Advances in the Study of Behavior, Volume 32 (Advances in the Study of Behavior)

Advances in THE STUDY OF BEHAVIOR VOLUME 32 Advances in THE STUDY OF BEHAVIOR Edited by Peter J. B. Slater Jay S. R...

Author: Peter J.B. Slater | Jay S. Rosenblatt | Charles T. Snowdon | Timothy J. Roper

91 downloads 993 Views 6MB Size Report

This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!
Report copyright / DMCA form

DOWNLOAD PDF

Advances in

THE STUDY OF BEHAVIOR VOLUME 32

Advances in THE STUDY OF BEHAVIOR Edited by

Peter J. B. Slater Jay S. Rosenblatt Charles T. Snowdon Timothy J. Roper

Advances in THE STUDY OF BEHAVIOR Edited by Peter J. B. Slater School of Biology University of St. Andrews Fife, United Kingdom

Jay S. Rosenblatt Institute of Animal Behavior Rutgers University Newark, New Jersey

Charles T. Snowdon Department of Psychology University of Wisconsin Madison, Wisconsin

Timothy J. Roper School of Biological Sciences University of Sussex Sussex, United Kingdom

VOLUME 32

Amsterdam

Boston London New York Oxford Paris San Francisco Singapore Sydney Tokyo

San Diego

This book is printed on acid-free paper. Copyright ß 2003, Elsevier Science (USA). All Rights Reserved. No part of this publication may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopy, recording, or any information storage and retrieval system, without permission in writing from the Publisher. The appearance of the code at the bottom of the first page of a chapter in this book indicates the Publisher’s consent that copies of the chapter may be made for personal or internal use of specific clients. This consent is given on the condition, however, that the copier pay the stated per copy fee through the Copyright Clearance Center, Inc. (222 Rosewood Drive, Danvers, Massachusetts 01923), for copying beyond that permitted by Sections 107 or 108 of the U.S. Copyright Law. This consent does not extend to other kinds of copying, such as copying for general distribution, for advertising or promotional purposes, for creating new collective works, or for resale. Copy fees for pre-2003 chapters are as shown on the title pages. If no fee code appears on the title page, the copy fee is the same as for current chapters. 0065-3454/2003 $35.00 Permissionions may be sought directly from Elsevier’s Science & Technology Rights Department in Oxford, UK: phone: (+44) 186 5843830, fax: (+44) 1865 853333, e-mail: [email protected]. You may also complete your request on-line via the Elsevier Science homepage (http://elsevier.com), by selecting ‘‘Customer Support’’ and then ‘‘Obtaining Permissions.’’

Academic Press An Elsevier Science Imprint. 525 B Street, Suite 1900, San Diego, California 92101-4495, USA http://www.academicpress.com

Academic Press 84 Theobald’s Road, London WC1X 8RR, UK http://www.academicpress.com International Standard Book Number: 0-12-004532-X PRINTED IN THE UNITED STATES OF AMERICA 03 04 05 06 07 08 9 8 7 6 5 4 3

2

1

Contents

Contributors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Preface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

ix xi

Self-Organization and Collective Behavior in Vertebrates IAIN D. COUZIN AND JENS KRAUSE I. II. III. IV. V.

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Group Shape and Motion . . . . . . . . . . . . . . . . . . . . . . . . . . . Group Internal Structure . . . . . . . . . . . . . . . . . . . . . . . . . . . Group Size and Composition . . . . . . . . . . . . . . . . . . . . . . . . Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1 6 35 48 66 67

Odor-Genes Covariance and Genetic Relatedness Assessments: Rethinking Odor-Based Recognition Mechanisms in Rodents JOSEPHINE TODRANK AND GIORA HETH I. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . II. Odor-Genes Covariance . . . . . . . . . . . . . . . . . . . . . . . . . . . . III. Mechanisms Underlying Differential Behavioral Responses to Individual Odors. . . . . . . . . . . . . . . . . . . . . . . IV. Rethinking Terminology Associated with Odor-Based Mechanisms Underlying Differential Behavior Responses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . V. Speculations on the Origin and Evolution of Preferential Responses Based on G-Ratios and Their Function as a Premating Isolating Mechanism . . . . . . . . . . . . . . . . . . . . . . VI. Prospects for Future Studies Relating to G-Ratios . . . . . . VII. Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

77 80 92

106

114 119 123 125

Sex Role Reversal in Pipefish ANDERS BERGLUND AND GUNILLA ROSENQVIST I. Mate Competition and Sex Roles . . . . . . . . . . . . . . . . . . . . II. Female Ornaments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . v

131 132

vi

CONTENTS

III. IV. V. VI. VII. VIII. IX. X. XI. XII. XIII.

Syngnathic Phylogeny. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Sex Roles in Syngnathids . . . . . . . . . . . . . . . . . . . . . . . . . . . . The Two Pipefish Species . . . . . . . . . . . . . . . . . . . . . . . . . . . Parental Investment, Potential Reproductive Rates, the Operational Sex Ratio, and the Bateman Gradient . . . . . . Female Competition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Male Choosiness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Mate Choice and Parasites . . . . . . . . . . . . . . . . . . . . . . . . . . Mate Choice and Offspring Quality . . . . . . . . . . . . . . . . . . . Ornament in Female Syngnathus typhle . . . . . . . . . . . . . . . . Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

134 135 136 141 145 146 149 150 151 160 161 162

Fluctuating Asymmetry, Animal Behavior, and Evolution JOHN P. SWADDLE I. II. III. IV.

What Is Fluctuating Asymmetry and Why Is It Interesting? . Fluctuating Asymmetry and Fitness . . . . . . . . . . . . . . . . . . . Methodology Issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Important Gaps in Our Knowledge about Fluctuating Asymmetry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . V. A Revised Look at Fluctuating Asymmetry and Sexual Selection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . VI. Fluctuating Asymmetry, Animal Behavior, and Evolution VII. Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

169 171 173 178 187 194 196 198

From Dwarf Hamster to Daddy: The Intersection of Ecology, Evolution, and Physiology That Produces Paternal Behavior KATHERINE E. WYNNE-EDWARDS I. II. III. IV. V. VI. VII.

Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Natural History . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Evolution of Biparental Care . . . . . . . . . . . . . . . . . . . . . . . . Endocrine Evolution in Phodopus campbelli . . . . . . . . . . . . Men Becoming Fathers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Sex Specificity in Endocrinology . . . . . . . . . . . . . . . . . . . . . . Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

207 208 219 227 243 251 252 254

CONTENTS

vii

Paternal Behavior and Aggression: Endocrine Mechanisms and Nongenomic Transmission of Behavior CATHERINE A. MARLER, JANET K. BESTER-MEREDITH, AND BRIAN C. TRAINOR I. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . II. Testosterone: An Aggression Hormone, A Nurturing Hormone, or Both? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . III. Arginine Vasopressin: Functionally Similar to Testosterone? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . IV. Cross-Generational Transmission of Aggression through Behavioral Mechanisms and the Role of Arginine Vasopressin . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . V. Role of Plasticity in Paternal Behavior and Arginine Vasopressin in the Nongenomic Transmission of Aggression across Multiple Generations in Peromyscus . . VI. Summary of Nongenomic Transmission of Aggression and Paternal Behavior across Generations and the Role of Arginine Vasopressin . . . . . . . . . . . . . . . . . . . . . . . VII. Conclusions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . VIII. Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

263 267 284

298

304

307 308 311 312

Cognitive Ecology: Foraging in Hummingbirds as a Model System SUSAN D. HEALY AND T. ANDREW HURLY I. II. III. IV. V.

Learning and Memory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Spatial Learning and Memory . . . . . . . . . . . . . . . . . . . . . . . Timing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Conclusions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

329 339 349 352 353

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

354

Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

361

Contents of Previous Volumes . . . . . . . . . . . . . . . . . . . . . . .

377

Contributors Numbers in parentheses indiate pages on which the authors’ contributions begin.

ANDERS BERGLUND (131), Department of Animal Ecology, Uppsala University, Uppsala SE-752 36, Sweden JANET K. BESTER-MEREDITH (263), Department of Psychology, University of Wisconsin—Madison, Madison, Wisconsin 53706 IAIN D. COUZIN (1), Department of Ecology and Evolutionary Biology, Princeton University, Princeton, New Jersey 08644 SUSAN D. HEALY (325), Institute of Cell, Animal, and Population Biology, University of Edinburgh, Edinburgh EH9 3JT, United Kingdom GIORA HETH (77), Institute of Evolution, University of Haifa, Haifa 31905, Israel T. ANDREW HURLY (325), Department of Biological Sciences, University of Lethbridge, Lethbridge, Alberta T1K 3M4, Canada JENS KRAUSE (1), Centre for Biodiversity and Conservation, School of Biology, University of Leeds, Leeds LS2 9JT, United Kingdom CATHERINE A. MARLER (263), Department of Psychology and Department of Zoology, University of Wisconsin—Madison, Madison, Wisconsin 53706 GUNILLA ROSENQVIST (131), Department of Biology, Norwegian University of Science and Technology, Trondheim N-7491, Norway JOHN P. SWADDLE (169), Institute for Integrative Bird Behavior Studies, Biology Department, College of William and Mary, Williamsburg, Virginia 23187 JOSEPHINE TODRANK (77), Institute of Evolution, University of Haifa, Haifa 31905, Israel BRIAN C. TRAINOR (263), Department of Psychology, University of Wisconsin—Madison, Madison, Wisconsin 53706 KATHERINE E. WYNNE-EDWARDS (207), Department of Biology, Queen’s University, Kingston, Ontario K7L 3N6, Canada ix

Preface

The aim of Advances remains as it has been since the series began: to serve the increasing number of scientists who are engaged in the study of animal behavior by presenting their theoretical ideas and research to their colleagues and to those in neighboring fields. We hope that the series will continue its ‘‘contribution to the development of cooperation and communication among scientists in our field,’’ as its intended role was phrased in the Preface to the first volume in 1965. Since that time, traditional areas of animal behavior research have achieved new vigor by the links they have formed with related fields and by the closer relationship that now exists between those studying animal and human subjects. Scientists studying behavior today range more widely than ever before: from ecologists and evolutionary biologists, to geneticists, endocrinologists, pharmacologists, neurobiologists, not forgetting the ethologists and comparative psychologists whose prime domain the subject is. It is our intention not to focus narrowly on one or a few of these fields, but to publish articles covering the best behavioral work from a broad spectrum. The skills and concepts of scientists in such diverse fields necessarily differ, making the task of developing cooperation and communication among them a difficult one. But it is one that is of great importance, and one to which the Editors and publishers of Advances in the Study of Behavior are committed. We will continue to provide the means to this end by publishing critical reviews, by inviting extended presentations of significant research programs, by encouraging the writing of theoretical syntheses and reformulations of persistent problems, and by highlighting especially penetrating research that introduces important new concepts. The present volume addresses a wide range of topics. The chapters by Marler and her colleagues and by Wynne-Edwards provide two different perspectives on paternal behavior in mammals, a subject that is challenging from both the mechanistic and functional viewpoints. Paternal behavior is also featured in the chapter by Berglund and Rosenqvist, who discuss the remarkable reproduction of pipefish in which male care of the offspring is taken to an extreme that can reasonably be described as ‘‘pregnancy.’’ Chapters that deal with especially persistent problems are the critical look at fluctuating asymmetry provided by Swaddle, the description that Healy and Hurly provide of the rules of foraging in hummingbirds, a nice example of controlled experiments carried out on animals in the wild, and the discussion of odor recognition mechanisms in rodents by Todrank and xi

xii

PREFACE

Heth. Couzin and Krause, on the other hand, describe a relatively young field, that of collective behavior: the remarkable organization and patterning that occurs in the behavior of groups of animals. There are striking emergent properties when individuals making their own decisions come together in groups, raising general principles of significance to many animal species, including humans.

ADVANCES IN THE STUDY OF BEHAVIOR, VOL. 32

Self-Organization and Collective Behavior in Vertebrates Iain D. Couzin1 and Jens Krause2 1

department of ecology and evolutionary biology princeton university princeton, new jersey 08544 2 centre for biodiversity and conservation school of biology university of leeds leeds ls2 9jt, united kingdom

I. Introduction A. Overview As a ripple of light the fish turn. Like some animate fluid, the school glides and turns again. The synchrony of motion is captivating. A similar integration of behavior can be seen in a bird flock. The volume and shape of the group change as the group turns and arcs overhead, and yet the aggregate remains cohesive. Many group-living vertebrates exhibit complex, coordinated, spatiotemporal patterns, from the motion of fish and birds, to migrating herds of social ungulates and patterns of traffic flow in human crowds. The common property of these apparently unrelated biological phenomena is that of interindividual interaction, by which individuals can influence the behavior of other group members. It is on how these interactions result in the collective behaviors of vertebrate animal groups that we focus here. Specifically, we consider systems in which insights from self-organization theory have been useful in improving our understanding of the underlying mechanics. Self-organization theory suggests that much of complex group behavior may be coordinated by relatively simple interactions among the members of the group. According to this theory, the form, and therefore often the function, of the collective structure is encoded in generative behavioral rules. Self-organization has been defined 1 Copyright 2003 Elsevier Science (USA). All rights reserved. 0065-3454/03 $35.00

2

IAIN D. COUZIN AND JENS KRAUSE

as ‘‘a process in which pattern at the global level of a system emerges solely from numerous interactions among the lower-level components of a system. Moreover, the rules specifying interactions among the system’s components are executed using only local information, without reference to the global pattern’’ (Camazine et al., 2001). It should be noted that often in nature, pattern-forming processes may not strictly conform to this classification: in some instances, such as animal migration, individuals may modify their local (self-organizing) interactions with others with reference to global information, such as a general desire to move in a certain direction. This type of system therefore self-organizes within the context of global cues. There has been expanding interest in pattern formation in biological systems (Gerhard and Kirshner, 1997; Maini and Othmer, 2000; Camazine et al., 2001). The study of pattern formation covers a wide range of areas, including attempting to explain fetal development (Keynes and Stern, 1988), patterns on the coats of mammals (Murray, 1981), the structure of social insect nests (Theraulaz and Bonabeau, 1995), and the collective swarms of bacteria (Ben-Jacob et al., 1994), army ants (Deneubourg et al., 1989), and locusts (Collett et al., 1998). In particular there is growing interest in the relationship between individual and population-level properties. A fundamental question is how large-scale patterns are generated by the actions and interactions of the individual components. Many pattern-forming processes in biological systems, such as cellular sorting or the collective organization of group-living (particularly eusocial) insects, are dynamic mechanisms whereby the large-scale patterns [e.g., clustering of cell types (Glazier and Graner, 1993) or periodic activity cycles in ant colonies (Boi et al., 1999)] can be accounted for by the interactions among the individual components of the system (e.g., differential adhesion among cells; ants responding locally to the activity of others). Applying such a self-organization viewpoint to vertebrate groupings is a more recent development, and despite the importance of understanding group dynamics for ecological processes (Levin, 1999), many collective behaviors are still only qualitatively understood. Vertebrates often have superior cognitive abilities and more complex behavior patterns than organisms such as social insects. Consequently it may appear that this approach may be less able to account for the collective behaviors of these organisms. However, the self-organization approach is applicable to even the most complex of organisms, such as humans, but is restricted to certain aspects of their behavior, such as the motion of pedestrians within crowds (see Sections II.B.1 and II.C), where interactions may be (mechanistically) relatively simple. A further reason that vertebrate groups have been less well studied in this context is that for many vertebrate groups, such as

SELF-ORGANIZATION AND COLLECTIVE BEHAVIOR IN VERTEBRATES

3

ungulate herds, pelagic fish schools, or human crowds, the interactions among the individuals are much harder to study than those in group-living insects, or bacterial swarms, where the manipulative experiments required to understand the underlying mechanisms better are easier to perform (and replicate). Here we review progress in this newly emerging area of study: that of applying self-organization theory to mobile vertebrate groups composed of many interacting individuals (such as bird flocks, ungulate herds, fish schools, and human crowds) in an attempt to improve our understanding of underlying organizational principles. B. Understanding the Dynamics of Collective Behavior Mathematical modeling is becoming increasingly recognized as an important research tool when studying collective behavior. This is because it is usually not possible to predict how the interactions among a large number of components within a system result in population-level properties. Such systems often exhibit a recursive, nonlinear relationship between the individual behavior and collective (‘‘higher order’’) properties generated by these interactions; the individual interactions create a larger scale structure, which influences the behavior of individuals, which changes the higher order structure, and so on. Consider the movement of ungulates across grassland, or over snow-covered terrain. The motion of an individual is likely to change the environment through which it moves (by compression of the grass or snow). This local change influences the motion of other individuals passing near that point: they exhibit a tendency to maximize their comfort of travel (and hence minimize energy expenditure) and thus have a greater propensity to move over the ground previously walked on. This results in further changes to the environment at that point (further compression of the substratum), which in turn increases the probability of others to choose to move over that point if close to it. Taken over a larger area, this feedback results in the generation, and use of, trail structures. Thus individuals change the local properties of their environment, which influences the motion of others, which further alters the environment, and so on. The generation of animal (including human) trails is discussed in more detail later, and the results of computer models are used to reveal the dynamics of this system. When modeling population-level processes, continuum approaches (‘‘Eulerian’’ models) have typically been used. These abstract the movement of large populations to population densities, and movement is usually represented by diffusion and advection processes. Such approximation procedures are useful, because there are well-developed

4


mathematical tools for their analysis. Although well suited to the movement of large populations (e.g., bacterial, planktonic, and certain insect aggregations), they are less suitable for most vertebrate animal groups, which consist of a relatively small number of individuals. Furthermore, the analysis of such models is typically greatly complicated when social interactions, or interactions between individuals and their environment, are an important organizing mechanism. Consequently, here we consider primarily the motion of groups as resulting from interactions among the individual group members and use, where appropriate, individual-based (or ‘‘Lagrangian’’) models of animal motion to elucidate certain (often generic) principles. This approach to modeling shares certain properties with techniques developed in nonlinear statistical physics to simulate the motion of particles, as for example in gases, fluids, or magnets. While particles may be subject to physical forces, animal behavior can conceptually be considered to result from individuals responding to ‘‘social forces,’’ for example, the positions and orientations of neighbors, internal motivations (e.g., degree of hunger), and external stimuli (such as the positions of obstacles). In understanding the movement decisions of animals we must better understand how and why motivations exist, and how these translate to collective patterns. The global level (‘‘emergent’’) dynamics of the group are usually not explicitly encoded: there is often no global blueprint or template for the pattern (although the formation of trails, as described, may to some degree be considered as the generation of an interactive, labile template). The form of the collective structure, and hence often the function, is usually encoded in generative behavioral rules. Such rules, being subject to natural selection, allow the generation of self-organized adaptive patterns at the group level. Because the costs and benefits to individuals when grouping may change dynamically, even as a function of the position of an individual relative to other group members, changes in individual rules are likely to occur as group members attempt to maximize their individual fitness. This can result in groups adopting different shapes, or motions, as well as being a potential driving force for internal structuring within vertebrate groups. Such properties are also discussed here. Environmental factors, such as physical habitat structure or temperature, may influence the behavior of individuals within groups, and consequently their motion and structure. These factors may affect the cohesion of groups, or act as ‘‘seeds’’ for self-organized aggregation processes. Individuals may balance global goal-oriented behavior (such as a desire to move up a temperature gradient) with local conditions, such as avoidance of isolation from a group, or alignment with group members. Such a balance of external and internal social forces may underlie the


5

motion of certain vertebrate groups, such as migrating fish schools. The structure of the environment through which individuals move is also important. In some cases, the spatial heterogeneity in the environment may be temporally stable (relative to the timescale over which grouping mechanisms function), such as the positions of trees, rocks, and other landmarks. Such heterogeneity may influence both the suitability of the environment for locomotion, and the effective range of interaction among individuals. This variability is likely to have a strong influence on both individual movement patterns and interaction range. In other cases spatial variation in habitat is dynamic, such as the flows and turbulent eddies within certain aquatic environments. A further important factor to consider when understanding the collective behaviors of animal groups (and self-organized pattern-forming processes in general) is the influence of stochastic (random) events. Animal behavior is inherently probabilistic, and stochastic properties of animal movement are likely to strongly influence the structure of many vertebrate groups. It is becoming increasingly evident that self-organized patterns often arise because of the amplification of random fluctuation (Nicolis and Prigogine, 1977; Seeley, 1995), as is discussed here when we consider the shape of migrating wildebeest herds. By developing stochastic computer models of animal groups the essential statistical mechanics of the system may be captured. The aim of modeling is often not to attempt to include all the known properties of a system, but rather to capture the essence of the biological organizing principles. One of the principal aims of self-organization theory is to find the simplest explanation for complex collective phenomena. A commonly perceived problem when modeling animal behavior, especially that of humans, is that of the representation of complex organisms through simple behavioral rules. The apparent complexity of the entities to be represented in a computer model may be misleading, however. To gain insight into the dynamics of a collective phenomenon, all of the complex details may not be necessary or even relevant. For example, much of human behavior within crowds is carried out almost automatically with little conscious decision making, and although the organism is complex, the interactions need not necessarily be so. Furthermore, when exploring potential grouping mechanisms it is often useful to deliberately explore a simplified representation of the system that characterizes a broader range of general mechanisms. That a biological population is described as being self-organized does not suggest that all individuals within the population are simple, identical, or have the same influence on one another. Of course, this is not to say that more specific representations of certain systems are not important. On the contrary, developing models of specific cases of a broader mechanism is extremely

6


valuable. However, there are currently often limitations in the quality of empirical information available, and thus creating a generalized model can often be more appropriate. Also, without an understanding of the behavior of the simplest system we cannot possibly know how changes made to the model affect its behavior. Even with relatively few parameters, the exploration of parameter space can be time-consuming and complex. A further point to bear in mind is that with collective systems, understanding the behavior of an individual in isolation does not necessarily provide information about the properties of that individual within a collective situation, where nonlinear interactions may determine much of the group dynamics.

II. Group Shape and Motion A. Wavelike Front of Migrating Wildebeest Herds Many collective behaviors result in complex, and coordinated, spatiotemporal patterns, from an undulating flock of birds to mobile herds of social ungulates and lanes of traffic flow in human crowds. One of the most dramatic examples of collective motion in vertebrates is that of migrating wildebeest (Connochaetes taurinus) that form huge herds that cross the Serengeti grasslands, moving to the north in May to June, and returning south in November. A single herd may include in excess of 100,000 individuals that, viewed from above, exhibit a common direction of motion, and a broad front that exhibits a characteristic wavelike form (Fig. 1). Interestingly, the wavelength of this front pattern is much larger than the possible interaction range of an individual. To gain a better understanding of how this group shape may be generated, Gueron and Levin (1993) developed a mathematical model of the herd front. They made the simplifying assumptions that individuals have a common directional preference and that it is likely to be the motion of individuals at the front of the group (leaders) that primarily explains this pattern. How certain individuals within a group may become leaders, and the influence of leadership within vertebrate groups, is discussed more fully in Sections III.B and III.D. Because these migrating wildebeest herds are so large, Gueron and Levin abstracted the herd front to a curve evolving in time and space, making the system tractable to mathematical analysis. Given that the phenomenon of interest, the wavy front, has a periodicity much greater than that of the supposed interaction range, individuals within the model were restricted to modify their motion in their desired direction only as a


7

Fig. 1. Herd of wildebeest showing a wavelike herd front. (From Sinclair, 1977; used with permission.)

function of that of neighbors within a specified range (the average location of individuals in a local neighborhood). Thus it was possible, using the model, to investigate the potential influence of the range of interaction. Using the following simple rule set, the model was found to generate a wavelike front: 1. Individuals have an intrinsic speed and accelerate or decelerate in response to the positions of neighbors within a local neighborhood. 2. Those lagging behind others in their local neighborhood can fall further behind, until the gap reaches a specified maximum distance. When this distance is reached individuals speed up to reduce the gap. 3. Those ahead of neighbors can speed up until the gap reaches a specified maximum distance, where the behavior is reversed and they slow down. According to these rules, an initial uniform herd front (straight line) is unstable. Small perturbations (stochastic irregularities) in the curve representing the herd front appear and tend to grow (amplify), yielding the irregular ‘‘wavy’’ fronts (Fig. 2), similar in appearance to those seen in nature (Fig. 1). Thus a simple and local set of behavioral rules can explain the long-range pattern. Importantly, the solutions to the model were ‘‘semistable,’’ meaning that although they did change over time, the characteristic feature of the system (the presence of the irregular wavy

8


Fig. 2. Model of a wildebeest herd front that produces wavelike herd fronts from initially aligned individuals (along the x axis). (Modified from Gueron and Levin, 1993.)

front) was persistent. The exact shape of the front was found to be largely dependent on the range of interaction, and the model was better able to represent the waveform seen in reality with relatively local interactions (the front becoming flatter as the range of interaction increases). Consequently the model makes broad-level predictions about the type of behavioral interactions herding wildebeest may exhibit, and about how interaction range affects group shape in this system. Furthermore, the model makes it clear that long-range patterns need not be explained by long-range interactions. However, it is unclear how such predictions may be tested in practical terms given the huge spatial scale of the system in question, and therefore, whether alternative local rules may also explain this phenomenon. In this case it may be difficult to test the predictions of computer models. Ideally it would be desirable to manipulate the system such that it would be possible to investigate the consequences of changing the parameters of the model (e.g., manipulate the interaction range of real organisms) or, more realistically perhaps, to track the motion of individuals within a subset of the herd to see whether they conform to the type of local interactions assumed by the model. This may be achieved by recording how the velocity of individuals depends on the velocities of neighbors (bearing in mind that velocity incorporates the position of an individual, its direction of motion, and its speed). We discuss the analysis of such groups in more detail in Section III.A. Gueron and Levin also point out that their type of approach may be relevant to understanding the motion of narrow bands of animals only one or a few individuals in width, such as thin streams of ungulates, birds, or bats. The direction of travel would then be considered to be perpendicular to that in their wildebeest model, and it would be assumed that individuals tend to adjust their position to either side, relative to individuals ahead of and behind them, for example, to avoid collisions or perhaps to improve


9

visual range. Given this formulation, individuals would tend to move to one side (e.g., to avoid collisions) until they reach a maximum distance, at which time they would tend to move back toward the group (representing a tendency to avoid being isolated). In this context, as suggested by Gueron and Levin, it may be beneficial to modify the rules of interaction such that individuals tend to predominantly respond to those ahead (as opposed to equally to those ahead and behind, if the original model were to be abstracted exactly). However, similar predictions are likely to result: that perturbations tend to grow, resulting in winding, as opposed to straight, lines and that the exact form of the wave will similarly depend on the interaction range. B. Generation and Use of Collective Trail Systems by Animals, Including Humans A further property that influences the motion of organisms such as herding ungulates is their ability to change the environment through which they move, and to respond to such changes. This recursive feedback loop may also be an important determinant of the types of patterns that form at the population level. Consider the type of situation outlined previously in Section I.B, in which individuals change local properties of their environment as they move through it, such as by trampling grass or snow. As well as responding to the positions of other group members, individuals respond to their environment. We are not aware of any mathematical approach that has been applied to this problem for organisms that actively aggregate, and thus the theoretical consequence of the balance of these forces has not yet been investigated. However, progress has been made in cases in which the effects of direct interactions between individuals can be ignored, such as when they are rare. Initially this may seem irrelevant in a chapter on animal groups, in which interactions are known to be important. However, because the only work on this topic has made this assumption, it is still beneficial to understand what pattern-forming processes occur when direct interactions are a trivial influence. Furthermore, even when individuals themselves do not interact directly, the pattern-forming mechanism is still collective through indirect interactions by environmental modification. We therefore make suggestions about suitable modifications of this approach to include direct interactions, and the potential outcome of such modifications, to this type of model. 1. Human Trails Helbing et al. (1997a,b) developed a model of trail formation by mobile individuals (or ‘‘active walkers’’). These walkers were considered to have

10


Fig. 3. Computer simulation of human trail systems. (A) Initially walkers will take moreor-less direct routes between the four entry/exit points in the corners. The instantaneous velocities of walkers are shown as arrowheads. (B) After a period of time a shared trail system forms. (From Helbing et al., 1997b; used with permission.)

the potential to modify the environment through which they move. In their model these changes represent the trampling of substratum as described previously. Such a model can be used to investigate the influence of the degree to which individuals change their environment (and consequently the effect this change has on others) and the lifetime of the environmental changes (simulating the local durability of a change; e.g., the regrowth of vegetation or further falls of snow will act to return the environment to its former state). Although Helbing et al. (1997a,b) restrict their discussion of vertebrate trails to those generated by humans, the type of model is applicable to any system in which individuals can modify their environment and respond to such modifications. For humans, a situation can be considered in which people move between certain points in space, each, for example, representing doors to buildings. Within the model developed by Helbing et al. (1997a) it is assumed that people will tend to take the shortest route to their destination but tend to reconcile this global goal-oriented behavior with a relatively local preference to walk on previously used (less bumpy) ground. They considered the movement of simulated pedestrians over initially homogeneous ground from, and to, specific points in space. Figure 3A shows a trail system forming near the beginning of a simulated run in the case in which there are four entrance/destination points, one in each corner of the simulated domain. As can be seen, pedestrians initially tend to take the direct route to their destinations. Over time, however (Fig. 3B), frequently


11

used trails become more comfortable (and hence attractive) and this influences the characteristics of the trail system: sections of trails become shared by walkers with different desired routes, creating a trail system in which the overall length is reduced. Increasing comfort means that a given section of trail is more likely to be used in future, which further increases its comfort, and so on (autocatalysis). Trails that are not sufficiently reinforced will decay through processes such as the regeneration of vegetation and weathering effects, thus providing a negative feedback within the system. The exact type of collective trail system that forms will depend on the properties of the system that affect these feedbacks. For example, increasing the number of pedestrians within the environment, or the degree to which pedestrians influence the comfort of the ground over which they walk, will increase the positive feedback. Increasing the rate of recovery of the environment, by contrast, will amplify negative feedback. This will influence how individuals reconcile their global and local behavioral tendencies. If the desire of pedestrians to use existing trails is great, the final trail system will be a minimal way system (the shortest system that connects all the points). Conversely, if there is no advantage to using trails (as in most urban environments) individuals will use a direct route system (similar to that seen in Fig. 3A). In between these extremes, the simulation suggests that pedestrians collectively will find a compromise between short and comfortable ways. 2. Extending Trail-Laying/Response Concepts to Other Animals This type of modeling approach is similar to earlier studies investigating the generation of trails by ants (Deneubourg et al., 1989; Franks et al., 1991). In the latter case positive feedback (amplification) can occur when the orientation of a trail-laying ant depositing chemical pheromone at a certain location influences the direction taken by a further ant passing that point. The latter ant may reinforce the pheromone trail, which can further influence the direction taken by subsequent ants, and so on. This can lead to a selection of trail orientation at that location. Pheromones decay, causing negative feedback. A decaying trail is less likely to be followed and will therefore be subject to further decay. As in the case of human trails, trail persistence depends on the balance between reinforcement (positive feedback) and decay (negative feedback). Thus at a certain level of description, vertebrate and insect trail laying may share some fundamental properties of organization. Consequently, the results of research on ant trail systems may also shed light on fundamental properties of trail formation by vertebrates. For example, it has been found that ants can find the shorter of two routes between the colony and a food source. Ungulates may create and use trails as they

12


Fig. 4. (A) A model of trail formation from Helbing et al., 1997b (used with permission), and (B) natural trails made by ungulates. (Copyright ß 2002, Iain Couzin.)

move between feeding areas or watering holes and, as described, humans may move between buildings. Figure 4A shows trails forming in accordance with the model of Helbing et al. (1997a), in which individuals have a desire to move from the top to the bottom of the domain, and vice versa. Figure 4B shows similar natural trail systems used by ungulates in Australia. The potential consequences of such dynamics can be considered in a hypothetical, and deliberately simple, situation as shown in Fig. 5A, in which individuals are considered to desire to move between just two points labeled 1 and 2, and vice versa, but in doing so must move around an impassable landmark in the center. Individuals create and follow trails as described. Initially, in the absence of any trails, individuals first reaching the landmark, having limited and local visual information, will randomly select a route around it (Fig. 5B). However, those individuals that take the


13

Fig. 5. (A) Hypothetical scenario in which individuals move between points 1 and 2, around an obstacle in between. (B) Initially individuals will randomly select a direction around the obstacle. However, the shorter route is reinforced more quickly. (C) The feedback in the trail formation-following system means that the shortest route can be selected collectively.

shorter route will reach their destination more quickly. This causes that route to be more rapidly reinforced. This means further individuals reaching the point at which trails bifurcate to the right and left around the landmark will tend to be more ‘‘attracted’’ to the shorter route, which will become even more attractive, and so on. Thus the counterbalance of positive and negative feedback could be expected to facilitate the collective selection of the shortest trail to a specific point, without individual decision making being invoked (Fig. 5C). A further property highlighted by research on ant recruitment mechanisms that may also relate to vertebrate trails is that of a trade-off between accurate and rapid decision making. Consider our earlier simplistic scenario, involving the navigation of organisms between two points in space around an asymmetric obstacle. If positive feedback is high (trails are attractive, as would be the case, e.g., if the ground is difficult to move over unless a path made previously is followed) then the trailforming system is susceptible to initial conditions. For example, if the first individual to pass the obstacle were, by chance, to go the long way around, then it would be likely that the next individual would also take that route. This would cause rapid fixation of the longer route. Thus the system would be dependent on the initial (random) choices of individuals. If each individual were to have a weak effect on the ground, or be only weakly attracted to the trails of others, then it will take a much longer time for one particular route around the obstacle to be dominant, but it is likely that it will be the shorter route that is ‘‘chosen.’’ A further point to be made is

14


that even given a symmetric obstacle, if positive feedback is relatively strong, then the organisms are still likely to select a (randomly determined) single route around the obstacle. We are not suggesting here that vertebrates behave just like ants, and the predictions we have made are deliberately speculative. Caution should be used when extrapolating the results between different systems. However, the presence of similar fundamental feedback mechanisms may mean that, as demonstrated by human trail formation, some collective processes exhibited by vertebrate populations may be explained without invoking complex decision-making abilities at the level of the individual. A further potential property of vertebrate trail systems that has yet to be investigated is the influence of direct interactions among the components of the system. Intuitively, it may be expected that herding behavior would tend to increase the amplification processes involved in trail formation because individuals would tend to remain in the proximity of others and would tend to follow one another. However, in the case of bidirectional traffic on a trail, congestion may cause trails to bifurcate more readily, creating a system with a series of anastomosing trails, as opposed to a single trail. However, it would be important to further develop models of these processes, and to find systems in which it would be possible to test the predictions of computer models. For example, it may be possible to compare trails made by organisms moving over vegetation that offers different resistance to locomotion (and consequently the ease of creation and relative comfort of trails). It may also be possible to set up experiments similar to that shown in Fig. 5 and investigate the collective solutions ‘‘found’’ by the organisms in question. In some cases vertebrates may deposit chemicals that can facilitate trail formation (or complement the mechanisms discussed previously). For example, Norway rats, Rattus norvegicus (Galef and Buckley, 1996), and naked mole rats, Heterocephalus glaber (Judd and Sherman, 1996), having found a food source, can deposit odor that can bias the direction taken by other individuals, somewhat analogous to trail deposition and following that by ants and termites. Being central-place foragers, such trails can facilitate information transfer to other (naı¨ve) individuals about the location of resources. It is known in rats that deposition of scent in urine is used as a trail marker (Wallace et al., 2002), whereas in naked mole rats it is unclear how the scent is deposited by the individual (Judd and Sherman, 1996). For rats it has also been shown that the attractiveness of a trail increases as a function of how many times a trail section has been traversed, and that rats deposit trail markers only when moving away from the food source (Galef and Buckley, 1996). Furthermore, it has been shown that the odor-discriminatory ability of rats allows them to distinguish between


15

self-generated trails and those of conspecifics (Wallace et al., 2002) Such an ability is also known to occur in certain ant species (Ho¨lldobler and Wilson, 1990) and could be useful for rats when searching within dark environments, in that it may allow them to retrace their trajectory. Further studies of Norway rats and naked mole rats are likely to provide an excellent basis for future research efforts because they can be more readily kept in captivity, and their experimental conditions can be more easily manipulated than those of larger organisms such as ungulates. It is possible that trail deposition and following may be widespread in rodents and may be combined with environmental modification such as trampling or removal of obstacles from the environment. Scent deposition and detection may also be important in other vertebrate trail systems, such as those already discussed for ungulates. A further extension of the trail formation concepts discussed here can be made to include collective burrow systems, such as those made by naked mole rats. Here, individuals modify their environment by digging, and an unmodified environment would need to be considered as resistant to motion. A further modification of the previous concepts would be that the environment would not return to its former state once modified (or would do so only extremely slowly). 3. Collective Generation of Home Ranges through Deposition of, and Response to, Scent Another collective biological phenomenon that relies on the modification, and response to modification, of the environment is the generation of home ranges by vertebrates such as carnivores (Gosling and Roberts, 2001; MacDonald, 2001) and rodents (Brown and MacDonald, 1985; Viitala et al., 1995; Gray and Hurst, 1997). In this case individuals, or groups of individuals in the case of pack-living canids (such as coyote or wolf), mark their territory with scent, which diffuses over time. The motion of individuals is dependent on the scent they detect as they move: they will tend to turn around (and hence not occupy space) in which they detect the scent of another individual (or group of individuals in the case of pack animals). Long-range patterns of space use result from these local interactions. Moorcroft et al. (1999) developed a mechanistic model based on these basic principles. Individuals were assumed to increase the degree to which they scent marked after interaction with the scent of another individual/group. Encounters with such foreign scent marks would also bias the trajectory of individuals toward the center of their own home range. Using an Eulerian approach [using partial differential equations (PDEs)], they showed that these rules were sufficient to explain the general properties of territory generation, and were better suited for their

16


experimental system (the coyote, Canis latrans, in the Hanford Site Arid Lands Ecology Reserve, southeast Washington State) than models in which individuals did not bias their movement after discovering foreign marks. C. Collective Behavior of Humans within Crowds In our earlier discussion of humans we considered the case in which people interact through environmental modification (trail formation), and largely ignored the influence of direct interactions among pedestrians. However, within an urban setting individuals can seldom influence their surroundings in this way. Furthermore, when walking down a busy street, or corridor, a balance is struck between global goal-oriented behavior (desire to reach a certain point) and local conditions created by the motion and positions of other nearby pedestrians. Each member in such a crowd is likely to have a limited perceptive radius in which information to determine future movement must be gathered. Consequently, larger scale patterns in crowds are seldom evident from the viewpoint of an individual pedestrian. However, if viewed from above crowds often do display obvious and consistent patterns. One of the most common of these can be seen when there is bidirectional traffic, as, for example, when people are trying to move both ways along a walkway, or crossing the road at a crosswalk. Under such circumstances ‘‘bands’’ of pedestrians form: each band is composed of a number of pedestrians with a common directional preference (Milgram and Toch, 1969). See Fig. 6A. The flow of pedestrians under conditions of crowding was likened by Henderson (1971) to the motion of fluids or gases. Henderson used a wellknown technique for the mathematical analysis of such materials, the Navier–Stokes equation for fluid dynamics, to simulate a crowd. Although providing an insight into how individual-level (microscopic) properties lead to large-scale (macroscopic) properties, such an approach is difficult to implement because the conservation of energy and momentum assumptions for a physical system do not apply to a biological system in which the individual components are ‘‘self-driven.’’ Despite this, Helbing (1992) was able to modify such equations with respect to some of these properties, but analytic solutions proved difficult to find. The most promising approach to studying crowd behaviors comes from individualbased modeling. 1. Influence of Repulsion: Collision Avoidance Helbing and Molna´r (1995) developed a simple individual-based model of pedestrian motion in which they consider people moving in opposite


17

Fig. 6. (A) Flow of pedestrians in a crowd. (B) Simulation of pedestrain dynamics, showing lane formation (from Couzin, 1999). The successive positions (trajectories) of individuals with a desire to move to the left are shown in gray. The positions of those individuals intending to move to the right are shown in black.

directions along a corridor. This simple geometric representation of space allows the assumption that all individuals have a desire to move only in one direction or another along the walkway. However, pedestrians will also tend to avoid collisions by decelerating and turning away if they come into close contact with one another. When no other individuals are within a specified local range, individuals will tend to accelerate to a desired speed, and orient toward their destination. This simple behavioral response alone can account for the formation of bands when there is bidirectional traffic. Individuals meeting others head on will have ‘‘strong’’ interactions, in which they are likely to slow down and move aside to avoid collisions. Initially this occurs frequently. However, individuals who find themselves behind others moving in the same direction are less likely to have to perform such extreme avoidance maneuvers, and in turn they ‘‘protect’’ others behind them from head-on avoidance moves. Given a sufficiently long corridor (and a sufficiently high traffic flow for interactions among pedestrians to be an important factor) the system will self-organize into lanes. Individuals entering the corridor (at random positions) move around in the direction perpendicular to their desired direction of travel when they interact with oncoming pedestrians. However, if by

18


chance they fall in behind another individual moving in the same direction this is a more ‘‘stable’’ state. Thus the system naturally selforganizes into a situation in which pedestrians are in the ‘‘slipstream’’ of others moving in the same direction as themselves, thus creating bands, and reducing movement in the direction perpendicular to desired motion (Fig. 6B). Helbing and Molna´r (1995) also demonstrated in their model that the number of bands that tend to form scales linearly with the width of the walkway. This demonstrates that there is a characteristic length scale to the pattern-forming process: that is, from any point in the system statistically similar motions occur one wavelength away. 2. Influence of Attraction to Other Pedestrians Clearly it is not necessary to invoke complex individual behavior to explain the banding patterns found in human crowds. The preceding model shows how individuals would ‘‘naturally’’ occupy space (in the dimension perpendicular to desired direction of travel) in which others ahead and behind them tend to have a similar direction of motion. It is possible in real crowds, however, that individuals actively (as well as passively) seek such positions. That is, instead of finding such positions by chance, as in the previous model, they will tend deliberately to walk behind individuals moving in the same direction as themselves. For example, Couzin (1999) simulated the motion of pedestrians crossing a road at a crosswalk. Given the type of rules described in Section II.C.1 the system requires some time to ‘‘find’’ the collision-minimization state. Consequently, in the crosswalk situation, although some banding does occur, congestion is still relatively high (Fig. 7A). However, if a supplementary rule is added such that an individual will exhibit a propensity to follow other individuals moving in their desired direction, then bands tend to form much more readily, thus reducing head-on collisions and increasing the rate of flow (Fig. 7B). On a crosswalk, such bands begin to form even before the pedestrians moving in different directions meet. Thus the groups act as ‘‘wedges’’ when they come into contact with one another, allowing the bands to interlace more readily when they reach the central area of the walkway. Thus, although attraction is not a necessary condition for bands to form in crowds, it decreases the time taken for bands to develop, and increases the flow rate more rapidly than does avoidance alone. 3. Influence of the Geometry of the Environment In these pedestrian models, the geometry of the environment is simple. However, what happens when an obstacle is introduced into the environment? Helbing and Molnar (1995) investigated how their model behaved when they placed a doorway in the corridor. What they found was that,


19

Fig. 7. Simulation of pedestrians attempting to move across a crosswalk. Gray arrows indicate individuals intending to move left; black arrows indicate individuals attempting to move right. (A) Where individuals exhibit only repulsion from others, flow is less smooth than when (B) they exhibit repulsion from some but also attraction toward others who desire to move in a similar direction.

given a sufficient density of pedestrians, oscillations in alternating flows of passing direction at the doorway occur. These occur because the ‘‘pressure’’ of pedestrians at one side of the door eventually results in an individual being able to make it through the door. This makes it easier for individuals with the same desired direction to follow, resulting in a unidirectional flow of individuals through the doorway, as shown in Fig. 8. This reduces the pressure of pushing pedestrians at that side of the door, which will then result in a situation in which the flow is stopped, and then individuals moving in the other direction are able to pass through (because the pressure on their side is now greater), and so on. If the doorway is widened, changes in direction of flow become more rapid. It was also found that, given the same total width of doorway, two halfsized doors near the walls of the corridor increase the rate of flow of pedestrians relative to a single door. This is because, due to the mechanism of band formation described above, each door becomes used by pedestrians flowing in a common direction for relatively long periods of time. Individuals leaving their respective doorway in one direction clear the space ahead of the door for their successors.

20


Fig. 8. Simulation of pedestrains at a doorway exhibiting oscillations of flow. Here individuals moving to the left have temporarily monopolized the doorway. The decrease in ‘‘pressure’’ to the right of the door, caused by this exodus, will shortly allow those standing to the left of the doorway to block and then to temporarily monopolize the doorway, and so on. Image modified from that available from the simulation at http://www.helbing.org/ Pedestrains/Door.html.

4. Crowd Behavior and Emergency Situations Under certain extreme conditions, such as when people are evacuating from a crowded building, panic can result in pedestrians being injured or killed through crushing or trampling. In some cases crushing can occur in the absence of any external factor (e.g., fire), resulting instead from the impatience of queuing individuals who, having predominantly only local information, push forward. The physical interactions among members of a crowd can add up to cause dangerous pressures up to approximately 4 450 N m1, which can cause brick walls to collapse, bend steel barriers, and result in a large number of fatalities (Elliott and Smith, 1993). In an attempt to understand better such collective situations, Helbing et al. (2000a,b) extended their models of pedestrian behavior to include a ‘‘body force,’’ which counteracts the compression of bodies, and a ‘‘sliding friction force,’’ which impedes relative tangential motion within crowds. Furthermore, they assume that, within such crowd situations, people exhibit a greater degree of stochasticity (fluctuations) in their movement, and a higher desired velocity, because of the psychological effects of panic (Kelly et al., 1965). The model showed that increasing the value of either, or both, of these parameters caused an increase in evacuation time from a building by increasing the degree of interpersonal friction. This resulted in blockages, which occurred especially in the vicinity of bottlenecks. Thus, people fleeing from a building can decrease their chances of survival by attempting to move as fast as possible, or by performing uncoordinated movement through nervousness or panic. Under conditions in which individuals have restricted information about their local surroundings, such as in a smoke-filled room, Helbing


21

et al. (2000a) investigated the possibility that people may respond not only individualistically, but also in response to the motion of individuals near them, which they term a ‘‘herding effect.’’ Under such conditions neither pure individualism nor herding behavior performs well. If following just the individualistic rule, the discovery of an exit becomes a largely random process for each individual. Although herding can result in groups of individuals escaping if an exit is found, it is more likely that the crowd will move in the same, blocked direction. However, if people are assumed to use an intermediate strategy combining both individualism and herding, then the rate of escape is maximized, given the assumptions of the model. These models of human crowds are based on a simplified set of plausible interactions, and as such provide useful insights into the general behavior of such groups under a variety of conditions. There is, however, a need for further empirical studies, which are lacking despite the economic and/or social benefits of such research (e.g., in designing facilities so as to reduce risk during evacuation). We encourage initial studies to be made of crowds within relatively simple environments, such as on walkways, where an individual’s desired direction of travel can be better judged than, for example, in a crowded street, where motivations may change dynamically and be influenced by many more factors. Gathering data during genuine evacuation procedures will always be problematic (practically, and in some cases ethically), but data gathered from practice evacuations may be useful in testing, and further improving, current models. The importance of such safety issues has been further emphasized by the events of September 11, 2001, during which large, highly populated buildings (the World Trade Center in New York and the Pentagon in Washington, D.C.), and the streets around these buildings, had to be evacuated. D. Fish Schools and Bird Flocks In other animal aggregates, such as fish schools and bird flocks, group shape is often less constrained by environmental structure than in the human crowd examples discussed in Section II.C. In open space these groups can display clear cohesion and structural order, with the behavior of the individuals resulting in such ordered patterns of motion that they appear to move as a single coherent entity. When perturbed, as for example when a predator is detected, rapid waves of turning can propagate across the group (Radakov, 1973; Davis, 1980; Partridge, 1982; Potts, 1984). Many of these kinds of collective behaviors can be understood only by considering the large number of interactions among individual group members.

22


Early work on such collective phenomena hypothesized that birds can transmit information about turning almost instantaneously to other group members by ‘‘thought transference’’ (Selous, 1931), or by the generation by muscles or the brain of an electromagnetic field that could be detected by other group members (Presman, 1970). Heppner and Haffner (1974) argued that that a ‘‘leader’’ must coordinate the motion of such groups whereas Radakov (1973) concentrated on the possibility that fish schools may interact through the propagation of relatively local information among group members. Radakov made important steps in moving toward quantifying certain aspects of collective motion in these groups, including the propagation of ‘‘waves of excitation’’ that spread across his experimental schools when disturbed. Such waves of turning were shown to share certain properties with physical waves in that they attenuated, potentiated, reflected off the tank walls, and even seemed to cancel out if they met midschool. The essential advance here is that Radakov realized that collective behavior need not be explained as a phenomenon coordinated by a leader, or by global information, but by the rapid propagation of local information about the motion of near neighbors. 1. Models Some of the most conceptually simple models of the coordination of such animal groups have focused on explaining how a propensity to align with near neighbors can result in a longer range alignment within a population of mobile individuals (Vicsek et al., 1995; Cziro´k et al., 1997, 1999). In these models it is assumed that individuals move at a constant speed and assume the average direction of motion (this direction being subject to error) of those within a local neighborhood. Such models are useful because their minimalism allows them to be analyzed by techniques developed for nonequilibrium statistical physics (for a review see Cziro´k and Vicsek, 2001). This comes at the cost of biological realism, however. For example, the mobile particles in these simulations neither avoid collisions nor exhibit attraction toward others. Consequently they cannot form a self-bounded group (such as the bird flocks and fish schools described previously) when individuals exhibit any error in decision making, and thus cannot fully explain the clearly defined animal groups seen in many species. Here we focus on more biologically realistic (yet still much simplified) models of animal motion, based on generic abstractions of the aggregation tendencies evident in fish schools and bird flocks (Partridge, 1980, 1982; Partridge and Pitcher, 1980; Heppner, 1997). Several authors have developed models in which grouping results from individuals exhibiting local repulsion, alignment, and attraction tendencies based on the positions


23

and orientations of individuals relative to one another (Aoki, 1982; Reynolds, 1987; Huth and Wissel, 1992; Couzin et al., 2002). Repulsion simulates individuals avoiding collisions if they come close to one another. Alignment reduces collisions among mobile individuals within a group and facilitates collective directional motion of large groups. Attraction allows groups to retain cohesion, and simulates an individual tendency to join groups and to avoid becoming isolated (Hamilton, 1971). In these models the individual behavioral rules result in group formation and cohesion, rather than fixing individual density within a periodic domain1 (as in the simplest models described previously). The Reynolds model (1987) simulated the motion of computer-animated flocking ‘‘boids’’ within three-dimensional space, and demonstrated how local interactions among individuals can lead to realistic-looking collective behaviors such as polarization within groups, and cohesion of groups, even when moving around environmental obstacles. Incorrectly, this model is sometimes thought to have included global information, perhaps due to the use by Reynolds of the term ‘‘flock centering.’’ However, it is clear from the original model description that information is restricted to local regions around each boid. Although capable of simulating motion similar to that of real birds, this model is somewhat complicated, including properties such as banking during turns to make it ‘‘look better’’ (this was intended as an animation tool for computer games and films). This makes it difficult to interpret and analyze from a more rigorous scientific perspective. Somewhat simpler models have been developed in twodimensional space by Aoki (1982) and Huth and Wissel (1992), and in three-dimensional space by Couzin et al. (2002). Aoki (1982) demonstrated that simple stimulus–response behaviors, similar to those used by Reynolds (1987), could account for the coordinated movement of groups of fish. Extending this model, Huth and Wissel (1992) investigated in more detail the potential interaction processes involved in coordinating such collective motion. They explored the possibility that individuals use a ‘‘decisionmaking process’’ from which they determine a single near neighbor with which they then interact, or an alternative ‘‘averaging process’’ whereby individuals average the influence of a different number of neighbors. Averaging models in which individuals combined the influence of several nearest neighbors were found to account better for the behavior of real fish, because they produced groups that were better aligned and less likely

1

A periodic domain is one with no boundaries. Individuals leaving the domain at one side reappear at the appropriate position at the opposite side. This is a standard technique in computer modeling to minimize the influence of ‘‘edge effects.’’

24


to fragment. Decision models, by contrast, could not account for the type of highly coordinated motion seen in real groups. 2. Individual Behavior and Group Shape Couzin et al. (2002) developed a model of animal aggregations in threedimensional space (e.g., flocking and schooling) similar to those previously described. They demonstrated how relatively minor changes in individual behavior can result in dramatic changes in group shape. They also investigated some properties of the transitions between group shapes that may highlight some fundamental properties of animal groups. In following the approach of Aoki (1982) and of Huth and Wissell (1992) they assumed, for tractability, that individuals respond to each other within specified behavioral ‘‘zones’’ (see Fig. 9). The highest priority for individuals was assumed to be maintenance of a minimum distance between themselves at all times to avoid collisions (Krause and Ruxton, 2002). They achieved this by moving away from other individuals within a close-range spherical ‘‘zone of repulsion,’’ with radius rd. If not performing an avoidance maneuver, individuals were assumed to align with others within a ‘‘zone of orientation,’’ ro, and to be attracted to other individuals within a ‘‘zone of attraction,’’ ra. These latter two zones were spherical, except for a volume behind the individual in which neighbors were undetectable. All behavioral zones in this model were nonoverlapping. An individual would

Fig. 9. Representation of an individual in the model of grouping in three-dimensional space, centered at the origin and pointing in the direction of travel. zor, Zone of repulsion; zoo, zone of orientation; zoa, zone of attraction. The possible ‘‘blind volume’’ behind an individual is also shown. , Field of perception. (From Couzin et al., 2002.)


