nuances of primate technological behaviors, gradually building an evolutionary tree of primate technology (Haslam et al., 2017; Campbell and Carvalho, 2017). In the following sections, we discuss the main methodological approaches of primate archaeology and present some of the key findings revealed thus far.

Methods in Primate Archaeology There is a wide selection of archaeological methods that have been applied in primatology, ranging from full scale excavations, surveys and surface documentation of tools and raw material sources, chaînes operatóires (behavioral sequences during tool-use or tool-making), analysis of use-wear (tool damage), analyses of density and distribution of tools at sites, mechanical properties of raw materials used and etc. The methods chosen for a particular study will rely largely on the research question and the type of information the investigation seeks to obtain. Here we summarize the main approaches and their applications.

Archaeological Excavations and Surface Surveys Ultimately, excavation and surface surveys deal with the distribution of artefacts in space and time within a technological assemblage. Archaeological excavations are a vital and important component of traditional archaeological approaches: they provide in-situ material, stratigraphic and depositional contexts, time depth, and chronology through absolute (e.g., radio-carbon, OSL, Ar/Ar) and relative (e.g., biochronology or via diagnostic tool morphologies) dating. Essentially, they help date the behavior, trace its history, track changes over time, and reconstruct the possible social and ecological contexts of the fossilized activity (Renfrew and Bahn, 2015). In primate archaeology, the excavation of tool sites is useful for dating the earliest evidence of a technological behavior, and investigating how the behavior may have changed over time (Mercader et al., 2007; Mercader et al., 2002; Haslam et al., 2016b). Excavation protocols vary widely between archaeological projects and largely depend on the geological context, the nature of the evidence (e.g., what materials will be excavated), and the resolution that the research question demands. Nevertheless, all include a methodology for systematic excavation (e.g., grid-system or trenches, by spits or stratigraphic layers), spatial data collection, as well as documenting and labelling all the collected material. There are several archaeology manuals, textbooks and online material that describe the different methods, some of which we have included in further reading at the end of this article. While an archaeological excavation has its merits in primatology, it requires time, manpower, and expensive, heavy, equipment. One must also consider the implications of disturbing, and potentially destroying, an active tool site and its impact on future activities by the tool-using population in that area. A less invasive and less time-consuming approach is to carry out surface surveys. As mentioned previously, unlike human archaeology which deals with fossilized behaviors, primate archaeology studies the technological behaviors of a living population. Thus, while surface surveys cannot provide a date for the origin of the behavior per se, they can document how these living archaeological sites are formed and record the changes in archaeological signatures in real time (Carvalho et al., 2008, 2011). This can be achieved by tracking, over time, the position and movements of tools, the appearance of new tools, the modification of tools, and their eventual discard. In this way, surface surveys provide a life history of primate tools and the assemblages they are part of. Furthermore, studying these living assemblages generates important tools for the identification of fossilized tool sites, as well as their behavioral interpretations (Almeida-Warren et al., 2018b).

Spatial data collection and GIS Most archaeological research, whether of humans or non-humans, involves spatial data collection, map-making and subsequent spatial analysis. One of the reasons for this is that it allows the record of the exact location of an artefact, but also its relative position to other artefacts, raw material sources, and ecological features that may be associated to the behavior (e.g., nut-trees and watercourses in the case of the chimpanzee nut-cracking; shorelines in the case of long-tailed macaque sea-snail extraction/foraging). This provides rigorous data which is essential in the study of site formation and tool movement over time, raw material sourcing, but also the spatial relationships between technological behaviors and other activities. There are several available tools and approaches for spatial data collection, which vary in cost, accuracy, and portability. To date, the most frequently used in Primate Archaeology research, is the traditional method which requires two measuring-tapes, string, water level, plumb bob and compass (digital or analog) (e.g., Carvalho et al., 2008; Pascual-Garrido et al., 2016; AlmeidaWarren et al., 2017). Known as the distance-azimuth method, one tape is used to measure the horizontal distance from a fixed Datum, the other for the vertical distance between the tape and the object, and the compass serves to determine the direction of the object from the Datum relative to North. The collected data can be converted into standard X, Y, Z coordinates using basic trigonometry. While this method is the most affordable and portable, it is by far the least accurate or precise. While the tapes can provide millimetric resolution, the bowing of the tape at greater distances and the fact that vertical measurements are limited between ground level (measuring upslope) and the measurer’s height (measuring downslope), can cause significant distortions to the data.

Cognition j Primate Archaeology

401

Furthermore, most compasses only have degree resolution, which causes greater error as the distance increases. This is augmented by the fact that all the relative measurements are taken manually, which leaves a lot of room for human error both in reading the measurements as well as recording. However, if the study in question does not require extremely detailed resolution, this is a good and practical option, especially suited to handle the complexities of surveying in the tropical rainforest. Another method, which is the current state-of-the-art in archaeological work, and employed in a few primate studies, is the Total Station. The Total Station was originally built for use in construction and engineering, and provides highly accurate spatial measurements relative to a fixed Datum. This technology uses a laser-beam and integrated digital compass to measure distance, azimuth, vertical and horizontal angles, which are automatically converted into X, Y, Z coordinates. The output can be read directly from the device, or can be downloaded as a digital file. To date, this is the most accurate method of spatial data collection, but it is also the least portable, and the most expensive. While it is the best tool for large scale excavations, it is heavy and cumbersome equipment to carry in remote locations with difficult terrain and closed vegetation, particularly if the research requires data from multiple sites. A very recently adopted, and less know option, is the so-called “paperless survey”, i.e., the use of the Disto X2 (Almeida-Warren et al., 2018b; Almeida-Warren et al., 2018a; Almeida-Warren et al., in prep). It is a portable hand-held laser distance measurer with an integrated digital compass and clinometer, which can be paired to an android device through the open-access software, TopoDroid. Originally developed for mapping cave systems to replace the traditional time-consuming and labor-intensive method of manual measurements and paper sketches (Heeb, 2008). This method started to gain recognition recently in archaeology for data collection of archaeological sites located in caves, where Total Stations are not viable, and GPS signals are absent (Trimmis, 2018, 2015). While it is less accurate as the Total Station method, it has the advantage of being highly portable and considerably more affordable (Redovnikovic et al., 2016). Current empirical work comparing the results of using the three described methods, suggests the DistoX2 as the optimal choice, overall, for the most common Primate Archaeological surveys (Almeida-Warren et al., in prep). The methods above describe several options of relative spatial data collection. Absolute coordinates, for global geographic mapping in GIS, can be determined with the help of a GPS/GNSS device. The GPS/GNSS device is used to provide the geographic coordinates of the Datums, which are then used to translate the relative X, Y, Z distances of points into actual geographic coordinates. There is a wide variety of GPS/GNSS devices available on the market. GPS technology is advancing at a fast pace, with some commercial GNSS systems boasting sub-cm accuracy (e.g., Arrow series by EOS; Trimble R-Series), and even the average smartphone or tablet can now provide GIS data with a five-meter accuracy under open sky (van Diggelen and Enge, 2015). The choice of which to use will largely depend on project budget, and the accuracy required for data collection. However, due to its versatility and portability most Primate Archaeology research uses hand-held GPS devices (e.g., Garmin, Magellan), which typically have an absolute accuracy of three meters under optimal conditions (e.g., minimal canopy cover), although the use of tablets is becoming increasingly popular. To map and analyze the data there are several GIS software applications available, such as ArcGIS and the open source equivalent QGIS. Both provide tools for generating site maps, as well as performing refined analysis of potential spatial relationships between individual artefacts, between assemblages, and other associated features.

Raw Material Studies Raw material studies primarily serve to investigate evidence of selectivity, i.e., if individuals in a population are targeting specific types of materials for their tools. There are two main approaches that can help answer this question. The first is to compare the raw material of the tools that are being used relative to their general availability in the surrounding ecological habitat. This can be achieved by documenting proportions of the different types of materials used, and comparing them to the natural proportions of available materials collected from the tool-site vicinity and/or sample areas (Almeida-Warren et al., 2017; Carvalho et al., 2008; Braun et al., 2008; Braun et al., 2009; Stout et al., 2005). If the proportions are different, it is a good indication that certain raw materials are being sourced over others. The second is to look at the physical and mechanical properties of the raw materials. For primate stone tools these may include size, weight, hardness, friability (Carvalho et al., 2008; Visalberghi et al., 2009). For plant tools these may include length, width, pliability (Almeida-Warren et al., 2017). Typically, this involves taking quantitative measurements of material dimensions and other relevant properties, and experimentally testing the performance of the materials used as tools, relative to other available materials that have not been used. Another form of experimental testing can be achieved by artificially supplying a selection of raw materials, of known physical properties, to a tool-use location and documenting, through observation, which types of materials are most frequently selected by the tool-users for the activity (see Carvalho et al., 2008; Carvalho et al., 2012; Visalberghi et al., 2009). This can not only help determine whether individuals are targeting specific material properties, but also assess cognitive skills, technological performance, and social learning mechanisms (Fig. 2). When the raw material source is known, another useful indicator is the distance between source and tool-site. This may be recorded through direct behavioral observation, or in the case of plant materials it is often possible to identify source plants from removal scarification (see Pascual-Garrido et al., 2012; Almeida-Warren et al., 2017). If raw materials for tools are sourced some distance away from the activity area when other types of suitable materials are available nearby, it suggests that individuals are travelling further afield to search for the best or preferred materials.

402

Cognition j Primate Archaeology

Fig. 2 Chimpanzee stone tool-use. Left, chimpanzee nut-cracking at the experimental outdoor lab in Bossou (Guinea); middle, weighing stone tools used by chimpanzees for nut-cracking in Bossou (Guinea); right, chimpanzee nut-cracking tools and nut shell debris from a site in Bossou (Guinea).

Raw material studies like these can not only help understand the underlying cognitive framework of a tool-based activity, but can also help uncover potential cultural traits of a primate population (Almeida-Warren et al., 2017).

Use-Wear The study of use-wear of tools is useful for understanding the techniques and manual/manipulative skills (e.g., action, precision, grips) involved in the technological behavior (Haslam et al., 2013). Furthermore, in the absence of direct behavioral observation, it can also help isolate the diagnostic features of a tool, which in turn can provide a protocol for identifying materials that have been used as tools from those which have not (Haslam et al., 2013; Caruana et al., 2014; Benito-Calvo et al., 2015). Use-wear can be studied at a macroscopic (macro-wear) or microscopic (micro-wear) level. This can be achieved through producing 3D-models (via photogrammetry or 3D-scanning) of the tools and analyzing their virtual surfaces through morphometric analysis (see Benito-Calvo et al., 2015). In the case of tools used in pounding activities, such as nut-cracking or processing marine invertebrates, this involves documenting pitting, crushing and fracture wear on the tool surface (Proffitt et al., 2018). For plant-based tools such as termite-fishing or ant-dipping, this would involve documenting features like fraying and clipping, although this has yet to be explored. Refined quantitative methods like these are providing the first building-blocks that allow for technological comparisons between different primate species, including hominins.

Tool Modification The ability to modify an object and the way in which this object is modified in order to produce a tool, is often used in Paleolithic archaeology as a measure of technological and cognitive complexity (Shea, 2016). One measure, for example, would be the number of flake detachments that are successfully removed from a cobble to achieve the desired tool (Delagnes and Roche, 2005). While primates only use stone tools that are modified by use, not prior to use, and do not intentionally flake stones in the wild, they do present several instances of tool modification, particularly in chimpanzee plant technology. Plant materials are often curated prior to use by removing leaves and off-shoots, and in some populations the functional ends of termite-fishing tools are modified to produce a brush tip that increases the surface catchment area of the tool, thus making the tool more efficient (e.g., Sanz et al., 2009). As with human archaeology, studying the processes of primate tool modification, can provide valuable insight into the cognitive frameworks of primate technology.

Case Studies While still in its infancy, Primate Archaeology is developing at a very fast pace. In the past decade, the field has been extremely prolific, expanding from chimpanzee archaeology, to studies including capuchins and macaques, and new populations of nonhuman primates have been reported to use tools. A wealth of information has been uncovered, which is continuously building our knowledge on primate technological repertoires and cognitive abilities, but also providing valuable insight into our evolutionary past. Here we present some of the key findings of these studies.

Chimpanzees (Pan Troglodytes) Chimpanzee populations throughout Africa use stone and plant materials for extractive foraging, social interactions and selfmaintenance (McGrew, 2004; Whiten et al., 1999). They use stones and hardwoods as hammers and anvils to crack open nuts;

Cognition j Primate Archaeology

403

wads of chewed-up leaves as sponges to soak up drinking water; and long twigs and strips of bark to harvest social insects (ants, termites, bees) and their products (e.g., honey) (Boesch and Boesch, 1984; McGrew, 2004; Sanz and Morgan, 2010; Whiten et al., 1999). They present the largest and most diverse repertoire compared to any other non-human species (Carvalho et al., 2013). The first non-human archaeological excavations were conducted at a chimpanzee nut-cracking field site where it was found that this behavior dated back to at least 4,300 years (Mercader et al., 2007). This provided the first evidence that chimpanzees, like our human ancestors, are also capable of producing a durable and archaeologically recognizable record. A recent study on use-wear analysis has further supported this, by revealing that stones used by chimpanzees for nut-cracking are quantitatively distinguishable from stone natural damage, and can, therefore, be identified without direct observation of the activity, i.e., in the archaeological record (Benito-Calvo et al., 2015). Subsequent research at the two long-term sites for wild chimpanzee nut-cracking, Taï (Cote d’Ivoire) and Bossou (Guinea), revealed that chimpanzees are selective of the rocks they use as tools (Sirianni et al., 2015; Carvalho et al., 2008) and show preference for re-using the same pair of stones (Carvalho et al., 2009). Furthermore, research on transport distances at a landscape-wide scale in Taï, suggests that hammerstone transport by chimpanzees follows a distance-decay model akin to that of some early hominin sites, in which the weight of the tool decreases with the increase in distance from the nearest raw material sources (Luncz et al., 2016). This pattern was also found in the spatial analysis of the excavated material from Noulo, suggesting that this strategy has probably remained unchanged in the last 4,300 years. Thus, selection for nut-cracking tools, seems to be conditioned by a combination of the physical properties of the raw material (e.g., size, weight, type), the ecology of the technology (e.g., availability and density of the resource), and the energetics of the behavior (costs of sourcing and transportation versus energetic gains) demonstrating that chimpanzees have an in-depth understanding of the physical and mechanical properties of different materials, and are also able to weigh up a series of complex information in order to achieve an optimal foraging strategy (Sirianni et al., 2015; Carvalho et al., 2008). This does not appear to be a universal condition of nut-cracking technology, however, as a recent study comparing different chimpanzee communities of Taï have revealed that there is considerable intra-specific variation in nut-cracking efficiency (Luncz et al., 2018). This variation occurs due to the differential selection of raw materials for tools and differences in the manual techniques used between communities, despite the availability of the same raw materials throughout the forest. These so-called cultural differences persist even with the transfer of immigrant individuals from neighboring communities, who adopt the technological norms of the new group, regardless of its efficiency (Luncz et al., 2018). This suggests, that although chimpanzees may have intimate technical knowledge, their technology is bounded by pressures imposed by the social norms of the community they belong to. While most studies have focused on stone tool use, interest has gradually been expanding towards the use of plant-based tools. Research into chimpanzee termite fishing in the Issa Valley (Tanzania), has also revealed that chimpanzees are targeting specific plant species and plant materials, where bark is the only material used, despite the greater abundance of other types of suitable raw materials (Almeida-Warren et al., 2017). The fact that other populations elsewhere use several kinds, provides tentative evidence of a cultural tradition by the chimpanzees of Issa. Studies of transport distances combined with resource availability data, also suggests that there is some degree of forward planning, in which chimpanzees may source raw materials for tools on the way to the targeted mound (Almeida-Warren et al., 2017). Similar evidence has also been found in the plant-based ant-dipping and honey-dipping technologies of the chimpanzees at Gashaka-Kwano (Nigeria) (Pascual-Garrido et al., 2012). Beyond extractive foraging technologies, researchers have also started studying the ethoarchaeology of chimpanzee nesting patterns (Hernandez-Aguilar, 2009; Stewart et al., 2011). Nest-building is the only technological behavior that manifests in all ape and all chimpanzee populations (Stewart et al., 2011). It’s study, both in terms of architecture, re-use, and spatial distribution may hold important clues for understanding the evolution of spatial cognition, as well as reconstructing the evolution of hominin shelter.

Bearded Capuchins (Sapajus Libidinosus) Capuchins are the most encephalized of the New World monkeys and range Central and South America (Fragaszy et al., 2004). They are known for their manipulative skills and are the only New World monkeys known to use tools, and, out of all capuchin species, only the bearded capuchin (S. libidinosus) is a habitual tool-user (Ottoni and Izar, 2008). This species lives predominantly in the dry savannah-like habitats of Brazil, and engages in a variety of technological activities including cashew nut-cracking with stones, and using plant tools to access small prey (insects, lizards) from tree hollows and rock-cracks, and for honey extraction (Ottoni and Izar, 2008). The also practice stone-on-stone percussion, possibly to access mineral dietary supplements (Proffitt et al., 2016). Recent excavations of capuchin nut-cracking sites in Serra da Capivara (Brazil) have begun to tap into the capuchin archaeological record, revealing that this activity dates back to at least 700 years (Haslam et al., 2016b). In Boa Vista (Brazil), studies on raw material selection in an experimental setting have provided evidence of hammerstone selection based on size, weight and friability (Visalberghi et al., 2009). It also found that this community of capuchins was often able to assess the suitability of a raw material based on visual cues alone. When they were not able to do so (e.g., all tools were visually similar, but different weight), they would manually handle the stones before carrying one to the nut-cracking location. This suggest that capuchins engage in pre-tool-use planning and raw material selection, much like chimpanzees and our early ancestors (Visalberghi et al., 2009). More recently, it has been found that capuchins are capable of producing stone flakes as unintentional by-products of stone-onstone percussion activities (Proffitt et al., 2016). These flakes exhibit the same conchoidal fracturing characteristic of the Oldowan assemblages of our ancestors, which were long thought only possible through intentional and calculated knapping. This discovery

404

Cognition j Primate Archaeology

challenges current methods for identifying intentionality in the archaeological record, highlighting the potential for error when evaluating artefacts without further context. It also gives further support to the older hypothesis that early stone flaking may have first emerged as a bi-product of pounding activities, later developing into an intentional and systematic technology (Mora and De la Torre, 2005; Carvalho et al., 2009). To date, no primate archaeology studies have explored capuchin perishable technology. Plant tools-use by capuchins is often a one-off activity (e.g., flushing a lizard out of a crack in a rocky outcrop, using a nasal probe or toothpick), different from the intensive extractive foraging of chimpanzees’ termite-fishing or ant-dipping in which chimpanzees may be engaged for hours at a time. It is perhaps more comparable to the use of sharpened sticks by the chimpanzees of Fongoli (Senegal) to immobilize and capture bushbabies sleeping in tree hollows (Pruetz and Bertolani, 2007). Nevertheless, the study of this more infrequent type of tool use can be beneficial to understand whether it also produces recognizable archaeological signatures and how the behavioral sequences compare to those of more intensive technological behaviors.

Burmese Long-Tailed Macaques (Macaca Fascicularis Aurea) The newest of the non-human primate tool-users, is the long-tailed macaque. Tool-use has been mostly reported for one subspecies – M. fascicularis aurea – endemic to the coastal areas in southwestern Thailand and Myanmar (Gumert and Malaivijitnond, 2012). However, there are new reports adding the Macaca fascicularis umbrosus to this list (Pal et al., 2008). Their marine-dependent ecological niche has shaped the development of unique behavioral adaptations for the extractive foraging of an impressive variety of different aquatic resources, such as oysters, snails, crabs, and sea-almonds, using stone tools (Gumert and Malaivijitnond, 2012; Falótico et al., 2017). But inland, and like chimpanzees, they have also developed a lithic technology to crack oil palm nuts. Oil palms are not endemic to Thailand and were only introduced in the last decade. This demonstrates that the Burmese macaques have been able to quickly transfer their former technological skill-set, used in the foraging of marine resources, for the exploitation of novel resources located away from the shoreline (Luncz et al., 2017a). Their lithic technology comprises two main techniques: axe-hammering of sessile oysters, and pounding of unattached hardshelled foods such as gastropods, crabs and nuts. For each technique, these macaques select tools with different morphological features and employ different manual skills, i.e., for prying oyster from rock, they use flat, axe-like tools held in a precision grip; for pounding, they use rounder and heavier stones held in a power grip (Gumert et al., 2009). A primate archaeology study on the use-wear patterns of the axes and hammer tools have revealed that these macaque technologies produce distinct damage patterns, and can be reliably identified through the analysis of impact marks left on the tools (Haslam et al., 2013). This also means that this method can be used to study macaque surface tool assemblages and tool life-histories without direct observation, as well as interpret excavated remains. Although transport distances of tools used in marine resource exploitation were recorded as far as 87 m, overall, stones were transported relatively short distances, and were mostly collected within one meter of the target resources (Haslam et al., 2016c). Nevertheless, tools were often reused during sequential foraging bouts, and frequently transported between different prey items, indicating that individuals have a strong tendency to keep hold of a tool throughout a feeding session. Larger tools, such as hammers used for pounding, were transported much shorter distances and were often collected and abandoned within the immediate vicinity of the anvil stones. This behavior produces patterns of spatial association of hammers with anvils, as well as oysterbearing rocks, which also increases the likelihood of tool re-use between individuals and foraging sessions (Haslam et al., 2016c). As with chimpanzees and capuchins, these patterns of accumulation and re-use, produce a recognizable archaeological footprint. The first excavated remains were uncovered in 2016, revealing an assemblage largely composed of axe-hammers used for processing mollusks, and one stone likely used for pounding activities (Haslam et al., 2016a). While the remains were dated to only 65 years ago, the authors speculate, based on geological evidence and sea-level reconstruction, that sites found in the area may have been formed within the last 1,000 years or be older than 7,000 (Haslam et al., 2016a). Future excavations will provide a clearer picture of the site formation processes in this area.

Future Directions Primate archaeology is building a solid profile of tool-use and technology that transcends our anthropocentric existence. It has demonstrated that, like human populations in the past and present, other primates like chimpanzees, capuchins and long-tailed macaques produce a durable and recognizable archaeological record, and that processes of tool selection, reuse, tool-function, transport and accumulation in discrete locations are traits manifested amongst all of the major primate tool-users. We now have comprehensive data on the stone tool behaviors and sites of three extant non-human primates, including dated archaeological records for all species (Haslam et al., 2017). The field is now starting to expand to include analyses of perishable tools, which represents the most prevalent material used to make tools and is the only tool type that primates modify prior to use. In the future, it will be important to include studies of the perishable tools used by another great ape, the arboreal orangutan (Pongo sp.). The unusual challenges of the primate environments, as well as the nature of the archaeological evidence have also spurred methodological innovations (e.g., using GIS to map tool surfaces, use of DistoX2 to map tool sites), which, in turn, are benefiting Paleolithic archaeological research by opening the way to standardized data collection across sites that will allow reliable

Cognition j Primate Archaeology

405

comparisons of technological records within the primate order (see Almeida-Warren et al., in prep.). Likewise, ethoarchaeology of primate technology is providing valuable insights into how archaeological sites are formed and the dynamic processes shaping patterns of accumulation, and how these fit within the broader contexts of landscape use and resource exploitation (McGrew, 2010). But Primate Archaeology is also pushing forward the studies of the Early Stone Age. Archaeologists focusing on the origins of technology started surveying Pliocene (>2.6 Ma) deposits in Africa by 2008. Based on a common denominator – pounding tools – the ubiquitous tool type shared by human and non-human primates and present across all chronological periods (Caruana et al., 2014), parsimony suggested that the oldest assemblages would be composed, by large, of pounding tools. The finding of the Lomekian tools dating from 3.4 Ma (Harmand et al., 2015) confirmed this premise and current work indicates that Pliocene Archaeology continues to expand (Tuosto et al., 2017). As the earliest tool-makers moved from our genus Homo to earlier human ancestors, including the small brained Australopithecines, and archaeologists continue to excavate archaeological records of modern nonhuman primates, it becomes a real possibility to find (and recognize) records with millions of years that may belong to nonhuman primates. Moreover, with the contribution of Primate Archaeologists, paleoanthropologists are putting forward new hypotheses, challenging old queries about the origins of hunting and scavenging behavior (Thompson et al., in press). Importantly, the field of Primate Archaeology has also raised issues that can have an important effect on primate conservation. Recently, a call for studying threatned ape populations was made, urging the documentation and preservation of their cultural heritage (Hockings et al., 2015). Finally, while archaeological methods have so far only been applied within the primate order, they are by no means restricted to it. The goals of primate archaeology and the methodologies described in this article have equal potential in the study of many other tool-using species whether mammal, bird or fish. It will provide an invaluable tool to unveiling the true richness of animal cognitive, technological, and cultural behaviors, as well as the ecological contexts and evolutionary paths that have led to the emergence of technology throughout the animal kingdom.

References Almeida-Warren, K., Sommer, V., Piel, A.K., Pascual-Garrido, A., 2017. Raw material procurement for termite fishing tools by wild chimpanzees in the Issa valley, Western Tanzania. American Journal of Physical Anthropology 164, 292–304. Almeida-Warren, K., Braun, D.R., Matsuzawa, T., Carvalho, S., 2018a. Mapping primate archaeology: New tools for standardizing spatial data collection across the primate order. In: Proceedings of the XIVIIth Congress of the International Primatological Society, Nairobi. Almeida-Warren, K., Braun, D.R., Regala, F.T., et al., 2018b. The ecology of chimpanzee technology: New approaches for modelling early hominin ranging patterns and resource exploitation. In: Proceedings of the XIVIIth Congress of the International Primatological Society, Nairobi. Almeida-Warren, K., et al. in prep. The DistoX2 survey method: towards the standardisation of spatial data collection in primate archaeology. Beck, B.B., 1980. Animal Tool Use: The Use and Manufacture of Tools by Animals. Garland, New York. Benito-Calvo, A., Carvalho, S., Arroyo, A., Matsuzawa, T., De LaTorre, I., 2015. First GIS analysis of modern stone tools used by wild chimpanzees (Pan troglodytes verus) in Bossou, Guinea, West Africa. PLOS ONE 10, 1–22. Boesch, C., Boesch, H., 1984. Possible causes of sex differences in the use of natural hammers by wild chimpanzees. Journal of Human Evolution 13, 415–440. Braun, D.R., Plummer, T., Ditchfield, P., et al., 2008. Oldowan behavior and raw material transport: Perspectives from the Kanjera Formation. Journal of Archaeological Science 35, 2329–2345. Braun, D.R., Plummer, T., Ferraro, J.V., Ditchfield, P., Bishop, L.C., 2009. Raw material quality and Oldowan hominin toolstone preferences: Evidence from Kanjera South, Kenya. Journal of Archaeological Science 36, 1605–1614. Campbell, R., Carvalho, S., 2017. Tool use and manufacture in the last common ancestor of Pan and Homo. In: Muller, M., Wrangham, R., Pilbeam, D. (Eds.), Chimpanzees and Human Evolution. Harvard University Press, Harvard, pp. 602–644. Caruana, M.V., Carvalho, S., Braun, D.R., et al., 2014. Quantifying traces of tool use: A novel morphometric analysis of damage patterns on percussive tools. PLOS ONE 9, 1–18. Carvalho, S., Sousa, C., Matsuzawa, T., 2007. New nut-cracking sites in Diecké forest, Guinea: An overview of the etho-archaeological surveys. Pan African News 14 (1), 11–13. Carvalho, S., Cunha, E., Sousa, C., Matsuzawa, T., 2008. Chaînes opératoires and resource-exploitation strategies in chimpanzee (Pan troglodytes) nut cracking. Journal of Human Evolution 55, 148–163. Carvalho, S., Biro, D., McGrew, W.C., Matsuzawa, T., 2009. Tool-composite reuse in wild chimpanzees (Pan troglodytes): Archaeologically invisible steps in the technological evolution of early hominins? Animal Cognition 12, 103–114. Carvalho, S., Cunha, E., Sousa, C., Matsuzawa, T., 2011. Extensive surveys of chimpanzee stone tools: From the telescope to the magnifying glass. In: Matsuzawa, T., Humle, T., Sugiyama, Y. (Eds.), The Chimpanzees of Bossou and Nimba, first ed. Springer Japan, Tokyo, pp. 145–155. Carvalho, S.C., Braun, D.R., Harris, J.W.K., McGrew, W.C., Matsuzawa, T., 2012. Wild Chimpanzee (Pan troglodytes) and Pleistocene hominin raw material preference compared: Evidence from archaeological data and natural experiments. PaleoAnthropology 2012, A7–A8. Carvalho, S., McGrew, W.C., 2012. The origins of the Oldowan: Why chimpanzees (Pan troglodytes) still are good models for technological evolution in Africa. In: DomínguezRodrigo, M. (Ed.), Stone Tools and Fossil Bones: Debates in the Archaeology of Human Origins. Cambridge University Press, Cambridge, pp. 222–244. Carvalho, S., Matsuzawa, T., McGrew, W., 2013. From pounding to knapping: How chimpanzees can help us to model hominin lithics. In: Sanz, C.M., Call, J., Boesch, C. (Eds.), Tool Use in Animals: Cognition and Ecology. Cambridge University Press, Cambridge, pp. 225–241. Carvalho, S., Beardmore-herd, M., in press. Technological origins: Primate perspectives and Early hominin tool use. Oxford Research Encyclopedia of African History. Delagnes, A., Roche, H., 2005. Late Pliocene hominid knapping skills: The case of Lokalalei 2C, West Turkana, Kenya. Journal of Human Evolution 48, 435–472. De la Torre, I., 2011. The origins of stone tool technology in Africa: A historical perspective. Philosophical Transactions of the Royal Society of London B: Biological Sciences 366, 1028–1037. Falótico, T., Spagnoletti, N., Haslam, M., et al., 2017. Analysis of sea almond (Terminalia catappa) cracking sites used by wild Burmese long-tailed macaques (Macaca fascicularis aurea). American Journal of Primatology 79, 1–8. Fragaszy, D., Izar, P., Visalberghi, E., Ottoni, E. B., De Oliveira, M. G. 2004. Wild capuchin monkeys (Cebus libidinosus) use anvils and stone pounding tools. American Journal of Primatology, 64(4), 359–366. Goodall, J., 1963. Feeding behaviour of wild chimpanzees. Symposia of the Zoological Society of London 10, 39–48.

406

Cognition j Primate Archaeology

Gumert, M.D., Malaivijitnond, S., 2012. Marine prey processed with stone tools by burmese long-tailed macaques (Macaca fascicularis aurea) in intertidal habitats. American Journal of Physical Anthropology 149, 447–457. Gumert, M., Kluck, M., Malaivijitnond, S., 2009. The physical characteristics and usage patterns of stone axe and pounding hammers used by Long-tailed Macaques in the Andaman Sea region of Thailand. American Journal of Primatology 608, 594–608. Harmand, S., Lewis, J.E., Feibel, C.S., et al., 2015. 3.3-million-year-old stone tools from Lomekwi 3, West Turkana, Kenya. Nature 521, 310–315. Haslam, M., Hernandez-Aguílar, A., Ling, V., et al., 2009. Primate archaeology. Nature 460, 339–444. Haslam, M., Gumert, M.D., Biro, D., Carvalho, S., Malaivijitnond, S., 2013. Use-wear patterns on wild macaque stone tools reveal their behavioural history. PLOS ONE 8, 1–8. Haslam, M., Hernandez-Aguilar, R.A., Proffitt, T., et al., 2017. Primate archaeology evolves. Nature Ecology and Evolution 1, 1431–1437. Haslam, M., Luncz, L., Pascual-garrido, A., et al., 2016a. Archaeological excavation of wild macaque stone tools. Journal of Human Evolution 96, 134–138. Haslam, M., Luncz, L.V., Staff, R.A., et al., 2016b. Pre-columbian monkey tools. Current Biology 26, R521–R522. Haslam, M., Pascual-garrido, A., Malaivijitnond, S., Gumert, M., 2016c. Stone tool transport by wild Burmese long-tailed macaques (Macaca fascicularis aurea). Journal of Archaeological Science: Reports 7, 408–413. Heeb, B., 2008. Paperless caving – An electronic cave surveying system. In: Proceedings of the IV European Speleological Congress. Lans-en-Vercors, France, pp. 130–133. Hernandez-Aguilar, R.A., 2009. Chimpanzee nest distribution and site reuse in a dry habitat: Implications for early hominin ranging. Journal of Human Evolution 57, 350–364. Hernandez-Aguilar, R.A., Moore, J., Pickering, T.R., 2007. Savanna chimpanzees use tools to harvest the underground storage organs of plants. Proceedings of the National Academy of Sciences of the United States of America 104 (49), 19210–19213. Hockings, K.J., McLennan, M.R., Carvalho, S., et al., 2015. Apes in the Anthropocene: Flexibility and survival. Trends in Ecology and Evolution 30, 215–222. Hovers, E., Braun, D.R., 2009. Interdisciplinary Approaches to the Oldowan. Springer, New York. Joulian, F., 1994. Culture and material culture in chimpanzees and early hominids. In: Roeder, J.J., Thierry, B., Anderson, J.R., Herrenschmidt, N. (Eds.), In: Proceedings of the XIVth Congress of the International Primatological Society, Current Primatology Vol. II: Social Development, Learning and Behaviour. Universite Louis Pasteur, Strasbourg, pp. 404–497. Joulian, F., 1996. Comparing chimpanzee and early hominid techniques: Some contributions to cultural and cognitive questions. In: Mellars, P., Gibson, K.R. (Eds.), Modelling the Early Human Mind. McDonald Institute for Archaeological Research, Cambridge, pp. 173–189. Ling, V., Hernandez-Aguilar, A., Haslam, M., Carvalho, S., 2009. The origins of percussive technology: A smashing time in Cambridge. Evolutionary Anthropology 18, 48–49. Luncz, L.V., Proffitt, T., Kulik, L., Haslam, M., Wittig, R.M., 2016. Distance-decay effect in stone tool transport by wild chimpanzees. Proceedings of the Royal Society B: Biological Sciences 283, 20161607. Luncz, L.V., Svensson, M.S., Haslam, M., Malaivijitnond, S., 2017a. Technological response of wild macaques (Macaca fascicularis) to anthropogenic change. International Journal of Primatology 38, 872–880. Luncz, L.V., Tan, A., Haslam, M., et al., 2017b. Resource depletion through primate stone technology. Elife 6, e23647. Luncz, L.V., Sirianni, G., Mundry, R., Boesch, C., 2018. Costly culture: Differences in nut-cracking efficiency between wild chimpanzee groups. Animal Behaviour 137, 63–73. Malaivijitnond, S., Lekprayoon, C., Tandavanittj, N., et al., 2007. Stone-tool usage by Thai long-tailed macaques (Macaca fascicularis). American Journal of Primatology 69, 227–233. McGrew, W.C., 2004. The Cultured Chimpanzee: Reflections on Cultural Primatology. Cambridge University Press, Cambridge. McGrew, W.C., Marchant, L.F., Hunt, K.D., 2007. Etho-archaeology of manual laterality: Well digging by wild chimpanzees. Folia Primatologica 78, 240–244. McGrew, W.C., 2010. In search of the last common ancestor: New findings on wild chimpanzees. Philosophical Transactions of the Royal Society of London B 365, 3267–3276. McPherron, S.P., Alemseged, Z., Marean, C.W., et al., 2010. Evidence for stone-tool-assisted consumption of animal tissues before 3.39 million years ago at Dikika, Ethiopia. Nature 466, 857–860. Mercader, J., Barton, H., Gillespie, J., et al., 2007. 4300-year-old chimpanzee sites and the origins of percussive stone technology. Proceedings of the National Academy of Sciences of the United States of America 104, 3043–3048. Mercader, J., Panger, M., Boesch, C., 2002. Excavation of a chimpanzee stone tool site in the African rainforest. Science 296, 1452–1455. Mora, R., De la Torre, I., 2005. Percussion tools in Olduvai Beds I and II (Tanzania): Implications for early human activities. Journal of Anthropological Archaeology 24, 179–192. Ottoni, E.B., Izar, P., 2008. Capuchin monkey tool use: Overview and implications. Evolutionary Anthropology 17, 171–178. Pal A., Kumara H.N., Mishra P.S., Velankar A.D., Singh M., 2008. Extractive foraging and tool-aided behaviors in the wild Nicobar long-tailed macaque (Macaca fascicularis umbrosus). Primates, 59(2), 173–183. Panger, M.A., Brooks, A.S., Richmond, B.G., Wood, B., 2002. Older than the Oldowan? Rethinking the emergence of hominin tool use. Evolutionary Anthropology 11, 235–245. Pascual-Garrido, A., Buba, U., Nodza, G., Sommer, V., 2012. Obtaining raw material: Plants as tool sources for Nigerian chimpanzees. Folia Primatologica 83, 24–44. Pascual-Garrido, A., Sommer, V., Almeida-Warren, K., 2016. Raw material procurement for termite fishing tools in wild chimpanzees. PeerJ Preprints 4, e1844v1. Proffitt, T., Luncz, L.V., Falótico, T., et al., 2016. Wild monkeys flake stone tools. Nature 539, 85–88. Proffitt, T., Luncz, V.L., Malaivijitnond, S., et al., 2018. Analysis of wild macaque stone tools used to crack oil palm nuts. Royal Society: Open Science 4, 171904. Proffitt, T., Haslam, M., Mercader, J., Boesch, C., Luncz, L.V., in press. Revisiting Panda 100, the first archaeological chimpanzee nut cracking site. Journal of Human Evolution. Pruetz, J.D., Bertolani, P., 2007. Savanna chimpanzees, Pan troglodytes verus, hunt with tools. Current Biology 17, 412–417. Redovnikovic, L., Stancic, B., Cetl, V., 2016. Comparison of different methods of underground survey. In: Proceedings of the International Symposium on Engineering Geodesy. Varazdin, Croatia, pp. 465–472. Renfrew, C., Bahn, P., 2015. Archaeology Essentials: Theories, Methods, and Practice, third ed. Thames & Hudson. Roche, H., 2016. Behavioural biology: Stones that could cause ripples. Nature 539, 34–35. Sakura, O., Matsuzawa, T., 1991. Flexibility of wild chimpanzee nut-cracking behavior using stone hammers and anvils: An experimental analysis. Ethology 87, 237–248. Sanz, C., Call, J., Morgan, D., 2009. Design complexity in termite-fishing tools of chimpanzees (Pan troglodytes). Biology Letters 5, 293–296. Sanz, C.M., Morgan, D.B., 2009. Flexible and persistent tool-using strategies in honey-gathering by wild chimpanzees. International Journal of Primatology 30, 411–427. Sanz, C.M., Morgan, D.B., 2010. The complexity of chimpanzee tool-use behaviours. In: Lonsdorf, E.V., Ross, S.R., Matsuzawa, T., Goodall, J. (Eds.), The Mind of the Chimpanzee: Ecological and Experimental Perspectives. University of Chicago Press, Chicago, pp. 127–140. Sept, J., 1992. Was there no place like home? A new perspective on early hominid archaeological sites from the mapping of chimpanzee nests. Current Anthropology 33, 187–207. Shea, J.J., 2016. Stone Tools in Human Evolution: Behavioral Differences Among Technological Primates. Cambridge University Press, Cambridge. Sirianni, G., Mundry, R., Boesch, C., 2015. When to choose which tool: Multidimensional and conditional selection of nut-cracking hammers in wild chimpanzees. Animal Behaviour 100, 152–165. Stewart, F.A., Piel, A.K., McGrew, W.C., 2011. Living archaeology: Artefacts of specific nest site fidelity in wild chimpanzees. Journal of Human Evolution 61, 388–395. Stout, D., Quade, J., Semaw, S., Rogers, M.J., Levin, N.E., 2005. Raw material selectivity of the earliest stone toolmakers at Gona, Afar, Ethiopia. Journal of Human Evolution 48, 365–380. Sugiyama, Y., Koman, J., 1979. Tool-using and -making behavior in wild chimpanzees at Bossou, Guinea. Primates 20, 513–524. Thompson, J., Carvalho, S., Marean, C., Alemseged, Z., in press. The origins of the human predatory pattern: The transition to large animal exploitation by early hominins. Current Anthropology. Toth, N., Schick, K., 2009. The Oldowan: The tool making of early hominins and chimpanzees compared. Annual Review of Anthropology 38, 289–305. Toth, N., Schick, K., Semaw, S., 2006. A comparative study of the stone tool-making skills of Pan, Australopithecus, and Homo sapiens. In: Toth, N., Schick, K. (Eds.), The Oldowan: Case Studies Into the Earliest Stone Age. Stone Age Institute Press, Gosport, IN, pp. 155–222. Trimmis, K.P., 2015. Fear of the light: Mapping modern cave use strategies in Kythera Island caves. Ethnoarchaeology 7, 141–156.

Cognition j Primate Archaeology

407

Trimmis, K.P., 2018. Paperless mapping and cave archaeology: A review on the application of DistoX survey method in archaeological cave sites. Journal of Archaeological Science: Reports 18, 399–407. Tuosto, K., Ascoli, S., Braun, D.R., et al., 2017. Selectivity among Early Pleistocene hominins: New evidence from the Koobi Fora Formation. PaleoAnthropology 2017, A7–A8. van Diggelen, F., Enge, P., 2015. The world’s first GPS MOOC and worldwide laboratory using smartphones. In: Proceedings of the 28th International Technical Meeting of the Satellite Division of the Institute of Navigation (ION GNSSþ 2015). Tampa, Florida, pp. 361–369. van Schaik, C.P., Ancrenaz, M., Borgen, G., et al., 2003. Orangutan cultures and the evolution of material culture. Science 299, 102–105. Visalberghi, E., Addessi, E., Truppa, V., et al., 2009. Selection of effective stone tools by wild bearded capuchin monkeys. Current Biology 19, 213–217. Whiten, A., Goodall, J., McGrew, W.C., et al., 1999. Cultures in chimpanzees. Nature 399, 682–685. Wynn, T., McGrew, W.C., 1989. An ape’s view of the Oldowan. Man 383–398. Wynn, T., Hernandez-Aguilar, R.A., Marchant, L.F., McGrew, W.C., 2011. “An ape’s view of the oldowan” revisited. Evolutionary Anthropology 20, 181–197.

Further Reading Biro, D., Haslam, M., Rutz, C., 2013. Tool use as adaptation. Philosophical Transactions of the Royal Society of London B Biological Sciences 368 (1630). https://doi.org/10.1098/ rstb.2012.0408. Biro, D., Inoue-Nakamura, N., Tonooka, R., et al., 2003. Cultural innovation and transmission of tool use in wild chimpanzees: Evidence from field experiments. Animal Cognition 6, 213–223. Buchanan, B., O’Brien, M.J., Eren, M.I., 2018. Convergent Evolution in Stone-Tool Technology. MIT Press, Cambridge, MA. Marchant, L.F., McGrew, W.C., 2005. Percussive technology: Chimpanzee baobab smashing and the evolutionary modeling of hominid knapping. In: Roux, V., Bril, B. (Eds.), Stone Knapping: The Necessary Conditions for a Uniquely Human Behaviour. McDonald Institute for Archaeological Research, Cambridge, pp. 341–350. Matsuzawa, T., 2011. Field experiments of tool-use. In: Matsuzawa, T., Humle, T., Sugiyama, Y. (Eds.), The Chimpanzees of Bossou and Nimba, first ed. Springer Japan, Tokyo, pp. 145–155. Mercader, J., 2002. Under the Canopy: The Archaeology of Tropical Rain Forests. Rutgers University Press, New Jersey. Muller, M.N., 2017. Chimpanzees and Human Evolution. Harvard University Press, Harvard. Renfrew, C., Bahn, P., 2012. Archaeology: Theories, Methods, and Practice, sixth ed. Thames and Hudson, London. Toth, N., Schick, K., 2006. The Oldowan: Case Studies into the Earliest Stone Age. Stone Age Institute Press, Gosport, IN. Sanz, C.M., Call, J., Boesch, C., 2013. Tool Use in Animals: Cognition and Ecology. Cambridge University Press, Cambridge.

Relevant Websites http://www.oum.ox.ac.uk/visiting/current.htm–O.U.M.N.H. Exhibitions –Oxford University Museum of Natural History. https://primobevolab.web.ox.ac.uk–Primate Models for Behavioural Evolution Lab. https://www.scientificamerican.com/article/ancient-stone-tools-force-rethinking-of-human-origins/–Scientific American –Ancient Stone Tools Force Rethinking of Human Origins. https://www.sciencenews.org/article/aping-stone-age–Science News –Aping the Stone Age.

Problem-Solving in Tool-Using and Non-Tool-Using Animals AM Seed and J Call, Max Planck Institute for Evolutionary Anthropology, Leipzig, Germany © 2010 Elsevier Ltd. All rights reserved.

Abstract Physical problem-solving is defined as the use of a novel means to reach a goal when direct means are unavailable. Problemsolving in the wild (approximated by reports of innovation) correlates with relative forebrain size in mammals and birds. In the laboratory, various cognitive and noncognitive factors influence problem-solving, making species comparisons difficult, but there is no evidence that tool-users outperform nontool-users. Most work has focused on large-brained primates and corvids, which show fast and flexible problem-solving in many contexts. Several theories concerning the cognitive processes are involved, including roles for predispositions, learning, representation, causal knowledge, and inference.

Keywords Cognition; Corvids; Object manipulation; Primates; Problem-solving; Representation; Tools

The majority of mobile animals need to locate their food in space (and in some cases, time). In addition, some kinds of food need to be extracted or processed before they can be consumed, and the pressure to exploit these resources, through extractive foraging or tool-use, has been hypothesized to have selected for advanced cognitive abilities. This article will focus on physical problem-solving in mammals and birds and review the evidence for the cognition underpinning it. A recurring question is whether animals that solve complex problems in their environment differ from those that do not, in terms of what, or how, they learn. Early research on the ability of animals to solve physical problems focused on tool-users, but these are not the only species to exploit embedded or otherwise defended resources, and so our review will encompass evidence from tool-using and nontool-using animals alike.

When Does a Problem Become a Problem? Animals face many problems in their lives; they need to find food, avoid predators, reproduce and, in some cases, care for their offspring. This article is concerned with the problems animals face in the course of accessing and processing food. Some species never face these sorts of physical problems in their natural habitat; their food can be processed directly through morphological adaptations. However, some species take advantage of embedded or otherwise defended foods, despite the fact that evolution has not equipped them to process such resources directly. For example, although rodents, such as agoutis, can gnaw through tough nuts to get access to their kernels, other species must use tools to smash them open (such as chimpanzees and capuchin monkeys), or drop them from a height onto a hard surface (a tactic employed by several corvid species). Similarly, the aye-aye, a species of prosimian, native to Madagascar, has an elongated middle digit which it uses to probe tree holes for insect larvae. However, species without this adaptation, such as chimpanzees, woodpecker finches, and New Caledonian crows use stick tools to fish for invertebrates instead. Physical problems then, such as extracting a larva from a tree or opening a nut, are not simply a feature of the environment, but are modulated by the nature of the animal encountering them. Kohler, one of the first to research the physical problem-solving abilities of animals, described a problem as follows: Something is to be achieved with regard to . a situation; but, as the situation is given, it cannot be achieved. How must we change the situation so that the difficulties disappear and the problem is solved? (Kohler, 1969, p. 134)

One of the most conspicuous candidates for problem-solving is tool-use, when an animal can be seen to reach beyond the limitations of its own body to gain an otherwise inaccessible resource. Animals in the wild also solve problems through direct manipulation of the environment; this can be identified when an expert in a particular species witnesses an individual encountering a new problem and finding a solution to it outside of the usual behavioral repertoire of that species. However, some behavioral solutions have a large genetically determined component, and are therefore, comparable to a morphological adaptation. Further study is needed to identify behavior as novel, and because of the interplay between genes and the environment, this may not be an easy task. For example, chimpanzees and orangutans are known to use tools customarily in the wild, but gorillas and bonobos do not. However, in captivity, all four species of great apes use tools. Nevertheless, regardless of whether tool-using per se is genetically predetermined in some or all of the ape species, the uses to which they are put are very variable and do not seem to be genetically fixed. Some of the difficulties in defining innovation and problem-solving can be circumvented by presenting animals with novel problems in the laboratory that prevent them from using morphological or behavioral adaptations; their problem-solving abilities can then be studied, although this approach is not without its caveats. This article will address four questions about problem-solving

408

Encyclopedia of Animal Behavior, 2nd edition, Volume 1

https://doi.org/10.1016/B978-0-12-813251-7.00090-0

Cognition j Problem-Solving in Tool-Using and Non-Tool-Using Animals

409

in animals. What selective pressures favor its evolution? What sort of problems can animals solve, and what makes one species better at solving problems than another? Finally, what cognitive mechanisms underpin problem-solving in different animal species?