25

perform a correlated random walk if it did not detect any neighbors. In accordance with these rules, individual trajectories were integrated over time at discrete intervals (timesteps), set as 0.1 s apart, representing the response latency of fish (Partridge and Pitcher, 1980). Couzin et al. (2002) investigated the consequences to group shape of changing the size of the zone of orientation (its width defined as ro = ro rd), and the zone of attraction (width defined as ra = ra ro), given random starting conditions. They calculated two properties from the simulation (after it had reached a dynamic equilibrium) that could allow group shape to be quantified: 1. Group polarization (pgroup), ranging from 0 to 1. This increases as the degree of alignment of group members increases. 2. Group angular momentum (mgroup), ranging from 0 to 1. This measures the degree of rotation of a group about the group center, increasing as degree of rotation increases. It was found that, if individuals exhibited attraction to others but little, or no, orientation tendency, they formed a ‘‘swarm’’ group type (Fig. 10A), characterized as having low pgroup and low mgroup values (even though individuals do rotate around the group center they do so in different orientations, thus resulting in low group angular momentum). As the size of the zone of orientation increased, however, the group was found not to adopt the swarm formation, but instead would form a ‘‘torus’’ with low pgroup and high mgroup values, in which the individuals perpetually rotated around an empty core (even though individuals are locally polarized, overall group polarization is low) (Fig. 10B). The direction of rotation was random. If the zone of orientation was increased further, however, the group initially adopted a ‘‘dynamic parallel’’ conformation (higher pgroup, low mgroup) (Fig. 10C), and then a ‘‘highly parallel’’ arrangement (highest pgroup, low mgroup) (Fig. 10D). This model predicts that these are the four fundamental types of collective state that individuals within such groups can adopt, and between these states the collective behavioral transitions are sharp (Fig. 10E and F). It also demonstrates that large changes in group properties and organization can result from relatively minor changes in local individual response, and that animal groups are likely to change rapidly between these states because intermediate group types are unstable. Biologically the ability of groups to change between structural types could be important in allowing individuals to maximize fitness as conditions change. This may occur, for example, as a response to hunger, or to external stimuli such as the presence of predators. It is known that fish and birds tend to become more

26


Fig. 10. The collective behaviors exhibited by the model developed by Couzin et al.: (A) swarm; (B) torus; (C) dynamic parallel group; and (D) highly parallel group. Also shown are group polarization ( pgroup; E) and angular momentum (mgroup; F) as a function of changes in the size of the zone of orientation ro and zone of attraction ra. The areas denoted as a–d correspond to the area of parameter space in which the collective behaviors (A–D), respectively, are found. (From Couzin et al., 2002.)


27

polarized within groups (individuals become better aligned) when predators are detected (Wilson, 1975; Partridge, 1982). This is important because it not only allows the individuals within the group to avoid colliding with one another, but also facilitates the transfer of indirect information about the presence of a predator. For example, if only one, or a subset of, individuals turns in response to such a stimulus, the alignment tendency allows this change in direction to be transmitted over a range much larger than the individual interaction radius. This property of groups is discussed in more detail in Section II.D.4. If alignment range in the model is reduced the individuals adopt a torus conformation. This group shape may initially appear uncharacteristic of real animal groups, but is in fact adopted by many species of fish including barracuda, jack, and tuna (see Parrish and Edelstein-Keshet, 1999, for a photograph of jack performing this behavior). This behavior results in a quasi-stationary group, yet allows the continual motion of individuals that is required by certain fish species for respiration, while permitting individuals to benefit from local polarization. Furthermore, it may allow individuals to save energy because each is in the slipstream of another. If individuals exhibit attraction, but little or no alignment, they form a swarm. This behavior is often seen in aggregates of insects, such as midges (Okubo and Chiang, 1974) and mosquitoes (Ikawa and Okabe, 1997), but can also be exhibited by fish schools (Pitcher and Parrish, 1993). Although cohesive, this group type does not benefit from the advantages of polarization discussed previously. 3. Behavioral Transitions, Collective Memory, and Hysteresis After initially exploring the types of group shape that form from random starting conditions, as described previously, Couzin et al. (2002) investigated the consequences of different starting conditions to the collective behavior within their model. In nature, groups are likely to move between collective states as conditions change, and as a consequence of this the previous history of individual orientations and positions may have an influence on the collective behavior as behavioral parameters change. To investigate this possibility, the same simulation was used but the starting conditions were nonrandom. Keeping the size of the zone of attraction constant, the influence of individuals modifying the size of their zone of orientation was investigated. Starting with no alignment tendency (ro = 0) the model was run to dynamic equilibrium (resulting in a swarm). Then, without resetting the model to the random starting conditions, the size of the zone of orientation (ro) was increased slightly, the model was allowed to run to dynamic equilibrium again, and the process was repeated until the group entered the dynamic polarized state. Then, the size of the zone

28


of orientation was decreased sequentially in the same way, until eventually the model was returned to the original parameter settings (ro = 0). Intuitively it would be assumed that this would simply mean the collective state moves from being a swarm, to a torus, to a parallel group type (as ro is incrementally increased) and then back to a torus and finally a swarm (as ro is decreased). As shown in Fig. 11, when the zone of orientation was increased the model behaved as assumed, but if moving through the same parameter space in the opposite direction (as ro is decreased) the collective behavior was different. The group did not adopt the torus conformation, and instead eventually returned only to the swarm configuration. This demonstrates an important principle: that two completely different

Fig. 11. The change in group polarization (pgroup; A) and angular momentum (mgroup; B) as individuals within a group increase (solid line) or decrease (dotted line) the size of the zone of orientation, ro. The group patterns that form depend on the previous history of the group (hysteresis). (From Couzin et al., 2002.)


29

collective behavioral states can exist for identical individual behavioral rules, and that the transition between behavioral states depends on the previous history (structure) of the group, even though the individuals have no explicit knowledge of what that history is. Thus the system exhibits a form of ‘‘collective memory.’’ Intuitively it might be assumed that group-living animals only need evolve a direct interaction between individual internal state (resulting from internal and external stimuli) and behavior (here the rules of interaction are employed). Our results suggest, however, that the situation is not so simple, and that the evolution of collective (extended) phenotypes may be more complex. Importantly, this kind of behavior is likely not to be specific to this model, or even this class of model, but rather may be a generic property of transitions between collective behaviors. 4. Group Shape and Motion in the Presence of External Stimuli Although the fundamental organizing principles defining the shape of aggregates such as fish schools and bird flocks do not rely on external stimuli, such stimuli may also be important in explaining shape under certain circumstances. For example, as suggested in Section II.D.2, local interactions allow information (here encoded as the positions and orientations of neighbors) to be propagated across the group. Thus individuals within such groups can perform avoidance maneuvers without direct detection of an incoming signal. Simulating predator attack allows the response of groups to transient disturbance to be investigated. For example, Fig. 12 shows a time series from an animation of a simulation of grouping developed by Couzin et al. (2002), in which a predator is included (shown in gray). Here the predator follows a simple rule: it moves toward the highest perceived density of individuals (Milinski, 1977). A supplementary rule is included for the behavior of prey individuals in the model described in Section II.D.2, which allows them to detect and move away from a predator. The model exhibits the characteristic collective patterns that have been described in natural groups under attack (Partridge, 1982), including ‘‘flash expansion,’’ in which individuals rapidly move away from the predator as it strikes (Fig. 12a); ‘‘vacuolation,’’ in which the expansion results in a cavity forming around the attacker (Fig. 12c and d); and the ‘‘split effect,’’ in which a group may be fragmented (Fig. 12h). The size of the volume in which individuals respond to others is also important in coordinating collective avoidance behaviors (Fig. 13). If this volume is small an individual will behave more or less independently of those around it. This increases the tendency of individuals to become nonaligned, and for groups to become fragmented (Fig. 13a). As the size of this zone increases, an individual will respond to a greater number of

30


Fig. 12. Computer simulation of 1000 grouping individuals (white) responding to attack by a predator (gray). (a–h) Successive snapshots of the simulation as the predator attacks.

neighbors. This increases the quality of information to which an individual has access, and decreases the variance (through averaging over a greater number of influences). The group becomes capable of transferring


31

Fig. 13. Theoretical influence of the size of the zone in which individuals respond to others. The predator is shown in dark gray, individuals that directly perceive the predator are shown in light gray, and all others are shown in black. (a) One individual detects the predator, but the small size of the behavioral zones does not allow other individuals to respond to its change in orientation. (b) Only two individuals detect the predator directly, but the behavioral zones are sufficiently large to allow group cohesion and the spread of relevant information about the location of the predator (the change in direction of light gray individuals) to other individuals nearby. (c) If behavioral zones are too large, individuals are swamped with information from both near and distant sources. This reduces the ability of individuals to respond to local perturbation.

information (Fig. 13b). The response of individuals not only to nearest neighbors, but also to neighbors further away (but still in a relatively local volume), also increases the speed of information transfer. This can explain the high speed of maneuver waves in birds (Potts, 1984) and fish (Radakov, 1973). If this zone continues to grow, however, the quality of information an individual acquires from the movement of others may decrease. The orientation and position of individuals further away are less likely to encode relevant information (Fig. 13c). Large behavioral zones

32


increase the homogeneity of response within the group. If all individuals respond equally to all others within a group, for example, they can collectively select a direction that may be detrimental to all, or almost all, of them. Group members would therefore be expected to respond only to those individuals that are most likely to have information that would benefit them. Proponents of self-organization theory often stress that animals do not need long-range information to coordinate group behavior (Bonabeau et al., 1997). However, localizing information input may provide significant adaptive benefits to an individual within a group, allowing sensitive response not only to predators but also to environmental obstacles. Similarly, Inada and Kawachi (2002) investigated how directly changing the number of neighbors that an individual responds to affects the information transfer within such groups. Their model was also able to emulate the escape responses of fish within real schools, and showed that in groups of 50 individuals, responding to a relatively small number of neighbors (3) was the best strategy for escape. However, their model requires individuals to be able to count the number of neighbors. Currently it is unclear whether fish perform such counts, or whether they perform behaviors such as changing the range over which they respond to others (which would indirectly change the number of neighbors with which they interact). In addition to allowing collective avoidance behaviors, the rapid changes in turning and group shape in such animal groups may also act to confuse the sensory system of predators, thus making it difficult to isolate and catch any given individual (Landeau and Terborgh, 1986). Information transfer among individuals is also likely to influence their response to other stimuli, such as the positions of resources, or favorable regions within a heterogeneous environment. In aquatic habitats, for example, resources such as phytoplankton, the temperature or salinity of the water, and concentrations of dissolved gases are all known to vary in a nonuniform way, and over both small and large length scales. Individuals are therefore expected to modify their positions with respect to these properties so as to maximize resource intake and minimize physiological stress. However, this is a nontrivial task: unpredictability and local fluctuations make finding and moving up or down such environmental gradients (taxis) difficult when an individual has only local knowledge on which to base its motion. In many cases, such as phytoplankton or gaseous concentration, the gradients occur over such large spatial scales (on the order of kilometers) that local detectable gradients are inherently extremely shallow. In the case of thermal gradients, for example, even the steepest horizontal gradients in the open sea are 0.01– 0.1 C/100 m (Dizon et al., 1974).


33

Gru¨nbaum (1998) used computer simulation to investigate the theoretical consequences of grouping to such taxis behavior. He assumed individuals use a simple form of taxis, known as klinotaxis, whereby a moving individual modifies its probability of making a turn as a function of whether conditions are perceived to improve or deteriorate over a given time interval. Such behavior is known to facilitate taxis in even simple organisms, such as bacteria (Keller and Segel, 1971; Alt, 1980; Tranquillo, 1990). Although they are not directly detecting the gradient, individuals performing such taxes will, on average, spend more time moving in favorable directions than in unfavorable ones. By simulating groups of individuals performing this behavior under conditions in which they do not interact with one another (asocial taxis) and do interact by balancing the tendency for taxis with a simple schooling behavior (social taxis), Gru¨nbaum (1998) demonstrated that such social interactions improve the motion of individuals up a gradient. The alignment of individuals, and thus transfer of information, when schooling, allows averaging of individual errors in gradient detection, and therefore results in reduced deviations in motion from the desired direction of travel. This information sharing within schools of fish has been likened to a ‘‘sensory array’’ (Kils, 1986), which allows information to be gathered over a wider spatial range than would be possible for a solitary or noninteracting individual, and dampens the influence of small-scale fluctuations in the environment. The model also predicts that the benefits of such information sharing are dependent on group size. As group size is increased the efficiency of taxis shows an asymptotic increase: initially it increases steeply, but then the rate of increase reduces over time, leading to a plateau where further increases in group size have little effect on taxis accuracy. Owing to the deliberately abstract nature of the model (to characterize a generic property), the absolute group sizes are less important than the general prediction of the type of relationship that should be expected in natural groups. A further property of individual behavior that Gru¨nbaum (1998) explored was the balance of the taxis and interaction ‘‘social forces’’ within the simulation, which demonstrated a trade-off between these two tendencies. Individuals that interacted only weakly with others (the taxis response is weighted strongly) would benefit little from averaging of information. At the opposite extreme, where the interactions of individuals with one another are strong relative to taxis, the group will take a long time to adjust to changes in gradients. Thus, in reality, individuals may be expected to evolve an intermediate strategy. Niwa (1998a) developed a conceptually similar model to investigate how large fish schools could use klinotaxis to move up heterogeneous temperature gradients when migrating. Migrating pelagic fish such as

34


sardines, anchovies, and mackerel form cohesive groups that can extend over kilometers, and contain in excess of 106 fish (Pitcher and Parrish, 1993). The memory necessary for thermal klinotaxis (comparing previous and current temperature) may be obtained from internal core temperatures that provide fish with information about previous thermal history (from internal sensors) and their current temperature detected by sensors in the skin (Neill et al., 1976). In the Niwa model individuals were considered to have a desired internal temperature and to behave like individuals in the Gru¨nbaum model (1998) described previously. Even for such large groups, simple and local response behaviors are able to account for the collective migration behavior. Group shape has also been found to be influenced by parasites. In wild schools of banded killifish (Fundulus diaphanus), for example, group geometry is dependent on the overall prevalence of a trematode parasite (Crassiphiala bulboglossa) among group members: groups with high parasite prevalence tended to exhibit a broad phalanx-like shape, whereas those with low parasite prevalence tended to be elliptical, with the major axis aligned with the direction of travel (Ward et al., 2002). Interestingly, a similar change in group shape has been exhibited in a general model of grouping behavior (of zebra herds, Equus burchellii, but also applicable to other group types) developed by Gueron et al. (1996). The difference in this model resulted not from changes in the interactions among individuals, but simply from a difference in individual speed. Groups in which individuals moved more slowly tended to proceed as a phalanx, whereas groups in which individuals tended to move more rapidly formed a more columnar structure, elongated in the direction of travel (see Section III.B for further discussion of this model). If parasite load affects swimming speed, such a difference may be able to account for the difference in group shape in killifish. Ward et al. (2002) suggest that the trematode cysts may reduce swimming performance by affecting the dorsal musculature of infected individuals, and/or by reducing the hydrodynamic streamlining of individuals. It should be noted, however, that other changes to behavior in schooling models can also change group shape. For example, decreasing the angle of perception (increasing the frontal bias) will also result in a group more elongated in the direction of travel (I. D. Couzin, unpublished data). 5. Parabolic Groups of Predatory Fish Group shape may also be important to predatory fish. Partridge et al. (1983) analyzed the structure of Atlantic bluefin tuna (Thunnus thynnus) schools in the wild from aerial photographs. Such groups are well suited to this kind of analysis because they swim just under the water surface, so the third spatial dimension is not required for the analysis of their positioning


35

behavior. Schools varied in size from 2 to 79 individuals, and group members tended to occupy defined positions relative to one another. In small schools ( 10 members) nearest neighbors tended to be alongside (90 ), and consequently groups tended to be more or less straight lines (perpendicular to the direction of travel). For larger groups, however, nearest neighbors tended to occupy positions of 45 and 135 . Perhaps the most interesting type of group shape was a ‘‘parabola’’ with the deflection point in the center of the group with respect to the direction of travel. Partridge et al. (1983) suggest that this group shape allows the school to act like a seine net, funneling or encircling the prey fish. They also hypothesize that individuals at the edge of the parabola would be less likely to catch prey (due to increasing overlap of strike zones from the group center), and thus it is possible that some alternative benefit may be associated with these positions, or that individuals change position within the group between hunting events.

III. Group Internal Structure A. Analyzing Spatial Positions in Natural Groups Despite the ubiquity of animal aggregations, there is limited quantitative information about the internal structure of most vertebrate groups. Groups moving in three-dimensional space present a particular challenge to study because there are significant technical complications involved in recording accurately the spatial positions and orientations of group members. Consequently, attempts to characterize such structure are often limited to qualitative observations (Radakov, 1973), although through the use of inventive camera-based techniques it has been possible to make accurate recordings of spatial positions in fish schools (Cullen et al., 1965; Partridge, 1980; Partridge and Pitcher, 1980; Partridge et al., 1980) and bird flocks (Major and Dill, 1978; Davis, 1980; Pomeroy and Heppner, 1992; Heppner, 1997). Partridge et al. (1980) used photographic techniques to record the positioning of individuals within fish schools in three-dimensional space within a large circular channel (1.8 m wide and 31 m in circumference) in the laboratory. A moving gantry projecting from the center allowed fish schools to be filmed from above. Fish were trained to swim over a ‘‘speckled spot of light’’ projected onto the floor of the tank, and thus by rotating the gantry at a constant speed they could film the school as it swam to keep pace with the spot. To reconstruct the three-dimensional positions and orientations of the fish they used a shadow method (Cullen et al., 1965), in which a secondary light shone at a known angle onto the

36


school casts shadows of the fish. The area filmed included that where the shadows were cast, and a calculation involving the known position of the light and the depth of the water allowed the height of the fish in the water column to be estimated, thus providing the third spatial dimension. There are, however, limitations to this method. First, it is time consuming, and somewhat subjective, to relate a shadow to an individual fish within the video sequence. Second, when fish become closely packed, it is not possible to record the positions of all fish because individual fish occlude one another. A third limitation to this particular study is that an extraneous stimulus controlled the position and speed of the group. Such a stimulus is likely to have an influence on the grouping behavior of fish by constraining their natural movement tendencies and by forcing fish to balance two social forces: their motion with respect to one another and their motion with respect to the stimulus. Nevertheless, this technique is still vastly superior to qualitative observations, and Partridge et al. (1980) were able to investigate positioning behavior in groups of up to 30 individuals for three species of fish: Atlantic cod (Gadus morhua), saithe (Pollachius virens), and Atlantic herring (Clupea harengus). As well as being commercially important, these fish possess different degrees of schooling tendency. Cod are weakly facultative schoolers, whereas saithe, although facultative schoolers, spend the majority of their time in polarized groups (individuals within the group are aligned). Herring are obligate schoolers, and form highly polarized groups. To examine internal group structure, both nearest neighbor distances and the elevation and bearing of group members to their nearest neighbors were recorded. Elevation and bearing correspond to the angle between the current orientation of the reference individual and the position of the nearest neighbor, in the vertical and horizontal plane, respectively. The angles 0 and 180 refer to directly ahead of, and behind, the reference individual, respectively. Individuals were shown to exhibit a minimal approach distance (analogous to the ‘‘zone of repulsion’’ described in Section II.D.2). To ascertain whether positioning was nonrandom, the elevation and bearing distributions were compared with those generated by a random (null) model in which individuals were assigned positions at random within a volume equal to that of the real school. Fish within the real schools were found to occupy nonrandom positions in all experiments. Among cod and saithe, the distribution of bearings had a peak at 90 , showing that individuals tended to be closest to lateral individuals. The distribution was more peaked for the relatively more polarized saithe groups. In herring schools, however, the distribution of bearings was found to be bimodal, with nearest neighbors found most frequently at angles of 45 and 135 ,


37

showing that these fish adopt a more lattice-like structure. Dill et al. (1997), however, questioned the use of the null model in these analyses, arguing that the results may actually be an artifact of using such a simplistic model with which to compare the data. They demonstrate how more complex, and perhaps more biologically meaningful, null models can be constructed. There are relatively poor data available to quantify the internal structure of bird flocks. By using stereo photography it was possible for Major and Dill (1978) to record the positions of birds within flocks of European starlings, Sturnus vulgaris, and dunlin, Calidris alpina. They concluded that there were ‘‘striking similarities’’ between the internal organization of bird flocks and fish schools, although the large variance in the data from bird flocks and the limited number of species investigated makes rigorous comparison difficult. Furthermore, they were unable to record the flight paths of individual birds over time. Davis (1980) observed the coordinated turns in flocks of dunlin, but had a small sample size (nine ‘‘analyzable incidents’’), and recorded only the number of birds with light plumage visible over short time periods (700 ms). The main problem with this type of analysis, as Heppner (1997) points out, is that as a flock moves relative to an observer, a ‘‘wave’’ of brightness (through revealing light plumage, or the reflection of light from the body) may appear to cross the flock, indicative of a turn. However, such an effect is likely to be an artifact of the change in position of individuals relative to the stationary observer and the light source. In both bird flocks and fish schools it appears that the internal structure of groups is usually dynamic, with individuals frequently shifting position. For example, Pomeroy and Heppner (1992) filmed a flock of 11 pigeons in flight and found that during a turn birds in the front of the flock can readily fall to the back, or those on one side change to the other. This is a consequence of the birds seeming to employ a relatively constant turning rate during a turn, resulting in positions being rotated. Sinclair (1977) (in conjunction with J. M. Cullen) used aerial photographs to analyze the spatial positions of individuals within grazing African buffalo (Syncerus caffer) herds. They used a manually operated plotting machine (in a nuclear physics laboratory, designed to plot the tracks of particles in bubble chambers) to record the positions and orientations of adults and calves. From these data the distances and bearings to nearest neighbors were calculated. To search for nonrandom patterns they compared the nearest neighbor data with those generated by a model in which individuals were randomly assigned positions within the same area. In all five herds analyzed, the distance of individuals from their nearest neighbor was significantly higher than expected. This suggests that the

38


animals are overdispersed when grazing. When the angles to nearest neighbors were analyzed there appeared to be no consistent pattern. However, in sheep it appears that grazing herds display more evident structure, with nearest neighbors tending to be at a bearing of about 55 ahead (Crofton, 1958). These data suggest that in sheep herds individuals become progressively more crowded toward the front of the group, because, if this were not the case, nearest neighbors should be expected to be as often behind as in front. Until further studies are made, however, it is difficult to interpret what these results mean with regard to the interactions among individuals within herds. We encourage any researchers who have any relevant data on the distribution and/or orientation of individuals within herds to make this information publicly available so that more rigorous analysis of the grouping behavior, and comparisons between species, can be made. Throughout this chapter we have emphasized the need for empirical studies to test existing models and develop new theoretical approaches. One of the principal limitations to the study of collective behavior is the difficulty of recording and analyzing the movement of many organisms concurrently. However, only by obtaining accurate recordings of the movement of individuals from which behavioral properties such as the interactions among individuals, and between individuals and their environment, can be made can we begin to understand the processes that underlie collective behaviors. As described previously, the manual recording of the positions of individuals over time is extremely laborious. Through recent technological advances, however, a new possibility has been introduced: that a computer can be programmed to ‘‘see’’ and record the movement of animals automatically. In this way it is possible to track a large number (tens or hundreds) of organisms simultaneously in two dimensions (e.g., fish within shallow water) by analyzing film made from above (Couzin, 1999; Roditakis et al., 2000) or fewer individuals within three-dimensional space (where occlusion of individuals in the center of large groups is inevitable) from film made by two or more cameras (e.g., one camera filming from directly above a group, and another from the side) (see Osborn, 1997; Parrish and Turchin, 1997). Once trajectories have been obtained it is possible to perform time-series analyses of the velocities of each individual with respect to other group members, investigating crosscorrelations between the velocities of individuals, as well as autocorrelation of the velocity of the focal individual (see Okubo, 1980; Partridge, 1980; Parrish and Turchin, 1997). Parrish and Turchin (1997), for example, examined a range of potential ‘‘foci’’ that may influence the behavior of individual fish (juvenile blacksmith, Chromis punctipinnis). Such foci were assumed to be either attractive, repulsive, or neutral depending on the


39

distance separating them from the fish and ranged from the individual’s nearest neighbor to the centroid (center) of the entire group. They found that individuals appear to pay most attention to their nearest neighbor and to the school as an entire unit, although it is currently difficult to determine from these results what behavioral rules are being used by the fish. It would be interesting to develop analysis techniques for such groups further. One attractive avenue of research may be to use computer models of grouping to make predictions about where individuals would be expected to move, from one instance (in time) from a video sequence to a future instance. Thus a search could be made for theoretical behavioral rules that have the highest predictive power (over a range of time intervals) when compared with a real data set. It would also be possible, using the type of computer vision systems mentioned previously, to recreate the visual information available to each individual at an instance in time. This may provide further insight into the actual information available to individuals within groups (e.g., for fish that predominantly use vision, such as the stickleback, Gasterosteus aculeatus; or ungulate herds filmed from above). Because the imaging software can calculate the size, orientations, and positions of individuals, it would be possible to create a program to calculate where each individual’s eyes are, and generate an impression of what visual information is available when making movement decisions. This is important because there may often be limits (which may vary with environmental conditions and the degree of local crowding) to the distances at which individuals are able to detect, and respond to, neighbors. The influence of external stimuli (such as the perception of obstacles) could also be investigated in this way. Thus, we believe further developments of imaging and behavioral analysis systems could provide new, and important, insights into the mechanisms of grouping behavior. B. Differences among Group Members and the Internal Structure of Groups Radakov (1973) considered fish within schools as being behaviorally identical and interchangeable with regard to position. In the simulations described, it is also assumed that individuals are identical. This is necessary to demonstrate how patterns form with the simplest possible assumptions and information input. However, the positions that individuals take within groups, relative to others, have important evolutionary and ecological consequences (Hamilton, 1971; Okubo, 1980; Krause and Ruxton, 2002) (see Krause, 1994, for a review). In many cases group members are not intrinsically equivalent (Pitcher et al., 1985; Parrish, 1989a; DeBlois and Rose, 1996; Krause et al., 1996).

40


Differences among individuals, such as age, nutritional status, and sex, may all influence the position adopted by an individual within a group. For example, Krause (1993a), in a study of schooling fish (roach, Rutilus rutilus), showed that starved individuals would tend to occupy positions toward the front of the group. It is likely that this increases food capture rate by these individuals because they are more likely to be able to detect and consume floating food items than are individuals toward the rear of the group. However, being at the front also increases the chance that these individuals will be the first to encounter predators (Bumann et al., 1997). Furthermore, there is evidence for certain mobile fish schools that individuals occupying positions whereby they are in the slipstream of others may need less energy for locomotion (Herskin and Steffensen, 1998). Consequently, there may be both benefits and costs to spatial positions. Individuals within groups may therefore be expected to modify their positions relative to neighbors as a function of their internal state; hungry individuals, for example, being more willing to risk dangerous positions if that will benefit their resource intake. When the advantage of being in a frontal position is outweighed by the perceived risk of predation, however, it may be expected that individuals will avoid the group front, perhaps occupying positions closer to the group center (Hamilton, 1971). Krause (1993b) found that minnows, Phoxinus phoxinus, respond to perceived danger by moving to positions where they tend to be surrounded by near neighbors on all sides. However, the center need not necessarily be the safest position within a group. Individuals in the center may not be able to detect a threat directly and may also be constrained in their escape movement by the proximity of others. Parrish (1989a), for example, found that Atlantic silversides, Menidia menidia, suffered higher predation from black sea bass, Centropristis striata, if they occupied central positions within the school. McFarland and Okubo (1997) suggested that central positions in fish schools may also be detrimental for another reason. Individual fish consume dissolved oxygen and increase local ammonium concentrations. In the center of large groups the modification of dissolved gases may be such that respiration is inhibited, which may be a group-structuring factor in large (particularly stationary) groups. To investigate the influence of individual behavioral heterogeneity on grouping dynamics, Gueron et al. (1996) developed a simple model of herding animals. Their model is conceptually similar to those described previously to investigate fish schools and bird flocks, in that individuals are assumed to respond to others within local zones. In their model, individuals attempt to maintain a minimal separation distance. This behavior has the highest priority. Outside this zone is a ‘‘neutral zone’’


41

extending to the sides, and ahead, of the focal individual. In this zone, individuals do not respond to neighbors unless all neighbors are on the same side, in which case the focal individual will move toward neighbors, but not change speed (representing avoidance of isolation). Lower in priority is the ‘‘attraction zone,’’ which extends beyond the neutral zone, again to the sides and front. The rule employed if individuals are found within this zone is to bias both direction and speed to maintain proximity to neighbors. If no individuals are detected in any of these zones, then individuals respond to neighbors (if any) that are present to their rear. An individual that has neighbors only to the rear is termed a ‘‘leader.’’ Note that this definition emerges from the relative positions of individuals within the group, as opposed to being explicitly specified. A leader is assumed to reduce its speed to remain in proximity to other group members. Individuals that do not detect any neighbors within their behavioral zones are termed ‘‘trailers.’’ Such individuals speed up to represent their attempting to maintain contact with the group. When all individuals are assumed to be identical, it was found that for a wide range of walking speeds, large groups (up to 100 individuals) could maintain cohesion for long time periods. At low walking speeds the group adopted a phalanx-like structure, with individuals forming a flat moving front. As individual speed was increased, however, the group structure became more columnar. As a next step, groups were considered to be composed of two subpopulations, each with a different speed. Not surprisingly, individuals in the faster subpopulation tended to occupy positions at the front of the groups, becoming leaders irrespective of their positions within the group when the simulation was started. Without individuals responding to others behind them (in the ‘‘rear zone’’) these subpopulations will inevitably separate, given time. If individuals did respond to those behind them, it was possible to retain cohesion, but as the difference in speed between the two subgroups increased this became less likely, and subgroup fission occurred more rapidly. However, given that the rear zone was set as the ‘‘lowest priority’’ in this model, it is possible that fragmentation of subpopulations would occur even when the difference in speed was relatively low, because it is possible that individuals would not respond to those to the rear for sufficiently long periods of time. Gueron et al. (1996) suggest that differences in speed between lactating and nonlactating zebra (lactating individuals moving more slowly) may explain the segregation of these individuals into subpopulations. This segregation occurs particularly under circumstances in which individuals tend to move quickly, such as when the perceived threat of predation is high (e.g., when moving near waterholes).

42


Couzin et al. (2002) also investigated the consequences of behavioral heterogeneity on the spatial positions individuals occupy within mobile animal groups, by modifying the model of grouping in three-dimensional space outlined in Section II.D.2. Unlike Gueron et al. (1996), they assumed a continuous distribution of individual phenotype within the population, as opposed to just two classes of individual. To simulate variation they modified the behavioral parameter under investigation by adding a Gaussian-distributed random deviate centered on 0 (independently drawn for each individual). Therefore the standard deviation of this distribution determined the degree to which individuals within the group differed with respect to that parameter. They investigated the consequence of variation in individual properties, including speed, turning rate, error, and the size of each of the three behavioral zones. To quantify the spatial positioning behavior of individuals, the correlation between these parameters and the distance between individuals and the group center, and the distance to the front of the group, were measured (Fig. 14A–D). The speed of individuals was positively correlated with their being at the front of the group, and slightly further away from (negatively correlated with) the group center (Fig. 14A). Those with a higher rate of turning tended to be at the rear, and slightly closer to the center, of the group (Fig. 14B) and individuals with higher degrees of error in movement tended to occupy the rear of groups. The size of the immediate personal space around individuals, represented as the zone of repulsion (rr), was important in structuring groups: individuals with low values of rr tended to occupy positions at the center, and toward the front, of the group (Fig. 14C and E). For all parameters investigated the strength of the correlation (degree of sorting) increased as the variation within the population increased. These results suggest that behavioral and/or motivational differences among individuals may constitute an important organizational principle within animal groups. As explained previously, there may be many reasons why individuals within groups may be expected to modify their positions relative to others. This model provides potential self-organizing mechanisms whereby this may occur. Importantly, the sorting within the model depends on ‘‘local rules of thumb,’’ that is, not on absolute parameters but rather on relative difference between individuals. Thus an individual decreasing its zone of repulsion relative to near neighbors will tend to move toward the center of the group, even if it has no knowledge of where the center actually is. This is important, because in many naturally occurring large collectives of vertebrates it is extremely unlikely that individuals have the cognitive or sensory capabilities to calculate their absolute position within the group (individuals are often closely packed,


43

Fig. 14. Sorting as a function of variation in (A) speed s, (B) turning rate , (C) zone of repulsion rr, and (D) zone of orientation ro. A typical group sorted by rr is shown in (E). Sorting is measured as the Spearman rank correlation coefficient (rho) of individuals calculated from the front (solid line) or center (dotted line) of the group. (From Couzin et al., 2002.)

44


restricting perception range). For example, pelagic fish schools can extend over kilometers and may consist of hundreds of thousands of individuals (DeBlois and Rose, 1996), so it would be impossible for individuals to measure their distance from the edge. Thus natural selection is likely to favor local self-organized mechanisms that individuals can use to modify their position relative to others, without necessitating a specific destination or knowledge of current location. Another point to note about this type of sorting mechanism is that, because of the interaction mechanics involved, individuals with similar phenotypes will become more closely associated within groups. In many fish schools individuals tend to be close to others similar in size (Pitcher et al., 1985; Parrish, 1989a; DeBlois and Rose, 1996; Ward et al., 2002) or, in multi species groups, to conspecifics (Parrish, 1989a). The Couzin et al. model suggests a mechanism whereby this would occur without invoking complex individual recognition capabilities: if size, or species, is correlated with behavioral response, then this could account for the assortment seen. A further property of this self-organized sorting is that given consistent differences among individuals, the system will reassemble to form the same configuration (statistically) after it has been perturbed from that state. Self-organized sorting may also improve our understanding of the spatial positions taken up by parasitized individuals within groups. For example, Krause and Godin (1994) in the laboratory, and Ward et al. (2002) in the field, studied the influence of parasitism on the positioning behavior of individuals within natural fish schools (banded killifish). They found that individuals parasitized by the digenean trematode Crassiphiala bulboglossa tended to occupy peripheral positions in the group (see also Barber and Huntingford, 1996, for a similar host–parasite system). It has been suggested that the parasite may be manipulating the behavior of the host, resulting in its modifying the position of its host with respect to others within the group, to increase the chances of propagation of the parasite to its definitive host, the belted kingfisher, Megaceryle alcyon (Barber et al., 2000). Investigating this behavior in the context of the type of sorting mechanisms we have outlined here may improve the understanding of the behavioral modifications that occur in such parasitized individuals. C. Social Dominance Relationships and Structuring within Groups Several authors have suggested that self-organized structuring may also occur within animal groups as a result of dominance interactions among individuals (Hogeweg and Hesper, 1983, 1985; Theraulaz et al., 1995;


45

Bonabeau´ et al., 1996, 1999a; Hemelrijk, 1998, 1999). Dominance relationships among individuals within a group have been recorded for many animals, such as birds (Schjelderup-Ebbe, 1913, 1922), primates (Kummer, 1968; Mendoza and Barchas, 1983; Barchas and Mendoza, 1984; Thierry, 1985), ungulates (Tyler, 1972; Barton et al., 1974), fish (Francis, 1983; Beaugrand and Zayan, 1984; Hsu and Wolf, 1999), and insects (Franks and Scovell, 1983; Heinze, 1990; Bourke, 1988; Oliveira and Ho¨lldobler, 1990). Dominance interactions are typically considered to be ‘‘pairwise’’: that is, most contests involve just two individuals at a time. Individuals that tend to ‘‘win’’ such interactions (termed high-rank individuals) are often thought to increase their access to resources (such as mates or food), so individuals should be expected to strive to increase their rank within the group (Datta and Beauchamp, 1991). However, dominance interactions with others are often aggressive, and thus may be energetically costly and time consuming. Individuals may therefore also be expected generally to avoid conflict, instead relying on passive recognition mechanisms once the hierarchy has been established (Karavanich and Atima, 1998). When such a hierarchy (network of dominance–submission relationships) persists, it should therefore be expected to organize the group in such a way that the costs of the dominance interactions do not offset the benefits of group membership. Within the context of self-organization theory, it has been proposed that a double reinforcement mechanism may explain certain properties of the dominance hierarchies seen in natural groups. Simplistically, such a mechanism assumes that winners of interactions increase their probability of winning future interactions, whereas losers increase their future probability of losing (Chase, 1982a,b). If it is assumed that all individuals are initially similar with regard to their probability of winning interactions, then the outcome of early contests will be relatively unpredictable. However, if by chance an individual wins, this increases its chance of winning a future contest. Similarly, a losing individual is more likely (probabilistically) to lose in future. Thus, this process of feedback and amplification of initial stochastic events can result in progressive differentiation of the group. Such effects have been reported in real animal groups (Ginsburg and Allee, 1975; Chase, 1980, 1982a,b, 1985; Francis, 1983; Beaugrand and Zayan, 1984). However, in reality it is likely that differences among individuals affect their real propensity to be successful in such contests (Slater, 1986), and the initial assumption that all individuals are similar in this respect merely acts to show that inherent differences are not essential to explain the generation of a hierarchy. Of course, a model could be constructed in which individuals have different intrinsic rates of feedback: that is, for some individuals positive and/or

46


negative reinforcement may be stronger than in others. It should be noted, however, that experimentally it may be difficult to differentiate a model based on preexisting differences, a so-called ‘‘correlational model’’ (Chase, 1980), from self-organized alternatives that either do, or do not, incorporate inherent heterogeneity in response to interactions. Even when the data are relatively detailed, as for some social insects such as Polistes wasps (Theraulaz et al., 1989, 1995), it is currently not possible to determine to what degree self-organized reinforcement structures dominant hierarchies because the empirical data can be explained by both correlational and selforganized approaches (Bonabeau et al., 1999a). Interestingly, evidence suggests that even simple organisms such as Polistes may be able to recognize nest mates, and that this ability influences the intensity of dominance interactions (Tibbetts, 2002). Hogeweg (1988) and Hemelrijk (1998, 1999) extended the selforganized models of hierarchy formation to investigate potential spatial effects that may emerge in populations of individuals that exhibit the type of feedback mechanism described previously. Individuals exhibit a simple grouping tendency and can perform double-reinforcement dominance interactions. It should be noted that although a centripetal force (tendency to move toward the group center) has not been explicitly encoded in these models, the propensity of individuals to approach others if they become isolated would result in a mean acceleration of individuals toward the group center (see Okubo, 1980, 1986). Without such an inward-oriented force (relative to the current group center) the group would tend to dissipate by randomness of motion. After a dominance interaction, both the winner and loser of such interactions turn a randomly determined angle of 45 either clockwise, or counterclockwise, and move forward. The loser is assumed to move farther in a given time interval, simulating its being ‘‘chased’’ away (thus it moves more rapidly). The model therefore assumes that the dominance rank of an individual influences the mobility of individuals; more submissive individuals being more mobile. In these models this rule set results in subordinates occupying peripheral positions, and dominant individuals occupying the group center. Given a physical system in which particles move at different rates, similar spatial structuring often results. A commonly known example of this is that more active particles within a liquid or gas often rise (e.g., hot air rising). In this case gravity can be considered analogous to the net centripetal force. The degree to which groups are structured in this manner is related to the difference in motion among individuals, and hence the ‘‘steepness’’ of the hierarchy, with the distance from the group center and dominance level being increasingly negatively correlated as the steepness increases. Thus


47

the model developed by Hemelrijk suggests that this spatial structure will be more defined for despotic societies (a steep hierarchy gradient) than for egalitarian societies (a relatively shallow gradient). Further extensions of this model allow investigations of other properties, including the importance of memory of previous interactions and different strategies with regard to the perceived risk of encounters, and thus provide a useful tool with which to make predictions about dominance relationships in real animal groups (Hemelrijk, 2000). However, it may still be difficult to differentiate between different explanations for the same phenomenon, as discussed by Bonabeau et al. (1999a). D. Leadership The models we have introduced demonstrate that leadership is not a necessary requirement for collective organization of groups. We have also described how leadership may ‘‘emerge’’ within mobile groups, as a result of the interactions among individuals. In some cases the behavioral properties of an individual may bias its probability of being a leader (Gueron et al., 1996; Couzin et al., 2002) (see Krause et al., 2000a, for a review of leadership in fish), whereas in the case of essentially identical individuals the probability of being a leader of a group may be largely random, or be dependent on the initial starting conditions. This concept of leadership is different from that used by early researchers such as Selous (1931) and Presman (1970), who assumed a leader has control of all other group members. This is clearly not the case. However, individuals that happen to be at the front of a group, or whose behavior increases their probability of occupying frontal positions, are likely to have a stronger influence on the motion of the group than are individuals further back, even if all individuals are identical and follow exactly the same rules. Consider a simplistic situation. Assume you are walking at the front of a group that is moving forward in a straight line. If you were to stop suddenly, this would be likely to impact on the motion of other group members, who must now avoid you to continue their journey. If a group is sufficiently fast moving and tightly packed, this can cause great disturbance. However, consider that you perform the same stop behavior, but instead of starting at the front of the group you were to start at the back. Your behavior would have little, or no, effect on other group members, who are not impeded by you. Similarly, within a herd, school, or flock an individual changing speed at the front of a group will have a larger influence on other group members than if it were at the rear. If an individual at the front performs a turn, for example, this also reduces its speed relative to the direction of group

48


motion. Thus it is likely to interact with a large number of other group members, and its orientation is much more likely to be propagated to other group members than if it were at the rear of the group. Such leadership effects may be further enhanced if individuals have a tendency to interact more strongly with those ahead, so-called frontal bias (Huth and Wissel, 1992), which has found support from empirical work on fish shoals (Bumann and Krause, 1993). This may be a result of having a blind area to the rear in which they cannot detect others, or individuals having evolved to bias their movement decisions more heavily to those ahead of themselves in moving groups. This makes sense because the individuals at the front of a group are more likely to encounter stimuli, such as environmental obstacles, sit-and-wait predators, or resources.

IV. Group Size and Composition In the previous sections of this chapter we have considered how the interaction dynamics among individuals result in the formation, internal structuring, and collective behaviors of vertebrate groups. In this section we consider the distribution of grouping individuals over larger spatial and temporal scales, and discuss how individual behaviors lead to population-level dynamics. At an ecological scale, the distribution of social organisms (such as schooling fish and herding ungulates) results from the processes of amalgamation (fusion) and splitting (fission) of groups (here we consider isolated individuals as being in a group of size 1) within the context of their environment. Understanding these properties is essential if we are to understand better disease transmission and the transfer of information among individuals (e.g., social learning). In Section IV.A we discuss how the time scale over which fission and fusion occur can result in stationary frequency distributions of group size within a population, and how modeling may help determine the underlying mechanisms of such processes when only group size distributions are available (as in many natural systems, where the distribution of group sizes is easier to record than the underlying interactions among groups). In Section IV.B we explore how the spatial dimension through which individuals move, and habitat properties such as fragmentation, may affect the distribution of grouping organisms. We then consider in Section IV.C how phenotypic differences among individuals may influence fission– fusion systems. We conclude by considering how the theory of optimal group size can be considered from a self-organization perspective, in Section IV.D.


49

A. Fission, Fusion and Group Size Distributions The fission–fusion processes described previously may often occur sufficiently rapidly (relative to the temporal and spatial scale over which ecological properties may change) that the group size distribution is stable (stationary) (Okubo, 1986; Gueron and Levin, 1995; Bonabeau and Dagorn, 1995; Niwa, 1998b; Bonabeau et al., 1999b; Sjo¨berg et al., 2000). The type or shape of group sizes found within a population is shown in Fig. 15 for African buffalo. Okubo (1986) discusses some of the behavioral and ecological constraints that may result in the equilibrium distribution of group size being unstable within given intervals of time, including sudden changes in the behavior of grouping individuals, or of the environment (e.g., availability of resources, or visibility). We return to the influence of these variables later. However, for simplicity it is reasonable at the outset to assume stability in these properties. When fusion is high relative to fission, then the number of groups with few individuals should tend to decrease (larger groups will be more likely to persist) and the group size distribution would be expected to have a relatively long tail. If fusion is low relative to fission, however, groups tend to be unstable, and large groups are less likely to form. Consequently, group size distributions would be more rapidly decreasing. Several studies have recorded exchange rates between groups of fish. Hilborn (1991), for example, studied skipjack tuna, Katsuwonus pelamis, and found that 16– 63% of individuals changed shoals within 1 day, although Bayliff (1988) found much more stable groupings in the same species. Klimley and Holloway (1999) for yellowfin tuna, Thunnus albacares, and Bayliff (1988) for skipjack tuna, found that cohesion of schools was high, and the half-life of schools was likely to be on the order of weeks. Krause (1993a) found in roach shoals (Rutilus rutilus) that a turnover of more than 50% of the individuals occurred within 2 days. Among killifish (Fundulus diaphanus) shoal encounters were observed frequently (on average every 1.1 min for a

Fig. 15. Frequency distribution of group sizes of African buffalo, Syncerus caffer. (From Sinclair, 1977; used with permission.)

50


given shoal), resulting in high rates of exchange of individuals between shoals and a complete mixing of fish in the population within 24 h (Krause et al., 2000c). A similar result was found by Seghers (1981) in spottail shiners (Notropis hudsonius), and Helfman (1984) reported an absence of shoal fidelity in the yellow perch (Perca flavescens). In summary, these data indicate that fission–fusion processes occur more frequently in freshwater, where density of individuals is high and thus encounters more common, than for pelagic marine species. Thus we would expect that the group size distributions would reflect the different fission–fusion dynamics of these groups. It is important to note, when looking at group size distribution data, that the pattern seen results from a dynamic process. Even though such a distribution can often be relatively stationary, it is the continuous splitting and fusion of groups that makes it so. Consequently it represents the probability distribution of a given individual being in a group of a certain size at any given moment of time. Considering the large scale over which fission–fusion dynamics takes place, some modelers (Okubo, 1986; Gueron and Levin, 1995) have made the assumption that the population properties can be described by a nonspatial approach, in which an attempt may be made to define the average rate of fusion or fission of groups. However, if we consider the mechanisms at the level of the individual it will become clear that both the probability of a group encountering another (and fusing) and the probability of a group spontaneously fragmenting (fission) will be dependent on the size of the group in question. For example, the range over which individuals can interact (and/or the strength of interaction) is likely to influence group cohesion. As described in Section II.D, groups much larger than the range of individual interactions can form. However, as group size increases, it will be increasingly likely that a group will fragment (because of the inherent stochastic nature of interactions and motion). Such fragmentation may be exacerbated by interactions with other groups and/or the environment (see Section IV.B). As group size increases other properties may also change, such as the velocity of the group, and the probability of a group encountering other groups (Flierl et al., 1999). Thus, although a system can be relatively easily described in terms of a time-dependent dynamic function of the number of groups of a given size incorporating size-dependent fission and fusion rates (see Okubo, 1986; Flierl et al., 1999), defining (and verifying) realistic fission–fusion functions used can be a complex task. Flierl et al. (1999) use individual-based models of fish schooling to estimate some of these functions (e.g., fission rates as a function of school size), and group size distributions in their model tend to be nearly exponentially distributed.


51

Using an alternative technique involving a ‘‘maximum entropy’’ principle, Okubo (1986) also predicted that all group size distributions should be exponentially decreasing (see Okubo, 1986, for a detailed mathematical description of the model), and fitted this model to a range of experimental data from fish species, including the spottail shiner, Notropis hudsonius, and ungulates including American bison, Bison bison, and desert bighorn sheep, Ovis canadensis. However, Bonabeau et al. (1998, 1999b) argue that the Okubo model (1986) assumes that there is a fixed average size to animal groups (and thus that there is a well-defined mean to such distributions), and that in reality when maximal group size is large, group size distributions may exhibit longer tails than predicted by a decreasing exponential function. Furthermore, they argue that such longtailed distributions are likely to be truncated because populations are ultimately finite, and are rapidly decreasing at large sizes. Consequently they propose that animal group size distributions may conform to a ‘‘truncated power law,’’ in which the number N(s) of groups of size s is proportional to sb, where b is the scaling exponent, up to a cutoff group size C. If group size distributions do follow a power law, then Bonabeau et al. (1998, 1999b) suggest that biotic factors that may influence the stability of groups, such as resource availability, should be expected to affect the cutoff size, but not the power index b (which corresponds to the slope of the function when plotted on a log–log scale), which is scale invariant. They were able to show that experimental data on group size distributions from fish schools (tuna and sardinella) and ungulate herds (African buffalo) exhibit long-tailed distributions characteristic of the truncated power law (indicating that such species form relatively cohesive, stable groups). Where data were available, for tuna fish, they also demonstrated that cutoff size does vary between years, but b appears relatively constant. If this model does indeed fit the observed data, Bonabeau et al. (1998, 1999b) also suggest that this may indicate that the underlying aggregation mechanism may be relatively simple (at least in terms of the join–leave probabilities), whereas the cutoff size could be used to reflect biologically important properties, such as changes in individual behavior, individual density, or the environment. The ability to determine the cutoff point, which may represent a critical, and biologically meaningful, aggregate size, is one of the potential strengths of applying the truncated power law model, as opposed to the other methods described previously, to group size distributions. However, the prediction that cutoff size would vary with density-dependent properties was not explicitly tested by Bonabeau et al. (1998, 1999b), presumably because of insufficient data. Sjo¨berg et al. (2000) demonstrated that they could fit truncated power laws to data from gray seals (Halichoerus grypus) aggregating on haul-out sites (Fig. 16),

52


Fig. 16. Frequency distribution of group sizes of gray seals, Halichoerus grypus, resting on an island. The fit shows a truncated power law with a cutoff at group size 21 [whole model: b1 (slope of line from group size 1 to 21) = 0.35, b2 (group size > 21) = 0.93, R2 = 0.91, p < 0.001]. (From Sjo¨berg et al., 2000; used with permission.)

larvae of tephritid flies (Paroxyna plantaginis) clumping in flower heads, and aphids (Aphidiodea spp.) aggregating on stems. However, they also provide some evidence that cutoff size does vary as a function of densitydependent effects (resource density and individual density) for some insects (aphids), although in their system the resource distribution is also likely to influence the group-size distribution. They also provide evidence that, for aphids and tephritid fly larvae, the exponent of the power law (the slope b) may be influenced by biotic factors. Thus they conclude that the simple mechanistic approach to understanding aggregation phenomena proposed by Bonabeau et al. (1998, 1999b) may not necessarily be suitable for other biological systems. Niwa (1998b) modified a balance equation model by Gueron and Levin (1995), in which the fission–fusion processes were shown to result in a stationary solution. Like Bonabeau et al. (1998, 1999b), he applied this modeling approach to understanding the group size distributions, inferring (as did Flierl et al., 1999) the fission–fusion rates from models of schooling dynamics (see Niwa, 1994, 1996). This model was fitted to data from freeswimming tuna (as used by Bonabeau et al., 1998, 1999b), Japanese sardine (Sardinops melanostictus), northern anchovy (Engraulis mordax), and flying fish (Cypselurus opisthopus hiraii and Cypselurus heterurus do¨derleini). Niwa argued that the school size distributions fit a truncated power law with a crossover to an exponential decay around a certain cutoff


53

size. This cutoff was dependent on the total population size, and is a result of fission–fusion within a finite population. Thus when the cutoff size is small (when populations are relatively small compared with these pelagic marine examples, as may be the case for some freshwater species: see Seghers, 1981, and Okubo, 1986), the exponential decay may be the only part of the function evident. As the population size increases one may therefore predict that the group size distribution would better fit a power law. The Niwa model (1998b) also predicts that the exponent of the power law does depend on population size, and that as population size increases the exponent should approach 1 over a wider range of group sizes (the population exhibiting a longer tail in the group size distribution). B. Influence of Habitat Structure Continuing the discussion of the truncated power law description of animal group size distributions proposed by Bonabeau et al. (1998, 1999b), a further prediction was made that the power index, b, should vary predictably as a function of the spatial dimension of the system in question (also see Takayasu, 1989). The reason b increases is because at a low spatial dimension the spatial constraints mean that groups have a higher probability of meeting and fusing, relative to that at a higher dimension (given all other properties of the system are constant). Thus the truncated power law model predicts that b increases with effective dimension, but that when the effective spatial dimension is less than 3 (as in all biologically reasonable cases) then b < 3/2. Specifically b = 4/3 (1.33 . . .) for d = 1, b = 1.465 0.003 for d = 2, b = 1.491 0.007 for d = 3 (although it should be noted that Niwa, 1998b, and Sjo¨berg et al., 2000, question whether b must be less than 3/2). Despite the fact that some animals, such as fish, move in threedimensional space, the individuals may not actually use the space available, as in the case of the schools of Atlantic bluefin tuna (Thunnus thynnus) discussed in Section II.D.3, which may predominantly occupy two-dimensional space, cruising just below the water surface (Partridge et al., 1983). Bonabeau et al. (1998, 1999b) tested this prediction with data from fish species that differ in their space use. Free-swimming tuna (a mixed population of yellowfin tuna, Thunnus albacares; skipjack tuna Katsuwonus pelamis; and bigeye tuna, Thunnus obesus) often move in open ocean, but are still likely not to use fully the space available to them. For example, they are more likely to be parallel, rather than perpendicular, to the surface (because of the influence of gravity). Bonabeau et al. (1998, 1999b) term the dimensionality of space actually used by the organisms as the ‘‘effective dimensionality’’ of the system. In the case of free-swimming

54


tuna their effective space use will be somewhere between two and three dimensions. Environmental structure is likely to influence this space use. Sardinellas (Sardinella maderensis and S. aurita), for example, tend to follow the coastline of West Africa, above the continental shelf (which limits water depth from 1 to 200 m) (Bonabeau et al., 1998, 1999b). Therefore the effective spatial dimension would be somewhere between 1 and 2. Space use may be reduced even further in some instances, such as when pelagic fish gather under artificial buoyant objects, known as fishaggregating devices (FADs). Bonabeau et al. (1998, 1999b) argue that the effective dimension of tuna fish schools caught in the vicinity of an FAD is less than 1 because the FAD is a point (relative to the large-scale spatial movements of tuna fish). It should also be noted that such a device also affects the aggregation dynamics by introducing an attractive focus to individuals. Comparing the predictions of their model with the data from sardinellas and tuna, described previously, Bonabeau et al. (1998, 1999b) demonstrated a qualitative, but not a quantitative, fit (b = 1.49 for freely swimming tuna fish, b = 0.95 for free-swimming sardinellas, and b = 0.698 for tuna fish caught in the vicinity of an FAD). This shows that, as expected, the exponent b of the power law is inversely related to the spatial dimension used by the animal. Because these animals (as described previously) use an unknown dimensionality of space it is perhaps not surprising that the fit to three, two, and one dimension for the free-swimming tuna, sardinellas, and aggregated tuna is only qualitative, although Bonabeau et al. (1988, 1999b) argue that the lack of perfect agreement with the empirical data with which they tested their model may result from biases in available data. First, their data came from schools caught by purse seine nets, and consequently a catch made may include only a subsection of a school (and this would be more likely as school size increased). Furthermore, because fishermen are not necessarily interested in small schools (and may use technology such as acoustic imaging to focus on larger groups) these are likely to be under represented in these data. Another limitation to these data is that school size is based not on an actual count of the number of individuals, but on an estimate made from the weight of each haul. As discussed in Section III.B, phenotypically similar individuals become associated within such groups, and this can result in groups, when they fragment, becoming phenotypically assorted (see Section IV.C). This means that in a fishery sample (which is assessed by weight) a small group of large individuals would be indistinguishable from a larger group of smaller individuals, given that the two are of similar weight. In all grouping animals the effective space is likely to be an important consideration when attempting to understand their distribution in space


55

and time. Animals are likely to live in heterogeneous habitats, and their behavior may often be influenced by habitat type. To some degree the effective dimensionality of the environment may be characterized as the fractal dimension of the spatial distribution of patches suitable for movement. However, the situation becomes complicated when there is more complex variation in habitat type (rather than a binary classification of merely suitable and unsuitable habitat). Some habitats may act to attract individuals (e.g., areas with high food abundance), whereas others may restrict motion (e.g., where there is structural complexity, as in forest). Habitat structure is also likely to affect other properties important in determining the fission–fusion dynamics. For example, some habitat (e.g., forest) may restrict the range over which individuals can respond to one another, and hence limit the interaction range. This is likely to have the effect of increasing the fragmentation of groups. In the case of animal groups that are mobile, motion around obstacles in the environment is also likely to increase the probability of splitting, and so the detailed nature of the habitat structure (such as the size and distribution of obstacles) may be expected to be an important influence on group size distribution. In some cases habitat structure may change rapidly. Flierl et al. (1999), for example, used computer modeling to investigate the consequences of turbulent flow in aquatic environments on the grouping dynamics of fish. In many fluid environments changes in flow regime may be rapid (e.g., the volume of water in a given stretch of a freshwater waterway may change rapidly as a result of flooding). Furthermore, where conditions are turbulent there are likely to be rapidly changing shear and strain fields that will exert physical forces on animals. Under turbulent conditions it may be expected that groups will be more likely to fragment because individuals will have insufficient control over their locomotion (relative to the strong physical forces exerted on them by the flow conditions) to maintain cohesion. At the very least it may be expected that even weak turbulence will act to impose largely stochastic physical forces that would decrease cohesion. In the model developed by Flierl et al. (1999), which incorporated a simple schooling tendency similar to that used by Gru¨nbaum (1997, 1998), strong turbulence was shown to fragment groups, but where turbulence is weaker groups can form in temporary ‘‘refuges.’’ Once formed, turbulence actually acts to increase the size of groups that form because the flow acts to increase the encounter rates between groups as individuals tend to occupy the spaces in between high-turbulence areas. This suggests that the turbulent regime of aquatic environments is likely to have a strong influence on fission–fusion processes of grouping and consequently will affect the resulting group sizes in the population. This model also highlights the importance of considering physical

56


properties of the environment that are potential pattern-forming processes. It is clear that there is still much debate about the processes involved in fission–fusion systems, yet the diverse range of mathematical techniques used has provided constructive and thought-provoking discussion of a topic that is relevant not only to our understanding of collective behaviors and ecological questions, but also of conservation issues, concerning which the models may allow a better understanding of how changes to the environment, or to the density of organisms, may affect group- and population-level processes. C. Phenotypic Assortment: Active or Passive? There is considerable empirical evidence that animals (most of the data come from fish shoals) tend to be assorted by phenotype between groups (Krause et al., 2000b). This includes sorting by body length, species, parasite load, and body color. In Section III.B, we discussed how differences among individuals within a group can lead to ‘‘natural’’ sorting: individuals with similar behaviors tend to become more closely associated as a result of the interaction mechanics. We hypothesized that, if properties such as body size, or species, are correlated with behavioral response, this could explain the fact that fish within schools tend to be close to others of similar size (Pitcher et al., 1986; Parrish, 1989a; DeBlois and Rose, 1996; Ward et al., 2002) or, in multispecies groups, to conspecifics (Parrish, 1989a). However, we did not consider the consequences of such self-sorting processes to population-level properties. Here we consider such sorting processes within the context of a fission– fusion system, and show how we can make some predictions about how these mechanisms are likely to influence group-level properties. The close association of individuals with similar phenotypes within a group means that, when groups fragment, individuals will tend to remain with others that are more similar to themselves (Croft et al., 2002). Thus phenotypically heterogeneous groups, when they fragment, will tend to do so into more homogeneous groups. This phenotypic assortment is counteracted to some degree by the merging of groups of different phenotypes. To better understand how this process works at a population level we construct a deliberately simplistic model of grouping. This model is similar to that outlined in Section II.D.2 (see Couzin et al., 2002, for further details), except here we simplify the model further: individuals have a close-range zone of repulsion that simulates their tendency to maintain a minimum distance between themselves, and a single 360 zone of perception that extends beyond this and in which they can detect others.


57

As described previously, the zone of repulsion has highest priority, but if there are no individuals within this zone, individuals will align with, and be attracted to, neighbors within the zone of perception (for simplicity these forces are assumed to have equal weighting). This grouping behavior is subject to slight error (stochastic effects). If no others are detected, individuals perform a correlated random walk. Further details of the formulation of this model can be found in Hoare et al. (2002). In our model, individuals are 4 cm long, move at 5 cm s1, and have a maximum turning rate of 100 /s, corresponding to the killifish Fundulus diaphanus. However, this model is generic, and the processes we describe are not dependent on the exact parameters used. We assume individuals have localized perception, and respond to others within two body lengths. Again, changing this parameter changes the results quantitatively but not qualitatively. Our simulated organisms move in continuous space on a twodimensional plane with periodic boundary conditions. Within this model individuals form mobile groups that exhibit fission and fusion. To investigate how differences among individuals can change group composition, we assume that there are two subpopulations that may, or may not, differ with respect to their behavior. This may correspond, for example, to two species, or to two classes of individuals (such as hungry vs satiated individuals). Clearly further modifications, such as simulating continuous variation in behavior, would be interesting, but for simplicity here we assume just two ‘‘types’’ of individual within the population. Within this model, even slight behavioral differences between the two subgroups results in groups becoming phenotypically assorted (see Fig. 17). One of the strongest sorting influences is a difference in speed between the two groups (Fig. 17A and B). Figure 17C shows how individuals assort when there is a difference in the size of the zone of repulsion between the two subpopulations (in Section III.B we discuss how this property affects the positioning of individuals within groups). Thus if different species, or types of conspecific, differ with respect to their behaviors (e.g., tend to move at different speeds, or tend to respond to others over different ranges), this may result in their becoming ‘‘naturally’’ self-sorted within a population. It may not be necessary to invoke complex recognition and decision-making capabilities on behalf of the organism, although to the human observer it may appear that individuals are behaving in a more complex way. For example, within our computer model a naı¨ve observer watching the individuals move around on the computer screen is likely to assume that they have been programmed to make complex decisions about whether to associate with others. This misconception results from biases in our perception because we often tend to consider behavior from too anthropocentric a point of

58


Fig. 17. Simulation model demonstrating how self-organized sorting can result in phenotypic assortment within groups in a population. (A) Typical snapshot of the simulation at dynamic equilibrium: Total number of individuals = 300, with the two ‘‘types’’ of individual (150 of each) shown in black (moving at 5 cm s1) and gray (moving at 7 cm s1); the domain size is 5 5 m with completely periodic boundaries. (B and C) The mean degree of assortment


59

view. We suggest that natural selection may act on such local rules of thumb to facilitate phenotypic assortment within groups if that confers a benefit to individuals. Our model suggests that heterogeneity within animal groups may make them more susceptible to fragmentation: groups of individuals that differ with respect to certain behaviors (e.g., speed or desired personal space) will tend to be less stable (and therefore more likely to fragment) than those in which individuals are phenotypically similar. Our results also suggest that the population will become more phenotypically assorted as the difference between the two subpopulations (behavior types) increases (Fig. 17B and C). Typically when researchers have observed phenotypic assortment within populations, they have assumed that individuals are making active choices with whom they group. A preference for conspecifics over heterospecifics is assumed to have functional significance. First, conspecifics may be more likely to be in spatial proximity after hatching, and thus may be expected to develop antipredator maneuvers with conspecifics as opposed to with heterospecifics (Krause et al., 2000b). Furthermore, individuals are thought to avoid being phenotypically ‘‘odd’’ individuals within a group because this may enhance their risk of predation (through the predator being more likely to ‘‘lock’’ its attention on the odd individual; Landeau and Terborgh, 1986; Theodorakis, 1989). There is some evidence that in mixed-species shoals the less common species may leave when the perceived threat of predation is high (Wolf, 1985). Furthermore, Allan and Pitcher (1986) reported that multispecies shoals separated into their component species when under predation threat. A similar explanation (avoidance of oddity) has been put forward to explain body length sorting in single-species groups (reviewed in Ranta et al., 1994, and Krause et al., 2000b). It may seem from our explanation of self-sorting that we contradict this view by suggesting a passive sorting strategy and not an active decision-making process. However, this is not the case. We do, however, point out that in some instances it may be difficult to determine whether individuals are actively sorting (i.e., making an active decision to leave a group), or whether this is an inevitable consequence of different behavior types that have evolved for another reason. However, we must stress that the type of self-sorting processes we have described may result from selection to allow assortment of individuals for the functional benefits we describe here. (where 1 = all individuals identical) within groups consisting of three or more individuals as a function of (B) the difference in speed between the two types (kept constant at 5 cm s1 in one type and increased in the other), and (C) the difference in size of the zone of repulsion (kept constant at 4 cm, 1 BL [body length], in one group and decreased in the other).

60


Thus, just as we argue that this type of self-organizing mechanism may be selected to allow individuals to modify their position within a group without necessitating complex cognitive abilities (or knowledge that would be difficult or impossible to obtain), natural selection may also select rules of thumb that individuals could use to become assorted by phenotype. The case, described by Wolf (1985), in which less common species in mixed species groups leave under threat of predation is consistent with a self-sorting type of mechanism. The perturbed group is likely to fragment, with odd individuals being ‘‘shed’’ as the group performs avoidance maneuvers. Another interesting point is that the empirical literature so far appears to be contradicting itself. The Landeau and Terborgh study (1986) clearly showed that individuals are at a higher risk in a group where they are phenotypically odd compared with one comprising phenotypically similar group members (provided both groups are of the same size). The higher predation risk explains why an individual should switch from a group where it is odd to one where it ‘‘fits in’’ and do so particularly under predation threat. However, Landeau and Terborgh (1986) also convincingly demonstrated that, if there is no such alternative, then an odd individual does much better by staying in a group where it is odd than by being on its own because when alone the predation risk is even higher. Furthermore, they reported that no cost due to oddity occurred, provided shoals were larger than about 15 fish, because the antipredator effects of grouping became so efficient at this group size that the predator could not make any captures regardless of whether the group did or did not contain an odd fish. In this context it seems surprising that Allan and Pitcher (1986) and Wolf (1985) found that different species separated under predation threat. We should expect to see the opposite, namely different species merging into shoals so that all individuals benefit from a large shoal size that renders oddity irrelevant. We suggest that multispecies groups split into single-species groups in such situations because of constraints imposed by species-specific behaviors (including potential differences in response latency, speed of locomotion, and interaction rules). Thus the split into single-species groups is not an adaptive behavior that lowers predation risk when under attack but a result of a constraint that is likely to increase risk but that fish cannot overcome in this situation. Interestingly, another benefit of phenotypic assortment may be that information transfer (in terms of changes in individual velocity being propagated across the group) may be more efficient in homogeneous than in heterogeneous groups (see Section II.D.4) and we encourage further research in this area. In the case of individuals within groups being assorted by size, it would be interesting to determine whether this is a consequence of individuals


61

somehow assessing the size of others relative to themselves (which introduces the problem of how individuals know their own size, and assess the size of others in the absence of stereo vision, as is the case over much of the field of view of most grouping animals) and choosing to associate with similar individuals. An alternative possibility is that size is correlated with a behavior that results in groups becoming assorted by the type of mechanism we propose here. Similarly, groups within populations in which there are parasitized and nonparasitized individuals may be expected to become assorted by parasite prevalence, and/or load (Ward et al., 2002). However, in some instances body length and parasite load may be correlated (Hoare et al., 2000), so it may be difficult to determine what causes the sorting. In reality it may also be difficult to determine whether the behavioral difference that results in a population being self-sorted has evolved for that purpose, or whether sorting is an epiphenomenon that merely does not incur a cost. Researchers should perhaps bear in mind that assortment may result from self-sorting processes. It should be noted that we are not saying that grouping individuals cannot, or do not, use more complex recognition and response behaviors. Rather, we aim here to introduce the possibility that complex phenomena at the level of the population may also be explained by alternative (and sometimes simpler) mechanisms. D. Optimal Group Size There are costs and benefits to being in groups (Ritz, 1997; Krause and Ruxton, 2002). Grouping individuals may, for example, decrease their chances of being consumed by a predator by positioning themselves near others (Hamilton, 1971). This is sometimes known as the ‘‘dilution effect’’ because, if a predator randomly selects prey, then an individual having near neighbors may dilute its chances of being consumed. As discussed in Section II.D.4, individuals within a group may benefit from information exchange about the positions of predators, and perhaps the ‘‘confusion effect’’ if individuals perform synchronized escape maneuvers (Partridge, 1982). A potential disadvantage of aggregation is that a group of individuals is more likely to be conspicuous to predators than a single individual. With regard to foraging behavior, grouping may benefit individuals by allowing transfer of information about resources (see Section II.D.4), but costs may also result from individuals within groups competing for resources once they are found (Krause and Ruxton, 2002). Because the costs and benefits of grouping vary as a function of group size, we may expect that individuals will modify their choice of group size (by joining or leaving groups) as ecological conditions change to maximize

62


their fitness (Pulliam and Caraco, 1984). One problem with this possibility is that individuals within a group may not be able to exclude solitary individuals, and consequently individuals will continue to join a group exceeding the ‘‘optimal group size’’ (Sibly, 1983). Thus solitary individuals, by joining a group, may increase their own fitness, but decrease that of all other group members. When the costs of grouping become greater than the benefits individuals should be expected to leave (see Krause and Ruxton, 2002, for further discussion of the costs and benefits of group size). Laboratory studies involving fish have shown that individuals, when presented with a simple binary choice of associating with one of two stimulus shoals (within containers so the perceived group size can be modified experimentally), usually ‘‘select’’ the larger group (Krause and Ruxton, 2002). However, see Van Havre and FitzGerald (1988) for an exception. There are several potential problems with such studies. First, test fish may be stressed when alone in the test compartment and, second, the range of shoal sizes that can be presented is limited because of the confined space in the laboratory, and rarely comprises more than 20 individuals. In nature, however, fish can often be found in shoals of hundreds or thousands of individuals. Thus there is a real need for more field work to be carried out in this area (see Hensor et al., 2002). In a laboratory study in which individuals could freely associate with others, Hoare et al. (2002) investigated the influence of ecological factors (perceived food availability and predation risk) on the schooling behavior of banded killifish. Because natural group sizes will result from the interactions among all individuals it is important, as they point out, to consider what group sizes result when all individuals can make membership decisions. They subjected groups of 10 size-matched fish to four treatments: (1) food, (2) control, (3) food and alarm, and (4) alarm. The food treatment involved adding food odor to the water (but to prevent competition for food items themselves no food particles were introduced), and predation risk was simulated through the use of killifish skin extract (which contains chemicals that cause alarm in fish and are naturally released when fish are injured or captured by predators). In the control treatment no odor was added, and in treatment (3) both food and alarm odors were added to the water. Group sizes were shown to be context dependent, with individuals tending to be in the smallest group sizes in the presence of food odor (Fig. 18a, part i), and the probability of individuals being found within larger groups increasing under control conditions (yet groups of five or more individuals are still rare; Fig. 18a, part ii). In the presence of alarm substance, however, the fish tended to form large groups, with individuals spending the majority of time in the maximum group size of 10 (Fig. 18a,


63

Fig. 18. Percent frequency distribution of median group sizes. (a) Model results demonstrating that increasing the range of interaction produces changes in group size distribution. These are similar to those obtained experimentally (b). (b) Experimental data compared with the results of the model. (i) Interaction radius = 1.2 BL [body length]; food treatment. (ii) Interaction radius = 1.5 BL; control treatment. (iii) Interaction radius = 1.6 BL; food + alarm treatment. (iv) Interaction radius = 2.9 BL; alarm treatment.

64


part iv). When alarm substance and food odor were present, fish exhibited a response intermediate to that in the presence of food odor or alarm odor alone (Fig. 18a, part iii). This demonstrates that group size in this organism is context dependent. Individuals increase their probability of being within small groups when food odor is detected because this may reduce intragroup competition for resources when found. When alarm odor is detected, however, individuals would be expected to form large groups, because by doing so fish may reduce per capita predation risk. In the case of both food odor and alarm odor, the fish seemed to have conflicting tendencies, resulting in their performing an intermediate behavior. This is interesting, as the fish seem to be trading off foraging benefits and safety from predators in their movement decisions. It may at first be thought that these killifish regulate group size by assessing how many individuals they are currently schooling with, and making a decision to stay or leave (or approach others) on the basis of that. However, this is a rather anthropocentric view of the behavior of these animals that assumes they can count the number of individuals within their group. Hoare et al. (2002) suggest that we need not invoke such complexity in the individual decision-making process, and that group size may be an emergent property resulting from fish following relatively simple rules of thumb. To support this conjecture, they developed an individual-based model of their experiment, similar to that we used previously to examine self-sorting within animal populations (see Section III.B). In their model they assumed that fish change the range over which they interact with others as environmental conditions change: individuals that detect food tend to respond only to very near neighbors, whereas those that experience alarm odor will increase the range over which they interact with others, so that they aggregate, thus avoiding isolation. By modifying this range of interaction within their model they could investigate its consequence on the group size distribution at the level of the population (Fig. 18b). This change in local response was shown to be able to account well for the shift in group size distributions from small to large groups recorded experimentally (see Fig. 18a, where the model data are compared with experimental data). This type of simple model demonstrates how individuals can modify their probability of being within a group of a certain size by changing a local behavior and, as Hoare et al. (2002) point out, their aim was not to determine the exact rules used by their fish, but rather to show that individuals can modify their probability of being in a group of a certain size without making explicit decisions about membership of particular groups. Thus their model demonstrates the logical consistency of their argument.


65

An individual fish need not know the range of possible group sizes available to it; rather, the group size distribution can be an emergent property of local interactions. This approach is appealing (and plausible) for other systems where it is unlikely that individuals can assess the size of the group they join, such as in pelagic fish populations where group sizes may frequently be on the order of thousands or even hundreds of thousands, or where interactions must be local, as in turbid water. As this model demonstrates, the ‘‘decisions’’ made by individuals may be much simpler than they may initially appear. Interestingly, in a study of African buffalo, Syncerus caffer, Sinclair (1977) found that the size of herds changes throughout the year. During the wet season herd sizes tend to be large, but in the dry season groups tend to be much smaller. Sinclair (1977) suggests that the groups may become larger during the rut, which begins at the start of the wet season. Furthermore, aggregation may act to protect the young produced in the wet season. However, most conceptions occur at the end of the wet season in this species, so Sinclair also argues that changes in resource availability (productivity is greater in the wet season) may also be important. Thus, the buffalo may be responding to resource availability for similar reasons as do the fish described previously, reducing competition for resources when food is limited and increasing group size when the productivity of their environment increases. In the study by Hoare et al. (2002), described previously, all individuals should be expected to have the same motivations (all individuals had the same preexperiment feeding regime, and all had the same stimulus and stimulus intensity). In reality, however, the situation is likely to be more complex than this. Satiated individuals, for example, would be expected to respond less strongly (if at all) to food odor, when compared with hungry fish. In addition to variation in individual state, there is also likely to be variation in the perceived stimuli, and also in the inherent propensity of individuals to respond (e.g., variation in general schooling tendency is known to occur within populations; Magurran et al., 1995). Individuals within the types of fission–fusion systems we have considered here would be expected, therefore, to change their behavioral response to others dynamically in order to increase their probability of being in groups of a size that approximates their current ‘‘optimal’’ group size. We therefore encourage further research into understanding whether, and how, potential self-organizing mechanisms can result in individuals maximizing their fitness by changing their probability of being in groups of a certain size. Such an approach should consider both current experimental evidence that shows how grouping individuals can regulate group sizes with the properties we have discussed previously in this chapter, such as intragroup

66


self-sorting processes, and the resulting group size distributions seen at the level of the population.

V. Summary We have looked at different taxonomic groups to reveal where selforganization theory can make an important contribution to explaining collective behavioral patterns. Because this is a newer area of research, and because vertebrate groups may be difficult to study, developing theories of self-organization for these groups (which can then be tested empirically) is particularly challenging. Consequently we focused on how modeling approaches (particularly those that are individual based) have been, and are being, used to help reveal the organizational principles in human crowds (Sections II.B.1 and II.C), ungulate herds (Section II.A), fish schools, bird flocks (Section II.D), and primate groups (Section III.C). The collective behavior of such systems is largely characterized by the interactions among individual components, and thus is well suited to an approach that seeks to elucidate generative behavioral rules. We also discussed the evolution of collective behaviors (Section II.D.3). Here, theory has been important in demonstrating that different collective behaviors can exist for identical individual behaviors, suggesting that the evolution of collective (extended) phenotypes may be more complex than it may, at first, appear. Behavioral differences among individuals within a group may have an important internal structuring influence, and by using simulation models we showed how individuals can modify their positions relative to other group members (e.g., to move relative to the front or center of a group) without necessitating information about their current position within the group (Section III.B). This is important because it is unlikely that individuals within large groups (e.g., pelagic fish schools) can determine their absolute position relative to all other group members; thus we argue that natural selection is likely to act on the kind of local rules we discussed. In Section IV we discussed how local self-organized interactions result in the distribution of animals at a larger spatial and temporal scale, showing how mathematical studies of group size distributions are being used to make testable predictions about how individual behavior translates to that at the level of a population (Sections IV.A and IV.B) and how differences among individuals within a population may lead to phenotypically assorted groups within a population (Section IV.C). We also addressed the ‘‘optimal group size’’ concept (Section IV.D). As an alternative to the view in which individuals explicitly assess the size of groups and then make


67

a decision to leave or join, we showed how local rules of thumb could be used by individuals to modify their probability of being within a group of a given size. We demonstrated that in real organisms (schooling fish) group size distributions (and hence the probability of an individual being within a group of certain size) is context dependent, and that this behavior is entirely consistent with a self-organized mechanism whereby individuals change local interactions as conditions change. In considering self-organization within vertebrate groups it is evident that the organization at one level (e.g., that of the group) relates to that at higher levels (e.g., that of the population). For example, self-sorting processes that lead to internal structuring within groups also result in population-level patterns when such groups fragment (e.g., phenotypic assortment), thus affecting the probability that an individual will be in a group of a given size and composition at any moment in time. These population properties then feed back to the individual interactions by changing the probability of encounters among different members of a population. Thus, to understand collective behaviors fully these properties cannot necessarily be considered in isolation.

Acknowledgments We thank Nicola Grimwood, Nigel Franks, Ana Sendova-Franks, Simon Levin, Ashley Ward, Dick James, Darren Croft, Liz Hensor, Ian Rozdilski, and Dan Hoare for helpful discussions. We also thank Peter Slater, Jay Rosenblatt, Tim Roper, and Sally Ward for reading the manuscript and making many helpful suggestions. I.D.C. acknowledges financial support from the Leverhulme Trust, the Pew Program in Biocomplexity at Princeton University, and the NSF (award 0201307 to Martin Wikelski).

References Allan, J. R., and Pitcher, T. J. (1986). Species segregation during predator evasion in cyprinid fish shoals. Freshwater Ecol. 16, 653–659. Alt, W. (1980). Biased random walk models for chemotaxis and related diffusion approximations. J. Math. Biol. 9, 147–177. Aoki, I. (1982). A simulation study on the schooling mechanism in fish. Bull. Jpn. Soc. Sci. Fish. 48, 1081–1088. Barber, I., and Huntingford, F. A. (1996). Parasite infection alters schooling behaviour: Deviant positioning of helminth-infected minnows in conspecific groups. Proc. R. Soc. Lond. B Biol. Sci. 263, 1095–1102. Barber, I., Hoare, D. J., and Krause, J. (2000). The effects of parasites on fish behavior: A review and evolutionary perspective. Rev. Fish Biol. Fish. 10, 131–165. Barchas, P. R., and Mendoza, S. D. (1984). Emergent hierarchical relationships in rhesus macaques: An application of Chase’s model. In ‘‘Social Hierarchies: Essays toward a

68


Sociophysiological Perspective’’ (P. R. Barchas, Ed.), pp. 81–95. Greenwood Press, Westport, CT. Barton, E. P., Donaldson, S. L., Ross, M. A., and Albright, J. L. (1974). Social rank and social index as related to age, body weight and milk production in dairy cows. Proc. Indian Acad. Sci. 83, 473–477. Bayliff, W. H. (1988). Integrity of schools of skipjack tuna, Katsuwonus pelamis, in the eastern Pacific Ocean, as determined from tagging data. Fish. Bull. 86, 631–643. Beaugrand, J. P., and Zayan, R. C. (1984). An experimental model of aggressive dominance in Xiphophorus helleri (Pisces: Poeciliidae). Behav. Processes 10, 1–53. Ben-Jacob, E., Shochet, A., Tenenbaum, I., Cohen, A., Czirok, A., and Vicsek, T. (1994). Generic modelling of cooperative growth in bacterial colonies. Nature 368, 46–49. Boi, S., Couzin, I. D., Del Buono, N., Franks, N. R., and Britton, N. F. (1999). Coupled oscillators and activity waves in ant colonies. Proc. R. Soc. Lond. B Biol. Sci. 266, 371–378. Bonabeau, E., and Dagorn, L. (1995). Possible universality in the size distribution of fish schools. Phys. Rev. E 51, R5220–R5223. Bonabeau, E., Dagorn, L., and Freon, P. (1998). Space dimension and scaling in fish schoolsize distributions. J. Phys. A 31, L731–L736. Bonabeau, E., Dagorn, L., and Freon, P. (1999b). Scaling in animal group-size distributions. Proc. Natl. Acad. Sci. USA 96, 4472–4477. Bonabeau, E., Theraulaz, G., and Deneubourg, J.-L. (1996). Mathematical models of selforganizing hierarchies in animal societies. Bull. Math. Biol. 58, 661–719. Bonabeau, E., Theraulaz, G., Deneubourg, J.-L., Aron, S., and Camazine, S. (1997). Selforganization in social insects. Trends Ecol. Evol. 12, 188–192. Bonabeau, E., Theraulaz, G., and Deneubourg, J.-L. (1999a). Dominance orders in animal societies: The self-organization hypothesis revisited. Bull. Math. Biol. 61, 724–757. Bourke, A. F. G. (1988). Dominance orders, worker reproduction, and queen-worker conflict in the slave-making ant Harpagoxenus sublaevis. Behav. Ecol. Sociobiol. 23, 323–333. Brown, R. E., and MacDonald, D. W. (1985). ‘‘Social Odours in Mammals.’’ Clarendon Press, Oxford. Bumann, D., and Krause, J. (1993). Front individuals lead in shoals of three-spined sticklebacks (Gasterosteus aculeatus) and juvenile roach (Rutilus rutilus). Behaviour 125, 189–198. Bumann, D., Krause, J., and Rubenstein, D. I. (1997). Mortality risk of spatial positions in animal groups: The danger of being in the front. Behaviour 134, 1063–1076. Camazine, S., Deneubourg, J.-L., Franks, N. R., Sneyd, J., Theraulaz, G., and Bonabeau, E. (2001). ‘‘Self-Organization in Biological Systems.’’ Princeton University, Princeton, NJ. Chase, I. D. (1980). Social process and hierarchy formation in small groups: A comparative perspective. Am. Sociol. Rev. 45, 905–924. Chase, I. D. (1982a). Dynamics of hierarchy formation: The sequential development of dominance relationships. Behaviour 80, 218–240. Chase, I. D. (1982b). Behavioral sequences during hierarchy formation in chickens. Science 216, 439–440. Chase, I. D. (1985). The sequential analysis of aggressive acts during hierarchy formation: An application of the ‘‘jigsaw’’ puzzle approach. Anim. Behav. 33, 86–100. Collett, M., Despland, E., Simpson, S. J., and Krakauer, D. C. (1998). The spatial scales of desert locust gregarisation. Proc. Natl. Acad. Sci. USA 95, 13052–13055. Couzin, I. D. (1999). ‘‘Collective Animal Behaviour.’’ Ph.D. thesis. University of Bath, Bath, UK.


69

Couzin, I. D., Krause, J., James, R., Ruxton, G. D., and Franks, N. R. (2002). Collective memory and spatial sorting in animal groups. J. Theor. Biol. 218, 1–11. Croft, D. P., Arrowsmith, B. J., Bielby, J., Skinner, K., White, E., Couzin, I. D., Magurran, A. E., Ramnarine, I., and Krause, J. (2003). Mechanisms underlying shoal composition in the Trinidadian guppy (Poecilia reticulata). Oikos (in press). Crofton, H. D. (1958). Nematode parasite populations in sheep on lowland pasture. VI. Sheep behaviour and nematode infections. Parasitology 48, 251–260. Cullen, J. M., Shaw, E., and Baldwin, H. A. (1965). Methods for measuring the threedimensional structure of fish schools. Anim. Behav. 13, 534–543. Cziro´k, A., Stanley, H. E., and Vicsek, T. (1997). Spontaneously ordered motion of selfpropelled particles. J. Phys. A 30, 1375–1385. Cziro´k, A., and Vicsek, T. (2001). Collective motion. In ‘‘Fluctuations and Scaling in Biology’’ (T. Vicsek, Ed.), pp. 177–242. Oxford University Press, Oxford. Cziro´k, A., Vicsek, M., and Vicsek, T. (1999). Collective motion of organisms in three dimensions. J. Phys. A 264, 299–304. Datta, S. B., and Beauchamp, G. (1991). Effects of group demography on dominance relationships among female primates. 1. Mother–daughter and sister–sister relations. Am. Nat. 138, 201–226. Davis, J. M. (1980). The coordinated aerobatics of dunlin flocks. Anim. Behav. 28, 668–673. DeBlois, E. M., and Rose, G. A (1996). Cross-shoal variability in the feeding habits of migrating Atlantic cod (Gadus morhua). Oecologia 108, 192–196. Deneubourg, J. L., Goss, S., Franks, N. R., and Pasteels, J. M. (1989). The blind leading the blind: Modelling chemically mediated army ant raid patterns. J. Insect Behav. 2, 719–725. Dill, L. M., Holling, C. S., and Palmer, L. H. (1997). Predicting the three-dimensional structure of animal aggregations from functional considerations: The role of information. In ‘‘Animal Groups in Three Dimensions’’ (J. K. Parrish and W. M. Hamner, Eds.), pp. 207–219. Cambridge University Press, Cambridge. Dizon, A. E., Stevens, E. D., Neill, W. H., and Magnusson, J. J. (1974). Sensitivity of restrained skipjack tuna (Katsuwonus pelamis) to abrupt increases in temperature. Comp. Biochem. Physiol. 49A, 199–291. Elliot, D., and Smith, D. (1993). Football stadia disasters in the United Kingdom: Learning from tragedy? Ind. Environ. Crisis Q. 7, 205–229. Flierl, G., Gru¨nbaum, D., Levin, S., and Olson, D. (1999). From individuals to aggregations: The interplay between behavior and physics. J. Theor. Biol. 196, 397–454. Francis, R. C. (1983). Experimental effects on agonstic behavior in the paradise fish, Macropodus opercularis. Behaviour 85, 292–313. Franks, N. R., Gomez, N., Goss, S., and Deneubourg, J.-L. (1991). The blind leading the blind in army ant raid patterns: Testing a model of self-organization (Hymenoptera: Formicidae). J. Insect Behav. 4, 583–607. Franks, N. R., and Scovell, E. (1983). Dominance and reproductive success among slavemaking worker ants. Nature 304, 724–725. Galef, B. G., Jr., and Buckley, L. L. (1996). Use of foraging trails by Norway rats. Anim. Behav. 52, 765–771. Gerhard, J., and Kirshner, M. (1997). ‘‘Cells, Embryos, and Evolution’’ Blackwell Science, Oxford. Ginsberg, B., and Allee, W. C. (1975). Some effects of conditioning on social dominance and subordination in inbred strains of mice. In ‘‘Social Hierarchy and Dominance’’ (M. W. Schein, Ed.). Dowden, Hutchinson, & Ross, Stroudsburg, PA.

70


Glazier, J. A., and Graner, F. (1993). Simulation of the differential adhesion driven rearrangement of biological cells. Phys. Rev. E 47, 2128–2154. Gosling, L. M., and Roberts, S. C. (2001). Scent marking by male mammals: Cheat-proof signals to competitors and mates. Adv. Study Behav. 30, 169–217. Gray, S. J., and Hurst, J. L. (1997). Behavioral mechanisms underlying the spatial dispersion of commensal Mus domesticus and grassland Mus spretus. Anim. Behav. 53, 511–524. Gru¨nbaum, D. (1997). Schooling as a strategy for taxis in a noisy environment. In ‘‘Animal Groups in Three Dimensions’’ (J. K. Parrish and W. M. Hamner, Eds.), pp. 257–281. Cambridge University Press, Cambridge. Gru¨nbaum, D. (1998). Schooling as a strategy for taxis in a noisy environment. Evol. Ecol. 12, 503–522. Gueron, S., and Levin, S. A. (1993). Self-organization of front patterns in large wildebeest herds. J. Theor. Biol. 165, 541–552. Gueron, S., and Levin, S. A. (1995). The dynamics of group formation. Math. Biosci. 128, 243–264. Gueron, S., Levin, S. A., and Rubenstein, D. I. (1996). The dynamics of herds: From individuals to aggregations. J. Theor. Biol. 182, 85–98. Hamilton, W. D. (1971). Geometry of the selfish herd. J. Theor. Biol. 31, 295–311. Heinze, J. (1990). Dominance behavior among ant females. Naturwissenschaften 77, 41–43. Helbing, D. (1992). A fluid-dynamic model for the movement of pedestrians. Complex Syst. 6, 391–415. Helbing, D., and Molna´r, P. (1995). Social force model for pedestrian dynamics. Phys. Rev. E 51, 4282–4286. Helbing, D., Farkas, I., and Vicsek, T. (2000a). Simulating dynamical features of escape panic. Nature 407, 487–490. Helbing, D., Farkas, I., and Vicsek, T. (2000b). Freezing by heating in a driven mesoscopic system. Phys. Rev. Lett. 84, 1240–1243. Helbing, D., Keltsch, J., and Molna´r, P. (1997b). Modelling the evolution of human trail systems. Nature 388, 47–50. Helbing, D., Schweitzer, F., Keltsch, J., and Molna´r, P. (1997a). Active walker model for the formation of human and animal trail systems. Phys. Rev. E 56, 2527–2539. Helfman, G. S. (1984). School fidelity in fishes: The yellow perch pattern. Anim. Behav. 32, 663–672. Hemelrijk, C. K. (1998). Spatial centrality of dominants without positional preference. In ‘‘Artificial Life,’’ Vol. 6 (C. Adam, R. Belew, H. Kitano, and C. Taylor, Eds.), pp. 307–315. MIT Press, Cambridge, MA. Hemelrijk, C. K. (1999). An individual-orientated model of the emergence of despotic and egalitarian societies. Proc. R. Soc. Lond. B Biol. Sci. 266, 361–369. Hemelrijk, C. K. (2000). Towards the integration of social dominance and spatial structure. Anim. Behav. 59, 1035–1048. Henderson, L. F. (1971). The statistics of crowd fluids. Nature 229, 381–383. Hensor, E. M. A., Hoare, D. J., Godin, J.-G. J., and Krause, J. (2003). Effects of nutritional state on the shoaling tendency of banded killifish (Fundulus diaphanus) in the field. Anim. Behav. (in press). Heppner, F. H. (1997). Three-dimensional structure and dynamics of bird flocks. In ‘‘Animal Groups in Three Dimensions’’ (J. K. Parrish and W. M. Hamner, Eds.), pp. 68–87. Cambridge University Press, Cambridge. Heppner, F. H., and Haffner, J. (1974). Communication in bird flocks: An electro-magnetic model. In ‘‘Biological and Clinical Effects of Low Intensity Magnetic Fields’’ (J. G. Llaurado, A. Sances, and J. H. Batto-Cletti, Eds.). Charles C Thomas, Springfield, IL.


71

Heppner, F. H., Convissar, J. L., Moonan, D. E., Jr., and Anderson, J. G. T (1985). Visual angle and formation flight in Canada geese (Branta canadesis). Auk. 102, 195–198. Herskin, J., and Steffensen, J. F. (1998). Energy savings in sea bass swimming in a school: Measurements of tail beat frequency and oxygen consumption at different swimming speeds. J. Fish Biol. 53, 366–376. Hilborn, R. (1991). Modelling the stability of fish schools: Exchange of individual fish between schools of skipjack tuna Katsuwonus pelamis. Can. J. Fish. Aquat. Sci. 48, 1081–1091. Hoare, D. J., Couzin, I. D., Godin, J.-G. J., and Krause, J. (2003). Context-dependent groupsize choice in fish. Anim. Behav. (in press). Hoare, D. J., Krause, J., Ruxton, G. D., and Godin, J.-G. J. (2000). The social organization of free-ranging fish shoals. Oikos 89, 546–554. Hogeweg, P., and Hesper, B. (1983). The ontogeny of the interaction structure in bumblebee colonies: A mirror model. Behav. Ecol. Sociobiol. 12, 271–283. Hogeweg, P., and Hesper, B. (1985). Socioinformation processes: MIRROR modelling methodology. J. Theor. Biol. 113, 311–330. Ho¨lldobler, B., and Wilson, E. O. (1990). ‘‘The Ants.’’ Springer-Verlag, Berlin. Hsu, Y., and Wolf, L. L. (1999). The winner and loser effect: Integrating multiple experiences. Anim. Behav. 57, 903–910. Huth, A., and Wissel, C. (1992). The simulation of the movement of fish shoals. J. Theor. Biol. 156, 365–385. Ikawa, T., and Okabe, H. (1997). Three-dimensional measurements of swarming mosquitoes: A probabilistic model, measuring system, and example results. In ‘‘Animal Groups in Three Dimensions’’ (J. K. Parrish and W. M. Hamner, Eds.), pp. 90–104. Cambridge University Press, Cambridge. Inada, Y., and Kawachi, K. (2002). Order and flexibility in the motion of fish schools. J. Theor. Biol. 214, 371–387. Judd, T., and Sherman, P. (1996). Naked mole rats recruit colony mates to food source. Anim. Behav. 52, 957–969. Karavanich, C., and Atima, J. (1998). Individual recognition and memory in lobster dominance. Anim. Behav. 56, 1553–1560. Keller, E. F., and Segel, L. A. (1971). Model for chemotaxis. J. Theor. Biol. 30, 225–234. Kelley, H. H., Condry, J. C., Jr., Dahlke, A. E., and Hill, A. H. (1965). Collective behavior in a simulated panic situation. J. Exp. Social Psychol. 1, 20–54. Keynes, R. J., and Stern, C. D. (1988). Mechanisms of vertebrate segmentation. Development 103, 413–429. Kils, U. (1986). Verhaltensphysiologische Untersuchungen an pelagischen Schwa¨rmen, Schwarmbildung als Strategie zur Orientierung in Umweltgradienten, Bedeutung der Schwarmbildung in der Aquakultur. Ber. Inst. Meeresk. Kiel. 163, 1–168; as cited in Pitcher, T. J., and Parrish, J. K. (1986). Functions of shoaling behavior in teleosts. In ‘‘Behaviour of Teleost Fishes’’ (T. J. Pitcher, Ed.), pp. 416–418. Chapman & Hall, London. Klimley, A. P., and Holloway, C. F. (1999). School fidelity and homing synchronicity of yellowfin tuna. Marine Biol. 133, 307–317. Krause, J. (1993a). The relationship between foraging and shoal position in a mixed shoal of roach (Rutilus rutilus) and chub (Leuciscus leuciscus): A field study. Oecologia 93, 356–359. Krause, J. (1993b). The effect of ‘Schreckstoff’ on the shoaling behaviour of the minnow: A test of Hamilton’s selfish herd theory. Anim. Behav. 45, 1019–1024. Krause, J. (1994). Differential fitness returns in relation to spatial positions in groups. Biol. Rev. 69, 187–206.

72


Krause, J., and Godin, J.-G. J. (1994). The influence of parasitism on the shoaling behavior of the killifish (Fundulus diaphanus). Can. J. Zool. 72, 1775–1779. Krause, J., and Ruxton, G. D. (2002). ‘‘Living in Groups.’’ Oxford University Press, Oxford. Krause, J., Butlin, R. K., Peuhkuri, N., and Pritchard, V. L. (2000b). The social organization of fish shoals: A test of the predictive power of laboratory experiments for the field. Biol. Rev. 75, 477–501. Krause, J., Godin, J.-G. J., and Brown, D. (1996). Phenotypic variability within and between fish shoals. Ecology 77, 1586–1591. Krause, J., Hoare, D. J., Krause, S., Hemelrijk, C. K., and Rubenstein, D. I. (2000a). Leadership in fish shoals. Fish Fish. 1, 82–89. Krause, J., Hoare, D. J., Croft, D., Lawrence, J., Ward, A. J. W., James, R., Ruxton, G. D., and Godin, J.-G. J. (2000c). Fish shoal composition: Mechanisms and constraints. Proc. R. Soc. Lond. B Biol. Sci. 267, 2011–2017. Kummer, H. (1968). ‘‘Social Organization of Hamadrayas Baboons.’’ University of Chicago Press, Chicago, IL. Landeau, L., and Terborgh, J. (1986). Oddity and the confusion effect in predation. Anim. Behav. 34, 1372–1380. Levin, S. A. (1999). ‘‘Fragile Dominion: Complexity and the Commons.’’ Perseus Books, Cambridge, MA. MacDonald, D. (2001). ‘‘The New Encyclopedia of Mammals.’’ Oxford University Press, Oxford. Magurran, A. E., Seghers, B. H., Shaw, P. W., and Carvalho, G. R. (1995). The behavioral diversity and evolution of guppy, Poecilia reticulata, populations in Trinidad. Adv. Study Behav. 24, 155–202. Maini, P. K., and Othmer, H. G. (2000). ‘‘Mathematical Models for Biological Pattern Formation.’’ Springer-Verlag, New York. Major, P. F., and Dill, L. M. (1978). The three-dimensional structure of airborne bird flocks. Behav. Ecol. Sociobiol. 4, 111–122. McFarland, W., and Okubo, A. (1997). Metabolic models of fish school behaviour: The need for quantitative observations. In ‘‘Animal Groups in Three Dimensions’’ (J. K. Parrish and W. M. Hamner, Eds.), pp. 301–312. Cambridge University Press, Cambridge. Mendoza, S. D., and Barchas, P. R. (1983). Behavioral processes leading to linear status hierarchies following group formation in rhesus macaques. J. Hum. Evol. 12, 185–192. Milgram, S., and Toch, H. (1969). Collective behavior: Crowds and social movements. In ‘‘The Handbook of Social Psychology’’ (G. Lindzey and E. Aronson, Eds.). Addison-Wesley, Reading, MA. Milinski, M. (1977). Experiments on the selection by predators against spatial oddity of their prey. Z. Tierpsychol. 43, 311–325. Moorcroft, P., Lewis, M., and Crabtree, R. (1999). Analysis of coyote home ranges using a mechanistic home range model. Ecology 80, 1656–1665. Murray, J. D. (1981). A prepattern formation mechanism for animal coat markings. J. Theor. Biol. 88, 161–199. Neill, W. H., Chang, R. K. C., and Dizon, A. E. (1976). Magnitude and ecological implications of thermal inertia in skipjack tuna, Katsuwonus pelamis (Linnaeus). Environ. Biol. Fish. 1, 61–80. Nicolis, G., and Prigogine, I. (1977). ‘‘Self-Organization in Non-Equilibrium Systems.’’ John Wiley & Sons, New York. Niwa, H. S. (1994). Self-organizing dynamic model of fish schooling. J. Theor. Biol. 171, 123–136. Niwa, H. S. (1996). Newtonian dynamical approach to fish schooling. J. Theor. Biol. 181, 47–63.


73

Niwa, H.-S. (1998a). Migration dynamics of fish schools in heterothermal environments. J. Theor. Biol. 193, 215–231. Niwa, H.-S. (1998b). School size statistics of fish. J. Theor. Biol. 195, 351–361. Okubo, A. (1980). Vol. 10: ‘‘Diffusion and Ecological Problems: Mathematical Models.’’ ‘‘Lecture Notes in Biomechanics.’’ Springer-Verlag, New York. Okubo, A. (1986). Dynamic aspects of animal grouping: Swarms, schools, flocks, and herds. Adv. Biophys. 22, 1–94. Okubo, A., and Chiang, H. C. (1974). An analysis of the kinematics of swarming behavior of Anarete pritchardi Kim (Diptera: Cecidomyiidae). Res. Popul. Ecol. 16, 1–42. Oliveira, P. S., and Ho¨ lldobler, B. (1990). Dominance orders in the ponerine ant Pachycondyla apicalis (Hymenoptera, Formicidae). Behav. Ecol. Sociobiol. 27, 385–393. Osborn, J. (1997). Analytical and digital photogrammetry. In ‘‘Animal Groups in Three Dimensions’’ (J. K. Parrish and W. M. Hamner, Eds.), pp. 194–206. Cambridge University Press, Cambridge. Parrish, J. K. (1989a). Layering with depth in a heterospecific fish aggregation. Environ. Biol. Fishes. 26, 79–85. Parrish, J. K. (1989b). Re-examining the selfish herd: Are central fish safer? Anim. Behav. 38, 1048–1053. Parrish, J. K., and Edelstein-Keshet, L. (1999). Complexity, pattern, and evolutionary tradeoffs in animal aggregation. Science 284, 99–101. Parrish, J. K., and Hamner, W. M. (1997). ‘‘Animal Groups in Three Dimensions.’’ Cambridge University Press, Cambridge. Parrish, J. K., and Turchin, P. (1997). Individual decisions, traffic rules, and emergent pattern in schooling fish. In ‘‘Animal Groups in Three Dimensions’’ (J. K. Parrish and W. M. Hamner, Eds.), pp. 194–206. Cambridge University Press, Cambridge. Partridge, B. L. (1980). The effect of school size on the structure and dynamics of minnow schools. Anim. Behav. 28, 68–77. Partridge, B. L. (1982). Structure and function of fish schools. Sci. Am. 245, 114–123. Partridge, B. L., and Pitcher, T. J. (1979). Evidence against a hydrodynamic function for fish schools. Nature 279, 418–419. Partridge, B. L., and Pitcher, T. J. (1980). The sensory basis of fish schools: Relative roles of lateral line and vision. J. Comp. Physiol. 135, 315–325. Partridge, B. L., Johansson, J., and Kalisk, J. (1983). Structure of schools of giant bluefin tuna in Cape Cod Bay. Environ. Biol. Fishes. 9, 253–262. Partridge, B. L., Pitcher, T. J., Cullen, J. M., and Wilson, J. (1980). The 3-dimensional structure of fish schools. Behav. Ecol. Sociobiol. 6, 277–288. Pitcher, T. J., and Parrish, J. K. (1993). Functions of shoaling behaviour in teleosts. In ‘‘Behaviour of Teleost Fishes’’ (T. J. Pitcher, Ed.), pp. 363–439. Chapman & Hall, London. Pitcher, T. J., Magurran, A. E., and Edwards, J. I. (1985). Schooling mackerel and herring choose neighbors of similar size. Mar. Biol. 86, 319–322. Pitcher, T. J., Magurran, A. E., and Allan, J. R. (1986). Size-segregative behavior in minnow shoals. J. Fish Biol. 29, 83–95. Pomeroy, H., and Heppner, F. (1992). Structure of turning in airborne rock dove (Columba livia) flocks. Auk. 109, 256–261. Potts, W. K. (1984). The chorus line hypothesis of manoeuvre coordination in avian flocks. Nature 309, 344–345. Presman, A. S. (1970). ‘‘Electromagnetic Fields and Life.’’ Plenum, New York.

74


Pulliam, H. R., and Caraco, T. (1984). Living in groups: Is there an optimal group size? In ‘‘Behavioural Ecology: An Evolutionary Approach’’ (J. R. Krebs and N. B. Davies, Eds.), pp. 122–147. Blackwell Scientific, Oxford. Radakov, D. (1973). ‘‘Schooling and Ecology of Fish.’’ John Wiley & Sons, New York. Ranta, E., Peuhkuri, N., and Laurila, A. (1994). A theoretical exploration of antipredatory and foraging factors promoting phenotype-assorted fish schools. Ecoscience 1, 99–106. Reynolds, C. W. (1987). Flocks, herds, and schools: A distributed behavioral model. Comput. Graphics 21, 25–33. Ritz, D. A. (1997). Costs and benefits as a function of group size: Experiments on a swarming mysid Paramesopodopsis rufa Fenton. In ‘‘Animal Groups in Three Dimensions’’ (J. K. Parrish and W. M. Hamner, Eds.), pp. 194–206. Cambridge University Press, Cambridge. Roditakis, E., Couzin, I. D., Baldrow, K., Franks, N. R., and Charnley, A. K. (2000). Improving secondary pick up of insect fungal pathogen conidia by manipulating host behavior. Ann. Appl. Biol. 137, 329–335. Schjelderupp-Ebbe, T. (1913). Ho¨nsenses stemme: Bidrag til ho¨nsenses psykologi. Naturen 37, 262–276. Schjelderupp-Ebbe, T. (1922). Beitra¨ge zur Sozialpsychologie des Haushuhns. Z. Psychol. 88, 225–252. Seeley, T. D. (1995). ‘‘The Wisdom of the Hive.’’ Harvard University Press, Cambridge, MA. Seghers, B. H. (1981). Facultative schooling behavior in the spottail shiner (Notropis hudsonius): Possible costs and benefits. Environ. Biol. Fishes. 6, 21–24. Selous, E. (1931). ‘‘Thought-Transference (or What?) in Birds.’’ Constable, London. Sibly, R. M. (1983). Optimal group size is unstable. Anim. Behav. 31, 947–948. Sinclair, A. R. E. (1977). ‘‘The African Buffalo: A Study of Resource Limitation of Populations.’’ University of Chicago Press, Chicago, IL. Sjo¨berg, M., Albrectsen, B., and Hja¨lteń, J. (2000). Truncated power laws: A tool for understanding aggregation patterns in animals? Ecol. Lett. 3, 90–94. Slater, P. J. B. (1986). Individual differences and dominance hierarchies. Anim. Behav. 34, 1264–1265. Takayasu, H. (1989). Steady-state distribution of generalized aggregation systems with injection. Phys. Rev. Lett. 63, 2563–2565. Theodorakis, C. W. (1989). Size segregation and the effects of oddity on predation risk in minnow shoals. Anim. Behav. 38, 496–502. Theraulaz, G., and Bonabeau, E. (1995). Coordination in distributed building. Science 269, 686–688. Theraulaz, G., Bonabeau, E., and Deneubourg, J. L. (1995). Self-organization of hierarchies in animal societies: The case of the primitively eusocial wasp Polistes dominulus (C.). J. Theor. Biol. 174, 313–323. Theraulaz, G., Gervet, J., Thon, B., Pratte, M., and Semenoff-Tian-Chansky, S. (1992). The dynamics of colony organization in the primitively eusocial wasp Polistes dominulus (C.). Ethology 91, 177–202. Theraulaz, G., Pratte, M., and Gervet, J. (1989). Effects of removal of -individuals from a Polistes dominulus (C.) wasp society: Changes in behavioural patterns resulting from hierarchical changes. Actes Colloq. Insectes Sociaux 5, 169–179. Thierry, B. (1985). Patterns of agonistic interactions in three species of macaques (Macaca mulatta, M. fascicularis, M. tonkeana). Aggressive Behav. 11, 223–233. Tibbetts, E. A. (2002). Visual signals of individual identity in the wasp Polistes fuscatus. Proc. R. Soc. Lond. B Biol. Sci. 269, 1423–1428.