The Evolution of Problem-Solving Problem-solving is costly, for two main reasons. First, interacting with new objects leads to risks, both of actual physical harm (from resources such as poisonous caterpillars and nettles), and from the prospect of investing valuable foraging time into a nonprofitable venture (some problems may prove insoluble). In environments where these risks are large and the incentives small (because of the direct availability of reliable resources) it will be more adaptive to be cautious and conservative. Second, in both birds and primates, those species for which higher rates of innovation and tool-use are reported have relatively enlarged areas of the brain involved in cognitive processes such as inhibition and rule-learning (the mesopallium and nidopallium in birds and the isocortex and striatum in primates). Neural tissue is expensive in terms of energy demands (requiring around nine times more energy than the average requirements of body tissue), and animals maintaining a surplus would be selected against. Given the costs of problem-solving, why would it evolve? One theory involves direct selection for problem-solving itself, in environments where the benefits of accessing defended food outweigh the costs of the risk of injury and a large brain. Candidates for the important features of these environments are: low availability of directly accessible foods; high availability of defended foods; availability of defended foods with high nutritional quality; and unpredictability (either through seasonal variation or rapid environmental change such as occurs in urban environments). Of course, these features are not mutually exclusive; need and opportunity can act in combination, and different features may have been acting during the evolution of different problem-solving species. Evidence for the importance of different environmental features comes from two sources. The first is the distribution of problem-solving behavior across populations of a single species occupying different habitats. For example, woodpecker finches use tools to probe tree holes for invertebrates more frequently in arid habitats in the dry season, where the availability of surface prey is low and the availability and nutritional content of embedded prey is high, compared both to the wet season and to other habitats, where the opposite is true (Tebbich et al., 2002). Similarly, capuchin monkeys use stones to dig for high-quality tubers in arid environments with low availability of easily accessible foods, but have not been reported to do so in habitats rich in other food sources (Moura and Lee, 2004). Another source of evidence comes from the survival rates of different species of bird during introduction events to new environments. Large-brained species show higher survival rates, through higher rates of innovation. Such studies give clues as to the selective pressures that originally favored the evolution of problem-solving and the associated neural structures. Alternatively, selection in another domain may have resulted in neural adaptations that, as a by-product, also allowed animals to solve physical problems. There are several hypotheses for the evolutionary pressures favoring increased brain size and intelligence. Some theories cite ecological factors other than those mentioned above, such as a patchy distribution of food in time and space. Other theories emphasize the challenges and opportunities stemming from social living, such as increased competition; cooperating to secure resources unavailable to individuals; and learning from others. Evidence supporting the Social Brain hypothesis and its variants comes from positive correlations between relative brain size and group size in mammals. Social system is also related to brain size in birds, with cooperative breeders and long-term monogamous species having the largest brains. Amici et al. (2008) report evidence for a link between the social environment and adaptations that impact on problem-solving. They found from a comparative study of inhibitory skills among seven species of primate that the degree of fission-fusion dynamics (splitting and merging into subgroups) predicted success on tasks in which a prepotent response, such as reaching directly for food, had to be inhibited. Inhibition is certainly a cognitive skill that would benefit a species with high fission–fusion dynamics (if group mates are not seen for some time, relations among them can change), but that also would be important for problem-solving, which requires the repression of an ineffective direct approach. Whether direct or indirect selection is responsible (and the two alternatives are not mutually exclusive), comparing animal problem-solving in the wild certainly reveals differences between species, with some groups of animals emerging as the most frequent innovators and tool-users; the great apes amongst primates (such an analysis has not been done with other mammals), and the corvids amongst birds. However, are these differences ascribable to a property of the animals themselves or the habitats in which they are found? Of course, the relatively large forebrains of these groups suggest that it is the former, but it could be the case that infrequent tool-users and innovators lack the incentive, rather than the ability, to solve problems. To comprehensively address whether there are differences between species in the types of problems they can solve, comparative tests under controlled conditions are needed.

What Sorts of Physical Problems Can Animals Solve? Observations of animal problem-solving in nature have inspired experiments to try and uncover what problems animals can solve under controlled conditions, and to characterize the cognitive processes involved. Most of this work has been conducted with chimpanzees, inspired by the striking observations of tool use and manufacture by this species in the wild and its close relationship to humans. However, other species have been incorporated, mostly from those highly innovative groups, the corvids and primates. A few studies have also been conducted on other mammals and birds. Table 1 summarizes the results of the main studies on which

410

Cognition j Problem-Solving in Tool-Using and Non-Tool-Using Animals

Table 1

The main physical problem-solving studies conducted in the laboratory on more than one species (species are shown in the legend) Apes

Property Shape Form Length Width Contact Connection

Support Material Rigidity Continuity/solidity

Weight

Primates

Mammals

Corvids

Birds

Task

1

2

3

4

5

1

2

3

4

5

6

7

8

1

2

3

1

2

3

1

2

3

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

Y Y Y – Y Y –

Y – Y – Y – –

Y – Y Y Y Y –

– – – Y – Y –

– – – – – – –

Y Y Y – Y Y Y

– – – – – Y –

– – – – – – –

– – – – – – –

– – – – – – –

– – – – – Y –

– – – – – – –

– – – – – – –

– – – – – – –

– – – – – – –

– – – – – – –

Y Y Y Y – Y Y

– – – – – – –

– – – – – – –

– – Y Y – – –

– – – – – – –

– – – – – – –

8. Inverted rake 9. String pull 10. Parallel 11. Crossed 12. Broken 13. Contact tube 14. Cloth

N Y Y Y Y – Y

– Y Y Y Y – Y

– Y Y Y Y Y Y

– Y Y Y Y – Y

– Y Y – – – –

– Y Y Y – – Y

– Y Y Y – – Y

– Y Y Y – – –

– Y Y N – – –

Y – – – Y – Y

N Y Y – Y – Y

– Y Y Y – – –

– – – – – – –

– Y Y N – – –

– Y Y – Y – Y

Y – – – – – –

– – – – – – –

– Y – – – Y –

– Y Y Y – – –

– – – – – – –

– Y Y Y – – Y

– Y – – Y – –

15. 16. 17. 18. 19. 20. 21.

Y Y Y Y – Y Y

– Y – – – Y Y

– Y – – – Y –

– N – – – Y –

– – – Y – – –

– Y – Y Y – Y

– – – – – – Y

– – – – – – –

– – – – – – Y

Y – – Y – – –

Y – – N – – –

– – – – – – –

– – – – – – Y

– – – – – N –

– – – – – – –

– – – – – – –

N Y Y Y – – –

– Y Y – – – –

– – – – – – –

– Y – – – – –

– – – – – – –

– – – – – – –

Bundled sticks Unbend wire tool Modify tool Tool length choice Assemble tool Meta-tool Tool diameter

Flimsy tool Trap tube Two-trap task Trap table Obstacles Boards Discrimination

Y indicates that at least one individual solved at least one configuration of the problem. N indicates that no individuals solved the problem. A dash indicates that the species has not been tested. Species shaded in gray are habitual tool-users in the wild. A (Apes): 1 – Chimpanzee; 2 – Bonobo; 3 – Orangutan; 4 – Gorilla; 5 – Gibbon. P (Primates – other): 1 – Capuchin; 2 – Macaque (various); 3 – Spider monkey; 4 – Squirrel monkey; 5 – Vervet monkey; 6 – Callatrichid; 7 – Baboon; 8 – Long-tailed lemur. M (Mammals – other): 1 – Dogs; 2 – Elephants; 3 – Degus. C (Corvids): 1 – New Caledonian Crow; 2 – Rook; 3 – Raven. B (Birds – other): 1 – Woodpecker finch; 2 – Parrot; 3 – Pigeon.

more than one species has been tested, organized in terms of the physical principle involved. In each case, the problem is the retrieval of a food reward, to which direct access has been blocked. The subject either needs to modify the situation (e.g., creating a long stick tool by putting three smaller ones together), or make a choice between two or more options (e.g., pulling an effective tool rather than an ineffective one, which might be long, complete, rigid or connected to food, rather than short, broken, floppy, or unconnected to food). The problems and their solutions are shown pictorially in Figure 1. For a detailed description of the results of most of these studies, see Tomasello and Call (1997). From this overview, it is clear that animals are able to solve problems of a great many varieties involving the physical principles of shape, connectedness, and material properties (such as rigidity, continuity, and weight), including some that were thought to be the unique preserve of humans, such as meta-tool use. However, Table 1 merely indicates which species were able to find the solution to the problem over the course of an experiment, and it must be emphasized that species differ considerably in other performance measures, such as the amount of experience they need to solve a problem, and the proportion of successful individuals. The broken string paradigm has been used to test several species, and a more detailed analysis of this problem can illustrate how performance can differ. When presented with two objects attached to food rewards (strings, or other objects such as strips of cloth or prepositioned tools), one intact and one with a clear break in the middle, great apes, vervet monkeys, cotton-top tamarins, elephants, and pigeons, are able to pull the connected, continuous object to bring the food within reach (Table 1; Figure 1(12)). However, whilst some species performed significantly above chance, right from the start of the experiment, in at least some configurations (great apes, vervet monkeys, and elephants), pigeons and cotton-top tamarins required extensive training (they required over a hundred trials before the correct solution was learned). It is tempting to infer that the species that solved the problem spontaneously used a qualitatively different cognitive mechanism, involving an appreciation of the principle of connectedness such as would underpin an adult human’s behavior. However, an animal’s performance depends on a number of both cognitive and noncognitive processes. For example, those animals that take longer to solve a particular task, or even fail it completely, may be less motorically dexterous, less motivated, more easily distracted, find the task at hand harder to perceive, or find irrelevant features of the task more attention-grabbing, compared to the species that solve it quickly. Even the same individuals can perform very differently on two tests supposedly probing the same ability. For example, although the great apes tested by Herrmann et al. (2008) on the broken string problem were able to solve it spontaneously when the material involved was string or cloth, they performed at chance (in the 6 trials given) when the objects were two wooden canes, prepositioned around the food rewards. Similarly, Girndt et al. (2008) tested a group of chimpanzees on the trap table test (in

Cognition j Problem-Solving in Tool-Using and Non-Tool-Using Animals

411

Figure 1 The 21 problems as numbered in Table 1. The arrow shows the transformation from one state to another or the direction in which the food should be moved for the problem to be solved.

which food should be raked in over a solid surface rather than one with a trap in it, Figure 1(18)). When tested with two prepositioned tools, they failed to find the solution in 20 trials, but they passed when given one tool and required to choose which reward to rake towards themselves. These small differences in task presentation had such a large effect on performance that had only one configuration been given, a very misleading picture of the animal’s abilities would have emerged. Given that many species comparisons rely on comparing studies done using not only different materials but also completely different research setups, it is clear that caution must be exercised. Importantly, even an identical setup may not be equivalent for different species. Ideally, results of many tests probing the same ability in different ways should be employed to build confidence in the interpretation, a process referred to as ‘triangulation.’ Although several processes (both cognitive and noncognitive) go into the make-up of a successful problem-solver, comparing performance measures across a range of problems can provide us with an idea of different species’ proclivity to solve new problems. Fine-grained analyses of performance data provide little evidence that apes outperform monkeys in problems involving space and objects, despite hypotheses to the contrary (Call, 2000). Furthermore, animals that customarily use tools in the wild do not systematically outperform nontool-users on the tasks in which both have been tested. Interestingly, on the tests conducted so far, corvids have performed comparably to primates, suggesting that there has been a convergent evolution of problem-solving skills in these

412

Cognition j Problem-Solving in Tool-Using and Non-Tool-Using Animals

two distantly related groups (Seed et al., 2009). However, to study the cognition underpinning problem-solving we need to go beyond absolute performance differences, and look at the results of tests that aim to isolate the cognitive processes from each other (e.g., perception, inhibition, learning, memory), and from the noncognitive ones (e.g., motor skills, motivation, attention, and temperament).

What Cognitive Mechanisms Underpin Animal Problem-Solving? When animals solve problems, such as the broken string task described above, do they use a knowledge of object properties (only materials that are continuous and connected to the reward are worth pulling), or have they simply associated certain perceptual cues with rewarding outcomes? The question of whether animals solve problems through sophisticated cognitive processes, such as mental representation and reasoning, is a century old debate that is still on-going. At one pole of the debate is the view expounded by Thorndike, an American psychologist working at the beginning of the twentieth century. Thorndike presented animals with puzzle boxes that they had to learn to open, either to gain their freedom or gain access to food. From these studies, he concluded that whilst the animals were able to learn through trial-and-error to associate perceptual features of the stimuli with behavioral responses, repeating those combinations that led to reward and ceasing those that did not, there was no evidence for any human-like reasoning. Kohler, a contemporary of Thorndike, objected to this conclusion on the grounds that the methodology employed did not allow the animals to engage in their powers of reasoning, because the causal mechanisms controlling the boxes’ release-mechanisms were made too obscure. In his studies with chimpanzees, he made food inaccessible, but the means to reach it clearly visible (e.g., putting food too high for the chimpanzees to reach it and leaving sticks to knock the food down, or boxes that could be stacked). From the rapid success of some of his subjects, he concluded that chimpanzees used insight, and not blind trialand-error, to solve problems. Importantly, both accounts contain a hypothesis about two facets of the cognitive process, namely what animals see in a problem (shallow perceptual features or more abstract representations of causally relevant structural features) and how animals solve problems (associative learning or more complex ‘human-like’ processes, such as reasoning and insight). These two facets will be discussed in turn.

What Do Animals See in a Problem? Objects in the environment have physical properties that dictate the possible ways in which they can interact with one another (e.g., solid objects cannot pass through one another). These properties, such as solidity, continuity, weight, and rigidity can be directly sensed, but the principles themselves can also be represented at a deeper level of abstraction (where more ‘abstract’ means that the information is not equivalent or reducible to concrete, analogue sensory input, but rather has undergone further processing in which meaning is extracted). This would enable them to be used to make predictions in novel situations that do not share any surface perceptual features with those in which they were first encountered. This is a critical feature of flexible problem-solving in humans, but in the absence of verbal report, when an animal transfers its solution to a new context, it is often impossible to know if it has inferred anything about causal properties, or if its behavior can be explained by a combination of past associative learning and generalization based on surface-level perceptual characteristics. Controlled experimentation that pits one account against another is needed. Once paradigms have been found that subjects can at least learn to solve, different features can be varied systematically in transfer tasks, to ascertain which of the possible features the subjects used to solve the original problem. For example, in the broken string task, cotton-top tamarins, having learned the original discrimination (pull the unbroken rather than the broken cloth), were given transfer tasks that varied functionally irrelevant features, such as the cloth’s color, shape, texture, and the shape and size of the gap. The tamarins readily transferred their solution across the majority of these changes, suggesting that they had used functionally relevant properties to solve the original discrimination (Hauser et al., 1999). In contrast, pigeons that had learned to solve one version of the broken string task failed to transfer to a new version in which the shape and color of the material was changed, suggesting that they had relied on perceptual cues to solve the original task (Schmidt and Cook, 2006). However, even the transfer tasks solved by the tamarins could be solved by learning to avoid the gap as a perceptual feature, and then generalizing along this parameter. Animals certainly do rely on perceptual cues to solve some problems. Povinelli (2000) conducted a series of experiments with chimpanzees, including several of the tasks depicted in Figure 1(8, 12, 14–16, 18). The results of many of these studies were commensurate with the idea that the chimpanzees had used a perceptually based rule, rather than one that encompassed an abstract notion of object properties. For example, in the trap tube task, in which the subject needs to push a piece of food out of a horizontal tube away from a trap (Figure 1(16)), the one subject that learned to do so continued to use this strategy even when the tube was inverted, and therefore nonfunctional. It seemed therefore that she had treated the trap as a perceptual cue but had not encoded its functional significance. But does the use of perceptual information in one task mean that structural information cannot be encoded by members of this species? It must be emphasized that the two sorts of knowledge are not mutually exclusive, and indeed human researchers have argued for the existence of two cognitive processes working in parallel during problem-solving. Furthermore, the many factors affecting performance mentioned at the end of the last section also impact the results of transfer tests, and so the results of one study (especially, if they are negative) need to be interpreted cautiously. Seed et al. (2009) found that eight chimpanzees solved a version of the trap problem that did not require them to use a tool. The performance of these subjects was then compared with naïve subjects on a perceptually distinct transfer test

Cognition j Problem-Solving in Tool-Using and Non-Tool-Using Animals

413

made of new materials. Chimpanzees without experience on the first problem performed poorly on this task (only one subject was successful), but all of the experienced subjects in a group tested without a tool solved the new test in very few trials, suggesting that, in contrast to the results of the original trap tube study, they had encoded information about the functional properties of the objects involved in the initial testing phase. However, another group of experienced subjects tested with a tool required many more trials, and only half of the subjects were successful, revealing the critical importance of the manner of task presentation. Seed et al. (2006) made some changes to the trap tube task in a study of rooks, and aimed to pit a perceptual and structural account directly against one another. Eight birds were tested on a version that featured two ‘traps’ along a horizontal tube (Figure 1(17)). One of the traps was sealed with a black disc at the bottom and would trap the reward if the rooks pulled the food over it. The other was nonfunctional; in Design A, it had a black disc at the top, which the food could pass across; in Design B, it had no black disc, so the food could fall through it. Once the birds had solved these tasks, they were given two transfer tasks, each featuring both previously nonfunctional traps (pass-across and fall-through). In Design C, both ends of the tube were blocked with bungs, and in Design D, the tube was lowered to the surface of the testing shelf. Crucially, therefore, both tasks featured the same familiar cue, but each required the opposite response to it (pull away from the black disc in Task C, pull towards it in Task D). One of the seven rooks was able to solve these transfers, suggesting that she did not simply use the appearance of the functional trap as an arbitrary, surface-level cue. Seed et al. (2009) recently conducted a similar experiment with chimpanzees. All of the eight chimpanzees tested learned to avoid the trap. Furthermore, one chimpanzee passed both designs C and D. Like the successful rook, this chimpanzee could not have been using a rule based on an arbitrary perceptual cue to solve the task. Hanus and Call (2008) also pitted perceptual and abstract representational accounts against one another in an experiment designed to probe what chimpanzees know about weight. Chimpanzees were shown a seesaw balance, with an empty cup at either end. The experimenter surreptitiously placed the banana in one of the cups behind a screen. The chimpanzees saw the balance tip, and strikingly, they chose the lower cup significantly more often than the higher cup from the first trial. Perhaps most interestingly, they chose at chance when the experimenter’s action tipped the balance, revealing that the shallow-level perceptual information, namely the downward movement of the lower cup, was not sufficient to elicit the chimpanzees’ choice. Even subjects tested first on the casual condition that had been choosing the lower cup reverted to random responding when the weight of the banana was no longer the cause of the movement. From these and other experiments, it seems that some animals do form abstract representations of some object properties. However, this assertion is still controversial, and a number of questions arise: which species; which specific properties; how do the abilities develop; and how (if at all) do they differ from those of humans?

How Do Animals Translate Information and Knowledge into Action? Associative learning as put forward by Thorndike to explain the performance of mammals in his puzzle boxes is based purely on covariation, and considerable trial-and-error is needed before the emergence of a correct solution. Experiments in the laboratory have shown that animals are adept at learning to associate a response to a given cue if doing so reliably leads to a certain outcome (e.g., pressing a lever or pecking a key when a light is turned on, to gain access to food). Putting aside the question of the level of abstraction at which the learned knowledge is represented (perceptual or structural), is this the only mechanism available for animals to discover the solution to a problem? In the real world, events are not random and arbitrary, but are instead underpinned by predictable causal structures (e.g., heat causes water to evaporate, and not the other way round). Adult humans are able to infer the solution to a completely novel problem without interacting with it at all, drawing on their knowledge of causality, logic, and other applicable concepts. Processes such as inference are little understood and lack rigid definitions; they are simply referred to as ‘nonassociative processes.’ What evidence is there for inference in animals? Call (2007) tested great apes on a task in which subjects needed to locate a food reward based on the inclination of two boards lain on a table. One board had a food reward underneath it, and was therefore at an incline, whilst the other was flat (Figure 1(20)). All great apes were able to locate the reward at above chance levels. They did not show a preference for the inclined board when the cause of the incline was a wedge, despite being rewarded if they did so. It therefore seems that apes were able to use their knowledge of solid objects and the effect they have on one another to infer the location of food, because a learned preference for inclined boards could not explain the results. Similarly, they were also able to infer the presence of food when two cups were shaken and only one made a rattling sound, but did not choose a tapped cup in another condition even though this choice was rewarded (Call, 2004). In the experiment by Hanus and Call (2008) described earlier, chimpanzees inferred the location of food on the basis of its weight, choosing the lower of two cups on a see-saw balance when the baiting of one of the cups was the cause, but not when the experimenter moved the balance by hand. Call (2004) also found evidence for inference by exclusion in apes. Subjects were shown two cups, and the hiding of a piece of food in one of them behind a screen. Apes were then shown the empty cup, or in another condition, the empty cup was shaken and produced no rattling sound. The apes chose the untouched, baited cup significantly above chance levels.

Conclusion Questions concerning the cognitive processes involved in animal problem-solving are still hotly debated, but they have relevance for fundamental questions concerning thinking in nonverbal animals, and the evolution of the human mind. The debate has moved

414

Cognition j Problem-Solving in Tool-Using and Non-Tool-Using Animals

on from a stark dichotomy between simple associative learning and more complex processes, and most modern researchers acknowledge the interplay between inherited predispositions, learning, and reasoning. Indeed, even Kohler, the great exponent of insight in animals, recognized that ‘. some previous learning is often needed not only for the solution of a problem but also for its understanding as a problem’ (Kohler, 1969, p. 135). Physical problem-solving provides researchers with an excellent window onto these questions, because unlike the social domain, objects and events in the physical world can be tightly controlled by experimenters. However, because so many variables impact on performance on these tasks, experiments must be carefully designed and interpreted with caution, especially when comparing species. Evidence so far suggests that for large-brained, innovative species such as primates and corvids, associative learning of perceptual rules is not the only process available for solving problems. However, further work is needed to characterize these nonassociative processes in more detail, and the precise role they play in animal problem-solving. We have taken some steps towards an understanding of animals’ knowledge of object properties, and their inferential abilities, but we know little about how these two facets interact. A still greater challenge is to explore the distribution of these cognitive skills amongst animals, to address the question of the evolutionary pressures selecting for problem-solving skills.

See also: Cognition: Animal Arithmetic; Categories and Concepts: Language-Related Competences in Non-Linguistic Species. Evolution: Development, Evolution and Behavior. Learning and Teaching: Apes: Social Learning; Costs of Learning.

Further Reading Amici, F., Aureli, F., Call, J., 2008. Fission-fusion dynamics, behavioral flexibility, and inhibitory control in primates. Current Biology 18 (18), 1415–1419. Beck, B., 1980. Animal Tool Behavior: The Use and Manufacture of Tools by Animals. Garland, New York. Bitterman, M.E., 1965. Phyletic differences in learning. American Psychologist 20, 396–410. Blaisdell, A.P., 2008. Cognitive dimension of operant learning. In: Roediger III, H.L. (Ed.), Cognitive Psychology of Memory. Vol. 1 of Learning and Memory: A Comprehensive Reference, 4 vols. Elsevier, Oxford, pp. 173–195 (J. Byrne Editor). Call, J., 2000. Representing space and objects in monkeys and apes. Cognitive Science 24 (3), 397–422. Call, J., 2004. Inferences about the location of food in the great apes (Pan paniscus, Pan troglodytes, Gorilla gorilla, and Pongo pygmaeus). Journal of Comparative Psychology 118 (2), 232–241. Call, J., 2007. Apes know that hidden objects can affect the orientation of other objects. Cognition 105 (1), 1–25. Call, J., Tomasello, M., 2005. Reasoning and thinking in non-human primates. In: Holyoak, K.J., Morrison, R.G. (Eds.), The Cambridge Handbook of Thinking and Reasoning. Cambridge University Press, New York, NY, pp. 607–632. Fujita, K., Kuroshima, H., Asai, S., 2003. How do tufted capuchin monkeys (Cebus apella) understand causality involved in tool use? Journal of Experimental Psychology Animal Behavior Processes 29, 233–242. Gibson, K.R., 1986. Cognition, brain size and the extraction of embedded food resources. In: Else, J.G., Lee, P.C. (Eds.), Primate Ontogeny, Cognition and Social Behaviour. Cambridge University Press, Cambridge, pp. 205–218. Girndt, A., Meier, T., Call, J., 2008. Task constraints mask great apes’ ability to solve the trap-table task. Journal of Experimental Psychology Animal: Behavior Processes 34 (1), 54–62. Hanus, D., Call, J., 2008. Chimpanzees infer the location of a reward on the basis of the effect of its weight. Current Biology 18 (9), R370–R372. Hauser, M.D., 1999. perseveration, inhibition and the prefrontal cortex: A new look. Current Opinion in Neurobiology 9, 214–222. Hauser, M.D., 2003. To innovate or not to innovate? That is the question. In: Reader, S.M., Laland, K.N. (Eds.), Animal Innovation. Oxford University Press, Oxford, pp. 329–338. Hauser, M.D., Kralik, J., Botto-Mahan, C., 1999. Problem solving and functional design features: Experiments on cotton-top Tamarins, Saguinus oedipus oedipus. Animal Behaviour 57 (3), 565–582. Herrmann, E., Wobber, V., Call, J., 2008. Great apes’ (Pan troglodytes, Pan paniscus, Gorilla gorilla, Pongo pygmaeus) understanding of tool functional properties after limited experience. Journal of Comparative Psychology 122 (2), 220–230. Heyes, C.M., Huber, L., 2000. The Evolution of Cognition. MIT press, Cambridge, MA. Hurley, S., Nudds, M., 2006. Rational Animals. Oxford University press, Oxford. Kohler, W., 1969. The Task of Gestalt Psychology. Princeton University Press, Princeton, NJ. Mackintosh, N.J., Wilson, B., Boakes, R.A., 1985. Differences in mechanisms of intelligence among vertebrates. Philosophical Transactions of the Royal Society of London B Biological Sciences 308, 53–65. Moura, A.C.d.A., Lee, P.C., 2004. Capuchin stone tool use in caatinga dry forest. Science 306, 1909. Penn, D.C., Holyoak, K.J., Povinelli, D.J., 2008. Darwin’s mistake: Explaining the discontinuity between human and non-human minds. Behavioral and Brain Sciences 31, 109–130. Povinelli, D.J., 2000. Folk Physics for Apes: The Chimpanzee’s Theory of How the World Works. Oxford University Press, Oxford. Reader, S.M., Laland, K.N., 2002. Social intelligence, innovation and enhanced brain size in primates. Proceedings of the National Academy of Sciences USA 99, 4436–4441. Roberts, W.A., 1998. Principles of Animal Cognition. McGraw-Hill, Boston. Schmidt, G.F., Cook, R.G., 2006. Mind the gap: Means-end discrimination by pigeons. Animal Behaviour 71 (3), 599–608. Seed, A.M., Call, J., 2009. Causal knowledge for events and objects in animals. In: Watanabe, S., Blaisdell, A.P., Huber, L., Young, A. (Eds.), Rational Animals, Irrational Humans. Keio University press, Tokyo, pp. 173–187. Seed, A.M., Call, J., Emery, N.J., Clayton, N.S., 2009. Chimpanzees solve the trap problem when the confound of tool use is removed. Journal of Experimental Psychology: Animal Behavior Processes 35 (1), 23–34. Seed, A., Emery, N., Clayton, N., 2009. Intelligence in corvids and apes: A case of convergent evolution. Ethology 115, 401–420. Seed, A.M., Tebbich, S., Emery, N.J., Clayton, N.S., 2006. Investigating physical cognition in rooks, Corvus frugilegus. Current Biology 16 (7), 697–701. Shettleworth, S.J., 1998. Cognition, evolution, and the study of behavior. In: Shettleworth, S.J. (Ed.), Cognition, Evolution, and Behavior. Oxford University Press, New York, NY, pp. 3–48. Sol, D., Timmermans, S., Lefebvre, L., 2002. Behavioural flexibility and invasion success in birds. Animal Behavior 63, 495–502.

Cognition j Problem-Solving in Tool-Using and Non-Tool-Using Animals

415

Tebbich, S., Taborsky, M., Fessel, B., Dvorak, M., 2002. The ecology of tool-use in the woodpecker finch (Cactospiza pallida). Ecology Letters 5, 656–664. Tomasello, M., Call, J., 1997. Primate Cognition. Oxford University Press, New York. Visalberghi, E., Fragaszy, D.M., 2006. What is challenging about tool use? The capuchin’s perspective. In: Wasserman, E.A., Zentall, T.R. (Eds.), Comparative Cognition: Experimental Explorations of Animal Intelligence. Oxford University press, New York, NY, pp. 529–552. Visalberghi, E., Limongelli, L., 1994. Lack of comprehension of cause- effect relations in tool-using capuchin monkeys (Cebus apella). Journal of Comparative Psychology 108, 15–22.

Rational Choice Behavior: Definitions and Evidence M Bateson, Newcastle University, Newcastle upon Tyne, UK © 2010 Elsevier Ltd. All rights reserved.

Abstract Rationality is a property of individual choice behavior, and is therefore relevant to the analysis of a range of animal decisions, including foraging, nest selection, and mate choice. Students of behavior use the term ‘rational’ in two senses: first, to describe the process of making a choice, in which case it implies use of logical reasoning in decision-making; and second, to describe the outcome of choice, in which case, it implies internally consistent choices resulting from maximization of a currency. The animal cognition literature has extensively explored rationality in the choice process. However, it is often difficult to draw definite conclusions about the nature of cognitive processes. Consideration of rationality in choice outcomes originated in economics but also has relevance to animal behavior. Research in behavioral ecology starts with the assumption that animals maximize inclusive fitness and are therefore ultimately biologically rational. The optimality approach tests specific hypotheses about the proximate currencies that animals maximize when they make decisions. Several recent studies have documented examples of economically irrational choice behavior in animals, such as violations of transitivity and regularity. However, this behavior can be interpreted as biologically rational, when it is considered in its broader ecological context.

Keywords Choice; Cognition; Decision-making; Economics; Foraging; Irrationality; Optimal foraging theory; Rationality; Regularity; Transitivity; Utility

Introduction: The Problem of Choice As humans, we are constantly faced with choices between alternative options. We have to decide which products to buy in the supermarket, what sort of house we want to live in, and even who we would like to meet again following a speed-dating event. These choices are made difficult not only by the sheer number of different options we are faced with, but also by the fact that the options differ in multiple attributes that may affect our decisions. Buying something as seemingly simple as a box of eggs recently, I realized that I had to make a choice based on price, box size, egg size, egg color, freshness, whether the hens were free-range or kept in battery cages, and whether they were fed a conventional or organic diet! Although this kind of problem might at first seem unique to humans living in the modern world, many non-human animals are also faced with complex choices about what to eat, where to live, and who to mate with. For example, a foraging rufous hummingbird (Selasphorus rufous) must choose between flowers of different species differing in corolla length and the sweetness and volume of nectar contained; a colony of rock ants (Temnothorax albipennis) moving house must choose between potential nests differing in the size of the entrance hole and the darkness of the interior; and a female mouse (Mus musculus) looking for a mate has to choose between males differing in genetic relatedness and genetic quality. Studying the choices made by animals in such situations is a major area of research in animal behavior. Researchers want to understand both the proximate mechanisms of choice and the ultimate evolutionary explanations for the choices animals make. The study of proximate mechanisms is predominantly the domain of ethologists and comparative psychologists, whereas the study of the adaptive significance of the choices animals make is the domain of behavioral ecologists. Rationality is a property of choice behavior that, as we will see shortly, has been used to describe both the mechanisms of choice and the outcome of the choice process. The ancient Greek philosopher, Aristotle, saw rationality as a property unique to human decision making, setting us apart from other animals. However, modern-day biologists and psychologists have extended the concept of rationality to animal choice, and are actively pursuing research into whether animals can be considered rational. In this article, I review the study of rationality in animals and examine what the evidence says about whether animals are indeed rational. However, before we address these questions, we first need to understand exactly what it means to describe a choice as rational.

What Is Rationality? It is difficult to provide a concise definition of rationality, because it has been used to refer to different properties of choice in different academic disciplines. Biologists, economists, philosophers, and psychologists all use the term ‘rationality’ to describe choice behavior, but they define rationality in many different ways. Any student new to the area will be horrified at the bewildering typologies produced by researchers in different fields, and this artilce cannot review the many subtly different definitions of rationality. Instead, I suggest that we can map many, if not most, of the existing definitions of rationality onto two broad categories: first, descriptions of the process of choice, and second, descriptions of the outcome of choice, that is, which option is actually chosen. Thus, in analyzing our hummingbird’s choice of flowers, we can focus on either the mechanisms it uses to choose one flower from

416

Encyclopedia of Animal Behavior, 2nd edition, Volume 1

https://doi.org/10.1016/B978-0-12-813251-7.00222-4

Cognition j Rational Choice Behavior: Definitions and Evidence

417

a set of three, or alternatively on which flower it actually chooses. We can ask whether both the process of choosing and the outcome of the bird’s choice can be described as rational. Thus, the two uses of rationality map neatly onto two of Tinbergen’s four questions: those of proximate mechanism and ultimate function.

Rationality of Choice Processes If you look up the adjective ‘rational’ in a standard English dictionary, you will find definitions such as: ‘Using reason or logic in thinking out a problem’ or ‘Endowed with the capacity to reason.’ Both these definitions refer to reasoning, which in turn is defined as the ability to think, or to draw conclusions from known facts. These layman’s definitions of rationality correspond quite closely with the way philosophers and cognitive psychologists use the term. Interestingly, they are also similar to the definition adopted by Charles Darwin who wrote in his notes, “Rational actions . are actions which are required to meet circumstances of comparatively rare occurrence in the life-history of the species, and which therefore can only be performed by an intentional effort of adaptation . rational actions . serve to meet novel exigencies which may never before have occurred even in the life-history of the individual.” Darwin went on to argue that rational action, “Implies the conscious knowledge of the relation between means employed and ends attained” (Darwin cited in Romanes (1882)). In defining rational actions, Darwin contrasts them with what he refers to as ‘reflexes’ and ‘instinctive actions.’ Thus, in describing the process of choice as rational, Darwin and others are implying the use of cognitive mechanisms that we might describe as ‘clever’ or ‘intelligent.’ By this, we mean mechanisms that represent information about the state of the world and the goals of the animal, and use this information in a flexible way to solve novel problems effectively. Darwin’s definition also implies that conscious intention has to be present for rationality. However, most modern biologists are not happy with the notion of ascribing conscious intentions a causal role in the generation of behavior. Consciousness is a private experience, and consequently we can never objectively observe or measure it in animals. Therefore, most modern research in animal cognition distinguishes between the study of information processing in animals and the study of consciousness. We can ask how animals acquire, represent, and use information in the generation of behavior without asking whether or not this happens via some conscious process. Thus, for the purposes of this study, we will define a choice process as rational if the resulting behavior displays evidence of flexible, goal-directed information processing based on representations of the state of the world.

Rationality of Choice Outcomes The second use of the term ‘rationality’ focuses on the alternatives an animal actually chooses, as opposed to the processes responsible for choice. This use of rationality therefore refers to directly observable behavior rather than unobservable cognitive processes. An individual’s behavior is defined as rational if it is compatible with the individual maximizing a currency of some type, resulting in internally consistent decisions. This definition of rationality has its roots in microeconomic theory and has only relatively recently been explicitly considered in the context of animal behavior. I will therefore start by describing what rationality means in economics before exploring how we can apply the concept in biology.

Economic rationality The theory of individual decision making developed in microeconomics starts by considering the problem of choosing from among a set of mutually exclusive alternatives (similar to the egg-choice problem with which I opened this study). Economic models of choice assume that when making such choices human consumers maximize a quantity called ‘utility.’ One can think of utility as a measure of the relative satisfaction an individual derives from a specific resource. However, it is important to realize that utility cannot be measured independent of what people actually choose. Rational choice is simply defined as choice behavior that is compatible with the maximization of utility. If an individual maximizes utility, or indeed any other currency, their choice behavior will be internally consistent in various ways that are considered to be hallmarks of rational choice. These hallmarks include the properties of transitivity, independence from irrelevant Alternatives, and regularity. I will briefly describe each of these properties in the following paragraphs. Transitivity is a property that applies specifically to binary choices. Preferences are transitive between the three options A, B, and C if A is preferred to B, B is preferred to C, and A is preferred to C. For example, if binary choices reveal that I prefer a cherry to a pear, and a pear to an apple, then if my choices are transitive I should prefer a cherry to an apple (see Figure 1(a) for an example). If I showed the opposite preference and preferred the apple to the cherry, this would constitute a violation of transitivity (see Figure 1(b) and 1(c)). Independence from irrelevant alternatives is a property that applies when a choice set is expanded. It implies that the preference between two options should be independent of the presence of additional inferior alternatives. For example, if A is preferred to B in the binary choice of A versus B, then the introduction of option C should not affect the preference for A over B (see Figure 2(a) for an example). If the relative preference for A over B is altered by the addition of C, this is referred to as a ‘violation of the constant ratio rule’ (Figure 2(b)). If the absolute preference for either A or B increases when C is added to the choice set, this is referred to as a ‘violation of regularity’ (Figure 2(c)). Thus, we can summarize the economists’ definition of rationality as follows. In economics, rationality describes the internal consistency in an individual’s choices that results if they are maximizing a currency known as ‘utility.’ Economists consider

418

Cognition j Rational Choice Behavior: Definitions and Evidence

Figure 1 (a) An example of transitive choice: A is preferred to B, B to C, and A to C; (b) an example of a violation of strong stochastic transitivity: the preference for A over C is less than the preference for A over B or B over C; (c) an example of a violation of weak stochastic transitivity: C is preferred to A.

transitivity and regularity to be fundamental features of rational choice. Given that one cannot measure utility directly, assessing transitivity and regularity will often be the only way to test whether human consumers maximize utility.

Biological rationality and optimal foraging theory Given the strong superficial similarity between the kinds of choices faced by humans and animals, it is perhaps not surprising that research on animal choice has drawn heavily on the theories developed to model the behavior of human consumers in microeconomics. However, a major difference between biological and economic models of choice is that they assume different currencies of maximization. In animal behavior, we start with the basic assumption that an animal’s behavioral repertoire is ultimately the product of evolution by natural selection. Natural selection favors genetic variants with the highest inclusive fitness; thus, the behavior of an animal observed in the context in which it has evolved should ultimately maximize its inclusive fitness. By analogy with the economists’ definition earlier, we can think of an individual that behaves in a way that maximizes its inclusive fitness as biologically rational. Since we assume that all animals should be ultimately biologically rational, this is in some sense a trivial definition. Ultimately, however this does not matter, because behavioral ecologists take biological rationality as their starting point for more detailed analyses of behavior; the basic assumption of ultimate biological rationality is not under test. Unlike utility, biologists can, in principle, measure inclusive fitness. However, inclusive fitness is unlikely to be an appropriate currency for computing the costs and benefits of alternative decisions in many circumstances. For example, assume that we want to understand the moment-to-moment flower choices of a foraging hummingbird. The inclusive fitness consequences of the bird choosing a specific flower type are hard for us to measure, because in order to estimate these it would be necessary to record the lifetime reproductive success of birds that fed on this flower type compared with birds that fed on another flower type. Similarly, the hummingbird cannot use its inclusive fitness as the currency it is maximizing when it is making foraging decisions, because the consequences of its choices are not immediately translated into detectable changes in fitness. We therefore assume that foraging animals must use currencies for decision making that are accessible to them over the time scale of a single foraging bout. The

Cognition j Rational Choice Behavior: Definitions and Evidence

419

Figure 2 (a) An example of independence from irrelevant alternatives. When the inferior option, C, is added to the binary choice of A and B, the relative preference for A over B remains unchanged; (b) an example of a violation of independence from irrelevant alternatives: the addition of C leaves the proportion of choices for A unchanged but reduces the proportion of choices for B, leading to a violation of the constant ratio rule; (c) two examples of violations of regularity: the addition of option C either increases the proportion of choices for B (first panel) or A (second panel).

assumption is that natural selection will have favored animals that use currencies for decision making that correlate well with inclusive fitness over the life of the animal. One of the most common currencies assumed in optimal foraging models is net rate of energy intake. Maximization of this currency is likely to lead to maximization of fitness because both time and energy have clear relationships with fitness: the more energy taken in the more can be used for growth and reproduction, and the less time spent foraging the more is available for other fitness enhancing activities. However, different currencies will be appropriate for different behavioral decisions. For example, in the case of a small bird in winter choosing between safe and risky foraging options, the best currency might be the probability of surviving the night, whereas for a worker bee choosing how much nectar to carry, it might be the ratio of energy gained to energy spent (known as ‘efficiency’). The specific currency that best predicts inclusive fitness in a given instance will depend on a number of factors including the biology of the species concerned and the exact behavioral decision being modeled. Research in optimal foraging commonly asks which proximate currency best predicts animal decision making. A classical study by Alex Kacelnik investigating foraging decisions in breeding starlings illustrates this approach. Given the parallels between economic models and optimal foraging models, it is interesting that the economists’ definition of rationality was until recently not explicitly mentioned in the foraging literature; indeed, the word does not appear in the index of Stephens and Krebs’ (1986) classic text on foraging theory. There are a number of explanations for this omission. The first is that economists and behavioral ecologists ask different questions: economists want to know whether or not we are rational, whereas behavioral ecologists assume that animals are ultimately rational and want to know which of various alternative proximate currencies they are maximizing. The second explanation is that economists and behavioral ecologists analyze different types of behavioral decisions. The hallmarks of rational decision-making analyzed by economists apply to a one-off simultaneous choices between mutually exclusive alternatives. Whereas the classic prey choice and patch leaving problems analyzed in foraging theory consider situations involving sequential, nonmutually exclusive choices, for which measurements of transitivity and regularity are hard to apply. Finally, behavioral ecologists can measure directly the currencies that they make hypotheses about, so they do not have to rely on indirect measures of rationality (such as transitivity and regularity) that characterize the economic approach to choice. Although animals seldom face simultaneous choices between mutually exclusive outcomes, arguably there are some situations in which they face such choices. For example, a peahen might assess the qualities of the peacocks displaying on the lek before choosing one of them to mate with, and a hummingbird might weigh up the benefits of two clumps of flowers of different species before committing to one of them for its next bout of foraging. In these cases, we can ask whether the animals’ choices display the economists’ hallmarks of rational decision making. If the proximate mechanisms underlying animal decision making involve the maximization of absolute currencies, such as for example rate of energy intake, then animal choices should be rational in the economists’ sense and display the properties of transitivity and regularity.

420

Cognition j Rational Choice Behavior: Definitions and Evidence

Thus, we can summarize the biologists’ approach to rationality as follows. Biologists assume that behavior is ultimately rational in that it maximizes inclusive fitness. Behavioral ecologists assume that animal decision-making mechanisms maximize proximate currencies which in turn maximize inclusive fitness. If animals make decisions by maximizing proximate currencies, then their choices should also be rational in the economic sense.

Are Animals Rational? Having established what biologists mean by rational behavior, in this section we will proceed to identify what kinds of evidence we need to determine whether or not animals are rational. Given the different definitions of process and outcome rationality explained earlier, we will consider each separately.

Testing Process Rationality in Animals The definition of process rationality implies mechanisms of choice that we cannot observe directly. Therefore, tests of process rationality must use observable behavior to draw inferences about unobservable mental states and processes. The challenge in testing whether animals can be described as rational is to find behavioral evidence for flexible, goal-directed information processing based on mental representations of the state of the world. This is extremely difficult, because simple rules and associative learning can often explain behavior that superficially appears to be rational. Investigators have used a number of different species and behavioral tasks to address the question of animal rationality including: tool use and tool construction in New Caledonian crows (Corvus moneduloides), scatter-hoarding behavior in western scrub jays (Aphelocoma californica), metacognition in Rhesus monkeys (Macaca mulatta) and pigeons (Columba livia), gaze following and mind reading in chimps (Pan troglodytes), to name a few systems where recent progress has been made. In this study, I use transitive inference (not to be confused with transitivity of choice) to illustrate the problems reaching unequivocal conclusions about process rationality in animals.

Transitive inference Many animals need to rank things in a stable series. The ranked entities could be group mates (who is dominant to whom) or food items. For example, within a primate troupe, there may be a stable dominance hierarchy of individuals, or within the territory of a hummingbird, some flower species always have more nectar than others. When an animal has knowledge of such a series, ‘transitive inference’ refers to the ability to deduce the relationship between two items in the series that the animal has not previously compared directly. For example, if a baboon knows that individual A is dominant to individual B, and that individual B is dominant to individual C, then she is capable of transitive inference if she can deduce that A should be dominant to C. Performing transitive inference apparently involves reasoning using prior knowledge about the relationships between entities in the world to form a conclusion about a novel situation. It therefore captures many of the attributes of process rationality identified earlier. As a consequence, attempts to demonstrate transitive inference in animals have had a central place in comparative psychologists’ attempts to test animal rationality. Investigators have tested transitive inference in a wide range of species including squirrel monkeys, rhesus monkeys, chimpanzees, pigeons, pinyon, and scrub jays, and even cichlid fish (Astatotilapia burtoni). Most tests begin by training experimental subjects on what is known as an n-term series task. A three-term task would involve three distinct stimuli, A, B, and C (Figure 3(a)). The procedure presents successive adjacent pairs of stimuli from the series (i.e., AB and BC for a three-term series, ABC) to the subject. For each pair, choosing one stimulus produces reinforcement (þ) while the other stimulus is unreinforced (). Thus, for the threeterm series, the two trained pairs are A þ B and B þ C. The assumption is that this training will create the linear series A > B > C in the animal’s mind. The critical test trial presents a nontrained, nonadjacent pair (in the case of the three-term task, this is AC). If the subject is capable of transitive inference, it should choose stimulus A, on the grounds that it can infer from inspecting its mental representation of the series that A > C. When trained on such a type of task, most animals do indeed prefer A. However, this preference could equally be explained by a very simple associative mechanism, because during training A has always been rewarded and C never. Hence, the animal could simply be picking the stimulus previously associated with reinforcement, as opposed to reasoning based on inspecting a mental representation of the linear order of the stimuli. For this reason, the standard procedure in tests of transitive inference is to train the subject on a five-term series (see Figure 3(b)). This allows a test trial with the novel pair BD. This test has the advantage over the three-term task that both B and D have been rewarded 50% of the time during training, removing the asymmetry present in the three-term task. When trained on this version of the task most animals prefer B, as predicted if they are capable of transitive inference. Successful performance on this task has been interpreted as evidence for transitive inference in animals. However, this is not the end of the story. In the course of training, B may have acquired a higher value than D because B is sometimes paired with A, which is always a winner, whereas D is sometimes paired with E, which is always a loser. Thus, if value transfers to B from A and to D from E by virtue of their sometimes being presented together, this could explain why B is preferred to D. It seems that however well-designed the test, it is always possible to come up with an associative account for the animals’ behavior that does not require reasoning based on a representation of the series. In studies of comparative cognition, it is usual to apply Lloyd

Cognition j Rational Choice Behavior: Definitions and Evidence

Figure 3

421

Reinforcement patterns used in n-term transitive inference tasks: (a) shows a three-term task, and (b) a five-term task.

Morgan’s Canon which states, “In no case is an animal activity to be interpreted in terms of higher psychological processes, if it can be fairly interpreted in terms of processes which stand lower in the scale of psychological evolution and development” (Morgan, 1903, p. 59). Following this rule, there is little evidence for process rationality in animals that we could not explain by some simpler mechanism.

Evidence for Outcome Rationality in Animals Outcome rationality is simpler to test than process rationality because tests rely on direct observations of what animals choose. As explained earlier, the assumption of biological rationality underpins the whole of behavioral ecology and is generally not directly tested. Instead, behavioral ecologists have focused on testing specific hypotheses about the proximate currencies animals maximize, and this approach has been extremely successful in showing how animal behavior is evolutionarily rational. However, a small number of more recent studies have set out to test whether animals are rational in the economists’ sense. Tests of economic rationality in animals have been inspired by examples of human irrationality. Experiments on human decision making have shown that we tend to make irrational choices when alternative options differ in more than one attribute (as in the egg example with which I started this article). When faced with decisions of this type, a rational decision maker should combine all the attributes into a single currency and choose the alternative that yields the highest value. For example, a hummingbird might choose from a set of flowers that differ in nectar volume, nectar concentration, and handling time. Under an optimal foraging account, we can summarize all these attributes in the single currency of net rate of energy intake. Using this currency, the hummingbird could compare the flowers and make a choice that maximizes net rate of energy intake. However, when humans face complex, multidimensional decisions, they often show violations of transitivity and tend to be influenced by the presence of irrelevant alternatives. For example, an experiment found that purchases of large cans of a high-quality, high-price brand of baked beans increased, and purchases of large cans of a low-quality low price brand decreased, when smaller, relatively more expensive cans of the same high-quality brand are added to the choice set. The small-but-expensive option is an irrelevant alternative, being more expensive and of no better quality than one of the other options, making this result a clear violation of regularity. One explanation for this irrationality is that rather than combining the attributes into a single currency, we instead resort to simple heuristics for decision making. For example, we might simply choose the option that ranks highest on the greatest number of attributes, ignoring the absolute values of the various attributes. Such heuristics have the benefit of being fast and easy to compute, but they sometimes result in economically irrational choices. Therefore, experiments designed to look for economic

422

Cognition j Rational Choice Behavior: Definitions and Evidence

rationality in animals have specifically focused on situations in which animals face choices between options that simultaneously differ in multiple attributes of interest.

Tests for transitivity of choice Transitivity (not to be confused with transitive inference) is a property of a series of binary choices made between pairs of simultaneously presented mutually exclusive alternatives. Thus, tests of transitivity typically present animals with pairs of choices and study which option the animal prefers. In the first experiment explicitly designed to test economic rationality in animals, Sharoni Shafir presented foraging honeybees (Apis melifera) with a series of binary choices between pairs of artificial flowers varying in two attributes both known to affect bees preference: the corolla length and the nectar volume. He found some individual bees that preferred flower A to B, B to C, C to D, but also D to A. Preferring D to A violates what is known as ‘weak stochastic transitivity.’ Bees that violated weak stochastic transitivity also violated strong stochastic transitivity, meaning that the strength of preference between two flowers adjacent on the scale of utility (e.g., A and B) was larger than that between two more widely separated flowers (e.g., A and C). Similar results have also been found in foraging gray jays (Perisoreus Canadensis).

Tests for independence from irrelevant alternatives and regularity Independence from irrelevant alternatives and regularity are properties of choice that emerge when we increase the number of alternatives in the choice set. Tests of regularity ask how adding a third alternative to the choice set (a ternary choice) affects preference between two options (a binary choice). My colleagues and I tested the preferences of foraging rufous hummingbirds presented with artificial flowers that offered different volumes and concentrations of nectar. In the binary treatment, birds chose between a high concentration flower (20 ml of 40% sucrose) and a high volume flower (40 ml of 20% sucrose), whereas in two ternary treatments we added a third flower type that was worse than either the high concentration or the high volume flower (10 ml of 30% sucrose and 30 ml of 10% sucrose, respectively) (see Figure 4(a)). The additional flowers should be irrelevant to the birds’ preference between the high concentration (C) and high volume flowers (V) because they are clearly worse than one of these flowers on both dimensions. However, we found that the third flower type affected both the relative and the absolute preferences for the two options

Figure 4 (a) The four flower types used by Bateson et al. (2003) in their study of hummingbird foraging decisions; (b) results from the experiment: relative preference for V increases in the treatment with DV, whereas relative and absolute preference for C increases in the treatment with DC. Redrawn from Bateson M, Healy SD, and Hurly TA (2003) Context-dependent foraging decisions in rufous hummingbirds. Proceedings of the Royal Society B 270: 1271–1276.

Cognition j Rational Choice Behavior: Definitions and Evidence

423

compared in the binary treatment (Figure 4(b)). Thus, the birds’ preferences violated both independence from irrelevant alternatives and regularity. Interestingly, our results support the idea that the birds use a simple heuristic that ranks concentration and volume dimensions independently, because in both ternary choices, preferences shifted toward the flower with the highest relative ranks on both dimensions. Studies using foraging gray jays, honeybees, and starlings, as well as in female green swordtails (Xiphophorus helleri) and fiddler crabs (Uca mjoebergi) choosing their mates have reported similar results.