75

Tranquillo, R. T. (1990). Models of Chemical Gradient Sensing by Cells. In ‘‘Lecture Notes in Biomathematics’’: Vol. 89: ‘‘Biological Motion.’’ (W. Alt and G. Hoffman, Eds.), pp. 415–441. Springer-Verlag, New York. Tyler, S. J. (1972). The behaviour and social organization of the new forest ponies. Anim. Behav. Monogr. 48, 223–233. Van Havre, N., and FitzGerald, G. J. (1988). Shoaling and kin recognition in the threespine stickleback (Gasterosteus aculeatus L.). Biol. Behav. 13, 190–201. Vicsek, T., Cziro´k, A., Ben-Jacob, E., Cohen, I., and Shochet, O. (1995). Novel type of phase transition in a system of self-driven particles. Phys. Rev. Lett. 75, 1226–1229. Viitala, J., Korpima¨ki, E., Palokangas, P., and Koivula, M. (1995). Attraction of kestrels to vole scent marks visible in ultraviolet light. Nature. 373, 425–427. Wallace, D. G., Gorny, B., and Whishaw, I. Q. (2002). Rats can track odors, other rats, and themselves: Implications for the study of spatial behavior. Behav. Brain Res. 131, 185–192. Ward, A. J. W., Hoare, D. J., Couzin, I. D., Broom, M., and Krause, J. (2002). The effects of parasitism and body length on the positioning within wild fish shoals. J. Anim. Ecol. 71, 10–14. Wieser, W., Forstner, H., Medgyesy, N., and Hinterleitner, S. (1988). To switch or not to switch: Partitioning of energy between growth and activity in larval cyprinids (Cyprinidae: Teleostei). Funct. Ecol. 2, 499–507. Wilson, E. O. (1975). ‘‘Sociobiology.’’ Harvard University Press, Boston. Wolf, N. (1985). Odd fish abandon mixed-species groups when threatened. Behav. Ecol. Sociobiol. 17, 47–52.

This Page Intentionally Left Blank


Odor–Genes Covariance and Genetic Relatedness Assessments: Rethinking Odor-Based ‘‘Recognition’’ Mechanisms in Rodents Josephine Todrank and Giora Heth institute of evolution university of haifa haifa 31905, israel

I. Introduction The theoretical evolutionary benefits of discriminative responses toward kin and nonkin that motivated Hamilton’s ideas (1964a,b) about kin selection and inclusive fitness as the evolutionary basis of social behavior were sufficiently appealing to inspire several decades of research on kin recognition (see reviews by Fletcher and Michener, 1987; Waldman et al., 1988; Barnard, 1990; Barnard et al., 1991; Hepper, 1991). The adaptive value of discriminative responses toward kin and nonkin led to much speculation about possible mechanisms to enable it (e.g., Holmes and Sherman, 1983; Sherman and Holmes, 1985; Hepper, 1986, 1991; Waldman, 1987; Halpin, 1991; Sherman et al., 1997; Tang-Martinez, 2001). Much ambiguity remains, however, about how kin recognition mechanisms actually work because the supportive evidence is far from unequivocal (see review by Tang-Martinez, 2001). Clearly, being able to discriminate effectively between kin and nonkin can add substantially to the adaptiveness of individual mate choices and to the selective targeting of altruistic behavior. This in turn could have a crucial impact on their inclusive fitness (e.g., Slater, 1994; Barnard et al., 1991). Perhaps other comparable types of discrimination, such as those between individuals from neighboring populations or closely related species, may be based on similar types of cues. This could also have a significant impact on adaptiveness of social behavior. ‘‘Kin’’ are, indeed, closely related genetically and share a high degree of similarity in their genome, but ‘‘nonkin’’ are not genetically unrelated. 77 Copyright 2003 Elsevier Science (USA). All rights reserved. 0065-3454/03 $35.00

78

JOSEPHINE TODRANK AND GIORA HETH

Nonkin are simply less closely related, that is, less genetically similar, by virtue of their having fewer genes in common. Siblings, cousins, conspecifics from other families in the population or other populations within the species, and heterospecifics from closely related species are individuals that represent a spectrum of relatedness. They have different degrees of similarity in their genome. Studying discriminations between kin and nonkin may have excluded equally important behavioral distinctions between individuals of differing degrees of relatedness. Sometimes the theoretical grounding of a hypothesis seems so solid and the putative mechanisms underlying it seem so reasonable that these ideas channel the thinking of both researchers and theoreticians in ways that ultimately complicate the understanding of the wider picture. Thinking based on categorical distinctions, such as between ‘‘kin’’ and ‘‘nonkin’’ or even between ‘‘conspecifics’’ and ‘‘heterospecifics,’’ that are both obvious and meaningful to the researcher may misrepresent the perceptual experience of the animal during interactions in its natural social world. Animal social behavior may involve much more subtle discriminations across a spectrum of genetic similarity rather than classification into a few distinct categories. In addition, reasonable questions have been raised about the discriminatory processes, such as Grafen’s query (1990) about whether animals really recognize kin rather than groups or species. These doubts are widely cited but largely dismissed in terms of guiding experimental designs (see Komdeur and Hatchwell, 1999). In our view, there are questions not only about whether it is kinship that is recognized but also about whether the process necessarily involves ‘‘recognition’’ at all. In the case of kin recognition, focusing on mechanisms that ‘‘should’’ have evolved specifically for discrimination between kin and nonkin may have misdirected some previous research efforts and constrained both the experimental designs used and the interpretations of the data collected. Tang-Martinez (2001) addressed much of the lingering controversy in a critical reevaluation of the mechanisms of kin discrimination and suggested dispensing with two of the four most widely proposed mechanisms. Spatial distribution may be a reliable indicator of relatedness in some social systems, and it may function as a means of differential responses toward kin and nonkin, but no one would contend that it represents a true means of actually recognizing kinship (Halpin, 1991; Tang-Martinez, 2001). Efforts to demonstrate ‘‘recognition alleles,’’ hypothesized by Hamilton (1964a,b), have produced equivocal results and have also been plagued by methodological flaws (see Tang-Martinez, 2001). The theoretical basis of this particular mechanism could also be suspect because of the difficulty of distributing alleles in a way that could code successfully for relatedness among both maternal and paternal cousins that originate from different

ODOR–GENES COVARIANCE AND GENETIC RELATEDNESS ASSESSMENTS

79

genetic lines. Even if there were such alleles available in sufficient unique combinations to identify particular immediate families effectively, a mechanism that is suitable only for identifying immediate forbears, full siblings, and direct descendants would be of limited usefulness. TangMartinez (2001), like Porter (1988) previously, suggests relinquishing the distinctions between the other two proposed mechanisms (recognition by association or familiarity and recognition by phenotype matching) because of apparent overlaps (in ‘‘learning’’ and ‘‘matching’’) between them and argues for a single ‘‘genuine’’ mechanism based on learning through familiarity. Evidence, some of which has been accumulated since the review by Tang-Martinez (2001), has substantially changed our way of thinking about recognition systems in general and kin recognition in particular. This evidence, which comes from studies of differential responses to odors in various rodent species, suggests that there are no odor-based mechanisms that have evolved specifically for recognition of kin vs nonkin. Rather, there are two mechanisms underlying discriminative responses based on individual genotypes: ‘‘individual recognition through association’’ enables true recognition of specific familiar individuals and their odors, irrespective of relatedness, and ‘‘genetic relatedness assessment through individual odor similarities’’ (G-ratios) enables differential treatment based on degrees of genetic relatedness throughout the spectrum from siblings to across species, irrespective of prior association. This chapter provides a summary of the findings that led to our reevaluation of mechanisms underlying differential responses to odors (including a description of the methodologies used to collect the data), suggests new ways of interpreting the terms used in describing discriminative responses to odors, proposes a hypothesis about the origin and evolutionary basis of differential responses based on genetic relatedness assessments, and details prospects for future studies that could deepen the understanding of how these mechanisms work. Although we deal exclusively with data based on the responses of rodents to odors, the principles explored in the experiments summarized here could apply equally well to other taxa and to sensory modalities other than olfaction, particularly vision and audition. To the extent that variability among individual genotypes is expressed as variability in phenotypic characteristics, be they individual odor, facial, or vocal characteristics, similarities in these traits across individuals are indicative of genetic similarity between them. To the extent that animals are able to detect such phenotypic differences and respond differentially to them, animals are responding on the basis of genetic similarity assessments.

80


II. Odor–Genes Covariance Sometimes experimental evidence not only speaks to the theoretical model that motivated the investigation but also inadvertently points to a new way of thinking about the research area. Such was the case with our efforts to extend the growing evidence of chemical communication in subterranean blind mole rats of the superspecies Spalax ehrenbergi (Heth and Todrank, 1995; Heth et al., 1996a,b; Todrank and Heth, 1996; Nevo et al., 2001) to an investigation of odor-based kin recognition. Like many other solitary subterranean rodent species, blind mole rats do not breed in captivity and thus all animals to be studied must be captured in the field. Because drawing blood without killing the animal is difficult, there is no easy way to determine the genetic relatedness between captured animals. It was known, however, that blind mole rats, like other rodents, produce individually distinctive chemosensory cues in their urine (Heth et al., 1996a). It was also known that urine of blind mole rats from the same population has distinctive qualities that are typical of that population and mole rats from the same species have urine odors that are distinctive for that particular species (Heth et al., 1996a). We hypothesized that close relatives may have distinctive qualities in their urine odors that are typical of their particular family. If this were true, we speculated that a mole rat may treat the odors of two closer relatives as similar in quality when compared with the odor of a third, less closely related individual. We thought of using this hypothetical odor similarity as a way to assess relatedness among mole rats collected from the field. Thus, we developed a variation on the standard habituation–discrimination technique [which was designed to assess discriminability of odors (Halpin, 1986)] that was designed, instead, to assess odor similarities. If the chemosensory cues are in the urine (which is easy to collect) and hence the hypothesized odor similarity makes sense, a question may arise concerning why it would not make more sense to analyze the urine than to go to the trouble of developing a new behavioral test. It is certainly true that analytic chemistry techniques enable accurate quantification of the chemical constituents even of complicated mixtures of compounds such as those in urine. The difficulty arises in trying to translate the chemical analysis into an understanding of how the odor is actually perceived (see also Section II.F). There are a wide variety of problems to overcome. Only some of the chemicals will have odors. Even having identified which components are odorous, the amount of any particular odorant does not indicate reliably how it will contribute to the overall quality of the odor: the quality of the particular component can change substantially depending on the concentration, and variations in proportions of chemicals


81

in mixtures can change the way the various odorants interact to produce the perceptual quality. The presence of a small amount of some odorants is sufficient to mask the perception of other odorants that are present in much higher concentrations. Given these complications, researchers must rely on carefully designed behavioral tests to understand the information in odors as well as how animals perceive and respond to them. As is described later, follow-up experiments using the habituation– generalization technique led, serendipitously, to theorizing about odor– genes covariance and the mechanisms for differential responses it implies.

A. Habituation–Discrimination Technique In a standard habituation–discrimination procedure, the subject is presented with an odor (termed the ‘‘habituation odor’’) either for an extended period of time or over repeated trials (see Halpin, 1986; Johnston et al., 1993). During this period of extended or repeated exposure, the subject’s interest in the odor decreases through the process of habituation. In the test phase, which follows the habituation phase, a second odor (termed the ‘‘test odor’’) is introduced. In some procedures the test odor and the habituation odor are presented during a single test trial and the time spent investigating each of the odors is measured. If the subject perceives the test odor as different from the habituation odor, the subject spends more time investigating the test odor and statistical tests comparing the two investigation times are significant. In other procedures the test odor is presented alone during the test trial and the time spent investigating the test odor is compared with the time spent investigating the habituation odor in the last habituation trial. Again, statistically significant increases in the investigation time indicate that the subject perceived the test odor as different from the habituation odor. Researchers who are unfamiliar with these techniques may underestimate the appeal of ‘‘novelty following boredom,’’ but for a rodent in the laboratory the dishabituation is often substantial. These methods (irrespective of the number or length of the habituation trials or the variety of odors used) provide consistent results that demonstrate reliably when subjects respond spontaneously to the differences between the two odors. (Of course, failure to respond to the differences does not mean that the odors are not discriminable, only that the subject does not respond to the differences in this context. But these tests are typically conducted with discriminable odors, and subjects typically demonstrate that they respond to the differences.) These tests are designed solely to demonstrate the ability of animals to discriminate spontaneously between odors with different

82


qualities; they do not enable other types of inferences, such as conclusions about the preferences of subjects, from their behavior during such tests. B. Habituation–Generalization Technique In the habituation–generalization variation on this technique (Todrank et al., 1998), the habituation procedure is the same, but during the test phase two test odors are presented rather than one. Although originally conceived as a method to assess similarity of odors from individuals whose relatedness was unknown, the technique was developed in tests with odor donors of known relatedness. When two test odors are used, one of the test odors is usually from an individual known to be more closely related to the habituation odor donor. This test odor should thus be more likely to have a quality similar to the habituation odor. The other test odor is usually from a less genetically similar donor. These two test odors can be presented either together in a single test trial or separately in two sequential test trials. When the test odors are presented together, significant differences in the time spent investigating the odors indicate that the two test odors are discriminated as different from one another and that the less investigated odor is treated as more similar in quality to the habituation odor. (The two test odors are always presented in a single test trial when the relatedness of the odor donors is not known.) When the two test odors are presented on separate test trials, the odor that is more likely to be perceived as similar to the habituation odor is typically presented on the first test trial and the other odor is presented second. This order is used to ensure that the subject has an opportunity to compare the two similar odors sequentially without the disruption of a dissimilar odor in between. This maximizes the subject’s chances of discriminating between the similar odors. In some cases, the odors are sufficiently similar that subjects do not discriminate between them, sometimes even spending less time investigating the first test odor than the habituation odor on the last habituation trial. It would be easier to mistake a similar odor for the habituation odor if a different odor is presented between them. Thus presenting the dissimilar odor first could undermine confidence in the results in terms of the ability of subjects to discriminate between similar odors. When the test odors are presented sequentially, the investigation time for the first test odor can be compared with the time spent investigating the habituation odor on the last habituation trial, to determine whether the subject discriminates between the two similar odors (Fig. 1A) or not (Fig. 1B). If subjects spend significantly more time investigating the second test odor than the first test odor, this indicates that the two test odors are discriminated as different from one another and that the first test odor is


83

Fig. 1. Schematic diagram depicting the mean ( SE) time, during different trials of an habituation–generalization experiment, that animals spend investigating the odor from the habituation odor donor (light gray columns) and then the odors from two other donors, one that is more genetically similar (medium gray columns; test 1) to the habituation odor donor and another that is less genetically similar (dark gray columns; test 2). In (A), a significant increase in investigation time in test 1 compared with the third habituation trial indicates that subjects discriminated between the two similar odors; a significant increase in investigation time in test 2 compared with test 1 indicates that subjects treated the test 1 odor as similar to the habituation odor. In (B), no significant increase in investigation time in test 1 compared with the third habituation trial indicates that subjects generalized between the two similar odors; a significant increase in investigation time in test 2 compared with test 1 indicates that subjects discriminated between the two dissimilar odors, demonstrating further that subjects treated the test 1 odor as similar to the habituation odor.

treated as more similar in quality to the habituation odor. This habituation– generalization variation is based on the principle that the more similar the test odor is to the habituation odor, the less likely it is that an increase in investigation would be observed, because more similar odors are less likely to be discriminated as different. It was the application of the same principle in previous research on infant visual attention and cognition (see, e.g., Cohen, 1976; Spelke, 1985) that inspired this habituation–generalization technique using odors. In addition, after the subject has been exposed to the habituation odor for a period of time, the more different a test odor is from the habituation odor the greater the response to that test odor is likely to be, and thus it is reasonable to expect that the subjects would spend significantly more time investigating the more different test odor (see Fig. 2). [Although the same principle would apply with other types of odors, such as food odors, thus far this procedure has been used exclusively for assessing similarities in the qualities of odors from individual animals. Because the assessment of interest concerns individual odor similarities that depend on genotypic similarities, other factors that may affect the

84


Fig. 2. Schematic diagram depicting odors of different individuals (A, B, C, D, E, and F) as variations in proportions of different compounds, as if all compounds contributed equally to the overall odor gestalt. The odors of A and B would be more similar than any of the other pairs because they have the largest area of overlap. The odors of C or D would be treated as similar to A or B in comparison with E or F because C and D overlap more with A and B than do E and F.

composite quality of odor, such as diet (e.g., Schellinck and Brown, 1999) and reproductive status (e.g., Heth et al., 1996b), are carefully controlled.] Although when the test odors are presented sequentially the observer knows that the first test odor ‘‘should’’ be perceived as more similar to the habituation odor, observer bias is unlikely to be an issue in these experiments for several reasons. The pattern of investigation is consistent across subjects, but there is often more variability in investigation times across subjects than across trials for a particular subject. Because, to be efficient, several subjects are typically tested during a single session, with one subject being tested during the intertrial interval of the others, it would be difficult to remember the investigation time of any particular subject in a previous trial and to know how to subtly bias the direction of the subsequent results. Furthermore, we obtained the same pattern of subtle differences (see Section III.A) between tests with familiar odor donors (discrimination) and unfamiliar odor donors (generalization) before we knew what pattern to expect. The habituation–generalization procedure cannot provide direct information about the chemical structure or composition of the odorants, but it does provide useful information about how animals perceive the qualities of odors in terms of the similarities and the discriminable differences between them. The habituation–generalization procedure was designed solely to assess the spontaneous responses of subjects to odors with differing degrees of similarity without the expectation that the results would provide other insights into odor-based social behavior. It is important to distinguish between differential investigation in the context of habituation experiments and in two-choice tests designed to assess differential interest. In habituation–generalization experiments, when two test odors are presented simultaneously in the test trial, differential investigation in this context does not signify a preference. Although it is theoretically clear that the experience of habituation alters


85

what may otherwise be shown as a preference, this may seem counterintuitive to those who are less familiar with the technique, but it can be verified experimentally by testing the same group of animals with the same pairs of odor categories, using both procedures. Using the habituation– generalization technique, mice tested with odors of conspecifics and heterospecifics after habituating to the odor of a conspecific spent more time investigating the odor of the heterospecific in the test trial (because it was less similar rather than because they preferred it). When given a choice between odors from a conspecific and a heterospecific without prior habituation, the same mice spent more time investigating the conspecific odor, indicating greater interest in the conspecific odor (G. Heth, J. Todrank, N. Busquet, and C. Baudoin, unpublished data). C. Evidence of Odor Similarities Based on Kinship In pilot testing of the habituation–generalization technique, blind mole rats habituated to the urine odor from a same-sex conspecific and were then tested with the odors of same-sex conspecifics and heterospecifics: males and females treated the odor of the second conspecific as similar to the habituation odor and the odor of the heterospecific as different (G. Heth and J. Todrank, unpublished data). The technique was further developed in experiments with a solitary, semifossorial rodent species, the golden hamster, Mesocricetus auratus (Todrank et al., 1998, 1999). In the experiments with hamsters, many tests were conducted and then repeated to assess similarities between the qualities of flank gland odors from siblings and half-siblings in comparison with the flank gland odors of hamsters that were not related to the other two odor donors. In every experiment, whether the test odors were presented together or in separate trials, the hamsters consistently investigated the odor of the nonsibling longer in the test trial than the odor of the sibling or half-sibling of the habituation odor donor (see Table I). This was true whether the odor donors were the subjects’ own siblings or pairs of siblings from another family; the results were also consistent whether the subjects were familiar with the odor donors or not (Todrank et al., 1998, 1999). Similar results were also found with Turkish hamsters, Mesocricetus brandti (Heth et al., 1999). The responses of hamsters to the odors of sibling and half-sibling hamsters suggested that the subjects perceived these odors as having similar qualities and thus provided evidence of odor similarity based on kinship. Subsequently similar tests (Heth et al., 2002), conducted with different colonies of giant mole rats (Cryptomys mechowi), a eusocial, subterranean rodent species (Burda, 1999), provided additional support for the odor

TABLE I Odor–Genes Covariance Dataa Genetic similarity

Subject species

Odor donorsb

Habituationc

Test 1

Test 2

Ref.

Mesocricetus auratus Mesocricetus auratus Mesocricetus auratus Mesocricetus auratus Mesocricetus auratus Mesocricetus auratus Mesocricetus auratus Mesocricetus auratus Mesocricetus auratus Mesocricetus auratus Mesocricetus auratus Mesocricetus brandti Mesocricetus brandti Cryptomys mechowi Cryptomys mechowi

Familiar sibs—males (1A) Familiar sibs—females (1A) Unfamiliar sibs—males (1B) Unfamiliar sibs—females (1B) Unfamiliar half-sibs—males (1B) Familiar sibs—males (1A) Unfamiliar subject’s sibs—males (1B) Familiar foster sibs—males (1A) Unfamiliar sibs—males (1B) Unfamiliar sibs—males (1B) Familiar sibs—males (1A) Familiar sibs—males (1A) Unfamiliar sibs—males (1B) Familiar sibs—females or males (1A) Unfamiliar sibs—females or males (1A)

22, 13, 8, 3 19, 12, 7, 5 14, 8, 6, 4 18, 12, 7, 6 8, 5, 4, 2 12, 7, 4, 2 26, 11, 6, 5 22, 8, 6, 3 17, 9, 8, 6 11, 6, 4, 5 11, 6, 6, 2 13, 4, 3, 2 17, 10, 3, 4 12, 5, 2 12, 8, 6

7 8 5 5 3 4 3 7 3 3 4 6 2 5 4

12 13 10 11 6 10 8 10 8 7 6 7 6 12 22

Todrank et al., 1998 Todrank et al., 1998 Todrank et al., 1998 Todrank et al., 1998 Todrank et al., 1998 Todrank et al., 1999 Todrank et al., 1999 Todrank et al., 1999 Todrank et al., 1999 Todrank et al., 1999 Todrank et al., 1999 Heth et al., 1998 Heth et al., 1998 Heth et al., 2002 Heth et al., 2002

Population species

Mus spicilegus Mesocricetus auratus Mesocricetus brandti

Unfamiliar—males Unfamiliar—males Unfamiliar—males

>20 13, 9, 6, 4 12, 10, 5, 3

17 4 2

26 19 8

Heth et al., 2001 Heth et al., 1999 Heth et al., 1999

>40 >40 >40 >40 >40 >40

13 13 14 15 12 20

22 24 22 31 25 29

Heth Heth Heth Heth Heth Heth

86

Kin group

Across species Spalax Spalax Spalax Spalax Spalax Spalax

ehrenbergi ehrenbergi ehrenbergi ehrenbergi ehrenbergi ehrenbergi

Unfamiliar—females Unfamiliar—females Unfamiliar—females Unfamiliar—females Unfamiliar—females Unfamiliar—females

or or or or or or

males males males males males males

and and and and and and

Todrank, Todrank, Todrank, Todrank, Todrank, Todrank,

2000 2000 2000 2000 2000 2000

87

Cryptomys anselli Cryptomys mechowi Cryptomys kafuensis Mus spicilegus, M. macedonicus Mus musculus, M. domesticus a

Unfamiliar—females Unfamiliar—females Unfamiliar—females Unfamiliar—females or males

>20 >20 >20 >20 >20

Unfamiliar—females or males

Mean investigation time rounded to full seconds. 1A, see Fig. 1A; 1B, see Fig. 1B. c Times over repeated trials (separated by commas) or during a single extended trial. b

6 6 7 15 15

15 17 19 21 26

Heth Heth Heth Heth Heth

et et et et et

al., al., al., al., al.,

2002 2002 2002 2001 2001

88


similarity based on kinship findings (see Table I). Giant mole rats also consistently treated the anogenital odors of sibling giant mole rats as having similar qualities in comparison with the anogenital odor of a giant mole rat from another colony, and this was true whether the odor donors were the subjects’ own siblings or unfamiliar siblings from a different colony (Heth et al., 2002). Comparable results were obtained in similar tests (S. Begall, personal communication) with another subterranean rodent species, the South American coruro (Spalacopus cyanus). A similar conclusion could be inferred from the results of earlier studies using a standard habituation–discrimination procedure in which laboratory rats did not discriminate between the odors of urine from two members of the same strain (Brown et al., 1990).

D. Evidence of Odor Similarities within Populations and Species The habituation–generalization technique was then used to assess similarities in the qualities of the flank gland odors from one hamster species in comparison with flank gland odors of hamsters from a different but closely related species (see Table I). In these experiments, males from two species of hamsters (M. auratus and M. brandti) habituated over repeated trials to the flank gland odor of an unfamiliar conspecific and then they were presented with the flank gland odors of a different unfamiliar conspecific and an unfamiliar heterospecific: males consistently treated the flank gland odors of the conspecifics as similar in comparison with the odor from the heterospecific hamster (Heth et al., 1999). These findings are consistent with the results from the original pilot study with urine odors from blind mole rats (G. Heth and J. Todrank, unpublished data). In addition, other data from blind mole rats indicate that they treat urine odors of conspecifics as similar and urine odors from heterospecifics as different (Heth et al., 1996b). The data that inspired the idea of using similarities in odor qualities to evaluate degrees of genetic relatedness came from a study in which the similarities in urine odors from blind mole rats within populations and species were detected by laboratory rats in a ‘‘go–no go’’ operant conditioning task (Heth et al., 1996a), and thus it was not clear whether mole rats would respond to the odor similarities in the same way. Tests of responses of blind mole rats to urine odors from two contiguous populations of the same species, using the habituation–generalization technique, indicated that mole rats treat the urine odors of individuals from the same population as similar in comparison with the urine odor of an individual from the other population. This provides evidence that mole


89

rats also respond to the similarities between odors of individuals from the same population (G. Heth and J. Todrank, unpublished data). Another test was conducted, using the habituation–generalization technique to assess similarities in anogenital odors within two populations of wild mice, Mus spicilegus: males treated odors of individuals from the same population as similar compared with the odor of an individual from a different population (see Table I), providing further evidence of odor similarities at the population level (Heth et al., 2001). Thus it became increasingly clear that not only was there something common to the quality of the individual odors of siblings and half-siblings that was different from other conspecifics, but there were also qualities in common among individuals from the same population that were distinguishable from the odors of conspecifics from a different population and qualities in common among conspecifics that were distinguishable from the odors of individuals from a different but closely related species. E. Evidence of Odor Similarities across Species A better understanding of the implications of the previous data emerged from a study of individual odor similarities across four sibling species in the S. ehrenbergi superspecies of blind mole rats, using the habituation– generalization technique (Heth and Todrank, 2000). In investigations of similarities between odors across species, both of the test odors are from different odor categories in that both test odor donors are from different sibling species than the habituation odor donor. There are two possible outcomes in such habituation tests: (1) if subjects investigate the odor of the heterospecific that is more genetically similar to the habituation odor donor, this indicates that the same principle obtains across species as within species discrimination; (2) on the other hand, if neither test odor is perceived as similar in quality to the habituation odor, as would be true if different species had unique species-specific odor markers, there would be no predictable differences between the investigation times of the odors in the test trial. In this case there would be no consistent pattern of results across subjects, and the statistical tests would not show significant differences between them. In the first study of individual odor similarities across species, mole rats were presented during habituation with the urine odor of a same-sex individual from one species and then were tested with the urine odors of same-sex individuals from two different species in the superspecies. Because the genetic relationships among sibling species and the phylogeny of this superspecies have been determined by, among other methods, DNA–DNA hybridization, allozymic, and mitochondrial DNA (mtDNA)

90


analyses (Nevo et al., 2001), it is possible to determine whether the similarities in urine odor qualities across species parallel the established phylogenetic relationships. In a series of six experiments, mole rats treated the urine odors of more closely genetically related species as similar in quality compared with the urine odor of an individual from a less closely related species (see Table I). This provides the first evidence of similarities between individual odors of heterospecifics and further evidence of parallels between genetic similarities and those in individual odors, this time at the across-species level. Furthermore, the odors of the descendant species were perceived as similar to those of their closest ancestral species, suggesting that genetic similarities that remain between the ancestral and the descendant species are evident in the similarities between their odors (Heth and Todrank, 2000). The habituation–generalization technique was also used to assess similarities in anogenital odors across species of Zambian mole rats from the genus Cryptomys: mole rats from three species (C. mechowi, C. anselli, and C. kafuensis) habituated to the odor of a conspecific and then were tested with odors of individuals from the other two species. Mole rats treated anogenital odors of genetically closer heterospecifics as more similar to the odors of conspecifics than to odors of less closely related heterospecifics (Heth et al., 2002). Again, the odor similarities paralleled genetic similarities as determined by analysis of allozymic and karyotypic distances (Filippucci et al., 1997; Burda et al., 1999). A third study assessing odor similarities across species, conducted with two pairs of genetically close species from the Mus musculus species complex (M. musculus and M. domesticus; M. spicilegus and M. macedonicus) showed (see Table I) that mice treated anogenital odors within each sibling species pair as similar compared with an odor from the other species pair (Heth et al., 2001). Consistent findings across species in three different genera confirm that the predictable relationship between genotypes and individual odors extends across species as well as within species. This also indicates that there are no salient species-specific odor markers to enable unequivocal distinctions between species. F. Understanding Odor–Genes Covariance Previous studies of individual odors in rodents indicate that each individual produces a distinctive odor that is determined in part by the unique genotype of that individual (e.g., Yamazaki et al., 1992, 1999a; Brown, 1995). The process by which particular genes encode particular individual odors remains unknown, but the genotype of each individual is


91

manifest in the odor of that individual. The cumulative evidence from studies with several rodent species indicates that for various degrees of genetic ‘‘relatedness,’’ or similarity [from siblings (Todrank et al., 1998, 1999; Heth et al., 1999, 2002) to populations (Heth et al., 2001) and species (Heth et al., 1999) to close species (Heth and Todrank, 2000; Heth et al., 2001, 2002)], rodents consistently treat odors from more closely related individuals as similar compared with odors from less genetically similar individuals. Because individual genotypes are expressed in individual odors, individuals that have different proportions of genes in common share proportional similarities in the qualities of their individual odors, a phenomenon termed ‘‘odor–genes covariance’’ (Heth and Todrank, 2000). In other words, although each individual has a unique combination of alleles, all individuals share genes in their genome with other individuals to the extent that they share a common ancestry (which is greater among members of the same population than different populations of conspecifics and among conspecifics than heterospecifics, and diminishes with genetic distance among heterospecifics). The commonality of their genotypes is expressed in common qualities in their individual odors. Before turning to the implications of this predictable relationship between odors and genes, it is important to have a clear understanding of what odor–genes covariance is and is not. The metabolic process by which individual genotypes become evident in individual odors is extraordinarily complex: even a small difference in the genotypes of congenic mice [pairs of inbred strains that are genetically identical except for one particular part of the genome, such as genes in the major histocompatibility complex (MHC)] produces differences in the proportions of at least eight volatile carboxylic acids expressed in their urine as detectably different odors (Singer et al., 1997). (Note that the differences are expressed in proportions and not in individually distinctive compounds!) Consequently, it would be virtually impossible, at least at the current state of knowledge, to decipher a predictable relationship between particular genes and particular chemical components or their proportions in individual odors. Odor–genes covariance does not refer to a relationship between particular genes and particular odorous compounds or even particular proportions of compounds. On the contrary, odor–genes covariance refers to a general relationship between individual genotypes and individual odors, namely, the greater the genetic similarity between two individuals, the more similar their individual odors will be. In other words, closer genetic relatedness between individuals (whatever the degree of closeness from kin to across species) is associated with greater similarity in the qualities of their individual odors. In this odor–genes covariance it is the variation in the genotypes that drives the variations in individual odors and

92


not the other way around. Thus, odor–genes covariance is a way of looking at ‘‘relatedness’’ (and how that ‘‘relatedness’’ is manifest as similarity in individual odors) without referring either to particular genes or to particular coefficients of relatedness. Identifying specific genetic similarity in terms of genes or coefficients of relatedness is not necessary when assessing relationships among individuals (and their odors) that are known to be from distinct gene pools. It is unlikely that scientists who study chemical communication would argue against the idea that there would be a predictable relationship between individual genotypes and individual odors; perhaps many would even think the relationship too self-evident to spend time on demonstrating it experimentally. But, as becomes clear in the next section, the implications from the findings of studies that did investigate this relationship between genotypes and individual odors have led to a deeper understanding of odor-based ‘‘recognition’’ mechanisms in rodents.

III. Mechanisms Underlying Differential Behavioral Responses to Individual Odors A. Individual Recognition by Association One of the surprising and fortuitous findings from the original hamster studies using the habituation–generalization technique was that the hamsters responded differently to the odors when they were familiar with the odor donors than when they had never previously encountered the odor donors (Todrank et al., 1998, 1999; Heth et al., 1999). As mentioned previously, all the hamsters tested treated the odors of siblings as having similar qualities; the differences came in the specifics of the responses to the test odors. When the two test odors are presented in separate sequential trials, it is possible to assess spontaneous discrimination between the similar odors as well as the similarity between them. Comparison of the investigation times in the first test trial and the last habituation trial can indicate whether or not the subjects discriminated between the similar odors. Significantly more time spent investigating the odor in the first test trial than the last habituation trial demonstrates that the subjects discriminated between the odors (see Fig. 1A and Table I). If subjects do not consistently perceive the odor in the first test trial as different from the habituation odor, the investigation time in the first test trial can be the same or even shorter than during the last habituation trial (see Fig. 1B and Table I).


93

Golden hamsters that were familiar with the odor donors consistently discriminated between the odors of sibling donors. This was true whether they were familiar with the donors through nesting together before weaning or had been exposed to the donors during brief encounters as adults (Todrank et al., 1998, 1999). By contrast, hamsters that were not familiar with the sibling odor donors did not spontaneously discriminate between their odors (Todrank et al., 1998, 1999). When tested as adults, hamsters that had been raised in mixed litters of siblings and nonsiblings did discriminate between the odors of their familiar foster siblings but did not discriminate between the odors of their unfamiliar biological siblings, which had been raised in another litter (Todrank et al., 1999). Similar results were also found with Turkish hamsters (Heth et al., 1999) and giant mole rats (Heth et al., 2002). The results from these studies (see Table I) include eight experiments in which subjects discriminated between the odors of familiar sibling donors (whether they were related to the subject or not) and seven experiments in which subjects did not discriminate between the odors of unfamiliar sibling donors (whether they were related to the subject or not). Although it is not possible to draw direct conclusions about individual recognition from these experiments, one inference from these findings is clear: for all three species tested, some sort of interaction with closely related individuals, such as siblings, was necessary for subjects to spontaneously discriminate between the odors of these donors. It is reasonable to surmise that during the social interaction, either through nesting together or through brief encounters as adults, individuals make associations between particular individuals and their specific individual odors, that is, rodents learn which odor belongs to which individual. There is additional supportive evidence for this notion because in other experiments golden hamsters learned to recognize particular individuals and not merely their individual odors (Johnston and Jernigan, 1994). The implication from these findings is that, when subjects spontaneously discriminate between odors of sibling donors in these habituation tests, they do so because they have made an association between the particular familiar individual and the odor of that individual, and this enables them to recognize the odors in the test as belonging to particular familiar individuals. Similarly, the contrasting implication would be that subjects do not discriminate between the odors of unfamiliar sibling donors because they have not associated those odors with particular individuals and because these unfamiliar donor odors are sufficiently similar that the differences between them are not salient without prior association. While acknowledging that the evidence is not definitive, we are ready to consider the demonstrated discrimination between odors of familiar sibling donors

94


as tantamount to recognition of those individual odors. We are equally prepared to conclude from the results of the cross-fostering study (Todrank et al., 1999) that the type of individual recognition that occurs as a consequence of nesting together before weaning is not recognition that is confined to biological siblings but is applicable to any familiar individual. Nest mates learn about each others’ odors and can recognize which odor belongs to which individual whether they are related to those other individuals or not. We are also prepared to conclude that the nature of the association made between individuals and their odors is the same whether the individuals are nest mates encountered early in life or acquaintances encountered as adults. Thus individual recognition through association should not be considered a mechanism for ‘‘kin recognition,’’ that is, a mechanism exclusively for the recognition of kinship, but rather a mechanism for recognizing any familiar individual whether kin or not. B. Genetic Relatedness Assessments Through Individual Odor Similarities: G-Ratios Although the odor–genes covariance tests investigated similarities within one distinct gene pool in comparison with another (i.e., siblings vs nonsiblings, one population vs another, one species vs another, and close species vs less closely related heterospecifics), one implication of the combined findings is that the covariance between genotypes and their representation in individual odors occurs across a continuous proportional spectrum of degrees of relatedness or similarity rather than being confined to particular relatedness categories. In other words, the similarities between odors do not fall into discrete genetically similar groups, such as kin, population, species, and superspecies. Rather, when tested at enough different degrees of relatedness, it would become clear that there is a graded spectrum even within the tested groups. Full siblings would have more similar odors than half-siblings because they are more genetically similar, the odors of half-siblings would be more similar than those of cousins, first cousins would have more similar odors than second cousins, and so on. Even within subpopulations of relatively unrelated conspecifics, for example, some individuals would share more genes in common and thus have more similarities in the qualities of their individual odors simply as a consequence of the shuffling of genes within populations from one generation to the next. The logical consequence of this relationship between genes and odors is that an individual rodent could determine its degree of genetic relatedness to any other individual if it were able to compare the degree of similarity between the odor of the other individual and its own odor. When this type


95

of process was originally proposed as a mechanism for distinguishing between kin and nonkin, it was termed the ‘‘armpit effect’’ (Dawkins, 1982) and ‘‘self-referent matching’’ (Holmes and Sherman, 1982, 1983). There was suggestive evidence of such a mechanism underlying differential responses to kin and nonkin in ground squirrels (Spermophilus beldingi; Holmes and Sherman, 1982), but it was first demonstrated in rodents in the context of a kin recognition by phenotype-matching study that included an experiment using mixed litters of cross-fostered hamsters (Heth et al., 1998). 1. Evidence of Differential Responses to Odors from Kin and Nonkin Based on Genetic Relatedness The evidence that was suggestive of a mechanism for individual recognition by association in golden hamsters (Todrank et al., 1998, 1999) raises the question of whether hamsters were also able to make discriminative responses to conspecifics based on genetic relatedness as well. Hamsters engage in two types of scent-marking behavior: flank marking, which is associated with competition and aggression, and vaginal marking, which females use for sexual solicitation (Johnston, 1990). Flank scent marking is stimulated by odors of conspecifics, particularly flank gland odors (Johnston, 1975, 1977). Thus flank and vaginal scent marking were selected as a measure to assess differential responses to flank gland odors from kin and nonkin. In the laboratory, hamsters scent mark at low rates both while in their home cage and initially on being introduced into a new area, but scentmarking rates are higher when a hamster is returned to its home cage after a brief period of removal. To capitalize on this tendency, hamsters were tested in pairs, with each acting as the scent donor for the other: both hamsters were removed from their home cages for 5 min while the stimuli were being prepared and then reintroduced into their home cage after the flank gland odor of the other hamster had been placed there. Using this strategy, hamsters marked at sufficiently high rates to distinguish between responses to odors from individuals of differing degrees of relatedness. Both male and female hamsters marked at significantly lower rates when exposed to the flank odors of brothers or sisters rather than nonsibling hamsters (Heth et al., 1998). The average number of scent marks during 8min trials was graded on the basis of the degree of relatedness in response to flank odors from brothers (males’ flank marks per trial, 4; females’ flank marks per trial, 3; vaginal marks per trial, 5), paternal half-brothers (males’ flank marks per trial, 6; females’ flank marks per trial, 5; vaginal marks per trial, 10), and nonsibling male hamsters (males’ flank marks per trial, 11; females’ flank marks per trial, 7; vaginal marks per trial, 17) (Heth et al.,

96


1998). [Not using the strategy of removing the subjects from their cages for some time before testing could explain the differences between the results in the original study and those of the follow-up study by Mateo and Johnston (2000a), in which many hamsters did not scent mark at all and those that did scent mark did so at low rates.] There is other behavioral evidence, although not solely in response to odor cues, for differential treatment of kin and nonkin based on genetic relatedness in many rodent species (e.g., Grau, 1982; Holmes and Sherman, 1982; Kareem and Barnard, 1982; Porter et al., 1983; Halpin and Hoffman, 1987; Hepper, 1987; D’Amato, 1994). 2. Evidence of Self-Referent Matching in Differential Responses to Odors from Kin and Nonkin Although the differential scent marking described previously clearly indicated differential responses to odors of kin and nonkin (Heth et al., 1998), it was not clear whether the hamsters were using their own odor as a referent or whether they were using a composite template learned from their litter mates while nesting together before weaning. The experiment designed to determine whether discrimination between the similar odors of siblings is based on learning through association or on an innate tendency to distinguish between one’s own siblings but not between siblings from another family (Todrank et al., 1999) demanded a cross-fostering procedure that involved transferring halves of litters [as was done by Kareem (1983) and D’Amato (1994)]. This cross-fostering arrangement was necessary to provide pairs of familiar foster siblings and pairs of unfamiliar biological siblings as odor donors for these tests. This procedure ensured that all the pups were familiar with both siblings and foster siblings. It also ensured the availability of genetically similar but unfamiliar individuals, that is, biological siblings raised in another litter and siblings of foster siblings raised in another litter. This cross-fostering scheme was advantageous because it provided a pool of subjects and odor donors for the scent-marking experiment that would make it possible to determine whether differential scent marking was based on familiarity with the odor donors or on genetic relatedness to them (and hopefully not on an indecipherable combination of the two). Four groups of odor donors are needed to conduct these tests: familiar siblings, familiar foster siblings, unfamiliar siblings, and unfamiliar siblings of foster siblings. (It is also advisable to have a control group of unfamiliar odor donors that are not related either to the subjects or their foster siblings.) Gathering data on the responses of subjects to odor donors in each of these categories enables several different types of comparisons: significant differences in responses to odors of (1) familiar as opposed to unfamiliar


97

siblings and (2) familiar foster siblings as opposed to unfamiliar siblings of foster siblings would demonstrate that familiarity with the odor donors affects subsequent responses to their odors. In other words, subjects’ responses to the odors are mediated by familiarity with the donor rather than by the quality of the odor itself. In contrast, no significant differences in these comparisons would indicate that previous interaction with the odor donors does not influence later responses to their odors. In other words, subjects’ responses to the odors are mediated by the quality of the odor rather than by prior familiarity with the donor. Similarly, significant differences in responses to odors of (3) familiar siblings as opposed to familiar foster siblings and (4) unfamiliar siblings as opposed to unfamiliar siblings of foster siblings and other unfamiliar nonsiblings would indicate that the response to the odor is modulated by the degree of relatedness between the subject and the odor donor, whether the odor donors are familiar to the subject or not. (By contrast, when only one individual is cross-fostered into another litter in the more commonly used crossfostering design, this individual is familiar only with nonkin phenotypes, except through self-inspection. Differential responses to familiar foster siblings and unfamiliar biological siblings do not distinguish adequately between the roles of familiarity and genetic relatedness because both factors are changed in the same comparison rather than controlled. It would be possible to distinguish between a template learned from nest mates and self-referencing with such a design if subjects were tested with unfamiliar relatives and unfamiliar unrelated individuals of a third phenotype that is different from the foster family phenotype. This is not usually done, however [see Todrank and Heth, 2001]). Tests with males raised in mixed litters of siblings and foster siblings demonstrated that differential responses based on degree of relatedness were not affected by familiarity with the odor donors: average flank scent marking during an 8-min trial was low after being exposed to flank gland odors of familiar (5 marks per trial) and unfamiliar (4 marks per trial) siblings whereas marking rates were significantly higher after exposure to the flank gland odors of familiar foster siblings (9 marks per trial), unfamiliar brothers of foster siblings (10 marks per trial), and other unfamiliar unrelated hamsters (10 marks per trial) (Heth et al., 1998). These findings demonstrate that the presence of foster siblings in the nest does not change the differential responses of hamsters to odors of kin and nonkin as adults and thus that hamsters use their own odor as a referent in modulating their scent-marking responses to odors from kin and nonkin rather than a template learned from their nest mates. Mateo and Johnston (2000a) came to the same conclusion in their study with hamsters. Selfreferencing in responses to kin and nonkin has since been demonstrated in

98


first-generation offspring from crossed strains of laboratory mice (Isles et al., 2001). Graded responses based on degrees of genetic relatedness to more and less closely related kin in laboratory rats (Hepper, 1987) are also consistent with a self-referencing process. Mate preferences that were consistent with a self-referencing process were also found in mice (Lenington et al., 1992). Self-referencing could also explain differential responses of mothers to mixed groups of congenic mouse pups, that is, their preferences for pups of their own genotype as opposed to congenic pups (Yamazaki et al., 1999b). 3. Evidence of Self-Referent Matching in Differential Responses to Odors from Individuals from Different Populations and Species A more recent follow-up study to the odor–genes covariance studies in mice involved assessing differential responses of cross-fostered individuals from two species of wild mice, M. spicilegus and M. musculus, to the anogenital odors of unfamiliar mice from different populations and species. Adult male and female mice that had been reared from birth until weaning with siblings and age-matched heterospecifics were presented with anogenital odors of unfamiliar females in two-choice tests: mice showed greater interest in the individual odors of female mice that were more genetically similar to themselves as opposed to odors of less genetically similar individuals. This was true whether the choice was between individuals from different populations of their own species (males: mean, 39 vs 25 s; females: mean, 10 vs 5 s), between conspecifics and heterospecifics from the foster species (males: mean, 23 vs 9 and 37 vs 10 s, respectively; females: mean, 25 vs 10 and 28 vs 15 s, respectively), or between individuals from other species (M. macedonicus, a sibling species of M. spicilegus, and M. domesticus, a sibling species of M. musculus) in the Mus musculus complex (males: mean, 38 vs 15 s; females: mean, 8 vs 4 s) (Heth et al., 2003). In this series of experiments, mice responded on the basis of the degree of similarity between the other individual’s odor and their own odor, and these responses were not affected by common rearing with heterospecifics (Heth et al., 2003). If phenotypes of nest mates, whether conspecifics or heterospecifics, were incorporated into the composite family template of the individual, being raised in a mixed litter with heterospecifics would have disrupted differential responses to individual odors that were based on genetic relatedness to the odor donor. Because the results suggest that differential interest in these odors was not affected by common rearing but followed degrees of relatedness, subjects seemed to evaluate the test odors on the basis of the degree of overlap between the other individual’s odor and their own odor rather than on a composite template that could have been


99

learned from nest mates. [Of course, it is not possible to determine from these experiments how or when the ‘‘self’’ template, if indeed one is used at all, is acquired (see Section VI.C), but it is clear that the template is not acquired during interactions with nest mates.] By relying on the unchanging genotype of an individual, which is expressed in the odor of that individual, responses based on one’s own odor may be more accurate (and thus more adaptive) than responses based on a matching template learned from nest mates. These findings are consistent with a graded self-referencing mechanism for differential responses across a wide spectrum of genetic relatedness from kin through populations to heterospecifics. There is evidence from mice (Gilder and Slater, 1978) of preferences for odors of their own as opposed to another strain and evidence from hamsters (e.g., Murphy, 1978, 1980; Johnston and Brenner, 1982) and blind mole rats (Heth et al., 1996b) of preferences for conspecifics as opposed to heterospecifics that is consistent with differential responses based on self-referencing, although cross-fostering was not used in those studies and thus the data were not interpreted in such terms. In tests using three species of Mesocricetus hamsters of differing degrees of genetic similarity (the more similar Turkish, M. brandti, and Romanian, M. newtoni, and the less similar Syrian, M. auratus), Murphy (1977) found that estrous female Turkish and Romanian hamsters prefer the more genetically similar (Romanian and Turkish) heterospecific males to the less similar Syrian males, providing a clear demonstration that their sexual preferences were based on degrees of genetic similarity. These findings also demonstrate that the sexual preferences were not based on a speciesspecific marker, because the estrous females would not have shown sexual interest in males of different species if a species-specific marker mediated the process. 4. Understanding Genetic Relatedness Assessments through Individual Odor Similarities: G-Ratios Cumulative evidence from the studies designed to investigate differential responses based on degrees of genetic relatedness indicates that rodents are making graded responses to individual odors, using their own odor as a referent. Although the process by which rodents make comparisons between the other individual’s odor and their own odor cannot be observed directly, behavioral evidence suggests that the rodent species tested are using a self-referencing process (implied by the odor– genes covariance findings) based on ‘‘genetic relatedness assessments through individual odor similarities,’’ a process we refer to as ‘‘G-ratios.’’ The evidence indicates that, in studies evaluating differential responses to kin and nonkin, subjects showed stronger responses (more scent

100


marking) to nonsiblings than siblings, that is, less genetically similar individuals (Heth et al., 1998). In studies evaluating differential responses to individuals from different populations or species, however, subjects showed stronger responses (greater interest) to members of their own population than to another population and their own species than to different species, that is, more genetically similar individuals (Heth et al., 2001, 2003). This demonstrates that the differential responses are not due to a process that involves habituation to one’s own odor and graded responses to novel individuals and their odors based on differences between the odor of the other individual and one’s own odor. The assessment is based on the degree of overlap between the two odors, but the response (such as differential interest) may not be graded in a parallel way. Instead, similar odors (such as siblings and half-siblings) and less similar odors (such as heterospecifics) could elicit minimal interest whereas varying degrees of moderately similar odors (such as those from one’s own population as opposed to a different population of conspecifics) could elicit varying degrees of stronger interest. The tuning of these responses may have important evolutionary consequences (see Section V). In this process, ‘‘genetic relatedness’’ refers to genotypic similarity rather than to specific coefficients of relatedness or to particular categories of relatedness, such as kin and conspecifics. Genetic relatedness does not imply immediate common descent or any particular degree of genetic similarity but is equally applicable to siblings and to heterospecifics. The term ‘‘assessment’’ was selected as having the least cognitive connotations of the possible terms available to refer to a response that is somehow discriminatory without the association of judgment or recognition. ‘‘Individual odor’’ refers to the composite quality of the odorous compounds that an individual produces in its secretions and excretions. In relation to G-ratios, it is the composite qualities that are associated with the genotype of the individual, that is, the trait characteristics in the odor (see Section IV.A), that mediate the process. ‘‘Similarity’’ is the degree of overlap between the individual odor of the observed individual and the individual odor of the observer, which could depend on the extent to which the various odorous compounds occur in similar proportions for the two individuals. One reason we prefer to use the term ‘‘G-ratios’’ to distinguish this process from the more conventional ‘‘self-referent matching’’ is because self-referent matching has traditionally been used in reference to categorical distinctions, usually between kin as opposed to unrelated conspecifics [but also more recently in reference to species discrimination in cowbirds, Molothrus ater, using visual (Hauber et al., 2000) and auditory (Hauber et al., 2001) cues]. Just as odor–genes covariance occurs across a continuous spectrum (from maximally similar genotypes and odors of


101

identical twins or inbred strains to less similar genotypes of individuals from different species that share similarities in the qualities of their individual odors to the extent that they share genes in common), the Gratios process appears to enable continuous differential responses across the spectrum of genetic relatedness. It is convenient for researchers to make categorical distinctions such as between kin and nonkin, but a rodent in its natural habitat may have to make discriminative responses, such as between a half-sibling and a cousin, that would be impossible if all close relatives were lumped together as ‘‘kin’’ and all others as ‘‘nonkin.’’ Similarly, the distinctions among species are necessary for taxonomic classification. Given the demonstrated similarity between odors of close heterospecifics (see Section II.E), however, it is unlikely, when one rodent encounters another, that they are aware of these somewhat arbitrary distinctions between species (see discussion in Section IV.B). They would benefit more by responding on the basis of an assessment of degrees of genetic relatedness per se. In addition to such subtle distinctions, tuning of the responses to differing degrees of similarity enables what functions as categorical distinctions between, for example, kin and nonkin or between conspecifics and heterospecifics whenever such broad distinctions are useful to the animal. The term ‘‘self-referent matching’’ implies some sort of cognitive match between the observed phenotype and a learned ‘‘self’’ template. Although it is not yet known how the assessments underlying the G-ratios mechanism are made, it is entirely possible that the assessments are made without perceptual matching at all. Accumulating evidence is leading to a better understanding of the development, coding, and neurobiological bases of the olfactory system (e.g., Gheusi et al., 1994; Buck, 1996; Lin and Ngai, 1999). Evidence in rodents of individual odor maps in the olfactory system that vary as a consequence of differences in individual odors of congenic mice (Schaefer et al., 2001) raises the possibility that there may be a neurophysiological basis for assessments of similarities between the odor of another individual and one’s odor (G-ratios). These findings are consistent with the hypothesis that individual odors could be screened for degree of overlap with the observer’s own odor at the sensory level, precluding the need for a higher perceptual matching process involving long-term memory of an odor template. This speculation awaits confirmation through further investigation (see Section VI). C. Two Separate Mechanisms: Two Separate Functions Taken together, the findings described here point to the conclusion that there are, in fact, two distinct odor-based mechanisms underlying

102


differential responses to the odors of other individuals. One mechanism, individual recognition by association, involves learning an association between an individual and its individual odor and remembering this association to enable identifying the individual and its odor in a future encounter with either. The other mechanism, G-ratios, involves responding on the basis of an assessment of one’s degree of relatedness to another individual by comparing the degree of overlap between the odor of that individual and one’s own odor. This process is accomplished without the necessity of previously encountering the individual and learning characteristics of the other individual’s odor and, possibly, without long-term memory of one’s own odor. The first mechanism functions to enable ongoing discriminative responses to familiar individuals and their odors irrespective of relatedness. The second mechanism enables differential responses based on genetic relatedness irrespective of familiarity. Holmes and Sherman (1982) speculated about the separateness of these two types of mechanism from their findings in ground squirrels. These ideas were later substantiated in parallel studies in hamsters (Heth et al., 1998; Todrank et al., 1999). It was previously proposed (see, e.g., Porter, 1988; Tang-Martinez, 2001) that animals could learn to recognize their kin or other individuals through association and later generalize to unfamiliar individuals by comparing unknown phenotypes with known phenotypes. Of course, the processes may vary with different social systems, but there is evidence against such a mechanism in hamsters. As noted previously, males did not respond differently to odors of unfamiliar brothers of foster siblings (whose phenotypes were similar to those of their familiar foster brothers) than to those of unfamiliar males from a third, unrelated litter (whose phenotypes were not similar either to those of familiar biological siblings or foster siblings) (Heth et al., 1998). In any case, to be useful, such a generalization process would require some basis for sorting among the known phenotypes before generalization. This sorting would probably have to occur by use of a G-ratios or some self-referencing process. It would thus be more efficient to use G-ratios directly rather than to learn individual phenotypes, sort them, and then generalize. In a critical reevaluation of kin discrimination mechanisms, TangMartinez (2001) suggests eliminating the distinction between recognition by association and by phenotype matching because all discriminations entail both an association and a matching component. It is true that when two individuals become acquainted with one another they must learn an association between the other individual and its particular characteristics in order to recognize each other as acquaintances on a later encounter. It is also true that recognizing this individual in a subsequent meeting requires


103

matching the observed characteristics with those remembered from the previous encounter. Likewise, it is true that phenotype matching assumes an association with kin phenotypes in order to develop a template against which unfamiliar phenotypes can later be matched. This is sufficient to raise confusion about the dichotomy between the ‘‘association’’ and the ‘‘matching’’ processes. Although there is some overlap in the processes as Tang-Martinez mentioned, there are also legitimate distinctions between these processes that have not been clearly articulated in past descriptions. Clarifying these distinctions leads to a different conclusion about how many mechanisms are available for animals to use in differential responses to other individuals. It is not yet clear whether the template of what constitutes self is innate or acquired through learning, but assume for this discussion that there is a template and that it must be acquired. It is clear that the information learned to enable individual recognition is substantively different from the information learned to enable phenotype matching. To recognize a familiar individual the observer must learn a connection between a particular individual and its particular characteristics, but learning this connection does not have an impact on the observer’s view of self or the kin group to which it belongs. Making associations between the characteristics of various individuals around the observer and the observer itself in order to develop a composite template of what constitutes kin may involve attention to the same characteristics (cues) involved in individual recognition but the information is processed in a different way. For phenotype matching, the connection is between the characteristics of the others and the observer itself. For individual recognition, the connection is between specific encountered individuals and their particular distinctive characteristics. There is a similar dichotomy at the matching phase. Individual recognition involves matching a remembered template of characteristics associated with a particular individual to characteristics perceived during a current encounter. In this case the matching is used to determine whether or not the encountered individual is the same individual that had been encountered on a previous occasion. The characteristics of the observer are not involved in any way in this process (see Fig. 3). By contrast, phenotype matching involves a comparison between a template of what constitutes self or kin with the characteristics of an encountered individual, which may or may not have been encountered previously, to determine whether this individual is a relative or not. Although both processes may involve matching, comparing oneself or a kin template with an encountered other is different from comparing a ‘‘remembered other’’ with the ‘‘currently encountered other.’’ It is entirely possible that both types of

104


Fig. 3. Schematic diagram representing the individual recognition process. Odors of different individuals (A, B, C, D, E, and F) are depicted as variations in proportions of different compounds, as if all compounds contributed equally to the overall odor gestalt. A, B, and C represent acquaintances and A, B, and C represent their associated remembered traits. D, E, and F represent strangers. When the observer encounters another individual, it compares the observed traits with the remembered traits. If the individual is an acquaintance, such as A, there are some differences between the observed traits and the remembered traits B and C, but there are no differences from the remembered traits A, and thus A is recognized as A. When strangers are compared with the remembered templates, there are some similarities, but the differences indicate that the strangers are not any of the known acquaintances.

processes occur simultaneously, that is, the observer may evaluate whether the encountered individual is recognized individually at the same time that the observer responds on the basis of whether this individual is a relative or not, but this makes them parallel processes rather than a single process. The data from the recognition studies in hamsters (Heth et al., 1998; Todrank et al., 1999) provide a good example of the separateness of the two systems because the same group of hamsters was tested in two different types of experiment (Section III.A). Studies conducted with three species of Zambian Cryptomys mole rats indicate comparable results in terms of individual recognition of odors from familiar individuals (Heth et al., 2002) and differential responses to kin and nonkin apart from familiarity (G. Heth, J. Todrank, S. Begall, and H. Burda, unpublished data). We concur with Tang-Martinez (2001) that individual recognition by association is a genuine mechanism for discriminative responses to different individuals, but in light of the evidence described and discussed previously, we would add G-ratios as a second genuine mechanism for discriminative responses to different individuals. Neither of these mechanisms could properly be termed ‘‘kin recognition’’ because neither is used exclusively for discriminative responses to kin and nonkin. Each may function as a means of differential responses to kin and nonkin, however, and, in this respect, individual recognition by association is encumbered by the necessity of restricting familiarity whereas the G-ratios process is not.


105

Fig. 4. Schematic diagram representing the G-ratios process and the degree of overlap of odors and genotypes for individuals of different degrees of relatedness (siblings A and B, conspecifies C and D, and heterospecifics E and F). The genotype of the observer, and thus its odor, overlaps substantially with those of A and B, enabling the observer, O, to treat them as siblings although each differs slightly from O. O’s odor overlaps sufficiently with those of C and D to treat them both as conspecifics, even though they differ from O. The same is true of the degree of overlap with E and F, despite their more substantial differences from O.

There is yet another important distinction between these two processes. The G-ratios process relies on assessing proportional similarities between the encountered individual and oneself whereas individual recognition relies on discriminating particular differences across individuals. When an observer is attempting to assess its degree of genetic relatedness to an encountered individual, the observer makes a comparison between the encountered characteristics and its own characteristics and must determine the degree of overlap or similarity between them (see Fig. 4). Recall that it is similarities between genotypes that are manifest in similarities between individual odors, and thus individual odors are used to assess genetic relatedness by evaluating the degree of similarity between the odor of another individual and one’s own odor. In this instance the particular differences are unimportant. (Of course, individuals of comparable genetic similarity to the observer would have comparable proportional differences in their odors but the particular differences would vary. Calculating the proportional differences from the particular differences would be more cumbersome than assessing the proportional similarity directly.) By contrast, discrimination can be accomplished only by noticing differences. Noticing the similarities that individuals share is useful for

106


categorizing them as members of a group, but it is not helpful in discriminating among them. Individual recognition can be accomplished only by discriminating the differences between the particular recognized individual and other known individuals. When attempting to recognize a particular individual, the observer must compare the encountered characteristics with the remembered characteristics of various individual templates, noting the differences to sort among them (see Fig. 3). Noticing only the similarities would easily lead to mistaken identity, not only because the known individuals may share many characteristics but also because there are similarities between known and unfamiliar individuals. Differences between the encountered characteristics and all the remembered characteristics indicate that the encountered characteristics belong to an unfamiliar individual (see Fig. 3). By finding among all the remembered characteristics the one that minimizes the differences from the encountered characteristics, it is then possible to determine whether there are sufficiently few differences to enable identifying the encountered individual correctly. In this process the similarities across individuals are unimportant. The two separate mechanisms described here for odors should operate on the same principle in other sensory modalities. Learning, for example, the association between a particular face or voice and a particular individual enables later recognition of that individual from the face or voice. This is possible by distinguishing the differences among particular features of faces or voices of known individuals. Facial or vocal resemblances may also covary with genetic similarity, however. Unfamiliar faces or voices may be mistaken for one another if they are similar, such as among same-sex members of the same family. Members of the same population have more facial features in common than do individuals from different populations. Analogously, dialects are often shared within populations and vary across populations. Thus assessing similarities in facial characteristics or dialects could provide a means of genetic similarity assessments based on degrees of similarity or overlap with oneself in modalities other than olfaction.

IV. Rethinking Terminology Associated with Odor-Based Mechanisms Underlying Differential Behavioral Responses Tang-Martinez (2001) wrote correctly that a lack of clarity in terminology can result in a lack of clarity in formulating questions, designing experiments, and interpreting results. In light of more recent


107

evidence and new interpretations, we would like to revisit some of the terminology as more clarity about these mechanisms is emerging.

A. Individual Odors Because each individual (except monozygotic twins) is genetically unique and because it is clear from previous studies that the odors of individual secretions and excretions can be discriminated individually because of differences in individual genotypes, it is not surprising that the traditional view of the term ‘‘individual odor’’ is associated with individually distinctive odor qualities that were thought to distinguish one individual from another member of the same species (Yamazaki et al., 1991, 1999a; Brown, 1995). Initial investigations of individual odors focused on the role of the highly polymorphic major histocompatibility complex (MHC) of genes in determining the distinctive qualities in individual odors (e.g., Beauchamp et al., 1986; Brown et al., 1987; Yamazaki et al., 1991). It was particularly intriguing that a single mutation in a gene of the MHC enabled discrimination between the urine odors of congenic mice when subjects were trained to make the distinctions (Yamazaki et al., 1990). Other studies of mice and rats have identified the importance of genes on the X and Y chromosomes (Yamazaki et al., 1986; Schellinck et al., 1993) and the role of the rest of the genome apart from the MHC in individual odor discrimination (Beauchamp et al., 1990; Eggert et al., 1996). In addition, studies using habituation–discrimination techniques (see, e.g., Halpin, 1986, and references therein) demonstrated that individual odors of unrelated conspecifics are readily discriminable even when the odor donors are not familiar to the subjects. All the evidence seemed to suggest the salience of the individually distinctive qualities as constituting the individual odor. The results from studies of sibling odors in golden and Turkish hamsters (Todrank et al., 1998, 1999; Heth et al., 1999) and giant mole rats (Heth et al., 2002) indicated, however, that sibling odors were sufficiently similar in quality that subjects did not discriminate between them spontaneously in habituation– generalization tests when the odor donors were not familiar to the subject (see Table I and Fig. 1B). This raises questions about just how distinctive the individually unique components are. Because sibling odors are discriminated when the odor donors are familiar to the subjects, there must be distinguishable differences between them (see Table I and Fig. 1B). But the unique qualities in the odors of these individuals would have to be subtle, at least in these species, for subjects to fail to discriminate spontaneously between them when they are unfamiliar.

108


Chemical analysis of the anal scent secretions of Indian mongoose, Herpestes auropunctatus (Gorman et al., 1974; Gorman, 1976), and urine from congenic strains of mice (Singer et al., 1997) indicates that variations in individual odors may depend on differences in proportions of odorous compounds rather than on uniquely distinctive odorants. The variations in proportions of odorous compounds in individual odors have not been assessed across the spectrum of relatedness, but the odor–genes covariance findings suggest that similarities in proportions of odorous compounds could account for the similarities in odor qualities of individuals within groups and differences in proportions of odorous compounds could account for the discriminability of odor qualities of individuals across groups. An individual odor is not merely a particular odor or combination of odorants that is unique to each individual. Rather, individual odors are composites of odor qualities that, in addition to having individually distinctive qualities, provide other types of information about the individual, including the kin group, population, and species to which the individual belongs (Todrank et al., 1998, 1999; Heth et al., 1999, 2001; Heth and Todrank, 2000), by virtue of sharing common parts of their genome with other members of the various groups. It may be useful to think of individual odors as gestalts that are analogous to faces, whose characteristics present information about individual genotypes, but simultaneously identify individuals as belonging to a particular family, ethnic group, race, and species. Just as it is the configuration of features rather than a specific feature that identifies particular individual faces, it is the odor gestalt rather than a specific odorous compound that identifies individuals in their odors. Because the odor–genes covariance and G-ratios studies have been investigating the relationship between genotypes and odors and how this relationship serves as a mechanism for differential responses based on relatedness, the investigators in these studies have been careful to control other transient factors known to affect the qualities of mammalian body odors, such as diet (e.g., Brown, 1995; Schellinck and Brown, 1999) and reproductive status (e.g., Heth et al., 1996b). Just as genotypic traits are manifest in individual odors as they are in individual faces, changing biological states affect the qualities of individual odors in a manner similar to how emotions affect facial expressions. The happiness or anger that is evident in a face does not, however, inhibit the recognition and identification of the particular individual or the other genotypic characteristics displayed there. The health, reproductive state, or dominance status of a rodent adds information to what is known about the particular individual at the particular time, but this should not mask the genetically determined characteristics. Although the changing states of the animal are also evident in variations in the way the composite odor quality


109

is perceived at any point in time, it is the unchanging genotypic traits expressed in the odor that enable the assessment of genetic similarity. The extent to which changing biological states affect differential responses to odors from individuals of differing degrees of relatedness is a fruitful area for future studies (see Section VI.A). B. Markers, Signals, and Cues Rodents are able to discriminate, on the basis of odors, between different individuals (e.g., Johnston and Jernigan, 1994), between kin and nonkin (see Halpin, 1991), and between conspecifics and heterospecifics (see Doty, 1986). One explanation for this behavioral evidence was that these discriminations were based on distinctions between particular odor markers, signals, or cues that characterize each individual or group. Other behavioral evidence indicated, however, that discriminative responses to odors of individual siblings (Section II.C), kin and nonkin (Sections II.C, III.B.1, and III.B.2), members of different populations and species (Sections II.D and III.B.3), and even different heterospecifics (Sections II.E and III.B.3) do not rely on specific markers that identify each category (Section II.F). They seem to rely on the covariance between genes and odors and are probably based on the G-ratios process. The chemical analysis (Gorman et al., 1974; Gorman, 1976, Singer et al., 1997) also suggests that the odor qualities shared among individuals that share genes in common provide the cue to the relatedness. Individual genotypes are expressed in individual odors, and thus the extent to which individuals have proportions of their genomes in common is evident in shared proportional similarities in the qualities of their individual odors. What were generally considered kin-, population-, or species-specific odor markers are actually similarities in the composite quality of the odors of individuals that share varying proportions of genes (Heth and Todrank, 2000). In other words, qualities of specific odorous compounds are not used to identify different groups, instead, odor similarities among individuals that share genes (and thus odor qualities) provide the basis for identifying members as belonging to various groups. The idea of specific odor markers, particularly for identifying kinship and species, seems theoretically plausible, especially given the necessity of discriminating between kin and nonkin and between conspecifics and heterospecifics, for example, for appropriate mate choices. On more careful reflection, however, it is difficult to hypothesize how such markers would have evolved. Any marker of kinship would have to change with every mating couple to indicate specific sibships and yet maintain sufficient similarity to indicate kinship with half-siblings and cousins on both the

110


mother’s and the father’s side. Any species marker would have to spread immediately to all members of the population of an incipient species and yet the individual carrying the initial mutation and its associated odor would not be considered a suitable mate and would tend to be removed by selection (Butlin, 1995). For such markers to be effective, they would need to be both distinctive and nonoverlapping so that the marker for one degree of relatedness was not confused with or masked by another. That is, the species marker must not be confused with the kin marker or the sibling marker confused with that for half-siblings and first and second cousins, the individual marker must not mask the species marker, and so on. This may be especially complicated because of the biological constraints on the variety of odorous compounds that animals are able to produce and because, as mentioned previously, the variations between the composite qualities of individual odors rely on differences in proportions of odorous compounds (Gorman et al., 1974; Gorman, 1976; Singer et al., 1997) rather than on distinctive odorants. Although the definitions of marker and signal are sufficiently broad as to be compatible with the findings discussed previously, the connotations of these terms suggest a degree of specificity that the data do not support. If the identifying qualities for particular groups were to be considered ‘‘markers,’’ it would have to be acknowledged that there is substantial overlap in these markers: the composite qualities shared by all individuals of a particular group subsume the qualities that are shared by all members of wider but still genetically similar groups. This is true whatever the degree of genetic similarity from kin groups through populations and species to across species. The same distinction between a marker and a gestalt applies in interpreting information in faces. On occasion an individual is ‘‘marked’’ by an identifying aberration, but generally it is the overall facial gestalt that enables the identification not only of the individual but also of the population and species to which he or she belongs. Thus, it is less confusing to forgo the terms ‘‘marker’’ and ‘‘signal’’ when referring to the trait characteristics that identify individuals as individuals or as members of particular groups and to refer instead to the overlap in the odor gestalts that enables the classification. This is advisable particularly because it is the degree of overlap rather than particular markers that enables genetic relatedness assessments on the basis of individual odors. The terms ‘‘marker,’’ ‘‘signal,’’ and particularly ‘‘cue’’ are useful, however, when referring to changes in the composite quality of the individual odor due to variations in biological or motivational state. To the extent that reproductive readiness, dominance status, and even diet are evident in the composite odor quality, these qualities are more likely to be representative of odor cues.


111

C. Recognition ‘‘Recognition’’ refers to identifying an individual or characteristic as having been known previously. Recognition is a cognitive process that implies an association between the particular individual and the circumstances through which it became known. Because recognition is a cognitive process, it cannot be observed directly and it is often difficult to determine from the observable behavior whether behavior is indicative of recognition or not. For this reason many researchers prefer the term ‘‘discrimination,’’ because behavioral discrimination is clear (see TangMartinez, 2001). In individual recognition by association, the term ‘‘recognition’’ is appropriate because it is clear from the behavior that the individuals are identified and recognized individually even though the cognitive association between the individual and its characteristics must be inferred. With elegant experimental designs, such as that used by Johnston and Jernigan (1994) with hamsters, this recognition can also be inferred from responses to individual odors. The idea of recognition becomes more complicated in reference to categories of individuals, such as kin or species, rather than to particular individuals. When an observer recognizes particular acquaintances who happen to be kin, it is difficult to determine whether recognition of the kinship is part of the process of recognizing the particular individual. The same can be said for recognition of acquaintances who are members of the observer’s own population or species. Even when particular acquaintances are treated differently and this differential treatment is attributable to differences in relatedness, it cannot be inferred that the observer ‘‘recognizes’’ these distinctions. When a person meets an acquaintance, he or she can recognize that individual as the one who was previously met. When a person meets a stranger, he or she may recognize particular characteristics about that stranger, such as the family, ethnic group, or race to which he or she belongs, or the observer may be unaware of the resemblance to these previously encountered features. Either way, if the stranger is another person, the stranger is recognized as a person. It is difficult to determine, however, even in human interactions, the extent to which responses to other individuals depend on characteristics that are learned and recognized, except when the recognition is of particular remembered individuals. Probably, given the many levels of finely tuned discriminative responses that rodents are able to make at so many different degrees of relatedness, rodents do not recognize these distinctions, but rather simply respond to them. When a hamster scent marks more vigorously in response to the

112


scent of a half-sibling than a full sibling (Heth et al., 1998), it is unlikely that the hamster recognizes the differences in its degree of relatedness to the two scent donors, particularly when the differences between those two odors (as in the case of siblings and half-siblings) are so subtle that they may not be discriminated as different in habituation tests (Todrank et al., 1998, 1999). Similarly, when a mouse shows more interest in the anogenital odor of an unfamiliar individual from its own population than the odor of an unfamiliar individual from a different population of conspecifics (Heth et al., 2001, 2003), it is unlikely that the mouse recognizes the difference between its population and another. The same can be said about the likelihood of recognizing ‘‘my species’’ or ‘‘my close heterospecific’’ as opposed to another. The beauty of the G-ratios mechanism is that it enables adaptive discriminative responses across a wide spectrum of degrees of relatedness without recognizing anything about those degrees of relatedness at all. Differential responses that are indicative of genetic relatedness assessments do not imply that animals recognize their degree of relatedness to other individuals. Similar responses to all individuals of comparable degree of relatedness, such as close relatives, members of one’s own population, or conspecifics, need not indicate that animals categorize or classify these individuals together or that an individual recognizes them as belonging to its own kin group, population, or species. The G-ratios process provides a means of differential responses to others without the necessity for recognizing or knowing about the degree of relatedness to those individuals. Thus, G-ratios would fall under the ‘‘discrimination’’ rubric because it enables discriminative responses based on degrees of relatedness, which may, but probably do not, involve recognition. D. Familiarity ‘‘Familiar’’ means ‘‘knowing about’’ or ‘‘closely acquainted with,’’ making the term and its opposite suitable for distinguishing between certain types of individuals and odors in studies of mechanisms enabling differential responses to individuals with whom the animals may or may not have previously been acquainted. The problem with this term is that it is not always clear what exactly the animal is familiar with. For example, a ‘‘familiar individual odor’’ could be an odor that is known because it was presented repeatedly during an habituation experiment, but the individual that produced the odor may or may not be known. The term ‘‘unfamiliar,’’ when used to refer to individuals and odors that have never been encountered, is problematic because although the particular odor may be unfamiliar, there are certainly qualities in the odor that are familiar to the


113

extent that they resemble those of the subject or other previous acquaintances. It is also problematic when animals have encountered the individual or the odor but long enough previously that the level of memory trace is unclear. The idea of ‘‘indirect exposure’’ (Holmes, 1986) or ‘‘indirect familiarity’’ (Porter, 1988) further complicates matters because it is not clear what, exactly, is familiar from previous acquaintances and identified in the newly encountered individual or odor, which may in some respects be appropriately described as unfamiliar. Assuming that degree of genetic relatedness is controlled in the following comparisons, an unfamiliar individual or odor may seem somewhat familiar because it can be identified as belonging to some particular category or because it is reminiscent of a previously encountered familiar individual or odor. When a rodent treats one nonacquaintance or its odor differently from another it does not seem possible to determine experimentally which of two explanations is responsible for the result. It could be because one of the nonacquaintances is reminiscent of an acquaintance or because the animal mistakes the nonacquaintance for an acquaintance. Similarly, when a rodent treats a current acquaintance differently from a former acquaintance that the rodent has not encountered for some period of time, what does that say about familiarity? If the rodent treats the former acquaintance the same as a newly encountered individual, it would indicate that some of the previously known-about qualities of the former acquaintance have been forgotten. It is not possible, however, to determine what aspects have been lost or whether any have been retained. There may be some reluctance to consider the former acquaintance unfamiliar even though that description is consistent with the behavior of the animal. No foolproof solutions leap to mind, but confusion can be diminished somewhat by describing the exact nature of any experience with the target animals before testing. Hamsters, for example, behave as though they ‘‘forget’’ briefly encountered conspecifics and their odors within 1 month of the encounters (Todrank et al., 1999), but ‘‘remember’’ the odors of their nest mates after even 9 months of separation, making the long-lost siblings behaviorally familiar and the more recent acquaintances behaviorally unfamiliar. Different rodent species with different social systems have been shown to forget their siblings after various short periods of isolation [see, e.g., spiny mice, Acomys cahirinus (Porter and Wyrick, 1979), Ansell’s mole rats, Cryptomys anselli (Burda, 1995), and prairie voles, Microtus ochrogaster (Paz y Minõ and Tang-Martinez, 1999b)], and ground squirrels, like hamsters, appear to remember their siblings after 9 months of separation (Mateo and Johnston, 2000b). Periodic exposure to

114


Fig. 5. Schematic diagram depicting hypothetical mean (SE) times during multiple twochoice tests of differential responses of subjects to pairs of individuals (or their odors) of differing degrees of relatedness. Note that although the assessment of relatedness is based on degrees of similarity, the response can be stronger to less similar conspecifics than to the more similar siblings. The response can also be stronger to more similar conspecifies than to the less similar conspecifics and heterospecifics.

the odor of a sibling in lieu of the sibling itself is sufficient in prairie voles to maintain recognition or familiarity with that sibling beyond the time that the animal would usually remember its sibling (Paz y Minõ and TangMartinez, 1999a). An insightful anonymous reviewer of an earlier study suggested referring to the unfamiliar odor donors as ‘‘strangers.’’ A stranger is an individual that the individual is not acquainted with irrespective of whether some aspects of the stranger’s phenotype may be familiar. This goes a long way toward minimizing confusion about the use of the word ‘‘unfamiliar.’’ Similarly, ‘‘former nest mates’’ can be used to refer to individuals that were reared together but later separated and ‘‘cage mates’’ can refer to subjects and target animals that are currently housed together. Further identification based on relatedness, such as ‘‘sibling strangers’’ or ‘‘unrelated cage mates,’’ can also add to the clarity of describing the experimental setup.

V. Speculations on the Origin and Evolution of Preferential Responses Based on G-Ratios and Their Function as a Premating Isolating Mechanism It is well known that varying genotypes are expressed in varying phenotypic traits, but the idea of using this relationship to make adaptive genetic relatedness assessments had not been explored previously. Thus it


115

is worth considering how the process of responding differentially on the basis of such assessments could have evolved. The covariance between genotypes and phenotypic traits not only suggested a mechanism for genetic relatedness assessments based on comparisons between the traits of another individual and one’s own characteristics; it also implied a hypothetical process by which this type of genetic similarity assessment mechanism and the differential responses based on those assessments may have emerged during evolution. Bearing in mind that genotypes are manifest in individual trait gestalts (such as odors, faces, and voices) rather than as specific markers, consider the process that occurs during speciation. As a population (e.g., of rodents) that will become an incipient species colonizes a new ecogeographical area and undergoes the associated environmental challenges, genetic variations occur that change the characteristics (e.g., physiological and behavioral) of individuals within that population. Some of these changes in individual genotypes are also manifest in changes in individual phenotypic characteristics. Individuals that retain the characteristics of the original population will have diminished reproductive success, irrespective of their mate preferences, to the extent that their characteristics are less adaptive in the new ecogeographical area. The reproductive success of individuals with adaptive genetic variations and the associated characteristics, however, will vary on the basis of the strategy by which they choose their mates. Individuals with adaptive characteristics will enhance their reproductive success by choosing to mate with other individuals with adaptive characteristics like their own rather than with individuals with less adaptive characteristics. Thus it would be advantageous to be able to discriminate between individuals on the basis of genetic similarity and then to prefer individuals with characteristics similar to one’s own. How could individual rodents with adaptive genetic variations and characteristics find each other when the characteristics, such as higher tolerance for heat or cold, may not be evident in particular distinguishable phenotypic characteristics? Because some variations in genotypes are manifest in variations in individual phenotypic characteristics, among them odors, individuals with similar phenotypic traits would be more genetically similar compared with individuals with less similarity in their phenotypic traits. Individuals that are able to respond discriminatively to others on the basis of their degree of genetic relatedness by responding to the degree of similarity between their phenotypic traits, such as odors, may be able to use this discriminative ability to choose more suitable mates in terms of the fitness of their descendants. Although the adaptive characteristics that are not manifest in variable phenotypic characteristics

116


may not initially be correlated with evident variable phenotypic traits, over time the correlations would be likely to increase to the extent that choosing on the basis of phenotypic similarity resulted in reproductive success. Thus discriminating on the basis of similarity and then preferring to mate with an individual with relatively similar traits could ultimately lead to enhanced reproductive success. These discriminative and preferential behavior patterns would evolve because of the greater reproductive success of individuals with adaptive characteristics and those that choose to mate with others possessing similar characteristics. (The preference would have to be for similar but not too similar, however, to optimize inbreeding/outbreeding while avoiding the consequences of inbreeding depression [Bateson, 1983].) Thus animals should capitalize on this covariance between genotypes and phenotypes and use their abilities to discriminate and respond preferentially in ways that enhance their reproductive success. Mate selection in rodents is driven by chemical communication, and thus mate choices depend mainly on odor preferences. (The neurophysiological basis underlying odor preferences is not yet known, but it is likely, given the importance of such ‘‘decisions,’’ that the preference process has prerodent and perhaps even premammal evolutionary origin, and this ‘‘decision-making’’ process probably need not reevolve during each speciation process.) Whatever the evolutionary root of the odor preference that is manifest in mate choice, individuals with adaptive genotypes that choose to mate with more genetically similar partners would leave more offspring, and the pattern of choosing on the basis of odor similarity could develop and spread as well. As adaptive genotypes and characteristics spread through the population of an incipient species [following the allopatric or peripatric models of speciation (Mayr, 1963, 1982; Fig. 16.1 in Futuyma, 1998)], the individual odor qualities associated with those genotypic changes also spread. The adaptive advantage of choosing ‘‘genotypic similarity’’ by using the G-ratios process should result in the spread of that process. At the same time, the preference to mate with an individual possessing the new similar odor qualities should spread in the population in parallel. Once this genetic relatedness assessment and the associated preferential responses become fixed in a population, they could serve to maintain the genetic isolation of the population. When the genotypic variations and changes of individual odors between two neighboring populations of a species are small, they would remain distinct populations but still members of the same species. In this case, the differences between the genotypes and the individual odors of the two populations would not be sufficient to avoid mating and inhibit gene flow between them in case of encounters between members of the two


117

populations and thus, despite the preference for similarity to oneself, there would be no splitting into separate species. When, on the other hand, the genotypic changes are more substantial, as would occur, for example, when geographic barriers prevent the gene flow between populations for a sufficiently long period of time and the selection pressures in the new environment are sufficiently strong, the concomitant changes in the individual odors would be more pronounced as well. As these substantial differences in genes and odors spread through the population of an incipient species, at some point the individual odors may become sufficiently different in quality from those shared with members of the original population that individuals from the new population would not find enough similarities between their own odors and those of individuals from the ancestral population to treat them as suitable mates should secondary contact occur between them. Thus, after specific individual odor preferences, which may have originated as a byproduct of fortuitous mate choices (i.e., assortative mating based on selecting mates with similar individual odors and thus a greater likelihood of similar adaptive characteristics), become fixed in the incipient species, they could function as a barrier to gene flow between the incipient and ancestral species. In this way, the finely tuned preferences based on G-ratios could serve as a premating reproductive isolating mechanism and, perhaps in conjunction with other [e.g., ecological, seasonal, ethological (vocal and visual), and postmating] isolating mechanisms that are byproducts of genetic divergence, could result in a cessation of gene exchange between the populations that would consequently lead to their becoming separate species. In the context of reproductive isolating mechanisms, the G-ratios process would be manifest in preferential responses to prospective mates. As noted previously, the findings indicate (see Fig. 5) graded stronger responses to unrelated conspecifics than to half-siblings or siblings (Heth et al., 1998) but also graded stronger responses to more genetically similar conspecifics and to more genetically similar heterospecifics (Heth et al., 2003). Although thus far the G-ratios mechanism has been tested only in the context of differential responses to odors rather than in mate preferences, the findings suggest preferences for degrees of genetic closeness that could lead to adaptive mate choices. Previous mate choice tests indicate preferences for genetic dissimilarity at the level of close kin (Lenington et al., 1992; Potts et al., 1991, 1992) but genetic similarity at the level of conspecifics (Murphy, 1978, 1980) and heterospecifics (Murphy, 1977). The odor preferences based on G-ratios and the previously demonstrated mate preferences are consistent with behaviors that could ultimately lead to reproductive isolation.

118


In areas where the incipient and ancestral species overlap after speciation, G-ratios and preferences for individuals with odors similar to one’s own could result in preferential responses toward conspecifics and their individual odors as opposed to those of heterospecifics. The discriminative behaviors could be expressed as a tendency to be less aggressive toward and more sexually responsive to conspecifics than heterospecifics. Heth and Nevo (1981) were among the few to test the origin and evolution of ethological isolating mechanisms in rodents (e.g., Godfrey, 1958; Moore, 1965; Smith, 1965). They tested the four species of blind mole rats from the S. ehrenbergi superspecies (discussed in detail in Nevo et al., 2001) and concluded that premating isolating mechanisms developed as a by-product of genetic divergence during the allopatric, ecological mole rat speciation. This is compatible with the hypothesis that G-ratios–based preferences could underlie the odor-based aspect of the reproductive isolation. Ethological isolation among rodent species that depend on chemical communication was commonly thought to rely on the development of species-specific odor markers [i.e., specific odorous compound(s) or discrete mixtures of odorants] that unequivocally distinguish individuals of one species from those of another species and that are subsequently used as the basis of species recognition. In such cases, salient differences between the specific marker signals of an incipient species and an ancestral species would enhance their efficacy in maintaining reproductive isolation because these differences would minimize identification errors that would result in inappropriate mate choices. Thus salient differences between odor markers of close species would provide an adaptive advantage by strengthening the reproductive isolation of the species. The odor–genes covariance evidence (described in Section II.E) militates against the salience of any such species-specific odor markers, at least in blind mole rats (Heth and Todrank, 2000), mice (Heth et al., 2001), and Zambian mole rats (Heth et al., 2002) because, if each species had evolved a distinctive species-identifying odor marker, the odors of heterospecifics, including closely related species, would have been treated as distinctly different rather than similar. The differential responses to members of different populations or species that are enabled by the G-ratios process and the associated preferences are based on the relative similarities or degrees of overlap between individual genotypes that are manifest in individual odors. Thus no ‘‘species-specific markers’’ (see Section IV.B) are required for the discriminative behaviors toward conspecifics and heterospecifics, and no ‘‘species recognition’’ process (see Section IV.C) is implied. The G-ratios process and the associated preferences, however, could enable rodents to


119

respond adaptively and differentially to individuals of differing degrees of relatedness, including conspecifics and heterospecifics, without the necessity of markers to discriminate between groups. Numerous studies indicate that rodents have olfactory preferences for conspecifics over heterospecifics (see Doty, 1986, for review). In these studies, the odor cues were whole body odors or other secretions (such as vaginal and scent gland) and excretions (such as urine) that are known to contain information about individual identity. Consequently, in these reports, the demonstrated preferences for conspecifics over heterospecifics could also depend on G-ratios. The odor–genes covariance evidence called the idea of species-specific markers into question, but the G-ratios mechanism, which odor–genes covariance implied, offers a satisfactory theoretical substitute. The G-ratios process not only can account for differential responses to odors of individuals of differing degrees of genetic relatedness in rodents but also suggests a process by which it can serve in reproductive isolation between biological species.

VI. Prospects for Future Studies Relating to G-Ratios The studies summarized here reveal important information about the relationships between genotypes and individual odors and about odorbased mechanisms underlying differential behavioral responses to individuals of varying degrees of genetic relatedness. At the same time, these findings raise a series of important questions for further investigation: How widespread is odor–genes covariance in the animal kingdom and how far does it extend across more divergent degrees of relatedness, such as genera within a family? How extensively is the G-ratios mechanism used in modulating differential behavioral responses in mammals and across other groups that rely on odors as a basis of communication between individuals? To what extent do social systems (e.g., solitary, social, or eusocial) in different species affect the relative importance of individual recognition by association and differential responses based on G-ratios? Does social behavior in some social species rely entirely on individual recognition or will new experiments demonstrate that individual recognition can mask mechanisms for differential responses based on G-ratios in these species? To what extent does the principle that variations in genotypes are expressed in variations in phenotypic traits enable genetic similarity assessments and differential responses based on those in other taxa and sensory modalities other than olfaction? The answers to these questions could emerge from broader and more extensive studies, some of which are described in the next section.

120


A. Distinguishing between the Importance of Traits and States in Differential Responses to Individual Odors Previous research has indicated that rodents are able to distinguish between transient biological states, such as diet quality (e.g., Ferkin et al., 1997) and dominance status (e.g., Rich and Hurst, 1998), and that subjects treat target animals differently depending on these biological indicators. A fruitful area for future research would involve investigating the interaction between previously demonstrated ‘‘state’’ preferences and the ‘‘trait’’ preferences suggested by the G-ratios studies, particularly to determine whether changing biological states interfere with responses to the trait characteristics that underlie the G-ratios mechanism. Previous studies of state preferences (such as testing effects of diet on odors of spiny mice; Porter et al., 1989) controlled for degrees of genetic relatedness and previous studies of trait preferences (described in Section III.B) controlled for transient biological factors. A series of studies that varied the genetic relatedness of target animals in a paradigm in which differential responses based on changing states, such as diet preferences, had already been demonstrated would indicate the relative importance of genetic relatedness in more complex behavioral interactions. For example, after demonstrating that conspecifics consuming a higher protein diet are preferred to conspecifics of comparable genetic similarity but ingesting lower amounts of protein, it would be possible to assess relative preferences for siblings or close heterospecifics on the preferred diet to nonsibling conspecifics on the less preferred diet. In previous tests, cutting the vomeronasal nerve had no impact on the preferences for conspecifics as opposed to heterospecifics in male golden hamsters (Murphy, 1980), so that it is unlikely that the accessory olfactory system plays a role in the responses in the suggested experiments. If, however, the state characteristics in odors trump the trait characteristics that are the basis of G-ratios preferences, it may be necessary to reinvestigate the interactions between the main and the accessory olfactory systems (see Johnston, 1998) in differential responses to individuals and their odors.

B. G-Ratios vs Species-Specific Sex Attractants or Excitants Conducting a study that involves male subjects and estrous females from different species as odor donors would put the habituation–generalization technique to the test and could produce interesting results that shed additional light on species-specific sex attractants or excitants. It is possible, given previous evidence of the stronger attraction of males to


121

estrous conspecifics than to heterospecifics (e.g., Murphy, 1980), that the attraction of the males to the odor of an estrous female would be sufficiently sustained that repeated or extended exposure would not result in habituation. This would preclude the possibility of assessing the perceived similarity of odors of estrous females from different close species by habituation–generalization techniques. It would be possible to do an experiment to distinguish between the importance of odor–genes covariance and species-specific sex attractants, however, if subjects in habituation–generalization tests fulfill two requirements. First, males that are exposed repeatedly to odors of estrous females must show habituation to the odor. Second, they must show the typical results, that is, greater investigation of the less genetically similar odor donor, when tested with odors of estrous females under two conditions: (1) conspecifics and genetically similar heterospecifics and (2) more and less similar heterospecifics. The conspecifics vs genetically similar heterospecifics control experiment would confirm that males do not investigate odors of reproductive conspecifics longer than those of heterospecifics in habituation– generalization tests. The more vs less similar heterospecifics control experiment would confirm that close heterospecifics do not have distinctive sex attractants that distinguish them categorically from less genetically similar heterospecifics. In the crucial follow-up test, males would habituate to the odor of an estrous female from a genetically similar heterospecific species. They would then be tested with odors of estrous females that were either conspecifics or from a less genetically similar heterospecific species. If males investigate the odor of the heterospecific longer than the odor of the conspecific, this would indicate that they are investigating the less similar odor and that there is no species-specific sex attractant or excitant drawing them to the odor of the conspecific and that this was not due to previous habituation to a conspecific species-specific sex attractant. A similar series of tests with estrous females as subjects and male odor donors from different species would indicate whether the female’s estrous state affects her responses to male odors in habituation– generalization experiments. C. Effects of Prenatal Learning on G-Ratios It is clear from our studies (described in Section III.B) involving subjects raised in mixed litters with siblings and unrelated individuals that postnatal association does not influence the olfactory concept of ‘‘self.’’ It is not clear, however, whether this self is determined exclusively by the genotype of the individual or whether the concept of self is influenced by exposure to odors of the mother and other fetuses in the womb during gestation. It is

122


clear from studies of prenatal learning that the prenatal environment can have a profound influence on postnatal behavior (e.g., Hepper, 1990). The results from some studies assessing prenatal odor learning [e.g., in rats (Terry and Johason, 1996), lambs (Schaal et al., 1995), and human infants (Schaal et al., 1998; Porter and Winberg, 1999)] can also be explained equally well by a G-ratios process as by prenatal odor learning, but those studies were not designed with the distinction between learning and an innate genetic relatedness assessment mechanism in mind. Previous studies (summarized by Schaal et al., 1994) have indicated that newborn pups orient toward the odor of their mother’s amniotic fluid as opposed to the odor of the amniotic fluid from another female that has just given birth. It is not clear whether the pups learned the odor of their own amniotic fluid, which includes odors from the mother and any other fetuses in the womb as well as each pup’s own odor (Schaal et al., 1994), during gestation, and thus preferred it, or whether they preferred it because this composite odor is more similar to their own odor. Of course, the only way to be sure that a newborn is responding on the basis of G-ratios rather than on matching based on odor learning during gestation would be to transplant single embryos shortly after conception. It would then be possible to test the response of the pup to the amniotic fluid of the foster mother’s sister (which would be an odor from an unrelated individual but one that would be similar in quality to the familiar odor of the foster mother) and the father’s sister (which would be an odor from a related individual but one that would not be familiar in quality through exposure during gestation). If the pup prefers the odor of the father’s sister, this would indicate preference based on genetic relatedness (which could be achieved only through comparison with the pup itself), whereas a preference for the odor of the foster mother’s sister would indicate a preference for the more familiar odor. This would be an extremely laborintensive study, but it is possible to get a sense of the likelihood of responding based on G-ratios or on odor cues learned during gestation in an easier way by testing this distinction with rodent species that have regular estrous cycles and that produce litters on an equally reliable schedule. By mating females of known relatedness with males of known relatedness, it would be possible to produce litters of subjects that are born at the same time and to have amniotic fluid from new mothers of varying degrees of relatedness to the subject pups. If newborn pups when tested with odors of amniotic fluid taken from their paternal aunt and from an unrelated conspecific female show more interest in the odor from the paternal aunt, this would suggest that they were choosing on the basis of their genetic relatedness to the aunt, which is genetically similar to


123

themselves but not related to their mother. If, in addition, newborn pups showed comparable interest in the odors of their maternal aunt and paternal aunt, this would suggest that pups do not incorporate the odors from their mother’s amniotic fluid into their self template because the maternal aunt’s amniotic fluid would be a better match with the amniotic fluid the pup gestated in than the amniotic fluid from the paternal aunt. Of course, if there were a preference for the maternal aunt’s amniotic fluid this still would not indicate that pups do incorporate their mother’s amniotic fluid odors into their self template because it is not clear from this result whether the preference is because the odor of the maternal aunt is closer to the familiar odor or whether it is a better match with their learned template. Similar tests could be conducted with rodent species in which females produce mixed paternity litters. Differential responses to odors of paternal aunts as opposed to paternal aunts of half-sibling littermates would provide additional evidence about whether or not rodents incorporate odors of their littermates into their concept of self during gestation. D. Neurophysiological Basis of the G-Ratios Mechanism Because mechanisms for assessing genetic relatedness based on a comparison of manifestations of the genotype of another individual with one’s own would probably be biologically adaptive, it is plausible to speculate that such mechanisms have a discernible neurophysiological basis. The neural architecture of the main olfactory bulb (Buck, 1996) is such that lateral inhibition could fine tune a process of individual odor similarity or overlap at the sensory level before even perceiving the odor. Conducting experiments that involve exposing subjects to odors from individuals of varying degrees of relatedness from siblings to across species would provide at least two types of interesting results. It would be possible to assess spatial patterns of odor-evoked activity in the main olfactory bulb (following Schaefer et al., 2001) that related to genetic similarity between the odor donor and the subject. It would also be possible to assess differential neurophysiological activity in other parts of the olfactory and limbic systems that may be indicative of differential processing based on G-ratios.

VII. Summary Theoretical distinctions among proposed kin recognition mechanisms in rodents are difficult to reconcile with some available data. Ambiguity remains because research on recognition mechanisms was originally driven

124


by kin selection theory but never adequately grounded in behavioral data that could inspire principles to explain observed responses. There is a tendency to design experiments in terms of categorical distinctions, such as kin vs nonkin or conspecifics vs heterospecifics, which may be more useful for researchers than meaningful to the animals. Serendipitous findings helped clarify practical aspects of odor-based mechanisms underlying differential responses to individuals of varying degrees of genetic relatedness and their individual odors. In experiments using habituation– generalization techniques, subjects from multiple species of hamsters, mole rats, and mice consistently, across degrees of relatedness from siblings to different close species, treated the individual odors of two more closely related individuals as similar in quality in comparison with the odor of less closely related individuals. The process by which particular genes are manifest in particular proportions of compounds in individual odors remains unknown, but the genotype of each individual is clearly evident in the odor of that individual. This predictable relationship between genotypes and individual odors, namely, the greater the proportion of genes that two individuals share, the greater the similarity between their individual odors, is termed ‘‘odor–genes covariance.’’ There were two important consequences of these studies for understanding recognition mechanisms. First, differential responses to odors of familiar and unfamiliar individuals indicated that rodents learn to associate particular individuals with their individual odors and can recognize the odors of familiar individuals irrespective of genetic relatedness. Thus ‘‘individual recognition’’ is a mechanism for responding both to kin and nonkin rather than a ‘‘kin recognition’’ process. Second, in conjunction with evidence for self-referencing in graded responses based on degrees of genetic relatedness to odors of kin, populations, and species, the odor–genes covariance findings raised the intriguing possibility that such selfreferencing would be the most practical means of assessing degrees of genetic relatedness to any other individual. Differential responses could occur throughout the spectrum from siblings to across species by comparing the degree of similarity between the odor of the other individual and one’s own odor, that is, ‘‘genetic relatedness assessments through individual odor similarities’’ or G-ratios. Individual odors are individually distinctive composites that also share common qualities with other genetically similar individuals from the same kin group, population, and species. These shared qualities in the odor gestalt enable relatedness assessments rather than specific odor markers of each group. Particular preferences for individuals with similar odors and genotypes that emerge with genetic divergence could serve as a premating ethological isolating mechanism during rodent speciation. Such a mechanism may help incipient


125

rodent species remain genetically distinct without the necessity of speciesspecific odor signals. Future studies should determine the breadth of these mechanisms, the neurophysiological basis of differential responses, the extent to which they are innate or learned, and their robustness in the face of transient factors, such as diet and motivational state, that may alter the qualities of individual odors.

Acknowledgments We thank R. Porter, Z. Tang-Martinez, A. Templeton, C. Snowdon, and T. Roper for insightful comments and suggestions on the earlier draft of this manuscript.

References Barnard, C. (1990). Kin recognition: Problems, prospects and the evolution of discrimination systems. Adv. Study Behav. 19, 29–81. Barnard, C., Hurst, J. L., and Aldous, P. (1991). Of mice and kin: The functional significance of kin bias in social behaviour. Biol. Rev. 66, 379–430. Bateson, P. (1983). Optimal outbreeding. In ‘‘Mate Choice’’ (P. P. G. Bateson, Ed.), pp. 257–277. Cambridge University Press, Cambridge. Beauchamp, G. K., Gilbert, A. N., Yamazaki, K., and Boyse, E. A. (1986). Genetic basis for individual discriminations: The major histocompatibility complex of the mouse. In ‘‘Chemical Signals in Vertebrates,’’ Vol. 4: ‘‘Ecology, Evolution, and Comparative Biology’’ (D. Duvall, D. Mu¨ller-Schwarze, and R. M. Silverstein, Eds.), pp. 413–422. Plenum, New York. Beauchamp, G. K., Yamazaki, K., Duncan, H., Bard, J., and Boyse, E. A. (1990). Genetic determination of individual mouse odour. In ‘‘Chemical Signals in Vertebrates,’’ Vol. V (D. W. MacDonald, D. Mu¨ller-Schwarze, and S. E. Natynczuk, Eds.), pp. 244–254. Oxford University Press, Oxford. Brown, R. E. (1995). What is the role of the immune system in determining individually distinct body odours? Int. J. Immunopharmacol. 17, 655–661. Brown, R. E., Singh, P. B., and Roser, B. (1987). The major histocompatibility complex and the chemosensory recognition of individuality in rats. Physiol. Behav. 40, 65–73. Brown, R. E., Roser, B., and Singh, P. B. (1990). The MHC and individual odours in rats. In ‘‘Chemical Signals in Vertebrates,’’ Vol. V (D. W. MacDonald, D. Mu¨ller- Schwarze, and S. E. Natynczuk, Eds.), pp. 228–243. Oxford University Press, Oxford. Buck, L. B. (1996). Information coding in the vertebrate olfactory system. Annu. Rev. Neurosci. 19, 517–544. Burda, H. (1995). Individual recognition and incest avoidance in eusocial common mole-rats rather than reproductive suppression by parents. Experientia. 51, 411–413. Burda, H. (1999). Syndrome of eusociality in African subterranean mole-rats (Bathyergidae, Rodentia), its diagnosis and aetiology. In ‘‘Evolutionary Theory and Processes: Modern Perspectives’’ (S. P. Wasser, Ed.), pp. 385–418. Kluwer Academic, Dordrecht, The Netherlands. Burda, H., Zima, J., Scharff, A., Macholan, M., and Kawalika, M. (1999). The karyotypes of Cryptomys anselli sp. nova and Cryptomys kafuensis sp. nova: New species of the common mole-rat from Zambia (Rodentia, Bathyergidae). Z. Saugetierkd. 64, 36–50.

126


Butlin, R. (1995). Genetic variation in mating signals and responses. In ‘‘Speciation and the Recognition Concept’’ (D. M. Lambert and H. G. Spencer, Eds.), pp. 327–366. Johns Hopkins University Press, Baltimore, MD. Cohen, L. B. (1976). Habituation of infant visual attention. In ‘‘Habituation’’ (T. J. Tighe and R. N. Leaton, Eds.), pp. 207–238. Lawrence Erlbaum, Hillsdale, NJ. D’Amato, F. R. (1994). Physiological evidence for genetically mediated sibling recognition in mice. Behav. Genet. 24, 493–496. Dawkins, R. (1982). ‘‘The Extended Phenotype.’’ W. H. Freeman, San Francisco. Doty, R. (1986). Odor-guided behavior in mammals. Experientia. 42, 257–270. Eggert, F., Holler, C., Luszyk, D., Mu¨ller-Ruchholtz, W., and Ferstl, R. (1996). MHCassociated and MHC-independent urinary chemosignals in mice. Physiol. Behav. 59, 57–62. Ferkin, M. H., Sorokin, E. S., Johnston, R. E., and Lee, C. J. (1997). Attractiveness of scents varies with protein content of the diet of meadow voles. Anim. Behav. 53, 133–141. Filippucci, M. G., Kawalika, M., Macholan, M., Scharff, A., and Burda, H. (1997). Allozyme differentiation and systematic relationship of Zambian giant mole-rats, Cryptomys mechowi (Bathyergidae, Rodentia). Z. Saugetierkd. 62, 172–178. Fletcher, D. J. C., and Michener, C. D. (Eds.) (1987). ‘‘Kin Recognition in Animals.’’ John Wiley & Sons, Chichester, UK. Futuyma, D. J. (1998). ‘‘Evolutionary Biology.’’ Sinauer Associates, Sunderland, MA. Gheusi, G., Bluthe, R., Goodall, G., and Dantzer, R. (1994). Social and individual recognition in rodents: Methodological aspects and neurobiological bases. Behav. Proc. 33, 59–88. Gilder, P. M., and Slater, P. J. B. (1978). Interest of mice in conspecific male odours is influenced by degree of kinship. Nature 274, 364–365. Godfrey, J. (1958). The origin of sexual isolation between bank voles. Proc. R. Phys. Soc. Edinb. 27, 47–55. Gorman, M. L. (1976). A mechanism for individual recognition by odour in Herpestes auropunctatus (Carnivora: Viverridae). Anim. Behav. 24, 141–145. Gorman, M. L., Nedwell, D. B., and Smith, R. M. (1974). An analysis of the contents of the anal scent pockets of Herpestes auropunctatus (Carnivora: Viverridae). J. Zool. 172, 389–399. Grafen, A. (1990). Do animals really recognize kin? Anim. Behav. 39, 42–54. Grau, J. H. (1982). Kin recognition in white-footed deermice (Peromyscus leucopus). Anim. Behav. 30, 497–505. Halpin, Z. T. (1986). Individual odors among mammals: Origins and functions. Adv. Study Behav. 16, 40–70. Halpin, Z. T. (1991). Kin recognition cues in vertebrates. In ‘‘Kin Recognition’’ (P. G. Hepper, Ed.), pp. 220–258. Cambridge University Press, Cambridge. Halpin, Z. T., and Hoffman, M. D. (1987). Sibling recognition in the white-footed mouse, Peromyscus leucopus: Association or phenotype matching? Anim. Behav. 35, 563–570. Hamilton, W. D. (1964a). The genetical evolution of social behaviour. I. J. Theor. Biol. 7, 1–16. Hamilton, W. D. (1964b). The genetical evolution of social behaviour. II. J. Theor. Biol. 7, 17–52. Hauber, M. E., Sherman, P. W., and Paprika, D. (2000). Self-referent phenotype matching in a brood parasite: The armpit effect in brown-headed cowbirds (Molothrus ater). Anim. Cogn. 3, 113–117. Hauber, M. E., Russo, S. A., and Sherman, P. W. (2001). A password for species recognition in a brood-parasitic bird. Proc. R. Soc. Lond. B Biol. Sci. 268, 1041–1048. Hepper, P. G. (1986). Kin recognition: Functions and mechanisms. Biol. Rev. 61, 63–93.


127

Hepper, P. G. (1987). The discrimination of different degrees of relatedness in the rat: Evidence for a genetic identifier? Anim. Behav. 35, 549–554. Hepper, P. G. (1990). Foetal olfaction. In ‘‘Chemical Signals in Vertebrates,’’ Vol. V (D. W. MacDonald, D. Mu¨ller-Schwarze, and S. E. Natynczuk, Eds.), pp. 282–288. Oxford University Press, Oxford. Hepper, P. G. (Ed.). (1991). ‘‘Kin Recognition.’’ Cambridge University Press, Cambridge. Heth, G., and Nevo, E. (1981). Origin and evolution of ethological isolation in subterranean mole rats. Evolution 35, 259–274. Heth, G., and Todrank, J. (1995). Assessing chemosensory perception in subterranean mole rats: Different responses to smelling versus touching odourous stimuli. Anim. Behav. 49, 1009–1015. Heth, G., and Todrank, J. (2000). Individual odour similarities across species parallel phylogenetic relationships in the S. ehrenbergi superspecies of mole rats. Anim. Behav. 60, 789–795. Heth, G., Beauchamp, G. K., Nevo, E., and Yamazaki, K. (1996a). Species, population, and individual specific odortypes in urine of the blind subterranean mole rat (Spalax ehrenbergi) detected by laboratory rats. Chemoecology 7, 107–111. Heth, G., Nevo, E., and Todrank, J. (1996b). Seasonal changes in urinary odors and in responses to them by subterranean mole rats. Physiol. Behav. 60, 963–968. Heth, G., Todrank, J., and Johnston, R. E. (1998). Kin recognition in golden hamsters: Evidence for phenotype matching. Anim. Behav. 56, 409–417. Heth, G., Todrank, J., and Johnston, R. E. (1999). Similarity in the qualities of individual odors among kin and species in Turkish (Mesocricetus brandti) and golden (Mesocricetus auratus) hamsters. J. Comp. Psychol. 113, 321–326. Heth, G., Todrank, J., Busquet, N., and Baudoin, C. (2001). Odour–genes covariance and differential investigation of individual odours in the Mus species complex. Biol. J. Linn. Soc. Lond. 73, 213–220. Heth, G., Todrank, J., and Burda, H. (2002). Similarity in the qualities of individual odors within colonies and across species of African eusocial mole rats (Cryptomys spp.). J. Mammal. 83, 569–575. Heth, G., Todrank, J., Busquet, N., and Baudoin, C. (2003). Mice smell degrees of relatedness: Genetic relatedness assessment through individual odour similarities. (in press). Holmes, W. G. (1986). Kin recognition by phenotype matching in two species of ground squirrels. Anim. Behav. 34, 129–136. Holmes, W. G., and Sherman, P. W. (1982). The ontogeny of kin recognition in female Belding’s ground squirrels. Am. Zool. 22, 38–47. Holmes, W. G., and Sherman, P. W. (1983). Kin recognition in animals. Am. Sci. 71, 46–55. Isles, A. R., Baum, M. J., Ma, D., Keverne, E. B., and Allen, N. D. (2001). Genetic imprinting: Urinary odour preferences in mice. Nature 409, 783–784. Johnston, R. E. (1975). Scent marking in male hamsters. II. The role of flank gland odor in the causation of marking. Z. Tierpsychol. 37, 138–144. Johnston, R. E. (1977). The causation of two scent-marking behaviour patterns in female hamsters (Mesocricetus auratus). Anim. Behav. 25, 317–327. Johnston, R. E. (1990). Chemical communication in golden hamsters: From behaviour to molecules and neural mechanisms. In ‘‘Contemporary Issues in Comparative Psychology’’ (D. A. Dewsbury, Ed.), pp. 381–409. Sinauer, Sunderland, MA. Johnston, R. E. (1998). Pheromones, the vomeronasal system, and communication. Ann. N. Y. Acad. Sci. 855, 333–348. Johnston, R. E., and Brenner, D. (1982). Species-specificity in scent marking in hamsters. Behav. Neural Biol. 35, 46–55.