What Does It Mean if Behavior Is Irrational? The experiments described earlier show that animals are sometimes irrational in the economists’ sense. These results imply that animals do not necessarily assign absolute values to alternatives options, but instead the values assigned can depend on the specific set of alternatives available at the time of choice. However, it is important to understand that this in no way threatens our view as behavioral ecologists that animals are ultimately biologically rational. Context dependency could occur for a number of different reasons. Earlier, I suggested that simple heuristics – such as preferring the option with the highest rank on all dimensions – could explain human and non-human irrationality. Although these heuristics might sometimes lead an animal to prefer a poor alternative (e.g., one that yields a lower rate of energy intake), we assume that natural selection has favored these heuristics because on balance they benefit the animal. Benefits could occur either via increased speed of decision making or a reduced requirement for computational resources in the brain. Other explanations for context dependency have also been suggested. In most studies, for example, the animal makes a sequence of choices, so the options chosen early in the sequence could change the animals state (e.g., reduce its hunger) and thus change the nature of optimal decisions later in the sequence. Thus, it might be possible to accommodate some apparently irrational behavior within a conventional optimal foraging framework. In summary, although animal behavior can sometimes appear economically irrational, when we consider it in its full ecological context, the biological rationality should become apparent. The value of studying economic irrationality in animal decision making lies in what these studies can tell us about the proximate mechanisms underlying animal choices.

See also: Evolution: The Importance of Animal Behavior for Charles Darwin and Other 19th Century Theorists. Landmark Studies: Niko Tinbergen. Social Behavior: Kin Selection and Relatedness.

Further Reading Bateson, M., 2004. Mechanisms of decision-making and the interpretation of choice tests. Animal Welfare 13, S115–S120. Bateson, M., Healy, S.D., Hurly, T.A., 2003. Context-dependent foraging decisions in rufous hummingbirds. Proceedings of the Royal Society B 270, 1271–1276. Hurley, S., Nudds, M., 2006. Rational Animals. Oxford University Press, Oxford. Kacelnik, A., 1984. Central place foraging in starlings (Sturnus vulgaris). I. Patch residence time. Journal of Animal Ecology 53, 283–299. McGonigle, B.O., Chalmers, M., 1992. Monkeys are rational! The Quarterly Journal of Experimental Psychology 45B, 189–228. Morgan, C.L., 1903. An Introduction to Comparative Psychology, 2nd edn. W. Scott, London. Romanes, G.J., 1882. Animal Intelligence. Kegan Paul Trench & Co, London. Schuck-Paim, C., Pompilio, L., Kacelnik, A., 2004. State-dependent decisions cause apparent violations of rationality in animal choice. Public Library of Science Biology 2, e402. Shafir, S., 1994. Intrasitivity of preferences in honey bees: Support for ‘comparative’ evaluation of foraging options. Animal Behaviour 48, 55–67. Stephens, D.W., 2008. Decision ecology: Foraging and the ecology of animal decision making. Cognitive, Affective, & Behavioral Neuroscience 8, 475–484. Stephens, D.W., Krebs, J.R., 1986. Foraging Theory. Princeton University Press, Princeton. Vasconcelos, M., 2008. Transitive inference in non-human animals: An empirical and theoretical analysis. Behavioural Processes 78, 313–334. Wynne, C.D.L., 2004. Do Animals Think? Princeton University Press, Princeton, NJ/Oxford.

Responses to Death James R Anderson, Kyoto University Graduate School of Letters, Kyoto, Japan © 2019 Elsevier Ltd. All rights reserved.

Abstract How animals respond to dead and dying companions is discussed with reference to four components of mature humans’ concept of death: irreversibility, non-functionality, inevitability, and causality. Knowledge that death is irreversible seems especially likely to emerge in large-brained species with relatively small home ranges and close intra-group social relations. In primates and cetaceans, mothers may continue to transport and care for dead infants before eventually abandoning them. These experiences, along with exposure to corpses of other conspecifics and other species probably contribute to learning that death is irreversible. In many species corpses may be explored and manipulated in various ways, which provides opportunities for learning that dead individuals never react to stimulation and that they remain inert (non-functionality). There is less evidence for understanding that all individuals will die (inevitability), but again, cognitively advanced species might infer that individuals of various species can die or be killed. Although animals generally strive to stay alive, it is unclear whether they fear death, rather than injury. Finally, some species appear likely to possess some core knowledge about the types of events (e.g., accidents, attacks) that can result in death. Social insects have evolved several behavioural mechanisms to manage corpses in ways to avoid danger; these mechanisms appear genetically pre-programmed. Among avian species, some corvids are particularly responsive to dead conspecifics and use the situation to assess environmental conditions. More research combining observational, experimental and non-invasive physiological monitoring promise to provide further insights into the comparative phenomenology of death, with potential implications for slaughter and euthanasia practices.

Keywords Cetaceans; Corvids; Death; Death concept; Dying; Grief; Primates; Social insects; Thanatology

Introduction The study of how both humans and nonhuman species (hereafter: “animals”) respond to dying and dead conspecifics falls within the broad discipline of comparative thanatology (Anderson, 2016). A wide range of behavioural responses to dead individuals has been described in animals, from exploration, affiliation, caretaking and grief reactions, to avoidance, sexual interest, abusive treatment, and even cannibalism. Some species simply abandon whereas others actively dispose of their dead; as described below, reactions vary across taxa and specific circumstances. Comparative (and evolutionary) thanatology is of interest to many scientists, philosophers and laypersons, with a frequently raised question being the extent to which other species’ understanding of death resembles that of humans (Anderson et al., 2018). However, given the extreme variability in different species’ lifestyles as well as emotional and cognitive abilities, thought must be given to the best way to frame this question. Research on humans can provide useful starting points for comparative thanatologists. For example, many psychologists view typical human adults’ concept of death as comprising at least four sub-concepts or cognitive “components,” namely irreversibility, non-functionality, inevitability, and causality. These cognitive components undergo changes during children’s development before converging with those of adults in late childhood (see Slaughter, 2005; Speece, 1995). Below, these four conceptual components are used to consider some aspects of animals’ behaviour towards death.

Irreversibility An individual who understands that death is irreversible knows that dead organisms will not recover; they will show no further signs of animacy. Unlike adults, young children often believe that dead individuals can come back to life; i.e., being dead is similar to being asleep. However, young children with personal experience of bereavement (including death of a family pet) may show a more mature of concept of death than those without such experience (e.g., Bonoti et al., 2013). Do any other animals understand that death cannot be undone? Species-typical lifestyle and learning abilities may be crucial in this context. In species that form large, anonymous aggregations and that move around within vast home ranges (e.g., many fish, birds, and herbivorous mammals), the death of a conspecific might go largely unnoticed. If noticed, it may elicit little or no observable response, the survivors being focused on moving, foraging, and avoiding predation. In these circumstances individuals might never get to see a dead conspecific for any length of time, which limits the opportunity to learn that once dead, always dead. Species that live in more restricted home ranges and with closer intra-group social relations based on individual recognition appear better placed to achieve an understanding of the irreversibility of death, especially those with more advanced cognitive abilities (e.g., large-brained mammals and possibly some birds). Mothers of dead infants in some mammalian species are reported to remain near, care for, and attempt to stimulate their dead offspring for various lengths of time (e.g., dingo: Appleby et al., 2013;

424

Encyclopedia of Animal Behavior, 2nd edition, Volume 1

https://doi.org/10.1016/B978-0-12-809633-8.90157-2

Cognition j Responses to Death

425

giraffe: Bercovitch, 2012; elephant: Payne, 2003). Particularly in primates and cetaceans these behaviours may co-occur with transport of the corpse (Goodall, 1971; Sugiyama et al., 2009; Biro et al., 2010; Reggente et al., 2016). Together, these behaviours likely contribute to mothers learning that a dead infant does not recover. Exposure to the absence of recovery in corpses of other age-sex classes and other species probably also contributes towards a notion of the irreversibility death. Corpses of conspecifics are usually eventually abandoned, often never to be seen again as scavengers and biodegradation effectively remove visible traces from the environment. However, there is a lack of information about animals’ reactions upon returning to sites where death was witnessed. In the case of elephants, skeletal remains may become a long-term feature of the surviving groupmembers’ landscape; repeated encounters with these permanent reminders probably reinforce knowledge about the permanence of death in these long-lived, intelligent mammals (see Moss, 1988; McComb et al., 2006).

Non-Functionality Some developmental thanatologists have reported that for young children dead individuals may still be sentient; for example, they can hear when people talk to them and feel when people touch them and may be in a sleep-like state (see Barrett and Behne, 2005 for contradictory evidence). Is this true for other species? It seems reasonable to suggest that experience of investigating and manipulating corpses can foster awareness of the non-sentience of dead individuals in cognitively capable species. Primates, elephants, giraffes, cetaceans, and dingos have all been reported to mouth, manipulate, or otherwise physically contact corpses (see references above and Dudzinski et al., 2003; Douglas-Hamilton et al., 2006; Merte et al., 2008/2009; Anderson et al., 2010; Buhl et al., 2012; Stewart et al., 2012; Yang et al., 2016). Several kinds of rough treatment towards corpses have been described in monkeys and apes, including dragging, hitting, biting, pulling hair/fur, and jumping on them, possibly in attempts to elicit a reaction (Anderson et al., 2010). By manipulating corpses, particularly in unusual or rough ways, an individual can learn that dead conspecifics remain unresponsive to anything that gets done to them; bystanders may also take in this information. Additionally, of course, corpses show no signs of agency. Thus, for at least some nonhuman species, absence of spontaneous behaviour and non-responsiveness are likely to lead to the inference that dead individuals are inanimate - behaviorally and psychologically non-functional. Intra-specific aggression is a cause of death in many species, and it is sometimes followed by further extreme acts against the corpse, including cannibalism. For example, infanticidal attacks in chimpanzees are commonly followed by cannibalism of the victim (Bygott, 1972; Goodall, 1977; Kirchhoff et al., 2018), including by bystanders who played no role in killing the infant. Cannibalism has been interpreted as a way to obtain animal protein (Fox, 1975; Derocher and Wiig, 1999; Kor nan and Macek, 2011), as parental manipulation (Hrdy, 1979; Culot et al., 2011), and abnormal aggressiveness (Ryazanov et al., 2018). Although widespread in nonprimates (Hrdy, 1979), the deliberate killing and consumption of their own infants by free-ranging primate mothers appears rare. Killing and occasional cannibalism of adult chimpanzees, including within a group, also occurs (Pruetz et al., 2017). In addition to true cannibalism (sometimes called “intraspecific predation”), which involves both killing and eating conspecifics (Polis, 1981; Richardson et al., 2010), consumption without evidence of prior killing by conspecifics may occur; to distinguish the two phenomena the term “intraspecific necrophagy” has been proposed for the latter (e.g., Brown and Norris, 2004). Possible examples in primates include two adult female orangutans starting to consume their dead infants 4 days and 8 days post-mortem; this behaviour was described as “aberrant” (Dellatore et al., 2009). Bonobos have shown similar behaviour (Tokuyama et al., 2017). The authors of a recently described case in macaques (De Marco et al., 2018) suggested that the switch from caretaking and transportation to consuming the corpse suggested a fundamental change in the mother’s psychological representation of her infant. Regardless of the cause of death, for the living, the consumption of dead conspecifics is a potential further source of information about both non-functionality of corpses and the irreversibility of death.

Inevitability Young children often believe that other people (the old and sick) or animals might die, but this does not apply to significant others in their life (e.g., family members), or themselves (see Slaughter, 2005). The understanding that all living things die is typically attained by middle to late childhood. Compared to irreversibility and non-functionality, it is more difficult to attribute the inevitability component of the death concept to other species, given that we cannot interrogate them. Nevertheless, species that regularly witness – or even bring about – death in others (conspecifics and allospecifics), might come to understand that many if not most other animals can die (or be killed). Along with chimpanzees, for example (Anderson, 2018), some large carnivores with a wide range of prey might plausibly possess a concept of “killable” species. For example, subadult and even adult elephants might be seen as potential prey and therefore killable by lions (Hayward and Kerley, 2005; Power, 2009) but not leopards (Hayward et al., 2006). The knowledge that death is universal and inevitable means that one is aware of the inevitability of one’s own death. Although animals clearly fear situations in which they might die (e.g., they react against predators, and take measures to avoid getting injured or attacked, such as when swimming (terrestrial mammals) or climbing in trees (primates)). But such behaviours might reflect socially learnt or “evolutionary” fears that help animals to survive without explicit knowledge that in those situations they might die (Gray, 1987; Mobbs et al., 2015; Anderson, 2018). It has been argued that suicide – the voluntary and purposeful killing of oneself – is unique to humans (Humphrey, 2018); there is no compelling evidence for suicide or even attempted suicide in our

426

Cognition j Responses to Death

nearest evolutionary neighbours, the great apes, possibly indicating limitations to their self-reflective capacities despite their selfrecognition abilities (Anderson, 2018). Various injuries or pathologies can be identified as contributing factors in most cases of cetaceans that die stranded on beaches (e.g., Arbelo et al., 2013; Domiciano et al., 2016).

Causality By the age of 10–11 years the typically developing child is approaching Piaget’s stage of “formal operations” and has increased competence with abstract concepts (Inhelder and Piaget, 1958). Children at this stage show improved understanding of the biological causes of death (e.g., irreparable damage to vital organs); before this stage children prioritize external causes, such as accidents or guns (Slaughter and Griffiths, 2007; Speece and Brent, 1984). Again, we cannot directly ask nonverbal species about their knowledge of what causes death, but their responses to various dangerous events and situations (e.g., taking care to avoid falling from high places, non-swimmers’ avoidance of deep water, fleeing from predators, avoiding ingesting poisonous substances), suggest possible knowledge about external dangers and death, or at least injury. In predatory mammals such as felids, preferred killing techniques indicate knowledge of how to efficiently kill or at least subdue prey (e.g., Eaton, 1970). Chimpanzees show flexibility in adjusting their killing techniques to prey items of different size and strength; they may even use tools to kill or disable some prey species (summarized in Anderson, 2018). Although nonhuman animals are unlikely to have explicit biological knowledge of causes of death (for example, there is little evidence than great apes typically shun or avoid conspecifics with potentially fatal contagious diseases), at least some species probably possess some core knowledge about what events (e.g., falls, attacks, bites to specific regions) might result in death.

Experimental Approaches Most accounts of responses to dead conspecifics in the wild, including those mentioned so far in this article, are from unplanned observations. For example, among the most common causes of death in nonhuman primates are disease, predation, hunting and other activities by humans, accidents, and intra-specific killing (Anderson, 2018), but observers are not always present or well prepared to record events when deaths occur. Although valuable qualitative and quantitative data can obtained from such occasions, especially with appropriate technical backup (e.g., video cameras, radio tracking), the lack of control over relevant variables is a clear limitation. By contrast, many experimental studies in smaller species including insects, fish, birds, and small mammals have provided insights into mechanisms underlying responses to dead conspecifics. Examples of such studies are presented below.

Social Insects For social insects such as bees, ants and termites, decaying corpses are a potential source of disease within the hive or nest; therefore corpse management is important. These species have evolved several behaviours aimed at rapidly and effectively disposing of dead conspecifics, with some individuals taking on the role of “undertakers” who detect, remove and dispose of dead colony members (e.g., Sun and Zhou, 2013). Four major disposal methods have been described (López-Riquelme and Fanjul-Moles, 2013): (1) In necrophoresis, corpses are transported and either abandoned outside the nest or deposited at refuse sites within the nest. (2) Intraspecific necrophagy (commonly called cannibalism) consists of the consumption of dead conspecifics. (3) In burial behaviour, corpses are covered with soil or other matter. (4) Necrophobia is the simple avoidance of corpses or associated areas, as observed in ants and termites. The value of corpse disposal was recently demonstrated experimentally in red ants (Diez et al., 2014). Survival rates were compared between colonies that could freely remove corpses and colonies in which the size of the nest entrance reduced to impede corpse removal. Freshly killed nest-mates were placed in each nest and left over a 7-week period. From the eighth day after corpse placement, adult worker survival rates in colonies with restricted removal possibilities were significantly reduced compared to colonies with normal-size nest entrances. Ants in control nests performed typical necrophoric behaviours, whereas those in restrictedentrance nests cut up corpses and ejected the pieces, or they moved corpses to areas of the nest as far away from the larvae as possible. By the end of the study larvae survival rates were also lower in the restricted-entrance nests. Social insects’ corpse disposal behaviours are innate – or pre-programmed – patterns triggered especially by chemical cues from the corpses. An early breakthrough study showed that any objects – even live, healthy nest-mates - that were daubed with oleic acid – one of the substances produced by decaying corpses – were transported by healthy ants to refuse piles (Wilson et al., 1958). There is now general agreement that chemical cues (“necromones”) alone or in combination elicit undertaking behaviour in social insects (Lopez-Riquelme and Fanjul-Moles, 2013). Furthermore, two chemical compounds found in live ants may inhibit inappropriate necrophoric responses, by masking pre-existing necromones; the rapid fading of such compounds in freshly dead workers may trigger their removal even before necromones take effect (Choe et al., 2009). Research on necromones in social insects gave rise to experimental work in rodents, leading to the discovery of the role of the chemical compounds cadaverine and putrescine in eliciting burial of dead conspecifics by rats (Pinel et al., 1981). Experiments have shown that various chemical cues including cadaverine cause diverse aquatic species to avoid areas associated with injured, diseased, and dead conspecifics (e.g., Adams and Claeson, 1998; Bryer et al., 2001; Hussain et al., 2013; Oliviera et al., 2014; Candia-Zulbarán et al., 2015).

Cognition j Responses to Death

427

Birds In contrast to the growing focus on chemical cues to death in experimental thanatological research on insects and aquatic organisms, studies on birds have emphasized visual stimuli. When the dried skin and feathers of a dead conspecific or similarly coloured non-social objects were placed near a feeder, Western scrub-jays emitted more loud vocalizations (“cacophonous reactions”) and took fewer nuts from the feeder in the presence of the dead stimulus. Playbacks of these calls attracted other jays to the site. Although a stuffed owl elicited similar cacophonous aggregations of jays, unlike the dead jay, it also got mobbed (Iglesias et al., 2012). (Experimental research on responses to model predators and dead conspecifics has recently been extended to passerine birds; see Carlson et al., 2017). In a follow-up study of scrub-jays’ use of carcasses as an indicator of risk, Iglesias et al. (2014) showed that dead birds of other species of similar size to scrub-jays also elicited calling and cacophonous aggregations, and suppressed foraging, whereas carcasses of smaller-sized birds were less effective. Experiments with American crows showed broadly similar responses as scrubjays towards dead conspecifics, along with negative reactions to either a hawk alone or (especially) a hawk or a human in association with a dead crow. Crows were notably less responsive to the carcass of a pigeon, and pigeons also took little notice of dead conspecifics (Swift and Marzluff, 2015). Furthermore, positron emission tomography revealed particular activation of the hippocampal region of the brain in crows watching a human holding a dead crow; other risk stimuli activated other forebrain regions (Cross et al., 2013). Brain imaging is likely to be increasingly valuable for revealing neural circuits involved in perceiving cues related to death in many animal species, including humans.

Conclusions Death, and behaviour toward death are of profound importance in the study of animal life. Animal behaviour researchers ask many different questions about death and dying, including mechanisms underlying detection of death and dealing with corpses, survival value of responses to the dead, learning about death, emotional correlates of bereavement, and different species’ (or, in humans, age-related) understanding of death (see Anderson et al., 2018). The issue of emotional consequences of experiencing death in conspecifics is of relevance not least from the perspective of the ethics of animal slaughter and euthanasia practices. Short-term alarm and distress responses to the sight or odour of dead conspecifics have already been reviewed above, but it is also important to consider whether some species can experience longer-term grief in response to death. There is strong evidence for this. Several decades ago many studies of involuntary separation-induced depression in nonhuman primates showed that permanent loss of an attachment figure could induce intense, negative behavioural and emotional responses including depression (Anderson, 2016, 2017). Behavioural and hormonal study of female baboons who had recently lost a close relative following predatory attacks by leopards and lions revealed significant stress hormone elevations in bereaved females but not in unrelated females, even though the latter also witnessed the predatory attacks (Engh et al., 2006). Conceivably, grief may affect not only recently bereaved mothers, but also close relatives and friends of deceased individuals. Clearly, further studies combining behavioural and physiological measures would help clarify the emotional consequences of witnessing death of a conspecific in a wide range of species (King, 2013). Death will not go away, and it seems likely to be a focus of increased observational and research efforts as we seek to better understand the evolution and significance of the diverse responses it elicits.

See also: Cognition: Mental Time Travel: Can Animals Recall the Past and Plan for the Future?; Time: What Animals Know.

References Adams, M.J., Claeson, S., 1998. Field responses of tadpoles to conspecific and heterospecific alarm. Ethology 104, 955–961. Anderson, J.R., 2016. Comparative thanatology. Current Biology 26, R553–R556. Anderson, J.R., 2017. Comparative evolutionary thanatology of grief, with special reference to nonhuman primates. Japanese Review of Cultural Anthropology 18, 173–189. Anderson, J.R., 2018. Chimpanzees and death. Philosophical Transactions of the Royal Society B, 373, 20170257. Anderson, J.R., Biro, D., Pettitt, P., 2018. Evolutionary thanatology. Philosophical Transactions of the Royal Society B, 373, 20170262. Anderson, J.R., Gillies, A., Lock, L.C., 2010. Pan thanatology. Current Biology 20 (8), R349–R351. Appleby, R., Smith, B., Jones, D., 2013. Observations of a free-ranging adult female dingo (Canis dingo) and littermates’ responses to the death of a pup. Behavioural Processes 96, 42–46. Arbelo, M., et al., 2013. Pathology and causes of death of stranded cetaceans in the Canary Islands (1999–2005). Diseases of Aquatic Organisms 103, 87–99. Barrett, H.C., Behne, T., 2005. Children’s understanding of death as the cessation of agency: A test using sleep versus death. Cognition 96, 93–108. Bercovitch, F.B., 2012. Giraffe cow reaction to the death of her newborn calf. African Journal of Ecology 51, 376–379. Biro, D., Humle, T., Koops, K., Sousa, C., Hayashi, M., Matsuzawa, T., 2010. Chimpanzee mothers at Bossou, Guinea carry the mummified remains of their dead infants. Current Biology 20, R351–R352. Bonoti, F., Leonardi, A., Mastora, A., 2013. Exploring children’s understanding of death: Through drawings and the Death Concept Questionnaire. Death Studies 37, 47–60. Brown, M.W., Norris, M.E., 2004. Survivorship advantage of conspecific necrophagy in overwintering boxelder bugs (Heteroptera: Rhopalidae). Annals of the Entomological Society of America 97, 500–503. Bryer, P.J., Mirza, R.S., Chivers, D.P., 2001. Chemosensory assessment of predation risk by slimy sculpins (Cottus cognatus): Responses to alarm, disturbance, and predator cues. Journal of Chemical Ecology 27, 533–546. Buhl, J.S., Aur, B., Ruiz-Lambides, A., Gonzalez-Martinez, J., Platt, M.L., Brent, L.J.N., 2012. Response of rhesus macaques (Macaca mulatta) to the body of a group member that died from a fatal attack. International Journal of Primatology 33, 860–871.

428

Cognition j Responses to Death

Bygott, J.D., 1972. Cannibalism among wild chimpanzees. Nature 238, 410–411. Candia-Zulbarán, R.I., Briones-Fourzán, P., Lozano-Álvarez, E., Barradas-Ortiz, C., Negrete-Soto, F., 2015. Caribbean spiny lobsters equally avoid dead and clinically PaV1-infected conspecifics. ICES Journal of Marine Science 72 (Suppl. 1), S164–S169. Carlson, N.V., Pargeter, H.M., Templeton, C.N., 2017. Sparrowhawk movement,calling, and presence of dead conspecifics differentially impact blue tit (Cyanistes caeruleus) vocal and behavioral mobbing responses. Behavioral Ecology and Sociobiology 71, 133. https://doi.org/10.1007/s00265-017-2361-x. Choe, D.H., Millar, J.C., Rust, M.K., 2009. Chemical signals associated with life inhibit necrophoresis in Argentine ants. Proceedings of the National Academy of Sciences of the United States of America 106, 8251–8255. Cross, D.J., Marzluff, J.M., Palmquist, I., Minoshima, S., Shimizu, T., Miyaoka, R., 2013. Distinct neural circuits underlie assessment of a diversity of natural dangers by American crows. Proceedings of the Royal Society B 280, 20131046. Culot, L., Lledo-Ferrer, Y., Hoelscher, O., Muñoz-Lazo, J.J., Huynen, M.-C., Heymann, E.W., 2011. Reproductive failure, possible maternal infanticide, and cannibalism in wild moustached tamarins, Saguinus mystax. Primates 52, 179–186. Dellatore, D.F., Waitt, C.D., Foitova, I., 2009. Two cases of mother-infant cannibalism in orangutans. Primates 50, 277–281. De Marco, A., Cozzolino, R., Thierry, B., 2018. Prolonged transport and cannibalism of mummified infant remains by a Tonkean macaque mother. Primates 59, 55–59. Derocher, A.E., Wiig, Ø., 1999. Infanticide and cannibalism of juvenile polar bears (Ursus maritimus) in Svalbard. Arctic 52, 307–310. Diez, L., Lejeune, P., Detrain, C., 2014. Keep the nest clean: Survival advantages of corpse removal in ants. Biology Letters 10, 20140306. Available at: https://doi.org/10.1098/ rsbl.2014.0306. Domiciano, I.G., Domit, C., Broadhurst, M.K., Koch, M.S., Bracarense, A.P.F.R.L., 2016. Assessing disease and mortality among small cetaceans stranded at a world heritage site in southern Brazil. PLOS ONE 11 (2), e0149295. https://doi.org/10.1371/journal.pone.0149295. Douglas-Hamilton, I., Bhalla, S., Wittemeyer, G., Vollrath, F., 2006. Behavioural reactions of elephants towards a dying and deceased matriarch. Applied Animal Behaviour Science 100, 87–102. Dudzinski, K.M., Sakai, M., Masaki, K., Kogi, K., Hishi, T., Kurimoto, M., 2003. Behavioural observations of bottlenose dolphins towards two dead conspecifics. Aquatic Mammals 29, 108–116. Eaton, R.L., 1970. The predatory sequence, with emphasis on killing behavior and its ontogeny, in the cheetah (Acinonyx jubatus Schreber). Zeitschrift für Tierpsychologie 27, 492–504. Engh, A.L., Beehner, J.C., et al., 2006. Behavioural and hormonal responses to predation in female chacma baboons (Papio hamadryas ursinus). Proceedings of the Royal Society B 273, 707–712. Fox, L.R., 1975. Cannibalism in natural populations. Annual Review of Ecological Systems 6, 87–106. Goodall, J., 1971. In the Shadow of Man. Collins, London. Goodall, J., 1977. Infant killing and cannibalism in free-living chimpanzees. Folia Primatologica 28, 259–289. Gray, J.A., 1987. The Psychology of Fear and Stress, second ed. Cambridge University Press, Cambridge. Hussain, A., Saraiva, L.R., Ferrero, D.M., et al., 2013. High-affinity olfactory receptor for the death-associated odor cadaverine. Proceedings of the National Academy of Sciences of the United States of America 110, 19579–19584. Hayward, M.W., Henschel, P., O’Brien, J., et al., 2006. Prey preferences of the leopard (Panthera pardus). Journal of Zoology 270, 298–313. Hayward, M.W., Kerley, G.I.H., 2005. Prey preferences of the lion (Panthera leo). Journal of Zoology 267, 309–322. Hrdy, S.B., 1979. Infanticide among animals: A review, classification, and examination of the implications for the reproductive strategies of females. Ethology and Sociobiology 1, 13–40. Humphrey, N., 2018. The lure of death: Suicide as a uniquely human phenomenon. Philosophical Transactions of the Royal Society B. Iglesias, T.L., McElreath, R., Patricelli, G.L., 2012. Western scrub-jay funerals: Cacophonous aggregations in response to dead conspecifics. Animal Behaviour 84, 1103–1111. Iglesias, T.L., Stetkevich, R.C., Patricelli, G.L., 2014. Dead heterospecifics as cues of risk in the environment: Does size affect response? Behaviour 151, 1–22. Inhelder, B., Piaget, J., 1958. The Growth of Logical Thinking from Childhood to Adolescence. Basic Books, New York. King, B.J., 2013. How Animals Grieve. University of Chicago Press, Chicago. Kirchhoff, C.A., Wilson, M.L., Mjungu, D.C., et al., 2018. Infanticide in chimpanzees: Taphonomic case studies from Gombe. American Journal of Physical Anthropology 165, 108–122. Kornan, M., Macek, M., 2011. Parental infanticide followed by cannibalism in golden eagles (Aquila chrysaetos). Journal of Raptor Research 45, 95–96. López-Riquelme, G.O., Fanjul-Moles, M.L., 2013. The funeral ways of social insects: Social strategies for corpse disposal. Trends in Entomology 9, 71–129. McComb, K., Baker, L., Moss, C., 2006. African elephants show high levels of interest in the skulls and ivory of their own species. Biology Letters 2, 26–28. Merte, C.E., Gough, K.F., Schulte, B.A., 2008/2009/. Investigation of a fresh African elephant carcass by conspecifics. Pachyderm 45, 124–126. Mobbs, D., Hagan, C.C., Dalgleish, T., Silston, B., Prevost, C., 2015. The ecology of human fear: Survival optimization and the nervous system. Frontiers in Neuroscience 9, 55. https://doi.org/10.3389/frins.2015.00055. Moss, C., 1988. Elephant Memories: Thirteen Years in the Life of an Elephant Family. Fontana/Collins, Glasgow. Oliviera, T.A., Koakoski, G., Costa da Motta, A., et al., 2014. Death-associated odors induce stress in zebrafish. Hormones and Behavior 65, 340–344. Payne, K., 2003. Sources of social complexity in the three elephant species. In: de Waal, F.B.M., Tyack, P.L. (Eds.), Animal Social Complexity: Intelligence, Culture, and Individualized Societies. Harvard University Press, Cambridge, MA, pp. 57–85. Pinel, J.J.J., Gorzalka, B.B., Ladak, F., 1981. Cadaverine and putrescine initiate the burial of dead conspecifics by rats. Physiology and Behavior 27, 819–824. Polis, G.A., 1981. The evolution and dynamics of intraspecific predation. Annual Review of Ecological Systems 12, 225–251. Power, R.J., 2009. Lion predation on elephants in the Savuti, Chobe National Park, Botswana. African Zoology 44, 36–44. Pruetz, J.D., Boyer Ontl, K., Cleaveland, E., Lindshield, S., Marshack, J., Wessling, E.G., 2017. Intragroup lethal aggression in west African chimpanzees (Pan troglodytes verus): Inferred killing of a former alpha male at Fongoli, Senegal. International Journal of Primatology 38, 31–57. Reggente, M.A.L., Alves, F., Nicolau, C., et al., 2016. Nurturant behavior toward dead conspecifics in free-ranging mammals: New records for odontocetes and a general review. Journal of Mammalogy 97, 1428–1434. Richardson, M.L., Mitchell, R.F., Reagel, P.F., Hanks, L.M., 2010. Causes and consequences of cannibalism in noncarnivorous insects. Annual Review of Entomology 55, 39–53. Ryazanov, S.D., Kirillova, A.D., Laskina, N.B., Burkanov, V.N., 2018. Infanticide and cannibalism in Steller sea lions (Eumetopias jubatus). Marine Mammal Science 34, 200–207. Slaughter, V., 2005. Young children’s understanding of death. Australian Psychologist 40 (3), 179–186. Slaughter, V., Griffiths, M., 2007. Death understanding and fear of death in young children. Clinical Child Psychology and Psychiatry 12, 525–535. Speece, M.W., 1995. Children’s concepts of death. Michigan Family Review 1, 57–69. Speece, M.W., Brent, S.B., 1984. Children’s understanding of death: A review of three components of a death concept. Child Development 55, 1671–1686. Stewart, F.A., Piel, A.K., O’Malley, R.C., 2012. Responses of chimpanzees to a recently dead community member at Gombe National Park, Tanzania. American Journal of Primatology 74, 1–7. Sugiyama, Y., Kurita, H., Matsui, T., Kimoto, S., Shimomura, T., 2009. Carrying of dead infants by Japanese macaque (Macaca fuscata) mothers. Anthropological Science 117, 113–119. Sun, Q., Zhou, X., 2013. Corpse management in social insects. International Journal of Biological Sciences 9, 313–321. Swift, K.N., Marzluff, J.M., 2015. Wild American crows gather around their dead to learn about danger. Animal Behaviour 109, 187–197. Tokuyama, N., Moore, D.L., Graham, K.E., Lokasola, A., Furuichi, T., 2017. Cases of maternal cannibalism in wild bonobos (Pan paniscus) from two different field sites, Wamba and Kokolopori, Democratic Republic of the Congo. Primates 58, 7–12. Wilson, E.O., Durlach, N., Roth, L., 1958. Chemical releasers of necrophoric behaviour in ants. Psyche 65, 108–114. Yang, B., Anderson, J.R., Li, B.G., 2016. Tending a dying adult in a wild multi-level primate society. Current Biology 26, R403–R404.

Time: What Animals Know JD Crystal, University of Georgia, Athens, GA, USA © 2010 Elsevier Ltd. All rights reserved.

Abstract Time is a fundamental dimension of life, and many animals have evolved the ability to time intervals from a fraction of a second to hours. Traditionally, interval timing is the ability to time short intervals and is distinguished from circadian timing, the timing of the rhythm of day and night. Recent data show that interval timing shares many properties with circadian timing. Converging approaches document that timing of targets from milliseconds to days are subserved by similar mechanisms that are conserved throughout this wide range. The conclusion is that timing is based on multiple oscillators, each with a different period.

Keywords Endogenous oscillator; Fourier analysis; Linear timing; Nonlinear timing; Oscillator; Periodogram analysis; Phase shift; Scalar timing; Temporal information processing; Temporal representation; Time perception; Time series analysis; Timing

Introduction Time is a fundamental dimension of life, and many animals have evolved the ability to time intervals lasting from a fraction of a second to hours. Events unfold in time, and we experience events over time. The ability to time events is ubiquitous, as illustrated by the assortment of events that are affected by temporal processing – speech, music, motor control, foraging, decision-making, sleep–wake cycles, and appetite. Thus, an understanding of learning, cognition, and performance requires an analysis of temporal information processing. A distinction is traditionally made between interval timing and circadian timing. Interval timing is the ability to time shorter intervals, typically in the range of seconds to minutes. Circadian timing is the ability to adjust to the daily cycle, which has a period of 24 h. A circadian clock repeats itself approximately every 24 h, and this period is set by external cues such as daylight or large meals. This article concerns interval timing and the mechanisms that subserve interval timing.

Timing in Natural Settings Three examples of timing in animal behavior are described. Free-living hummingbirds time the interval between successive visits to flowers that they visited throughout the day. In experiments in which the flowers were replenished after different intervals of time (e.g., 10 vs. 20 min), the revisits to flowers tracked the replenishment rate. Hummingbirds apparently update their memories of when and where food was encountered for each flower and how long ago they last visited each location. Bees also adjust the time of visiting food sources on the basis of the amount of time elapsed since the last visit. Scavenging birds appear to anticipate food availability and arrive at locations before food has reached its peak amount.

Representations of Time How are intervals timed? The basic idea is that an interval elapses with respect to the occurrence of some event. With a mechanical artifact such as a stopwatch, we can track the elapsing interval. When the event begins, we reset the stopwatch to zero and start the stopwatch. Reading the stopwatch provides an estimate of time to complete the event. A basic question about temporal information processing concerns the mechanism by which time is represented. Two types of temporal representations are described. One mechanism that may be used to time intervals is a pacemaker-accumulator. A pacemaker emits pulses as a function of time; the accumulator counts or integrates the number of pulses emitted. An hourglass is a familiar example of a pacemaker-accumulator – the elapsed duration, since the hourglass was turned over (i.e., reset), is indexed by the amount of sand in the bottom chamber of the hourglass. A second mechanism that may be used to time intervals is an oscillator mechanism. Circadian timing is the best known biological oscillator. Circadian timing is based on the completion of a periodic process that is approximately a day. Time of day is indexed by the phase of a circadian periodic process. Circadian timing is widespread. Oscillator properties of timing intervals of approximately a day may be extended to intervals in the range of seconds to minutes as discussed in the section ‘Oscillator Properties of Interval Timing’.

Encyclopedia of Animal Behavior, 2nd edition, Volume 1

https://doi.org/10.1016/B978-0-12-813251-7.00092-4

429

430

Cognition j Time: What Animals Know

Formal Properties of Interval and Circadian Timing In 1997, Gibbon and colleagues provided a classic description of the operating characteristics of interval and circadian timing systems. Gibbon’s outline is briefly described in this section. According to the classic description, the interval timing system is based on a pacemaker-accumulator mechanism, and the circadian system is based on an endogenous-oscillator mechanism. Endogenous means that the oscillator does not require continued periodic input to produce ongoing periodic output. For example, when an animal is exposed to daily periodic light cycles such as 12 h of light followed by 12 h of darkness, activity patterns occur at species-typical times of day. After termination of the periodic light cycle (e.g., constant dim illumination), behavior ‘free runs’ with a period that typically departs slightly from 24 h. Free running behavior after the termination of periodic stimuli provides evidence that the timing system is endogenous because this pattern of data is important to rule out the possibility that observed behaviors are linked to the occurrence of daily environmental changes (e.g., temperature or noise fluctuations). In contrast, Gibbon and colleagues outlined the interval timing system as requiring resetting. The timing system operates on an elapsing interval timed with respect to the occurrence of some stimulus; a single presentation of the stimulus is necessary and sufficient to reset the interval timing system (i.e., one shot reset). The circadian system operates within a limited range of entrainment. In particular, presentation of a periodic input entrains the endogenous oscillator only if the periodic input is within a limited range of periods near 24 h. By contrast, the interval timing system has a broad training range covering 3–4 orders of magnitude from seconds to hours. A hallmark feature of the circadian system is its slow adjustment to a phase shift. A phase shift is an abrupt change in the initiation of a periodic process. Jet lag is a familiar example; we experience a phase shift in the unusual wake-up times after flying to a destination across several time zones. It usually requires several days before activities are synchronized to the new time zone. By contrast, the interval timing system undergoes immediate adjustment to a phase shift; the response to a single shift in a cycle is complete adjustment or complete resetting of the timing processes (i.e., one-shot reset). Temporal performance based on a circadian oscillator is highly precise as measured by cycle-to-cycle variation. Precision is typically measured relative to the interval being timed. In particular, the coefficient of variation (CV) is the standard deviation of time estimates divided by the mean of time estimates. The CV of circadian performance is
1–5%. By contrast, interval timing performance is characterized by a much low level of precision (CV of 10–35%). Thus, a characteristic of a circadian oscillator is relatively high timing precision. In particular, a consequence of having an endogenous oscillator dedicated to timing select values within a limited range appears to be relatively high sensitivity to timing these target durations. The variance properties of timing have played an important role in understanding interval timing. By contrast, the analysis of variance properties has had less impact in the study of circadian timing.

Oscillator Properties of Interval Timing The sections that follow describe a series of empirical tests that were designed to evaluate the hypothesis that interval timing is based, at least in part, on oscillatory processes.

Resetting Properties of Short-Interval Timing Although the phase-shift manipulation is a classic experimental design for the diagnosis of a circadian oscillator, the same manipulation may be used to assess a short-interval timing mechanism. Figure 1 shows an example of a phase-shift manipulation in shortinterval timing. Rats were trained to time 100 s using a fixed-interval procedure. In a fixed-interval procedure, the first response after the fixed interval elapses produces a food pellet. To produce a phase shift, an early, response-independent pellet was delivered (i.e., free food). Four food cycles were required before adjustment was complete, which is consistent with the hypothesis that shortinterval timing of 100 s is based on an oscillator mechanism.

Endogenous Oscillations in Short-Interval Timing A critical diagnostic property of a timing mechanism may be assessed by discontinuing periodic input (i.e., extinction) and assessing subsequent anticipatory behavior. The defining feature of an oscillator is that periodic output from the oscillator continues after the discontinuation of periodic input. Similarly, a defining feature of a pacemaker-accumulator system is that elapsed time is measured with respect to the presentation of a stimulus, according to the classic description of this system described earlier. Thus, output of a short-interval system is periodic only when driven by periodic input. Moreover, periodic output is expected to cease if periodic input is discontinued. To test pacemaker-accumulator and endogenous-oscillator mechanisms, rats were trained with a variety of short intervals (e.g., 48, 96, and 192 s). The critical manipulation was the suspension of food delivery. Figure 2 shows data from a representative individual rat, and Figure 3 shows group data. Periodic delivery of food produced periodic behavior during training (Figure 3, left column), as predicted by both mechanisms. Behavior continued to be periodic after termination of periodic input (Figure 3, right column), consistent with an endogenous-oscillator, but not a pacemaker-accumulator, mechanism. The

Cognition j Time: What Animals Know

431

Figure 1 A phase shift produces gradual adjustment in short-interval timing. Left panel: Schematic representation of training, phase-shift manipulation, predictions, and data (double plotted to facilitate inspection of transitions across successive intervals; consecutive 100-s fixed intervals are plotted left to right and top to bottom). Rats (n ¼ 14) timed 100-s intervals, and the last five intervals before the phase shift are shown (F ¼ food pellet, S ¼ start time of response burst). A 62-s phase advance (i.e., early pellet) on average was produced by the delivery of a response-independent food (FFREE). All other food-to-food intervals were 100 s (FPS ¼ food post phase shift). Dashed lines indicate predictions if rats are insensitive (0% adjustment, purple) or completely sensitive (100% adjustment, pink) to the most recently delivered food pellet. A pacemaker-accumulator mechanism predicts 100% adjustment on the initial interval after the phase shift on the assumption of complete reset. An oscillator mechanism predicts initial incomplete adjustment. Data (D) indicate incomplete adjustment on the first three trials. Right panel: Start times on the initial three trials were earlier than in preshift baseline. Resetting was achieved on the fourth trial. Each 45-mg food pellet was contingent on a lever press after 100 s in 12-h sessions. The start of a response burst was identified on individual trials by selecting the response that maximized the goodness of fit of individual responses to a model with a low rate followed by a high rate. The same conclusions were reached by measuring the latency to the first response after food. Baseline was the average start time on the five trials before the phase shift. Left panel: Zero on the y-axis (purple dashed line) corresponds to complete failure to adjust to the phase shift; 100% (pink dashed line) corresponds to complete resetting. Error bars represent 1 SEM. Reproduced from Crystal JD (2006b) Time, place, and content. Comparative Cognition & Behavior Reviews 1: 53–76, with permission.

Figure 2 Many small interresponse times in short-interval timing are punctuated by much longer interresponse times, and punctuation by relatively long interresponse times continued after termination of periodic food delivery. Interresponse time (i.e., times of responses Rnþ1  Rn) is plotted as a function of response time for a representative rat. During training, food was delivered on a fixed interval 96-s schedule. During testing, food was not delivered (i.e., extinction). Extinction began at a randomly selected point in the session. The response measure was the time of occurrence of photo beam interruptions in the food trough. Reproduced from Crystal JD and Baramidze GT (2007) Endogenous oscillations in short-interval timing. Behavioural Processes 74: 152–158, with permission from Elsevier.

periodic behavior in extinction appears to be based on entrainment to the periodic feeding in training given that the period in extinction increased as a function of the period in training. Short-interval timing is, at least in part, based on a self-sustaining, endogenous oscillator.

Timing Long Intervals A critical hypothesis about an oscillator is that it functions to confer improved sensitivity to time intervals near the period of the oscillator. To test this hypothesis, a series of experiments investigating meal anticipation was undertaken to identify a local peak in sensitivity to time 24 h. To examine anticipation of long intervals, food was restricted to 3-h meals, which rats earned by breaking a photo beam in a food trough. Critically, the rats tend to inspect the food trough before meals start, thereby providing a temporal

432

Cognition j Time: What Animals Know

Figure 3 Endogenous oscillations in short-interval timing continue after the termination of periodic input. Short time Fourier transforms are shown for training (left panels) and testing (right panels) conditions using fixed interval 48-, 96-, and 192-s procedures. The three-dimensional images show frequency (period ¼ 1/frequency) on the vertical axis as a function of time within the session along the horizontal axis; the color scheme represents the amount of power from the Fourier analysis. Concentrations of high power occur at a frequency of
0.02, 0.01, and.0005 which correspond to periods of
50, 100, and 200 s in top, middle, and bottom panels, respectively. Adapted from Crystal JD and Baramidze GT (2007) Endogenous oscillations in short-interval timing. Behavioural Processes 74: 152–158, with permission from Elsevier.

anticipation function for each intermeal interval condition. Figure 4 shows anticipation functions for intermeal intervals in the circadian range (22–26 h) and well outside this range (14 and 34 h). Note that response rates increased later into the intermeal interval for intervals near the circadian range than for intervals outside this range. The response distributions were used to estimate sensitivity to time (i.e., relatively small spreads in the distributions correspond to relatively high sensitivity to time). As shown in Figure 5, intermeal intervals in the circadian range produced spreads that were smaller (i.e., lower variability) compared to intervals outside this range. Note that the data in Figure 5 document a local maximum in sensitivity to time near 24 h, consistent with the hypothesis that a function of a circadian oscillator is improved sensitivity to time.

Endogenous Oscillations in Long-Interval Timing The examples of timing noncircadian long intervals in Figures 4 and 5 indicate that rats can time intervals outside the circadian range, but they do not identify the mechanism. In particular, these data could be based on an endogenous oscillator mechanism or a pacemaker-accumulator mechanism reset by meals. By contrast, these examples document endogenous oscillations in timing

Cognition j Time: What Animals Know

433

Figure 4 Response rate increased later into the interval for intermeal intervals near the circadian range (unfilled red symbols) relative to intervals outside this range (filled blue symbols); dashed lines indicate width of response rate functions. Anticipatory responses increase immediately prior to the meal for all intermeal intervals except 34 h. Each 45-mg food pellet was contingent on a photo beam break after a variable interval during 3-h meals. Intermeal intervals were tested in separate groups of rats (n ¼ 3–5 per group). The end of the meal corresponds to 1 on the x-axis. Testing was conducted in constant darkness. Adapted from Crystal (2001a). Reproduced from Crystal JD (2006b) Time, place, and content. Comparative Cognition & Behavior Reviews 1: 53–76, with permission.

Figure 5 Intervals near the circadian range (red symbols) are characterized by higher sensitivity than intervals outside this range (blue symbols). Variability in anticipating a meal was measured as the width of the response distribution prior to the meal at 70% of the maximum rate, expressed as a percentage of the interval (N ¼ 29). The interval is the time between light offset and meal onset in a 12–12 light-dark cycle (leftmost two circles) or the intermeal interval in constant darkness (all other data). The percentage width was smaller in the circadian range than outside this range. The width/interval did not differ within the circadian or noncircadian ranges. The same conclusions were reached when the width was measured as 25%, 50%, and 75% of the maximum rate. The data are plotted on a reversed-order y-axis so that local maxima in the data correspond to high sensitivity, which facilitates comparison with other measures of sensitivity (e.g., Figure 7). Mean SEM ¼ 2.4. Adapted from Crystal (2001a). Reproduced from Crystal JD (2006b) Time, place, and content. Comparative Cognition & Behavior Reviews 1: 53–76, with permission.

434

Cognition j Time: What Animals Know

short intervals (1–3 min) by demonstrating that behavior continued after the termination of periodic input. The same experimental approach is used in this section to document endogenous oscillations in long-interval timing (16 h). Rats earned food by interrupting a photo beam in the food trough during 3-h meals using a 16-h intermeal interval. After approximately a month of experience with the intermeal interval, the meals were discontinued. Figure 6 (top panel) shows that the response rate increased as a function of time prior to the meals, documenting that the rats timed 16 h, consistent with either

Figure 6 Endogenous oscillations in long-interval timing continue after the termination of periodic input. Response rate increased as a function of time within the 16-h intermeal interval cycle during the first and second nonfood cycle. Response rate (frequency of responses expressed as a proportion of the maximum frequency within the cycle) is plotted as a function of time within the cycle. The cycle included meals (indicated by the solid rectangle) during training (top panel). The meals were omitted (indicated by the dashed rectangles) in the first (middle panel) and second (bottom panel) nonfood cycles. Reproduced from Crystal JD (2006a) Long-interval timing is based on a self sustaining endogenous oscillator. Behavioural Processes 72: 149–160, with permission from Elsevier.

Cognition j Time: What Animals Know

435

oscillator or pacemaker-accumulator mechanisms. When two successive meals were skipped, the rats anticipated the arrival of two successive 16-h intervals (Figure 6 middle and bottom panels), consistent with the use of an endogenous oscillator. In particular, the response rate was reliably higher during the 3-h omitted meal relative to the earlier 13 h for both first and second nonfood cycles. If timing was based on a pacemaker-accumulator reset by meals, then the rats would be expected to time the first, but not the second, skipped meal. A pacemaker-accumulator does not predict an increase in response rate prior to the second skipped meal because elapsed time since the last meal is larger than the intermeal interval during the second nonfood cycle (i.e., time since the last meal is unusually long at this point). The reliability of a periodic trend was assessed and observed periods were estimated with a periodogram analysis; a periodogram analysis involves wrapping a response rate function around different proposed periods to identify the period that best fits the observed data. A reliable periodic trend was observed for each rat, and the mean period in extinction (20.4  0.9 h, mean  SEM) was reliably different from 16 and 24 h. These data suggest that the natural period of the oscillator that drove behavior was 20.4 h, which is distinct from the circadian oscillator; according to this hypothesis, the two oscillatory systems are dissociated by their different characteristic periods. However, the data are also consistent with the hypothesis that the circadian oscillator’s freerunning period is modified by the periodic input to which it was previously exposed. According to both of these hypotheses, longinterval timing is based on a self-sustaining, endogenous oscillator; the hypotheses differ in specifying the characteristic period of the oscillators. In either case, long-interval timing is based on a self-sustaining, endogenous oscillator.