128


Johnston, R. E., and Jernigan, P. (1994). Golden hamsters recognize individuals, not just individual scents. Anim. Behav. 48, 129–136. Johnston, R. E., Derzie, A., Chiang, G., Jernigan, P., and Lee, H.-C. (1993). Individual scent signatures in golden hamsters: Evidence for specialization of function. Anim. Behav. 45, 1061–1070. Kareem, A. M. (1983). Effect of increasing periods of familiarity on social interactions between male sibling mice. Anim. Behav. 31, 919–926. Kareem, A. M., and Barnard, C. J. (1982). The importance of kinship and familiarity in social interactions between mice. Anim. Behav. 30, 594–601. Komdeur, J., and Hatchwell, B. J. (1999). Kin recognition: Function and mechanisms in avian species. Trends Ecol. Evol. 14, 237–241. Lenington, S., Coopersmith, C., and Williams, J. (1992). Genetic basis of mating preferences in wild house mice. Am. Zool. 32, 40–47. Lin, D. M., and Ngai, J. (1999). Development of the vertebrate main olfactory system. Curr. Opin. Neurobiol. 9, 74–78. Mateo, J., and Johnston, R. E. (2000a). Kin recognition and the ‘‘armpit effect’’: Evidence of self-referent matching. Proc. R. Soc. Lond. B Biol. Sci. 267, 695–700. Mateo, J., and Johnston, R. E. (2000b). Retention of social recognition after hibernation in Belding’s ground squirrels. Anim. Behav. 59, 491–499. Mayr, E. (1963). ‘‘Animal Species and Evolution.’’ Harvard University Press, Cambridge, MA. Mayr, E. (1982). Processes of speciation in animals. In ‘‘Mechanisms of Speciation’’ (C. Barigozzi, Ed.), pp. 1–19. Alan R. Liss, New York. Moore, R. E. (1965). Olfactory discrimination as an isolating mechanism between Peromyscus maniculatus and Peromyscus polionotus. Am. Mid. Nat. 73, 85–100. Murphy, M. R. (1977). Intraspecific sexual preferences of female hamsters. J. Comp. Psychol. 91, 1337–1346. Murphy, M. R. (1978). Oestrus Turkish hamsters display lordosis toward conspecific males but attack heterospecific males. Anim. Behav. 26, 311–312. Murphy, M. R. (1980). Sexual preferences of male hamsters: Importance of preweaning and adult experience, vaginal secretion, and olfactory or vomeronasal sensation. Behav. Neural Biol. 30, 323–340. Nevo, E., Ivanitskaya, E., and Beiles, A. (2001). ‘‘Adaptive Radiation of Blind Subterranean Mole Rats: Naming and Revisiting the Four Sibling Species of the Spalax ehrenbergi Superspecies in Israel: Spalax galili (2n = 52), S. golani (2n = 54), S. carmeli (2n = 58), and S. judaei (2n = 60).’’ Backhuys Publishers, Leiden, the Netherlands. Paz y Minõ, G., and Tang-Martinez, Z. (1999a). Effects of exposures to siblings or sibling odors on sibling recognition in prairie voles. Can. J. Zool. 77, 118–123. Paz y Minõ, G., and Tang-Martinez, Z. (1999b). Effects of isolation on sibling recognition in prairie voles, Microtus ochrogaster. Anim. Behav. 57, 1091–1098. Porter, R. H. (1988). The ontogeny of sibling recognition in rodents: The superfamily Muroidea. Behav. Genet. 18, 483–494. Porter, R. H., and Winberg, J. (1999). Unique salience of maternal breast odors for newborn infants. Neurosci. Biobehav. Rev. 23, 439–449. Porter, R. H., and Wyrick, M. (1979). Sibling recognition in spiny mice (Acomys cahirinus): Influence of age and isolation. Anim. Behav. 27, 761–766. Porter, R. H., Matochik, J. A., and Makin, J. W. (1983). Evidence for phenotype matching in spiny mice (Acomys cahirinus). Anim. Behav. 31, 978–984. Porter, R. H., McFadyen-Ketchum, S. A., and King, G. A. (1989). Underlying bases of recognition signatures in spiny mice, Acomys cahirinus. Anim. Behav. 37, 638–644.


129

Potts, W. K., Manning, C. J., and Wakeland, E. K. (1991). Matings of seminatural populations of mice influenced by MHC genotype. Nature 352, 619–621. Potts, W. K., Manning, C. J., and Wakeland, E. K. (1992). MHC-based mating preferences in Mus operate through both settlement patterns and female controlled extra-territorial matings. In ‘‘Chemical Signals in Vertebrates,’’ Vol. 6 (R. Doty and D. Mu¨ller-Schwarze, Eds.), pp. 183–187. Plenum, New York. Rich, T. J., and Hurst, J. L. (1998). Scent marks as reliable signals of the competitive ability of mates. Anim. Behav. 56, 727–735. Schaal, B., Orgeur, P., and Marlier, L. (1994). Amniotic fluid odor in neonatal adaptation: A summary of recent research in mammals. Adv. Biosci. 93, 239–245. Schaal, B., Orgeur, P., and Arnould, C. (1995). Olfactory preferences in newborn lambs: Possible influences of prenatal experience. Behavior 132, 351–365. Schaal, B., Marlier, L., and Soussignan, R. (1998). Olfactory function in the human fetus: Evidence from selective neonatal responsiveness to the odor of amniotic fluid. Behav. Neurosci. 112, 1438–1449. Schaefer, M. L., Young, D. A., and Restrepo, D. (2001). Olfactory fingerprints for major histocompatibility complex determined body odors. J. Neurosci. 21, 2481–2487. Schellinck, H. M., and Brown, R. E. (1999). Searching for the source of urinary odor individuality in rodents. In ‘‘Advances in Chemical Communication in Vertebrates’’ (R. E. Johnston, D. Mu¨ller-Schwarze, and P. Sorensen, Eds.), pp. 267–280. Plenum, New York. Schellinck, H. M., Monahan, E., Brown, R. E., and Maxson, S. C. (1993). A comparison of the contribution of the major histocompatibility complex (MHC) and Y chromosomes to the discriminability of individual urine odors of mice by Long-Evans rats. Behav. Genet. 23, 257–263. Sherman, P. W., and Holmes, W. G. (1985). Kin recognition: Issues and evidence. In ‘‘Experimental Behavioral Ecology and Sociobiology’’ (B. Holldobler and M. Lindauer, Eds.), pp. 437–460. G. Fischer Verlag, Stuttgart, Germany. Sherman, P. W., Reeve, H. K., and Pfenning, D. W. (1997). Recognition systems. In ‘‘Behavioural Ecology: A Behavioural Approach’’ (J. R. Krebs and N. B. Davies, Eds.), pp. 69–96. Blackwell, London. Singer, A., Beauchamp, G. K., and Yamazaki, K. (1997). Volatile signals of the major histocompatibility complex in male mouse urine. Proc. Natl. Acad. Sci. USA 94, 2210–2214. Slater, P. J. B. (1994). Kinship and altruism. In ‘‘Behaviour and Evolution’’ (P. J. B. Slater and T. R. Halliday, Eds.), pp. 193–222. Cambridge University Press, Cambridge. Smith, M. H. (1965). Behavioral discrimination shown by allopatric and sympatric males of Peromyscus eremicus and Peromyscus californicus between females of the same two species. Evolution 19, 430–435. Spelke, E. S. (1985). Preferential-looking methods as tools for the study of cognition in infancy. In ‘‘Measurement of Audition and Vision in the First Year of Post-Natal Life: A Methodological Overview’’ (G. Gottlieb and N. Krasnegor, Eds.), pp. 323–363. Ablex, Norwood, NJ. Tang-Martinez, Z. (2001). The mechanisms of kin discrimination and the evolution of kin recognition in vertebrates: A critical re-evaluation. Behav. Proc. 53, 21–40. Terry, L. M., and Johanson, I. B. (1996). Effects of altered olfactory experiences on the development of infant rats’ responses to odors. Dev. Psychobiol. 29, 353–377. Todrank, J., and Heth, G. (1996). Individual odors in two chromosomal species of blind, subterranean mole rat (Spalax ehrenbergi): Conspecific and cross-species discrimination. Ethology 102, 806–811.

130


Todrank, J., and Heth, G. (2001). Re-thinking cross-fostering designs for studying kin recognition mechanisms. Anim. Behav. 61, 503–505. Todrank, J., Heth, G., and Johnston, R. E. (1998). Kin recognition in golden hamsters: Evidence for kinship odours. Anim. Behav. 55, 377–386. Todrank, J., Heth, G., and Johnston, R. E. (1999). Social interaction is necessary for discrimination and memory for odours of close relatives in golden hamsters. Ethology 105, 771–782. Waldman, B. (1987). Mechanisms of kin recognition. J. Theor. Biol. 128, 159–185. Waldman, B., Frumhoff, P. C., and Sherman, P. W. (1988). Problems of kin recognition. Trends Ecol. Evol. 3, 8–13. Yamazaki, K., Beauchamp, G. K., Matsuzaki, O., Bard, J., Thomas, L., and Boyse, E. A. (1986). Participation of the murine X and Y chromosomes in genetically determined identity. Proc. Natl. Acad. Sci. USA 83, 4437–4440. Yamazaki, K., Beauchamp, G. K., Bard, J., and Boyse, E. A. (1990). Single MHC gene mutations alter urine odour constitution in mice. In ‘‘Chemical Signals in Vertebrates,’’ Vol. V (D. W. MacDonald, D. Mu¨ller-Schwarze, and S. E. Natynczuk, Eds.), pp. 255–259. Oxford University Press, Oxford. Yamazaki, K., Beauchamp, G. K., Bard, J., Boyse, E. A., and Thomas, L. (1991). Chemsensory identity and immune function in mice. In ‘‘Chemical Senses: Genetics of Perception and Communication’’ (C. J. Wysocki and M. R. Kare, Eds.), pp. 211–225. Marcel Dekker, New York. Yamazaki, K., Beauchamp, G. K., Imai, Y., Bard, J., Thomas, L., and Boyse, E. A. (1992). MHC control of odortypes in the mouse. In ‘‘Chemical Signals in Vertebrates,’’ Vol. 6 (R. Doty and D. Mu¨ller-Schwarze, Eds.), pp. 189–196. Plenum, New York. Yamazaki, K., Beauchamp, G. K., Singer, A., Bard, J., and Boyse, E. A. (1999a). Odortypes: Their origin and composition. Proc. Natl. Acad. Sci. USA 96, 1522–1525. Yamazaki, K., Beauchamp, G. K., Curran, M., Bard, J., and Boyse, E. A. (1999b). Parent– progeny recognition as a function of MHC odortype identity. Proc. Natl. Acad. Sci. USA 97, 10500–10502.


Sex Role Reversal in Pipefish Anders Berglund1 and Gunilla Rosenqvist2 1

department of animal ecology uppsala university se-752 36 uppsala, sweden 2 department of biology norwegian university of science and technology n-7491 trondheim, norway

I. Mate Competition and Sex Roles Sexual selection operates on traits that increase reproductive success, either by increasing mating success or by increasing partner quality (Andersson, 1994). Sexual selection that increases mating success usually operates more strongly in males than in females, with males consequently possessing more elaborate sexually selected traits (Darwin, 1871). The main reason for this is that in most species females incur a higher cost for a single breeding attempt than males (Kokko and Monaghan, 2001), that is, females make a larger parental investment than males (Trivers, 1972). This in turn gives males a higher potential reproductive rate than females (i.e., males can reproduce faster than females given that number of mates is not limiting), with the consequence that the operational sex ratio (OSR) becomes biased toward an excess of males (Clutton-Brock and Parker, 1992; Clutton-Brock and Vincent, 1991; Kvarnemo and Ahnesjo¨, 1996; Parker and Simmons, 1996; Reynolds, 1996). The OSR is the proportion of males to females willing to mate (Emlen, 1976; Emlen and Oring, 1977), and provides a snapshot predicting mating competition at a certain time and place. Potential reproductive rates (or their equivalents, time in and out of reproduction; Parker and Simmons, 1996) better reflect mating competition in a population during a specific reproductive event, and are a population-level estimate as competition occurs within groups of animals. Therefore, the potential reproductive rates of all males and females in the population must be included if we wish to use this measure to predict mating competition [but see Ahnesjo¨ et al. (2001) for a mathematically 131 Copyright 2003 Elsevier Science (USA). All rights reserved. 0065-3454/03 $35.00

132

ANDERS BERGLUND AND GUNILLA ROSENQVIST

slightly different approach]. A male surplus in relation to number of females, or a male surplus in relation to other resources necessary for breeding, causes competition among males for the resource in short supply, and male reproductive success is therefore determined mainly by the number of copulations performed. Female reproductive success, on the other hand, generally depends on access to resources required to produce offspring and on the quality of the male (genetic or, when applicable, his ability to provide resources). Females hence may have good reasons to be choosy in selecting a partner. Thus, the form of sexual selection that increases reproductive success through partner quality usually applies more to females than males, resulting in females evolving preferences for specific male traits. Consequently, we usually expect a closer association between number of matings and number of progeny in males than in females. This so-called Bateman gradient (Bateman, 1948) was originally demonstrated in the fruit fly Drosophila melanogaster, using visible genetic markers. A strong correlation was evident in males, but not in females, that is, sexual selection operated on males in this species (Bateman, 1948). In some unusual and highly interesting species sex roles are reversed: here, females ready to mate outnumber willing males, and hence females compete over access to males. Can potential reproductive rates explain this biased OSR? In other words, are females of such species faster reproducers than males? In this chapter we answer this question with a ‘‘yes’’ by reviewing research targeted at exactly this problem in two sex rolereversed pipefish species. Moreover, we discuss why potential reproductive rates differ as well as the consequences of this ‘‘reversed’’ mating competition on sexually selected traits in females. We also outline the costs and benefits of these sexually selected traits and mating preferences to allow for a better understanding of sexual selection processes in sex role-reversed species. Through this we also intend to better understand the very same processes in species with conventional sex roles.

II. Female Ornaments Does sexual selection produce the same type of ornaments/weapons in sex role-reversed females as in males of ‘‘conventional’’ species? Not necessarily: concerning parental investment in offspring, females always invest energy and cytoplasm in the egg, whereas males produce a relatively low-energy (‘‘cheap’’) gamete. In fact, males may pass on nothing but genes to their offspring. Such nonpaternal males may suffer a reduction in condition and/or survival due to the development and maintenance of

SEX ROLE REVERSAL IN PIPEFISH

133

sexual characters. Females and brood-caring males, however, who both invest substantial amounts of energy into offspring, face yet another cost: a reduction in egg number, egg quality, and/or parental investment, because resources are taken from reproduction to that character (Fitzpatrick et al., 1995). Females that invest in sexual ornaments do this at the expense of their reproductive potential (fewer or smaller eggs), and males that choose such ornamented females may compromise their own reproductive success by doing so. In nonpaternal males the sexual selection process will be opposed and finally brought to an end by counteracting natural selection in terms of decreased condition (or, eventually, increased mortality) imposed by the sexually selected trait. Females and investing males suffer a second constraint on ornament evolution: mating advantages incurred from the ornament must compensate not only a decrease in condition but also costs to potential fecundity in terms of a reduction in egg and/or parental care quantity or quality (Fitzpatrick et al., 1995). In view of this trade-off, sexually selected female characters should be energetically cheap. On the other hand, we expect such characters to be costly if they are to honestly indicate female quality, so that males benefit by being choosy. If the character is not honest, then males will not benefit by selecting mates expressing this signal. Males not paying attention to dishonest signals no longer incur the costs associated with being choosy. Hence, female sexually selected traits need to be honest to attract attention, and honesty may require that the signal be associated with a cost (the handicap principle; Zahavi and Zahavi, 1997). Therefore, we must look for signals that directly indicate quality, or signals with costs in other forms than energy. We can imagine three kinds.

A. Body Size Signal Strictly speaking, body size need not be a signal at all, as it has not primarily evolved to change the behavior of others. Still, information from body size assessment can be of use, and such an informative trait is sometimes called a ‘‘cue’’ (Hasson, 1997) or a ‘‘revealing indicator’’ (Iwasa et al., 1991; Johnstone, 1995). Signals and cues need not represent clear-cut dichotomies, however: body size may have been further modified to signal dominance or attractiveness, and is hence, at least in part, also a signal. Characters such as body size, concerning which an allocation to the character may also be an allocation to reproduction, may be common cues/ signals in sex role-reversed species. Size is difficult to fake convincingly, and need not compromise fecundity; in fact, it actually correlates positively with fecundity in most cold-blooded and many other species.

134


B. Dietary Pigment Signal Signals can be both energetically cheap and honest if different components of the resource in question are used for the character and for reproduction. An example may be when color pigments in the food are used to produce sexual colorings, while the energetic bulk of the food goes into reproduction. Such a signal, honest by design and sometimes called an ‘‘index’’ (Hasson, 1997, 1999; Taylor et al., 2000), cannot be faked, as its expression depends on the level of food intake, and thus honestly signals nutritional status. As most of the energy in the food goes to reproduction the energetic cost of the pigment itself may be small. Alternatively, the allocation to reproduction and to the ornament can be partitioned in time, so that the production of the ornament does not interfere with the production of eggs or offspring care.

C. Socially Costly Signals Signals may be costly in terms of something other than energy. Color patches signaling dominance (‘‘status badges’’) may maintain honesty not through an energetic but through a social cost. A cheater pays the cost when it encounters an opponent with a similar-size badge, and must prove its worth in a real encounter. As the cheater is bound to lose that fight, status badges are honest (Boake and Capranica, 1982; Jones, 1990; Ja¨rvi and Bakken, 1984; Møller, 1987a; Qvarnstro¨m, 1997; Studd and Robertson, 1985). For example, male birds suffered badly in real contests when their badge was experimentally exaggerated (Møller, 1987b, 1988; Rohwer, 1977; Rohwer and Rohwer, 1978). Status signals can also increase disease susceptibility by reducing immunocompetence as a consequence of elevated testosterone levels (Folstad and Karter, 1992; Owens and Hartley, 1991; Zuk et al., 1990). Maynard-Smith and Harper (1988) concluded from a model that honest communication is evolutionarily stable even if the badge is energetically cheap, provided that a dishonest signaler pays the full cost of a contest. A dishonest mutant can invade the population only if it can escape from contests with a more aggressive opponent without fighting (but see Johnstone and Norris, 1993).

III. Syngnathid Phylogeny Because sex role reversal and female ornamentation seem intimately linked with paternal care, we briefly outline how that paternal care may have evolved in syngnathids (i.e., pipefish and seahorses). After that, we discuss


135

how sex role reversal in pipefish is ultimately a consequence of an extreme level of paternal care in combination with a polygamous mating pattern. Teleost fish are exceptional in both the variety of parental care provided and the fact that males usually do all the caring (Forsgren et al., 2002; Gross and Sargent, 1985). Within teleosts, male syngnathids are remarkable in having advanced adaptations for paternal care by carrying the eggs on their bodies while nursing them. This extreme and costly form of care confirms the male paternity (e.g., A. G. Jones et al., 1998a, 1999). Why did this exaggerated form of parental care evolve? As a member of the same order (Gasterosteiformes) as the syngnathids, stickleback males are known to glue their eggs within a nest. Under circumstances of high nest predation, males of a hypothetical ancestral species may have attached some eggs onto themselves, thereby securing the survival of at least a few young. This ‘‘innovation’’ may have provided the starting point for what we now observe as perhaps the most remarkable specialization for paternal care in any animal. Pipefish presently exhibit what seems to be a gradient of paternal care elaboration: eggs may be brooded openly on the male’s abdomen, brooded in individual membranous egg compartments, brooded in a partly closed pouch consisting of pouch plates, brooded in a fully closed pouch consisting of two folds, or brooded in a fully closed saclike pouch. This gradient has been used to construct a phylogeny (Herald, 1959), largely confirmed by molecular evidence (Wilson et al., 2001). The close link between speciation and brood pouch specialization identifies the brood pouch as a key evolutionary innovation (Wilson et al., 2001). In fact, there are some 230 syngnathid species in the world (Dawson, 1985) compared with only 7 stickleback species (Wootton, 1984), further suggesting a link between male care elaboration and speciation. Regardless of brooding type, all syngnathid males seem to transfer nutrients to their offspring (Berglund et al., 1986b; Haresign and Schumway, 1981; Linton and Soloff, 1964). In species with fully closed pouches (such as Syngnathus and seahorses, Hippocampus), males do this via a placenta-like arrangement and also oxygenate and osmoregulate their young. In species with a more ‘‘primitive’’ brooding type (e.g., Nerophis ophidion) males invest less in their young compared with species with more advanced pouches (e.g., Syngnathus typhle; Berglund et al., 1986b).

IV. Sex Roles in Syngnathids Sex role reversal seems to have evolved independently several times in syngnathids (A. B. Wilson, I. Ahnesjo¨, A. C. J. Vincent, and A. Meyer,

136


unpublished mitochondrial DNA-based phylogeny), and is independent of the degree of brood pouch development (Berglund et al., 1986a,b; Vincent et al., 1992; A. B. Wilson, I. Ahnesjo¨, A. C. J. Vincent, and A. Meyer, unpublished data). Thus, N. ophidion with its more ‘‘primitive’’ care is markedly sex role reversed, whereas the seahorse Hippocampus whitei exhibits conventional roles (Vincent, 1994) in spite of its elaborate care. Instead, sex roles correlate with mating patterns: polygamous species, such as Syngnathus pipefishes (Jones and Avise, 1997; Jones et al., 1999), are sex rolereversed, whereas monogamous species, such as seahorses (A. G. Jones et al., 1998; Kvarnemo et al., 2000; Masonjones and Lewis, 2000; Vincent and Sadler, 1995), have conventional sex roles (Masonjones and Lewis, 2000; Vincent, 1994). Thus, a substantial paternal expenditure in offspring seems to be a necessary but not sufficient prerequisite for sex role reversal. Seahorses, being monogamous, experience mating competition only at the onset of the breeding season. At this time, males have empty pouches while females need time to mature eggs. Thus, males can potentially reproduce faster than can females at the time of pair formation, and sex roles become conventional (Masonjones and Lewis, 2000; Vincent et al., 1992). In some species, factors such as low mobility or low mate encounter rates reduce the potential for polygamy. Instead, reproductive efficiency through monogamy may be promoted, allowing females to take time maturing eggs. This, in turn, affects potential reproductive rates and operational sex ratios at the start of the breeding season in such a way that sex roles become conventional. However, exceptions may exist, such as the apparently monogamous pipefish Corythoichthys haematopterus, which reportedly shows female-biased OSRs and sex role reversal due to high female mortality (Matsumoto and Yanagisawa, 2001). Thus, the link between parental investment and potential reproductive rate may be substantially modified by mating patterns, highlighting the need to understand the factors shaping such patterns.

V. The Two Pipefish Species Syngnathus typhle and N. ophidion are pipefish found in shallow eelgrass (Zostera marina) meadows along the coasts of Europe. These pipefish are suction feeders, hunting mainly small crustaceans such as mysids and shrimps by eye. Predators include sculpins (Myoxocephalus scorpius), eels (Anguilla anguilla), cod (Gadus morhua), and various sea birds. The pipefish live for 2 to 3 years, and mature in their first year. They are extremely cryptic in shape, color, and behavior, closely resembling the eelgrass where they live. Individuals align themselves vertically within the eelgrass, where their body colors closely approximate the surrounding


137

substrate (Vincent et al., 1994), but mate-seeking and courtship behaviors disturb the resemblance between pipefish and their visual background, rendering them conspicuous. Mating is thus potentially dangerous, exposing the fish to predation. After a conspicuous, lengthy, and ritualized mutual dance, the female may transfer eggs to the male’s body, where he fertilizes them. Males thus have a paternity confidence of 100% (Jones et al., 1999). In both species the resulting embryos are nourished by the male, and in S. typhle the male also oxygenates and osmoregulates them in his pouch until they are born several weeks later (Berglund et al., 1986a,b; Fiedler, 1954; Haresign and Schumway, 1981). The male pregnancy is accompanied by hormonal changes. As male brooding advances, plasma levels of 11ketotestosterone drop dramatically in S. typhle (Mayer et al., 1993). Such a drop is typical in teleost fish with male parental care. Moreover, the level of 17-estradiol is usually high in early breeding male S. typhle, but low in females, a partial reversal of the typical pattern in fish (Mayer et al., 1993). The two pipefish species differ in the degree of sexual dimorphism. Nerophis ophidion is more dimorphic than S. typhle: N. ophidion females are much larger than the males, whereas in S. typhle the sexes are more similar in size. Moreover, N. ophidion females have sexual characteristics in the form of a permanent blue coloring along their sides and a ventral skin fold that develops during the breeding season. In S. typhle, females display only a temporary color pattern. The question why these species differ in sexual dimorphism is complicated by differences in mating constraints between them. In pipefish species without a pouch, the male receives all the eggs he will carry at once from only one female. Such males should benefit from choosing a fecund female, which would then result in more intense sexual selection on the females of these species. Nerophis ophidion may be an example of this, and the large body size and vivid coloration of females may be the outcome of intense sexual selection. Among promiscuous species, a male with a brood pouch can receive eggs from several females, and a female may give eggs to several males (Berglund et al., 1988, 1989; Jones et al., 2000b). The penalty to males for making the ‘‘wrong’’ choice is in this case much lower, as males can at least partially compensate for matings with low-quality females by subsequently mating with high-quality females. This may reduce the intensity of sexual selection among females, generating the prediction that brood pouch species should be less dimorphic than open brooders, all else being equal. A. Syngnathus typhle In nature, male S. typhle actively choose among and reject some females, whereas females vigorously display, often in temporary groups in a leklike

138


fashion. Males typically swim within the eelgrass, searching out females who display by swimming up and down well above the eelgrass. Once such a displaying female group is found, males may or may not dance, and perhaps subsequently mate, with a particular, often large, female (Vincent et al., 1994, 1995). Females compete for matings during such group displays, and females may herd other females off from the male, thus prolonging the courtship phase (Vincent et al., 1995) and increasing predation risk (Berglund, 1993; Fuller and Berglund, 1996). The lengthy and ritualized mutual dance always precedes copulation. The dance includes wriggling and shaking movements as well as rising above the eelgrass (Fiedler, 1954). The dance may terminate by a copulation, in which the female transfers her eggs to the male’s brood pouch by means of a ‘‘penis,’’ while the couple ascends. Thereafter the male, assuming a stiff S-shaped posture while sinking to the bottom, fertilizes the eggs (Fiedler, 1954). Males may brood eggs from one or several females, and a female may, within a short time span, transfer eggs to several males. Large females mate with more males than do small females, possibly thereby reducing sib–sib competition within the pouch (Ahnesjo¨, 1996; Berglund et al., 1988), or simply because they are more successful in mating competition. Males, on the other hand, are more inclined to mate multiple times when encountering only small, as opposed to only large, females (Jones et al., 2000b). Males provide offspring with nutrients and oxygen during pregnancy. The pregnancy is costly to the males: they no longer hunt actively and consequently feed less and grow more slowly than females (Svensson, 1988). This decreases future reproductive success, as male fecundity is correlated to size (Berglund et al., 1986a,b). However, pregnant males do not fall victim to predators more than females (Svensson, 1988). During pregnancy, males seem to trade offspring growth with offspring numbers (Ahnesjo¨, 1992b), and male reproductive success becomes limited by the eggs they have received and their ability to brood them (Ahnesjo¨, 1992a). Offspring appear to compete for resources within the pouch, with the larger offspring being more successful than the smaller (Ahnesjo¨, 1996). The pregnancy ends with the young fish leaving the pouch, thereafter leading independent lives. The adult sex ratio is equal in our study populations (Berglund et al., 1986a,b). Generally, color patterns of fish are characterized by a darker dorsal than ventral side, which may improve crypsis (Schliwa, 1986). This pattern is also observed in S. typhle, despite the fact that they usually align themselves vertically. Furthermore, the natural variation in color in both females and males is high, ranging from light green over gray to nearly black. Individual fish may pale (decrease the color intensity), but not change their basic color (hue). The different colors closely match the


139

Fig. 1. Syngnathus typhle females without (top) and with (below) the ornament displayed. (Modified from original drawing by Maria Schtrutz.)

varying colors of living and decaying eelgrass. Besides the variation in color, different color patterns exist both in males and females: fish range from dull (uniformly colored with little contrast) to a contrasted pattern with darker stripes looking like the letter B. Fish with intense colors of high contrast are more conspicuous (at least to the human eye), especially when displaying or performing nuptial dances in open water (A. Berglund and G. Rosenqvist, personal observations). While performing reproductive activities, females as well as males may suddenly display a drastic blackening of their normally striped B pattern, making them considerably more conspicuous in appearance (Fig. 1). This is common among competing and courting females, who readily display to males as well as to other females, but uncommon in males, who may display to females but only rarely to other males (Berglund et al., 1997; Bernet et al., 1998; A. Berglund and G. Rosenqvist, personal observations). Reproductive performance depends on body size in both males and females, with the larger females producing more and larger eggs and larger males accommodating more offspring (Berglund et al., 1986a,b). Sexes are of roughly equal size at 1 year of age, when they reproduce for the first time, but females grow faster and become larger than males in their second year (Berglund and Rosenqvist, 1993). B. Nerophis ophidion The N. ophidion female deposits her eggs openly on the abdomen of the male (Fig. 2) after the courtship dance. The female coils around the male

140


Fig. 2. A Nerophis ophidion male with eggs. (Drawing by Maria Schtrutz.)

and squeezes the eggs onto him from his genital opening upward, much like piping cream onto a cake. The male remains passive during copulation, and subsequently the couple sinks to the bottom with the female still coiled around the male (Fiedler, 1954; A. Berglund and G. Rosenqvist, personal observations). The male probably fertilizes the female internally or deposits a mucous sperm package just above his genital opening immediately before copulation, and this package is then squeezed upward by the roe string while fertilizing the eggs. We have never seen males copulate again before eggs have hatched, whereas females may immediately court and copulate with more males in succession. Pregnancy is costly to the male in terms of predation risk: when pregnant, males fell victim to fish predators more often than females, whereas no such difference was evident outside the reproductive season (Svensson, 1988). However, pregnancy did not appear to impair feeding activity and growth in males (Svensson, 1988). Thus, N. ophidion males seem to employ a more risk-prone reproductive strategy compared with S. typhle males.


141

Female N. ophidion possess bright blue sexual colorings along the sides of the head as well as a dorsal skin fold. The fold develops during the breeding period, whereas the blue coloration is more permanent. Males may display a temporary sexual color, a bright yellow nose, while being courted and when copulating. Females are much larger than males, and fecundity correlates positively with body size in females but not in males (Berglund et al., 1986a). Moreover, female fecundity correlates positively with the area of the blue coloring (Berglund et al., 1986a). In our field populations, the adult sex ratio does not differ from equality (Berglund et al., 1986a,b).

VI. Parental Investment, Potential Reproductive Rates, the Operational Sex Ratio, and the Bateman Gradient Parental investment according to Trivers (1972) refers to any investment in an offspring that increases offspring fitness at the expense of a parent’s ability to invest in other offspring. The form of investment could be anything, such as nutrition, immunoglobulins, protection, and so on. We began our pipefish research by asking whether sex role reversal in our two species could be explained by males providing more energy to an offspring than females. Female provisioning is easily estimated in pipefish, as this is what the female puts into an egg. Male energetic provisioning is more complicated. An unnourished egg decreases in energy content until hatching as the embryo consumes energy. As any male energy provision opposes that decrease, we can calculate a male’s contribution by calculating the decrease had no provisioning occurred (by measuring embryo respiration and converting that to metabolic energy consumption), then calculating the actual change in energy from egg to newborn, finally subtracting the two (Fig. 3). With these measurements we revealed that S. typhle males and females each provided approximately equal amounts of energy to offspring. Furthermore, N. ophidion males provided much less than females (Berglund et al., 1986b). This clearly demonstrates that differences in energy provisioning by males and females cannot alone explain sex role reversal. However, these findings do not disprove Triver’s (1972) idea. Rather, the critical parental investment may take a form other than energy. Our next idea formed after discussions with George C. Williams: investment in time, rather than energy, could be crucial. This proved a fruitful idea, underpinning subsequent ideas on how differences in potential reproductive rates may shape mating competition (Clutton-Brock and Parker, 1992; CluttonBrock and Vincent, 1991). We began by experimentally measuring whether females on average could, during a certain time period, produce more eggs than males on average could process during the same time period (Berglund

142


Fig. 3. Change in energy content of an egg if the embryo is not nourished or if the embryo is nourished by the male, and how male energy contribution is calculated.

et al., 1989). If so, female–female competition for males would make sense. This is exactly the same thing as measuring potential reproductive rates, although that term was not yet in use when we performed these experiments (we called it ‘‘maximal reproductive rate’’ instead; Berglund and Rosenqvist, 1990; Berglund et al., 1989). This simple potential reproductive rate scenario has subsequently been further elaborated (and sometimes complicated) by invoking concepts such as collateral investment (Parker and Simmons, 1996), mating qualifications (Ahnesjo¨ et al., 2001), and so on, but for our species the simple approach is adequate. As adult sex ratios in nature are approximately equal in these two pipefish species, a female can expect on average to have access to the brood pouch of one male. If females during the time period of one male pregnancy can fill more than one pouch, it would suggest that an excess of eggs is produced. In other words, females have a higher potential reproductive rate than males. In a simple experiment designed to address this idea a female was provided with an excess of males and, indeed, more than one male was filled during an average pregnancy span (in fact, on average almost two males were filled in both species; Fig. 4). Thus, males were a resource in short supply (Berglund et al., 1989). In S. typhle, the difference between male and female potential reproductive rates increased with age, such that among 2-year-old individuals, females could fill almost three males (Berglund and Rosenqvist, 1990). In addition, temperature may modify but not reverse this sex difference. As the temperature rises, the male pregnancy becomes progressively shorter whereas female egg production is less affected. This difference does not, however, proceed to such a point that males become faster than females at processing eggs (Ahnesjo¨, 1995), at least not within the temperature range at our latitudes.


143

Fig. 4. Female Syngnathus typhle could fill on average two males during an average male pregnancy (45 days that year) when provided with an excess of mates. This means that, at an even adult sex ratio, the female potential reproductive rate is twice that of males. (After Berglund et al., 1989.)

Moreover, these sexual differences in potential reproductive rates influenced operational sex ratios: if males cannot reproduce as fast as females, females will by necessity be ready to remate sooner than males, and nubile females will therefore outnumber males. This was confirmed by field data: the OSR was always female biased, as most males had full pouches (Berglund and Rosenqvist, 1993), except sometimes early in the breeding season when many males may be simultaneously available (Vincent et al., 1994). If females are potentially faster reproducers than males, and females willing to mate outnumber males willing to mate, we expect a closer association between number of matings and number of progeny in females than in males, that is, a steeper Bateman gradient in females. Thus, the logic underlying sexual selection is that parental investment influences potential reproductive rates, which influence the OSR, which, in turn, influences the Bateman gradient (Table I). Using molecular techniques to assign parentage (four microsatellite loci) we were for the first time able to demonstrate that the Bateman gradient is reversed in a sex role-reversed species. This was done in an experiment in which numbers of freely mating males and females were equal or in which females outnumbered males. As predicted, female S. typhle exhibited a steeper Bateman gradient than did males. In both males and females the slope was significantly greater than

144


TABLE I Sex Role Research in Syngnathus typhle and Nerophis ophidion: A Schematic Representation of Results in Relation to the Logic behind Sexual Selection Parental investment: male pregnancy length and male brooding capacity (Ahnesjo¨, 1992a,b, 1995, 1996; Berglund et al., 1986a,b; Svensson, 1988) # Potential reproductive rates: males < females (Ahnesjo¨, 1995; Berglund and Rosenqvist, 1990; Berglund et al., 1989) # Operational sex ratio: males < females (Berglund, 1991, 1994; Berglund and Rosenqvist, 1993; Vincent et al., 1994, 1995) # Females compete (Berglund, 1991; Berglund and Rosenqvist, 2001b; Bernet et al., 1998; Rosenqvist, 1990), Males choose (Berglund and Rosenqvist, 1993, 2001a; Berglund et al., 1986a; Rosenqvist, 1990)

Influenced by predation (Berglund, 1993; Fuller and Berglund, 1996) Influenced by mate encounter rate (Berglund, 1995) Influenced by mate quality (Jones et al., 2000b; Sandvik et al., 2000) Influenced by parasites (Rosenqvist and Johansson, 1995)

# Fertility depends on number of mates more for females than for males (Jones et al., 2000a) # Sexual selection primarily on females # Female ornaments and status signals (Berglund, 2000; Berglund and Rosenqvist, 2001a,b; Berglund et al., 1997; Bernet et al., 1998; Fitzpatrick et al., 1995; Rosenqvist, 1990)

zero, but the slope for females was steeper than that for males. Thus, females exhibited a stronger association between number of mates and fertility than did males in S. typhle. In a treatment in which males were in excess this difference vanished (Fig. 5; Jones et al., 2000a). It must be noted that paternal care per se does not cause sex role reversals, as the majority of caring fish species have exclusive paternal care without reversed sex roles. This is so because the usual form of care in fish is guarding and fanning, which allows the male to accept several clutches and does not reduce his potential reproductive rate below that of the females. Moreover, not even the most extreme form of paternal care found in pipefishes and seahorses necessarily causes sex role reversal. Seahorses


145

Fig. 5. The relationship between mating success and fecundity in Syngnathus typhle. Open squares and a solid line represent males and closed circles with a dashed line represent females. Each square or circle represents a mean with 1 standard error drawn. Numbers of males and females are given for each mean. The sexual selection gradient (the Bateman gradient) is given by the weighted least-squares regression line relating mating success to fertility. The nonzero slope for male S. typhle is almost entirely due to six receptive males that failed to mate during the experiment. The Bateman gradient for females is significantly steeper than the gradient for males. (After Jones et al., 2000a.)

typically have conventional sex roles, as mentioned previously, in spite of their elaborate paternal care (Vincent, 1994; Vincent et al., 1992). Also note that the scheme presented in Table I (parental investment influencing potential reproductive rates, which influences the OSR, which influences the Bateman gradient, which influences mating competition, thus influencing sexual selection) is an oversimplification. Many other factors are influential in this process, such as mate search costs, mating costs, and mating pattern, which all, besides parental investment, can influence potential reproductive rates. The OSR is, of course, influenced by much besides potential reproductive rates, such as sexual differences in distribution, mortality, or reproductive life span and by sex ratio at birth. Variance in mate quality can, besides the OSR, also influence mating competition and choosiness directly, as can various ecological conditions such as mate encounter rate, breeding resources, and predation pressure (Clutton-Brock and Parker, 1992; Johnstone et al., 1996; Owens and Thompson, 1994).

VII. Female Competition If the conditions described previously invoke competition in female pipefish, in what form would the competition exist? Syngnathid fish

146


are characterized by their jointed jaws, by having no teeth, and by being slow moving. Therefore, any form of competition is nonviolent. One form it takes in S. typhle is mating interruption (described previously), in which a female may disturb and eventually interrupt the mating activities of a pair just by swimming in between them. Larger females were much more efficient at this than smaller females (Berglund, 1991), and females were commonly forced out of displaying groups (Vincent et al., 1994). Interruption efficiency can also be predicted from ornament display, with females more similar in display duration being more equal competitors and delaying mating longer than subdominant females (A. Berglund and G. Rosenqvist, unpublished data). Another form of competition is more indirect and occurs by means of between-female dominance. Large females may by their mere presence interfere with, and substantially decrease, reproduction in small females. This was demonstrated in an experiment in which females were provided with an ample supply of males. These females produced fewer and smaller eggs if they saw an enclosed, larger female, as compared with females seeing a similar-sized enclosed female. The enclosed females could not dance or mate with the males. By largely giving up reproduction the subdominant females instead grew faster, indeed as rapidly as females not reproducing at all (Berglund, 1991). By forfeiting current reproduction these small females could potentially return for the next breeding cycle at a larger and more competitive size. As winter survival seems high (Berglund, 1991), this apparent life history decision makes good sense. In N. ophidion, female–female competition is likewise nonviolent and occurs by status signaling: in all-female groups, only one female developed a large ventral skin fold (a sexually selected characteristic important in female–female interactions and also in male choice; see below), whereas females kept alone with males invariably developed the fold (Rosenqvist, 1990). Thus, females seem to dominate each other by means of a sexual signal, the fold, thereby effectively reducing reproductive success in subdominants in much the same way as in S. typhle.

VIII. Male Choosiness Body size is evidently an important trait in S. typhle. In fact, in experiments both males and females preferred to mate with a large partner if given a choice (Berglund et al., 1986a). They both accrued direct advantages from doing so. A large female produces larger eggs than does a small one, so males benefit from receiving these larger, energy-rich eggs. Large eggs give rise to larger offspring that grow and survive predation better than do smaller newborns (Ahnesjo¨, 1992a,b). Females benefit by


147

receiving the better paternal care a large male can provide, compared with what a small male can offer (Berglund et al., 1986a). Thus, body size in both males and females is under both natural (higher fecundity, better eggs) and sexual (increased attractiveness) selection. These experiments were staged so as to provide the focal animals with a choice of two partners differing in size and, given this opportunity to choose, both sexes did so. However, males were choosier than females, as we would predict if it is predominantly females that compete over males. When provided with a less attractive (i.e., small) partner, males were slower to copulate, copulated fewer times, and accepted fewer eggs, compared with females mating with a less attractive mate. Thus, males were reluctant to mate whereas females readily reproduced even with low-quality partners (Berglund and Rosenqvist, 1993). In addition, male S. typhle preferred to mate with ornamented females, whether the ornament was displayed naturally or manipulated (see Section XI.C). Choosiness is a plastic male trait in S. typhle, which can be experimentally modified by, for instance, manipulating predation threat. The nuptial dance and copulation occur largely above the protective eelgrass vegetation that these animals normally dwell within, so for the male to choose which female to favor and then to dance and copulate with her are potentially risky behaviors. Consequently, choosiness disappeared in the presence of a predator, which in effect decreased the time spent dancing and also decreased the number of copulations. However, the number of eggs transferred per copulation increased, so males were filled to capacity as quickly and safely as possible, but with eggs from a more random set of females (Berglund, 1993). Thus, predation threat will decrease the force of sexual selection in these fish. Moreover, the level of predation experienced by the males affected risk taking: a predator that was only seen had less effect than a predator both seen and smelled, which in turn had less effect than a predator seen, smelled, and felt in the water (Fuller and Berglund, 1996). In an extremely cryptic and slow-moving animal such as S. typhle, it makes good sense to be risk sensitive: it is probably a most hazardous endeavor to leave the vegetation to reproduce. Therefore, by reducing the level of choosiness and consequently mating quickly and indiscriminately, males may in effect reduce predation risk at the cost of mating with other than only high-quality females: males trade mate information for risk reduction. Male choosiness was also modified by the operational sex ratio: choosiness disappeared completely under male excess, compared with female excess (Berglund, 1994). Instead, males reproduced faster under male excess, obviously again treasuring speed at the expense of quality in this situation, just as in a situation with predation risk (Berglund, 1994).

148


Fig. 6. Male choice between large and small females in Syngnathus typhle in two treatments, one with a low mate encounter rate (left) and one with a high encounter rate (right). NS, Not significant. **p < 0.01 according to binomial p tests. (After Berglund, 1995.)

This was also reflected in the Bateman gradients: under male excess the slopes of Bateman gradients for males and females did not differ. This contrasts with the case in which OSRs were equal or female biased, with the slope for females greater than the slope for males (Fig. 5; Jones et al., 2000a). Furthermore, choosiness was affected by mate encounter rate: choosiness disappeared under experimentally staged low encounter rates, that is, when mates were difficult to find, as compared with high encounter rates (Fig. 6; Berglund, 1995). Nerophis ophidion males likewise preferred to mate with larger rather than smaller females when given the choice experimentally (Rosenqvist, 1990). Moreover, males also preferred ornamented females with larger areas of blue coloration along their heads, independent of female body size (Berglund et al., 1986a), and also females with larger skin folds, again independent of female body size (Rosenqvist, 1990). This suite of male preferences for female traits makes good sense, as body size, blue area, and skin fold size all correlate with female fecundity (Berglund et al., 1986a; Rosenqvist, 1990). Thus, a male mating with a female that is large in any of these respects will obtain more eggs to brood than a male without such preferences. Nerophis ophidion females were not choosy at all: no preference for larger rather than smaller males could be demonstrated in this species (Berglund and Rosenqvist, 1993).


149

Fig. 7. The life cycle of the trematode parasite (Cryptocotyle sp.): egg with miracidium (upper right), Littorina littorea (first host, lower right), cercariae (bottom), pipefish (second host, left), and gull (final host, top). (Drawings by Annika Robertson.)

IX. Mate Choice and Parasites Parasites are believed to have a profound effect on all sexually reproducing organisms; in fact, they are believed to be responsible for the preponderance of sexual over asexual reproduction. Since Hamilton and Zuk (1982) suggested that females should prefer showy males because such males were resistant to parasites, many studies have investigated the effect of parasites on ornaments (e.g., Burley et al., 1991; Chappell et al., 1997; Houde and Torio, 1992; Johnson and Boyce, 1991; Kennedy et al., 1987; Milinski and Bakker, 1990; Møller, 1990; Zuk, 1996; Zuk et al., 1990). Pipefish have numerous parasites and can also be artificially parasitized and therefore readily lend themselves to the study of parasite effects on sexual selection and life history trade-offs. In S. typhle a parasitic trematode (Cryptocotyle sp.) is especially conspicuous. It can be recognized by the black pigment that the pipefish develops around the parasite encysted in the skin. The parasite has a complicated life cycle with several hosts: snail, fish, and, finally, a bird or a mammal (Fig. 7). High levels of infection with the parasite have been reported to kill the fish host (Sindermann, 1966). In S. typhle the parasite can impair female fecundity: highly parasitized females produced significantly fewer eggs than did less parasitized females, and also survived less well (Rosenqvist and Johansson, 1995). Accordingly, males preferred a female with fewer parasites when given a choice (Rosenqvist and Johansson, 1995). Avoidance of parasitized partners has also been shown in many other fish species, for example, guppies (Houde and Torio, 1992; Kennedy et al., 1987) and sticklebacks (Milinski and

150


Bakker, 1990). However, one problem with using a natural parasite load to compare the attractiveness of infected and uninfected individuals is the difficulty in distinguishing effects of parasite infection from effects of general condition. To control for this, S. typhle females were tattooed with black ink to look infected, and then paired with matched females that had been tattooed with clear ink in a mate choice experiment. In this experiment, males again preferred females with fewer black spots. As the tattoo altered condition and behavior equally in either female category, this implies that males used visible signs to avoid parasitized females (Rosenqvist and Johansson, 1995). Parasites may influence the evolution of mating preferences in two ways: (1) avoidance of conspecifics with an infection that can be directly transmitted (Hoelzer, 1989) or (2) avoidance of individuals that will provide fewer benefits (Heywood, 1989). The complex life cycle of Cryptocotyle makes direct transfer from one pipefish to another impossible (Stunkard, 1930; Thulin, 1971). Furthermore, the possibility that the males were only trying to avoid areas with a high concentration of parasites is improbable: in experiments males did not avoid the company of other males with tattooed black spots or prefer males with clear spots (Rosenqvist and Johansson, 1995). Thus, males probably avoid females with parasites as these provide a less direct benefit in terms of fecundity. A more fecund female will transfer more eggs during each copulation (Berglund et al., 1986a), allowing the male to fill his pouch more quickly and safely. Moreover, males mating with nonparasitized females may increase offspring fitness if parasite resistance is heritable.

X. Mate Choice and Offspring Quality In pipefish choosiness is obviously costly, as mate assessment and courting reduce crypsis and thus increase exposure to predation (Berglund, 1993; Fuller and Berglund, 1996). So what are the benefits? Obviously choosiness should increase reproductive success, for instance, by resulting in better quality offspring. Few studies have experimentally investigated direct fitness effects from mate choice on the offspring, and conclusions regarding benefits gained are often ambiguous (references in Møller and Alatalo, 1999). We estimated the relative impact of male and female choice on offspring fitness correlates by setting up mate choice trials in which males or females could choose between two potential mates (Sandvik et al., 2000). Focal individuals were mated with either a partner of their own, previous choice, or with a nonpreferred, previously rejected partner. This experimental design ensured the detection of individual


151

preferences, which could otherwise have confounded results from matings based on estimates of general attractiveness to the opposite sex (see Drickamer et al., 2000). We measured two fitness correlates in the resulting offspring: growth during the first week after birth, and the ability to escape a common predator (a small sea anemone living on eelgrass, Sagartiogeton viduatus). As predicted, a constrained choice had an impact on the resulting offspring. Newborns from preferred matings were superior at escaping predation, when both males and females were allowed to choose a partner. However, only choosy females benefited in terms of faster growing offspring (Sandvik et al., 2000). This was the first study demonstrating that both sexes may produce fitter offspring if allowed to mate with their preferred partners (but see, e.g., Drickamer et al., 2000; T. M. Jones et al., 1998; Nicoletto, 1995; Partridge, 1980; Petrie, 1994; Reynolds and Gross, 1992, for effects of female choice on offspring quality). Female preferences potentially had a stronger effect on offspring quality than male choice, because only females were able to pick out males producing faster growing young. The recognition of the importance of mutual mate choice (i.e., that both sexes of a species can exhibit choice) and mutual competition may lead to different predictions for the evolution of both sexual traits and preferences. Given a generally female-biased OSR for this species (Berglund and Rosenqvist, 1993; Vincent et al., 1994), it seems probable that only high-quality females can afford to choose among males. Thus, to make predictions about the evolutionary consequences of mutual mate choice, we must also include constraints such as mate assessment abilities and the effect on offspring quality of choice. Our results show that being denied access to the preferred partner may reduce fitness in both sexes, and thus underline the importance of a dynamic approach to mate choice mechanisms in the study of evolution of preferences and sexual characters (Sandvik et al., 2000).

XI. Ornament in Female Syngnathus typhle Even if mating competition and mate choice are mutual processes between males and females, affecting individuals of differing quality differently, basically females are more competitive and males more choosy in our pipefish. Therefore, sexual selection acts more strongly on females than on males, producing sexually selected characters predominantly in females. As argued previously (Section II), female ornaments should not be energetically costly but should maintain honesty either by being honest by design or by having other associated costs. We have studied body size as

152


well as the temporary striped ornament in S. typhle extensively from this point of view. Regarding body size, we have already stated that both males and females prefer larger individuals, and that they both gain direct (and possibly indirect genetic) benefits by their choice (Section VIII). Regarding the striped pattern, which we now deal with specifically, what kind of information does this ornament convey to males? Does it honestly signal female quality? Do males find it attractive, independent of other features of female display? What are the costs and benefits associated with ornament display? By attempting to answer these questions we hope to better understand the end product of sexual selection, the sexually selected character itself. A. Ornament as an Amplifier A. Zahavi suggested that the stripes in female S. typhle may act as an amplifier (personal communication). An amplifier increases the resolution power of a signal, that is, it makes differences between two signals easier to detect (Hasson, 1989, 1990, 1991, 1997, 1999). Thus, the amplifier does not necessarily boost the signal or make it look stronger than it really is, it just improves readability. In other words, the amplifier acts on and enhances discrimination, rather than the strength of the signal itself. If a sexually selected signal is an honest quality indicator, it has been claimed that an amplifier acting on this particular signal need not be costly or attractive in itself, only the signal it amplifies (Hasson, 1989). However, as the signal receiver obviously perceives the amplifier, the amplifier may evolve to become attractive itself. Low-quality animals may be ‘‘forced’’ to display the amplifier, as not doing so may signal low quality. Also, an amplifier may increase conspicuousness both to competitors and predators, thus making the amplifier costly: it may be socially provocative during, for instance, mating competition, or reduce crypsis and therefore increase vulnerability to predators. Thus, amplifiers may easily evolve to become costly, condition-dependent quality signals themselves. Therefore, they may also become attractive to potential mates. If so, they can be termed ‘‘amplifying handicaps’’ (Fitzpatrick, 1998; Hasson, 1990, 1997). For example, tail markings in birds may be such amplifying handicaps, indicating feather quality more clearly (Fitzpatrick, 1998). Until now no signal has been demonstrated to function as an amplifier, but speculations abound (e.g., Bradbury and Vehrencamp, 1998; Hasson, 1991; Taylor et al., 2000). Moreover, there are not yet any empirical demonstrations of whether amplifiers are ‘‘pure,’’ cost-free and neutral (in terms of attractiveness) signals, or costly and attractive. If the striped ornament in female S. typhle is an amplifier of female body size, that is, if it


153

Fig. 8. The bars mimicking Syngnathus typhle pipefish females with three different degrees of ornamentation were used to let students estimate which bar, in a pair of same-patterned bars, was larger. Estimates of both length and width were more accurate the more ‘‘ornamented’’ the rectangles. (After Berglund, 2000.)

makes it easier for males to tell females of different sizes apart, this can be estimated by letting students try to tell differently ‘‘ornamented’’ rectangles, differing in size, apart (Berglund, 2000). First, to elucidate the impact of the ornament itself, students were asked to state which rectangle in a pair was longer (or wider) between rectangle pairs that differed in ‘‘ornamentation’’ (Fig. 8). As predicted, the stronger the ‘‘ornamentation,’’ the more accurately students could identify the larger rectangle in a pair (Berglund, 2000). Second, to elucidate why the ornament is cross-wise rather than length-wise striped, students were asked to tell cross- or length-wise striped rectangle pairs apart. Which way the stripes ran did not matter for ability to tell rectangles differing in length apart, but for width there was a difference: a cross-wise pattern facilitated discrimination (Berglund, 2000). As female width correlates with female fecundity more strongly than length (Berglund, 2000), this may explain why the ornament is cross-wise rather than length-wise in direction. This requires that pipefish see the world similarly to the way students do. Pipefish have good vision and hunt by eye (Fiedler, 1954), and the chromatophore-regulated melanin-based ornament ought to pose no problem with possible ultraviolet vision in pipefish. The assumption of enough similarity between human and pipefish vision for ornament perception therefore seems reasonable. Overall, our results suggest that the ornament functions as an amplifier. It increases the accuracy of body size estimates if pipefish see things the way humans do, that is, it makes it easier for the receiver of the information to tell differently sized females apart. Consequently, this will

154


make it easier for males to distinguish females carrying numerous and large eggs, as these properties both correlate positively with female size. So, is the ornament a ‘‘pure,’’ cost-free, and in itself unattractive amplifier? Does the ornament itself signal female quality? Is it provocative to other females? Further experiments were required to address these questions. B. The Ornament and Female Quality Females displaying the ornament more frequently in experiments were also more fecund than females displaying it less frequently (Berglund et al., 1997). More ornamented females also displayed more actively to males and were in closer contact with them (Bernet et al., 1998). Thus, ornamentation reliably predicted female quality and mating success in S. typhle: ornamented females danced more, mated more, and transferred more eggs than did nonornamented females (Berglund and Rosenqvist, 2001a; Berglund et al., 1997; Bernet et al., 1998). However, under predation threat these advantages disappeared. Females displayed the ornament less under predation risk, leaving males with less of an opportunity for mate choice (Bernet et al., 1998). By contrast, female–female competition encouraged ornament display: most females displayed the ornament under such competition (Bernet et al., 1998), thereby probably intimidating rivals (Berglund and Rosenqvist, 2001a,b). Males utilizing ornamentation as a guide in their mate choice may thus be able to perform their choice more quickly, reducing the time spent on potentially dangerous mate search. They may also more accurately identify the largest (i.e., most fecund) female, thanks to the amplifying function of the ornament. Moreover, males that mate with more ornamented and thus dominant females (Berglund and Rosenqvist, 2001a,b; A. Berglund and G. Rosenqvist, unpublished data; see below) may be less harassed by other females while mating, thereby further reducing the dangers associated with this behavior (Berglund, 1993; Fuller and Berglund, 1996), something also corroborated by field observations (Vincent, 1994; Vincent et al., 1995). Intrasexual contest competition and mate choice are often intertwined processes and can both influence ornamentation (e.g., Andersson, 1994; Berglund et al., 1996), so the pipefish ornament evidently serves a dual function: attracting males and repelling females, singling out the high-quality females for reproduction. C. Is the Ornament Attractive? As mentioned above, females spontaneously displaying the ornament for longer periods gain a higher mating success than females displaying less often, both by attracting males and deterring other females (Berglund and


155

Rosenqvist, 2001a,b; Berglund et al., 1997; Bernet et al., 1998). Is this because the ornament in itself is attractive to males, or because such females are in better condition and thus able to display better? To investigate this we need an experiment in which female size and behavior are held constant while ornamentation is manipulated. Few such experiments, rigorously testing one specific signal while controlling for others, including behavior, have been carried out (but see Jones and Hunter, 1999). We manipulated the ornament by painting females while controlling female behavior by sedating them and moving them up and down in a dancelike fashion by stringing them to a motor. We let males choose between two such equal-sized females in mate choice trials, one female painted black and the other sham painted. As predicted, males preferred ornamented females: they spent more time in front of such females (a good predictor of which female eventually gets to mate; Berglund, 1993) and tried to dance with them more often (Fig. 9; Berglund and Rosenqvist, 2001a). This was not because darkness in females itself was attractive: in another mate choice experiment, males preferred females painted crosswise over females painted length-wise, but otherwise of similar overall darkness (A. Berglund and G. Rosenqvist, unpublished data). Thus, the ornament itself is a signal attractive to males in their choice of partner. D. The Ornament and Female–Female Competition In nature, females display in groups by swimming up and down and in and out of the eelgrass, with their ornaments fully displayed. Females actively compete among themselves for matings during such group displays, and may try to herd other females away from the male (Vincent et al., 1995). When this happens, mating is interrupted, and indeed attempts to copulate can be thwarted for a long time by interference from other females (A. Berglund and G. Rosenqvist, field and aquarium observations; Berglund, 1991). In experiments, ornament display is promoted if females can compete (Bernet et al., 1998). Can males use information from ornament display under conditions of female–female competition? It may well pay males to select dominant females, which are able to discourage other females from courtship interference, as this will reduce the time spent on potentially risky mating behaviors. Moreover, there is no conflict between female signals of dominance and signals of caring ability, as females contribute no care, only eggs. Finally, we have never observed female aggression toward males. Thus, males should prefer dominant and competitive females, and if males can directly assess such abilities in females this may be more important than information from attraction displays to the male’s mate choice

156


Fig. 9. A Syngnathus typhle male could choose between a black-painted ‘‘ornamented’’ female and a sham-painted one. Females were sedated and mechanically moved up and down by a motor. The main figure represents the male’s view, and the inset is a view from above. Males preferred the black-painted (ornamented) females. (After Berglund and Rosenqvist, 2001a.)

process. Indeed, any female can display her ornament, but females displaying it under female–female competition have demonstrated their worth in a potentially costly situation. Therefore, males mating with dominant females should experience a smooth courtship and copulation process, gaining direct benefits for themselves in terms of reduced risk, as well as possible direct and/or genetic benefits for their offspring from mating with high-quality females. To investigate this, we staged a mate choice experiment in which an enclosed male could choose between two females (Berglund and Rosenqvist, 2001b). On the first experimental day, females could interact freely and compete, whereas on the second day they were isolated from


157

each other. When female–female competition was allowed, females displayed the ornament to one another more persistently than they did to the male: time spent competing, rather than time spent courting the male, correlated with ornament display duration. The female displaying most intensely is thus likely to be the dominant one within a female pair. Obviously, a more intense ornament display (or other information from watching competitive interactions) made this dominant female attractive to the male, as males used this information on the following day when females were no longer allowed to compete: the male spent the most time before the female that had displayed for longer in the interfemale competitive situation. In fact, this old information about the ornament or competitive displays completely overrode new information, and significantly influenced male mate choice. Furthermore, ornament display under competition did not correlate significantly with ornament display in the absence of competition (Berglund and Rosenqvist, 2001b). Consequently, on the day of the male’s choice, ornament display contained little or no information regarding female dominance. In a mirror image experiment, with females separated the first day and allowed to compete the second (Berglund and Rosenqvist, 2001a), ornament display on day 1 significantly predicted ornament display on the second day. Ornament display on either day also significantly predicted male mate choice. Both males and females chose and behaved similarly on both days in this mirror image experiment, so time effects, such as experience with the experimental setup, are unlikely to have confounded the results. Thus, when females were separated on the first day, their ornament display was the same on the second day with competition, but when competition preceded separation this was no longer true. Does intense competition make females less able, or less willing, to subsequently display their attractiveness? The fact that females competing more intensely on day 1 displayed their ornaments to a lesser extent on day 2 suggests that intense competition is tiring or otherwise discourages females from performing the display on the following day. Indeed, long-term reproductive inhibition has been demonstrated in this species previously: larger, and presumably dominant, females may decrease reproductive activity in smaller females (Section VII; Berglund, 1991). Alternatively, a female having proved her worth during female–female encounters may not need to display her attractiveness further, as males will in any case rank such information more highly than pure displays of attraction. Thus, male pipefish seemed to remember and make use of information regarding partner dominance, presumably according to the reliability of that information: competitive displays seemed more reliable as a signal of female quality than noncompetitive displays. The display under

158


competition may provoke other females and thus carries a cost, compared with noncompetitive displays, in which the cost is only a lost mating opportunity. Thus, male pipefish prefer beautifully ornamented females, but, above all, victorious ones. E. Costs and Benefits of The Ornament We commonly assume that a signal given when sender and receiver have conflicting interests needs to be honest, because otherwise cheaters would initially have an advantage and receivers would be selected to ignore the signal in question. Ornaments are classic examples of such conflict-laden signals. Signals can be honest by design or because of associated costs. In S. typhle, body size in females was, as we have seen, attractive to males, and males enjoyed direct benefits from mating with larger females. Body size is a signal/cue difficult to fake convincingly, but the striped pattern amplifying size is not. As this signal was attractive in itself (Berglund and Rosenqvist, 2001a), and thus can be regarded as more of a traditional ornament than only an amplifier, we expect some costs associated with ornament display to maintain its honesty. We have found no energetic costs for ornament display, however: in an experiment, starved females were as likely to display the ornament as well-fed females (Berglund et al., 1997). This was expected, both on theoretical grounds (see Section II) as well as physiological grounds: ornament display is simply a widening of melanin-containing chromatophores, which is not a very energy-consuming process. Another potential cost to these pipefish is mortality through increased predation, presumably due to lost crypsis. Indeed, females were more reluctant to display the ornament under predation threat (Bernet et al., 1998), thereby mating at a lower risk but also more randomly than otherwise. However, whether this mortality cost is sufficient to maintain ornament honesty is doubtful. A requirement is that high-quality females dare to display the ornament in spite of predation risk, something about which we have no evidence. Second, mortality may generally be of little importance to ornament evolution and maintenance, in spite of its being a classic textbook explanation of what causes trait evolution to come to a halt. This is for two reasons: first, we do not usually see excess mortality in the more ornamented sex. In fact, it has been difficult to demonstrate such mortality in any species; usually more ornamented individuals also survive better. Consequently, sex ratios are rarely heavily biased due to ornamentrelated mortality, which we would expect if such mortality were important. Second, if more ornamented individuals actually do suffer high mortality, a consequence is that more ornamented survivors will enjoy increased


159

mating success. This requires that the ornament be attractive and its bearer not too difficult to find. If so, the increased mating success of survivors may compensate for the decreased mating success of those that died, and, consequently, mortality would not be an efficient selective agent in ornament evolution and maintenance (A. Berglund and S. Fitzpatrick, unpublished data). Instead, we expect animals to experience a decrease in condition long before they die. So, returning to our pipefish, is there some cost to the female ornament besides mortality, related to condition? The ornament is used in female–female competition. In experiments we have shown that female competition may substantially prolong the time until mating occurs, and more so if females are equal competitors (A. Berglund and G. Rosenqvist, unpublished data). Furthermore, intense competition seems costly to a female’s subsequent performance (Berglund and Rosenqvist, 2001b). Such a social cost due to female–female harassment may well provide the key cost to signal honesty here (A. Berglund and G. Rosenqvist, unpublished data): only high-quality dominant females dare challenge others to engage in a costly and possibly prolonged courtship, and males should prefer such females to a reduced risk during mating. Thus, risk may enter into the cost but, for females, through social interactions and male mate choice rather than through mortality directly. F. Trade-offs with Life History Traits The effects of trade-offs between natural and sexual selection on the evolution of allocation strategies have been relatively neglected by researchers. Moreover, trade-offs between sexually selected traits that appear to differ in cost have rarely been considered. In S. typhle, female body size and female ornament display may constitute two such traits. Might females shift their pattern of investment in these two sexually selected traits if food becomes restricted? As variable environmental conditions should favor such conditional investment strategies in a species with more than one preferred trait with differing costs, we might expect such shifts under different resource budgets. Furthermore, as sexual displays should be regarded in the same way as any other life history trait, the costs of sexual selection can be studied only by manipulating individual resource budgets and then measuring display and allocation patterns. As the cost of sexual selection can be mediated in different ways, a focus on such mechanisms can produce novel insights into how individuals optimize allocation strategies. Specifically, traits such as body size, growth potential, and number and size of eggs, as well as remaining length of reproductive season and

160


reproductive life, may affect ornament display in females. We expect such trade-offs to be especially important in animals with indeterminate growth (such as fish), among which fitness consequences of various allocation patterns may reach far and last long. Does body growth trade off with egg numbers or egg size? How are these tentative trade-offs affected by a female’s resource budget? Does previous reproductive effort affect condition and ornament display? Does predation risk affect such tradeoffs, and, if so, how? Under environmental stress, to which these pipefish are sensitive, will allocation to body size, to ornament display, or to eggs suffer first? Similarly, do parasites affect the order of allocations? At present we have no answers to all these questions; rather, they outline future research.

XII. Conclusions Ideas regarding how parental investment, potential reproductive rates, operational sex ratios, and mate quality shape mating competition and sexual selection remain ambiguous if only species with ‘‘normal’’ sex roles are investigated, that is, species in which it is predominantly males that compete over access to females. The suspicion that there is something specific to maleness per se always lingers as a potentially confounding source. However, with the introduction of sex role-reversed species as study objects, a new light is shed on these matters (Williams, 1966): we have now gained a firmer understanding of why and how sexual selection operates, which applies also to species with ‘‘normal’’ sex roles. Moreover, the role of mutual mating competition/choosiness seems fruitful to explore in the context of sex role reversal (e.g., Amundsen, 2000): here, both sexes invest in offspring, and both can potentially be competitive as well as choosy. Finally, the similar but not identical role of female ornaments, compared with male ornaments, sheds light on what constrains ornament evolution: costs to fecundity probably hinder females from becoming as bizarre as some males in species with conventional roles. Consequently, energetically cheap ornaments predominate, as in S. typhle females, honesty instead being achieved through other, largely social, costs. Indeed, such ornaments should also prevail in species with ‘‘normal’’ sex roles, where males provide direct benefits for females or offspring, an insight largely inspired by research on sex role-reversed species. Thus, a seemingly small and unimportant group of animals, those with reversed sex roles, have played a large and important role in shaping our understanding of sexual selection and mating patterns.


161

XIII. Summary In a small but interesting minority of all animal species, sex roles are reversed: females ready to mate are in excess, and therefore they compete for males. Potential reproductive rates (how fast males and females can potentially remate, given that partner availability is not limiting) and operational sex ratios (the proportion of males and females willing to mate) are instrumental in understanding this pattern. In two sex rolereversed pipefish species, Syngnathus typhle and Nerophis ophidion, males devoted more time, but not more energy, than females to offspring production. The long male pregnancy (offspring are brooded and nourished by males on their bodies) lowered the potential reproductive rate of males below that of females. As females were faster reproducers the operational sex ratio became skewed toward an excess of females. Thus, males are choosy and females competitive. When given a choice, S. typhle males preferred large over small females, and parasite-free over parasitized females, and both S. typhle and N. ophidion males preferred ornamented over nonornamented females. Choosiness was a plastic male trait that in S. typhle could be modified by predation threat (choosiness disappeared in the presence of a predator), by the operational sex ratio (choosiness disappeared under male excess), and by mate encounter rate (choosiness disappeared under low encounter rates). Moreover, potentially dangerous mating activities decreased as the level of threat increased. Syngnathus typhle males, which fertilize the eggs inside their brood pouch, had a paternity confidence of 100% (confirmed by microsatellite analysis). A male received eggs from on average three females before his brood pouch was full. Moreover, females exhibited a stronger positive association between number of mates (as determined by microsatellite analysis) and fertility than did males, so the relationship between mating success and number of progeny, as characterized by the Bateman gradient, affected the strength and direction of sexual selection. Female ornaments ought to be honest signals of quality, but should also be energetically inexpensive (otherwise female fecundity would be reduced). Body size in pipefishes can be such a signal: in S. typhle and N. ophidion males preferred larger females that had more and larger eggs. Larger eggs gave rise to higher quality offspring. Larger females also dominated smaller females, as a result gaining a reproductive advantage. Furthermore, status badges may serve as honest signals carrying a social (but not energetic) cost to cheaters. A temporary color ornament used by S. typhle females in female–female interactions and during courtship was an example of this: it was used in female–female competition, attracted

162


males, was independent of nutritional level, and honestly predicted fecundity and mating success in females. Ornament design was not arbitrary: it facilitated the detection of size differences between females. Ornament display under female–female competition was a more reliable and important signal for males than was display without competition. Males mating with dominant females may gain direct benefits in terms of reduced risk during courtship as well as benefits for their offspring. Displaying females suffered harassment from other females and reduced crypsis to predators. The choice of partners conferred fitness benefits to the choosing individual, whether male or female. Broods from preferred matings were superior at escaping from predators, whether males or females performed the choice. Moreover, females, but not males, also benefited in terms of faster growing offspring when mated to a partner of their choice. Thus, the process of sexual selection of females, constrained by costs for female sexual characters, mediated by a plastic process of mutual mate choice and encouraged by fitness advantages through mating competition and offspring quality, is now better understood. Acknowledgments The authors thank Ingrid Ahnesjo¨ (neé Svensson) and Maria Sandvik for their long-term membership on the pipefish research team, and also our more temporary collaborators Patricia Bernet, Becky Fuller, Kerstin Johansson, and Amanda Vincent: the studies summarized here are indeed a joint effort. In addition, numerous field assistants have made invaluable efforts. Our research was carried out at the Klubban biological station and the Kristineberg marine research station, mainly with grants from the Swedish Natural Science Research Council (to A.B.) and the Norwegian Research Council (to G.R.). We also thank Ingrid Ahnesjo¨, Sarah Robinson, Tim Roper, Maria Sandvik, and an anonymous referee for comments on this manuscript.

References Ahnesjo¨, A., Kvarnemo, C., and Merilaitaa, S. (2001). Using potential reproductive rates to predict mating competition among individuals qualified to mate. Behav. Ecol. 12, 341–397. Ahnesjo¨, I. (1992a). Consequences of male brood care: Weight and number of newborn in a sex-role reversed pipefish. Funct. Ecol. 6, 274–281. Ahnesjo¨, I. (1992b). Fewer newborn result in superior juveniles in the paternally brooding pipefish Syngnathus typhle. J. Fish Biol. 41B, 53–63. Ahnesjo¨, I. (1995). Temperature affects male and female potential reproductive rates differently in the sex-role reversed pipefish, Syngnathus typhle. Behav. Ecol. 6, 229–233. Ahnesjo¨, I. (1996). Apparent resource competition among embryos in the brood pouch of a male pipefish. Behav. Ecol. Sociobiol. 38, 167–172. Amundsen, T. (2000). Why are female birds ornamented? Trends Ecol. Evol. 15, 149–155.


163

Andersson, M. B. (1994). ‘‘Sexual Selection.’’ Princeton University Press, Princeton, NJ. Bateman, A. J. (1948). Intra-sexual selection in Drosophila. Heredity 2, 349–368. Berglund, A. (1991). Egg competition in a sex-role reversed pipefish: Subdominant females trade reproduction for growth. Evolution 45, 770–774. Berglund, A. (1993). Risky sex: Male pipefishes mate at random in the presence of a predator. Anim. Behav. 46, 169–175. Berglund, A. (1994). The operational sex ratio influences choosiness in a pipefish. Behav. Ecol. 5, 254–258. Berglund, A. (1995). Many mates make male pipefishes choosy. Behaviour 132, 213–218. Berglund, A. (2000). Sex role reversal in a pipefish: Female ornaments as amplifying handicaps. Ann. Zool. Fenn. 37, 1–13. Berglund, A., and Rosenqvist, G. (1990). Male limitation of female reproductive success in a pipefish: Effects of body size differences. Behav. Ecol. Sociobiol. 27, 129–133. Berglund, A., and Rosenqvist, G. (1993). Selective males and ardent females in pipefishes. Behav. Ecol. Sociobiol. 32, 331–336. Berglund, A., and Rosenqvist, G. (2001a). Male pipefish prefer ornamented females. Anim. Behav. 61, 345–350. Berglund, A., and Rosenqvist, G. (2001b). Male pipefish prefer dominant over attractive females. Behav. Ecol. 12, 402–406. Berglund, A., Rosenqvist, G., and Svensson, I. (1986a). Mate choice, fecundity and sexual dimorphism in two pipefish species (Syngnathidae). Behav. Ecol. Sociobiol. 19, 301–307. Berglund, A., Rosenqvist, G., and Svensson, I. (1986b). Reversed sex roles and parental energy investment in zygotes of two pipefish (Syngnathidae) species. Marine Ecol. Prog. Ser. 29, 209–215. Berglund, A., Rosenqvist, G., and Svensson, I. (1988). Multiple matings and paternal brood care in the pipefish Syngnathus typhle. Oikos 51, 184–188. Berglund, A., Rosenqvist, G., and Svensson, I. (1989). Reproductive success of females limited by males in two pipefish species. Am. Nat. 133, 506–516. Berglund, A., Bisazza, A., and Pilastro, A. (1996). Armaments and ornaments: An evolutionary explanation of traits of dual utility. Biol. J. Linn. Soc. Lond. 58, 385–399. Berglund, A., Rosenqvist, G., and Bernet, P. (1997). Ornamentation predicts reproductive success in female pipefish. Behav. Ecol. Sociobiol. 40, 145–150. Bernet, P., Rosenqvist, G., and Berglund, A. (1998). Female–female competition affects female ornamentation in the sex-role reversed pipefish Syngnathus typhle. Behaviour 135, 535–550. Boake, C. R. B., and Capranica, R. R. (1982). Aggressive signal in ‘‘courtship’’ chirp of a gregarious cricket. Science 218, 580–582. Bradbury, J. W., and Vehrencamp, S. L. (1998). ‘‘Principles of Animal Communication.’’ Sinauer Associates, Sunderland, MA. Burley, N., Tidemann, S. C., and Halupka, K. (1991). Bill colour and parasite levels of zebra finches. In ‘‘Bird–Parasite Interactions’’ (J. E. Loye and M. Zuk, Eds.), pp. 359–376. Oxford University Press, Oxford. Chappell, M. A., Zuk, M., Johnsen, T. S., and Kwan, T. H. (1997). Mate choice and aerobic capacity in red junglefowl. Behaviour 134, 511–529. Clutton-Brock, T. H., and Parker, G. A. (1992). Potential reproductive rates and the operation of sexual selection. Q. Rev. Biol. 67, 437–455 . Clutton-Brock, T. H., and Vincent, A. (1991). Sexual selection and the potential reproductive rates of males and females. Nature 351, 58–60. Darwin, C. (1871). ‘‘The Descent of Man and Selection in Relation to Sex.’’ Murray, London. Dawson, C. E. (1985). ‘‘Indo-Pacific Pipefishes.’’ Allen Press, Lawrence, KS.

164


Drickamer, L. C., Gowaty, P. A., and Holmes, C. M. (2000). Free female choice in house mouse affects reproductive success and offspring viability and performance. Anim. Behav. 59, 371–378. Emlen, S. T. (1976). Lek organization and mating strategies in the bullfrog. Behav. Ecol. Sociobiol. 1, 283–313. Emlen, S. T., and Oring, L. W. (1977). Ecology, sexual selection and the evolution of mating systems. Science 197, 215–223. Fiedler, K. (1954). Vergleichende Verhaltensstudien an Seenadeln, Schlangennadeln und Seepferdchen (Syngnathidae). Z. Tierpsychol. 11, 358–416. Fitzpatrick, S. (1998). Birds’ tails as signaling devices: Markings, shape, length, and feather quality. Am. Nat. 151, 157–173. Fitzpatrick, S., Berglund, A., and Rosenqvist, G. (1995). Ornaments or offspring: Costs to reproductive success restrict sexual selection processes. Biol. J. Linn. Soc. Lond. 55, 251–260. Folstad, I., and Karter, A. J. (1992). Parasites, bright males, and the immunocompetence handicap. Am. Nat. 139, 603–622. Forsgren, E., Reynolds, J., and Berglund, A. (2002). Behavioural ecology of reproduction. In ‘‘Handbook of Fish and Fisheries’’ (P. Hart and J. Reynolds, Eds.). Blackwell Science, Oxford. Fuller, R., and Berglund, A. (1996). Behavioral responses of a sex-role reversed pipefish to a gradient of perceived predation risk. Behav. Ecol. 7, 69–75. Gross, M. R., and Sargent, R. C. (1985). The evolution of male and female parental care in fishes. Am. Zool. 25, 807–822. Hamilton, W. D., and Zuk, M. (1982). Heritable true fitness and bright birds: A role for parasites. Science 218, 386–387. Haresign, T. W., and Schumway, S. E. (1981). Permeability of the marsupium of the pipefish Syngnathus fuscus to [14C]- amino isobutyric acid. Comp. Biochem. Physiol. 69A, 603–604. Hasson, O. (1989). Amplifiers and the handicap principle in sexual selection: A different emphasis. Proc. R. Soc. Lond. B Biol. Sci. 235, 383–406. Hasson, O. (1990). The role of amplifiers in sexual selection: An integration of the amplifying and the Fisherian mechanisms. Evol. Ecol. 4, 277–289. Hasson, O. (1991). Sexual displays as amplifiers: Practical examples with an emphasis on feather decorations. Behav. Ecol. 2, 189–197. Hasson, O. (1997). Towards a general theory of biological signaling. J. Theor. Biol. 185, 139–156. Hasson, O. (1999). Knowledge, information, biases and signal assemblages. In ‘‘Animal Signals: Signalling and Signal Design in Animal Communication’’ (Y. Espmark, T. Amundsen, and G. Rosenqvist, Eds.), pp. 107–124. Tapir Publishers, Trondheim, Norway. Herald, E. S. (1959). From pipefish to seahorse: A study of phylogenetic relationships. Proc. Calif. Acad. Sci. 4th Ser. 29, 465–473. Heywood, J. S. (1989). Sexual selection by the handicap mechanism. Evolution 43, 1387–1397. Hoelzer, G. A. (1989). The good parent process of sexual selection. Anim. Behav. 38, 1067–1078. Houde, A. E., and Torio, A. J. (1992). Effect of parasitic infection on male color pattern and female choice in guppies. Behav. Ecol. 3, 346–351. Iwasa, Y., Pomiankowski, A., and Nee, S. (1991). The evolution of costly mate preferences. II. The ‘‘handicap’’ principle. Evolution 45, 1431–1442. Ja¨rvi, T., and Bakken, M. A. (1984). The function of the variation in the breast-stripe of the great tit (Parus major). Anim. Behav. 32, 590–596.


165

Johnson, L. L., and Boyce, M. S. (1991). Female choice of males with low parasite loads in sage grouse. In ‘‘Bird–Parasite Interactions: Ecology, Evolution and Behavior’’ (J. E. Loye and M. Zuk, Eds.), pp. 377–388. Oxford University Press, Oxford. Johnstone, R. A. (1995). Sexual selection, honest advertisement and the handicap principle: Reviewing the evidence. Biol. Rev. 70, 1–65. Johnstone, R. A., and Norris, K. (1993). Badges of status and the cost of aggression. Behav. Ecol. Sociobiol. 32, 127–134. Johnstone, R. A., Reynolds, J. D., and Deutsch, J. C. (1996). Mutual mate choice and sex differences in choosiness. Evolution 50, 1382–1391. Jones, A. G., and Avise, J. C. (1997). Microsatellite analysis of maternity and the mating system in the Gulf pipefish Syngnathus scovelli, a species with male pregnancy and sexrole reversal. Mol. Ecol. 6, 203–213. Jones, A. G., Kvarnemo, C., Moore, G. I., Simmons, L. W., and Avise, J. C. (1998). Microsatellite evidence for monogamy and sex-biased recombination in the Western Australian seahorse Hippocampus angustus. Mol. Ecol. 7, 1497–1505. Jones, A. G., Rosenqvist, G., Berglund, A., and Avise, J. C. (1999). The genetic mating system of a sex-role-reversed pipefish (Syngnathus typhle): A molecular inquiry. Behav. Ecol. Sociobiol. 46, 357–365. Jones, A. G., Rosenqvist, G., Berglund, A., and Arnold, S. J. (2000a). The Bateman gradient and the cause of sexual selection in a sex-role-reversed pipefish. Proc. R. Soc. Lond. B Biol. Sci. 267, 2151–2155. Jones, A. G., Rosenqvist, G., Berglund, A., and Avise, J. C. (2000b). Mate quality influences multiple maternity in the sex-role-reversed pipefish Syngnathus typhle. Oikos 90, 321–326. Jones, I. L. (1990). Plumage variability functions for status signalling in least auklets. Anim. Behav. 39, 967–975. Jones, I. L., and Hunter, F. M. (1999). Experimental evidence for mutual inter- and intrasexual selection favouring a crested auklet ornament. Anim. Behav. 57, 521–528. Jones, T. M., Quinnell, R. J., and Balmford, A. (1998). Fisherian flies: Benefits of female choice in a lekking sand fly. Proc. R. Soc. Lond. B Biol. Sci. 265, 1651–1657. Kennedy, C. E. J., Endler, J. A., Poynton, S. L., and McMinn, H. (1987). Parasite load predicts mate choice in guppies. Behav. Ecol. Sociobiol. 21, 291–295. Kokko, H., and Monaghan, P. (2001). Predicting the direction of sexual selection. Ecol. Lett. 4, 159–165. Kvarnemo, C., and Ahnesjo¨, I. (1996). The dynamics of operational sex ratios and competition for mates. Trends Ecol. Evol. 11, 404–408. Kvarnemo, C., Moore, G. I., Jones, A. G., Nelson, W. S., and Avise, J. C. (2000). Monogamous pair bonds and mate switching in the Western Australian seahorse Hippocampus subelongatus. J. Evol. Biol. 13, 882–888. Linton, J. R., and Soloff, B. L. (1964). The physiology of the brood pouch of the male seahorse. Bull. Marine Sci. 14, 56–61. Masonjones, H. D., and Lewis, S. M. (2000). Differences in potential reproductive rates of male and female seahorses related to courtship roles. Anim. Behav. 59, 11–20. Matsumoto, K., and Yanagisawa, Y. (2001). Monogamy and sex role reversal in the pipefish Corythoichthys haematopterus. Anim. Behav. 61, 163–170. Mayer, I., Rosenqvist, G., Borg, B., Ahnesjo¨, A., Berglund, A., and Schulz, R. W. (1993). Plasma levels of sex steroids in three species of pipefish (Syngnathidae). Can. J. Zool. 71, 1903–1907. Maynard-Smith, J., and Harper, D. G. C. (1988). The evolution of aggression: Can selection generate variability? Philos. Trans. R. Soc. Lond. B Biol. Sci. 319, 557–570.

166


Milinski, M., and Bakker, T. C. M. (1990). Female sticklebacks use male coloration in mate choice and hence avoid parasitized males. Nature 344, 331–333. Møller, A. P. (1987a). Social control of deception among status signalling house sparrows Passer domesticus. Behav. Ecol. Sociobiol. 20, 307–311. Møller, A. P. (1987b). Variation in badge size in male house sparrows Passer domesticus: Evidence for status signalling. Anim. Behav. 35, 1637–1644. Møller, A. P. (1988). Badge size in the house sparrow Passer domesticus: Effects of intra- and intersexual selection. Behav. Ecol. Sociobiol. 22, 373–378. Møller, A. P. (1990). Parasites and sexual selection: Current status of the Hamilton and Zuk hypothesis. J. Linn. Soc. 48, 187–211. Møller, A. P., and Alatalo, R. V. (1999). Good-genes effects in sexual selection. Proc. R. Soc. Lond. B Biol. Sci. 266, 85–91. Nicoletto, P. F. (1995). Offspring quality and female choice in the guppy, Poecilia reticulata. Anim. Behav. 49, 377–387. Owens, I. P. F., and Hartley, I. R. (1991). ‘‘Trojan sparrows’’: Evolutionary consequences of dishonest invasion for the badges-of-status model. Am. Nat. 138, 1187–1205. Owens, I. P. F., and Thompson, D. B. A. (1994). Sex differences, sex ratios and sex roles. Proc. R. Soc. Lond. B Biol. Sci. 258, 93–99. Parker, G. A., and Simmons, L. W. (1996). Parental investment and the control of sexual selection: Predicting the direction of sexual competition. Proc. R. Soc. Lond. B Biol. Sci. 263, 315–321. Partridge, L. (1980). Mate choice increases a component of offspring fitness in fruit flies. Nature 283, 290–291. Petrie, M. (1994). Improved growth and survival of offspring of peacocks with more elaborate trains. Nature 371, 598–599. Qvarnstro¨m, A. (1997). Experimentally increased badge size increases male competition and reduces male parental care in the collared flycatcher. Proc. R. Soc. Lond. B Biol. Sci. 264, 1225–1231. Reynolds, J. D. (1996). Animal breeding systems. Trends Ecol. Evol. 11, 68–72. Reynolds, J. D., and Gross, M. R. (1992). Female mate preference enhances offspring growth and reproduction in a fish, Poecilia reticulata. Proc. R. Soc. Lond. B Biol. Sci. 250, 57–62. Rohwer, S. (1977). Status signaling in Harris sparrows: Some experiments in deception. Behaviour 61, 107–129. Rohwer, S., and Rohwer, F. C. (1978). Status signalling in Harris sparrows: Experimental deceptions achieved. Anim. Behav. 26, 1012–1022. Rosenqvist, G. (1990). Male mate choice and female–female competition for mates in the pipefish Nerophis ophidion. Anim. Behav. 39, 1110–1115. Rosenqvist, G., and Johansson, K. (1995). Male avoidance of parasitized females explained by direct benefits in a pipefish. Anim. Behav. 49, 1039–1045. Sandvik, M., Rosenqvist, G., and Berglund, A. (2000). Male and female mate choice affects offspring quality in a sex role reversed pipefish. Proc. R. Soc. Lond. B Biol. Sci. 267, 2151–2155. Schliwa, M. (1986). Pigment cells. In ‘‘Biology of the Integument of Vertebrates’’ (R. HahnMatoltsy, Ed.), pp. 56–77. Springer-Verlag, Berlin. Sindermann, C. J. (1966). Diseases of marine fish. Adv. Mar. Biol. 4, 1–89. Studd, M. V., and Robertson, R. J. (1985). Evidence for reliable badges of status in territorial yellow warblers (Dendroica petechia). Anim. Behav. 33, 1102–1113. Stunkard, H. J. (1930). The life history of Cryptocolyle lingua. J. Morphol. Physiol. 50, 143–191. Svensson, I. (1988). Reproductive costs in two sex role reversed pipefish species (Syngnathidae). J. Anim. Ecol. 57, 929–942.


167

Taylor, P. W., Hasson, O., and Clark, D. L. (2000). Body postures and patterns as amplifiers of physical condition. Proc. R. Soc. Lond. B Biol. Sci. 267, 917–922. Thulin, J. (1971). Parasitiska plattmaskar hos torsk. Zool. Rev. 33, 76–84. Trivers, R. L. (1972). Parental investment and sexual selection. In ‘‘Sexual Selection and the Descent of Man, 1871–1971’’ (B. Campbell, Ed.), pp. 136–179. Aldine, Chicago. Vincent, A., Ahnesjo¨, I., Berglund, A., and Rosenqvist, G. (1992). Pipefishes and sea-horses: Are they all sex role reversed? Trends Ecol. Evol. 7, 237–241. Vincent, A., Ahnesjo¨, I., and Berglund, A. (1994). Operational sex ratios and behavioural sex differences in a pipefish population. Behav. Ecol. Sociobiol. 34, 435–442. Vincent, A., Berglund, A., and Ahnesjo¨, I. (1995). Reproductive ecology of five pipefish species in one eelgrass meadow. Environ. Biol. Fishes 44, 347–361. Vincent, A. C. J. (1994). Seahorses exhibit conventional sex roles in mating competition, despite male pregnancy. Behaviour 128, 135–151. Vincent, A. C. J., and Sadler, L. M. (1995). Faithful pair bonds in wild seahorses, Hippocampus whitei. Anim. Behav. 50, 1557–1569. Williams, G. C. (1966). ‘‘Adaptation and Natural Selection.’’ Princeton University Press, Princeton, NJ. Wilson, A. B., Vincent, A., Ahnesjo¨, I., and Meyer, A. (2001). Male pregnancy in seahorses and pipefishes (Family Syngnathidae): rapid diversification of paternal brood pouch morphology inferred from a molecular phylogeny. J. Hered. 92, 159–166. Wootton, R. J. (1984). ‘‘A Functional Biology of Sticklebacks.’’ Croom Helm, London. Zahavi, A., and Zahavi, A. (1997). ‘‘The Handicap Principle: A Missing Piece of Darwin’s Puzzle.’’ Oxford University Press, Oxford. Zuk, M. (1996). Disease, endocrine-immune interactions, and sexual selection. Ecology 77, 1037–1042. Zuk, M., Johnson, K., Thornhill, R., and Ligon, J. D. (1990). Parasites and male ornaments in free-ranging and captive red jungle fowl. Behaviour 114, 232–248.



Fluctuating Asymmetry, Animal Behavior, and Evolution John P. Swaddle biology department college of william & mary williamsburg, virginia 23187

I. What Is Fluctuating Asymmetry and Why Is It Interesting? Fluctuating asymmetry refers to small deviations from a prior expectation of symmetric development in morphological traits (Ludwig, 1932). Across a population, signed asymmetries (i.e., signed difference between left and right sides) tend to show a normal (or leptokurtic) distribution where the mean is zero (i.e., symmetry) and individuals with relatively large asymmetries are rare. These morphological asymmetries are hypothesized to result from imperfect development, and are thought to reflect the inability of the genome to buffer developmental processes against intrinsic, random noise (Ludwig, 1932; Waddington, 1957; Zakharov, 1992). There are many genetic and environmental factors that can disrupt developmental stability and increase noise and asymmetry (review in Møller and Swaddle, 1997). However, the response of asymmetry to stressors appears to be taxon and trait specific (Leung and Forbes, 1996) because, in the traits of some species, fluctuating asymmetry does not appear to be affected even by severe, mortality-inducing stresses (Bjorksten et al., 2001). It is therefore clear that the relationship between asymmetry and genetic and environmental stresses is not straightforward: asymmetry in one trait may result from different stressors than asymmetry in another trait. Despite the lack of generality of such responses to stress, fluctuating asymmetry has often been used as an indicator of developmental instability (reviews in Clarke et al., 1986; Parsons, 1992; Møller and Swaddle, 1997). Developmental stability is most commonly defined as the production of a predicted phenotype from a specified genotype in a particular environmental setting (Zakharov, 1992). Developmental instability reflects the 169 Copyright 2003 Elsevier Science (USA). All rights reserved. 0065-3454/03 $35.00

170

JOHN P. SWADDLE

inability of the genome and developmental pathways to suppress random noise during development (Palmer and Strobeck, 1986; Zakharov, 1992). Therefore, to assess developmental instability a way of recording the degree of noise associated with development needs to be devised. As fluctuating asymmetry indicates the population variance around expected symmetric development, many researchers have proposed that fluctuating asymmetry is a strong candidate for assessing developmental instability (Mather, 1953; Clarke et al., 1986; Zakharov, 1992). The strongest empirical evidence for developmental instability as a cause of fluctuating asymmetry comes from studies of the ability of single genotypes to resist perturbations at different magnitudes of environmental stress (Rettig et al., 1997; Swaddle and Witter, 1998; Shykoff and Møller, 1999). However, such experiments do not always reveal genetically related responses to stress (e.g., Perfectti and Camacho, 1999; Andalo et al., 2000). This topic is treated in more detail in Section IV.A. If fluctuating asymmetry does reflect developmental instability, it is possible that phenotypic asymmetry can reveal information about the fitness of populations (Jones, 1987) and individual quality (Møller, 1990). To this end, many behavioral ecologists have tried to describe and understand the possible relationships between fluctuating asymmetry and fitness indicators (as briefly reviewed in Section II). The increased interest in fluctuating asymmetry–fitness relations has inspired a wide range of evolutionary biologists to focus attention on fluctuating asymmetry. Studies have begun to elucidate the genetic and developmental origins of these small asymmetries. By marrying the interest in fitness associations and the origins of fluctuating asymmetry, we have started to understand the evolutionary potential and importance of these small developmental asymmetries in both a behavioral and evolutionary context. Although interest in fluctuating asymmetry has become intense (more than 200 studies of fluctuating asymmetry are indexed in Web of Science (http://www.isinet.com/isi/products/citation/wos) for the period January 1 to September 1, 2001), there is a general feeling that our level of understanding has not progressed at the same pace (Palmer, 1996b; Markow and Clarke, 1997; Møller and Swaddle, 1997; Houle, 1998; Van Dongen et al., 2001). This review is intended as a critical, yet constructive, analysis of the importance of fluctuating asymmetry to studies of animal behavior and evolution. For those new to this area of research, this chapter indicates the growing complexity of fluctuating asymmetry analysis and highlights common pitfalls (Section III). In addition to remarking on past flaws, this review also aims to indicate the major empirical and theoretical gaps in our knowledge about fluctuating asymmetry (Section IV). In many cases, fundamental information concerning the genetic and developmental

FLUCTUATING ASYMMETRY, ANIMAL BEHAVIOR, AND EVOLUTION

171

origins, and possible behavioral consequences, of fluctuating asymmetry is simply lacking. For example, despite intense interest in asymmetry as a cue during mate choice, we do not know whether animals can actually perceive small morphological asymmetries (discussed in Section V.B). As a particular focus for this review, one area of behavioral research in which fluctuating asymmetry has been frequently studied—sexual selection—is revisited, in light of our current knowledge, to indicate how work could progress in the near future (Section V).

II. Fluctuating Asymmetry and Fitness For asymmetry to be important to studies of the evolution of animal behavior, a crucial question to address is whether asymmetry has any adaptive significance. The most popular way to answer this question, thus far, has been to look for associations between asymmetry and fitness and to study how these relationships are determined. There have been numerous reviews of the relationship between fluctuating asymmetry and fitness indicators (Markow, 1995; Leung and Forbes, 1996; Palmer, 1996b, 2000; Møller, 1997; Møller and Swaddle, 1997; Clarke, 1998b; Thornhill and Møller, 1998; Simmons et al., 1999a; Swaddle, 1999b; Van Dongen, 2001; Zakharov, 2001). There have also been a series of metaanalyses of fluctuating asymmetry in relation to various parameters, including fitness (Leung and Forbes, 1996; Møller and Thornhill, 1997; Thornhill and Møller, 1998; Vøllestad et al., 1999), but these have generated much controversy (Markow and Clarke, 1997; Whitlock and Fowler, 1997; Palmer, 1999, 2000; Simmons et al., 1999a). A major cause of controversy is the considerable inconsistency and flaws in methods used in different studies. In many cases it is not clear that authors have quantified and analyzed actual fluctuating asymmetry (rather than their own measurement error). Additional problems have been the inaccurate representation of published data, and the underreporting of nonsignificant or ‘‘opposite’’ results (Whitlock and Fowler, 1997; Palmer, 1999, 2000; Simmons et al., 1999a). One particular complexity in employing metaanalyses is that researchers must go to great lengths to track down unpublished data (commonly referred to as the ‘‘file drawer problem’’). As much fluctuating asymmetry research appears to have been performed on an ad hoc basis (Section VI), this presents a substantial challenge to any metaanalytic technique investigating an asymmetry relationship. For these reasons, this chapter does not attempt a metaanalysis as there appears little possibility that such a review could be satisfactorily comprehensive.

172

JOHN P. SWADDLE

One element of the reviews on which there is wide agreement is the inconsistency in the relation between asymmetry and fitness. Fluctuating asymmetry of some traits is correlated with fitness indicators in some organisms, but not others (Leung and Forbes, 1996; Swaddle, 1999b). This pattern of inconsistency has not changed with the addition of more data. For example, feather and tarsal asymmetry are not related to levels of parasitism in the pigeon Columbia livia (Quek et al., 1999). However, floral asymmetry in Linum usitatissimum and Brassica rapa (Salonen and Lammi, 2001) and fin asymmetry in the male gobie Pomatoschistus microps (Sasal and Pampoulie, 2000) are positively related to levels of parasitism. There is a negative relation between horn asymmetry, female condition, and some (but not all) life history traits in the mountain goat Oreamnos americanus (Cote and Festa-Bianchet, 2001). However, feather and skeletal asymmetry are not related to body condition in the red-collared widowbird Euplectes ardens (Goddard and Lawes, 2000). Probability of survival is negatively related to tarsal asymmetry in the water boatman Callicorixa vulnerata (Nosil and Reimchen, 2001) and the striped dolphin Stenella coeruleoalba (Pertoldi et al., 2000). However, many life history traits are not related to morphological asymmetry in the winter moth Operophtera brumata (Van Dongen et al., 1999b) and the brook stickleback Culaea inconstans (Hechter et al., 2000). The size of spermatophore passed from male to female in the field cricket Gryllodes sigillatus is negatively related to female appendage asymmetry (Farner and Barnard, 2000), which may indicate that males make mating decisions in relation to female asymmetry (although probably indirectly). In addition, there is some evidence to indicate that asymmetric male bushcrickets (Requena vaticalis) invest more highly in the nutritional content of their spermatophore, which was interpreted as asymmetric males investing more in parental effort (Simmons et al., 1999b). However, neither male nor female limb asymmetry is related to ejaculate size in the moth Plodia interpunctella (Gage, 1998). The preceding descriptions skim the surface of more recent studies, but they illustrate the diversity of interest in fluctuating asymmetry, the range of techniques used, and the variability in traits and taxa studied. It appears that some of the relationships between fitness and asymmetry found arise from direct influences of asymmetry. For example, asymmetry in mechanical traits can directly decrease competitive ability (Sneddon and Swaddle, 1999), auditory abilities (Bosch and Marquez, 2000), mating ability (Blackenhorn et al., 1998), or predation (Swaddle, 1997b). Alternatively, asymmetry could be used as a direct visual cue in mate selection (Swaddle and Cuthill, 1994a; Morris and Casey, 1998). By


173

contrast, other relationships are mediated by indirect associations between asymmetry and fitness. For example, symmetric males may produce more attractive pheromones (Thornhill, 1992; Martin and Lopez, 2000). In many cases, authors have looked for broad correlations between asymmetry and fitness parameters, and so cannot distinguish between direct and indirect relationships (review in Swaddle, 1999b). As asymmetry can be related to trait size, many of the observational studies have also not accounted for how size–fitness relationships can give rise to relationships between trait asymmetry and fitness (Nachman and Heller, 1999). When there does appear to be a relationship between asymmetry and fitness, we rarely understand how this relationship is mediated. One report has highlighted a powerful experimental technique to study relationships between developmental instability and fitness (Shykoff and Møller, 1999). By comparing the within-individual change (between successive feather molts) in outer tail length asymmetry in the barn swallow Hirundo rustica with changes in reproductive success, they found that individuals that increased in asymmetry (i.e., those that experienced an increase in developmental stress between molts) had reduced success. By exploring within-individual changes in asymmetry, we may draw closer to measures of developmental instability (as there is more than one measure of asymmetry for a given genotype; see Section IV.A) and, hence, be able to relate this parameter to evolutionarily important characters (such as reproductive success). This type of experiment has also been adopted in studies of clonal plants, but unlike the case of barn swallows, these data indicate that there are no relations between genotypeassociated asymmetry and fitness indicators (e.g., Andalo et al., 2000). Repeated lines of Drosophila are also an excellent model for studying fluctuating asymmetry–fitness relations. By producing known mutations in flies against the same genetic background, Bourquet (2000) produced lines of flies with various degrees of fluctuating asymmetry in sternopleural bristle number. However, there were no associations between bristle fluctuating asymmetry and either of two fitness indicators: reproductive success or competitive male mating success (Bourquet, 2000). It will be important to expand these experimental paradigms to other biological systems. At present, it appears that the relations between fitness and fluctuating asymmetry cannot be generalized and are trait specific.

III. Methodology Issues As alluded to in the previous section, there are many methodological flaws among studies of fluctuating asymmetry. This is a major problem that

174

JOHN P. SWADDLE

has been discussed many times, but still goes unheeded in many instances. Some of the mandatory first steps that every study needs to follow are summarized below. For more detailed discussion readers should refer to the many helpful articles on this topic (e.g., Palmer and Strobeck, 1986; Cuthill et al., 1993; Palmer, 1994; Swaddle et al., 1994; Van Dongen, 1998a,b; Van Dongen et al., 1999a). A. Identifying Fluctuating Asymmetry Although identifying fluctuating asymmetry appears simple, it is important to distinguish it from directional asymmetry (where asymmetries are large and one side of the paired trait is predictably larger than the other, e.g., the left side is larger than the right in mammalian hearts) or antisymmetry (where asymmetries are also large but it is not possible to predict which side of the trait will be larger, e.g., either the left or right front claw can develop into the exaggerated signaling claw of the male fiddler crab Uca), as these other forms could lead to different predictions concerning the relationships between asymmetry and evolutionarily important parameters. For example, antlers of the roe deer Capreolus capreolus appear to display fluctuating asymmetry and this asymmetry is negatively correlated with survival (Pe´labon and van Breukelen, 1998). But, curiously, antlers of the fallow deer Dama dama display directional asymmetry, and this asymmetry is not related to quality indicators (Pe´ labon and Joly, 2000). Fluctuating asymmetry and directional asymmetry are fundamentally different forms of asymmetry with different developmental origins. Directional asymmetries are preprogrammed lateral differences, whereas fluctuating asymmetries are deviations from what is normally perfect symmetry. Some researchers have reported fluctuating asymmetry ‘‘turning’’ into directional asymmetry or antisymmetry under increasing stress (Mather, 1953; McKenzie and Clarke, 1988; Graham et al., 1993; Lens and Van Dongen, 2000). However, it is difficult to understand how such a massive restructuring of developmental programs would occur in such a short period of time. Antisymmetry also appears to result from developmental processes distinct from fluctuating asymmetry (Van Valen, 1962; Palmer, 1996a). Interestingly, Rowe and colleagues have shown that several of the early studies of fluctuating asymmetry, including many of Møller’s studies (Møller, 1990, 1992a; Møller and Eriksson, 1994), were apparently measuring antisymmetry–fitness relations (Rowe et al., 1997). Therefore, in some cases it appears that antisymmetry could be condition dependent. However, it is not clear whether these are isolated cases. It is also possible that fluctuating asymmetry could appear like antisymmetry at small sample sizes.


175

It has also been suggested that variation from a population mean directional asymmetry may be equivalent to fluctuating asymmetry, and there are ways to ‘‘adjust’’ measured directional asymmetry to become fluctuating asymmetry (Graham et al., 1998). However, it is not clear, in an asymmetry that is (or was recently) under directional selection, whether a positive-signed deviation from the predicted form (e.g., left bigger than right) is equivalent to a negative-signed deviation (e.g., right bigger than left). We can make this assumption with fluctuating asymmetry but it is far more difficult with directional asymmetry, as directional selection on asymmetry suggests there is a difference in reproductive success for leftbiased versus right-biased traits. Therefore, deviation from a mean directional asymmetry can often not be interpreted in the same way as variation in fluctuating asymmetry (e.g., Leamy et al., 2000). B. Trait Selection Most research has focused on traits that are straightforward to measure—which is understandable as this will tend to reduce measurement error and make asymmetry measures more accurate. However, we should ask whether the asymmetries that people measure are behaviorally meaningful and whether there are others forms of asymmetry that should be measured. There are ways of measuring more complex shape asymmetry, that are especially important in studying the development of integrated units (e.g., Klingenberg and McIntyre, 1998; Mardia et al., 2000; Klingenberg et al., 2001). In these particular examples, researchers analyzed overall wing shape asymmetry, which is likely to be relevant to flight behaviors. The techniques employed quantified discrepancies in twodimensional landmarks using established morphometric tools, such as Procrustes. The principles of these techniques could also be extended to three-dimensional morphometric assessments of asymmetry. It has been suggested that assessing asymmetry of integrated units (or developmentally correlated traits) will render a more accurate estimate of true developmental instability than single-trait measures (Polak and Starmer, 2001). This makes sense, as this technique samples the same developmental instability with more than two data points. Combining sizestandardized asymmetry measures from multiple traits into a single index (Windig and Nylin, 2000) could give an overall indicator of bodily asymmetry (Woods et al., 1999). However, as fluctuating asymmetry appears somewhat trait specific, it is not clear what such a measure would indicate if traits that are not developmentally linked were combined. Even if all asymmetry measures are related to developmental instability, different traits may have differing buffering capacities, varying

176

JOHN P. SWADDLE

susceptibility to environmental stressors (due to heterogeneity in biochemical pathways in development), and differing ontogenetic patterns and timing of asymmetry development. Therefore, interpreting multitrait indices of asymmetry (where traits are not developmentally linked) could be more problematic than interpreting single traits. When researchers have used multitrait indices, it is likely that they are sampling from a large part of the genome and that their data reveal responses to a combination of developmental conditions (Leung et al., 2000; Hewa-Kapuge and Hoffmann, 2001). Not all of development can be summarized by size and shape. Therefore, due consideration should be given to other forms of asymmetry. For example, position of traits on the left or right side, or coloration, could yield meaningful estimates of asymmetry in the appropriate systems (e.g., Polak, 1997; Martin and Lopez, 2000). Studies of these kinds have started to appear, but are still rare.

C. Measurement Error As fluctuating asymmetry is relatively small, measurement error can swamp accurate estimates of asymmetry. Therefore, performing repeated measurements on the same individuals is essential to measuring fluctuating asymmetry. Some authors experienced at measuring fluctuating asymmetry are still performing repeat measures on only a subset of samples in their study (Cuervo and Møller, 1999; Woods et al., 1999; Andalo et al., 2000; Klingenberg et al., 2001; Polak and Starmer, 2001). In most cases, it is not good enough to perform repeatability tests on a subset of the study sample. It is imperative to perform the repeats on all the samples to minimize measurement error. Several articles have discussed this (e.g., Palmer, 1994; Swaddle et al., 1994; Merila¨ and Bjo¨rklund, 1995; Van Dongen et al., 1999a).