Variance Properties in Circadian and Short-Interval Timing As mentioned earlier, the study of variance properties has historically played a significant role in the development of theories of short-interval, but not circadian, timing. However, the data summarized in Figure 5 suggest that a function of the wellestablished circadian oscillator is the relative improvement in sensitivity to time
24 h. Thus, other putative oscillators may be identified by documenting other local maxima in sensitivity to time. In addition, the observation that short-interval timing in the range of 1–3 min exhibits endogenous, self-sustaining patterns of behavior after the termination of periodic input reinforces the expectation that short-interval timing may be based on an endogenous oscillatory mechanism. To search for local peaks in sensitivity to time in the short-interval range, a series of experiments were conducted using many, closely spaced target intervals. Figure 7 shows sensitivity to time plotted as a function of stimulus duration from these experiments. Sensitivity to time short intervals is characterized by multiple local peaks. Each peak in sensitivity to time may identify the period of a short-period oscillator. The procedure involved presenting a short or long stimulus followed by the insertion of two response levers. Left or right lever presses were designated as correct after short or long stimuli. Accuracy was maintained at
75% correct by adjusting the duration of the long stimulus after blocks of trials. Sensitivity to time was approximately constant for short durations from 0.1 to 34 s. However, local maxima in sensitivity to time were observed at
0.3, 1.2, 12, and 24 s.

Figure 7 Sensitivity to time is characterized by local maxima at 12 and 24 s (left panel), 12 s (middle panel), and 0.3 and 1.2 s (right panel). Green symbols: average across rats. Red symbols: a running median was performed on each rat’s data and the smoothed data were averaged across rats to identify the most representative local maxima in sensitivity. Left panel: Rats discriminated short and long noise durations with the duration adjusted to maintain accuracy at
75% correct. Short durations were tested in ascending order with a step size of 1 s (n ¼ 5) and 2 s (n ¼ 5). Sensitivity was similar across step sizes, departed from zero, and was nonrandom. Mean SEM ¼ 0.03. Middle panel: Methods are the same as described in left panel, except short durations were tested in random order (n ¼ 7) or with each rat receiving a single interval condition (n ¼ 13); results from these conditions did not differ. Sensitivity departed from zero and was nonrandom. Mean SEM ¼ 0.02. Right panel: Methods are the same as described in left panel, except intervals were defined by gaps between 50-ms noise pulses and short durations were tested in descending order with a step size of 0.1 s (n ¼ 6). Sensitivity departed from zero and was nonrandom. Mean SEM ¼ 0.04. Sensitivity was measured using d 0 from signal detection theory. d 0 ¼ z[p(short response j short stimulus)] – z[p(short response j long stimulus)]. Relative sensitivity is d 0 – mean d 0 . Adapted from Crystal (1999, 2001b). Reproduced from Crystal JD (2006b) Time, place, and content. Comparative Cognition & Behavior Reviews 1: 53–76, with permission.

436

Cognition j Time: What Animals Know

Figure 8 Multiple local maxima in sensitivity to time are observed in the discrimination of time across seven orders of magnitude. The existence of a local maximum near a circadian oscillator (peak on right side; purple squares) and other local maxima in the short-interval range (peaks on left side; blue, red, and green circles) are consistent with the hypothesis that timing is mediated by multiple oscillators. Intervals in the blank region in the center of the figure have not been tested. Left side: Rats discriminated short and long durations, with the long duration adjusted to maintain accuracy at 75% correct. Short durations were tested in sequential order (blue and red circles; N ¼ 26) or independent order (green circles; N ¼ 20). Circles represent relative sensitivity using d 0 from signal detection theory and are plotted using the y-axis on the left side of the figure. Right side: Rats received food in 3-h meals with fixed intermeal intervals by breaking a photo beam inside the food trough. The rate of photo beam interruption increased before the meal. Squares represent sensitivity, which was measured as the width of the anticipatory function at 70% of the maximum rate prior to the meal, expressed as a percentage of the interval (N ¼ 29). The interval is the time between light offset and meal onset in a 12–12 lightdark cycle (leftmost two squares) or the intermeal interval in constant darkness (all other squares). Squares are plotted with respect to the reversedorder y-axis on the right side of the figure. Y-axes use different scales, and the x-axis uses a log scale. Adapted from Crystal (1999, 2001a, 2001b). Reproduced from Crystal JD (2006b) Time, place, and content. Comparative Cognition & Behavior Reviews 1: 53–76, with permission.

Figure 8 shows multiple local maxima in sensitivity to time across several orders of magnitude, using data from the experiments described earlier. The data on the right and left sides of Figure 8 come from Figures 5 and 7.Figure 8 suggests that multiple local peaks in sensitivity to time are observed in timing across several orders of magnitude.

Integration of Interval and Circadian Timing The summarized data suggest that the psychological representation of time is nonlinearly related to the interval being timed. The existence of a local maximum near a circadian oscillator (Figure 8, peak on right side) and local maxima in the short-interval range (Figure 8, peaks on left side) are consistent with timing based on multiple oscillators. According to multiple-oscillator proposals, each oscillator is a periodic process that cycles within a characteristic period. Each oscillator can be characterized by its period (i.e., cycle duration) and phase (i.e., current point with the cycle). Thus, each unit within a multiple oscillator system has its own period and phase. Sensitivity to time an interval near an oscillator is expected to be higher than timing an interval farther away from the oscillator because an oscillator functions to increase temporal sensitivity. Therefore, the multiple local peaks in sensitivity to time shown in Figure 8 suggest the existence of multiple short-period oscillators. The data reviewed in this section suggest that interval timing is based on an endogenous-oscillator, rather than a pacemakeraccumulator, mechanism according to the classic distinction discussed at the beginning of this article. The main findings are summarized as follows. The data in Figure 1 document that short-interval timing exhibits gradual phase adjustment, consistent with an oscillator mechanism. The data in Figures 2 and 3 suggest that short-interval timing is endogenous and self-sustaining, consistent with an oscillator mechanism. The data in Figures 4 and 5 document that many long, but noncircadian, intervals can be timed, and the data in Figure 6 suggest that long-interval timing is endogenous and self sustaining, consistent with an oscillator mechanism. The data in Figures 5, 7, and 8 show that both short-interval and circadian timing are characterized by local peaks in sensitivity to time.

Conclusion The data suggest continuity of mechanisms in short-interval, long-interval, and circadian-timing systems. The data reviewed in this article may prompt the development of a theory of timing that encompasses the discrimination of temporal intervals across several

Cognition j Time: What Animals Know

437

orders of magnitude, from milliseconds to days. Such a system is capable of representing when specific events occurred in time, as people describe when events occurred using calendar-date-time systems. This type of a system may underlie the temporal representation of episodic memory (i.e., the memory system that contains memories of unique events from one’s past).

See also: Cognition: Mental Time Travel: Can Animals Recall the Past and Plan for the Future?

Further Reading Aschoff, J. (Ed.), 1981. Handbook of Behavioral Neurobiology, vol. 4: Biological Rhythms. Plenum Press, New York. Boisvert, M.J., Sherry, D.F., 2006. Interval timing by an invertebrate, the bumble bee Bombus impatiens. Current Biology 16, 1636–1640. Church, R.M., Broadbent, H.A., 1990. Alternative representations of time, number, and rate. Cognition 37, 55–81. Crystal, J.D., 1999. Systematic nonlinearities in the perception of temporal intervals. Journal of Experimental Psychology: Animal Behavior Processes 25, 3–17. Crystal, J.D., 2001a. Circadian time perception. Journal of Experimental Psychology: Animal Behavior Processes 27, 68–78. Crystal, J.D., 2001b. Nonlinear time perception. Behavioural Processes 55, 35–49. Crystal, J.D., 2006a. Long-interval timing is based on a self sustaining endogenous oscillator. Behavioural Processes 72, 149–160. Crystal, J.D., 2006b. Time, place, and content. Comparative Cognition & Behavior Reviews 1, 53–76. Crystal, J.D., Baramidze, G.T., 2007. Endogenous oscillations in short-interval timing. Behavioural Processes 74, 152–158. Gallistel, C.R., 1990. The Organization of Learning. MIT Press, Cambridge, MA. Gibbon, J., Fairhurst, S., Goldberg, B., 1997. Cooperation, conflict and compromise between circadian and interval clocks in pigeons. In: Bradshaw, C.M., Szabadi, E. (Eds.), Time and Behaviour: Psychological and Neurobehavioural Analyses. Elsevier, New York, pp. 329–384. Henderson, J., Hurly, T.A., Bateson, M., Healy, S.D., 2006. Timing in free-living rufous hummingbirds, Selasphorus rufus. Current Biology 16, 512–515. Takahashi, J.S., Turek, F.W., Moore, R.Y. (Eds.), 2001. Handbook of Behavioral Neurobiology, vol. 12: Circadian Clocks. Plenum, New York. Wilkie, D.M., Carr, J.A.R., Siegenthaler, A., Lenger, B., Liu, M., Kwok, M., 1996. Field observations of time-place behaviour in scavenging birds. Behavioural Processes 38, 77–88. Zhou, W., Crystal, J.D., 2009. Evidence for remembering when events occurred in a rodent model of episodic memory. Proceedings of the National Academy of Sciences USA 106, 9525–9529.

COMMUNICATION: SIGNAL MODALITY Acoustical Signals – In Air and Water Jakob Christensen-Dalsgaard, University of Southern Denmark, Odense M, Denmark © 2019 Elsevier Ltd. All rights reserved.

List of Symbols c T f l u k a r r (rho) p p0 p1 v

Sound speed (air: ca. 340 m/s, water ca. 1500 m/s) Cycle time (s) Frequency (Hz) Wavelength (m) Angular frequency (radians/s)¼2pf Wave number (u/c) Radius (m) Distance (m) Density (kg/m3) Sound pressure (N/m2) Reference sound pressure (N/m2) Source level (N/m2) Particle velocity (m/s)

Abstract The chapter gives an introduction to acoustics and acoustical signals. It is argued that motion in air and water always produces sound, in addition to other types of acoustical signals. The chapter details sound emission, propagation and production by animals.

Keywords acoustics; dipole; far field; fluid wave; monopole; near field; sound transmission; sound wave

Acoustical Signals: Definition Acoustical signals very broadly defined are mechanical disturbances produced by the signaler. Some of these signals are produced as a part of well-defined communication behavior, loud and characteristic like the calls of crickets, frogs, birds and mammals; others are epiphenomena resulting from animals moving in a medium. It is impossible for moving animals not to produce acoustical signals; there is thus a wide spectrum of signals available, not only for communication, but also for interception of signals not intended for communication, and it is not surprising that all eumetazoan animals have mechanoreceptors that potentially can respond to acoustic signals. Some of these mechanoreceptors are incorporated in what we usually understand as ‘ears’ – tympanate structures, for example in insects and tetrapods. However, many other types of sensory organs respond to acoustical signals. These are, for example, otolith or statocyst organs also used as gravistatic receptors (Budelmann, 1988; Fritzsch and Straka, 2014), subgenual organs in the joints of arthropod legs, lateral line and neuromast organs in fishes and amphibians and even free standing sensory hairs found in many invertebrates. Usually, ‘Acoustical signals’ includes a diverse group of mechanical disturbances: Substrate vibrations, as well as fluid waves and sound waves. Since vibration signals are described in another article in this encyclopedia I will only focus on sound waves here. It should be noted, however, that a sound emitter always will generate other wave types than sound waves – for example fluid waves or substrate vibrations– that may also function in communication or be intercepted. Also, from the receiving animal’s point of view the distinction between types of acoustical signals may not be meaningful – the useful distinction will be of which types of receptors the signal excites and the subsequent neural processing. Conversely, parts of sound signals may not excite the receptors and so not have any function in sound communication. An example could be ultrasonic harmonics falling outside of the hearing range of the

438

Encyclopedia of Animal Behavior, 2nd edition, Volume 1

https://doi.org/10.1016/B978-0-12-809633-8.90736-2

Communication j Acoustical Signals – In Air and Water

439

animal. Thus, the sensitivity of the receiving animal defines which acoustical signals are important and the range of communication signals. However, I will not address the receiver part of acoustical signaling here; the reader is referred to the extensive literature on animal hearing, especially the Springer Handbook of Auditory Research series, for example (Hoy et al., 1998; Dooling et al., 2000; Fay and Popper, 1994; Narins et al., 2006). Rather, the aim of this article will be to introduce types of acoustical signals, their generation and transmission. The focus is on animal signal production and transmission, but some general properties of sound waves need to be introduced here. The article can only provide a very cursory overview of the large field of acoustics and bioacoustics, so I will recommend a few books for further study, all written for biologists: Fletcher (1992, 2007) has written excellent introductions to bioacoustics and the underlying physics, and Bradbury and Vehrencamp (2011) a very readable introduction to signal analysis and acoustics in the context of animal communication. The articles by Larsen and Wahlberg (Larsen and Wahlberg, 2016; Wahlberg and Larsen, 2017) on general acoustics and sound propagation are also good starting points, before consulting the general acoustics textbooks, such as Beranek (1983) or Kinsler et al. (1999).

The Nature of Sound – Wave Equation, Particle Motion and Pressure The sound wave is an elastic longitudinal wave propagating in a medium. This means that motion of the sound emitter generates alternating condensations and rarefactions of medium, and the resulting pressure gradient accelerates the medium particles in the propagation direction and produces a particle motion parallel and antiparallel to the propagation direction. The term ‘medium particle’ is not very precise, but medium particles are a ‘tiny bulk’ of the medium, small enough to have a well-defined location, and large enough to have macroscopic properties such as pressure and temperature. Because of medium elasticity and continuity the wave propagates in the medium with a characteristic velocity (c) that depends on the medium properties, i.e., on the ‘springiness’ of the medium. In air, c is around 340 m/s (depending on temperature), in water around 1500 m/s. The fundamental equation of mechanical waves is the wave equation that relates pressure (p) to time (t) and space, based on Newtons 2nd law (in Eulers formulation) and conservation of mass. In one dimension (x), the equation is: vvt2p ¼ c2 vvxp2 , where c is the propagation speed. A simple interpretation of the wave equation is that waves propagating in one dimension have the same shape as a function of time and as a function of space. In Fig. 1 this is shown by a common axis of time and location for a sinusoidal signal. The red bar shows the fundamental characteristic of a sine wave, which in space is the wavelength l (the distance between two wave maxima), in time the cycle time T. The inverse of T is frequency f, the number of cycles per second (more conveniently stated as the angular frequency u, the rate of change of phase in radians/s, where u¼2pf) and combining these leads to the equation c¼fl, which is very important for animal sound communication and sound reception. Especially the ratio of wavelength and size (measured by the radius a) a/ l, governs both sound production and sound emission. The ratio is usually stated as ka, the product of wave number k (¼2p/l) and radius, and the ratio can be understood as the acoustical size of an object. If ka is small (>1, as for large objects at high frequencies) the objects will reflect sound and be relatively efficient sound emitters (Larsen and Wahlberg, 2016). 2

2

Fig. 1 Sine wave; due to the wave equation wave functions have the same shape in time and space, as shown by the common x-axis. The bar shows the wavelength or cycle time.

440

Communication j Acoustical Signals – In Air and Water

Fig. 2 A single echolocation chirp of a pipistrelle bat (Pipistrellus pipistrellus). (a) Shows the time signal, (b) the amplitude spectrum of the chirp and (c) the spectrogram, using 128 point power spectra and 125 point overlap between consecutive spectra. The 128 point spectrum corresponds to a time slice of 0.48 ms, shown by the black bar below (c). All figures were generated in MATLAB. The call recording was provided by Annemarie Surlykke.

Sound as a Signal A sound signal is a waveform of pressure oscillations and can be characterized in terms of its duration, amplitude, envelope, and frequency content. In the Fig. 1 above, the signal is a pure tone, a sine wave, which allows a very clear demonstration of frequency and wavelength. However, animal sounds are never pure sine waves. An example is the bat echolocation call shown in Fig. 2(a). It is clear that the waveform is not a simple sinusoid. Fortunately, any waveform can be represented as a sum of sine waves of different frequency by a mathematical operation known as Fourier analysis (see Bradbury and Vehrencamp, 2011). The amplitudes, phases and frequencies of the sine waves constitute the frequency spectrum of the sound (Fig. 2(b)), in this case showing that the bat echolocation call covers the frequency range from 20 to 50 kHz, with maximal amplitude at 50 kHz. However, whereas the spectrum can show overall frequency content of sounds, it is not ideal for analysis of communication sounds, since most communication sounds show fast variations that are not captured by the spectrum. Therefore, animal sounds are usually displayed as spectrograms or sonograms, which are sequences of amplitude spectra. Each of these spectra (displayed as vertical lines) is made from a brief time slice moved along the original signal (the black bar below Fig. 2(c)), and the amplitudes are shown as a color scale. In the bat call, the spectrogram (Fig. 2(c)) shows that the frequencies change systematically during the call, creating a downward frequency-modulated sweep with a harmonic component, which is not apparent from the time signal or the amplitude spectrum. In general, different call patterns can be identified much more easily from spectrograms than from time signals or amplitude spectra. A word of caution, though: Spectrogram construction entails setting a number of parameters that greatly determines the appearance of the signals displayed, for example the size of the time slice used for spectra and the overlap of consecutive time slices. A spectrogram is impossible to interpret properly, if these parameters are not reported. For further reading on signal analysis, see Bradbury and Vehrencamp (2011).

The dB Scale Sound pressures are usually stated in the dB scale. This scale is logarithmic and based on a ratio, usually between the actual sound pressure p and a reference sound pressure p0, which in air is 20 mPa, and in water usually is 1 mPa: pdB ¼ 20log10 pp0 . Thus, in air a sound pressure at 0 dB is 20 mPa; a sound pressure at 20 dB is 10 times higher than the reference¼200 mPa. The reference value in air is chosen because it corresponds to the average human threshold at 1000 Hz. dB values are also used without a reference and are then interpreted as ratios between two values. For example, a 6 dB reduction in sound pressure is a halving of pressure amplitude.

Communication j Acoustical Signals – In Air and Water

441

The Sound Field: Monopoles, Dipoles and Pistons Different types of movement patterns of objects in air or water generate characteristic sound emission patterns or sound fields. Of these movement patterns, the simplest is the so-called acoustical monopole, a pulsating sphere. The monopole produces two kinds of spherical waves concurrently. One is the sound wave propagating in the medium. The other type of wave, which we could call the fluid wave, is created by displacement of the medium by the motion of the monopole. This wave is not a sound wave and does not propagate out from the source. Following the derivation by Kalmijn (1988) these properties follow from the analytical expression of the particle velocity produced at distance r by a monopole, in this case having a simple sinusoidal oscillation with velocity amplitude V0: ka2 a2 velocity v ¼  V0 sinðut  krÞþ 2 V0 cosðut  krÞ r r and pressure p ¼  rcka r V0 sinðut  krÞ, where r is the medium density From these expressions, it is clear 2

(1) That the expression for medium particle velocity has two components that are 90 out of phase (the sine and cosine terms). We can call the two components vs and vf, since they are associated with the active, propagating sound wave and the reactive, non-propagating fluid wave, respectively. (2) That vs and vf attenuates differently by distance (the r and r2 terms), so the fluid wave is more ‘local’, and at far distances vs will dominate. (3) That pressure only has one term, in phase with vs, and p¼rc* vs (Ohms acoustical law). (4) That the sound emission depends on size and frequency (the ka2 term in vs and p, reducing sound emission by small monopoles at low frequencies), whereas fluid wave amplitude depends on size only. (5) That both sound emission and fluid flow are oriented radially from the center of the monopole. From the equation for pressure follows the distance law, p(r)¼p1/r, where p1, the source level, is the sound pressure at 1 m distance. In the dB scale, the distance law is stated as pdB ðrÞ ¼ 20log10 pr1 (Note that the dB value here is relative to the source level). Since doubling the distance r results in a dB-value of approximately 6, the distance law is often stated as ‘6 dB/distance doubled’. The distance law is very important for sound communication, since it defines a communication range, depending on the source level and the sound sensitivity of the animal. Therefore, sound recordings where the source level is not known (because the distance to the signaling animals is not measured) are not very useful in animal communication research. There are exceptions to the distance law, mainly due to environmental effects, that I will outline below. I have described the behavior of the monopole in some detail, since this simplest type of sound emitter illustrates very general properties of sound signaling. Naturally, most sound emitters generate more complicated patterns. However, they can be understood as aggregates of monopoles; furthermore, at far distances (many wavelengths) sound sources generally behave as monopoles; in particular, far away, in the ‘far field’ (see below) Ohms acoustical law p¼rcvs holds for all sound emitters. Thus, the acoustical dipole, a sphere oscillating in a medium (Fig. 3), is equivalent to two closely spaced monopoles pulsating in antiphase. The radiation pattern is more complicated here, since sound does not radiate in directions perpendicular to the

Fig. 3 Sound and hydrodynamic flow generation by an acoustic monopole (pulsating sphere). The monopole expands and contracts with an amplitude shown by the blue arrow. The red arrows show the radial sound particle motion and the open arrows the hydrodynamic particle motion.

442

Communication j Acoustical Signals – In Air and Water

Fig. 4 Sound and hydrodynamic flow generation by an acoustic dipole (oscillating sphere). The two circles show the extreme positions of the sphere and the blue arrow the motion direction. The sound radiation (red arrows) has a figure-eight pattern with no radiation perpendicular to the direction of motion. The hydrodynamic flow (open arrows) is not aligned with the sound radiation, but flows around the dipole.

motion axis of the dipole, due to destructive interference of the two monopoles. Also, the concurrent fluid wave is not aligned with the sound propagation direction, but rather flows around the dipole. The sound emission also depends strongly on ka in acoustical dipoles. At low frequencies, small objects will be inefficient sound emitters due to the long wavelengths and short distances between the two monopoles, producing acoustical ‘short circuiting’, i.e., equalization of sound pressures. Objects showing approximately linear, bidirectional motion (loudspeaker membranes, insect wings) are to good approximation acoustical dipoles. Higher-order sound emitters can be produced by combinations of monopoles or dipoles, for example quadrupoles (two dipoles in alternating phase), that will correspond to sound emitted by objects rotating in the medium. These higher-order emitters will naturally have more complicated sound radiation and fluid wave patterns following from the combination of several monopoles. Another type of emitter, especially important for high-frequency emission, is the moving piston or a membrane mounted in a wall. The piston can produce a very directional beam pattern (Fig. 4) and is a good approximation of sound emission in for example bats (Jakobsen et al., 2013) and odontocetes. The moving piston, especially at high ka, produces a very directional sound emission, due to interference (the piston can be regarded as a continuum of sound emitters). At lower frequencies the emitter will be similar to a dipole, but because the wall limits acoustic short-circuiting sound emission efficiency will be increased (Fig. 5).

Near and Far Fields The concurrent emission of sound and fluid waves has led to some confusion in the literature (reviewed in Rogers and Cox (1988)). It is clear that when animal or microphone are close to the sound emitter the particle motion contribution from the fluid wave (vf ) is non-negligible. The proportionality factor between vs and vh amplitudes is vvfs ¼ kr for the monopole, i.e., l the two components have equal amplitudes; this distance will increasing with distance and frequency. At a distance of r ¼ 2p be four times longer in water than in air due to the longer wavelengths. This distance is often taken as a boundary between

Fig. 5 Sound radiation by a piston (a vibrating disk in a baffle, ka¼8). The multilobed radiation pattern is compared to an omnidirectional sound radiator emitting sound at the same intensity (blue curve). The directivity index is the relative increase in sound level at the best direction compared to the omnidirectional emitter. For a piston, the directivity index is 20logka¼18 dB in this example. It follows that the directivity is strongly frequency-dependent. Figure kindly provided by Lasse Jakobsen, redrawn Jakobsen, L., Brinkløv, S., Surlykke, A., 2013. Intensity and directionality of bat echolocation signals. Frontiers in Physiology 4 (89), doi: 10.3389/fphys.2013.00089.

Communication j Acoustical Signals – In Air and Water

443

the ‘near field’ and the ‘far field’, but it is clear that this boundary is arbitrary. Also, for more complicated sound emitters than the monopole the simple relationship between vs and vh does not hold. Here, sound emission will be more directional, and there can be directions where fluid waves dominate even at quite large (theoretically infinite) distances. The bottom line is that if the animal under study is sensitive to particle motion and communication distances are small (less than a wavelength) it is recommendable to measure the particle velocity components directly. Undoubtedly, much close-range acoustic communication is mediated by fluid waves (e.g., wingbeats, tail strokes) rather than sound, where motion detectors such as lateral lines in fish or antennal mechanoreceptors or sensory hairs in insects are the appropriate receptors. Conversely, when the sound sensitivity of an animal is studied in order to understand its long-range communication, stimulating the animal at too close ranges may overestimate its sound sensitivity and hence its communication range, because its auditory system is responding to the particle motion component of the hydrodynamic fluid wave. The terms ‘near-’ and ‘far field’ are used for a completely different phenomenon than the ‘fluid wave’ near field, unfortunately. This derives from the so-called Fresnel effect, where a sound source at large ka is seen as a ‘continuum of monopoles’, separate sound emitters, whose sound emission at very close ranges interfere, producing local minima ((see Larsen and Wahlberg, 2016) for a detailed description). For a circular piston sound emitter, which is a good model of most high-frequency emitters, the distance 2 limit of this interference field will be approximately r > 2pa l (Larsen and Wahlberg, 2016). In the Fresnel near field close to the emitter, sound level is not a monotonic function of distance (so the distance law does not hold). Thus, it is advisable to be in the Fresnel far field defined by the distance limit above. Note that the Fresnel near field and the fluid wave near field concern two opposite frequency ranges due to the dependence on wavelength: The fluid wave near field is mainly a low-frequency, the Fresnel near field a high-frequency phenomenon. However, in both cases the near fields extend further in water than in air, due to the longer wavelengths in water.

Environmental Effects Factors in the environment can change the propagation pattern of sound and produce excess attenuation in addition to the attenuation produced by spherical spreading. The main causes of excess attenuation are temperature gradients, salinity gradients (in water) reflecting or scattering objects, humidity and wind. Temperature and salinity gradients both produce a gradient in sound speed. For example, at nighttime the ground temperature will often be lower than the temperature in the air above. The resulting gradient will refract sound downwards, and an emitter on the ground will be able to communicate with receivers far away, but transmission at close ranges will be impeded. In contrast, if ground temperature is higher than in the air above, the temperature gradient will refract sound upwards, creating a shadow zone near the ground where very little sound energy is propagated. However, an emitter on the ground can be heard by receivers on elevated posts. Wind can have similar effects as temperature gradients, so transmission upwind will be refracted upwards, downwind downwards (Wahlberg and Larsen, 2017). In water, temperatures in open water usually decrease from the surface down. However, as pressure increases with depth, sound velocity increases. Thus, a sound velocity minimum is created at intermediate depths. This velocity profile will create a lens-type refraction, where sound energy will be concentrated in a channel (so-called SOFAR channel, (Rogers and Cox, 1988) ) and emitters placed in this channel will be able to transmit over long distances (several thousand kilometers). Salinity gradients can also result in refraction of sound in water.

Reflecting and Scattering Objects The main reflectors in the environment are the ground surface (for sound in air) and the bottom and (especially) the water surface for sound in water. These reflections are strong because of the very large ka ratios and differences in sound propagation in air, soil and water, which depend on the specific impedances, rc (the product of density and sound speed) in the different media. Generally, reflections are strong, when the difference in specific impedance is large. Thus, the received signal from an emission close to the ground will consist of the sound wave following the shortest (i.e., the direct) path and a strong reflection of the sound wave from the ground, the so-called ground effect. The phase difference between these sound components will lead to constructive and destructive interference with excess attenuation of some frequency components (see Wahlberg and Larsen, 2017) that depends on the height of emitter and receiver above ground. In water, strong reflections especially from the water surface will create similar interference patterns. A special and important environment is shallow water, water bodies where the water depth is equal to or smaller than the wavelength of sound and comprising almost all freshwater bodies and coastal waters. Shallow water has very special acoustical properties, due to the repeated reflections from surface and bottom, and very complicated acoustics (discussed in Rogers and Cox (1988); Larsen and Radford, 2018). Most importantly, the medium has a cutoff-frequency below which sound does not propagate, or rather propagates with a very large excess attenuation. The cutoff frequency depends on water depth and sound velocity in the bottom layers. For example, in 1 m deep shallow water with a sandy bottom, the cutoff frequency is close to 1 kHz (see formula in Rogers and Cox, 1988). At frequencies below 1 kHz in this environment the excess attenuation will be large, producing a steep gradient, so these frequencies are only usable for short range communication. However, for frequencies above the cutoff frequency sound propagates as in a duct with only 3dB attenuation/ distance doubled, or p ¼ pp1ffir . Therefore, high-frequency communication is advantageous here.

444

Communication j Acoustical Signals – In Air and Water

Depending on their ka ratio, objects in the sound field will interact in different ways with the propagating sound wave. Objects with high ka ratios (>>1) will be good reflectors, so most objects will reflect ultrasound, for example. Objects with intermediate ka ratios will scatter sound, i.e., distort the sound field. Thus, in a forest ultrasound will not propagate very far, and the best strategy for long-distance propagation is to make low-frequency sounds. Also, the sound received will be composed of the primary sound wave and reflected sound components arriving later, because their propagation path is longer. The sound received at any distance will therefore consist of the direct (shortest path) component followed by a tail of reverberations. If the reflectors are closely spaced and close to the sound emitter, as in a forest, the reflected components can interact with the direct component, producing frequency-dependent constructive and destructive interference. Therefore, both the frequency content and the temporal structure of signals will be changed, the more the farther the signal has propagated. Since these effects depend on habitat composition (which may change seasonally), as well as on temperature, humidity, and position of the sender, it is really difficult to predict the environmental effects on sound propagation, and it is recommended to measure the sound degradation in the habitat (by controlled sound emission and recording). Note, however, that any natural habitat is likely not homogenous, so sound propagation may depend on direction, and this ‘calibration’ of the habitat may be time-consuming. See Wahlberg and Larsen (2017) for further discussion of environmental effects. The effect of humidity is to produce excess attenuation (on top of the 1/r attenuation from spherical spreading) by absorption of sound energy. This absorption is strongly frequency dependent, as well as dependent on the percent humidity and temperature. As an example, model calculations (at the homepage provided in “Relevant Websites section” of the National Physics Laboratory, UK) show that at 500 Hz, 50 % humidity at 20 C, the absorption is 0.003 dB/m. At 5000 Hz, absorption has increased to 0.04 dB/m, but at 50,000 Hz, absorption is 1.7 dB/m, producing an excess attenuation of 85 dB at a 50 m distance, on top of the 6 dB/distance doubled. This severely limits the range of ultrasonic frequencies, particularly at high humidity (for example at nighttime in the tropical rain forest).

Noise Another very important aspect of sound communication is noise. Environmental noise can be biotic (sounds produced by other animals) or abiotic (for example wind, rain, running water), and, more recently anthropogenic; in all cases noise can be masking sound signals and decreasing communication ranges. The impact of noise depends on its frequency range, amplitude, and temporal structure, as well as on the auditory processing of the animals. All animals have to deal with noise and are using several different strategies. The simplest probably is to move away from noise sources, another is to increase signal amplitude (the Lombard effect, described for several animal groups (Hotchkin and Parks, 2013)), and a third strategy is to use frequency bands or time intervals that are relatively noise-free. For example, an assembly of frog and insect species will usually partition the frequency space with relatively small overlap, and frogs near waterfalls may even communicate at ultrasonic frequencies to avoid masking by environmental noise (or, in other cases, rely on visual instead of acoustic communication). See Larsen and Radford, (2018) for a full treatment of the subject.

Animals and Signals Animal Sound Emitters Animals make sound by drumming, scraping or buckling structures, by wing beats or by blowing air across membranes, and the sound producing organs are often acoustically coupled to resonant structures (Bennet-Clark, 1999). Most sound-producing insects, for example orthopterans (locusts, crickets and bushcrickets) stridulate, i.e., they scrape a plectrum along a comb-like structure (file) (Fig. 6(B)). Each tooth in the file produces a transient vibration upon scraping, and the frequency output will depend on the speed of scraping and the spacing of the teeth. Crickets and bush crickets stridulate by moving the wings, and the stridulation apparatus is coupled to highly resonant structures, in some cases producing almost pure tones (Bennet-Clark, 1999; Montealegre-Z et al., 2011). Stridulation is an efficient way to generate a high-frequency output from a relatively slow movement (of the scraper). Another efficient way is to buckle a membrane. The deformation of a stiff, cuticular membrane, for example in the tymbal organ of cicadas, produces an intense transient (Bennet-Clark, 1975) (Fig. 6(A)). Sound producing fishes like, e.g., the toadfish or midshipman are closest to the ‘idealized’ sound emitters described above, since they drum or scrape the air-filled swim bladder, which will radiate sound almost like an ideal monopole. The frequency content of sound emitted will depend on the resonance frequency of the swim bladder, its structure and the movements used to excite it. The swim bladder was traditionally described as a resonator, but it is highly damped, and the emission frequency appears to be controlled by extrinsic structures – drumming muscles in some species (toadfish) or movement of skeletal structures (for example in clownfish). The drumming muscles used by toadfish and others are so-called superfast muscles that are capable of contracting at very high rates (>100 Hz)(Parmentier and Fine, 2016). Most terrestrial vertebrates produce communication sounds by forcing air over a larynx and exciting the vocal cords (mammals, anurans, lizards) or forcing air through the syrinx and exciting the medial and lateral labia (birds)(Fig. 7). The frequency output of these structures can potentially be controlled by air pressure and tension of the vocal cords or labia. Anurans cannot control tension of their vocal cords, but birds and mammals have muscular control of the tension (and thus the frequency output) of their sound producing membranes. The sound production mechanism in these different structures is very similar, at least in birds and

Communication j Acoustical Signals – In Air and Water

445

Fig. 6 Insect sound radiators. A: (a) The cicada in transverse section with the powerful muscles attached to the tymbal organ shown in red. (b) and (c) shows the tymbal organ before (b) and after deformation (c). Drawings from Encyclopedia Britannica, 1911, redrawn. The call shown is from the cicada Okanagana canadensis, recording on the SINA pages maintained by Dr. Tom Walker, University of Florida. B. An example of a stridulatory organ, from the acridid grasshopper Pseudochorthippus parallelus showing the file located on the femur (the scraper (not shown) is a raised vein on the elytra). Drawing from Darwin 1871, p. 357, redrawn. Recording from Orthoptera.speciesfile.org, under the Creative Commons Licence.

mammals, producing a membrane motion wave that travels from one end of the membrane to the other (Elemans et al., 2015). The main shaping of the sound is in resonant chambers, lungs, larynx and the vocal tract (Herbst, 2016).

Animal Sound Radiation The propagating sound wave radiates energy from the sound source. The intensity (energy/(time*area)) is simply pvs¼p2/(rc), so the intensity of radiated sound at 1 Pa in air is around 3 mW; it is much less in water, 0.6 mW. For a small animal the expenditure of 3 mW is not negligible, especially since the efficiency (radiated intensity/expended power) for most animals is low, around 1%–10%. Clearly, the efficiency depends on ka and is low at low ka; for example, it is 1%–5% in crickets with a ka close to 1 (Bennet-Clark, 1975), and generally, in insects, anurans and birds efficiency does not exceed a few percent (Ophir et al., 2010). The directionality of sound emission of animals depends on properties of the air spaces and body shape, but the most important parameter in sound radiation is ka, the relation between size and frequency. For example, the spring peeper frog (Pseudacris crucifer), length 30 mm, calling at 2.8 kHz has a ka of 0.77 and is therefore an omnidirectional sound radiator (acoustic monopole), as also shown by direct measurements (Gerhardt, 1975). Conversely, the bat Myotis daubentonii, emitting 50 kHz echolocation calls through a mouth opening of 5 mm has a ka of 2.25 and a strongly directional sound emission that can be described by the piston model above (Fig. 4), and the ka’s of odontocetes is even larger, leading to very narrow emission beams (Jakobsen et al., 2013). Birds calling at 1–2 kHz have ka’s around 1 and will have a largely omnidirectional sound emission. However, the directionality of omnidirectional radiators can be modified by structures such as walls, either natural or constructed by the animal to improve sound radiation and directionality, as for example shown for tree crickets that construct a baffle by cutting a hole in leaves (Mhatre et al., 2017) and placing itself in the hole. This baffle prevents acoustical short circuiting, essentially increasing the acoustic size of the animal. Other examples are construction of burrows, for example the horn-shaped burrow constructed by mole crickets (Bennet-Clark, 1999) or resonant burrows constructed by frogs (Penna, 2004).

446

Communication j Acoustical Signals – In Air and Water

Fig. 7 Vertebrate sound radiators and sound signals. The figure shows a diagram of sound producing structures (black) and examples of calls in three terrestrial vertebrates: (A) Treefrog (Litoria infrafrenata), (B) canary (Serinus canaria) and C gibbon (Hylobates muelleri). (A) shows a transverse section of the frog larynx with the vocal chords (black) attached to the cricoid cartilages (C) and air flow from the trachea (T) through the glottal slit (G). Muscles (red arrows) are attached to the cricoid cartilages. B shows a simplified diagram of the bird syrinx. Air flows from the two bronchi (B) across the median and lateral labia (black) to the trachea (T). The position of the labia is controlled (indirectly) by syringeal muscles (red arrow). (C) shows the mammalian larynx with air flowing across the vocal chords (black) from the trachea to the epiglottis (EG). The tension of the vocal chords is controlled by musculature (red arrow). Animal sketches: A and B originals, C under Creative Commons License. The sound recording of the canary was provided by Dr. Michiel Vellema, the gibbon recording was downloaded from the homepage of Dr. Thomas Geissman (http://www. gibbons.de/main).

Structure of Animal Signaling The important parameters in animal signals are frequency content (fundamental frequency and harmonics), intensity, duration, pulse rate, and pulse shape. In addition, animals may control large-scale temporal structure by combining call elements sequentially. The fundamental frequency of animal sound emitters depend on size of the animal, but in addition, animals can vary the harmonic content by controlling the vibrating structures and modifying resonant structures. Animals have different degrees of control over their vocal apparatus, as outlined above, and very different sizes of their vocal repertoires. For example, most insects and frogs have very repetitive calls with one or two different call types, whereas the repertoire in birds and mammals can be very large, with more than 20 different calls. A word of caution here: Unless the communication behavior is studied any classification of different call types is tendentious. There is sometimes a tendency to identify anything that produces a visually complex spectrogram as a ‘complex’ signal. However, it is evident that lack of control due to an unspecialized call production mechanism (analogous to rubbing two stones together) can produce a noise-like, seemingly complex signal, whereas the production of a ‘simple’ sinusoid demands a highly tuned ‘instrument’. Also, driving a sound production mechanism to its physical limits (‘screaming on top of your voice’) will produce non-linearities and a seemingly more complex sound. Without behavioral evidence it is not certain, however, that this complexity is not just an epiphenomenon caused by attempt to maximize sound output to the animals’ physical limits.

Size and Frequency For animal sound production the general relationship between size and frequency, as described for the ideal monopole, also holds. Simple scaling of vocal organs based on allometry would suggest that the lowest emission frequency that animals can produce

Communication j Acoustical Signals – In Air and Water

447

efficiently was inversely related to linear size, i.e., f1a, which corresponds to f fM0:333 , where M is body mass. However, Fletcher (2004) has argued, based on optimization of communication range, that f fM0:4 (Fig. 4) and shown that the available data for insects, birds and mammals can be described reasonably well by such a power function. Due to this relationship between size and frequency, a listening animal can obtain cues to the size of the caller. This is also a cue that is difficult to ‘fake’, except by producing larger vocal structures. The enlargement of the larynx in many mammals may be caused by sexual selection for low-frequency vocalization (Charlton and Reby, 2016), and sexual selection for males with low-frequency calls have been described in mammals (primates: (Puts et al., 2016)), frogs (Gerhardt, 1994) and insects (Balakrishnan, 2016), but in all groups there are also exceptions, where low-frequency calls are not optimal. Also, generally in mammals selection is not for the lowest fundamental frequency (produced by larger vocal chords)(Garcia et al., 2014), but for the largest spacing of harmonics (produced by an enlarged vocal tract and/or other air spaces communicating with the vocal tract) (Charlton and Reby, 2016)

Conclusion Acoustic communication is the most far-reaching type of signaling in animals (extending to thousands of km in whales) and studying it has never been easier – with the revolution in digital sound and video recording equipment and analysis software. The aim of this article has been to give a brief outline of physical parameters related to sound signaling, since understanding how acoustical signals are generated and propagated is an essential part of understanding communication behavior and its evolution.

Acknowledgements I thank Ole Næsbye Larsen and Michael Greenfield for comments on the manuscript.

See also: Communication: Vibrational Signals: Sounds Transmitted Through Solids. Methodology: The Use of Playbacks in Behavioral Experiments. Neurons and Senses: Acoustic Communication in Insects: Neuroethology; Bat Neuroethology; Ears and hearing in vertebrates; Neuroethology of Sound Localization in Birds; Vocal–Acoustic Communication in Fishes: Neuroethology.

References Balakrishnan, R., 2016. Behavioral ecology of insect acoustic communication. In: Pollack, G.S., Mason, A.C., Popper, A.N., Fay, R.R. (Eds.), Insect Hearing. Springer International Publishing, Champaign, pp. 49–80. Bennet-Clark, H.C., 1975. Sound production in insects. Science Progress 62 (246), 263–283. Bennet-Clark, H.C., 1999. Resonators in insect sound production: How insects produce loud pure-tone songs. Journal of Experimental Biology 202 (23), 3347–3357. Beranek, L.L., 1983. Acoustics. Acoustical Society of America, New York. Bradbury, J., Vehrencamp, S., 2011. Principles of Animal Communication, second ed. Sinauer Associates, pp. 19–63. Budelmann, B.-U., 1988. Morphological diversity of equilibrium receptor systems in aquatic invertebrates. In: Atema, J., Fay, R.R., Popper, A.N., Tavolga, W.N. (Eds.), Sensory Biology of Aquatic Animals. Springer, New York, pp. 757–782. Charlton, B.D., Reby, D., 2016. The evolution of acoustic size exaggeration in terrestrial mammals. Nature Communications 7, 12739. https://doi.org/10.1038/ncomms12739. Dooling, R.J., Fay, R.R., Popper, A.N., 2000. Comparative Hearing: Birds and Reptiles. Springer Handbook of Auditory Research, vol. 13. Springer, New York. Elemans, C.P.H., Rasmussen, J.H., Herbst, C.T., et al., 2015. Universal mechanisms of sound production and control in birds and mammals. Nature Communications 6, 12739. https://doi.org/10.1038/ncomms9978. Fay, R.R., Popper, A.N., 1994. Comparative Hearing: Mammals. Springer Handbook of Auditory Research, vol. 4. Springer, New York. Fletcher, N., 1992. Acoustical Systems in Biology. Oxford University Press. Fletcher, N., 2007. Animal bioacoustics. In: Rossing, T.D. (Ed.), Springer Handbook of Acoustics. Springer, New York, pp. 785–804. Fletcher, N.H., 2004. A simple frequency-scaling rule for animal communication. Journal of the Acoustical Society of America 115 (5), 2334. https://doi.org/10.1121/1.1694997. Fritzsch, B., Straka, H., 2014. Evolution of vertebrate mechanosensory hair cells and inner ears: Toward identifying stimuli that select mutation driven altered morphologies. Journal of Comparative Physiology A 200 (1), 5–18. https://doi.org/10.1007/s00359-013-0865-z. Garcia, M., Charlton, B.D., Wyman, M.T., Fitch, W.T., Reby, D., 2014. Do red deer stags (Cervus elaphus) use roar fundamental frequency (F0) to assess rivals? PLOS ONE 8 (12), e83946. https://doi.org/10.1371/journal.pone.0083946. Gerhardt, H.C., 1975. Sound pressure levels and radiation patterns of the vocalizations of some North American frogs and toads. Journal of Comparative Physiology 102 (1), 1–12. Gerhardt, H.C., 1994. The evolution of vocalization in frogs and toads. Annual Review of Ecology and Systematics 25, 293–324. Herbst, C.T., 2016. Biophysics of vocal production in mammals. In: Suthers, R.A., Fitch, W.T., Fay, R.R., Popper, A.N. (Eds.), Vertebrate Sound Production and Acoustic Communication. Springer International Publishing, Champaign, pp. 159–189. Hotchkin, C., Parks, S., 2013. The Lombard effect and other noise-induced vocal modifications: Insight from mammalian communication systems. Biological Reviews 88 (4), 809–824. Hoy, R.R., Popper, A.N, Fay, R.R., 1998. Comparative Hearing: Insects. Springer Handbook of Auditory Research, vol. 10. Springer, New York. Jakobsen, L., Brinkløv, S., Surlykke, A., 2013. Intensity and directionality of bat echolocation signals. Frontiers in Physiology 4 (89). https://doi.org/10.3389/fphys.2013.00089. Kalmijn, A.J., 1988. Hydrodynamic and acoustic field detection. In: Atema, J., Fay, R.R., Popper, A.N., Tavolga, W.N. (Eds.), Sensory Biology of Aquatic Animals. Springer, New York, pp. 83–130. Kinsler, L.E., Frey, A.R., Coppens, A.B., Sanders, J.V., 1999. Fundamentals of Acoustics, fourth ed. John Wiley and Sons, New York, NY. Larsen, O.N., Radford, C., 2018. Acoustic conditions affecting sound communication in air and underwater. In: Slabbekoorn, H., Dooling, R., Popper, A.N., Fay, R.R. (Eds.), Effects of Anthropogenic Noise on Animals, Springer Handbook of Auditory Research. Springer International Publishing, Champaign, 109-144.

448

Communication j Acoustical Signals – In Air and Water

Larsen, O.N., Wahlberg, M., 2016. Sound and sound sources. In: Brown, C., Riede, T. (Eds.), Comparative Bioacoustics: An Overview. Bentham Science Publishers, pp. 3–61. Mhatre, N., Malkin, R., Deb, R., Balakrishnan, R., Robert, D., 2017. Tree crickets optimize the acoustics of baffles to exaggerate their mate-attraction signal. eLife 6, e32763. https://doi.org/10.7554/eLife.32763. Montealegre-Z, F., Jonsson, T., Robert, D., 2011. Sound radiation and wing mechanics in stridulating field crickets (Orthoptera: Gryllidae). The Journal of Experimental Biology 214 (12), 2105–2117. https://doi.org/10.1242/jeb.056283. Narins, P.M., Feng, A.S., Fay, R.R., Popper, A.N., 2006. Hearing and Sound Communication in Amphibians. Springer Handbook of Auditory Research, vol. 28. Springer, New York. Ophir, A.G., Schrader, S.B., Gillooly, J.F., 2010. Energetic cost of calling: General constraints and species-specific differences. Journal of Evolutionary Biology 23 (7), 1564–1569. Parmentier, E., Fine, M.L., 2016. Fish sound production: Insights. In: Suthers, R.A., Fitch, W.T., Fay, R.R., Popper, A.N. (Eds.), Vertebrate Sound Production and Acoustic Communication. Springer International Publishing, Champaign, Illinois, pp. 19–49. Penna, M., 2004. Amplification and spectral shifts of vocalizations inside burrows of the frog Eupsophus calcaratus (Leptodactylidae). Journal of the Acoustical Society of America 116 (2), 1254–1260. https://doi.org/10.1121/1.1768257. Puts, D.A., Hill, A.K., Bailey, D.H., et al., 2016. Sexual selection on male vocal fundamental frequency in humans and other anthropoids. Proceedings of the Royal Society B: Biological Sciences 283 (1829). https://doi.org/10.1098/rspb.2015.2830. Rogers, P.H., Cox, M., 1988. Underwater Sound as a Biological Stimulus Sensory Biology of Aquatic Animals. Springer New York, New York, NY, pp. 131–149. Wahlberg, M., Larsen, O.N., 2017. Propagation of sound. In: Brown, C., Riede, T. (Eds.), Comparative Bioacoustics: An Overview. Bentham Science Publishers, pp. 62–119.

Relevant Website http://resource.npl.co.uk/acoustics/techguides/absorption/–NPL Acoustics: Calculation of absorption of sound by the atmosphere.

Bioluminescent Signals Kathrin F Stanger-Hall, University of Georgia, Athens, GA, United States Todd H Oakley, University of California, Santa Barbara, CA, United States © 2019 Elsevier Ltd. All rights reserved.

Mysterious glows or flashes have fascinated human observers for over 3000 years and appear in the folklore of Europe, Asia, Africa, and the Americas, with the firefly as one of the most common insects in Mayan art (Roda, 2011). Whether in the depth of the ocean, in crashing waves at the beach, or as aerial light displays during mid-summer nights, what we are witnessing is nothing less than a quest for survival and reproduction by an astonishing diversity of organisms. Bioluminescence, the conversion of chemical energy into light by living organisms, is a widespread phenomenon across the tree of life. It may have evolved at least 40 times in a wide range of taxonomic groups, including bacteria, dinoflagellates, fungi, jellyfish, squid, crustaceans, snails, annelids, insects, and vertebrates (Haddock et al., 2010), with more than seven unique biochemical systems (Hastings, 1983; Petushkov and Rodionova, 2005; Viviani et al., 2002; Fallon et al., 2018) adopted for light production. In this article we introduce the diversity of bioluminescent signals in the context of animal communication, with a focus on the structure, function and evolution of bioluminescent signaling behavior.

Bioluminescence The Chemical Diversity of Bioluminescence Bioluminescence is generated by oxidation reactions, but the specific chemistry varies greatly among organisms (Hastings, 1983). Bioluminescence requires two principal components: a light emitter that is oxidized (generically called luciferins, although specific structures vary) and a catalyzing enzyme (generically called luciferases, although not members of one protein family). Some marine organisms use photoproteins, associated with specific triggers, instead, e.g., the photoproteins of cnidarians, ctenophores, and radiolarians require Ca2þ to trigger their luminescence (Shimomura, 1985). Luciferins are conserved across some light-emitting organisms. For example, all beetles share an identical beetle luciferin (Day et al., 2004), while another luciferin, coelenterazine, is the light emitter for nine different marine phyla (Haddock et al., 2010). Depending on the group, luciferins can be dietary, or are biochemically produced (Haddock et al., 2001). In contrast, luciferases are considerably more diverse and species–specific. Their high sequence variability is associated with differences in light emission colors among beetles (Viviani, 2002). The molecular biology of different luciferases, and especially the amino acids causing shifts in light-color of firefly luciferases, are exceptionally well characterized due to their importance in biotechnology and biomedical reporter systems (e.g., Fraga, 2008; Welsh and Kay, 2005).

Intrinsic Versus Symbiotic Bioluminescence Many bioluminescent organisms have evolved their own intrinsic bioluminescent chemistry, while others derive their bioluminescence from symbiotic relationships with bioluminescent bacteria (Haddock et al., 2010), which requires the evolution of specialized structures to permanently house the bacteria and intricate interactions between hosts and symbionts (e.g., McFall-Ngai, 2014, for potential bacterial benefits see Stabb, 2005). Symbiotic relationships with bioluminescent bacteria contribute greatly to the diversity of bioluminescent signaling in the ocean. For example, a recent study determined that bioluminescence in 1500 species of fish evolved independently at least 27 times (Davis et al., 2016). In 17 of these instances bioluminescent bacteria were taken up as symbionts to generate bioluminescence. There is a remarkable diversity of bioluminescent chemistry within genera and within species (Widder, 2010; Takenaka et al., 2012). Even more intriguing, there can be multiple chemistries within individuals. For example, adult females of the deep-sea anglerfish Linophryne coronata have two different light-emitting systems: bacterial luminescence in the dorsal lure, and an intrinsic, unidentified chemistry in the chin barbel (Widder, 2010).

Bioluminescent Body Parts Versus Secretions Adding yet another layer of complexity to the study of bioluminescent signals, luminescent reactions can take place in secretions outside the body, or in specialized cells (photocytes) or “light organs” (photophores) of varying morphological complexity, including some with added light reflectors and lenses to focus the direction of bioluminescence (Herring, 2000). In some instances (e.g., bobtail squid), the light production of bacterial symbionts is monitored by a visual pigment in the light organ, which functions as an “inner eye” that can produce bioluminescence as well as perceive it, and could enable the host to impose selection for bioluminescence on its symbiotic bacteria (McFall-Ngai, 2014). Interestingly, visual pigments are also expressed in the light organs of some fireflies (Fallon et al., 2018) with unknown function. Organisms that generate bioluminescence with both, light organs as well as bioluminescent secretions, include the shrimp Systellaspis and Oplophorus (Herring, 1983; Latz et al., 1988), and the deep-sea vampire squid Vampyroteuthis, which release glowing particles into the water, apparently to distract predators (Haddock et al., 2010).

Encyclopedia of Animal Behavior, 2nd edition, Volume 1

https://doi.org/10.1016/B978-0-12-809633-8.90717-9

449

450

Communication j Bioluminescent Signals

Why Use Bioluminescence? Bioluminescence is a very effective way to communicate in dark environments as long as the visual pathway is unobstructed (Haddock et al., 2010). Depending on the environmental conditions, a bioluminescent flash can be seen from tens to hundreds of meters away in the ocean (Warrant and Locket, 2004), as well as on land (KSH personal observation). The vast majority of bioluminescent organisms are found in marine environments (Fig. 1). The abundance of marine bioluminescence may be partly due to the older age of the lineages living in this environment, partly due to widespread bioluminescent symbionts, and/or to the expanse of the oceans, with light penetrating only the surface layers, leaving most of the ocean environment in perpetual darkness. Yet even shallow ocean waters, dark only at night, harbor luminous organisms. In an extensive study off the California coast, 76% of all individual animals (in 553 taxa) observed between the surface and a depth of 3900 m are known to be luminous, with little variation with depth (Martini and Haddock, 2017). In contrast, only about 25% of 746 bioluminescent genera analyzed to date are found in terrestrial environments, with bioluminescent beetles contributing most genera (Oba and Schultz, 2014). For extensive surveys of known bioluminescence see the excellent reviews by Lloyd (1983), Viviani (2002), Oba and Schultz (2014), Herring (2007), Haddock et al. (2010), Widder (2010), Verdes and Gruber (2017), among others.

Bioluminescent Signals Communication signals and crypsis (obscurement of potential cues for unintended receivers) mediate key interactions between organisms in ecological communities. In the communication framework a light signal has to elicit an observable change in the behavior of the intended receiver to be considered a signal. This definition separates bioluminescent communication signals from the use of bioluminescence for crypsis (elimination of body shape cues by counterillumination: Young and Roper, 1976), however many marine animals use their bioluminescence for both. For example, the firefly squid (Watasenia scintillans) ascends from the deep ocean at night to form large near-surface aggregations during the spawning season; it uses its ability to sense and to produce light for counter-illumination, making it difficult for predators to detect it from below, however at the surface it may light up its whole body to attract a mate (Barratt and Allcock, 2014). Just like any other communication signals, bioluminescent signals are shaped by natural and/or sexual selection in the specific context and environment they are used, ultimately impacting both the signaler’s and the receiver’s fitness (Endler, 1992, 1993). Other definitions of animal communication signals require mutual benefits to signalers and receivers (Alcock, 2010), essentially limiting communication to honest signaling. However, we use a somewhat wider definition here, including signals to manipulate the behavior of potential prey and potential predators to the main benefit of the signaler and at a cost to the intended receiver.

Fig. 1 The diversity of bioluminescent organisms as proportion of 746 genera analyzed. With permission from Oba, Y., Schultz, D.T., 2014. Ecoevo bioluminescence on land and in the sea. In: Bioluminescence: Fundamentals and Applications in Biotechnology, vol. 1. Berlin, Heidelberg: Springer, pp. 3–36.

Communication j Bioluminescent Signals

451

For successful communication a signal has to be generated, travel through the environment and be detected and processed by the intended receiver before the receiver can respond.

Signal Production Signal phenotypes Bioluminescent signals can be short (flashes) or long (glows), however, these are somewhat subjective distinctions along a temporal continuum; sometimes longer signals are labelled as flashes due to their initially high, but declining brightness. Ideally, when describing signal phenotypes, these description would be accompanied by time measurements, however this is not always possible, especially for marine animals. Therefore we propose the following working definition here: a “flash” is a short light emission (2 s) signal that may vary greatly in duration. For example, the single flashes of North American fireflies (Stanger-Hall and Lloyd, 2015), ostracods in the genus Photeros (Morin and Cohen, 2010), and lanternfishes (Mensinger and Case, 1997) last less than 1 s. Glows are emitted by the larvae of all firefly species, by larviform firefly females and by a few firefly males (DeCock et al., 2014), as well as the larvae and adults of other bioluminescent beetles (Viviani and Santos, 2012). In ostracods glows often last up to 15 s (Morin and Cohen, 2010). In addition to these descriptions of light emission phenotypes, “pulses” are used to describe individual signal components within complex flash patterns (Fig. 2), much like “pulses” are used as components of complex calls in the acoustic literature (Gerhardt, 1992). In the ocean, “pulses” describe combinations of individual flashes and/or glows generated by a series of bioluminescent secretions in ostracod courtship displays (Morin and Cohen, 2010).

Biochemistry of signal phenotypes All light signal phenotypes are a direct result of the underlying bioluminescence chemistry, specifically the rates of enzyme turnover and substrate availability. While in many bioluminescent taxa this chemistry is hidden within photocytes and/or light organs, it is exposed in the bioluminescent secretions of marine ostracods, providing a rich study opportunity for the quantification of bioluminescent signal traits. Although differences in direction and distance between the pulsed secretions of ostracods are determined by male swimming behavior, other aspects of courtship displays are dominated by the biochemistry of the luciferin:luciferase reaction, which occurs in secreted mucus outside the body and therefore is not under direct behavioral control. In particular, biochemistry determines the brightness, duration, and decay rate of individual pulses (Fig. 3).

Signal context Bioluminescent signals are visual signals that are produced for specific intended receivers in three main contexts: (1) Predator defense: startle predators, illuminate predators and increase their predation risk, or as an aposematic signal to warn predators of unprofitability; (2) Food acquisition: locate food by means of built-in headlights or attract prey with light or glowing lures, exploiting their phototropic behavior; (3) Reproduction: locate and/or attract mates with species-specific temporal and spatial patterns of light emissions. Despite these clear categories, many organisms use variants of their bioluminescent signals in more than one context. For example, some fish can control their light with the help of their nervous system, using it not just to lure prey, but also potentially for intra-specific communication (Haddock et al., 2010; Fig. 4(A)), and most luminous insects use their light signals for courtship, as well as anti-predator defense or prey attraction (Lloyd, 1983; Fig. 4(B)).

Fig. 2 Flash signal traits of fireflies. Shown is a 3-flash flash pattern that is repeated at regular flash pattern intervals. Inter-pulse intervals describe the temporal spacing of flashes within multi-flash flash patterns. From Stanger-Hall, K.F., Lloyd, J.E., 2015. Flash signal evolution in Photinus fireflies: Character displacement and signal exploitation in a visual communication system. Evolution 69(3), 666–682.

452

Communication j Bioluminescent Signals

Fig. 3 (A): A cardinalfish (Apogonidae) is illuminated by the anti-predator response of a cypridinid ostracod. The cypridinid released a large cloud of bioluminescent mucus and swam away safely after a predation attempt in a staged aquarium setting. (Photo captured from video by N. Hensley and T. Rivers). (B): Light pulses are created as large anti-predator displays and also as small light pulses that form components of courtship displays. The duration of light pulses in secreted luminescence is determined by three phases: rise, plateau, and decay. In the rise phase the substrate and enzyme are first secreted and light intensity increases over time. During the plateau phase, the enzyme is limiting and total pulse duration will be determined by the enzyme:substrate ratio. In the decay phase, the substrate is limiting, resulting in an exponential decay of light as the substrate is used up. From Hensley, N.M., Ellis, E.A., Gerrish, G.A., et al., 2018. Phenotypic evolution shaped by current enzyme function in the bioluminescent courtship signals of sea fireflies. (Submitted).

(1) Predator defense (intended receivers: potential predators). Phylogenetic and ontogenetic evidence suggests that bioluminescent signaling likely originated as anti-predatory response, often in immature life-stages, of many bioluminescent animal lineages, while bioluminescent signaling in adults is derived. For example in bioluminescent cypridinid ostracods worldwide, all instars of either sex produce antipredatory bursts of light when disturbed (Morin and Cohen, 2010), but only male cypridinid ostracods from reefs in the Caribbean Sea, phylogenetically nested within the broader luminous group, produce bioluminescent courtship displays (Morin, 1986; Cohen and Morin, 2003). Similarly, all known firefly (Lampyridae) larvae of 2000þ species worldwide emit green larval glows as aposematic warning signals, directed at potential predators (Sivinski, 1981; Underwood et al., 1997; DeCock, 2009), while the adults of many, but not all, firefly species generate glows or flashes as anti-predator and courtship signals (Lloyd, 1971). Honest aposematic defense glows advertise powerful toxins, e.g., the lucibufagins (cardiotoxic steroids) of Photinus, Ellychnia and Lucidota fireflies (Eisner et al., 1978; Smedley et al., 2017) or the cyanide of bioluminescent Motyxia millipedes (Marek et al., 2011), to a would-be vertebrate predator (Lloyd, 1973). In fireflies these defense glows are part of a multimodal aposematic signal syndrome: they simultaneously emit characteristic odors, described as musky, cabbage-like, fungus, peppermint and/or resin odors in different lampyrid species, reminiscent of the description of pyrazines (DeCock, 2009), which were identified as the volatile compounds of Photuris trivittata and associated with lower rates of ant predation, but without effect on vertebrates (Vencl et al., 2016). (2) Prey attraction (intended receivers: potential prey). Prey attraction with light likely exploits a natural tendency of many insects and other animals to orient towards light, but it can also expose the light-emitting species to its own predators (Lloyd, 1983; Haddock et al., 2010). Bioluminescent elaterid beetle larvae (Pyrearinus termitilluminans) excavate an intricate network of tunnels in the outer layers of Brazilian termite mounds, and use these as vantage points to attract and catch flying termites and ants with their green bioluminescence (Costa and Vanin, 2010). Among dipterans, the carnivorous larvae of species in three genera of fungus gnats (Mycetophilidae) use their bioluminescence to attract prey and trap them in their mucus webs that are covered with sticky or poisonous droplets (Sivinski, 1982, 1998; Viviani, 2002). Although mostly untested, many possible examples of prey attraction with light exist in marine environments, including octopods and many fishes (e.g., angler fish, dragon fish, hatchet fish) that seem to use light to lure in prey (Fig. 4(A)). (3) Reproduction (intended receivers: potential conspecific mates). There is limited information on the use of bioluminescence for mate attraction in the ocean (Herring, 2007). A few examples include the bioluminescent interactions of male and female ponyfish with male species-specific luminescent signaling (Sparks et al., 2005), and the species-specific male courtship displays of cypridinid seed shrimp (ostracods; Cohen and Morin, 2003). Some marine species exhibit sexual dimorphism of light-producing phenotypes, suggesting a use in mate-attraction (reviewed in Ellis and Oakley, 2016). In contrast, mate attraction is a well-characterized function of adult bioluminescent signaling on land. For example, fungus gnat females are luminous, even though they do not eat, suggesting that they use their light to attract males. This is supported by adult males orienting towards pupal and female lights, and numerous males clinging to and fighting over female pupae, waiting for them to eclose; if no male is attached at the time of eclosion, adult females may flash their light on and off until a male arrives (Sivinski, 1982, 1998). Among beetles, flashes are exclusively used by lampyrids (fireflies), and glows are used by lampyrids, phengodids (railroad worms), elaterids (click beetles), and rhagophthalmids (Asian glow worms) for mate attraction by one or both sexes (Fig. 4(B)). There is compelling evidence that the origin of bioluminescence as a sexual signal is associated with higher species abundance and speciation rates across animals both on land and in the ocean (Ellis and Oakley, 2016).

Communication j Bioluminescent Signals

453

Signal structure and function As in other signal modalities, specific light signal characteristics are associated with signal function and/or signaling environment. For example short signals (i.e., flashes) with a sudden onset and high intensity are often used to startle potential predators, while longer signals (i.e., glows) that allow the receiver to localize a stationary sender are often used to attract prey and potential mates. Continuous glows from moving senders may often be advantageous in environments such as forests, where intermittent flashes, temporarily obscured by vegetation, can reduce signal detection efficiency by a receiver. However, in predator-rich environments, continuous glows may also significantly increase the predation risk of signaling animals and it is likely that the intermittent luminescence in fireflies evolved under predation pressure (Lloyd, 1989). When signals are not generated within animals, but outside the body in bioluminescent secretions, as is the case for many marine animals, the limitations imposed by predators on signal traits are relaxed, or even may be reversed. Releasing a big, lasting

Fig. 4 Hypothesized functions of (A) (top): marine bioluminescence. From Haddock, S.H.D., Moline, M.A., Case, J.F., 2010. Bioluminescence in the sea. Annual Review of Marine Sciences 2, 443–493, and (B) (bottom): terrestrial bioluminescence in the contexts of defense (blue), offense (magenta), and intraspecific communication (gray). The organisms benefiting from these are listed to the right in each panel. Some animals are known to use their luminescence in several different roles.

454

Fig. 4

Communication j Bioluminescent Signals

(continued).

cloud of glowing mucus into the environment when threatened by a predator likely increases the escape opportunities for the potential prey by extending the time before a predator can reorient and pursue the prey. For example many copepods release their bioluminescent chemicals from glands on the tail or swimming legs, which generate enough thrust during escape to eject the bioluminescence beyond the boundary layer of the escaping animal (Widder, 2002). As an advantage of secreted bioluminescence, the attackers can potentially be covered by the bioluminescent mucus, making them easy targets for secondary predators (Morin, 1986; Haddock et al., 2010; Widder, 2010). Similarly, when secreted signals are used as male location cues for females during mate attraction, the respective males usually do not move too far from their secreted signals, e.g., by swimming in tight spirals, rather than in straight lines (Rivers and Morin, 2008). If signals are used for mate attraction and several species are active in the same area, sexual selection can cause these signals to evolve into complex, species-specific courtship displays that consist of temporal (flash duration, inter-pulse interval, flash pattern interval) and spatial (signaling area in environment and male movements while signaling) sequences of light-signals, as evident in the intriguing courtship displays of both male terrestrial fireflies (Lloyd, 1966) and male marine ostracods (Gerrish and Morin,

Communication j Bioluminescent Signals

455

Fig. 5 Species-specific light signals of male fireflies and male ostracods, the two most thoroughly studied bioluminescent groups on land and in the ocean. These sexual signals are emitted as repeated flash patterns (fireflies) or pulses of bioluminescent secretions (ostracods) in space and time, and result in female flash answers (not shown) from sedentary firefly females, and approach behavior (without light signals) in female ostracods. In addition to the temporal and spatial signal variation among species, complex courtship behaviors can provide additional variation for sexual selection in both groups. (A) (left). Fireflies: (1) Photinus consimilis (slow-pulse), (2) P. brimleyi, (3) P. carolinus, (4) P. collustrans, (5) P. marginellus, (6) P. consangineus, (7) P. ignitus, (8) P. pyralis, (9) P. granulatus. (B) (right). Ostracods: (1) MWU, (2) Photeros annecohenae, (3) Kornickeria hastingsi, (4) MSH, (5) Photeros morini and (6) ZZD; relative signal duration indicated by number of circles surrounding the center. Filled circles represent pulses that would simultaneously be visible during the display based on signal durations (Gerrish, G.A., Morin, J.G., 2016. Living in sympatry via differentiation in time, space and display characters of courtship behaviors of bioluminescent marine ostracods. Marine Biology 163(9), 190; illustration by Natalie Renier).

2016), the two best-studied terrestrial and marine bioluminescent groups (Fig. 5). However, unlike the reciprocal duets seen in fireflies where females respond to males with species-specific response delays, female ostracods are not known to signal during courtship, rather they simply approach the displaying male. Behaviorally the ostracod system is similar to acoustic signaling in amphibians and insects, where “silent” females approach signaling males (Gerrish and Morin, 2016).

Neural control In contrast to ostracods, which secrete their bioluminescent chemistry into the environment, fireflies retain neural control of their bioluminescent reactions (Timmins et al., 2001). Species-specific pattern generators are likely involved in generating the diverse courtship displays in both ostracods (secretion pattern) and fireflies (flash pattern) (Carlson and Copeland, 1985, Stanger-Hall and Lloyd, 2015). Variation in these neural controls would allow for evolutionary modifications of bioluminescent signaling behavior. Similarly, in brittle stars (echinoderms) a complex system of neurotransmitters modulates light output, with yet unknown function (Haddock et al., 2010).

Signal Detection: Brighter and High-contrast Signals are Detected More Easily Light signals tend to be used in dark environments with brighter signals allowing for higher probabilities of detection, and at greater distances. Ambient light, moonlight, and low visibility due to particles in the ocean or in the air, interfere with detection of light signals (reducing the signal/light noise ratio) and can impact the bioluminescent signaling behavior of animals both on land and in the ocean. For example, firefly species change the onset of their bioluminescent signaling behavior in response to cloudy skies and/ or moonlight (Lloyd, 1966), and ostracods only start signaling in the shallow ocean waters at astronomical twilight ( 1 h after sunset) or near moonset, whichever occurs last (Morin and Cohen, 2010). Therefore, when required for efficient signaling, darkness becomes an ecological resource (Gerrish et al., 2009). Artificial lighting and wide-spread light pollution can reduce flashing activity in both ostracods and fireflies (Lloyd, 2006; Hagen et al., 2015; Gerrish and Morin, 2016; Owens et al., 2018) and it significantly reduces firefly mating (Ineichen and Rüttimann, 2012; Firebaugh and Haynes, 2016). When male Aquatica ficta fireflies, are experimentally exposed to higher levels of ambient light (50 lbs.) on mammalian evolution. True mammals had existed for
100 My preceding this extinction event. Yet their evolutionary diversification had been unimpressive, with their modest taxonomic diversity primarily restricted to small, morphologically generalized (more or less rat-like) insectivores. This paucity of early mammalian variety has commonly been attributed to their competitive exclusion from most niches by ecologically dominant reptile taxa, such as dinosaurs. This hypothesis receives support from the spectacular phylogenetic and phenotypic ‘mammalian radiation’ that occurred during the
10 My immediately following the mass extinction. Most major mammalian orders first appear in the fossil record at this time, as do many of the most specialized mammalian forms

q

Change History: July 2018. Gil Rosenthal added in-text reference citations. This is an update of D.J. Funk, Specialization, Editor(s): Michael D. Breed, Janice Moore, Encyclopedia of Animal Behavior, Academic Press, 2010, Pages 315-321.

Encyclopedia of Animal Behavior, 2nd edition, Volume 2

https://doi.org/10.1016/B978-0-12-809633-8.20867-4

101

102

Evolution j Specialization

such as bats, whales, primates, and enormous herbivores. This belated mammalian profusion was seemingly promoted by access to suddenly unexploited resources/niches and their subsequent partitioning. A different, coevolutionary, scenario of adaptive radiation was offered by Ehrlich and Raven (1964) to explain the evolutionary proliferation of two ecologically associated groups of organisms: plant-feeding insects (hereafter referred to simply as ‘herbivores’) and flowering plants. The scenario begins with an herbivore lineage overcoming the defenses of an unexploited group of plants (e.g., by evolving digestive enzymes that neutralize plant toxins). The herbivore lineage exploits and partitions this new resource via the evolution of new herbivore species, each specialized on a different plant. Subsequently, a lineage of these plants evolves a newly impenetrable defense, allowing it to analogously diversify by filling the niches it had been excluded from by the presence of its herbivore attackers. Iteratively repeating this coevolutionary cycle could thus explain much about specialization and species richness in these two groups. This model helped establish the ecological specialization of insect herbivores as a major topic in evolutionary ecology (Strong et al., 1984; Tilmon, 2008), hence the focus on this phenomenon by the present article.

Herbivorous Insect Exemplars Herbivory has independently evolved dozens of times across nine insect orders. Compared to other taxa, insect herbivores frequently exhibit strong ecological specialization (Jaenike, 1990; Fig. 1). The degree of specialization of an herbivore species is

Fig. 1 Herbivorous insect species that have provided insights on the evolution of ecological specialization. Most are invoked in the article. (a) Rhagoletis pomonella fruit flies, whose host race formation via specialization on apple versus hawthorn trees has informed our study of speciation through work by Jeffrey Feder and colleagues. Courtesy of J. Feder. (b) Timema cristinae stick insects, whose two ecotypes exhibit specialized morphologies (body shape and stripe presence/absence) that render each ecotype best camouflaged from predators when on its native host plant. Studied especially by Chris Sandoval and Patrik Nosil. Courtesy of P. Nosil. (c) Neochlamisus bebbianae leaf beetles, whose populations each specialize on one of several unrelated tree species. These populations prefer, and survive best on, their native host plant and appear to be speciating as a consequence of this divergent host adaptation. Developed and employed as a study system by Daniel Funk. Courtesy of C. Brown. (d) Grammia crotalaria caterpillars, generalists that are predators on various small herbaceous plants, perhaps due to the nutritional value of such a mixed diet. Studied by Michael S. Singer, who provided the photo. (e) Jadera haematomola soapberry bugs, whose beak length has evolved markedly within a matter of decades in response to the introduction of plant species whose seeds (the bug food source) deviate in their depth from the fruit surface as compared to seed depth in the insect’s normal host plant. Research and photo provided by Scott Carroll. (f) Uroleucon ambrosiae aphids, whose evolution of generalization from specialized ancestors was the focus of work by Daniel Funk and colleagues, and is detailed in the last section of this article. (g) Ophraella notulata leaf beetle, one of the species studied by Douglas Futuyma to evaluate the role of genetic constraints on the evolutionary history of host shifts and specialization across this beetle genus. (h) Euphydryas editha butterflies, whose populations exhibit complex patterns of geographic variation in host plant specialization, including host preference changes that have evolved over a matter of years. Devotedly studied by Michael C. Singer, who provided the photo. (i) Enchenopa binotata treehoppers. The eggs of Enchenopa populations associated with different host plant species begin development at different times that specifically correspond to variation in initial spring sap flow among these hosts. Thus, treehoppers on different hosts mature at different times, reducing interpopulation mating opportunities and illustrating environmentally based temporal isolation. Investigated by the late Tom Wood.

Evolution j Specialization

103

commonly evaluated in terms of the variety of taxa used as host plants. Whether the criterion for specialist status is having hosts restricted to a single plant family, a single genus, or a single species, a sizeable proportion of herbivores qualify as ecologically specialized. At the same time, more ‘generalized’ herbivores that use a considerable diversity of host plants also exist. Thus, although most herbivores are specialists, considerable variation in the degree of specificity/generalism exists among herbivore taxa. For example, while Orthoptera (grasshoppers and kin) and certain Lepidoptera (butterflies and moths) families are dominated by highly polyphagous generalists that use many plant families as hosts, leaf beetles, and aphids exemplify groups dominated by monophagous specialists that use but a single host plant species. In between, are oliphagous species that use a modest number of host taxa. Other informative terms for characterizing herbivore specialization include ‘preference hierarchy’, a ranking of the relative degree to which an herbivore prefers each of its multiple hosts, and ‘host range’, which refers to the suite of particular plant species an herbivore uses as hosts. In thinking about herbivore specialization, it is important to recognize that for many herbivores, the host plant is effectively its entire habitat. There, all life activities occur, from oviposition through larval development and adult feeding and mating. For such insects, host specialization thus imposes extremely strong selection on diverse aspects of biology. This may influence the evolution of sensory structures and behaviors governing the location and acceptance of the host as a site of oviposition, feeding, and mating; of anatomical structures allowing adherence to the host substrate and feeding on host tissues; of the physiological capacity to tolerate or avoid host toxins and digest host tissues; of behavioral strategies for avoiding desiccation and natural enemies; and so on. With this in mind, imagine that a population specializing on host A becomes geographically isolated on novel plant B in an area devoid of host A. Since herbivore host-associations are often determined by behavioral preferences that belie a physiological capacity to develop on nonhost plants, our hypothetical population may plausibly survive on novel plant B. If indeed it manages to establish a population on B, strong selection would thereafter promote adaptation to it, resulting in its adaptive divergence from the ancestral population on host A in diverse phenotypic traits relevant to ecological specialization. Recent molecular studies reveal wide-ranging genomic effects of this divergent host adaptation, supporting the enormous influence of host-associated selection on herbivore evolution. For example, sampling hundreds of genetic loci from Neochlamisus bebbianae leaf beetle population specializing on maple versus willow trees indicated that 10%–15% of their genomes may be evolving under the influence of this ‘host-related selection’. And such selection can yield rapid results. For instance, within the last 100 years, soapberry bugs have evolved considerably longer beaks to penetrate the broader fruits of an introduced plant species, allowing them to feed on the seeds inside. Because of their great diversity, discrete hosts/habitats, and tendency to exhibit phenotypic divergence, as well as their small size, fast generation times, and ready propagation, many herbivores provide great opportunities for controlled and informative studies of ecological specialization. Such studies may even have practical benefits, given the great economic costs imposed by herbivorous pests of crops, most of which are not native hosts of these pests. A notable example is the sowing of an herbivore’s native host plants amidst crop plants. This provides the herbivore with a preferred alternative to the crop plant, thus ameliorating the otherwise intense selection pressures on such herbivores to adapt to crops grown in large monocultures. This strategy reduces the likelihood that a native herbivore will evolve into a (greater) pest.

Evaluating Host Preference: A Principle Determinant of Host Use A herbivore’s capacity to feed and develop on a plant is influenced by various anatomical and physiological factors, such as having appropriate mouthparts and digestive enzymes. However, host preference is governed by critical behavioral components. This section introduces some of these components and various means of studying them. Since most herbivores must locate new host individuals during their life cycle (e.g., for oviposition), a herbivore’s initial manifestation of host preference is often based on long-distance cues. These are generally olfactory and detected by receptors (e.g., on the antennae) sensitive to very small concentrations of air-borne plant volatiles. Locating the source of those volatiles eliciting interest brings the herbivore within range of potential visual cues. A plant acceptable to this point may be landed upon, such that touch (via mechanoreceptors) and taste (via chemoreceptors) come into play as the herbivore assesses external plant anatomy (e.g., surface waxes, plant hairs, etc.) and chemistry. This may be followed by the sampling of internal constituents via feeding and the assessment of postingestive feedback. Plants produce a diverse array of secondary metabolites that often play critical roles in plant acceptance versus rejection as a host. These include nonvolatile chemicals that contribute to the stimulation versus deterrence of herbivore feeding and oviposition. Herbivore researchers have evaluated the stages and factors involved in host identification, acceptance, and preference using a heterogeneous array of approaches: Wind tunnels are used to evaluate long to medium distance host orientation and preference. A herbivore flying upwind from one end of the tunnel toward odor sources blown from the other end allows the assessment of its orientation behavior in the odor plume and of odor attractiveness. Another common technique uses ‘Y-tube olfactometers’, whereby an insect is placed in the tube at the base of the ‘Y0 while alternative odors are wafted through the two distal tubes, providing the herbivore with a choice when it reaches the tube’s fork. Visual cues may be evaluated by investigating herbivore responses to variables such as alternative leaf shapes and colors. Behavioral experiments are often supplemented by fine-scale, electrophysiological assays. The electroantennogram technique involves removing a herbivore antenna, inserting silver wires in each end and measuring the antenna’s electrical output when exposed to an odor, thus quantifying relative neurological response. At an even finer scale, the responses of individual sensilla and even individual cells within sensilla can be evaluated. Particular plant

104

Evolution j Specialization

chemicals can be efficiently investigated as potential feeding stimulants or deterrents by adding them to artificial diets and evaluating feeding responses. Feeding and oviposition preferences can also be more ‘naturally’ evaluated by scoring feeding damage or eggs laid by a herbivore on actual plant foliage over a specified period of time. Finally, simple, continuous observation of individual herbivore behavior in the lab or field can prove very informative.

Evolutionary Hypotheses for Ecological Specialization Resource limitation and associated competition are commonly invoked to explain ecological specialization. Yet that explanation seems insufficient for herbivores. A summer walk through the woods or an old field reveals an abundance of green, healthy foliage. ‘The world is green’, and plants overcome by herbivore attack are relatively uncommon. So why do most herbivores specialize on a minority of the available resources? The very active area of research on this seeming paradox has produced a number of hypotheses. 1. Perhaps the best known is the tradeoffs hypothesis. Based on the notion that ‘a jack of all trades is master of none’, this hypothesis invokes the evolutionary difficulty of simultaneously adapting to multiple hosts. It suggests that generalists are likely to use each of their many hosts less effectively than specialists use their more limited hosts, resulting in lower generalist fitness. Further, any increase in the capacity to use a new host plant is likely to reduce the capacity to use an ancestral host. This hypothesis is intuitively very appealing, yet the concept of tradeoffs has received limited empirical support. 2. An alternative proposes that specialization is favored when a host is very abundant and easily located, thus representing a reliable resource. This ‘plant apparency’ hypothesis is supported by the tendency of specialists to use common hosts more frequently than do generalists. 3. A third hypothesis links herbivore feeding behavior to predation pressure. Generalist predators tend to feed on generalist prey, thus allowing specialists to effectively escape into enemy-free space. 4. One intriguing hypothesis suggests that specialized species are favored by sexual selection. Using a single host plant provides a reliable ‘mating rendezvous’ that increases the chances of potential mates finding one another. This hypothesis, which has not yet been rigorously tested, suggests that specificity is especially advantageous in low-density herbivore species for whom matefinding may be especially challenging. 5. A more recent hypothesis that has received strong experimental support posits that specialists live in a cognitively simpler world than generalists, having fewer host-associated cues to process when making decisions about host preference and thus facing simpler choices than generalists, which must choose among multiple hosts. This ‘cognitive constraints’ hypothesis predicts that specialists, free from the distractions of multitasking, should be more efficient in their host-use behaviors. This efficiency should result in higher growth rates, greater fecundity, and reduced exposure to predators, yielding higher fitness and selection for greater specificity. This area of research also explores why some herbivores are generalists. If particular host plants are rare or unpredictably available in time and space, specialists run a high risk of not finding a host plant at all. Indeed, multiple studies have shown a tendency for greater generalism in marginal environments. Generalism may also be advantageous in the minority of cases where the herbivore is effectively a plant predator, that is, where it completely consumes small plants and then must find more. Unless a given host is particularly common and the predation risks associated with moving between hosts are modest, generalism should be advantageous for such herbivores. Another potential advantage to generalism is nutritional. A single plant species is unlikely to provide the optimal balance of nutrients for insect growth, and thus there may be an advantage to ‘food mixing’, that is, feeding on multiple host species that cumulatively provide a balanced diet. Grammia caterpillars offer informative investigations of both of the latter two hypotheses.

Phylogenetic Insights on Historical Patterns of Host Association Understanding the evolution of ecological specialization is further advanced when phylogenetic approaches explore its historical aspects (Mitter et al., 1988; Nosil and Mooers, 2005). For example, one can evaluate the evolutionary history of host associations among a group of related herbivores by treating host plant taxon as a trait to be reconstructed (i.e., ‘mapped’ on a phylogeny), thus providing the most parsimonious explanation of this history. By phylogenetically tracing evolutionary changes in host use by herbivore lineages through time, specific inferences about the number, timing, nature, and directionality of host shifts can be made. Such inferences may further provoke investigations of issues such as the conservative versus labile nature of herbivore host use evolution, the causes of host shifts, and their role in speciation. Historical changes in the degree of host specificity can be similarly evaluated. In the simplest analyses, herbivore species are dichotomously characterized as specialists or generalists, and the history of changes between these strategies is reconstructed. Such analyses have rejected a long-standing hypothesis that specialists represent an evolutionary dead end, having become so specifically adapted to their host that they have lost the evolutionary capacity to use alternative plants. Such an irrevocably specialized herbivore would be incapable of spawning new species via changes in host association see Section “Ecological Specialization and Speciation” and be more prone to extinction, given its reliance on a single resource. This hypothesis thus predicts that while

Evolution j Specialization

105

generalist lineages can spawn specialists, the reverse is unlikely. In contrast, phylogenetic studies provide examples of both situations, indicating that host specialization is not quite so evolutionarily constraining. Indeed, the phylogenetic reconstruction of host use has been applied to a great many issues, including the use of alternative plant tissues and growth forms, the role of plant chemicals as determinants of host association, the geographic context of host range variation, etc. Further, the method of independent contrasts can extend the reconstruction approach and allow the statistical testing of whether nonrandom associations exist between the phylogenetic distributions of different traits (e.g., between particular host associations and plant toxins). This allows the evaluation of adaptive hypotheses on the causes of ecological specialization. Phylogenetic investigations can also be integrated with experiments (Futuyma and Moreno, 1988). One such study tested the hypothesis that genetic constraints had influenced host use evolution in Ophraella leaf beetles (Futuyma et al., 1995). This study used a quantitative genetic approach to evaluate genetic variation in the capacity of particular Ophraella species to use the host plants of congeners. It found evidence for genetic variation significantly more frequently for those Ophraella species tested on the host of a close phylogenetic relative or on a plant closely related to its own host. These findings suggested that the maintenance of appropriate genetic variation facilitates host shifts to closely related plants, thus helping explain the frequency of taxonomically conservative host shifts in other herbivores. These conservative tendencies were further supported by a reconstruction of host association at the taxonomic level of tribe that implied only one more than the minimum possible number of host shifts. Intriguingly, a complementary reconstruction of hosts at the lower taxonomic level of genus revealed four evolutionarily independent host shifts between Ambrosia and Iva. Especially given the close relationship of these two genera, this repeated pattern provided another, novel, form of evidence for conservative host use evolution in Ophraella.

Characterizing Host Range: Not as Simple as it Seems This article has implicitly assumed the accuracy of documented herbivore host ranges and the homogeneity of host range among individuals within a given herbivore species. However, reality is not so simple. First, reported host associations may be inaccurate due to the misidentification of host plants or host records based on observations of herbivores that are merely resting on a nonhost plant. Further, a herbivore’s host range is often incompletely documented, a fact promoted by a tendency to seek a herbivore on its previously recorded hosts. The existence of so many herbivores and plants makes such inaccuracies difficult to avoid. Doing so requires intensively systematic sampling of all possible plants, as is presently occurring via Dan Janzen’s large-scale Costa Rican studies on the host ranges of species of particular lepidopteran taxa. Second, many herbivore species use different host taxa in different parts of their geographic range. This may simply reflect geographic variation in the availability of hosts rather than in the herbivore’s actual host preferences. Alternatively, it may reflect genetically based variation in host preferences, perhaps reflecting divergent adaptation to local flora (Fox, 1981). Third, host range variation may exist within local populations. Such individual-level variation represents an understudied aspect of ecological specialization (Bolnick et al., 2003). Nonetheless, agricultural examples are provided by the multiple biotypes within various pest species, each genetically adapted to overcome the resistance alleles of a particular crop genotype. Likewise, such genetic variation in host preference has been demonstrated in natural populations. Individuals may also differ due to environmental influences. Particularly compelling examples are provided by aphids of the same genotypic clone that differ in induced host preference depending on what plant they have been reared on. Maternal effects provide a case where induction is cross-generational. For example, a mother feeding on plant A prior to oviposition might pass along some A-specific chemicals to her eggs that predispose the hatchling larvae to prefer plant A more than if she had fed on plant B. Thus, teasing apart the genetic versus environmental contributions to host use is critical to understanding the causes of host specialization. Fourth, recent methodological advances have increasingly revealed nominal species of herbivore to include multiple anatomically homogeneous but reproductively isolated and host-specific sibling species. For example, recent experiments show host-use, reproductive, and molecular differentiation between co-occurring N. bebbianae leaf beetle populations associated with red maple versus river birch, supporting their likely status as sibling species.

Ecological Specialization and Speciation Suspicions of an association between host specialization and species formation date to the nineteenth-century writings of Benjamin Walsh and are supported by Ehrlich and Raven’s (1964) coevolutionary model and the existence of host-specific sibling species. Further support was provided by a study documenting statistically elevated levels of species richness across many independently evolved lineages of herbivores as compared to their equally aged and nonherbivorous sister groups. This study provided the first rigorous evidence for a general association between herbivory and species-level diversity. The above study was published just as intriguing findings were being reported on a herbivore system that had inspired Walsh himself: populations of Rhagoletis fruit flies associated either with their ancestral hawthorn host or the recently adopted apple host (apple having been introduced to the United States less than 200 years previously). These ‘haw’ and apple fly populations are now known to be divergently adapted to their respective host plants. They also exhibit partial host-associated reproductive isolation. For example, their tendency to mate on the host plant reduces interbreeding between these host-associated populations. The modest gene flow between these populations is nonetheless sufficient to yield genetic homogenization across most

106

Evolution j Specialization

of their genome. However, these populations are strongly differentiated at several genetic regions that contain the genes underlying their divergent adaptation. The haw and apple populations thus seem to represent an intermediate state in an ongoing process of speciation promoted by divergent host adaptation to host plants. Eurosta fruit flies, Timema stick insects, Neochlamisus leaf beetles, Heliconius butterflies, and pea aphids are among the increasing number of additional herbivore taxa believed to illustrate speciation via divergent host specialization. These systems provide examples of ‘ecological speciation’, whose models explain how genetic adaptation to alternative environments incidentally results in reproductive isolation and species formation (Berlocher and Feder, 2002). In this article, we have considered (1) the tendency of herbivores toward ecological specialization, (2) the strong selective pressures imposed on populations for whom host plant equals habitat, (3) models arguing that divergent adaptation promotes reproductive isolation, (4) the documentation of host-specific sibling species, and (5) evidence that herbivorous lineages are especially species-diverse. In this light, consider again Darwin’s tangled bank. I would wager that if one inspected that bank closely, they would find a myriad of specialized insect herbivores!

An Integrated Example From Uroleucon Aphids I conclude this article by describing my postdoctoral investigations of ecological specialization in Uroleucon ambrosiae aphids (Funk and Bernays, 2001), research that involved many issues addressed above. Aphids are phloem sapsuckers that feed by penetrating host plant tissues with their beak-like mouthparts. Uroleucon is a genus of aphids whose species tend to be monophagous on hosts in the Asteraceae (the sunflower family). U. ambrosiae was known as a specialist on Ambrosia trifida (the giant ragweed), a common plant in disturbed habitats of the eastern United States. Thus, it was occasional reports of this aphid on other plants in the southwestern United States that first motivated my studies. These confirmed the specialist status of U. ambrosiae in the east where my fieldwork recovered it only on A. trifida, excepting a few collections on another Ambrosia species and a species of the closely related genus Iva. In Arizona, however, I collected U. ambrosiae from 16 host species, representing 11 genera and 4 tribes of Asteraceae, plus a species of Malvaceae, documenting its regionally generalist status. Individual clones of eastern and southwestern populations were subsequently collected and maintained on host plants in a greenhouse. (Most aphids alternate between periods of clonal and sexual reproduction, but clonality can be maintained using long-day photoperiods.) Four southwestern hosts (including A. trifida) were used for a variety of behavioral assays using wind tunnels, continuous observation of host acceptance behaviors, electrical penetration graph analysis (EPG; in which electrodes record aphid mouthpart activity within plant tissues), and longer-term choice trials of settling behavior. These experiments showed eastern aphids to be significantly more specialized and efficient than southwestern aphids. The use of replicate clonal genotypes established a genetic basis for this differentiation. These findings showed that southwestern aphids have evolved increased generalism and supported the cognitive constraints hypothesis on advantages of the specialization, while likewise raising the question of what advantages to generalism might outweigh the reduced efficiency of the southwestern aphids. A potential mechanism for this generalism was provided by the significantly and consistently reduced numbers of antennal sensilla across southwestern versus eastern populations. We hypothesized that the loss of sensory structures used in host plant selection could diminish the capacity to distinguish among alternative plants (cf. Bernays, 2001), resulting in the broader host range of the southwestern aphids. The capacity of these aphids to develop on this wide range of hosts was also demonstrated, by host performance studies showing that eastern and southwestern aphids each performed best on the ancestral A. trifida host, but exhibited no differences in their relative capacity to perform on the four test plants. Thus, eastern aphids appeared physiologically preadapted to use plants they did not locally encounter. Intriguingly, however, no further physiological adaptation to these new hosts had occurred in the southwest. Likewise, no host-associated molecular genetic differentiation was observed. Further findings contributed more pieces to the puzzle. Cross-continental DNA sequence variation proved astonishingly low, with no evidence of geographic structure. Population genetic analyses indicated that the former result could reflect the loss of variation under bottleneck-induced genetic drift. The latter result suggested that U. ambrosia may represent an approximately panmictic population across the entire United States. Indeed, the small size and limited flying ability of aphids make them ‘insect aeroplankton’ that may be transported great distances by air currents, thus potentially homogenizing populations via long-distance gene flow. A final suite of environmental observations potentially completes this puzzle. The riparian areas suitable for host plant growth proved to be rare and patchy in Arizona. Further, I experienced the unpredictability of desert environments in the form of dramatically varying levels of annual rainfall. Correspondingly, I observed great annual variation in the density of host plants, and associated aphid densities that varied by at least two orders of magnitude. Cumulatively, the above findings support the following scenario for the evolution of host generalism in southwestern U. ambrosiae aphids: Eastern specialists blown to the southwest initially failed to become established due to the patchiness and unpredictable presence of their host plant. Eventually, mutations occurred that yielded the reduction of antennal sensilla and a correspondingly reduced capacity to discriminate against plants they were actually capable of developing on. This ‘behavioral release’ freed southwestern aphids to adopt multiple hosts, and from the precarious existence formerly imposed by the difficult challenge of consistently locating one specific host plant. Selective forces favoring this generalism are apparently strong enough to maintain the low-sensilla mutation(s) at high frequency in the face of considerable gene flow. Thus, in the unpredictable southwest, the advantages of U. ambrosiae generalism apparently far outweigh the efficiency advantages offered by specialization.

Evolution j Specialization

107

See also: Evolution: Isolating Mechanisms and Speciation; Phylogenetic Inference and the Evolution of Behavior. Foraging: Habitat Selection; Optimal Foraging Theory: An Introduction. Hormones and Behavior: Maternal Effects on Behavior. Host-Parasite Interactions: Avoidance of Parasite. Learning and Teaching: Decision-Making and Learning: The Peak Shift Behavioral Response. Neurons and Senses: Invertebrate Vision; Taste: Invertebrates.

References Berlocher, S.H., Feder, J.L., 2002. Sympatric speciation in phytophagous insects: Moving beyond controversy? Annual Review of Entomology 47, 773–815. Bernays, E.A., 2001. Neural limitations in phytophagous insects: Implications for diet breadth and evolution of host affiliation. Annual Review of Entomology 46, 703–727. Bernays, E.A., Chapman, R.F., 1984. Host-Plant Selection by Phytophagous Insects. Chapman & Hall, New York, NY. Bolnick, D.I., Svanback, R., Fordyce, J.A., et al., 2003. The ecology of individuals: Incidence and implications of individual specialization. American Naturalist 161, 1–28. Ehrlich, P.R., Raven, P.H., 1964. Butterflies and plants: A study in coevolution. Evolution 18, 586–608. Fox, L.R., 1981. Specialization: Species property or local phenomenon? Science 211, 887–893. Funk, D.J., Bernays, E.A., 2001. Geographic variation in host specificity reveals host range evolution in Uroleucon ambrosiae aphids. Ecology 82, 726–739. Futuyma, D.J., Keese, M.C., Funk, D.J., 1995. Genetic constraints on macroevolution: The evolution of host affiliation in the leaf beetle genus Ophraella. Evolution 49, 797–809. Futuyma, D.J., Moreno, G., 1988. The evolution of ecological specialization. Annual Review of Ecology and Systematics 19, 207–233. Jaenike, J., 1990. Host specialization in phytophagous insects. Annual Review of Ecology and Systematics 21, 243–273. Mitter, C., Farrell, B., Wiegmann, B., 1988. The phylogenetic study of adaptive zones: Has phytophagy promoted insect diversification? American Naturalist 132, 107–128. Nosil, P., Mooers, A.O., 2005. Testing hypotheses about ecological specialization using phylogenetic trees. Evolution 59, 2256–2263. Schluter, D., 2000. The Ecology of Adaptive Radiation. Oxford University Press, Oxford. Strong, D.R., Lawton, J.R., Southwood, R., 1984. Insects on Plants: Community Patterns and Mechanisms. Harvard University Press, Cambridge, MA. Tilmon, K.J. (Ed.), 2008. Specialization, Speciation, and Radiation: The Evolutionary Biology of Herbivorous Insects. University of California Press, Berkeley, CA.

FORAGING Foraging: Section Overview Graham H Pyke, University of Technology Sydney, Ultimo, NSW, Australia; and Macquarie University, NSW, Australia © 2019 Elsevier Ltd. All rights reserved.

Abstract This section on foraging behaviour is warranted in this encyclopedia because foraging is fundamental to all life. The foraging process involves different stages (e.g., pre- and post-ingestion) and a number of types of decision (e.g., patch choice, diet choice, patch exploitation, movements, group membership, producer versus scrounger, foraging mode), leading to studies that focus on such different foraging stages and decisions. The chapters in this section therefore provide a comprehensive coverage of foraging behaviour by separately focusing on one or other of these stages and decisions. All chapters in this section rely on evolutionary theory to understand observed foraging behaviour. Foraging by animals (and other organisms) is such an important and fundamental aspect of animal behaviour, that we need to understand it as well as possible, and the best way to achieve this is to invoke evolutionary theory with animals hypothesised to forage in ways that maximise their biological fitness. This evolutionary approach leads to Optimal Foraging Theory (OFT), which includes Classic OFT, if individual animals do not respond directly to the foraging behaviour of others, and the more-general Foraging Game Theory, if such responses do occur. In either case, the basic hypothesis is that animals in a population will adopt a foraging strategy such that no individual can achieve higher biological fitness by deviating from the rest of the population. If individuals do not respond directly to the foraging behaviour of others, this hypothesis is equivalent to classic maximisation of some currency that acts as proxy for biological fitness. If such responses do occur, then individuals are collectively involved in foraging ‘games’ that are expected to be at equilibrium. Consequently, all chapters in this section embrace OFT, in one guise or the other. OFT has grown and developed enormously during its 50-year history, through expansion, extension, application and inspiration, with observations generally supporting expectations, to the point where it is now a ‘strong theory’ of behaviour and ecology. Several chapters in this section describe such growth and development. This section also aims to provide stories that you, our readers, will find easy to read and understand, as well as interesting, informative, compelling and memorable.

Introduction Foraging is undoubtedly the most fundamental and important aspect of animal behaviour, and so this Encyclopedia of Animal Behaviour, or any similar general treatment of this subject, warrants a section devoted to Foraging. In this chapter I provide an overview to this section. I begin with a discussion of the different stages of foraging up to and after food is obtained, which is followed in turn by discussions of the foraging decisions involved in obtaining food, observed patterns in foraging behaviour, the use of Optimal Foraging Theory (OFT) (which includes Classic OFT and the more-general Foraging Game Theory) to understand these patterns, and how OFT has grown and developed. The chapters in this section provide good coverage of all these issues and point to future research that addresses them. I finish by indicating how these chapters have been crafted to make them easy to read, interesting, informative, compelling and memorable.

Foraging – Stages and Decisions Foraging involves the following stages: obtaining food or other resource; storage or ingestion; and post-ingestion. However, this section on foraging focuses primarily on aspects of animal behaviour that are involved in obtaining food and its external storage, because they fall within the traditional realm of animal behaviour and been subject to the greatest amount of interest. On the other hand, post-ingestion processes also constitute behaviour, especially as individual animals exhibit considerable flexibility in how they are implemented and post-ingestion experience (such as through eating toxin-laden food) can influence subsequent food choice. Consequently, a few chapters consider internal food storage, other post-ingestion processes, and their interactions with subsequent foraging behaviour. Foraging, leading up to when food is obtained, is a process involving constant decision-making with respect to various alternative options. An animal may, for example, make choices in terms of where to forage (patch choice or habitat selection), what food items to eat (diet), when to leave one food location for another (patch departure), and how to move within and between food patches (movement patterns) (Pyke et al., 1977; Pyke, 1984). Animals may also choose whether to forage solitarily or socially (i.e., as part of a cohesive group), to transfer from one group to another, to forage for themselves or ‘scrounge’ food from other

108

Encyclopedia of Animal Behavior, 2nd edition, Volume 2

https://doi.org/10.1016/B978-0-12-809633-8.20879-0

Foraging j Foraging: Section Overview

109

group members (Giraldeau and Dubois, 2008), and what foraging mode to adopt (e.g., actively search versus ‘sit and wait’) (Visser and Fiksen, 2013).This section on foraging behaviour therefore includes chapters that cover this range of foraging decisions.

Optimal Foraging Theory – Classic and Foraging Game Theory The generally accepted explanation for why animals (and other organisms) forage in the ways they do, is that they have evolved certain foraging strategies or rules through natural selection, and attempts to compare expected and observed foraging strategies have led to what has been called Optimal Foraging Theory or OFT for short (Pyke et al., 1977; Pyke, 1984). However, OFT now includes what may be considered to be Classical OFT, where individuals do not respond directly to the foraging behaviour of other individuals, as well as Foraging Game Theory, where individual foragers respond to the observed behaviour of others and thus all individuals are involved in a ‘game’. Consequently, all the chapters in this section embrace OFT when they attempt to explain and predict observed foraging behaviour.

Optimal Foraging Theory – Expansion, Extension, Application and Inspiration OFT, which is now just over 50 years of age, has grown and developed enormously. It has grown at an ever increasing rate in terms of publications and geographic spread of their study locations and authors (Pyke, 2010). It has developed in ways that involve expansion, extension, application and inspiration and have been, in some areas, quite surprising. Consequently, several chapters in this section provide reviews of these developments of OFT.

Section Chapters and What They Cover The principal goal of this foraging section is to cover the issues identified above with chapters focusing on one issue or another, but not necessarily mutually exclusive of one another. We have therefore sought to address the following issues: different foraging stages; variety of decisions that foraging animals must make; use of OFT, including both Classic OFT and Foraging Game Theory; and growth and development of OFT. This is illustrated by the following annotated list of chapters in this section: Optimal Foraging Theory: An Introduction by Graham Pyke (OFT; patch departure). Habitat Selection by Ian Hamilton (Patch choice; OFT; foraging game theory; variance sensitive foraging). Hunger and Satiety: linking mechanisms, behaviour and evolution by David Raubenheimer & Stephen Simpson (OFT; diet; postingestion). Patch Exploitation by Peter Nonacs (OFT; patch choice; patch departure). Animal Movements: An Optimal Foraging Theory Approach by Graham Pyke (OFT; movements). Food Hoarding by Anders Brodin (OFT; food storage – external). Internal Energy Storage by Anders Brodin (OFT; food storage – internal). How Variance and Risk Affect Foraging Behavior by Andrew Hurly & Susan Healy (OFT; variance sensitive foraging). Aposematism as a Defence Against Predation by Christina Halpin & Candy Rowe (OFT; post-ingestion; diet). Group or Social Foraging by Luc-Alain Giraldeau & Graham Pyke (OFT; group membership; producer-scrounger game; foraging game theory). Foraging Behavior as a Cornerstone of Population and Community Ecology by Peter Abrams (OFT extension). Optimal Foraging and Plant Pollinator Co-Evolution by Graham Pyke (OFT extension; game theory). Optimal Foraging Theory: Application and Inspiration in Human Endeavours Outside Biology by Graham Pyke & David Stephens (OFT application & inspiration). The only aspect of foraging that is not included in this section is foraging mode.

Chapter Style and Structure A secondary goal of this section is to provide stories that you, our readers, will find easy to read and understand, and also interesting, informative, compelling and memorable. To make our chapters easy to read and understand, we have tried to adhere to the following simple rule: One point per paragraph, made in its initial sentence, such that anyone can understand the story just by reading the sequence of initial sentences (for further details, set in the context of research excellence and citation success, see Pyke, 2013, 2014). To make our chapters interesting and informative we have aimed to focus on particularly exciting and relevant topics, providing up-to-date perspectives, and have not tried to be totally comprehensive. To make our chapters compelling, we have sought to present clear and logical arguments, justifying significant conclusions. Of course,

110

Foraging j Foraging: Section Overview

all of this should make our chapters memorable but, in addition, we have endeavoured to bring our important ‘take-home’ messages together in our conclusions sections.

Conclusion As foraging is such an important and fundamental aspect of animal behaviour, we need to be able to explain and predict observed foraging behaviour, within the context of different foraging stages (e.g., pre- and post-ingestion) and a number of kinds of decision (e.g., patch choice, diet choice, patch exploitation, movements, group membership, producer versus scrounger, foraging mode). Studies of foraging behaviour, as illustrated by the chapters in this section, can therefore be categorised on the basis of these stages and decisions. To achieve such understanding of foraging behaviour requires adoption of Optimal Foraging Theory (OFT), which includes both Foraging Game Theory as well as Classic OFT, depending respectively on whether or not individual foragers respond directly to the behaviour of others. OFT, which is now just over 50 years old, has grown and developed enormously, and in ways that have sometimes been surprising. The chapters in this section therefore aim to cover all that is foraging by individually focusing on one kind of stage or decision or another, and they all rely on OFT, in one guise or the other, to understand observed foraging behaviour. They have been structured and styled so as to provide stories that you, our readers, will find easy to read and understand, and also interesting, informative, compelling and memorable.

See also: Cognition: Rational Choice Behavior: Definitions and Evidence. Evolution: The Importance of Animal Behavior for Charles Darwin and Other 19th Century Theorists. Foraging: Optimal Foraging Theory: An Introduction. Overview Essays: Game Theory and Animal Behavior.

References Giraldeau, L.-A., Dubois, F., 2008. Social foraging and the study of exploitative behavior. In: Brockmann, H.J., Roper, T.J., Naguib, M., et al. (Eds.), Advances in the Study of Behavior, Vol. 38. Academic Press, pp. 59–104. Pyke, G.H., 1984. Optimal foraging theory: A critical review. Annual Review of Ecology and Systematics 15, 523–575. Pyke, G.H., 2010. Optimal foraging theory – Introduction. In: Breed, M.D., Moore, J. (Eds.), Encyclopedia of Animal Behavior. Academic Press, Oxford, pp. 601–603. Pyke, G.H., 2013. Struggling scientists. Please Cite our Papers! Current Science 105, 1061–1066. Pyke, G.H., 2014. Achieving research excellence and citation success: What’s the point and how do you do it? Bioscience 64, 90–91. Pyke, G.H., Pulliam, H.R., Charnov, E.L., 1977. Optimal foraging: A Selective review of theory and tests. Quarterly Review of Biology 52, 137–154. Visser, A.W., Fiksen, O., 2013. Optimal foraging in marine ecosystem models: Selectivity, profitability and switching. Marine Ecology Progress Series 473, 91–101.

Optimal Foraging Theory: An Introductionq Graham H Pyke, University of Technology Sydney, Ultimo, NSW, Australia; and Macquarie University, NSW, Australia © 2019 Elsevier Ltd. All rights reserved.

Abstract Optimal Foraging Theory (OFT) uses techniques of mathematical optimization to make predictions about foraging behavior which is a fundamental aspect of animal behavior. As it has just turned 50, it is timely to review its foundations, what it has achieved and where it is headed. As an introduction to OFT, I discuss the classic model of patch exploitation, developed by Eric Charnov in 1973, which considers how long a forager should spend exploiting a patch before it moves to a fresh one. A graphical and mathematical approach to this problem led to predictions that were supported by early empirical studies, thus fueling enthusiasm for the optimality approach to understanding animal foraging behavior. OFT, from the outset, also considered foraging decisions regarding patch choice, diet and movements. OFT views foraging behavior as the outcomes of a set of decisions, made continuously, assumes fundamentally that animal foraging decision-making has evolved to the point that biological fitness (i.e., ability to contribute to the next generation) of an individual forager has been maximized, and seeks to understand such decision-making by matching predicted and observed behavior as closely as possible. OFT must therefore be based on models that mathematically describe the foraging processes involved. Because biological fitness is difficult to measure directly, it is generally necessary to adopt a surrogate ‘currency’ and to find the behavior that maximizes it. Examples include the net rate of energy intake and the likelihood of meeting total energy requirements during available foraging time. The classic OFT models have been expanded and extended in many ways. Expansions have included allowing for lack of perfect information, the risk of becoming someone else’s meal, and for decisions to vary with changes in an animal’s state. Extensions have occurred where OFT models have been incorporated into studies of related biological phenomena that involve foraging, such as population dynamics, food webs, and co-evolutionary relationships between nectar-feeding animals and the plants they visit. OFT has also been applied to areas outside the realm of animal feeding behavior and acted as inspiration for solving various optimization problems in human technology. Through all of this OFT has grown enormously and been successful in terms of qualitative predictions, but less so quantitatively. It has demonstrated its usefulness and emerged as a strong theory of behavior and ecology

Keywords Diet; Expansion; Extension; Game theory; Movements; Optimal foraging theory; Patch choice; Patch departure

Introduction Optimal Foraging Theory (OFT) is an approach to the study of foraging behavior that uses the techniques of mathematical optimization to make predictions about this critical aspect of animal behavior (Pyke, 2010). Foraging, which is the process by which animals obtain food, is a fundamental activity for animals, as they require food to sustain their metabolism, provide energy for a wide range of activities, and support reproduction. In some situations, foraging occupies a high proportion of available time, and since animals often cannot do two things at once, increasing the time spent on foraging may reduce the time available for other activities such as mating, resource defense, and predator avoidance. OFT began with consideration of four aspects of foraging: diet, patch choice, patch exploitation, and movement (Pyke et al., 1977; Pyke, 1984; Stephens and Krebs, 1986). It was early recognized that a foraging animal will generally encounter different kinds of food items and must decide whether to consume each encountered item or continue on to the next (Emlen, 1966; MacArthur and Pianka, 1966). It was also recognized that food is generally distributed in patches, rather than uniformly, which means that a forager must decide which food patches to visit (MacArthur and Pianka, 1966) and when to leave its present food patch and move to another (Krebs et al., 1974; Charnov, 1976). It was similarly recognized that when a foraging animal decides to move on from its present location, it must decide on direction, speed and possibly target destination, thus resulting in different patterns of movement while foraging (Siniff and Jessen, 1969; Cody, 1971). As OFT has just turned 50 years of age, it is timely to review its foundations, what it has achieved, and where it is headed. This approach to understanding the foraging behavior of animals began in 1966 with the above-mentioned published articles by MacArthur and Pianka (1966) and Emlen (1966), and has therefore had considerable time to grow and develop. As foraging often mirrors other kinds of search behavior, there is much scope for comparable development of approaches across different kinds of

q

Change History: July 2018. Graham H Pyke made changes to the text and references. This is an update of Graham H. Pyke, Optimal Foraging Theory: An Introduction, Reference Module in Life Sciences, Elsevier, 2017.

Encyclopedia of Animal Behavior, 2nd edition, Volume 2

https://doi.org/10.1016/B978-0-12-809633-8.01156-0

111

112

Foraging j Optimal Foraging Theory: An Introduction

search behavior. As foraging is integral to interactions between animal populations and their resources and to interactions between different species, a theoretical approach to animal foraging behavior such as OFT may also contribute to other areas of biology such as co-evolution, population dynamics and community structure. As an introduction to Optimal Foraging Theory, I shall begin with a discussion of what is now considered the classic model of patch exploitation.

Optimal Foraging: The Classic Model of Patch Exploitation One of the classical problems of foraging theory is how long a forager should spend exploiting a patch before it moves to a fresh one, a foraging situation presented and modelled by Charnov in 1973 (Krebs et al., 1974; Charnov, 1976). Consider, for example, a hummingbird drinking nectar from flowers. When our hummingbird arrives at a fresh flower it obtains nectar quickly, but as it spends more time the nectar becomes harder to obtain because the hummingbird has depleted the supply. Most food patches work this way; fresh patches provide food quickly, but the rate of intake declines as the forager depletes the patch. This simple observation presents a foraging problem, because it takes time and energy to move to a fresh patch. However, the rate of food gain will eventually decline to the point that it is better for the forager to seek a new patch than remain where it is. How a forager decides when it has reached this threshold point is the fundamental issue in this foraging problem. A simple graphical approach illustrates this foraging problem and how it may be solved (see also Charnov, 1976). Fig. 1 shows the idea of patch depletion such that the amount of energy e extracted from the patch increases with the time t an animal spends in the patch (i.e., e¼e[t]), but the instantaneous rate of energy gain (given by the slope of this function) declines steadily; so this gain function e[t] increases but bends down. Now, it takes T units of time for the animal to travel from one patch to another, which is indicated by a distance of -T along the x-axis (which is the time axis). Then the overall rate of energy gain for the forager would simply be the energy gained per patch divided by total time per patch, which is the sum of time spent at each patch and time taken to move to the next patch. In other words, this rate is e[t]/(Tþt), which is the slope of the line that connects the point -T on the x-axis to the point (t, e[t]) on the gain function. The classic patch model finds the optimal patch time, t, which gives the highest rate of energy intake. Of course, this approach assumes that rates of energy expenditure are no different for times spent at a patch and moving between patches, as then the net rate of energy gain, which may be more important to a forager than its gross rate of gain, is simply the gross rate of gain (i.e., gain function in Fig. 1) minus a constant. However, the good news is that relaxing this assumption does not affect the general qualitative conclusions that arise from the model and are explained below. From the simple graphical approach illustrated in Fig. 1, it can be seen that there is an ‘optimal patch time’, such that the rate e [t]/(Tþt) is maximized, when the above-mentioned line (between -T on the x-axis and the point (t, e[t]) on the gain function) is tangent to this function. If you imagine increasing t along the x-axis in Fig. 1, such that the point (t, e[t]) moves along the gain function, the slope of this line, which is the rate e[t]/(Tþt), will increase up to maximum when the line is tangent to the gain function, and then decline. One prediction that arises from this graphical model is that foragers should stay longer within patches when it takes longer to travel to fresh patches. Compare, for example, topt1 and topt2, which correspond to short (T) and long (4T) travel times respectively in Fig. 1. As seen in this figure, topt1 is less than topt2, which indicates generally that topt will increase as T increases, all else remaining constant.

Fig. 1

Graphical representation of optimal departure rule.

Foraging j Optimal Foraging Theory: An Introduction

113

Another prediction from the model is that, with increasing overall habitat quality, as measured by greater average patch quality (i.e., steeper gain function e[t]), foragers should spend less time per patch. To see this, just imagine counter-clockwise rotation, about the origin, of the gain function in Fig. 1. The tangent to this curve will likewise rotate such that the optimum time per patch decreases but the forager achieves a higher rate of energy gain. Early empirical studies supported these predictions both qualitatively and quantitatively, thus fueling enthusiasm for the optimality approach to understanding animal foraging behavior (Pyke et al., 1977). For example, researchers working together at Oxford University observed the foraging behavior of birds in aviaries set up with artificial food patches (Krebs et al., 1974; Cowie, 1977). They found that time spent per patch decreased as habitat quality increased (Krebs et al., 1974) and increased with increasing travel time between patches (Cowie, 1977), in both cases providing qualitative agreement with expectations. In addition, observed average times per patch agreed well with predicted times when allowance was made for different rates of energy expenditure for foraging at patches and moving between patches, and net rather than gross rate of energy intake was considered (Cowie, 1977). The theoretical and empirical results of patch exploitation are further reviewed elsewhere in this encyclopedia. The classic model of patch exploitation outlined above illustrates the following general issues in relation to animal foraging and the optimality approach to understanding its many facets.

Optimal Foraging Theory: The Basic Foundations We can view foraging behavior as the outcomes of a set of decisions, made continuously. As described earlier, an animal can decide whether to stay in a patch or leave it. Foraging animals make many other types of decisions, of course. For example, they decide what types of food to eat, where and when to search for food, and how to move between locations. Such decisions are made continuously, as a foraging animal can always stop what it is doing and do something else. These decisions result in the foraging behavior that we observe. OFT seeks to understand such decision-making by foraging animals by matching predicted and observed behavior as closely as possible, ideally in quantitative rather than qualitative terms (Sih and Christensen, 2001). We might, for example, assume that an animal can determine its average energy yields and its handling times associated with consuming various potential food types when encountered, as well as the average time it spends between successive food items (Pulliam, 1974; Carrillo et al., 2007). Then, based on our own measurements of these variables, we could predict which food types a forager should include in its diet (Pulliam, 1980; Carrillo et al., 2007). This would amount to a quantitative prediction that could be compared with the observed diet (Pulliam, 1980; Carrillo et al., 2007). On the other hand, we might settle for a qualitative prediction, such as that a forager’s diet should expand with decreasing overall food abundance. In general, the extent to which our observations match our predictions would indicate how well we understand the forager’s dietary decisions. OFT assumes fundamentally that animal foraging decision-making has evolved to the point that the biological fitness (i.e., ability to contribute to the next generation) of an individual forager associated with its foraging behavior has been maximized. In this sense, foraging behavior may be considered to have been optimized, and it is appropriate to describe such behavior as being ‘optimal’. Of course, it is always possible that evolution has been proceeding in the direction of increasing biological fitness but maximal fitness has not yet been reached, an issue that I shall return to later. For now, however, let us assume that time has been sufficient for biological fitness to be maximized, and then we can use the mathematical machinery of optimization to critically formulate our predictions about foraging behavior. To apply this logic, OFT must be based on foraging models that mathematically describe the foraging processes involved. In the patch model described earlier, for example, we explicitly imagined a foraging animal moving from one food patch to another and spending time collecting food within each patch. At the same time, we implicitly assumed that food patches are recognizably distinct from one another and that a forager does not re-cross its path as it moves between patches. We could then express the outcome of the foraging process as a simple mathematical function of its ingredients, namely the energy gain function, time spent per patch and travel time between patches, with differences between activities in energetic costs being easily incorporated as well. As I shall discuss below, it is possible to modify the assumptions of this model. It is also necessary, as part of the foraging model, to assume that certain behavioral variables are subject to choice and decision on the part of a foraging animal, as these variables will determine the foraging outcomes. In the case of patch exploitation, described earlier, the forager is assumed able to choose its patch residence time, upon which will depend the rate (net or gross) of energy intake. The next step in the process is to determine, or assume, how variation in foraging outcome affects biological or evolutionary fitness. Ideally, in our patch exploitation example, we would find the fitness, measured in terms of offspring production, associated with a given rate of energy intake and hence be able to relate the fitness of an individual forager directly to its patch residence time. Determining the patch residence time that maximized fitness would then be a straightforward exercise. However, it is generally difficult in practice to measure the relationship between the biological fitness of an individual forager and either the outcomes of its foraging or the underlying behavioral variable that affects foraging outcomes. It is hard, for example, to imagine or implement an experiment in which patch residence time is artificially varied and fitness consequences determined. On the other hand, it may sometimes be possible to vary an animal’s diet and see how this affects its survival, growth, development and reproduction (Wacker and Baur, 2004; Cook et al., 2012; Videla et al., 2012), all components of its biological fitness.

114

Foraging j Optimal Foraging Theory: An Introduction

It is therefore necessary, in general, to adopt a surrogate ‘currency’ as a proxy for biological fitness and to find the behavior that maximizes this currency. We did this, in the example of patch exploitation, by assuming (implicitly at the time!) that the appropriate currency is net rate of energy intake. The adoption of such a currency simplifies things considerably and allows one to estimate the optimal, and hence predicted, foraging behavior. Net rate of energy intake has been frequently adopted as currency for fitness of a foraging animal (Pyke, 2010), but it makes reasonable sense in some situations and probably little or no sense in others. It would seem quite reasonable for nectar-feeding animals that are foraging for floral nectar, which is essentially just an energy source, while doing little or nothing else at the same time and not subject to other factors such as becoming someone else’s meal. However, it would be much less reasonable for animals that forage on foods that differ in nutritional content, and may even contain toxins, or that are subject to varying levels of predation, depending on where, when and how they forage. Some researchers have therefore adopted nutrient combinations as foraging currencies (Schaefer et al., 2003; Jensen et al., 2012) or included predation risk in the currency (Kotler et al., 2016). It is also possible that the important thing to a foraging animal is not the rate of intake of energy or some other nutrient, but rather the amount collected in available time, in which case the forager might benefit by minimizing the probability of suffering a shortfall in this amount. It is possible, for example, that an individual forager’s fitness might hinge simply on whether or not it survives a period, such as night or winter, when it is unable to forage, and hence it should seek to ensure that it obtains sufficient food to survive this period (Carmel and Ben-Haim, 2005; Nolet et al., 2006). A model that incorporates minimization of the likelihood of starvation may lead to different predictions from one based on rate maximization (Carmel and Ben-Haim, 2005; Nolet et al., 2006).

Beyond the Classic Foraging Models Optimal foraging models can and do take many forms, but the classic models are important as starting points. Models can differ in the behavioral decision they consider (e.g., patch use, prey choice, movement), and they differ in how they model the environment (e.g., sequential encounter with resources vs. simultaneous encounter) and in which currency they maximize (e.g., rate of net energy intake vs. probability of survival). Notwithstanding this diversity, we recognize a classic set of foraging models, as illustrated by the patch exploitation model discussed in some detail above, that are important because they serve as starting points for further development. Investigators have expanded and improved the classical models in many ways. For example, a fairly large family of models considers tradeoffs between foraging and other aspects of behavior. The best location for foraging might, for example, be the worst location in terms of the risk of predation. Tradeoffs between foraging and predation risk have been the focus of many recent theoretical and empirical studies (Brown and Kotler, 2007). Also, some studies have considered situations where foraging animals encounter potential food items simultaneously, rather than sequentially (Lima et al., 2003), as was assumed in early models of optimum diet. The classical models also assume that the forager’s behavior is tuned to environmental conditions as if it has perfect information about the properties of the environment such as patch quality, prey quality or encounter rates. Realistically, however, variables like these will often change, and a forager will need to adjust its behavior in response to these changes. Several models have considered the problems of ‘incompletely informed foragers’ (Klaassen et al., 2006; Berger-Tal and Avgar, 2012). Commonly, these models make assumptions about how the environment varies, and consider how experience and information acquisition should influence foraging decisions (Klaassen et al., 2006; Berger-Tal and Avgar, 2012). This approach, therefore, provides an important bridge to other aspects of animal behavior such as learning, cognition, and decision making. Another important trend is the development of so-called dynamic foraging models. In the classical models, we imagine that the animal adopts, for example, a fixed patch residence time that represents the single best choice. Dynamic optimization models suppose, instead, that the best patch residence might change as the animal’s state (e.g., it’s hunger or recent experience) changes (Ydenberg and Houston, 1986; Visser et al., 1992). Instead of predicting a single optimal choice, dynamic models predict an optimal decision trajectory that predicts how decisions might change over the course of a day or some other period, and how this change covaries with a state-variable like hunger (Ydenberg and Houston, 1986). Optimal foraging models have also been extended through incorporation into models of related biological phenomena that involve foraging. Population dynamics, for example, generally depends on the foraging success, or otherwise, of the individuals that make up a population. Interactions between species likewise depend on the foraging behavior of individuals as, for example, when individuals compete for access to the same food resource or members of one species consume those of another species. Models of population dynamics and species interactions should therefore include models of foraging by individuals (GenkaiKato, 2007; Mougi and Nishimura, 2007). Optimal foraging models can also help us to understand co-evolutionary relationships between species, such as those between nectar-feeding animals and the plants they visit, where the foraging behavior of an individual of one species affects both its own fitness and the fitness of individuals of another species. Such co-evolution is clearly indicated by ‘pollination syndromes’, where a trait for various nectar-feeding animal species, such as proboscis length, is correlated with a trait, such as corolla-length, associated with flowers on the plant species visited by each animal species. In this example, the foraging behavior of an animal affects both its fitness through foraging outcomes as well as the fitness of the plants through the process of transferring pollen between flowers. It

Foraging j Optimal Foraging Theory: An Introduction

115

should therefore be possible to consider both optimal foraging and optimal plant traits, but surprisingly few studies have so far been carried out along these lines (Pyke, 2016). Another fascinating development is the application of the optimal foraging approach to phenomena that involve searching but are outside the realm of animal feeding behavior. Engineers, economists, computer scientists, sociologists, anthropologists, archaeologists, psychologists, and many others have all adapted foraging models for their purposes, and assumed that search processes are, in some sense, ‘optimal’ (Rode et al., 1999; Hanning et al., 2010; Wells, 2012). Investigators have, for example, adapted foraging models to consider criminal behavior (e.g., areas with relatively high incidence of burglaries; bird poaching) (Johnson and Bowers, 2004; Pires and Clarke, 2011), military search strategies (Andrews et al., 2007), and how human computer users distribute their time while searching within and among various web sites (e.g., online dating?) (Held et al., 2012). Optimal foraging models can also act as inspiration for solving human optimization processes. For example, imagining how social animals, such as ants, communicate information about foraging conditions and thus adjust their collective foraging behavior, has inspired the concept of ‘swarm intelligence’ which has been used in engineering and other applications (Saber, 2012; Tapkan et al., 2012). It is therefore hardly surprising that the classic set of foraging models has been replaced with a large and growing array of modified and extended models of animal foraging, and that such models have been used or acted as inspiration in a variety of other contexts. Such extensions, applications and inspirations of OFT are discussed in other articles in this encyclopedia.

How Successful has OFT Been? OFT has clearly been reasonably successful in terms of qualitative predictions, but less so when observations have been compared with quantitative expectations. Many studies have, for example, supported the qualitative predictions arising from the above patch exploitation model that time spent per patch should increase with increasing travel time between patches and decrease with increasing average patch quality (Stephens and Krebs, 1986). However, quantitative comparisons between observed and expected residence times per patch have often found significant differences between them, with animal foragers typically spending more time in patches than predicted by the model (Nonacs, 2001). The most common response to such quantitative failure of OFT has been to seek explanations in terms, for example, of model assumptions warranting revision. A foraging animal might, for example, optimally spend extra time in each patch if its risk of suffering predation were greater while moving between patches than while foraging within a patch (Nonacs, 2001). The original patch exploitation model could be modified to allow for this differential predation risk, leading to revised predictions and tests of them (Nonacs, 2001). This would be standard scientific practice. Another response would be to consider any model modification as unacceptable post hoc rationalization, warranting abandonment of the optimal foraging approach (Pierce and Ollason, 1987). However, this conclusion would fail to acknowledge the overall qualitative success of OFT and the fact that marked improvement in agreement between observations and predictions has generally followed reasonable model revision. It would thus be unscientific.

Growth and Prognosis The optimal foraging approach has grown enormously in terms of numbers of publications and continues to grow (see Fig. 2). Beginning in the mid-1960s, the annual number of publications considering foraging theory grew exponentially, especially during the late 1970s, and has continued to grow ever since. Unlike many other areas of research, OFT has not yet begun to show a decline in publication rate. Optimal foraging theory has survived a number of criticisms and passed all the reasonable tests that one could apply to any theoretical approach. Some have criticized it for being on overly simplistic and unrealistic; but most significant conceptual paradigms develop iteratively, improving assumptions and refining models as new data comes to light. Some critics argue natural selection has not had enough time to optimize foraging behavior (discussed in Pyke, 2010). For others, the premise of optimization is valuable and justified, because behavior can evolve relatively rapidly, and because it has manifestly improved our understanding of animal foraging behavior. Other critics, as mentioned above, point to quantitative disagreements between expectations and observations and pronounce the theory dead or a ‘waste of time’ (Pierce and Ollason, 1987). In contrast, proponents point to consistent qualitative agreement and reasonable (but more modest) quantitative agreements with the theory. Investigators have used ideas from optimal foraging theory in several other areas of biology. Ecologists have, for example, used the theory to predict (1) how food density affects consumer behavior (via the so-called functional response) (Abrams, 1992), (2) population dynamics of foraging animals (Svanback and Bolnick, 2005; Mougi and Nishimura, 2007), and (3) species coexistence (Gleeson and Wilson, 1986; Krivan and Diehl, 2005). It has also had a major impact on the area of psychology through its involvement with issues such as learning, memory, and decision-making (Epstein, 1985; Hamblin and Giraldeau, 2009). Optimal foraging theory has therefore demonstrated its usefulness and emerged as a strong theory of behavior and ecology (Marquet et al., 2014).

116

Fig. 2

Foraging j Optimal Foraging Theory: An Introduction

Number of scientific articles relating to Optimal Foraging Theory published per year vs Year period.

See also: Cognition: Rational Choice Behavior: Definitions and Evidence. Evolution: The Importance of Animal Behavior for Charles Darwin and Other 19th Century Theorists.

References Abrams, P.A., 1992. Adaptive foraging by predators as a cause of predator-prey cycles. Evolutionary Ecology 6, 56–72. Andrews, B.W., Passino, K.M., Waite, T.A., 2007. Social foraging theory for robust multiagent system design. IEEE Transactions on Automation Science and Engineering 4, 79–86. Berger-Tal, O., Avgar, T., 2012. The Glass is half-full: Overestimating the quality of a novel environment is advantageous. PLOS ONE 7. Brown, J.S., Kotler, B.P., 2007. Foraging and the ecology of fear. In: Stephens, D.W., Brown, J.S., Ydenberg, R.C. (Eds.), Foraging Behavior and Ecology. University of Chicago Press, Chicago & London, pp. 437–480. Carmel, Y., Ben-Haim, Y., 2005. Info-gap robust-satisficing model of foraging behavior: Do foragers optimize or satisfice? American Naturalist 166, 633–641. Carrillo, C.M., Moreno, E., Valera, F., Barbosa, A., 2007. Seed selection by the trumpeter finch, Bucanetes githagineus. What currency does this arid-land species value? Annales Zoologici Fennici 44, 377–386. Charnov, E.L., 1976. Optimal foraging, the marginal value theorem. Theoretical Population Biology 9, 129–136. Cody, M.L., 1971. Finch flocks in the Mohave Desert. Theoretical Population Biology 2, 142–158. Cook, S.C., Wynalda, R.A., Gold, R.E., Behmer, S.T., 2012. Macronutrient regulation in the Rasberry crazy ant (Nylanderia sp nr. pubens). Insectes Sociaux 59, 93–100. Cowie, R.J., 1977. Optimal foraging In Great Tits (Parus major). Nature 268, 137–139. Emlen, J.M., 1966. The role of time and energy in food preference. American Naturalist 100, 611–617. Epstein, R., 1985. Extinction-induced resurgence preliminary investigations and possible applications. Psychological Record 35, 143–154. Genkai-Kato, M., 2007. Macrophyte refuges, prey behaviour and trophic interactions: Consequences for lake water clarity. Ecology Letters 10, 105–114. Gleeson, S.K., Wilson, D.S., 1986. Equilibrium diet optimal foraging and prey coexistence. Oikos 46, 139–144. Hamblin, S., Giraldeau, L.-A., 2009. Finding the evolutionarily stable learning rule for frequency-dependent foraging. Animal Behaviour 78, 1343–1350. Hanning, C., Yunlong, Z., Kunyuan, H., 2010. Multi-colony bacteria foraging optimization with cell-to-cell communication for RFID network planning. Applied Soft Computing 10, 539–547. Held, C., Kimmerle, J., Cress, U., 2012. Learning by foraging: The impact of individual knowledge and social tags on web navigation processes. Computers in Human Behavior 28, 34–40. Jensen, K., Mayntz, D., Toft, S., et al., 2012. Optimal foraging for specific nutrients in predatory beetles. Proceedings of the Royal Society B-Biological Sciences 279, 2212–2218. Johnson, S.D., Bowers, K.J., 2004. The stability of space-time clusters of burglary. British Journal of Criminology 44, 55–65. Klaassen, R.H.G., Nolet, B.A., Van Gils, J.A., Bauer, S., 2006. Optimal movement between patches under incomplete information about the spatial distribution of food items. Theoretical Population Biology 70, 452–463. Kotler, B.P., Brown, J.S., Bleicher, S.S., Embar, K., 2016. Intercontinental-wide consequences of compromise-breaking adaptations: The case of desert rodents. Israel Journal of Ecology & Evolution 62, 186–195. Krebs, J.R., Ryan, J.C., Charnov, E.L., 1974. Hunting by expectation or optimal foraging a study of patch use by chickadees. Animal Behaviour 22, 953–964. Krivan, V., Diehl, S., 2005. Adaptive omnivory and species coexistence in tri-trophic food webs. Theoretical Population Biology 67, 85–99. Lima, S.L., Mitchell, W.A., Roth II, T.C., 2003. Predators feeding on behaviourally responsive prey: Some implications for classical models of optimal diet choice. Evolutionary Ecology Research 5, 1083–1102. MacArthur, R.H., Pianka, E.R., 1966. On optimal use of a patchy environment. American Naturalist 100, 603–609.

Foraging j Optimal Foraging Theory: An Introduction

117

Marquet, P.A., Allen, A.P., Brown, J.H., et al., 2014. On theory in ecology. Bioscience 64, 701–710. Mougi, A., Nishimura, K., 2007. A resolution of the paradox of enrichment. Journal of Theoretical Biology 248, 194–201. Nolet, B.A., Gyimesi, A., Klaassen, R.H.G., 2006. Prediction of bird-day carrying capacity on a staging site: A test of depletion models. Journal of Animal Ecology 75, 1285–1292. Nonacs, P., 2001. State dependent behavior and the Marginal Value Theorem. Behavioral Ecology 12, 71–83. Pierce, G.J., Ollason, J.G., 1987. Eight reasons why optimal foraging theory is a complete waste of time. Oikos 49, 111–118. Pires, S.F., Clarke, R.V., 2011. Sequential foraging, itinerant fences and parrot poaching in Bolivia. British Journal of Criminology 51, 314–335. Pulliam, H.R., 1974. On the theory of optimal diets. American Naturalist 108, 59–74. Pulliam, H.R., 1980. Do chipping sparrows forage optimally? Ardea 68, 75–82. Pyke, G.H., 1984. Optimal foraging theory: A critical review. Annual Review of Ecology and Systematics 15, 523–575. Pyke, G.H., 2010. Optimal foraging theory – Introduction. In: Breed, M.D., Moore, J. (Eds.), Encyclopedia of Animal Behavior. Academic Press, Oxford, pp. 601–603. Pyke, G.H., 2016. Plant–pollinator co-evolution: It’s time to reconnect with Optimal Foraging Theory and Evolutionarily Stable Strategies. Perspectives in Plant Ecology Evolution & Systematics 19, 70–76. Pyke, G.H., Pulliam, H.R., Charnov, E.L., 1977. Optimal foraging: A Selective review of theory and tests. Quarterly Review of Biology 52, 137–154. Rode, C., Cosmides, L., Hell, W., Tooby, J., 1999. When and why do people avoid unknown probabilities in decisions under uncertainty? Testing some predictions from optimal foraging theory. Cognition 72, 269–304. Saber, A.Y., 2012. Economic dispatch using particle swarm optimization with bacterial foraging effect. International Journal of Electrical Power & Energy Systems 34, 38–46. Schaefer, H.M., Schmidt, V., Bairlein, F., 2003. Discrimination abilities for nutrients: Which difference matters for choosy birds and why? Animal Behaviour 65, 531–541. Sih, A., Christensen, B., 2001. Optimal diet theory: When does it work, and when and why does it fail? Animal Behaviour 61, 379–390. Siniff, D.B., Jessen, C.R., 1969. A simulation model of animal movement patterns. Advances in Ecological Research 6, 185–219. Stephens, D.W., Krebs, J.R., 1986. Foraging Theory. Princeton University Press, Princeton, N.J. Svanback, R., Bolnick, D.I., 2005. Intraspecific competition affects the strength of individual specialization: An optimal diet theory method. Evolutionary Ecology Research 7, 993–1012. Tapkan, P., Ozbakir, L., Baykasoglu, A., 2012. Bees Algorithm for constrained fuzzy multi-objective two-sided assembly line balancing problem. Optimization Letters 6, 1039–1049. Videla, M., Valladares, G.R., Salvo, A., 2012. Choosing between good and better: Optimal oviposition drives host plant selection when parents and offspring agree on best resources. Oecologia 169, 743–751. Visser, M.E., Van Alphen, J.J.M., Nell, H.W., 1992. Adaptive superparasitism and patch time allocation in solitary parasitoids: The influence of pre-patch experience. Behavioral Ecology and Sociobiology 31, 163–171. Wacker, A., Baur, B., 2004. Effects of protein and calcium concentrations of artificial diets on the growth and survival of the land snail Arianta arbustorum. Invertebrate Reproduction & Development 46, 47–53. Wells, V.K., 2012. Foraging: An ecology model of consumer behaviour? Marketing Theory 12, 117–136. Ydenberg, R.C., Houston, A.I., 1986. Optimal trade-offs between competing behavioural demands in the great tit. Animal Behaviour 34, 1041–1050.

Habitat Selectionq Ian M Hamilton, The Ohio State University, Columbus, OH, United States © 2019 Elsevier Ltd. All rights reserved.

Abstract Habitat selection refers to the set of rules individuals use to choose among patches that differ in some way. The outcome of habitat selection is the spatial distribution of populations. The ideal free distribution (IFD) models habitat use when individuals have perfect information about relative fitness payoffs and can move freely. Later models incorporate different forms of competition and other influences on fitness into the IFD and related models. How these rules shape populations and communities can also be determined through curves that specify when to be selective (isolegs) and which habitat to use (isodars).

Keywords Agent-based model; Evolutionarily stable strategy; Game theory; Ghost of competition past; Ideal despotic distribution; Ideal free distribution; Ideal preemptive distribution; Interference competition; Isodar; Isoleg; Predation risk

Introduction Habitat selection refers to the rules used by organisms to choose among patches or habitats that differ in one or more variables that influence its fitness, such as food availability or safety from predators. Patches are relatively homogeneous areas that differ in some way from other parts of the landscape. Perceived patch quality typically increases with increasing food availability and with increasing safety. However, patches with high food availability are not necessarily safer, and so there can be trade-offs between food availability and safety that influence habitat selection. Habitat selection rules determine the spatial distribution of organisms, and can thereby influence population and community-level processes. For example, differences in habitat selection rules between species or competitive classes may allow these groups to coexist. Changes in habitat use in response to changes in resources or perceived risk of predation can result in a variety of direct and indirect effects to predator and prey species, with implications for community structure and dynamics and management. Much of this article focuses on the ideal free distribution (IFD) and its derivatives, which are described in the next section. These models make assumptions about the effects of habitat quality, such as availability of food, and competition with others on fitness and then predict how individuals that maximize fitness are expected to distribute themselves across the landscape.

Conceptual Background The theory of habitat selection is strongly linked with optimal foraging theory. Optimal foraging theory is based on the assumption that animals’ choice of food items or locations is based on rules that optimize some currency that directly influences fitness. In many cases, the currency used in optimal foraging models is the long-term rate of net energy intake (energy gained – energetic costs of foraging). Many habitat selection models are based on game theory. Game theory is used to analyze strategic behavior when the net benefits (or payoff) to a particular action depend on the expected actions of others, such as competitors, social partners or predators. A key concept in game theory is the Nash equilibrium. A strategy is a Nash equilibrium in a game if an individual cannot improve its payoff by switching to another strategy. Evolutionary game theory places game theory in an evolutionary context by modeling whether natural selection will favor a strategy, given the strategies used by others in a population (Maynard Smith and Price, 1973). A key concept in evolutionary game theory is the evolutionarily stable strategy (ESS). If a strategy is an ESS, then if most individuals in the population are using use the ESS strategy, any rare strategy other than the ESS has lower fitness than the ESS strategy.

q

Change History: July 2018. IM Hamilton reorganized the article by combining the summaries and empirical tests with the descriptions of each theoretical model, adding a figure on ideal free habitat use, modifying figure for isolegs and isodars, and adding more examples. This is an update of I.M. Hamilton, Habitat Selection, In Encyclopedia of Animal Behavior, edited by Michael D. Breed and Janice Moore, Academic Press, Oxford, 2010, Pages 38–43.

118

Encyclopedia of Animal Behavior, 2nd edition, Volume 2

https://doi.org/10.1016/B978-0-12-809633-8.90049-9

Foraging j Habitat Selection

119

Theory The Ideal Free Distribution The ideal free distribution (IFD) is a general mathematical model of habitat selection that was introduced by Steven Fretwell and Henry Lucas (Fretwell and Lucas, 1969; Fretwell, 1972). The ideal free distribution model predicts the expected distribution of foragers across resource patches when these patches differ in quality (for example, the productivity of food resources), and foragers compete with one another for resources. Because of competition for resources, patch quality decreases as the density of individuals in a patch increases (i.e., quality is density-dependent) (Fig. 1). The IFD model has as key assumptions that foragers are ‘ideal’: they have perfect information about relative patch qualities and the densities of foragers in each patch, and ‘free’: foragers are able to move between patches without cost or time delay and are not excluded from entering patches by the current inhabitants. As an example, consider two people at opposite ends of a small pond feeding pieces of bread to a large population of ducks. Suppose that one person provides twice as many pieces of bread per minute as the other. Where should ducks feed if they seek to maximize the amount of bread received? A piece of bread eaten by one duck is not available to the others, and so if all the ducks congregated at the more productive location, each individual duck would receive few pieces of bread. At the other end of the pond, the other feeder provides fewer pieces of bread overall, but there is also no competition for that bread. Therefore, some ducks could increase the amount of bread received by moving to that end. By doing so, however, they would increase competition for food at the new location. If each duck forages optimally, that is, it maximizes the number of pieces of bread that it receives over the long term given the behavior of other ducks, how many ducks, would be expected to occupy the more productive location and how many would be expected to occupy the less productive location? The answer, from

Fig. 1 Ideal free habitat selection between two patches. Fitness declines as competitor density increases in either the higher quality patch (solid line) or the lower quality patch (dashed line). In A, total density is low, and all individuals can maximize their fitness by using the higher quality patch exclusively (closed circle). In B, total density is higher, leading to a decline in the fitness of individuals choosing the higher quality patch (closed circle). At this density, some individuals can gain the same fitness by using the lower quality patch (open circle). The equilibrium densities in the two patches, at which fitness gained from using either patch is equal, are shown by the arrows.

120

Foraging j Habitat Selection

an optimal foraging perspective, depends on how food arrives in a patch and how the density of competitors influences the rate at which ducks ingest their food.

Continuous input The continuous-input IFD model is the original model of Fretwell and Lucas (1969). This model assumes that food items arrive randomly at different rates in different patches and that foragers consume them immediately with each getting an equal share of resources on the patch (Parker and Sutherland, 1986). If these assumptions are met, the long-term rate of net energy intake in patch i is the per capita encounter rate with food items, Ri. This rate is the ratio of the rate of resource (food) input (Qi) to the density of competitors in that patch (di) (i.e., Ri¼Qi/di). The model also assumes that foragers should maximize their long-term rate of net energy intake. By assumption, ideal free consumers can identify and move to any habitat that will provide them with a greater intake rate. In this model, there is a unique equilibrium distribution such that no forager can improve its intake rate by switching habitats unilaterally (i.e., it is a Nash equilibrium). If there are two patches that differ in quality, then this point occurs when the ratio of competitor densities in patches 1 and 2, d1/d2, equals the ratio of resource inputs Q1/Q2 (input matching). At this point, the rate of encounter with food items in patch 1 is equal to that in patch 2 (i.e., Q1/d1¼Q2/d2, so R1¼R2). Consequently, the continuous-input IFD model makes two testable predictions. The first is that all patches provide the same net intake rate at equilibrium. The second is that, at equilibrium, consumers should be distributed so that the ratio of consumer densities across patches equals the ratio of patch resource input rates, that is, there should be input matching. Game theoretical analyses of ideal free habitat selection models have shown that using input matching is an evolutionarily stable strategy. That is, if most of the population uses an input matching rule (i.e., the likelihood that an individual chooses a patch is equal to that patch’s quality relative to other patches), then any individual that uses a different habitat selection rule has lower fitness (reviewed in Krivan et al., 2008). Empirical support for the predictions of the continuous-input model of the IFD has been mixed. Several studies found support for the predictions of the IFD (sticklebacks: Milinski, 1979; mallards: Harper, 1982). However, several other studies have not found support for the predictions of input matching and equal intake rates across patches (a review of early studies is presented in Kennedy and Gray, 1994). For example, in another study on mallards, there were there were fewer ducks occupying high-quality patches than the theoretical expectation (Kennedy and Gray, 1994), a phenomenon known as ‘undermatching.’ In a review Kennedy and Gray (1994), the authors find that undermatching is frequently found in tests of the continuous-input IFD. When there is undermatching, foragers in the high-quality patch often have higher intake rates than do occupants of lower-quality patches. Kennedy and Gray (1994) also found some cases of overuse of high quality patches relative to expectation, or overmatching. As this pattern of undermatching is common, it suggests that one or more of the assumptions of this simple continuous-input model are often violated. For example, the assumption of continuous-input resource dynamics could be violated. The effects of changing this assumption are described in the “Interference” section below. Interference models can lead to undermatching when interference is very high, but also commonly predict the opposite effect, overmatching of resource inputs in high-quality patches, so interference alone is unlikely to explain the high frequency of undermatching in tests. Violation of the assumption of perfect information can lead to undermatching and differences in intake rate between patches. Animals might not have perfect information because they are limited by cognitive or perceptual abilities from distinguishing between alternatives that are very similar in payoffs (Gray and Kennedy, 1994). If this is so, there may be some minimal necessary difference in quality, below which animals choose patches with equal probability. This will result in more individuals moving to the poorer patch than would be expected based on patch quality. Deviation from input matching can also result from stochastic variation in input rates (Hakoyama, 2003). If resource input rates are variable and there is some risk of starvation, then the variance of resource inputs may be important influences on habitat use. For example, in one variance-sensitive model of the IFD, underuse of high-quality patches is predicted when the variance in resource inputs is much higher in high quality than low quality patches, and overuse of high quality patches when variance in these patches is low compared to that in lower quality patches (Hakoyama, 2003).

Interference In the continuous-input model, competition over resources results from food items consumed by one individual being unavailable to others. However, competition can occur through other mechanisms, such as aggressive defense of resources, theft of resources from others (kleptoparasitism), or changes in the behavior or defenses of prey such as hiding. All of these are forms of interference competition (Parker and Sutherland, 1986) in which resources are not removed from the environment, but the presence of one individual prevents a second from using the resource. Interference refers to a reversible decline in intake rate with density; this decline is reversible because removal of one individual would make some of the resource again available to others. One way to incorporate interference into a model that is similar to the continuous-input IFD is through the addition of an ‘interference constant’ (m) on searching rate, so that the intake rate is:

Ri ¼

Qi di m

Foraging j Habitat Selection

121

This model is based on a population model by Hassell and Varley (1969). Here, Qi is the foraging rate of a solitary forager. When m is greater than one, changes in density have a stronger effect on intake rate than in the continuous-input model; when m is lower than one, changes in density have less effect on intake rate than in the continuous-input model. The Nash equilibrium point in the two-patch interference model is reached when the ratio of the densities of competitors in patches 1 and 2, d1/d2, equals (Q1/Q2)1/m. Setting d1/d2¼(Q1/Q2)1/m and rearranging leads to the condition Q1/d1m¼Q2/d2m or R1¼R2; in other words, intake rates are again equal across patches at equilibrium. The Hassell-Varley interference model can predict undermatching or overmatching, depending on the value of m. If m>1, so the effects of density on intake rate are stronger than in the continuous-input model, this model predicts undermatching at equilibrium. This is because the effect of density on intake rate is higher in the more productive, and hence more densely occupied, patch than in the less productive patch. If m15 insects, with the colours distinguishing treatment groups (green ¼ high P:C, blue ¼ intermediate P:C and red ¼ low P:C). The first mean in the series for each treatment (i.e., that which falls on the respective nutritional rail) represents the nutrient intake following 48 h in which the animals were confined to the respective food, while subsequent means show the point of cumulative intake over subsequent 4 h periods in which the insects were able to self-select a diet from all three foods. All treatment groups headed in different directions, so as to compensate for their period of constrained imbalance, and by 48 h had converged on the same intake point. They maintained this grouping, and continued to select the same macronutrient balance, until the experiment was terminated at 120 h. Data from Raubenheimer, D., Jones, S.A., 2006. Nutritional imbalance in an extreme generalist omnivore: Tolerance and recovery through complementary food selection. Animal Behavior 71, 1253–1262.

This approach of constructing models that relate animal feeding and foraging to its consequences for the fitness of the animal is equivalent to optimal foraging theory (OFT) (Simpson et al., 2004). NGF differs from much of OFT, however, in several important details of model construction and modelling goals (Raubenheimer and Simpson, 2018). For example, OFT typically considers the important nutritional currency to be a single dietary component, most commonly energy (Stephens and Krebs, 1986), whereas in the NGF the foraging “goal” of animals is to compose a diet that has a particular blend of nutrients (the intake target) (Simpson and Raubenheimer, 2012; Lambert and Rothman, 2015). Further, in OFT models the animal is typically considered to aim at eating as much of the stipulated nutritional currency as it can (e.g., to maximize energy intake), or as fast as it can (maximize the rate of energy intake) (Stephens and Krebs, 1986). In NGF the goal of the animal is not to maximize, but to eat only as much as is needed to satisfy its requirements for the various nutrients (i.e., to reach the intake target). OFT models often do consider particular nutrients in addition to energy, but these are incorporated into the model as factors that might limit the amount of the nutritional currency that can be eaten (e.g., as constraints on energy intake) (Stephens and Krebs, 1986). For example, the opportunity for a moose to eat energy-rich plants might be limited because it needs also to eat energy-poor aquatic plants that enable it to achieve the minimum required intake of sodium (Belovsky, 1978). Finally, OFT models are primarily concerned with functional (i.e., adaptive) aspects of foraging (Stephens and Krebs, 1986), whereas the emphasis in NGF is on the integration of function with behavioural and physiological mechanisms (Raubenheimer et al., 2012).

Foraging j Hunger and Satiety: Linking Mechanisms, Behavior and Evolution

137

Fig. 5 Experiment investigating the link between macronutrient selection and performance in fruit flies (Drosophila melanogaster). Female flies were allocated to one of 28 dietary treatments, which resulted in a wide spread of macronutrient intakes (the small grey dots represent intakes of 1800 individual flies). The lifetime egg production of these flies was then plotted as a performance landscape, where red represents high values and blue low values. In a second experiment, three groups of 25 flies were each given a different combination of nutritionally complementary foods from which they could self-select a diet. Dashed arrows show how flies in the three self-selecting treatments adjusted their intake of separate yeast and sugar solutions to converge on a common nutrient intake trajectory that maximised lifetime egg production. In each case flies had access to a 180 g L-1 solution of sugar, and a yeast hydrolysate solution at either 180 g L-1 (green arrow), 90 g L-1 (white arrow) or 45 g L-1 (pink arrow). This experiment demonstrates dietary self selection, as seen for cockroaches in Fig. 4, and also that self-selected macronutrient balance corresponds with that which maximises fitness. Data are from Lee, K.P., Simpson, S.J., Clissold, F.J., et al., 2008. Lifespan and reproduction in Drosophila: New insights from nutritional geometry. Proceedings of the National Academy of Sciences of the United States of America 105, 2498–2503.

The integration of function with behavioural and physiological mechanisms using geometric models provides a powerful tool for understanding the costs and benefits for animals of achieving different nutritional states, thereby helping to reveal the factors that underlie the evolution of nutritional regulatory systems.

Conclusions “Hunger” and “satiety” can be important concepts in the study of animal behaviour, but care needs to be taken in how these terms are used. They should not be used in the anthropomorphic sense of “the subjective experience of hunger drives an animal to seek food, that animal is seeking food, therefore that animal is experiencing the subjective experience of hunger”. Such anthropomorphic interpretations tell us nothing about the processes that are actually involved in animal foraging. Furthermore, because hunger and satiety cannot be directly observed in animals but must be inferred, they make us vulnerable to the logical errors of fallacy of the converse and circular reasoning. If, however, our notion of “hunger” and “satiety” are based not on anthropomorphism, but on a systematic schema such as the intervening variable concept, then they can provide a useful model for unravelling the causes of foraging and feeding. In this case, these concepts are not considered realities in their own right, but rather frameworks for unravelling the actual causes of behaviour (mechanisms) and the consequences for the animal of having those mechanisms. The consequences to which we are referring are, firstly, behavioural consequences: How the component mechanisms influence behaviour. Secondly, functional (evolutionary) consequences pertain to the question of why the animal evolved the behaviour that it exhibits. Nutritional geometry implements the intervening variables approach in a framework that enables interactions between nutrient-specific appetites to be quantified and systematically linked to mechanisms, on the one hand, and evolutionary function on the other.

See also: Foraging: Optimal Foraging Theory: An Introduction. Host-Parasite Interactions: Nutrition and infectious disease.

138

Foraging j Hunger and Satiety: Linking Mechanisms, Behavior and Evolution

References Belovsky, G.E., 1978. Diet optimization in a generalist herbivore: The moose. Theoretical Population Biology 14, 105–134. Berridge, K.C., 2004. Motivation concepts in behavioral neuroscience. Physiology & Behavior 81, 179–209. Calow, P., 1976. Biological Machines. Arnold, London. Cooper, S.J., 2008. From Claude Bernard to Walter Cannon. Emergence of the concept of homeostasis. Appetite 51, 419–427. Dussutour, A., Latty, T., Beekman, M., Simpson, S.J., 2010. Amoeboid organism solves complex nutritional challenges. Proceedings of the National Academy of Sciences of the United States of America 107, 4607–4611. Fromentin, G., Darcel, N., Chaumontet, C., et al., 2012. Peripheral and central mechanisms involved in the control of food intake by dietary amino acids and proteins. Nutrition Research Reviews 25, 29–39. Garcia, J., Kimeldorf, D.J., Koelling, R.A., 1955. Conditioned aversion to saccharin resulting from exposure to gamma radiation. Science 122, 157–158. de Graaf, C., Boesveldt, S., 2017. The chemical senses and nutrition: The role of taste and smell in the regulation of food intake. In: Tepper, B., Yeomans, M. (Eds.), Flavor, Satiety and Food Intake. Wiley-Blackwell, Hoboken, pp. 35–56. Guthrie, S.E., 1997. Anthropomorphism: A definition and a theory. In: Mitchell, R.W., Thompson, N.S., Miles, H.L. (Eds.), Anthropomorphism, Anecdotes, and Animals. State University of New York Press, Albany, pp. 50–58. Johnson, C.A., Raubenheimer, D., Rothman, J.M., Clarke, D., Swedell, L., 2013. 30 days in the life: Daily nutrient balancing in a wild chacma baboon. PLOS ONE 8, e70383. Kennedy, J.S., 1992. The New Anthropomorphism. Cambridge University Press, Cambridge. Laeger, T., Henagan, T.M., Albarado, D.C., et al., 2014. FGF21 is an endocrine signal of protein restriction. Journal of Clinical Investigation 124, 3913–3922. Lambert, J.E., Rothman, J.M., 2015. Fallback foods, optimal diets, and nutritional targets: Primate responses to varying food availability and quality. Annual Review of Anthropology 44, 493–512. Lee, K.P., Simpson, S.J., Clissold, F.J., et al., 2008. Lifespan and reproduction in Drosophila: New insights from nutritional geometry. Proceedings of the National Academy of Sciences of the United States of America 105, 2498–2503. Lihoreau, M., Buhl, J., Charleston, M.A., et al., 2015. Nutritional ecology beyond the individual: A conceptual framework for integrating nutrition and social interactions. Ecology Letters 18, 273–286. Moran, T.H., Schulkin, J., 2000. Curt Richter and regulatory physiology. American Journal of Physiology - Regulatory Integrative and Comparative Physiology 279, R357–R363. Morrison, C.D., Laeger, T., 2015. Protein-dependent regulation of feeding and metabolism. Trends in Endocrinology & Metabolism 26, 256–262. Morrison, C.D., Reed, S.D., Henagan, T.M., 2012. Homeostatic regulation of protein intake: In search of a mechanism. American Journal of Physiology-Regulatory Integrative and Comparative Physiology 302, R917–R928. Piper, M.D.W., Soultoukis, G.A., Blanc, E., et al., 2017. Matching dietary amino acid balance to the in silico-translated exome optimizes growth and reproduction without cost to lifespan. Cell Metabolism 25, 610–621. Pyke, G.H., 1984. Optimal foraging theory: A critical review. Annual Review of Ecology and Systematics 15, 523–575. Raubenheimer, D., Jones, S.A., 2006. Nutritional imbalance in an extreme generalist omnivore: Tolerance and recovery through complementary food selection. Animal Behaviour 71, 1253–1262. Raubenheimer, D., Lee, K.-P., Simpson, S.J., 2005. Does Bertrand’s rule apply to macronutrients? Proceedings of the Royal Society B 272, 2429–2434. Raubenheimer, D., Simpson, S.J., 2018. Nutritional ecology and foraging theory. Current Opinion in Insect Science 27, 38–45. Raubenheimer, D., Simpson, S.J., Tait, A., 2012. Match and mismatch: Conservation Physiology, Nutritional Ecology and the timescales of animal adaptation. Philosophical Transactions of the Royal Society B: Biological Sciences 367, 1628, 164. Raubenheimer, D., Tucker, D., 1997. Associative learning by locusts: Pairing of visual cues with consumption of protein and carbohydrate. Animal Behaviour 54, 1449–1459. Richter, C.P., 1943. Total self-regulatory functions in animals and human beings. Harvey Lecture Series 38, 63–103. Rothman, J.M., Van Soest, P.J., Pell, A.N., 2006. Decaying wood is a sodium source for mountain gorillas. Biology Letters 2, 321–324. Schoener, T.W., 1971. Theory of feeding strategies. Annual Review of Ecology and Systematics 2, 369–404. Simpson, S.J., James, S., Simmonds, M.S.J., Blaney, W.M., 1991. Variation in chemosensitivity and the control of dietary selection behaviour in the locust. Appetite 17, 141–154. Simpson, S.J., Le Couteur, D.G., James, D.E., et al., 2017. The Geometric framework for nutrition as a tool in precision medicine. The Nutrition and Healthy Aging 4, 217–226. Simpson, S.J., Raubenheimer, D., 2012. The Nature of Nutrition: A Unifying Framework from Animal Adaptation to Human Obesity. Princeton University Press, Princeton. Simpson, S.J., Ribeiro, C., González-Tokman, D., 2018. Insect feeding behavior. In: Córdoba-Aguilar, A., González-Tokman, D., González-Santoyo, I. (Eds.), Insect Behaviour: From Mechanisms to Ecological and Evolutionary Consequences. Oxford University Press, Oxford, pp. 116–130. Simpson, S.J., Sibly, R.M., Lee, K.P., Behmer, S.T., Raubenheimer, D., 2004. Optimal foraging when regulating intake of multiple nutrients. Animal Behaviour 68, 1299–1311. Simpson, S.J., Simpson, C.L., 1992. Mechanisms controlling modulation by haemolymph amino acids of gustatory responsiveness in the locust. Journal of Experimental Biology 168, 269–287. Simpson, S.J., Sword, G.A., Lorch, P.D., Couzin, I.D., 2006. Cannibal crickets on a forced march for protein and salt. Proceedings of the National Academy of Sciences of the United States of America 103, 4152–4156. Simpson, S.J., White, P.R., 1990. Associative learning and locust feeding: Evidence for a “learned hunger” for protein. Animal Behaviour 40 (1990), 506–513. Solon-Biet, S.M., Cogger, V.C., Heblinski, M., et al., 2016. Defining the nutritional and metabolic context of FGF21 using the Geometric Framework. Cell Metabolism 24, 555–565. Sternson, S.M., Eiselt, A.K., 2017. Three pillars for the neural control of appetite. Annual Review of Physiology 79, 401–423. Stephens, D.W., Krebs, J.R., 1986. Foraging Theory. Princeton University Press, Princeton. Toates, F.M., 1986. Motivational Systems. Cambridge University Press, Cambridge. Tordoff, M.G., 2001. Calcium: Taste, Intake, and Appetite. In: Physiological Reviews, 81, pp. 1567–1597 (1459). Walker, S.J., Corrales-Carvajal, V.M., Ribeiro, C., 2015. Postmating circuitry modulates salt taste processing to increase reproductive output in Drosophila. Current Biology 25, 2621–2630.

Further Reading Raubenheimer, D., Simpson, S.J., Mayntz, D., 2009. Nutrition, ecology and nutritional ecology: Toward an integrated framework. Functional Ecology 23, 4–16. Simpson, S.J., Raubenheimer, D., 2000. The hungry locust. Advances in the Study of Behavior 29, 1–44.

Patch Exploitationq Peter Nonacs, University of California, Los Angeles, CA, United States © 2019 Elsevier Ltd. All rights reserved.

Glossary Bayesian Relative to animal behavior, the assumption that animals continually use new information to change their expectations of the environment (and therefore change their behavioral decisions). Giving-up density and time (GUD and GUT) Giving-up density is the amount of food or prey items still remaining in the patch, when a forager leaves it. Giving-up time is the length of time a forager will go without encountering a food item before it leaves a patch. Both are important metrics for testing predictions of the marginal value theorem. Marginal value theorem (MVT) A model within optimal foraging theory that predicts whether an animal should continue to exploit a given patch based on its current (marginal) value relative to the expected gain from moving to another patch. Optimal foraging theory A body of theory that predicts behavior relative to maximizing or minimizing one or a set of goals. Rules of thumb Simple measures that animals can use to approximate solutions to optimal foraging problems. An example would be using the number of prey items encountered to leave patches as predicted by the marginal value theorem. State-dependent model Models that use the techniques of stochastic dynamic optimization to predict animal behavior. Often used to model tradeoffs that animals face when having to decide between competing factors such as getting food and avoiding predators. Stochastic dynamic optimization A mathematical technique that predicts optimal behavior by having computers examine every possible set of behaviors. This produces a numerical rather than an analytical solution as found by the marginal value theorem.

Abstract How animals exploit patches is a basic question in foraging biology. The marginal value theorem (MVT) is particularly useful for addressing this problem. It predicts animals leave patches when immediate intake rates drop below the average across the entire habitat. Because foragers cannot directly solve the MVT, simpler rules of thumb (e.g., leave when no prey are encountered for set periods of time), approximate optimal solutions. Using the MVT allows easily measured metrics (e.g., giving-up densities and patch residence times), to give great insight into foraging decisions. Recent advances have improved on the MVT by adding more reality by including more complex behavior where the forager’s state (e.g., hunger or perceived predation risk) can also affect patch exploitation decisions.

Keywords Giving-up density; Marginal value theorem; Optimality; Patch use; State dependence

Introduction Animals must continually navigate through a heterogeneous world that requires making decisions sometimes on an almost moment to moment basis. A certain area (a “patch”) may have more or less food in it, or may be relatively safe or not from predators. Where the animal finds itself may be ideal for both feeding and safety, or it may need to move to get to a better place. However, one thing is certain – animals need to eat and, therefore, how to find food or how best to forage is under strong evolutionary pressure to be efficient. Natural selection, therefore, has honed a set of hierarchical decisions. Which patch should the animal choose to look for food in? Once there, what set of potential prey or food types should it eat? Finally, when should it leave a chosen patch to go look for food elsewhere?

q

Change History: July 2017. Peter Nonacs cited references in the text provided a reference list. In the previous version there was a section of Gilliam’s Rule. While not disproved or rejected, there does not seem to be much use of it over the last decade, even by Jim Gilliam, so that section and Fig. 5 that went with it (note that former Fig. 6 is now Fig. 5) has been deleted. The discussion of state-dependent behavior and learning which are still active areas of work has been increased and some new examples of what people are doing in these areas has been added. Otherwise, the manuscript is as before except for a bit of editorial polishing here and there. A couple of references have been deleted as they longer fit with the new order. Prose has been edited to better show that the MVT applies well to foragers that are searching for discrete items (e.g., seeds) or draining a food source (e.g., nectar from a flower, body fluids from a prey item. A number of relevant citations and, as requested, some more examples of phenomena have been added. This is an update of P. Nonacs, Patch Exploitation, In Encyclopedia of Animal Behavior, edited by Michael D. Breed and Janice Moore, Academic Press, Oxford, 2010, Pages 683-690.

Encyclopedia of Animal Behavior, 2nd edition, Volume 2

https://doi.org/10.1016/B978-0-12-809633-8.20821-2

139

140

Foraging j Patch Exploitation

Here we consider the last of these questions: The behavioral rules for patch use. This includes how foragers best exploit patches from the point of entering them to when they should leave them. Knowing how and why animals comport themselves in patches can help explain patterns of spatial distribution of populations and how species interact and coexist. So let’s consider how you might exploit your favorite restaurant as a patch. You are foraging on all your favorite dishes, prepared to perfection. Assume the restaurant’s food supplies replenish faster than you can eat and the bill never arrives. In the parlance of foraging, we call this a non-depleting patch with no foraging costs (Fig. 1). Although one might eventually leave the restaurant for reasons unconnected to eating, optimal foraging theory predicts one would never leave this patch expecting to find better food at a lower cost elsewhere. A number of organisms, such as barnacle larvae looking to settle on a rock, may indeed make only one patch choice decision in their entire lives. Getting your settlement timing right and correctly choosing the substrate will determine your survival and lifetime success (Blythe and Jesús Pineda, 2009). Thus for a barnacle, foraging equates to choosing the best non-depleting patch that can never be abandoned. Most animals, however, are not barnacle-like. They move from place to place often as patches deplete (Fig. 1), and in spite of costs that may be associated with moving. In the restaurant analogy, suppose the kitchen runs out of your favorite foods and starts serving less palatable entrees. Or lots of other customers arrive causing the service to deteriorate and increasing the time it takes for food to arrive. Or they start asking you to pay the bill. If one or all these things happen, you may decide to quit this particular establishment and go somewhere else to eat. In terms of foraging jargon, the net rate of energy intake (calculated as the value of the food, minus its cost, and divided by the time it takes to get and eat it) may be better at another restaurant. The difference between non-depleting and depleting patches can be viewed graphically. In a non-depleting situation, a forager’s total net energy intake goes up linearly (Fig. 1). In other words, if it spends twice as long in a patch, it gets twice as much food. In a depleting patch, on the other hand, the longer an animal spends in a patch, the less food it will collect per unit of time. Now let us consider a more natural situation: a bird gleaning insects in a bush. As the bird hunts successfully, it is emptying the bush of edible prey. At what point does it benefit the bird to quit searching a particular bush, and fly to another one? A number of variables could be involved in this decision. First, what is the current capture rate of insects in the occupied bush? Second, is there another bush off in the distance with even more bugs on it? Third, how much does it cost the bird in time and energy to move? Note that all these variables deal only with energetic, food-related issues. The bird could also take into consideration other factors such as does the current bush provide more or less cover from predators relative to some other bush? One solution for the bird is to leave the bush when it is empty and no insects are left. Practically, however, this solution is almost always unworkable. One could almost never be 100% certain that all the insects are gone, or that another has not just arrived. Furthermore, the bird might need to search for a very long time without any food reward to confidently conclude that patch is empty. Certainly, it could be doing better than that! Indeed, the mathematical solution for the best time to leave a depleting patch was derived by Charnov (1976) and is known as the Marginal Value Theorem (often abbreviated to MVT). The basic idea of the MVT is to compare of how well a forager is doing currently versus what its expected success would be if it moved on to be elsewhere. The genesis of the MVT and its future composes the remainder of this article.

Fig. 1 The relationship between time spent foraging and the total amount of energy collected. The slopes of the lines give the encounter rates. If the patch approximates an infinite food supply (i.e., the foragers never seriously deplete the amount of food), then the amount of food collected continues to increase in a linear manner (dashed line). If foraging depletes the patch, then a diminishing return function is expected and encounter rate decreases with time spent in the patch (solid line).

Foraging j Patch Exploitation

141

The Marginal Value Theorem The MVT may be the most predictive and useful equation in the history of the behavioral ecology of foraging. As of the day in 2018 that this sentence is written, Charnov’s (1976) paper has been cited by 2770 other papers (and continues to be cited well over a hundred times a year). This is quite an achievement for a short (7 pages), and primarily mathematical paper! What follows is a brief abstract of the math. We start with the realization that the resource value of a depleting patch (call it patch type i) is a decelerating function of energy gained per time spent in the patch (Fig. 2). Mathematically, we can represent the ever-changing resource gain function curve for patch type i as depending on the amount of time spent in that patch type, or gi(ti). Each patch type is encountered at a specific rate (li). For each patch type the animal spends time ti foraging in a patch before departing for another patch; this is known as the patch residence time. There is a search cost (s) in terms of energy expended looking through the patches. For the mathematically-inclined reader, summing across n patch types, the net rate of resource intake (R) for a forager is: R ¼ ½l1 g1 ðt1 Þþ l2 g2 ðt2 Þþ . þ ln gn ðtn Þ  s=ð1 þ l1 t1 þ l2 t2 þ . þ ln tn Þ

(1)

For the forager there exists a set of n values for all the ti that will maximize R. Rather than solving for all these t values simultaneously, Charnov simplified this problem by assuming that the animal spends the optimal amount of time in patches 2 through n, and solves for the more tractable question of how long it should spend foraging in patch type 1. Therefore, he could then gather all the non-patch 1 terms into two constants, c and k. This gives: R ¼ ½l1 g1 ðt1 Þþ k=ðc þ l1 t1 Þ

(2)

By basic calculus, R is maximized with respect to time, with the t1 value for which the first derivative (dR/dt) equals zero. This results when: g’1 ðt1 Þ ¼ ½l1 g1 ðt1 Þþ k=ðc þ l1 t1 Þ

(3)

where g0 1(t1) is the instantaneous net gain in energy in patch type 1. What this means, to all of us not as mathematically proficient as Eric Charnov, is that the optimal point at which to leave a patch (the left-hand side of Eq. (3)) is when the patch quality has declined to where it is producing food at a rate that is equal to the expected average net rate of energy intake from foraging across the entire habitat with all its patch type encounter rates, search costs and travel expenditures (the right-hand side of Eq. (3)). We can therefore verbally summarize the fundamental MVT prediction as this: “Foragers should stay in a patch until, on average, they would do better by leaving it to find another patch”.

Marginal Value Theorem Predictions: Patch Residence The MVT leads to predictions in terms of how long animals reside in patches, which is called the patch residence time. Another way of restating the above fundamental MVT prediction is that the richer an entered patch is, the longer it will take to deplete to the overall habitat average. This leads to a very intuitive prediction: Foragers should stay in better patches longer than in poorer patches. There is also, however, a somewhat less intuitive prediction. Foragers should leave patches even if prey are still likely to be present or ignore poorer than average patches altogether. To stay too long in a depleted patch or at all in a poor patch will reduce overall foraging return rates relative to moving on. One very useful feature of the above MVT equations is that the optimal solution translates into a beautiful graphical solution (Fig. 2). If the effect of catching and eating prey is to deplete the patch, it will take foragers longer and longer to collect equivalent amounts of energy. Hence, the declining return rate for staying within the patch should eventually make it better to go forage

Fig. 2 The optimal patch residence time when foragers travel between patches and patches decline in quality due to foragers’ actions. The maximum rate of energy intake can be found by drawing the line with the highest slope that intersects the gain function (i.e., the tangent). The intersection predicts the optimal patch residence time.

142

Foraging j Patch Exploitation

elsewhere. Going elsewhere, however, requires traveling between patches. Combining travel costs with the patch gain function creates a possible family of patch residence times which can be represented as lines drawn on the graph from where travel commences to their intersection with the gain curve. Note that lines that do not intersect the gain curve are impossible. Every line, except one, that does intersect with the gain curve will do so at two points. The only line that intersects at just one point is the one that is tangential to the gain curve. This unique line will also have the highest slope of all the lines that cross the gain curve. Mathematically, this is g0 1(t1) – the optimal solution! Using a ruler on Fig. 2, one can quickly visualize the effect that travel distance will be predicted to have on patch residence times. Reduce the travel time, and the tangential intersection point moves to the left on the curve. Increase the time and the intersection point moves to the right. Thus, for a patch of a given quality, the longer the travel time, the longer should the forager stay in that patch. As a second exercise, start with a fixed travel distance and imagine gain curves that either rise more or less steeply on Fig. 2 from where the forager enters the patch. The overall habitat is now either better or worse as the average patch gain function has increased or decreased. Drawing tangents to these new curves reveals another MVT prediction for patch residence: For a given travel distance, patch residence times should negatively correlate with overall habitat quality.

Marginal Value Theorem Predictions: Prey Density If, as the MVT predicts, animals do not reside in patches long enough to capture all potential prey or eat all the available food, how much is left uneaten gives us a valuable metric by which to infer how foragers exploit their environment. This uneaten, but eminently edible, food or prey composes what is known as the giving-up density (GUD). A GUD can come in several different forms. It can be the seeds buried in a tray of sand that were not found (Kotler, 1997); fruits left unparasitized by a fly (Roitberg and Prokopy, 1982); nectar left behind in a flower or plant (Pyke, 1978); or predator only eating a fraction of the prey item it has captured (Sih, 1980: Partial prey consumption). Using GUDs across multiple patches, we can immediately test if an animal adheres to maximizing marginal value. If so, it should leave every foraged-in patch with approximately the same amount of food still in it (i.e., equal to the habitat average). If not, we can further test for factors which may predict a deviation from purely maximizing energy intake rates. Using GUDs as a descriptor of foraging behavior brings several advantages. First off, measuring patch residence times directly may be impossible without disturbing the forager. Thus, the strong connection between time in patch and food eaten means that knowledge about the latter allows inferences about the former. Second, an experimenter can set out food patches with known quantities and then by subtracting the GUD, know exactly how much a forager ate. Finally, manipulation of travel times and habitat qualities should both predictably affect the easily measurable GUD. Thus, many experimental tests of the MVT measure prey densities before and after patch visits and then compare these values across patches or experimental manipulations.

Marginal Value Theorem Predictions: Decision Rules For some foraging type problems, the MVT applies very directly. Imagine you have captured a delicious milk shake and are consuming it through a straw. When the glass is almost empty it takes more time and effort to gain less shake. It is likely that at some point you stop trying and therefore you never consume every last drop. A bee or a hummingbird drinking nectar from a flower or plant faces the exact same problem (Pyke, 1978). For both you and the animals, the MVT predicts abandoning the glass/flower when that instantaneous gain (g0 1(t1) from Eq. (3)) drops below the threshold of it being better to forage elsewhere. We can see interesting examples of this in nature. Early in the flowering season, rufous hummingbirds (Selaphorus rufus) will defend and exploit certain patches of flowers. Unfortunately for the birds as the season progresses there are many more insects competing for that same nectar resource. All the flowers start to resemble milk shake glasses that are invariably almost empty– and the birds stop both foraging on and defending the flower patches (Heinemann, 1992). Similarly, Vucetich et al. (2012) looked at the phenomenon of partial prey consumption of moose by predatory wolves. It turns out that the more moose that wolves can kill, the less they eat from each kill. The wolves are not simply eating until they can eat no more. Instead, when moose kills are likely to be plentiful, and therefore the habitat quality is high, they specialize on only eating the most rewarding parts of the current moose, and then hunting the next one. Of course, a bird at a nectar-rich flower or wolf at a fresh kill will be experiencing a high feeding rate and will be very unlikely to consider leaving the patch at that particular moment. However, in between flowers or moose kills, g0 1(t1)¼0. It would obviously make for a poor overall foraging strategy for a forager to leave patches the moment it has stopped eating. Thus, we have to think of the MVT as not an instantaneous predictor of behavior, but as one where animals need to keep some running average of the amount of food they have collected over some longer time period. Now, although the MVT predicts the optimal economic solution, it is hardly likely to expect that animals can accurately keep track of both their running average for their ongoing feeding rate and the overall quality of an environment composed of many separate patches. Therefore, researchers have suggested that animals might use some sort of ‘rule of thumb’. This would be a simpler and easier to remember metric that may be a close approximation of the MVT values. For example, “Stop sucking on your milk shake straw when you are getting more air than shake.” Or for a hummingbird – abandon a plant if nectar obtained from last flower is less than some threshold amount; otherwise try another flower on the same plant (Pyke, 1978).

Foraging j Patch Exploitation

143

Fig. 3 Rules of thumb for leaving patches. With a fixed time rule (a), the forager stays for a set length of time in each patch. Thus, the number of prey eaten may vary across patches. With a fixed number of captures rule (b), the forager stays in each patch until it captures a set number of prey items (3 in this example). Thus, time spent in each patch may vary. With a giving up time rule (c), the forager leaves when it fails to encounter a prey item for a given period of time. In this case, both the number of prey captures and patch residence time may vary across patches.

Three general rules of thumb are proposed for how foragers could exploit patches. Each one posits a simple behavioral mechanism that could be relatively easily used by a wide range of species. They are: (A) A fixed time rule (Fig. 3(a)), where an animal spends a set length of time in each patch and then moves on, regardless of how successful it has been: (B) Leave after encountering a fixed amount of food (Fig. 3(b)) Leave if nothing of food value has been encountered over some critical time period of searching (Fig. 3(c)). If all the patches in a habitat are more or less about the same quality, then the MVT would predict that animals ought to spend about the same time in them resulting in about the same number of prey captures or amount of food in each patch. If an animal is faced with such a fairly uniform foraging problem, it is likely that simply allocating a fixed time spent foraging in each patch or leaving after a fixed number of prey or amount of food has been encountered could work very well. The forager would not have to track either its change in feeding rate or estimate average habitat quality. Fixed time and number rules, however, would have serious drawbacks in environments that have higher variance in quality across patches. A fixed time rule could cause foragers to leave too quickly from very good patches and a fixed amount rule could trap foragers for long periods of time in rather poor patches. In variable environments a better rule of thumb is to have a giving-up time (GUT), where the forager leaves the patch based on when it last encountered a prey item or ate. GUT models have several variants with this time expected to either be fixed or to change over time. The simplest expectation is a constant time rule: the animal leaves whenever a fixed period of time has passed without a prey capture. Whenever a capture is made the forager’s “clock” is reset to its original value and begins counting down again. However, if patches are likely to be depleted rapidly due to the foragers’ activities, it may be more advantageous for the animal to have a constantly decreasing fixed time interval. A prey capture resets the clock, but to new, shorter time. Conversely, if good patches are rare but not rapidly depleting, then it may be more advantageous for the fixed time to increase with every capture. Thus, a short run of bad luck would not cause a forager to abandon a still relatively rich patch. Foragers who do modify their giving-up times are acting like “Bayesian statisticians” in their decision-making behavior. In other words, such foragers are continually updating or reevaluating their expectations of environmental quality. A review by Valone (2006) suggests Bayesian abilities and behavior based on learning and updating one’s expectations about the environment is widely distributed across animal species.

144

Foraging j Patch Exploitation Table 1

Tephritid fruit fly foraging behavior as a function of host density

GUT (min) No. of leaves visited

1.6 m

3.2 m

N

16.4 59.3

22.1 89.3

32.0 103.4

Source: Flies were released into three types of habitats that differed by having the trees spaced 1.6, 3.2 m, or out of the range of the fly’s eyesight (N). The MVT predicts that as flies perceive greater travel times between patches, they stay longer in patches (with longer GUTs) and search more thoroughly. The results support both predictions.

Testing the Marginal Value Theorem: Patch Residence times Whichever rule of thumb an animal may use, a consistent prediction of the MVT is that longer travel times between patches should increase patch residence times. If nothing within the patches is changed, longer travel times by themselves would result in the overall reward rate of the habitat declining because the increased time spent traveling means fewer prey captures. As we have seen in the context of the MVT, poorer habitats select for longer patch residences. Roitberg and Prokopy (1982) tested the travel time prediction in the foraging behavior of tephritid fruit flies on hawthorn fruits. Female flies lay eggs on the mature fruit, which they use vision to locate. In the experiment, they released the females on hawthorn trees that had a fixed number of fruit. Surrounding the test tree the authors placed more trees that were either 1.6 m or 3.2 m distant. In a third treatment there were no other trees within sight for the flies. Thus the experiment gave females access to patch of known and constant quality, but with visibly different travel distances to the next patch. The MVT predicted that: (1) The females would stay longer with greater GUTs in a tree as their perception of the distance to the next patch increased, and (2) The females would spend this time searching more leaves, even though their success rate per time spent searching would decline. Both predictions were supported (Table 1): Flies had a longer giving-up time in individual trees and the number of leaves visited increased. Roche and Glanz (1998) found that black-capped chickadees would also adjust GUT’s depending on their evaluation of the initial patch quality. In a poor patch, the birds came to ‘expect’ that the intervals between prey encounters were likely to be long, while in good patches encounters occur in close succession. Thus, birds could more easily detect patch depletion in the better patches as the decrease in encounter rates is more easily perceived. The result was that birds had shorter GUT’s in the better patches than in the poorer ones.

Testing the Marginal Value Theorem: Giving up Densities The measurement of giving-up density of food is one of the main methodologies in behavioral ecology for inferring patch exploitation decisions. Simply knowing how much food an animal leaves behind in a patch can lead to inferences about overall habitat quality, predation risk, and competition might be resolved. As an example for how GUDs can give insight to how these three factors interact, Burt Kotler (1997) used two species of gerbils (genus Gerbillus). In a series of experiments, he set out trays of seeds mixed in sand that both species would happily eat. The gerbils had to dig to find the seeds. After a night of foraging, Kotler collected the trays and counted the seeds left behind (i.e., the GUD). In one set of trials, the trays were the only food patch available. In a second set of trials he provided “free” food at same time he set out the sand trays. This free food was piles of seeds without the sand. Such easily collected food made the sand-tray patches relatively less valuable: Why dig when you don’t have to? In the context of the MVT, free food increased the overall quality of the habitat. You can visualize this as increasing the steepness and height of the gain function for the habitat average in Fig. 2. The tangent drawn on such line would predict animals should decrease the time spent in each patch. Less time equals fewer seeds eaten equals higher GUD. Supporting this MVT prediction, both species of gerbils left significantly more seeds uncollected in the sandy trays (Fig. 4). Along with measuring the effects of overall habitat quality, giving-up densities demonstrated the effect of predation risk. Small nocturnal rodents tend to view open locations suspiciously because being in the open makes them more susceptible to attack from owls. Consistent with such a risk, Kotler found significantly higher giving up densities in trays in the open rather than under bushes. Interestingly, in the absence of free food (Fig. 4), one species (G. allenbyi) had a significantly higher giving-up density than the other species (G. pyramidum). This strongly suggests there is also a species-level difference in responding to risk, with the smaller species, G. allenbyi, placing a higher premium to foraging in safer areas. Thus, GUDs also provided Kotler with a significant insight into how these two competing species use and segregate themselves in their shared habitat.

Testing Patch Departure Rules Most environments are variable and therefore, one would expect that a giving-up time rule of thumb might be the most useful. It is difficult, however, to determine the degree to which animals actually measure or keep track of time itself. Thus, testing for GUTs is more along the lines of how times to encounter correlate with changes in behavior. A variety of studies have, indeed, found evidence that foragers behave at least as if they are keeping track of the time between food encounters. For example, Lefebvre et al. (2007) found that bumblebees increased their tendency to leave patches after each encounter with a rewarding flower (i.e., GUTs grew

Foraging j Patch Exploitation

145

Fig. 4 The mean giving-up densities for two species of foraging gerbils. Giving-up density (as grams of uneaten seeds) is measured in sand trays containing seeds under two conditions: when piles of seeds not buried in sand are either absent or present. Both species forage similarly, with higher GUDs, when free seeds are present. In the absence of free seeds, however, the larger species, Gerbillus pyramidum (solid bars), took significantly more seeds from the sand trays than did the smaller species, G. allenbyi (striped bars).

shorter). In essence this fits with the presumption that the bee views each success as depleting the patch and therefore making the patch less likely to be rewarding in the future. Wajnberg (2006) in a review across an extensive set of studies on parasitoid wasps searching for hosts to parasitize found the results to be supportive for a variety of patch departure rules. In some studies, there were no effects on the likelihood to leave a patch across sequential encounters – consistent with fixed number or time rules of thumb. However, more often encounters did affect the time wasps spent in patches, as would be consistent with GUT rules. Sometimes the effect was to increase patch residence – consistent with what would be expected if wasp perceived this to be a good patch in a world of generally poor patches. Other times, like Lefebvre‫׳‬s bees, encountering prey made the wasps more likely to leave – as if they were perceiving patch depletion. Interestingly, recent work has shown that parasitoid wasps can associate an actual time interval as the metric for an attractive stimulus (Parent et al., 2017). Thus, it appears that even insects may be able to keep track of time! To review all of the work done on how animals decide to leave patches is beyond what can be accomplished in this article. It is safe to say, however, that the “rules” foragers employ vary across species and likely within species relative to the habitats they encounter. However, as Wajnberg (2006) noted, the majority of studies have been done under laboratory or artificial conditions. Field work under natural conditions and in the presence of competitors is greatly needed.

Beyond the Basics of the Marginal Value Theorem The MVT is all about maximizing the rate of collecting food. Clearly, however, animals should care about a lot of other things, too. Patches may also vary in their safety from predators, presence of aggressive competitors, or opportunities to encounter future mates. Failure to consider that foragers might have simultaneous and perhaps conflicting goals can create substantive deviations in observed behavior in relation to MVT predictions. For example, I found that across 26 different studies from a wide range of taxonomic groups, MVT predictions were qualitatively strongly supported (Nonacs, 2001). Patch residence time tended to increase or decrease in the appropriate direction as the environment was changed. Quantitatively, however, only 3 of the 26 studies made accurate predictions. In 19 of the 23 misses, animals stayed longer in patches than predicted. If we expect that foragers cannot always be perfect in their decisions, then quantitative errors are to be expected. However, if they simply reflected perception errors they would also be expected to occur equally both directions – over and under staying. The fact that a significant majority of ‘misses’ are in animals staying too long suggests that foragers get something more from patches than just food. One such second benefit from being in a patch is that the longer a forager is in a patch without being attacked, the safer that patch is likely to be. When I included this added dimension of safety to an MVT food gain model, I found results that looked more like what animals have been observed to do (Nonacs, 2001). The fact that patch use and its metrics such as GUDs may reflect a variety of competing tradeoffs, does not mean the MVT and GUD’s are irretrievably flawed for explaining foraging behavior. Addressing both the costs and benefits of GUD data, Bedoya-Perez et al. (2013) presented a “Practical Guide” on how to use GUDs and avoid misinterpreting outcomes. For example, GUDs may be strongly affected by the immediate presence of competing foragers. One would not always expect the same outcome when a group forages together in comparison to where each group member foraged alone. A member of this research team, Alexandra Carthey,

146

Foraging j Patch Exploitation

then led a follow-up study that found groups of rats exploited patches to a greater degree than solo foragers, leaving behind lower GUDs (Carthey and Banks, 2015). This example illustrates that patch exploitation models are as applicable to group-foraging species as they are to solitary foragers. One simply must factor in the dynamic of within-group competition.

The Problem of State-Dependency Animals and their environments vary and are dynamic, but Charnov’s original marginal value theorem model is static in that it predicts the behavior of the average animal responding to the average state of the environment. However, animals come in all ages, sizes and conditions. Some are hungry, others are well-fed. Some have offspring to feed, others are foraging only for themselves. Moreover, the environment is dynamic, too. There are longer term seasonal changes and shorter term good and bad days. Clearly, the ‘state’ of the animal and its habitat ought to affect the foraging process. One can certainly add to Eq. (1) other factors such as the possible presence of competitors or predators within patches. Unfortunately, these factors may interact in complex and opposing ways (e.g., a safe patch may also be poor as regards food). If you already find Eqs. (1)–(3) daunting, then realize that adding more factors will increase the mathematical impenetrability by at least an order of magnitude. Thus, there is little hope that there will be a more inclusive equation that is as comprehensive and intuitive as our current MVT. With no hope of an analytical solution to guide us, we need a different type of modeling approach. Fortunately, we can harness the ever-increasing power of computers to solve complex problems numerically. So if animals and environments come in a variety of states of hunger, body condition, predation risk, food abundance, or whatever; the computer can then consider every possible combination of states and solve across all possible behaviors, as to what the best decision is. This approach has many names, but most often is known as state-dependent modeling or stochastic dynamic optimization (Clark and Mangel, 2000; Mangel and Clark, 1988).

Patch Exploitation With Predators, Time Constraints and Learning As described before, when I added the possibility that patches are valued for other benefits in addition to food, deviations from MVT predictions became more understandable (Nonacs, 2001). The model itself employed stochastic dynamic optimization, whereby foragers had particular states (e.g., hunger level, estimation of predation risk, etc.) and the goal of foragers was to maximize their survival or mating success rather than simply collect calories at the highest rate. A specific example of the difference between the predictions of static and dynamic models is evident in the oviposition behavior of parasitoid wasps. A female wasp lays her eggs on a host. The size and survivorship of her offspring depend on the size of the host and the number of eggs she lays on it. Hosts can easily support more than one parasitoid, but if mother puts too many eggs into a host, the offspring will compete too much to the detriment of all. This decision is, therefore, easily put into the context of the MVT and for each size of host Charnov and Skinner (1984) could predict the optimal number of oviposited eggs that maximizes the parasitoid wasp’s reproductive success. However, when the model was tested, the actual clutch size laid by Nasonia vitripennis females only partially supported MVT predictions. Females rarely exceeded the predicted optimal number of eggs per host, but quite often laid far fewer eggs than that optimum. When the same situation was modeled with a dynamic state-dependent model by Mangel and Clark (1988) and Clark and Mangel (2000), they also predicted the same maximum number of eggs laid as had Charnov and Skinner. However, their model also predicted a distribution of clutch sizes that was far more similar to the observed (Fig. 5). Mangel and Clark succeeded by being able to include two real-world complications: A finite number of eggs that could be produced by a female wasp per time period and time effects (each foraging session is ended by onset of nighttime; the wasps grow older; and winter approaches). Basically, if a wasp is nearing the end of her life or night time is fast approaching, it is better to lay fewer than the optimal number of eggs rather than not to lay at all. Another common state to consider is what does a forager know about its environment? Knowledge is useful when it can be used to make better decisions on how to exploit patches. A variety of experiments and models have shown the value and effects of having information about patch quality. In a classic experiment, Lima (1984) showed that downy woodpeckers are great Bayesian statisticians – they sample an empty food patch just long enough to generate a significant level of certainty that it is indeed empty. More recently, Katz and Naug (2017) showed that better-fed honey bees value immediate food rewards less and instead, devote more time and effort to exploring their habitat. They are willing to give up some immediate food payoff to learn more about their entire habitat. Berger-Tal and Avgar (2012) showed that for naïve foragers, optimism (initially over-estimating the overall quality of an unknown habitat) is a better learning strategy than pessimism. Optimists leave patches earlier and therefore encounter more patches more quickly than pessimists. This means that optimists gather information faster and learn the true habitat quality faster. One of the things that a forager may learn is what to expect in the future. For example, do you expect the environment to be fairly constant or will there be significant variance? Berger-Tal et al. (2014) showed that gerbils change their patch exploitation based on their lifetime experiences. At the beginning of the experiment, gerbils were presented with a generally rich environment. Then it was switched to being poor, and as predicted, the animals utilized the better patches relatively more. However, a switch back to the original conditions did not result in the original patch use patterns. The gerbils instead spent more time foraging and harvested more food from the patches. They had learned the world occasionally turns bad, and you had better stock up to prepare for it.

Foraging j Patch Exploitation

147

Fig. 5 The predicted and actual behavior of parasitoid wasps encountering prey of 4 different size ranges. Upon encountering a suitable host, a wasp can lay from 1 to over 40 eggs on the host. The grey bars are the percentages of optimal clutch sizes (¼ eggs per host) predicted by a dynamic model that takes into consideration time of day and season, and different possible states and previous foraging experiences across wasps. The striped bars are the percentages predicted by a static model for each prey size that assumes a constant behavioral and physiological state for the wasps. The black dots are the observed percentages of clutch sizes for ovipositions on 36 hosts. The dynamic model fits the observed data better by predicting a range of small clutch sizes.

Conclusion Optimal foraging theory has had a number of notable successes in helping behavioral ecologists understand how animals behave in terms of finding food or avoid becoming food for someone else. This article has focused on basically one aspect: How best to exploit a patch and the foundational model for understanding that problem, the marginal value theorem. The MVT provides two extremely useful metrics for studying animal behavior: Patch residence time and giving-up density. Both allow us to explore how animals value patches not only relative to food, but to other factors such as predation risk and the value of information. Decision rules based on simple measures, such as time from last prey capture, can reasonably approximate optimal solutions and do appear to match animal behavior. Finally, systematic errors in or deviations from MVT predictions have led to creating more sophisticated and accurate models that can incorporate predation risk, the need to learn about environments, and the animals’ current physiological state. The influence that the MVT has had throughout behavioral ecology shows no signs of abating. A Web of Science search turns up that the term “marginal value theorem” was used eight times in 1977 across journals that publish in evolutionary, ecological or behavioral sciences. In 1987, the number of citations in that year was 67. In 1997, 217 citations. In 2007, 530 citations. And finally in 2017, the per year number has climbed to 955. The MVT has even entered the classroom as a learning problem for budding ecologists. Clark and Begley (2015) developed a game where undergraduates must assume the roles of owls (predators) or skunks (prey). All players need to accumulate resources to survive by foraging in patches and at various points must decide to either stay where they are, or move to another patch. The students’ decisions are based on whether or not they think they are in a good place from their perceptions of their patch’s current resources, competitors and predators (and, of course, if the player is a predator or prey). The last surviving owl and skunk win the game. The increasing relevance of the MVT follows from the simple but important problem it addresses: How to best exploit a patch. This frames one of the most fundamental questions in all of behavioral biology: “Is it better to stay where you are now or to move on to elsewhere?” The many ramifications and complications of that question will keep behavioral ecologists quite busy for the foreseeable future!

See also: Foraging: Optimal Foraging Theory: An Introduction. Predator-Prey Interactions: Anti-Predatory Vigilance; Predation Risk and Life Histories.

References Bedoya-Perez, M.A., Carthey, A.J.R., Mella, V.S.A., McArthur, C., Banks, P.B., 2013. A practical guide to avoid giving up on giving-up densities. Behavioral Ecology and Sociobiology 67, 1541–1553. Berger-Tal, O., Avgar, T., 2012. The glass is half-full: Overestimating the quality of a novel environment is advantageous. PLOS ONE 7, e34578.

148

Foraging j Patch Exploitation

Berger-Tal, O., Keren, E., Kotler, B.P., Saltz, D., 2014. Past experiences and future expectations generate context-dependent costs of foraging. Behavioral Ecology and Sociobiology 68, 1769–1776. Blythe, J.N., Jesús Pineda, J., 2009. Habitat selection at settlement endures in recruitment time series. Marine Ecology Progress Series 396, 77–84. Carthey, A.J.R., Banks, P.B., 2015. Foraging in groups affects giving-up densities: Solo foragers quit sooner. Oecologia 178, 707–713. Charnov, E.L., 1976. Optimal foraging: The marginal value theorem. Theoretical Population Biology 9, 129–136. Charnov, E.L., Skinner, S.W., 1984. Evolution of host selection and clutch size in parasitoid wasps. Florida Entomologist 67, 5–21. Clark, J.M., Begley, M.T., 2015. Fight for life: A classroom game of foraging, patch selection, and risk. American Biology Teacher 77, 693–698. Clark, C.W., Mangel, M., 2000. Dynamic State Variable Models in Ecology. Oxford University Press, Oxford. Heinemann, D., 1992. Resource use, energetic profitability, and behavioral decisions in migrant rufous hummingbirds. Oecologia 90, 137–149. Katz, K., Naug, D., 2017. Energetic state regulates the exploration–exploitation trade-off in honeybees. Behavioral Ecology 26, 1045–1050. Kotler, B.P., 1997. Patch use by gerbils in a risky environment: Manipulating food and safety to test four models. Oikos 78, 274–282. Lefebvre, D., Pierre, J., Outreman, Y., Pierre, J.S., 2007. Patch departure rules in bumblebees: Evidence of a decremental motivational mechanism. Behavioral Ecology and Sociobiology 61, 1707–1715. Lima, S.L., 1984. Downy woodpecker foraging behavior: Efficient sampling in simple stochastic environments. Ecology 65, 166–174. Mangel, M., Clark, C.W., 1988. Dynamic Modeling in Behavioral Ecology. Princeton University Press, Princeton, NJ. Nonacs, P., 2001. State dependent patch use and the marginal value theorem. Behavioral Ecology 12, 71–83. Parent, J.P., Takasu, K., Brodeur, J., Boivin, G., 2017. Time perception-based decision making in a parasitoid wasp. Behavioral Ecology 28, 640–644. Pyke, G.H., 1978. Optimal foraging in hummingbirds: Testing the marginal value theorem. American Zoologist 18, 739–752. Roche, J.P., Glanz, W.E., 1998. Temporal characteristics of foraging movements in Black-capped chickadees. Journal of Field Ornithology 69, 603–613. Roitberg, B.D., Prokopy, R.J., 1982. Influence of intertree distance on foraging behaviour of Rhagoletis pomonella in the field. Ecological Entomology 7, 437–442. Sih, A., 1980. Optimal foraging – Partial consumption of prey. American Naturalist 116, 281–290. Valone, T.J., 2006. Are animals capable of Bayesian updating? An empirical review. Oikos 112, 252–259. Vucetich, J.A., Vucetich, L.M., Peterson, R.O., 2012. The causes and consequences of partial prey consumption by wolves preying on moose. Behavioral Ecology and Sociobiology 66, 295–303. Wajnberg, E., 2006. Time allocation strategies in insect parasitoids: From ultimate predictions to proximate behavioral mechanisms. Behavioral Ecology and Sociobiology 60, 589–611.

Animal Movements – An Optimal Foraging Theory Approach Graham H Pyke, University of Technology Sydney, NSW, Australia; and Macquarie University, NSW, Australia © 2019 Elsevier Ltd. All rights reserved.

Abstract Foraging animals, in addition to deciding where and when to forage, what to feed on, and how long to spend in one area before departing for another, must also decide how to get from one location to another or what movement strategy to adopt. If you observe foraging animals you will generally see them moving or changing location, exhibiting patterns in terms of distances and directions as they go and patterns in terms of where they end up spending their time. In attempting to understand such patterns, Optimal Foraging Theory (OFT) may help. Early attempts, during the 1970s and early 1980s, to use OFT to understand movements of foraging animals generally took a Cognitive Forager Approach (CFA) in which animals were assumed to be aware of their internal state, have a sense of direction and thus able to maintain directionality to their movements, an ability to sense potential food items at a significant distance, and memory regarding previous foraging. They also often assumed that food is patchily distributed, with multiple items tending to occur relatively close to one another (e.g., seeds on the ground) or nearby food locations (e.g., flowers) tending to have similar amounts of food. More recently, some attempts to use OFT to understand movements of foraging animals have adopted the Lévy foraging hypothesis (LFH), which is the antithesis of the CFA. According to the LFH, as originally conceived in the late 1990s, animals employ a simple random walk (i.e., no directionality) to search for randomly distributed food items in a featureless environment and in a completely uninformed manner, with no sense of direction, no ability to perceive food items unless they ‘bump’ into them, and no memory regarding previous circumstances. In other words, animal foragers are assumed to be clueless, senseless and uninformed. The LFH should be abandoned in favour of the CFA, despite the former receiving considerable acclaim from its proponents. The assumptions behind the Lévy foraging hypothesis bear little resemblance to biological reality, as no foraging animal is clueless, senseless and uninformed, and thus its predictions lack validity. The LFH also unrealistically omits directionality of movements, where successive movement segments tend to be in the same direction, and Area Restricted Search, where food encounter or encounter with a relatively high amount of food is followed by increased turning and shorter movement segments which tend to keep an animal in the vicinity of encountered food. In addition, Lévy-like behaviour is hardly surprising, arising as an emergent property of many natural processes, and hence tells us little to nothing about how and why animals forage in the ways they do. Since about 2005, the CFA has been reborn as a revamped ‘movement ecology’ with models labelled as agent-based models (ABM) or individual-based models (IBM). However, these models have been primarily descriptive rather than theoretical, and not so far resulted in the development and testing of predictions. Future research on movements by foraging animals should therefore focus on situations where foraging can be differentiated from other behaviour, what the animals can perceive and remember is reasonably clear, food encounter can be recorded, the abilities of animals to collect and consume encountered food are reasonably clear, the spatial and temporal distribution of food can be determined, and everything can be expressed as a mathematical model. Model assumptions can then be refined through comparing emergent patterns with observed patterns of foraging movement and optimality predictions can be compared with observations. In the end, we should better understand why foraging animals move the way they do.

Keywords Area restricted search; Cognitive forager approach; Directionality; Fitness maximization; Lévy foraging hypothesis; Movement strategy; Optimal foraging theory

Introduction Anyone can easily observe animals moving, and thus changing their location, and often foraging is clearly or likely to be involved. If, for example, you walk around a local garden or park you will probably see birds moving across areas of mown grass, pausing every so often to peck at the ground, sometimes grabbing and swallowing some morsel of food. In some places, such as Sydney, Australia, where I live, you may see large numbers of fruit bats, emanating at dusk from daytime roosting sites and spreading out across the sky as they head to feeding areas. Or if you are simply looking at flowering plants, you may see flower-visiting birds, bees and many other kinds of animals, move from flower to flower, and plant to plant, harvesting nectar, pollen or some other floral resource along the way. In all these examples, and many others, obtaining food is one reason for why these animals are moving. Of course, animals may also alter their posture and orientation, and such changes may also be food-related, but here I consider as movements only changes to the location of an animal’s entire body. If you continue watching your animals as they move about, you may observe patterns to their movements, especially in terms of movement distances and directionality. Bees, for example, generally move from one flower or plant to a near neighbour, rather to one more distant (Pyke, 1978c, 1979). Bees also seem to generally avoid reversing their ongoing direction and returning to the

Encyclopedia of Animal Behavior, 2nd edition, Volume 2

https://doi.org/10.1016/B978-0-12-809633-8.90160-2

149

150

Foraging j Animal Movements – An Optimal Foraging Theory Approach

flower or plant that they have just visited, and are more likely to leave a flower or plant in about the same direction as their arrival (Pyke and Cartar, 1992). In other words, there is directionality to their movements. Bees are fascinating and remarkable animals to watch! You may like to try watching your favourite animal, to see if you can observe such movement behaviour. It is then only natural to seek explanations for observed movement patterns and this is where Optimal Foraging Theory (OFT) may prove useful (Pyke, 1983). OFT, in the context of movements of foraging animals, posits that foraging animals adopt rules of movement, subject to various constraints, that maximise net gains in biological fitness. However, there may be constraints in terms, for example, of available information, cognitive abilities (e.g., perception, memory), morphology and energetics. For some animals, especially those that feed on floral nectar, it may be reasonable to equate biological fitness with Net Rate of Energy Intake (NREI). Such observations and considerations led, about 50 years ago, to what is known as ‘movement ecology’ where animal movements are considered in their ecological and behavioural context (Marchinton and Jeter, 1967). From the outset researchers sought to develop predicted movement patterns using computer simulations and to compare these predictions with observations of foraging animals (Siniff and Jessen, 1969; Cody, 1971). Over about the next decade, interest in applying OFT to understanding patterns of animal movements increased sharply (Smith, 1974; Jones, 1977; Krebs, 1978; Pyke, 1978a), especially with regard to nectar-feeding birds and bees (Pyke, 1978c, b; Hartling and Plowright, 1979; Pyke, 1979; Zimmerman, 1979; Pyke, 1981; Best and Bierzychudek, 1982; Zimmerman, 1982). Since about 2005 there has been a veritable explosion of interest in movement ecology (e.g., Allen et al., 2018; Torney et al., 2018), as illustrated in Fig. 1 and as observed in 2008 (Nathan, 2008; Nathan et al., 2008). Around 50 years ago, as OFT was rapidly developing (Pyke et al., 1977), it was recognised that foraging animals, in addition to deciding where and when to forage, what to feed on, and how long to spend in one area before departing for another, must also decide how to get from one location to another or, in other words, what movement strategy to adopt. OFT was seen as potentially explaining all such foraging decisions (Pyke et al., 1977; Pyke, 1984). Early attempts to use OFT to understand movements of foraging animals generally took what I call the Cognitive Forager Approach (CFA) in which animals were assumed to be aware of their own internal physiological state, have a sense of direction, hence able to maintain some directionality to their movements (Siniff and Jessen, 1969; Cody, 1971; Pyke, 1978a), an ability to sense potential food items at a significant distance (Pyke, 1978c), and various levels of memory regarding previous foraging (Pyke, 1978c). Similar assumptions were made with regard to applying OFT to other foraging decisions (Pyke et al., 1977). I adopt this, otherwise unnecessary, terminology to distinguish this approach from a more recent approach in which animals are imagined to move in uninformed random walks (see below). Early attempts to understand foraging decisions in general, and movement strategies in particular, also generally assumed that food is patchily distributed, with multiple items tending to occur in patches where they are relatively close to one another (e.g., seeds on the ground) or nearby food locations (e.g., flowers) tending to have similar amounts of food. This concept of patchiness has helped to explain animal decisions regarding when to leave one area and shift to another, especially through development of

Fig. 1 Interest in movement ecology of foraging animals as indicated by the Average number of articles re OFT and animal movement published per year vs Year period (References to articles were obtained by searching the Web of Science for articles relating to foraging theory and movement i.e., Topic ¼ [“foraging theory” OR “optim* forag*”] AND Topic ¼ “movement”).

Foraging j Animal Movements – An Optimal Foraging Theory Approach

151

the Marginal Value Theorem (Charnov, 1976), and adopting Area Restricted Search such that they tend to remain in the vicinity of encountered food (Krebs, 1978). More recently, attempts to use OFT to understand movements of foraging animals have included a new and quite different path, which I shall refer to as the Lévy foraging hypothesis (LFH) and which has split off from the cognitive forager approach (CFA). According to the LFH, as originally conceived (Viswanathan et al., 1996, 1999), animals employ a simple random walk (i.e., no directionality) to search for randomly distributed food items in a featureless environment and in a completely uninformed manner, with no sense of direction, no ability to perceive food items unless they ‘bump’ into them, and no memory regarding previous circumstances. As mentioned above, the CFA assumes the opposite in all these respects. LFH is, in my opinion, largely a waste of time, and something that should simply be abandoned in favour of the CFA (Pyke, 2015). However, it has generated a large amount of research interest (Pyke, 2015; Reynolds, 2018) and continues to receive considerable acclaim, from both theoretical and empirical perspectives (Wosniack et al., 2017; Reynolds, 2018). I shall therefore discuss it in some detail below, before returning to consider the CFA and what is has contributed to our understanding of movements of foraging animals. I shall end this article with discussion of the future for research on movements of foraging animals.

Modelling Movements of Foraging Animals Movements of foraging animals, and animal movements in general, arise from changes in direction and speed, as observations of any animal will indicate. Just watch your favourite animal for a while, and you will see that it is not always facing in the same direction, its speed of travel varies, and it ends up in a different location from where it began. This is the essence of animal movement. At the smallest spatial scales movements of foraging animals could reasonably be considered to consist of linear segments. For example, for a bird that is walking around on its two legs, shifting one leg at a time, we might consider the lifting of one leg off the ground and its placement back on the ground as a movement segment defined by the distance and direction between the two footpositions. The movement path of this bird would then consist of a sequence of these linear segments. Furthermore, at such small spatial scales movements are likely to be consistent in both direction and speed, with little or no change in either between successive movement segments. Animals, including us humans, generally ‘put one foot in front of the other’ and change speed gradually rather than abruptly. Thus, when considered at such small spatial scales, animal paths will often appear ‘meandering’ or ‘tortuous’, with continuous and gradual changes in direction and speed (Da Silveira et al., 2016). However, in order to incorporate animal movements into mathematical models, it is generally necessary to think of them as consisting of sequences of linear segments, possibly with relationships between segments (Pyke, 2015). Assuming that each successive segment begins where the previous one ended, then each segment will be determined by its length and direction, and either or both of these segment properties could depend on similar properties of the previous segment or of earlier segments. The frequency distribution of changes in direction between successive segments might, for example, be unimodal with a mean of zero, if the animal tends to maintain its direction of movement between successive segments. The frequency distribution of changes in speed between successive segments could be similar. Such frequency distributions could be incorporated into a mathematical model of animal movements. In some cases, animal movements may be naturally interrupted, resulting in successive segments that may be taken as linear. For example, some animals, such as herbivorous or carnivorous vertebrates, pause movement to consume or digest their food. Nectarfeeding animals, in particular, pause at flowers and plants while they obtain floral nectar. If such interruptions can be observed, then movements may be naturally considered and modelled as sequences of linear segments. Furthermore, if the interruptions can be related to food, then variations in possible movement strategies can be considered in relation to patterns of food distribution, leading to derivation of predicted optimal foraging strategies that can be tested empirically (more on this below). However, there are often situations where division of animal movement into linear segments must be arbitrary, independent of animal behaviour and carried out at relatively large spatial scale (Pyke, 2015). In some cases, for example, animal movements that have exhibited continuously changing direction have been divided into linear segments on the basis of a threshold change in direction. Here, an animal is considered to have changed direction if its direction relative to an initial direction reaches some threshold. Such thresholds have been arbitrary and independent of what the animal is actually doing as it moves along. In addition, when available technology is such that estimated animal locations are obtained at intervals of time and movements are assumed to consist of linear segments between location ‘fixes’, the resultant movement segments may be large, with little apparent relationship with feeding or other behaviour of the animal (Pyke, 2015). Albatross, for example, have been followed using GPS technology with locations recorded at interval of 90 s, in between which these birds would have traversed about 1000 body lengths, as their flight speed is about 10 m/s and their body length can approach about 1 m (Weimerskirch et al., 1993; Pyke, 2015). There are many other similar examples. In such situations, where an animal can potentially engage in many behaviours in between location fixes, it will clearly be difficult to interpret recorded animal movements in the context of foraging. As animals move around, some of the time foraging, they will generally experience a world that is rich in sensory perception, including in relation to potential food items or sources of food. Even a supposedly lowly bacterium can perceive much about its immediate chemical environment, detecting, for example, various chemical gradients, and thus navigating towards food or away from toxins (Berg, 2000). For those of us with eyes, ears, noses, hair or other sensory apparatus, we can gain an enormous amount

152

Foraging j Animal Movements – An Optimal Foraging Theory Approach

of information about our surroundings through our senses, and a lot of this information may be food-related. You can see this for yourself next time you eat at your favourite restaurant! In modelling movements of foraging animals, it is therefore reasonable and realistic to include assumptions about what an animal can and cannot perceive in terms of food. For nectar-feeding animals, for example, it might generally be reasonable to assume that an animal, at the point of leaving a flower or plant, can observe a number of other flowers or plants around it, and can perceive distances and directions to them, along with other attributes such as size, colour, shape, and so on (Pyke, 1978c). In some cases, such animals may also be able to detect floral odour. Different animal species are likely, through the course of evolution, to have good correspondence between sensory abilities and potentially available information concerning food. A nectar-feeding animal would be expected, for example, to be well able to assess its nectar intake in terms of volume, concentration and sugar composition, as these nectar attributes vary between flowers, plants and plant species, and affect the animals involved. In other words, if a foraging animal could benefit significantly from being able to perceive something, then it probably can. Movements of foraging animals will also, in general, be influenced by memory in relation to previous foraging events, through storing, retrieving and utilising information concerning earlier foraging. Even a bacterium may move differently in response to a chemical stimulus depending on previous chemical experience, and is thus able to utilise information regarding its previous foraging (Berg, 2000). Evidence abounds for other animals employing remembered information as they undergo foraging movement. In fact, it would be reasonable to assert that all of life utilises memory in moving around while foraging. Animals typically live and forage in patchy environments, possibly even with a hierarchy of patches within patches etc, and their foraging movements, coupled with sensory perception and memory, will likely reflect this. For example, a patchy distribution of food items would mean that, when a food item is encountered, others will likely be nearby. Consequently, an animal foraging in a patchy environment might benefit from responding to food encounter through increased changes in direction and shorter movement segments, which would tend to keep it in the vicinity of the encountered item. Such behaviour has been termed ‘area restricted search’ (Krebs, 1978) and has been commonly observed. Similarly, if an animal knows it is foraging in a patch, can keep track of its foraging success in the patch and remembers the average food quality of other patches, then it would be expected to depart its present patch and move to another when its foraging success in its present patch (e.g., rate of food gain) is less than what it would expect if it departed. This expectation is known as the Marginal Value Theorem (Charnov, 1976). As these examples illustrate, the movement strategy adopted by an animal will likely depend on the patchy nature of food distribution. It follows that movements of foraging animals can be modelled as sequences of linear segments, either arising naturally or arbitrarily, so long as care is taken in interpretation of such sequences, directionality of successive segments is included, perception and memory are also adequately included, and patchy food distribution is taken into account. As I shall explain below, the Lévy Foraging Hypothesis abandons almost all of this.

The Le´vy Foraging Hypothesis A central feature of the Lévy Foraging Hypothesis is that the lengths of movement segments are chosen in accordance with a Lévy probability function (Viswanathan et al., 1999). According to this probability function, different segment lengths (‘) occur with probabilities given by P(‘)¼a‘m where the exponent m lies within the range 1gene B-> behavior); reverse genetic studies of gene A can confirm that the gene is essential for the behavior of interest but gene B may contain functional mutations that cause differences in behavior between individuals, and/or differences in behavior between species. Here gene A and B are both essential in a genetic sense, but gene B is more ‘informative’ for understanding differences in behavior between individuals and species.

228

Genes and Behavior j Studying the Genetics of Behavior in the Genomics Era

Behavioral Genomics: Let the Genome ‘Bee’ Your Guide Genome sequencing is an important first step towards understanding what makes a species unique. It can be used to identify novel genes unique to a taxonomic group, to determine evolutionary relationships among organisms, and to infer protein structure and function (Danchin, 1995). The importance of DNA sequencing is reflected in the efforts to sequence the human genome, which was one of the largest collaborative research endeavours spanning 15 years and costing about $2.7 Billion (Lander et al., 2001; Venter et al., 2001). Over the decade following the human genome project the cost of sequencing halved about every 9 months and by 2014 a human genome could be sequenced for $1000 (van Dijk et al., 2014). This massive price reduction greatly facilitated genomic research on other organisms chosen because of their interesting behavior; a point that is easy to illustrate given the rise of social insect genomics over the past decade. The honey bee, Apis mellifera, is a model organism for understanding the genetics of eusocial behavior. In eusocial insects, individuals belong to distinct castes. Reproductive castes produce offspring, while non-reproductive castes build and maintain the nest, care for the brood, forage for food, and engage in nest defence (Michener, 1974). In the honey bee, the non-reproductive workers exhibit many fascinating behaviors such as dancing to communicate the location of new food sources (Von Frisch, 1967), removing sick or dead individuals from a hive to prevent spread of disease (Park et al., 1937), attacking invaders (Breed et al., 2004), grooming each other to remove pests and parasites (Aumeier, 2001), and clustering together in the winter to keep the colony warm (Seeley and Visscher, 1985). The behaviors of worker bees are vastly different from that of queens, yet both caste members share the same genome. It is these intriguing behaviors that led to the honey bee being the third insect to have its genome sequenced (The Honeybee Genome Sequencing Consortium, 2006), following the fruit fly D. melanogaster and the malaria carrying mosquito Anopheles gambiae. One of the most intriguing findings from the honey bee genome, with respect to behavioral genetics, was the expansion of odorant receptors genes. There are 108 more odorant receptor genes in the honey bee compared to the fly (The Honeybee Genome Sequencing Consortium, 2006), most of which are in tandem suggesting gene duplication as a mechanism for the expansion. Social insect colonies contain a plethora of odour cues and pheromones and this finding suggested an increased capacity for chemical communication. In fact, when several ant genomes were sequenced, similar expansions of chemoreceptors were found. The first two ant genomes to be sequenced were of a highly social species and a more primitively social species (Bonasio et al., 2010). The more social species had a larger repertoire of odorant receptors (139 genes versus 105), further suggesting that an increase in sociality is associated with an increase in chemoreceptors and the capacity for chemical communication. The sequencing of other ant species revealed an even larger expansion with 367 and 344 odorant receptor genes in the argentine ant and red harvester ant genomes, respectively (Smith et al., 2011a,b). Comparative genomic studies of social insects were further propelled by the availability of genome sequences for ten bee species that varied in social complexity with multiple independent transitions to sociality (Kapheim et al., 2015). The authors found an intriguing pattern where the evolution of regulatory sequences was strongly associated with the evolution and elaboration of sociality. Comparative genomic analysis of social insects will certainly improve over time as more eusocial insects are sequenced. In addition to comparative genomics, the availability of a genome sequence can be instrumental in studying the genetics of behavior by facilitating functional genomic research to identify genes and mutations that influence behavior, and to study the evolutionary forces acting on such genes, as we detail below.

Transcriptomics and the Neurogenomic States Underlying Behavior The availability of a reference genome allows for the development of tools to rapidly measure global gene expression via microarrays or direct sequencing of RNA (RNAseq). Unlike the genome, which – barring epigenetic modification – is static over an individual’s lifetime, the transcriptome (i.e., collection of gene transcripts found in a tissue at a specific time) is highly dynamic and tissue specific. An individual’s behavior, like transcription, is also dynamic over short time scales. Studying the brain transcriptome of animals that differ in behavior can thus lead to knowledge about the underlying patterns of gene expression that influence behavior. The typical experiment involves sampling individuals that differ in their behavioral norms or repertoires (e.g., aggressive bees vs. gentle bees), extracting mRNA from behaviorally-relevant tissues (e.g., the brain or a specific brain region), and either sequencing the mRNA or hybridizing it to a chip (i.e., microarray) that contains probes for all predicted genes in the genome. The resulting data consists of absolute or relative abundance of transcripts for most genes in a genome for each individual studied. Comparing the transcriptomes of individuals that perform different behaviors or have different behavioral norms will then lead to lists of differentially expressed genes associated with differences in behavior; for example, genes that are over-expressed or ‘upregulated’ in aggressive bees versus gentle bees, and vice versa. This experimental approach has been widely applied to quantify patterns of differential gene expression associated with behavior in many non-genetically tractable organisms, such as bees, wasps, sticklebacks, and zebra finches, as well as genetically tractable species such as the fly and mouse (Naurin et al., 2011; Sanogo et al., 2011; Zayed et al., 2012). The honey bee provides an excellent example of the utility of transcriptomics to study behavior. At any given time, worker bees take on a distinct role within the colony, which involves performing a set of behaviors (e.g., nursing young larva, foraging for pollen and nectar, or defending the colony). However, workers are not ‘locked’ into a specific behavioral state; they progress through a series of roles as they mature, and can switch between roles given cues found in their colony. The BeeSpace experiments

Genes and Behavior j Studying the Genetics of Behavior in the Genomics Era

229

(Zayed and Robinson, 2012) quantified the whole brain (or in some cases distinct regions) transcriptome of worker bees across 20 behavioral contrasts. Some of these contrasts involved analysis of individuals during different points of their behavioral maturation (e.g., young workers that nurse versus old workers that forage). Other contrasts utilized different subspecies of honey bees that differ in their behavior because of heredity, and still others involved analysis of bees exposed to specific environmental cues (i.e., pheromones) that are known to alter their behavior. These studies demonstrated a very close link between brain gene expression and behavior, and highlighted many instances of genes associated with similar behaviors in evolutionary distant species (e.g., bees and wasps, bees and flies, or bees and humans) (Zayed and Robinson, 2012). Behavioral transcriptomics is a very powerful approach for developing mechanistic hypotheses about the genes involved in regulating behavior. With the cost effectiveness of RNAseq, transcriptomics can essentially be applied to any species from which one can extract RNA. Given recent bioinformatic advances, it is now possible to carry out RNAseq experiments without first generating a high-quality reference genome – although it still helps – meaning researchers can quickly quantify the behavioral transcriptome of non-model species (Lozier and Zayed, 2016). One of the main benefits of transcriptomics is that it rapidly provides high resolution data; data on the genes, and even specific exons, that are differentially regulated in association with behavior. However, the transcriptomic approach does suffer from some drawbacks. On their own, transcriptomic studies highlight patterns of gene expression correlated with behavior and determining cause-and-effect is not trivial. Changes in the expression of some genes may cause behavior, while changes in behavior may cause changes in expression of other genes, or an unobserved factor (e.g., a hormone) may concurrently influence both behavior and gene expression. While functional knowledge gleaned from genetically tractable organisms can allow researchers to formulate hypotheses regarding the nature of the relationship between brain gene expression and behavior, additional follow up experiments are often needed. Moreover, because genes work in networks, it is not correct to assume that the genes associated with behavior based on evidence of differential expression are the same genes that lead to intra-specific differences in behavior within a population or differences in behavior between species. There is an emerging technique that can partially alleviate the above-mentioned drawbacks. The technique is called Expression QTL (eQTL) analysis. We discussed above how crosses between phenotypically divergent laboratory or natural populations can be used to map QTLs that causally influence the phenotype of interest, but such studies often lack gene-level resolution because QTL regions are very large. Brain gene expression is a phenotype, and thus the genetics of gene expression can be analysed using the QTL approach (Schadt et al., 2005). The identified eQTLs represent genetic mutations that causally influence the regulation of a specific gene. Quantitative geneticists often classify eQTLs as either cis or trans if the eQTL is physically close or far from the gene it regulates. It is also possible to highlight regulatory hot spots for brain gene expression; a small genomic locus with genetic variation that influences the expression of many genes in the brain. eQTL studies of brain gene expression have proved very useful in studying the regulatory architecture of complex behavioral traits, such as stress response in chickens, and schizophrenia in humans (Chen et al., 2008; Fallahsharoudi et al., 2017).

Genome Wide Association Studies and the Genetics of Behavior Genome-wide association studies (GWAS) is an emerging approach for uncovering candidate regions of a genome associated with complex and quantitative traits such as behavior. GWAS involves quantifying traits in a large number of individuals, which are then sequenced to high coverage or genotyped at a very large number of markers (Fig. 3) (Kruglyak, 2008; McCarthy et al., 2008). Both GWAS and QTL studies capitalize on linkage disequilibrium between the causal mutation and adjacent neutral mutations to identify regions of the chromosome that are associated with the phenotype of interest. Linkage disequilibrium decays in natural populations as a function of generation time and recombination rate between loci. In a large random-mating population, linkage disequilibrium blocks are relatively small in size, which substantially increases the resolution of GWAS over QTL studies; the former studies can typically result in gene-level, and in some cases, mutation-level, associations. Another major advantage of GWAS over QTLs is it can be applied to non-genetically tractable organisms; there is no need for crosses or breeding work. GWAS has been effectively applied to study the genetics of several behaviors in a growing number of animals (Li et al., 2016b; Parker et al., 2016; Spotter et al., 2016). GWAS does have some shortcomings. Careful consideration must be given to statistical power. Unfortunately, most studies are underpowered, especially for detecting mutations that have small effects on the phenotype (Chabris et al., 2012). Like with QTL studies, GWAS on complex traits that are controlled by a large number of loci each contributing a very small effect are less likely to be fruitful (Korte and Farlow, 2013). Multiple testing is also an issue given that most GWAS pipelines involve testing a very large number of mutations for phenotypic associations. Methods for post-hoc multiple test correction already exist that vary in their assumptions of test independence and stringency (Goeman and Solari, 2014). The problem of false positives can be rapidly compounded if the statistical analyses do not take into account confounding factors, spatial structure, and the distribution of the phenotypes studied (Korte and Farlow, 2013). Despite these limitations, GWAS still provides a lot of promise for identifying mutations that have a moderate to large influence on behavior in natural populations.

Population Genomics Ties the How’s and Why’s of Behavioral Evolution Behavior is a complex trait. A prime example of complexity in evolutionary theory during Darwin’s time was the eye – how could it possibly evolve to its current state (Gehring, 2005)? Yet now sequencing of opsins in birds and reptiles has enabled us to deduce the

230

Genes and Behavior j Studying the Genetics of Behavior in the Genomics Era

Fig. 3 Overview of Genome Wide Association mapping of behavioral traits. Vibrating behavior is measured for many bees in a natural population. The genomes of these individuals are sequenced at high coverage to genotype a large proportion of the mutations found in the population. In a large stable randomly mating population, linkage disequilibrium blocks tend to be small. Mutations that cause vibration behavior, or those that are very closely linked to causal loci, are expected to show a statistical association with vibrating behavior in the sampled individuals (e.g., allele G at a specific SNP on chromosome 4; inset bar plot). GWAS results are typically visualized with a Manhattan plot, showing chromosome position versus the significance of genotype-phenotype associations (log10(P)) for SNPs across the genome.

gene sequences of opsins in the eyes of dinosaurs. Expressing these genes in the laboratory has allowed us to find out which colours of light dinosaurs were sensitive to, and conversely how this changed as modern birds evolved (Chang et al., 2002). Deducing the evolutionary history of complex traits from gene sequences can also be carried out with behavioral traits, and population genomics can show us how specific parts of genes have changed in response to selection, as we summarize below. Population genomics is population genetics applied to entire genomes or transcriptomes (Hasselmann et al., 2015; Kent and Zayed, 2015). Population genetics provides us with quantitative tests of whether sequences have been under positive, negative, or balancing selection. In the past, population genetics was restricted to one or a few genes because of the cost and time involved in sequencing. Now, entire genomes of multiple individuals in a population can be sequenced quickly and for the same cost we used to pay for sequencing one gene. The most powerful population genomics tools for looking at recent evolution require a number of genome sequences of many individuals from multiple populations or from a target species and a sister species. Genetic diversity is the main input for population genomic analyses – so individuals should be sequenced to sufficient depth to make confident genotype calls possible, unless you are studying a model organism with an extensive haplotype database (Thorisson et al., 2005), in which case lighter coverage may be used with imputation methods (Marchini et al., 2007). We illustrate the utility of population genomics in understanding behavioral genetics and evolution using a few examples from model and non-model organisms. The fruit fly Drosophila melanogaster is the model organism with the longest history of reverse genetics. In 2012 a consortium of researchers created the Drosophila melanogaster Genetic Reference Panel (DGRP) (Mackay et al., 2012). This very powerful genomic tool includes full genome sequences of 205 inbred lines derived from a natural population, and at least 29 phenotypes have been measured for each line. Genetic differences found between the inbred lines allowed researchers to identify candidate genes involved in olfactory behavior (Swarup et al., 2013), sleep (Swarup et al., 2013), pheromones involved in mating behaviors (Dembeck et al., 2015), and aggression (Shorter et al., 2015), among many others. The mouse behavior community also has inbred lines and a very powerful resource (Churchill et al., 2004) – the Collaborative Cross – for studying the genetics of behavioral states such as stress, anxiety, and alcohol response (Goldowitz et al., 2006). The Mouse Phenome Database (Bogue et al., 2018) contains 35 behavioral

Genes and Behavior j Studying the Genetics of Behavior in the Genomics Era

231

measures on hundreds of strains of mice from various crosses, a dataset of mutations found within these strains, and many online tools for analysis. The DGRP and the Mouse Phenome Database illustrate the value of collaboration and shared online resources in population genomics. However, if you work on a non-model species, it is still possible to apply population genomics for behavioral research. New genomes or exomes can be sequenced at a modest cost, and interspecies comparisons of such new genomes can shed light on how genes associated with behavior evolve. For example, a consortium of bee researchers sequenced and compared genomes of 10 bee species ranging from solitary through primitively social to fully eusocial, and found signs of selection on regulatory features as the degree of sociality increased (Kapheim et al., 2015). Ant researchers used transcriptomes of 16 ant species to look for conserved “building blocks” involved in evolution of complex social behaviors (Morandin et al., 2016), while a study of 7 ant genomes revealed signs of positive selection both within the ants (Roux et al., 2014). Several groups have sequenced individual honey bee genomes to scan for signatures of selection (Harpur et al., 2014; Molodtsova et al., 2014; Wallberg et al., 2014). One of the major findings from this effort provided evidence that genes regulating worker division of labour and behavior were enriched for signs of positive selection (Harpur et al., 2014), and that such genes were often evolutionarily ‘novel’ (Jasper et al., 2014). A similar study on the primitively eusocial bumble bees found different patterns to those found in honey bee studies, suggesting the possibility that social complexity in bees is strongly related to patterns of adaptive genome evolution (Harpur et al., 2017).

The Sum of the Behavioral Tool Kit is Much More Powerful Than its Parts We have detailed several approaches to study the genetics of behavior in animals. While each method offers some advantages over the others, collectively, they make for a very powerful tool-kit for generating causal knowledge that links genetic diversity to behavior in populations, knowledge about the molecular biology underlying specific behavioral phenotypes, and how the genes, regulatory sequences and networks underlying behavior evolve. Additionally, there has been recent substantial improvement in the ability to carry out research on non-model organisms. The diminishing costs of genome sequencing and the emergence of CRISPR/ Cas9 will help efforts to study the genetics, molecular biology and evolution of behavior via GWAS, expression profiling and genome editing, and population genomic approaches. We anticipate that the major hurdle in studying the genetics of behavior will be quantifying behavior from a large number of individuals. We look forward to the development of new tools and methods to accurately and, ideally, passively quantify behavior; high throughput data on behavior would greatly leverage the power of genomics to study the genetics of behavior in animals.

See also: Genes and Behavior: Behavioral genetics of dog breeds; Evolutionary behavioral genetics; Genetics of animal and bird migration; MHC genes and mate choice; The Oxytocin System: Single Gene Effects on Social Behavior Across Species.

References Anholt, R.R.H., Mackay, T.F.C., 2004. Quantitative genetic analyses of complex behaviours in Drosophila. Nature Reviews Genetics 5, 838–849. Aumeier, P., 2001. Bioassay for grooming effectiveness towards Varroa destructor mites in Africanized and Carniolan honey bees. Apidologie 32, 81–90. Bogue, M.A., Grubb, S.C., Walton, D.O., et al., 2018. Mouse phenome database: An integrative database and analysis suite for curated empirical phenotype data from laboratory mice. Nucleic Acids Research 46, D843–D850. Bonasio, R., Zhang, G., Ye, C., et al., 2010. Genomic comparison of the ants Camponotus floridanus and Harpegnathos saltator. Science 329, 1068–1071. Breed, M.D., Guzmán-Novoa, E., Hunt, G.J., 2004. Defensive behavior of honey bees: Organization, genetics, and comparisons with other bees. Annual Reviews in Entomology 49, 271–298. Carlezon, W.A., Thome, J., Olson, V.G., et al., 1998. Regulation of cocaine reward by CREB. Science 282, 2272–2275. Chabris, C.F., Hebert, B.M., Benjamin, D.J., et al., 2012. Most reported genetic associations with general intelligence are probably false positives. Psychological Science 23, 1314–1323. Chang, B.S., Jönsson, K., Kazmi, M.A., Donoghue, M.J., Sakmar, T.P., 2002. Recreating a functional ancestral archosaur visual pigment. Molecular Biology and Evolution 19, 1483–1489. Chen, Y., Zhu, J., Lum, P.Y., et al., 2008. Variations in DNA elucidate molecular networks that cause disease. Nature 452, 429–435. Churchill, G.A., Airey, D.C., Allayee, H., et al., 2004. The collaborative cross, a community resource for the genetic analysis of complex traits. Nature Genetics 36, 1133–1137. Danchin, A., 1995. Why sequence genomes? The Escherichia coli imbroglio. Molecular Microbiology 18, 371–376. Dembeck, L.M., Böröczky, K., Huang, W., et al., 2015. Genetic architecture of natural variation in cuticular hydrocarbon composition in Drosophila melanogaster. eLife 4, e09861. Falconer, D.S., Mackay, T.F.C., 1996. Introduction to Quantitative Genetics. Longman, Essex, UK. Fallahsharoudi, A., de Kock, N., Johnsson, M., et al., 2017. Genetic and targeted eQTL mapping reveals strong candidate genes modulating the stress response during chicken domestication. G3 (Bethesda) 7, 497–504. Flint, J., Valdar, W., Shifman, S., Mott, R., 2005. Strategies for mapping and cloning quantitative trait genes in rodents. Nature Reviews Genetics 6, 271–286. Fu, X.-D., Ares Jr., M., 2014. Context-dependent control of alternative splicing by RNA-binding proteins. Nature Reviews Genetics 15, 689–701. Gehring, W.J., 2005. New perspectives on eye development and the evolution of eyes and photoreceptors. Journal of Heredity 96, 171–184. Goeman, J.J., Solari, A., 2014. Multiple hypothesis testing in genomics. Statistics in Medicine 33, 1946–1978. Goldowitz, D., Matthews, D.B., Hamre, K.M., et al., 2006. Progress in using mouse inbred strains, consomics, and mutants to identify genes related to stress, anxiety, and alcohol phenotypes. Alcoholism: Clinical and Experimental Research 30, 1066–1078. Harpur, B.A., Dey, A., Albert, J.R., et al., 2017. Queens and workers contribute differently to adaptive evolution in bumble bees and honey bees. Genome Biology and Evolution 9, 2395–2402.

232

Genes and Behavior j Studying the Genetics of Behavior in the Genomics Era

Harpur, B.A., Kent, C.F., Molodtsova, D., et al., 2014. Population genomics of the honey bee reveals strong signatures of positive selection on worker traits. Proceedings of the National Academy of Sciences 111, 2614–2619. Hasselmann, M., Ferretti, L., Zayed, A., 2015. Beyond fruit-flies: Population genomic advances in non-Drosophila arthropods. Briefings in Functional Genomics 14, 424–431. Hohenlohe, P.A., Bassham, S., Etter, P.D., et al., 2010. Population genomics of parallel adaptation in threespine stickleback using sequenced rad tags. PLOS Genetics 6, e1000862. Jasper, W.C., Linksvayer, T.A., Atallah, J., et al., 2014. Large scale coding sequence change underlies the evolution of post-developmental novelty in honey bees. Molecular Biology and Evolution. Jones, W.D., 2009. The expanding reach of the GAL4/UAS system into the behavioral neurobiology of Drosophila. BMB Reports 42, 705–712. Kapheim, K.M., Pan, H., Li, C., et al., 2015. Genomic signatures of evolutionary transitions from solitary to group living. Science 348, 1139–1143. Kaun, K.R., Hendel, T., Gerber, B., Sokolowski, M.B., 2007. Natural variation in Drosophila larval reward learning and memory due to a cGMP-dependent protein kinase. Learning and Memory 14, 342–349. Kearsey, M.J., 1998. The principles of QTL analysis (a minimal mathematics approach). Journal of Experimental Botany 49, 1619–1623. Kent, C.F., Daskalchuk, T., Cook, L., Sokolowski, M.B., Greenspan, R.J., 2009. The Drosophila foraging gene mediates adult plasticity and gene – Environment interactions in behaviour, metabolites, and gene expression in response to food deprivation. PLOS Genetics 5, e1000609. Kent, C.F., Zayed, A., 2015. Chapter nine – Population genomic and phylogenomic insights into the evolution of physiology and behaviour in social insects. In: Zayed, A., Kent, C.F. (Eds.), Advances in Insect Physiology. Academic Press, pp. 293–324. Korte, A., Farlow, A., 2013. The advantages and limitations of trait analysis with GWAS: A review. Plant Methods 9, 29. Kruglyak, L., 2008. The road to genome-wide association studies. Nature Reviews Genetics 8, 314–318. Lander, E.S., Linton, L.M., Birren, B., et al., 2001. Initial sequencing and analysis of the human genome. Nature 409, 860–921. Lapidge, K.L., Oldroyd, B.P., Spivak, M., 2002. Seven suggestive quantitative trait loci influence hygienic behavior of honey bees. Naturwissenschaften 89, 565–568. Lehner, B., 2013. Genotype to phenotype: Lessons from model organisms for human genetics. Nature Reviews Genetics 14, 168–178. Li, Y.I., van de Geijn, B., Raj, A., et al., 2016a. RNA splicing is a primary link between genetic variation and disease. Science 352, 600–604. Li, Z., Zheng, M., Abdalla, B.A., et al., 2016b. Genome-wide association study of aggressive behaviour in chicken. Science Reports 6, 30981. Lozier, J.D., Zayed, A., 2016. Bee conservation in the age of genomics. Conservation Genetics 1–17. Lynch, M., Walsh, B., 1998. Genetics and Analysis of Quantitative Traits. Sinauer Associates Inc, Sunderland, MA, United States. Mackay, T.F., Richards, S., Stone, E.A., et al., 2012. The Drosophila melanogaster genetic reference panel. Nature 482, 173–178. Marchini, J., Howie, B., Myers, S., McVean, G., Donnelly, P., 2007. A new multipoint method for genome-wide association studies by imputation of genotypes. Nature Genetics 39, 906. Matthews, D.B., Bhave, S.V., Belknap, J.K., et al., 2005. Complex genetics of interactions of alcohol and CNS function and behavior. Alcoholism: Clinical and Experimental Research 29, 1706–1719. McCarthy, M.I., Abecasis, G.R., Cardon, L.R., et al., 2008. Genome-wide association studies for complex traits: Consensus, uncertainty and challenges. Nature Reviews Genetics 9, 356–369. Michener, C.D., 1974. The Social Behavior of the Bees: A Comparative Study. Belknap Press of Harvard University Press. Moaddab, M., Dabrowska, J., 2017. Oxytocin receptor neurotransmission in the dorsolateral bed nucleus of the stria terminalis facilitates the acquisition of cued fear in the fearpotentiated startle paradigm in rats. Neuropharmacology 121, 130–139. Molodtsova, D., Harpur, B.A., Kent, C.F., Seevananthan, K., Zayed, A., 2014. Pleiotropy constrains the evolution of protein but not regulatory sequences in a transcription regulatory network influencing complex social behaviours. Frontiers in Genetics 5, 431. Morandin, C., Tin, M.M.Y., Abril, S., et al., 2016. Comparative transcriptomics reveals the conserved building blocks involved in parallel evolution of diverse phenotypic traits in ants. Genome Biology 17, 1–19. Naurin, S., Hansson, B., Hasselquist, D., Kim, Y.-H., Bensch, S., 2011. The sex-biased brain: Sexual dimorphism in gene expression in two species of songbirds. BMC Genomics 12, 37. Osborne, K.A., Robichon, A., Burgess, E., et al., 1997. Natural behavior polymorphism due to a cGMP-dependent protein kinase of Drosophila. Science 277, 834–836. Page, R.E., Fondrk, M.K., Hunt, G.J., et al., 2000. Genetic dissection of honeybee (Apis mellifera L.) foraging behavior. Journal of Heredity 91, 474–479. Park, O., Pellet, F., Paddock, F., 1937. Disease resistance and American foulbrood. The American Bee Journal 77, 20–25. Parker, C.C., Gopalakrishnan, S., Carbonetto, P., et al., 2016. Genome-wide association study of behavioral, physiological and gene expression traits in outbred CFW mice. Nature Genetics 48, 919–926. Roux, J., Privman, E., Moretti, S., et al., 2014. Patterns of positive selection in seven ant genomes. Molecular Biology and Evolution 31, 1661–1685. Salvi, S., Tuberosa, R., 2005. To clone or not to clone plant QTLs: Present and future challenges. Trends in Plant Science 10, 297–304. Sanogo, Y.O., Hankison, S., Band, M., Obregon, A., Bell, A.M., 2011. Brain transcriptomic response of threespine sticklebacks to cues of a predator. Brain, Behavior and Evolution 77, 270–285. Saunders, B.T., Richard, J.M., Janak, P.H., 2015. Contemporary approaches to neural circuit manipulation and mapping: Focus on reward and addiction. Philosophical Transactions of the Royal Society B: Biological Sciences 370. Schadt, E.E., Lamb, J., Yang, X., et al., 2005. An integrative genomics approach to infer causal associations between gene expression and disease. Nature Genetics 37, 710–717. Seeley, T.D., Visscher, P., 1985. Survival of honeybees in cold climates: The critical timing of colony growth and reproduction. Ecological Entomology 10, 81–88. Shalem, O., Sanjana, N.E., Zhang, F., 2015. High-throughput functional genomics using CRISPR – Cas9. Nature Reviews Genetics 16, 299. Shorter, J., Couch, C., Huang, W., et al., 2015. Genetic architecture of natural variation in Drosophila melanogaster aggressive behavior. Proceedings of the National Academy of Sciences 112, E3555–E3563. Sivanantharajah, L., Zhang, B., 2015. Current techniques for high-resolution mapping of behavioral circuits in Drosophila. Journal of Comparative Physiology A 201, 895–909. Smith, C.D., Zimin, A., Holt, C., et al., 2011a. Draft genome of the globally widespread and invasive Argentine ant (Linepithema humile). Proceedings of the National Academy of Sciences 108, 5673–5678. Smith, C.R., Smith, C.D., Robertson, H.M., et al., 2011b. Draft genome of the red harvester ant Pogonomyrmex barbatus. Proceedings of the National Academy of Sciences 108, 5667–5672. Sparta, D.R., Jennings, J.H., Ung, R.L., Stuber, G.D., 2013. Optogenetic strategies to investigate neural circuitry engaged by stress. Behavioural Brain Research 255, 19–25. Spotter, A., Gupta, P., Mayer, M., Reinsch, N., Bienefeld, K., 2016. Genome-wide association study of a Varroa-specific defense behavior in honeybees (Apis mellifera). The Journal of Heredity 107, 220–227. Swarup, S., Huang, W., Mackay, T.F., Anholt, R.R., 2013. Analysis of natural variation reveals neurogenetic networks for Drosophila olfactory behavior. Proceedings of the National Academy of Sciences 110, 1017–1022. Tallafuss, A., Washbourne, P., Postlethwait, J., 2014. Temporally and spatially restricted gene expression profiling. Current Genomics 15, 278–292. Tarantino, L.M., Eisener-Dorman, A.F., 2011. Forward genetic approaches to understanding complex behaviors. In: Behavioral Neurogenetics, pp. 25–58. Springer. The Honeybee Genome Sequencing Consortium, 2006. Insights into social insects from the genome of the honeybee Apis mellifera. Nature 443, 931–949. Thorisson, G.A., Smith, A.V., Krishnan, L., Stein, L.D., 2005. The international HapMap project web site. Genome Research 15, 1592–1593. Toth, A.L., Robinson, G.E., 2007. Evo-devo and the evolution of social behavior. Trends in Genetics 23, 334–341. Trible, W., Olivos-Cisneros, L., McKenzie, S.K., et al., 2017. Orco mutagenesis causes loss of antennal lobe glomeruli and impaired social behavior in ants. Cell 170, 727–735.

Genes and Behavior j Studying the Genetics of Behavior in the Genomics Era

233

van Dijk, E.L., Auger, H., Jaszczyszyn, Y., Thermes, C., 2014. Ten years of next-generation sequencing technology. Trends in Genetics 30, 418–426. Venter, J.C., Adams, M.D., Myers, E.W., et al., 2001. The sequence of the human genome. Science 291, 1304–1351. Von Frisch, K., 1967. The Dance Language and Orientation of Bees. Belknap Press of Harvard University Press. Wallberg, A., Han, F., Wellhagen, G., et al., 2014. A worldwide survey of genome sequence variation provides insight into the evolutionary history of the honeybee Apis mellifera. Nature Genetics 46, 1081–1088. Wang, Y., Mutti, N.S., Ihle, K.E., et al., 2010. Down-regulation of honey bee IRS gene biases behavior toward food rich in protein. PLOS Genetics 6, e1000896. West-Eberhard, M.J., 1989. Phenotypic plasticity and the origins of diversity. Annual Review of Ecology and Systematics 20, 249–278. Wolschin, F., Mutti, N.S., Amdam, G.V., 2011. Insulin receptor substrate influences female caste development in honeybees. Biology Letters 7, 112–115. Yang, L., Zhao, Y., Wang, Y., et al., 2015. The effects of psychological stress on depression. Current Neuropharmacology 13, 494–504. Yan, H., Opachaloemphan, C., Mancini, G., et al., 2017. Olfactory receptors are required for social behavior and neural plasticity in ants, as evidenced by CRISPR-mediated gene knockout. In: BioRxiv, p.142232. Zayed, A., Naeger, N.L., Rodriguez-Zas, S.L., Robinson, G.E., 2012. Common and novel transcriptional routes to behavioral maturation in worker and male honey bees. Genes, Brain and Behavior 11, 253–261. Zayed, A., Robinson, G.E., 2012. Understanding the relationship between brain gene expression and social behavior: Lessons from the honey bee. Annual Review of Genetics 46, 591–615.

Further Reading Ashbrook, D.G., Mulligan, M.K., Williams, R.W., 2017. Post-genomic behavioral genetics: From revolution to routine. Genes, Brain and Behaviour e12441 (epub ahead of print). Charney, E., 2017. Genes, behavior, and behavior genetics. Wiley Interdisciplinary Reviews: Cognitive Science 8, e1405. Hayes, B., 2013. Overview of statistical methods for genome-wide association studies (GWAS). In: Gondro, C., van der Werf, J., Hayes, B. (Eds.), Genome-Wide Association Studies and Genomic Prediction. Humana Press, Totowa, NJ, pp. 149–169. Mackay, T.F.C., Eric, A.S., Julien, F.A., 2009. The genetics of quantitative traits: Challenges and prospects. Nature Reviews Genetics 10, 565–577.

Relevant Websites dgrp2.gnets.ncsu.edu–Drosophila Genetics Reference Panel. phenome.jax.org–Mouse Phenome Database.

Evolution of Behavior: Genotype to Phenotype Jennifer R Merritt, Emory University, Atlanta, United States © 2019 Elsevier Ltd. All rights reserved.

Glossary Adeno-associated virus (AAV) A type of virus that can infect organisms and replicate its genetic material, with or without integrating its DNA into the host genome. AAV is used as a tool for researchers to deliver genes to tissues of interest. Alternative splicing A process that allows for a single gene to encode multiple protein products. mRNA transcribed from exons is differentially cut, joined, and/or skipped prior to translation. See alternative transcript. Alternative transcript An mRNA isoform from an alternatively spliced gene. See alternative splicing. Chromatin accessibility The winding of chromatin, which is a complex of DNA, RNA, and proteins called histones. Chromatin that is tightly wound (heterochromatin) is less accessible to transcriptional regulators than chromatin that is lightly wound (euchromatin). Accordingly, euchromatin is associated with actively transcribed genes, and heterochromatin is associated with less gene expression. Epigenetic modifications to histones and DNA contribute to chromatin accessibility. See epigenetics. Chromosomal inversion This large-scale chromosomal rearrangement is caused by a double break in a chromosome. The broken segment is reintegrated backwards such that all of the genes in the segment are in reverse order. Chromosomal inversions prevent recombination between the inversion and homologous non-inverted chromosome. As such, these rearrangements are thought to be important for evolutionary change because they allow for the co-inheritance of co-adapted genes for complex phenotypes. Cis-regulatory element A noncoding DNA sequence that regulates the transcription of a gene. These sequences can be found upstream or downstream of a protein coding sequence, introns, or even kilobases away from they gene they regulate. Promoters, enhancers, and silencers are types of cis-regulatory elements. For contrast, see trans-regulatory element. Clozapine-N-oxide (CNO) The drug that activates DREADDs. See DREADDs. Congenic strains Organisms that differ from another species or strain by only a single locus in the genome. Congenic strains are generated by backcrossing one strain onto another, and genotyping and/or phenotyping the animals between each generation. Cre recombinase An enzyme that recombines genes that are flanked by lox sites in the DNA. Cre-lox recombination can be used to create insertions, deletions, inversions, and translocations in transgenic organisms with lox sites inserted upsteam and downstream of the gene of interest. Cre-lox is a powerful tool for neuroscientists to target single genes in order to study the circuits and cell types in the brain contributing to cognition and behavior. See transgenic organism. CRISPR-Cas9 Abbreviation for clustered regularly interspaced short palindromic sequences. This technology harnesses the antiviral immune system of prokaryotes for genome editing. See genome editing. Deletion A mutation in which one or more nucleotides in a DNA sequence is lost, relative to another sequence. See insertion. Designer receptors exclusively activated by designer drugs (DREADDs) Disassortative mating A mating system in which an organism breeds with another organism that is different from its own genotype or phenotype. Enhancer A type of cis-regulatory element that can bind transcription factors in order to promote the expression of the associated gene. Because DNA is coiled in the nucleus, an enhancer can be kilobases away from its target gene, but be physically nearby. Enhancers facilitate gene expression by forming DNA-protein complexes with promoters, transcription factors, and RNA polymerase. Like silencers, these sequences can be adjacent to the gene that they regulate, or as far as a million base pairs away. See transcription factor, silencer, cis-regulatory element. Epigenetics The study of changes in gene expression that do not involve changes to the underlying DNA sequence. Epigenetic mechanisms are susceptible to environmental influence, reversible, and in some cases can be passed down to future generations. Methylation and histone modifications are examples of epigenetic mechanisms. See chromatin accessibility, methylation. Functional genomics The study of ‘omics data, such as genomics and transcriptomics, to predict the function, regulation, and interactions of genes, RNA, and protein. Genome editing A type of genetic engineering in which the DNA of a living organism is modified resulting in an insertion, deletion, or mutation. See SNP, insertion, deletion, CRISPR-Cas9. Hybrid An organism that is the result of a genetic cross between two parental strains or species. The first genetic cross between parental lines is the first filial generation (F1). Each F1 hybrid inherits exactly 50% of its genetic material from each parental strain, resulting in genetically identical organisms that are heterozygous at every loci in the genome. The genetic cross between F1 hybrids is called the second filial generation (F2). F2 hybrids are genetically unique due to recombination of F1 chromosomes during meiosis. Like F2 hybrids, subsequent generations have a mixed genetic background, and are termed F3, F4, etc. F2 hybrids are commonly used for QTL mapping.

234

Encyclopedia of Animal Behavior, 2nd edition, Volume 2

https://doi.org/10.1016/B978-0-12-809633-8.90785-4

Genes and Behavior j Evolution of Behavior: Genotype to Phenotype

235

Insertion A mutation in which one or more nucleotides into a DNA sequence is introduced, relative to another sequence. See deletion. Introcerebroventricular (i.c.v.) injection A technique of injecting substances into the cerebral ventricles of the brain. Methylation An epigenetic mechanism, in which a methyl group is added to a cytosine or adenine nucleotide. This process is reversible and does not alter the underlying DNA sequence. Methylation is thought to increase or decrease gene expression, depending on which parts of the gene (i.e. promoter, exon, intron, 5’ UTR, 3’ UTR) are methylated and the density of methylated nucleotides. Methylation in noncoding regions typically decreases expression, and methylation in protein coding sequences can increase expression. See epigenetics. Microsatellite A region of DNA with a short (1-6 bp) repeating motif (e.g. ACTACTACTACTACT). The number of repeats, or length of the microsatellite, can be used to genotype organisms. Functionally, this motif is thought to play a role in the regulation of gene expression when it occurs within cis-regulatory elements. See cis-regulatory element. Mutagenesis The generation of mutations within DNA. This process can occur naturally or can be induced by exposure to mutagens, such as radiation like x-rays or compounds like ethyl methanesulfonate. Neuropeptide Polypeptides that act as neurotransmitters. Examples of neuropeptides include oxytocin, vasopressin, neuropeptide-Y, a-melanocyte stimulating hormone, galanin, and encephalin. Next generation sequencing (NGS) A catch-all term for high-throughput sequencing techniques that allow for fast, accurate, and inexpensive sequencing of DNA and RNA. Examples of NGS platforms include the Illumina HiSeq, Roche 454, Ion Torrent, Life Sciences SOLiD4, and Pacific Biosciences SMRT, which can be contrasted with Sanger sequencing, the technique of the “first-generation” of DNA sequencing. Non-synonymous mutation A nucleotide mutation in the coding sequence of DNA that will alter the amino acid sequence of a protein. Ortholog A gene in related species that has evolved from the gene of a common ancestor by speciation. Over evolutionary time, orthologs typically retain a similar function. Pleiotropy The well-documented phenomenon of a single gene affecting more than one phenotype. Promoter A short region of DNA found immediately upstream of the transcription start site of a gene that contains binding sites for RNA polymerase to induce gene expression. Quantitative trait locus (QTL) mapping A statistical method that uses genetic variation (i.e. SNPs) to predict phenotypes in order to identify regions of the genome (called quantitative trait loci or QTL) that may contribute to the genetic basis of complex phenotypes. See SNP, hybrid. RNA-seq A NGS technique that can be used to sequence all of the RNA transcripts within a sample. RNA-seq data can provide information about RNA abundance and alternative transcripts, as well as the identity and alleles of the genes that are expressed. See next-generation sequencing, alternative transcript. Silencer A type of cis-regulatory element that can bind transcription factors in order to repress the expression of the associated gene. Silencers repress gene expression by forming DNA-protein complexes with promoters and transcription factors. Like enhancers, these sequences can be adjacent to the gene that they regulate, or as far as a million base pairs away. See enhancer, cis-regulatory element. Single nucleotide polymorphism (SNP) A site of genetic variation at a single position in the genome. See insertion, deletion. Transcription factor binding site A short stretch of DNA (typically