D. Sampling Fluctuating asymmetry measurements are error prone and may have weak associations with the parameter of real interest (i.e., developmental instability; see Section IV.A). Therefore, obtaining a large and unbiased sample is important. As people have suggested that fluctuating asymmetry is related to survival, sampling from natural populations at particular age intervals may unwittingly bias samples in terms of fluctuating asymmetry (Møller and Swaddle, 1997). Because of the inherent random nature of asymmetries, it will often require a large sample size ( >100 individuals) to


177

accurately estimate the variance associated with fluctuating asymmetry (Mogie and Cousins, 2001). Analysis of statistical power is often overlooked but is important to studies of fluctuating asymmetry, as many conclusions tend to be drawn from a lack of statistical significance (Palmer, 1996b). The power of many tests in the literature appears to be less than 50% (J. P. Swaddle, unpublished data), yet many positive conclusions are drawn from this form of weak evidence. Other than considering the power of individual statistical analyses, Van Dongen and colleagues have applied a number of simulation tests to explore how large effects could arise from more limited sample sizes (Van Dongen, 1999; Van Dongen et al., 2001). Most notably, increasing the number of repeat measurements on the same individuals (from two to nine repeats) can substantially increase the ability to discriminate asymmetry differences between two populations, even at relatively small sample sizes of between 20 and 40 individuals (Van Dongen, 1999). Future studies of fluctuating asymmetry should discuss the power of tests, especially when interpreting the lack of associations between fluctuating asymmetry and other parameters. E. Absolute versus Relative Asymmetry Often, size may be related to dependent variables in addition to asymmetry. Hence, size could mask or alter potential relationships between asymmetry and the dependent variables. Therefore, taking account of size could be necessary (depending on the data at hand). It is only relevant to use a relative measure of asymmetry when the relationship between size and asymmetry is isometric and intercepts the origin (Cuthill et al., 1993). In addition, the relationship between size and the dependent variable would have to be isometric (Nachman and Heller, 1999). This will rarely be the case, and therefore it is necessary to account for size variation with other statistical models, such as analysis of covariance or residual analysis (Palmer, 1994; Swaddle et al., 1994; Leung, 1998). F. Statistical Models As many of the processes associated with determination of fluctuating asymmetry are stochastic, Van Dongen has suggested the use of statistical and modeling methods that treat variables with uncertainty distributions— such as Bayesian hierarchical modeling—as opposed to methods that treat variables as fixed effects (e.g., Van Dongen, 2000). These techniques are being investigated further, but without empirical data to support the assumptions it is not clear what forms of probability distribution should be

178

JOHN P. SWADDLE

used in such modeling. A number of useful analytic tools and suggestions for forms of statistical models have been collated by A. R. Palmer (www.biology.ualberta.ca/palmer.hp/asym/asymmetry.htm). A review of this site before starting a new study would assist many authors.

IV. Important Gaps in Our Knowledge about Fluctuating Asymmetry There is great interest in fluctuating asymmetry as an indicator of developmental instability and fitness, but there are also significant gaps in our knowledge. If the gaps described below can be filled, it will help tremendously in interpreting the data already collected and galvanize future research.

A. Does Fluctuating Asymmetry Reflect Developmental Instability? The amount of asymmetry quantified in populations and individuals is an estimate of their developmental instability. The term ‘‘estimate’’ is important, as it is not known how accurately a single fluctuating asymmetry measure actually represents underlying instability (Whitlock, 1996). If an individual could develop a trait over and over again under identical environmental conditions there would be a certain degree of variability in the resulting phenotypic asymmetry. However, it is unclear how variable this repeated development in identical genetic and environmental conditions would be. At present, it is assumed that two data points (i.e., the left and right sides of a trait) sufficiently sample this theoretical distribution. Several authors have raised this as a substantial problem with interpreting fluctuating asymmetry data (Whitlock, 1996; Houle, 1997; Van Dongen, 1998a). One way of addressing this issue has been developed by Whitlock and Van Dongen, in which (with certain assumptions) it is possible to estimate how much of the variance observed in fluctuating asymmetry is attributable to developmental instability (Whitlock, 1996, 1998; Van Dongen, 1998a). Their techniques concentrate on calculating the repeatability (R) of fluctuating asymmetry production (given an underlying level of developmental instability). This has most recently has been summarized by Whitlock (1998) as 2 2 1 Rffi 2 C2


179

in which C is the coefficient of variation in signed absolute asymmetry, and 3.142. Essentially, R is greater (i.e., the measured asymmetry more closely resembles developmental instability) when fluctuating asymmetry is less variable, and when measurement error has been minimized by numerous repeated measures on both left and right sides. As C can be calculated from many published studies, R has been used to ‘‘correct’’ several estimates of developmental instability—especially in studies of heritability and individual asymmetry parameters (Van Dongen, 1998a; Whitlock, 1998; Gangestad and Thornhill, 1999). Although estimates of R have varied substantially from small values (0.072; Gangestad and Thornhill, 1999) to considerably larger values (0.36; Van Dongen and Lens, 2000), all these studies indicate that the evolutionary significance of developmental stability can be masked by the poor correlation between measured fluctuating asymmetry and true developmental stability. Analyses by Whitlock indicate that R is maximally 0.64, implying that fluctuating asymmetry (based on measurement of left and right sides of a bilateral trait) will never perfectly estimate developmental instability, even if measurement error is zero (Whitlock, 1998). As is true for any model, Whitlock makes a number of assumptions about fluctuating asymmetry and developmental instability. These include the following: (1) values from left and right sides are drawn from the same normal distributions; (2) production of the left and right sides is independent (i.e., the value drawn from the distribution for the left side does not influence the value for the right); and (3) developmental instability follows a normal distribution. These assumptions imply that the model will not hold when either directional asymmetry or antisymmetry is present, or when traits are developmentally integrated (Whitlock, 1996). A further application of the basic Whitlock model illustrates that some of these assumptions may have to be altered. Houle (2000) showed that the model predicts that the coefficient of variation in developmental instability would have to be ‘‘enormous’’ and far greater than for any other trait yet reported. However, a quantitative genetic analysis of bristle counts in Drosophila falleni indicates that genetically related fluctuating asymmetry may be extremely variable, yielding phenotypic coefficients of variation of approximately 85–100% (Polak and Starmer, 2001). Similar estimates of variation in asymmetry have been obtained for clonal cherimoya trees (Annona cherimola) (Perfectti and Camacho, 1999), in which the coefficient of variation for asymmetry was approximately 5 to 10 times greater than that for size of the same traits. Although this particular criticism may not hold, there may have to be other refinements of the model, as Houle also suggests that left and right sides may not be drawn from normal distributions (cf. Klingenberg and Nijhout, 1999) and that

180

JOHN P. SWADDLE

developmental instability may not follow a normal distribution (Houle applies a distribution). Although these modeling techniques are undoubtedly valuable, it will be crucial to explore empirically the repeatability of developmental stability estimates. Although many authors have categorically stated that it is not possible to generate repeated fluctuating asymmetry values for the same genotype under constant environmental conditions (Whitlock, 1996; Van Dongen, 1998a), I think there are productive ways of investigating the repeatability of fluctuating asymmetry. If genotypes and environmental conditions are held constant, it is possible to assess repeatedly bilateral symmetry across a broad taxonomic range by using clones, within- and among-population comparisons of asexual organisms, homogeneous strains of laboratory animals, and perhaps even organisms that show repeated growth of the same trait throughout life (e.g., feather traits regrown after molt in birds: Swaddle, 1997a; Swaddle and Witter, 1998). Interestingly, we already know that isolated lines of Drosophila falleni (originating from single pairs of virgin males and females) maintain consistent differences in asymmetry from each other over several generations (Polak and Starmer, 2001). This implies that each line has a somewhat consistent, genetically related expression of asymmetry. In contrast, the majority of variance in leaf and petal asymmetry in cherimoya trees can be accounted for by within-tree and within-clone variation. Little appears to be genetic in origin (Perfectti and Camacho, 1999). However, there were also significant environmental influences on asymmetry in that study, which may mask any genetic contributions to asymmetry. It would be interesting to repeat such a study under more controlled environmental conditions. In a similar study, there was just as much variation among genotypes as within genotypes for leaf and petal asymmetry in birdsfoot trefoil, Lotus corniculatus (Andalo et al., 2000). There may also be problems with studying asymmetry in plants, as sessile organisms are more likely to experience consistently directed environmental pressures than most animals. Hence, studying the production and genetic origins of asymmetry in clonal animals is particularly appealing. It will be interesting to explore the utility of radial and translational asymmetry in assessing developmental stability, as these forms of asymmetry will render substantially more than two data points per individual and, hence, give a better estimate of the distribution of developmental instability. Plants are well known for their translational symmetry (e.g., repeated development of leaves along an axis) and many invertebrate taxa show translational asymmetry (e.g., Fusco and Minelli, 2000) or pentaradial symmetry (e.g., echinoderms). Studies of this sort will allow researchers to inspect the predictions made by the Whitlock model


181

and give more insight into how reliably fluctuating asymmetry estimates developmental instability. B. Is Fluctuating Asymmetry a Trait? Many authors posit that developmental instability is a genome-wide phenomenon and so, if fluctuating asymmetry reflects developmental instability, there should be an organism-wide indication of asymmetry and among-trait correlations in asymmetry within the same individual (i.e., an ‘‘individual asymmetry parameter’’; Soule´ and Baker, 1968). In general, there is little evidence of an individual asymmetry parameter (review in Clarke, 1998a). Several (nonmutually exclusive) explanations for this have been developed. First, as described in the previous section, there could be a weak association between a single expression of fluctuating asymmetry and developmental instability (Whitlock, 1996). Second, the lack of among-trait correlations could indicate that developmental instability of one trait is not related to that of another. Another suggested explanation is that independent traits have sensitive phases of development at different times so that, if the environment is not constant throughout development, this would result in different levels of asymmetry (Swaddle and Witter, 1997; Hardersen, 2000). In addition, genetic effects could vary with stages of development (Clarke, 1998a). All of these hypotheses suggest that describing asymmetry in one trait may not reveal the same information as asymmetry in another, independent trait. The lack of an individual asymmetry parameter could suggest that independent units of the genome influence asymmetry of different traits separately. The Clarke and McKenzie investigations of asymmetry production in the blowfly Lucilia cuprina in response to specific insecticide resistance are consistent with this hypothesis (McKenzie and Batterham, 1994; Clarke, 1997; Clarke et al., 2000). There probably are not organismwide developmental stability genes. However, it could be that developmental stability is a pleiotropic effect and so is likely to show inter trait variance in properties and heritability. In summary, even if developmental stability is an organism-wide phenomenon, it seems highly unlikely (from both theoretical and empirical evidence) that fluctuating asymmetry can be viewed as a ‘‘trait’’ in itself. It appears to reflect different properties when expressed in different paired traits. This may be due to the inexact nature of fluctuating asymmetry, but it could also reflect the predominance of environmental factors in determining fluctuating asymmetry expression. As the environment will affect the development of independent traits to differing degrees, we should expect to see large variation in fluctuating asymmetry among traits.

182

JOHN P. SWADDLE

However, asymmetry of individual characters can reveal information about developmental conditions and the ability of individuals to buffer development during a specific period of growth. This argument is a relevant expansion of points made in Section III.B, and indicates that trait selection is an extremely important step in any study of fluctuating asymmetry.

C. What Are the Origins of Fluctuating Asymmetry? Not surprisingly, there have been repeated calls to understand the genetic underpinnings of fluctuating asymmetry (Palmer and Strobeck, 1986, 1992; Markow, 1995; Møller and Swaddle, 1997). There are selection experiments that demonstrate that fluctuating asymmetry can be selected both for and against (e.g., Mather, 1953), and limited demonstrations of the heritability of fluctuating asymmetry (see below). There are also continuing searches for genetic correlates of fluctuating asymmetry through quantitative trait locus (QTL) mapping (Leamy et al., 1997, 1998, 2000). However, a common suggestion is that fluctuating asymmetry has little (or perhaps no) genetic origin. For example, a simple pointsource morphogen diffusion-threshold model, which included fluctuating asymmetry as purely random noise associated with independent (heritable) development of left and right sides, indicated that many of the reported ‘‘apparent’’ genetic correlations of fluctuating asymmetry are consistent with nongenetic origins (Klingenberg and Nijhout, 1999). In particular, when relations between developmental variables and the expression of the phenotype were nonlinear, and the developmental variables and the expression of the phenotype were nonlinear, and the developmental variables (controlling the overall growth of left and right sides independently) possessed genetic variation, apparent genetic properties for random developmental noise (i.e., fluctuating asymmetry) could emerge as general properties of the model (Klingenberg and Nijhout, 1999). Trait size and trait asymmetry became correlated under several conditions. This model implies that (given certain assumptions) it is not necessary to invoke any special genetic mechanisms to explain fluctuating asymmetry other than the mechanisms controlling the general growth and development of the left and right sides of traits. That is, there need not be developmental stability genes. In a similar manner, there are suggestions that canalization may be largely ‘‘controlled’’ by nongenetic factors although they appear heritable in experimental situations (Amzallag, 2000). It could be that fluctuating asymmetry is an epigenetic phenomenon.


183

The Klingenberg and Nijhout (1999) model is valuable in showing that certain data sets are consistent with a simple developmental explanation, yet there may still be observations that cannot be explained in this way. The authors themselves point out that there are specific genes in the Australian sheep blowfly Lucilia cuprina that disrupt and also restore developmental stability, yet have no associated effect on overall trait size/ value (unpublished data from J. A. McKenzie referred to in Klingenberg and Nijhout, 1999). Repeated lines of Drosophila melanogaster, with known combinations of mutations against the same genetic background, have consistent among-mutation heterogeneity in sternopleural bristle asymmetry, implying that these mutations affect fluctuating asymmetry in a predictable and quantifiable manner (Bourquet, 2000). In addition, a quantitative genetic study of wild-caught Drosophila falleni lines demonstrated that additive and dominant genetic effects (perhaps with some localization to the X chromosome) influence positional asymmetry of bristles, and that trait size and asymmetry are largely genetically independent (Polak and Starmer, 2001). In addition, studies of identical twins indicate a strong concordance in asymmetry of dermal ridge patterns on the fingertips within genotypes compared with variation among genotypes (Kilgariff et al., 2000). However, common maternal and environmental effects were not accounted for in this study. Although the Klingenberg and Nijhout (1999) model is appealing in its simplicity and ability to explain much of the apparent variation in asymmetry, it appears possible that there are specific genes (or gene complexes) that affect fluctuating asymmetry. The Klingenberg and Nijhout (1999) model is also valuable in that it shows that selection against fluctuating asymmetry is a slow process and that asymmetry can commonly be maintained in populations even in the face of strong truncation selection. This is a topic that causes a number of researchers to doubt whether genetic variance for fluctuating asymmetry could be maintained in populations (Møller and Swaddle, 1997). If there are genes that affect developmental instability, which genes are they and how do they act? To relate this question to the current working models of developmental stability, it is necessary to consider both the production of stochastic ‘‘noise’’ and the regulation (or suppression) of that noise. The generation of developmental noise is not well understood at the genetic level (Roux-Rouquie, 2000), although there are suggestions that some gene expression could occur stochastically (McAdams and Arkin, 1999). In addition, there may be a whole suite of cell-signaling and activation factors that do not occur perfectly, and so introduce random noise at many levels (e.g., binding sites not working to the same degree of efficiency). In terms of regulation of noise, it could be that there are many

184

JOHN P. SWADDLE

gene products that affect the same developmental pathway (i.e., there is redundancy), and so there might be inherent back-up plans to maintain stability that only go awry in rare cases. In addition, genes could affect developmental instability through pleiotropic effects (i.e., a single gene having multiple phenotypic effects). Suitable candidates could be transcriptional factors or the large family of cytokines. At present, all of these hypotheses are purely speculative. In general, there is a growing need for the theoretical advances on the origins of fluctuating asymmetry to be tested by empirical investigations. Leamy and colleagues have encouraged researchers to think of general genetic correlates of asymmetry, and have searched for elements of the genome associated with variance in asymmetry through QTL mapping (Leamy et al., 1997, 1998, 2000). These data illustrate that there may be a small number of loci associated with fluctuating asymmetry and that these behave in a dominant manner (Leamy et al., 1998). However, this pattern could also be consistent with nongenetic origins of fluctuating asymmetry (Klingenberg and Nijhout, 1999). In future studies, it will be important to search for loci associated with fluctuating asymmetry (and developmental instability) under varying environmental stressors. It could be that gene expression varies with environmental conditions and it is only in circumstances of increased stress that genes are activated that increase developmental buffering (cf. Rutherford and Lindquist, 1998). These investigations will provide considerable insight into the direct effects of the environment on the origins of fluctuating asymmetry. It is also necessary to consider how any observable genetic variation in fluctuating asymmetry and developmental instability is related to fitness parameters, preferably within the same experimental situation. Understanding the genetic architecture of fluctuating asymmetry will also reveal levels of genetic and phenotypic redundancy and, hence, could help explain why asymmetry may persist in the face of strong selection against it (as in the barn swallow). My hypothesis is that there is a large amount of genetic redundancy. If so, then many genotypes could give rise to the same level of asymmetry, and this redundancy is a major factor maintaining asymmetry in populations (see Section IV.E). D. Is Fluctuating Asymmetry Heritable? The validity of assessing the heritability of fluctuating asymmetry is contentious. Møller and Thornhill (1997) reviewed the literature and stated that asymmetry shows a small, but significant, heritability across many taxa and traits. However, many researchers have criticized the Møller and Thornhill review and have pointed out flaws in their data,


185

analysis, and interpretations (Leamy, 1997; Markow and Clarke, 1997; Whitlock and Fowler, 1997; Palmer, 2000). In many cases, it would appear that there is little additive genetic variance for fluctuating asymmetry and it is, as yet, unclear how heterogeneous the heritability of fluctuating asymmetry is among taxa and traits (Van Dongen, 2000). If fluctuating asymmetry is not a trait per se, it may be meaningless to try to quantify whether it is generally heritable for particular characters (Section IV.B). Estimates of heritability in multitrait indices of fluctuating asymmetry suggest a greater additive genetic component to developmental instability in some cases (Swaddle, 1997c; Pechenkina et al., 2000) but not all (Bryden and Heath, 2000). By applying corrections for the loose correlation between fluctuating asymmetry and developmental instability, Gangestad and Thornhill have suggested that developmental instability is much more heritable than previously reported (Gangestad and Thornhill, 1999). However, subsequent estimates (Section IV.A) have indicated that the Gangestad and Thornhill estimate of R may not be representative (too low) and that the heritability of developmental instability is itself still low (Van Dongen et al., 2001). Using R as a ‘‘correction’’ factor to make it seem that fluctuating asymmetry is more heritable is inappropriate. R is an indication of how imprecisely fluctuating asymmetry reflects developmental instability and so it seems counterintuitive to use it to support the exactness of fluctuating asymmetry. Surely it adds variance to estimates rather than shifting the mean in any particular direction. It is true that more accurate estimates of developmental stability are needed. If researchers study traits that are believed to reflect developmental instability more closely, it appears that heritability can be substantial (Polak and Starmer, 2001). Researchers should concentrate on more careful experimental design rather than posthoc statistical manipulation of data that were not collected appropriately. Studies that incorporate replicate estimates of the heritability of fluctuating asymmetry in a developmentally linked morphological unit, in which heritability is assessed across a range of environments, will be particularly illuminating. Such experiments are underway (e.g., Polak and Starmer, 2001) but are presently too few for general conclusions. There have also been attempts to analyze reaction norms of asymmetry (Loeschcke et al., 1999; Shykoff and Møller, 1999; Andalo et al., 2000), which indicate that asymmetry changes across environmental gradients. As most reports assess fluctuating asymmetry in terms of absolute asymmetry (i.e., unsigned difference between left and right sides), a change in mean asymmetry will also be associated with a change in asymmetry variance. Altering variance implies that measures of heritability will be different across the environmental gradient. It would be interesting in subsequent

186

JOHN P. SWADDLE

investigations of reaction norms to include data on individual organisms and to show the variation in slopes of individual reaction norms and how much variance is associated with the ‘‘grouped’’ means reported thus far. As it seems that fluctuating asymmetry has (at best) low additive genetic variance, the variation reported in populations must be largely environmentally determined. This has important implications for studies of heritability—as asymmetry could have ‘‘apparent’’ heritability through any mechanism that links the genome with the propensity for individuals to develop and grow under favorable conditions. These could be maternal conditions, or consistent environmental conditions throughout various life cycle stages. If asymmetry is selected against through some proximate mechanism (e.g., mate preference, or predation), genes linked with developing under favorable environmental conditions would be favored. If so, we could observe apparent heritability of asymmetry in natural populations even though the asymmetry has no genetic component. However, this also suggests another area for heritability studies in relation to fluctuating asymmetry. If asymmetry is determined largely by the environment, it would be relevant to assess whether there is heritability of developmental conditions—which extend beyond studies of maternal effects (e.g., inheritance of territory quality, feeding locations, or breeding locations). E. Is Low Fluctuating Asymmetry Adaptive? As can be seen from the preceding treatment of fluctuating asymmetry– fitness relations (Section II), it is difficult to claim that low asymmetry is always adaptive. An even more challenging question to ask is whether fluctuating asymmetry of some traits could be a neutral phenotypic character (and so not related to adaptive behavior or morphology)? Perhaps fluctuating asymmetry is selected against through natural selection only when the asymmetries reach a threshold value (which would vary among traits, and perhaps reach these higher values only in a minority of traits). If this hypothesis is true, it would be predicted that fluctuating asymmetry in some traits, but not others, is related to fitness parameters. This is the pattern we see in nature. Interestingly, a study of fitness correlates in the wasp Trichogramma brassicae indicates that asymmetry must rise beyond a threshold value before it has a negative association with fitness (Hewa-Kapuge and Hoffmann, 2001). As in the case of developmental canalization, variation in fluctuating asymmetry may be exposed only under extreme developmental (genetic and/or environmental) stress. Therefore, only under extreme conditions could asymmetry be selected on—acting to restore developmental stability to


187

prestress levels in that particular environment. Thus, fluctuating asymmetry may have adaptive significance only under these conditions. Fluctuating asymmetry is, in part, determined by the random noise associated with development (e.g., inequalities of cell division, signal receptors not working at equal efficiency). Developmental noise is likely to have multiple causes. Similarly, a range of genotypes can produce the same degree of asymmetry in their phenotype, and the evidence published to date indicates that many genetic and environmental factors affect developmental stability. These observations raise the possibility that there is a large amount of genetic redundancy in determination of both noise and stability. If so, selection or drift could often act without there being any effect on resultant asymmetry, and fluctuating asymmetry has the capacity to be a neutral character. Alternatively, developmental stability may have a genetic component but low asymmetry appears adaptive only when that genetic component is located close to genes that affect fitness. Therefore, asymmetry–fitness relationships may be due to genetic linkage. Identifying QTLs that influence fluctuating asymmetry could be an approach used to test this hypothesis. Although this seems patently obvious, it is important to stress that if fluctuating asymmetry does have adaptive significance, it is highly likely that this adaptive value was not the original cause of the evolution of the asymmetry. Fluctuating asymmetry did not evolve to reveal fitness, but rather as a by-product of symmetric development. It arose without function, but may now affect fitness in some cases. These scenarios, speculative as they are, indicate that the patterns of documented asymmetry may be consistent with fluctuating asymmetry being a neutral character or one subject to infrequent natural selection. Discovering the genetic origins of fluctuating asymmetry and knowing how stressors affect developmental stability are necessary to test whether fluctuating asymmetry is adaptive. Searching for additional inconsistencies in asymmetry–fitness correlations without learning more about asymmetry production will not be the most productive use of research time and funding.

V. A Revised Look at Fluctuating Asymmetry and Sexual Selection In some cases, but by no means all, symmetric individuals are preferred over asymmetric competitors during mate choice or intrasexual competition (Thornhill and Møller, 1998; Swaddle, 1999b). As most of these studies have reported correlations between mate selection parameters and asymmetry, it is not possible to ascertain the role that asymmetry may

188

JOHN P. SWADDLE

actually play in sexual selection (review in Swaddle, 1999b; see Schlaepfer and McNeil, 2000 for a more recent example). Six studies have investigated a direct effect of fluctuating asymmetry in sexual selection processes by manipulating degrees of phenotypic asymmetry within a natural range. Of these cases, three indicate that individuals with low asymmetry in secondary sexual traits have an advantage over their more asymmetric counterparts (Swaddle and Cuthill, 1994a; Møller and Sorci, 1998; Morris and Casey, 1998). The remaining three studies indicate that fluctuating asymmetry in other sexually selected characters had no detectable influence on mate selection or social dominance (Swaddle and Witter, 1995; Jablonski and Matyjasiak, 1997; Tomkins and Simmons, 1998). Even in the much-cited study of the barn swallow, manipulations of phenotypic asymmetry were unnaturally large (Møller, 1992b) or created novel phenotypic conditions (e.g., high ultraviolet-reflective paint applied to tail feathers; see Swaddle and Cuthill, 1994a) that are difficult to interpret with respect to fluctuating asymmetry (Møller, 1993b). The jury is still out on whether fluctuating asymmetry plays a role in sexual selection. There have been too few tests of the role asymmetry may play. Given the substantial research effort devoted to fluctuating asymmetry, this may be surprising to those peripheral to this area of research. To me, it is indicative of how an idea became accepted too quickly without fundamental tests of the predictions initially proposed by Møller (1990). A ‘‘back-to-basics’’ approach is much needed and overdue. As described in Section IV, our present state of knowledge suggests that fluctuating asymmetry is largely nongenetic in origin. This has important consequences for the role asymmetry may play in sexual selection. How fluctuating asymmetry may be associated with nongenetic benefits in sexual selection, and how symmetry preferences may arise in the absence of any detectable benefits, are considered below. These observations have implications for how sexual selection should be studied in future. This is discussed in Section V.C. A. Nongenetic Benefits of Low Fluctuating Asymmetry It is commonly claimed that, if fluctuating asymmetry plays a role in sexual selection, preferences for symmetric individuals would provide strong support for ‘‘good genes’’ models of sexual selection (e.g., Møller and Pomiankowski, 1993; Gangestad et al., 1994; Scheib et al., 1999; Van Dongen et al., 2001). This is contrary to the evidence that fluctuating asymmetry is largely determined by environmental conditions during development. A preference for symmetry may be more related to direct environmental factors than to indirect inherited benefits through the


189

genes. It is an oversimplification to consider fluctuating asymmetry to be a tool to distinguish between the relative roles of direct and indirect benefits in models of sexual selection (Møller and Pomiankowski, 1993). One of the most cited cases in which asymmetry influences mate selection and reproduction is that of the barn swallow. In this species, female mate choice is affected by gross levels of tail feather asymmetry, resulting in symmetric males gaining higher reproductive success than asymmetric males (Møller, 1992b, 1993b), and females contributing relatively more parental care when mated to a symmetric male (Møller, 1994a). In addition, the symmetry of barn swallow tail feathers is a sensitive indicator of a range of suboptimal environmental conditions, ranging from parasitic infestation to radioactive contamination (Møller, 1992a, 1993a). Møller has used such relationships to support the hypothesis that fluctuating asymmetry of elongated tail feathers reveals ‘‘good genes’’ in male barn swallows. However, more recent studies have indicated that this model system may not be as straightforward as once reported. In a 3-year test of whether parental asymmetry is related to offspring quality, there appeared to be no relationship between the asymmetry of either parent with offspring size, immunocompetence, or condition (Cadeé, 2000a). In addition, fluctuating asymmetry of many traits, including length of the outer tail feathers, does not appear heritable (Cadeé, 2000b). Only in years when environmental variance is low do asymmetry measures approach significant heritability, which indicates that asymmetry values are dominated by environmental effects in most years. Hence, there appears to be little heritable benefit of low asymmetry to swallow offspring and thus little support for good genes models of sexual selection. If fluctuating asymmetry carries little heritable benefit, direct benefits of selecting against asymmetry should be considered. There are accounts of mate selection being mediated through environmental effects on asymmetry. For example, horn length asymmetry may be related to some aspects of phenotypic fitness in female mountain goats. This relationship appears to arise because more advantageous environmental conditions lead to increased female condition and, hence, a better chance of successful reproduction; more advantageous environmental conditions reduce horn asymmetry (Cote and Festa-Bianchet, 2001). Therefore, asymmetry can be a marker of phenotypic fitness because of the common effect of the environment on condition and asymmetry. Interestingly, in a laboratory study of Drosophila melanogaster, the only asymmetry measure (orbital bristles) that was consistently increased by environmental stressors was also the one with the smallest additive genetic component (Woods et al., 1999). It may be this type of trait that is important in determining mating preferences, as asymmetry in such traits could give conspecifics a

190

JOHN P. SWADDLE

window into the nutritional and energetic, developmental environment encountered by potential mates. This may be especially relevant in species whose breeding/nesting habitat shows some level of heritability, or in species that show repeated growth associated with the timing of breeding (e.g., winter/spring feather molt in birds may reveal the condition of individuals as they enter the breeding season). Another relevant study has shown that humans display a preference for symmetry when judging attractiveness of monozygotic twins. This not only emphasizes the significant influence of environmental effects on asymmetry production, but also explicitly shows that preferences based on asymmetry differences can exist when the asymmetries are wholly environmentally determined (Mealey et al., 1999). Understanding the relative roles of the genome and the environment in asymmetry production appears crucial to interpreting whether asymmetry could play a role in sexual selection. This is not a surprising statement; phenotypic traits must have a genetic basis in order for selection to result in change. However, researchers often plough ahead and record fluctuating asymmetry in relation to sexually selected behaviors or morphology without any knowledge of genetic causes. That approach adds more data points to review articles, but it would be more helpful if researchers married genetic, developmental, and behavioral approaches and studied the causes and consequences of asymmetry simultaneously. There are limited cases in which this has occurred. Perhaps the best two examples are barn swallows (see references) and the European earwig Forficula auricularia. In the latter, male forceps size is heritable and plays a role in mating preferences, but forceps asymmetry is not heritable, appears to result from developmental buffering, and does not influence female choice (Tomkins and Simmons, 1998, 1999; Tomkins, 1999). Preliminary investigations of zebra finch chest bar asymmetry indicate that this characteristic is not heritable (J. P. Swaddle, unpublished data), but females prefer males with symmetric bars (Swaddle and Cuthill, 1994a). In addition to making sure they are measuring a repeatable characteristic (Section IV.A) and understanding how the asymmetry is produced (Section IV.C), researchers need to expand their consideration of the benefits of low asymmetry to include immediate environmental factors in addition to heritable consequences. B. Perceptual Processes and Asymmetry There are two general issues I want to raise in terms of how perceptual processes affect studies of asymmetry and sexual selection: perceptual bias and asymmetry detection.


191

1. Perceptual Bias for Symmetry The notion that a preference for low asymmetry can arise as a byproduct of species or object recognition is discussed at length elsewhere (Swaddle and Cuthill, 1994a; Swaddle, 1999a). Below is a brief summary of the most salient points. Fluctuating asymmetry is characterized by a normal (or leptokurtic) distribution of signed asymmetry scores (i.e., left minus right) that is centered around a mean of zero (Palmer and Strobeck, 1986). Individuals in a population should always be exposed to this form of distribution of asymmetries in visual cues (or signals) they assess to judge potential mates and competitors. There are both theoretical and empirical data to show that animals (including humans) form a mental template that is the average of their previous experience against which they compare a new form (Kalish and Guttman, 1957; Blough, 1969; Dill and Heisenberg, 1995; Enquist and Johnstone, 1997). As the average expression of fluctuating asymmetry will always be close to symmetry (i.e., zero asymmetry), animals could possess symmetry preferences because a symmetric form is closer to their mental template than is an asymmetric form (Johnstone, 1994; Swaddle and Cuthill, 1994a). Hence, a symmetry preference can arise as a by-product of other, more general cognitive processes, such as species recognition. This particular process has yet to be tested explicitly, but is worthy of much greater consideration than it is currently given in the sexual selection literature. As fluctuating asymmetry is a byproduct of symmetric development (Section IV.E), this discussion of symmetry preferences suggests that the role of asymmetry in sexual selection results from the interaction of two byproducts: a by-product of development, and a by-product of perception. As the relationship between the selection pressure and the target trait (i.e., low fluctuating asymmetry) is mediated by at least two indirect links, perhaps it should not be surprising to see little evolutionary effect on fluctuating asymmetry. There is likely to be stronger selection on other genes/traits that affect fluctuating asymmetry as a by-product (i.e., selection on trait size, and/or selection on perceptual mechanisms for object recognition). 2. Ability to Detect Asymmetry If fluctuating asymmetry is used as a direct cue in sexual selection, it is important to demonstrate that animals can perceive small differences in the magnitude of phenotypic asymmetry. As most fluctuating asymmetry in nature tends to be small (commonly less than 1% relative asymmetry in most individuals in a population) it is questionable whether animals can reliably detect and respond to such minor variation (Swaddle, 1999c).

192

JOHN P. SWADDLE

Experiments that have demonstrated a direct, visual role for asymmetry in sexual selection have manipulated the asymmetry of traits that are unusually asymmetric in their natural state: 21% relative asymmetry in male swordtail fish (Xiphophorus cortezi) bar patterns (Morris and Casey, 1998); 10% asymmetry in the chest plumage of male zebra finches (Taeniopygia guttata) (Swaddle and Cuthill, 1994a). Even in exaggerated secondary sexual characters, mean fluctuating asymmetry is often much lower than these values (Balmford et al., 1993). When the population frequency distribution of signed asymmetry scores is more platykurtic, for example, in the barn swallow, it is possible that some individuals could have greater asymmetry even though the population mean is rather low. From Møller’s reports of fluctuating asymmetry in the outer tail feathers of male swallows (Møller, 1990, 1992a, 1994b), it seems that approximately half the population possesses tail feather asymmetry above the population mean of 2.3%. Some have asymmetries of more than 6%. In more leptokurtic distributions, however, most individuals will have asymmetry values below the population mean. Experiments have started to explore the capacity of European starlings, Sturnus vulgaris, to discriminate symmetry from asymmetry. Although these experiments have used unnaturalistic cues, they indicate that starlings cannot accurately detect length asymmetries of the size they would most commonly experience in the wild (Swaddle, 1999c). However, if the asymmetries are sufficiently conspicuous, the birds can categorize images as being symmetric or asymmetric (Swaddle and Pruett-Jones, 2001), but there appears a threshold effect in terms of asymmetry detection (Swaddle, 1999c). Therefore, if asymmetry is large enough it could play a direct role in sexual selection. There are some experiments that have manipulated asymmetry within the natural range, and the experimental asymmetry seems large enough to be detected. These studies show that sometimes asymmetry is an important cue (Swaddle and Cuthill, 1994a; Morris and Casey, 1998), whereas in other cases asymmetry does not affect sexual selection processes (Swaddle and Witter, 1995; Tomkins and Simmons, 1998). Experiments that have manipulated asymmetry to unrealistic levels have shown that if an asymmetry is large enough it will be avoided during mate selection: for example, Møller altered the relative asymmetry of male swallow tails from 2.3% to approximately 22% (Møller, 1992b). Once more naturalistic experimentation has been performed, it would be interesting to review the studies to see whether reported effect sizes are positively related to the magnitude of the asymmetry (both manipulated and unmanipulated). It may be that asymmetry can be used as a cue only in species where the asymmetries are large and variable.


193

One report indicates that, among humans, there are facial features correlated with overall facial symmetry that can be detected when only one side of the face is visible (Scheib et al., 1999). This form of correlation could, potentially, lead to an apparent preference for symmetry even when asymmetry differences are not detectable. Therefore, studying correlates of symmetry (even when the symmetry is artificially produced) is an important feature of experimental design as, at present, we do not know how most organisms will perceive the symmetry of any given trait. C. Recommendations for Studying Fluctuating Asymmetry and Sexual Selection In addition to the specific suggestions made in the previous section, there are several important issues relevant to studying the role of fluctuating asymmetry in sexual selection. First, researchers should develop a priori arguments as to why asymmetry of a particular trait may play a role in sexual selection. In general, the asymmetry of the trait will have to be conspicuous (large variance and mean absolute asymmetry) and related to other fitness indicators. Second, it is necessary to consider more fully how animals may perceive their environment (and each other) and how development progresses. Although it is convenient to assess fluctuating asymmetry in terms of two-dimensional lengths of traits, this may not be the way that an animal assesses asymmetry in that trait or accurately represents the variation in developmental trajectories. The potential role of fluctuating asymmetry in sexual selection lies at the intersection of perception and development. We need to consider variation in size, shape, and coloration—and demonstrate how these are related to perceptual abilities and developmental programs (e.g., measuring shape and size variation in integrated units). It is highly unlikely that asymmetry is judged independently of other cues. It is possible to design studies to investigate the relative effect of asymmetry in realistic ways, yet few have been reported thus far (Swaddle, 1999b,c; Swaddle and Pruett-Jones, 2001). Within such experiments it is important to realize that to alter asymmetry the size of the two sides is altered, which could influence mate selection processes independent of the asymmetry. Therefore, it is crucial to balance any manipulation of size (whether that is average of left and right, or independent assessment of left or right) across the experimental design (Swaddle, 1997d). An example of such a design is to present animals with a population of asymmetry values, and to balance presentation of asymmetry cues in terms of size characteristics (cf. Swaddle, 1999c). In this particular example, the intent

194

JOHN P. SWADDLE

was for birds to discriminate between symmetric and asymmetric patterns. By randomizing average trait size among paired (symmetric and asymmetric) presentations, a schedule was produced in which birds could not succeed by simply favoring average size, left size, or right size (Swaddle, 1999c). In terms of size–asymmetry relations, authors commonly report the relationship between size and asymmetry of a trait to interpret the mode of selection acting on trait size (Møller and Pomiankowski, 1993; Møller and Swaddle, 1997). Supposedly, a U-shaped relationship is indicative of stabilizing selection, whereas a negative relation between size and asymmetry indicates directional selection—which could imply active sexual selection for increased size. However, there is no clear prediction as to the relationship between developmental stability, phenotypic variance, and trait size under directional selection regimens (review in Swaddle et al., 2002), and possession of relatively larger traits in cases of sexual dimorphism is not a clear predictor of directional selection (Swaddle et al., 2000; Karubian and Swaddle, 2001). Therefore, interpreting size–asymmetry relations is problematic, and a negative relationship may not reliably indicate directional sexual selection. Future investigations of the role of fluctuating asymmetry in sexual selection could further address the relative direct and indirect effects of asymmetry on mate selection processes through matched observational studies and naturalistic phenotypic manipulations. Does fluctuating asymmetry directly affect performance, or is it indirectly related through its relationship with other parameters important to mate selection?

VI. Fluctuating Asymmetry, Animal Behavior, and Evolution Many researchers have stated that fluctuating asymmetry must be both heritable and related to fitness for it to be evolutionarily relevant (Markow, 1995; Van Dongen and Lens, 2000). Although this is obviously true for adaptive evolution, it need not be true if fluctuating asymmetry is a neutral character (Section IV.E). Fluctuating asymmetry data may be particularly challenging to evolutionary biologists, as experiments indicate that asymmetry can have direct fitness effects (indicating a selection pressure) and yet have little or no additive genetic variance. This scenario would indicate that fluctuating asymmetry would have a minimal effect on long-term evolution but would still appear relevant to behavioral ecologists and functional morphologists investigating the proximate mechanisms of behaviors and biomechanics. For example, analysis of the effects of within-individual variation of flight


195

feather asymmetry between subsequent molts has shown that the noise associated with the production of asymmetry has significant effects on flight performance (Swaddle, 1997a). Presumably, reduced flight performance will have physiological and behavioral costs, and hence will likely directly affect fitness (cf. Witter et al., 1994). This study indicates that small variation in asymmetry can have large direct consequences. Similarly, small variation in limb asymmetry negatively influences antipredatory performance in the house fly Musca domestica (Swaddle, 1997b) and the lizard Psammodromus algirus (Martin and Lopez, 2001), predation ability in the male yellow dungfly Scathophaga stercoraria (Swaddle, 1997b), and fighting ability in the male shore crab Carcinus maenas (Sneddon and Swaddle, 1999). In any of these cases, if asymmetry shows heritability, then there is the potential for natural selection against asymmetry. Early experiments clearly indicated that some birds are influenced by the symmetric appearance of artificial characters (Møller, 1993b; Swaddle and Cuthill, 1994b; Swaddle, 1996), which indicated active mate preference selection mechanisms against asymmetry, yet, these interpretations may have little relevance to the evolutionary design of sexually selected characters if there is no heritability for asymmetry. There are two well-studied systems in which there are heritability estimates, fitness correlations, and analyses investigating whether asymmetry influences behavior: the barn swallow and Drosophila (review in Møller and Swaddle, 1997). The barn swallow is a much lauded example of behavioral processes in sexual selection, and it also quickly became a model species for studying fluctuating asymmetry (Møller, 1994c). Møller has reported that outer tail feather asymmetry is both heritable and negatively related to a variety of fitness parameters (Møller, 1997; Møller and Thornhill, 1997). In addition, Møller has suggested that female swallows judge their mates on the basis of this asymmetry (Møller, 1992b), in association with a number of other characteristics (Møller, 1994c). Studies have indicated that parental asymmetry is not related to offspring asymmetry or offspring quality (Cadeé, 2000a). Also, it is not clear how much asymmetry would actually influence natural mate choice mechanisms given the strength of the relations between other visual and social cues with mate preferences in barn swallows (Møller, 1994c). The heritability of asymmetry in chaetae and sternopleural bristles appears low in Drosophila (Reeve, 1960; Scheiner et al., 1991), although it does appear possible to select for bristle fluctuating asymmetry (Mather, 1953). In addition, heritability of asymmetry between correlated developmental units of bristle morphology is greater than heritability of simple bristle count asymmetry, implying that developmental instability of bristles may have a significant additive genetic component but that previous

196

JOHN P. SWADDLE

measures of fluctuating asymmetry do not represent this (Polak and Starmer, 2001). However, there is little evidence that bristle asymmetry is related to fitness parameters in species of Drosophila (Markow, 1987; Markow and Ricker, 1992; Markow et al., 1996; Polak, 1997; Hoffmann et al., 1998; Hoikkala et al., 1998) Hence, even in these two well-studied systems, it is not clear how much fluctuating asymmetry would actually affect adaptive evolutionary processes. Perhaps, in many systems, fluctuating asymmetry is a neutral trait or a trait with minimal additive genetic variance. This does not mean that fluctuating asymmetry is not relevant to studies of animal behavior. One of the goals of this review is to illustrate how, if behavioral ecologists want to answer meaningful evolutionary questions about fluctuating asymmetry, they need to integrate approaches and tools from other disciplines and learn from their successes. In a complementary fashion, behaviorists have a great deal to contribute in terms of the direct effects of asymmetry on fitness, which behavioral ecologists are accustomed to studying through phenotypic manipulation experiments. At the moment, studies often appear sporadically as reports from researchers who decide to measure asymmetry in their favorite study organism and correlate it with any fitness measure at hand. While reporting such data has its merits in terms of raising awareness of the asymmetry debate, it fuels the fires of people with extreme views (both pro and con) and does little to bring resolution to any broader questions about fluctuating asymmetry and evolution. Researchers (and funding agencies) should consider longer-term studies of the impact and origins of fluctuating asymmetry, and they should follow through on correlations to understand what is really mediating those relationships. The study of fluctuating asymmetry suddenly became fashionable, and then was rapidly pilloried, but it is indeed a genuine and intriguing area of study for evolutionary biologists—including those interested in animal behavior.

VII. Summary The intention of this chapter has been to show why fluctuating asymmetry is relevant to behavioral ecologists, to illustrate where our knowledge about fluctuating asymmetry is lacking in important areas, and to offer suggestions that will help behavioral ecologists fill those gaps. In particular, relationships between asymmetry and fitness are reviewed. These relationships appear to be taxon and trait specific. The specificity of such relationships probably arises from the weak correlation of fluctuating asymmetry with developmental instability. This is one of the areas in which


197

more knowledge is needed, and several suggestions are made concerning how better to assay developmental instability. Common methodological problems of studying fluctuating asymmetry are discussed briefly. In addition, it is important to understand how asymmetries arise, whether fluctuating asymmetry can be described as a generalized trait, whether asymmetries have a significant additive genetic component, and whether fluctuating asymmetry can be viewed as having adaptive significance. A hypothesis is presented that fluctuating asymmetry may have significant genetic redundancy and, in many cases, could be viewed as a neutral character. Fluctuating asymmetry may have nonlinear associations with fitness and only when asymmetry exceeds a threshold value will an apparent relationship with fitness parameters be observed. Also, environmental influences may dominate the development of fluctuating asymmetry—although there are limited indications of specific genes influencing asymmetry production. Many behavioral ecologists have quantified fluctuating asymmetry in secondary sexual characters and investigated whether such asymmetry plays a role in sexual selection processes. Evidence of selection for symmetric individuals is commonly used to support ‘‘good genes’’ models of sexual selection. This position is refuted, as fluctuating asymmetry is largely affected by environmental influences, and a number of ways to expand our studies of sexual selection to incorporate direct, environmental benefits are suggested. In addition, some fundamental areas of sexual selection that require more attention before any conclusions can be drawn about the role of fluctuating asymmetry are highlighted. Notably, it is important to understand whether symmetry preferences exist and, if so, how they can arise independently of asymmetry–fitness associations. A better understanding of whether animals can detect and if so, how they respond to, natural asymmetry cues is also needed. In the final section, a link between behavioral studies of fluctuating asymmetry and broader evolutionary questions is made, and it is suggested that researchers adopt a more integrated approach that brings together a proximate understanding of the impact of asymmetry with more long-term studies of the origins of asymmetry and evolutionary consequences of selection against fluctuating asymmetry.

Acknowledgments I thank Rowan Lockwood, Dan Cristol, Paul Heideman, Charles T. Snowdon, Peter Slater, Tim Roper, and an anonymous reviewer for extensive, constructive comments on earlier drafts that dramatically improved the final version. I was financially supported by NSF Grant IBN-0133449.

198

JOHN P. SWADDLE

References Amzallag, G. N. (2000). Canalization as a non-genetic source of adaptiveness during morphogenesis: Experimental evidence from analysis of reproductive development in Sorghum bicolor. BioSystems 57, 95–107. Andalo, C., Bazin, A., and Shykoff, J. A. (2000). Is there a genetic basis for fluctuating asymmetry and does it predict fitness in the plant Lotus corniculatus grown in different environmental conditions? Int. J. Plant Sci. 161, 213–220. Balmford, A., Jones, I. L., and Thomas, A. L. R. (1993). On avian asymmetry: Evidence of natural selection for symmetrical tails and wings in birds. Proc. R. Soc. Lond. B Biol. Sci. 252, 245–251. Bjorksten, T. A., Pomiankowski, A., and Fowler, K. (2001). Temperature shock during development fails to increase the fluctuating asymmetry of a sexual trait in stalk-eyed flies. Proc. R. Soc. Lond. B Biol. Sci. 268, 1503–1510. Blackenhorn, W. U., Reusch, T., and Muhlhauser, C. (1998). Fluctuating asymmetry, body size and sexual selection in the dung fly Sepsis cynipsea: Testing the good genes— assumptions and predictions. J. Evol. Biol. 11, 735–753. Blough, D. S. (1969). Generalization gradient shape and summation in steady-state tests. J. Exp. Anal. Behav. 12, 91–104. Bosch, J., and Marquez, R. (2000). Typanum fluctuating asymmetry, body size and mate choice in female mid-wife toads (Alytes obstetricans). Behaviour 137, 1211–1222. Bourquet, D. (2000). Fluctuating asymmetry and fitness in Drosophila melanogaster. J. Evol. Biol. 13, 515–521. Bryden, C. A., and Heath, D. D. (2000). The heritability of fluctuating asymmetry for multiple traits in chinook salmon (Oncorhynchus tshawytscha). Can. J. Fish. Aquat. Sci. 57, 2186–2192. Cadeé, N. (2000a). Parent barn swallow fluctuating asymmetry and offspring quality. J. Avian Biol. 31, 495–503. Cadeé, N. (2000b). Genetic and environmental effects on morphology and fluctuating asymmetry in nestling barn swallows. J. Evol. Biol. 13, 359–370. Clarke, G. M. (1997). The genetic and molecular basis of developmental stability: The Lucilia story. Trends Ecol. Evol. 12, 89–90. Clarke, G. M. (1998a). The genetic basis of developmental stability. IV. Individual and population asymmetry parameters. Heredity 80, 553–561. Clarke, G. M. (1998b). Developmental stability and fitness: The evidence is not quite so clear. Am. Nat. 152, 762–766. Clarke, G. M., Brand, G. W., and Whitten, M. J. (1986). Fluctuating asymmetry: A technique for measuring developmental stress caused by inbreeding. Aust. J. Biol. Sci. 39, 145–153. Clarke, G. M., Yen, J. L., and McKenzie, J. A. (2000). Wings and bristles: Character specificity of the asymmetry phenotype in insecticide-resistant strains of Lucilia cuprina. Proc. R. Soc. Lond. B Biol. Sci. 267, 1815–1818. Cote, S. D., and Festa-Bianchet, M. (2001). Life-history correlates of horn asymmetry in mountain goats. J. Mammal. 82, 389–400. Cuervo, J. J., and Møller, A. P. (1999). Phenotypic variation and fluctuating asymmetry in sexually dimorphic feather ornaments in relation to sex and mating system. Biol. J. Linn. Soc. 68, 505–529. Cuthill, I. C., Swaddle, J. P., and Witter, M. S. (1993). Fluctuating asymmetry. Nature 363, 217–218. Dill, M., and Heisenberg, M. (1995). Visual pattern memory without shape recognition. Philos. Trans. R. Soc. Lond. B Biol. Sci. 349, 143–152.


199

Enquist, M., and Johnstone, R. A. (1997). Generalization and the evolution of symmetry preferences. Proc. R. Soc. Lond. B Biol. Sci. 264, 1345–1348. Farner, D. C., and Barnard, C. J. (2000). Fluctuating asymmetry and sperm transfer in male decorated field crickets (Gryllodes sigillatus). Behav. Ecol. 47, 287–292. Fusco, G., and Minelli, A. (2000). Developmental stability in geophilomorph centipedes. In ‘‘Progress in Studies on Myriapoda and Onychophora’’ (J. Wytwer and S. Golovatch, Eds.), pp. 73–82. Fragmenta Faunistica 43 (Suppl.), Polish Academy of Sciences, Warsaw, Poland. Gage, M. J. G. (1998). Influences of sex, size, and symmetry on ejaculate expenditure in a moth. Behav. Ecol. 9, 592–597. Gangestad, S. W., and Thornhill, R. (1999). Individual differences in developmental precision and fluctuating asymmetry: A model and its implications. J. Evol. Biol. 12, 402–416. Gangestad, S. W., Thornhill, R., and Yeo, R. A. (1994). Facial attractiveness, developmental stability, and fluctuating asymmetry. Ethol. Sociobiol. 15, 73–85. Goddard, K. W., and Lawes, M. J. (2000). Ornament size and asymmetry: Is the tail a reliable signal of male quality in the red-collared widowbird? Auk 117, 366–372. Graham, J. H., Emlen, J. M., and Freeman, D. C. (1993). Developmental stability and its applications in ecotoxicology. Ecotoxicology 2, 175–184. Graham, J. H., Emlen, J. M., Freeman, D. C., Leamy, L. J., and Kieser, J. A. (1998). Directional asymmetry and measurement of developmental instability. Biol. J. Linn. Soc. 64, 1–16. Hardersen, S. (2000). Effects of carbaryl exposure on the last larval instar of Xanthocnemis zealandica: Fluctuating asymmetry and adult emergence. Entomol. Exp. Appl. 96, 221–230. Hechter, R. P., Moodie, P. F., and Moodie, G.E.E. (2000). Pectoral fin asymmetry, dimorphism and fecundity in the brook stickleback, Culaea inconstans. Behaviour 137, 999–1009. Hewa-Kapuge, S., and Hoffmann, A. A. (2001). Composite asymmetry as an indicator of quality in the beneficial wasp Trichogramma nr. brassicae (Hymenoptera: Trichogrammatidae). J. Econ. Entomol. 94, 826–830. Hoffman, A. A., Hercus, M., and Dagher, H. (1998). Population dynamics of the Wolbachia infection causing cytoplasmic incompatibility in Drosophila melanogaster. Genetics 148, 221–231. Hoikkala, A., Aspi, J., and Suvanto, L. (1998). Male courtship song frequency as an indicator of male genetic quality in an insect species, Drosophila montana. Proc. R. Soc. Lond. B Biol. Sci. 265, 503–508. Houle, D. (1997). Comment on ‘‘A meta-analysis of the heritability of developmental stability’’ by Møller and Thornhill. J. Evol. Biol. 10, 17–20. Houle, D. (1998). High enthusiasm and low R-squared. Evolution 52, 1872–1876. Houle, D. (2000). A simple model of the relationship between asymmetry and developmental stability. J. Evol. Biol. 13, 720–730. Jablonski, P. G., and Matyjasiak, P. (1997). Chaffinch (Fringilla coelebs) epaulette display depends on the degree of exposure but not symmetry of intruder’s epaulettes. Behaviour 134, 1115–1121. Johnstone, R. A. (1994). Female preference for symmetrical males as a by-product of selection for mate recognition. Nature 372, 172–175. Jones, J. S. (1987). An asymmetrical view of fitness. Nature 325, 298–299. Kalish, H. I., and Guttman, N. (1957). Stimulus generalization after equal training on two stimuli. J. Exp. Psychol. 53, 139–144.

200

JOHN P. SWADDLE

Karubian, J., and Swaddle, J. P. (2001). Selection on females can create ‘‘larger males.’’ Proc. R. Soc. Lond. B Biol. Sci. 268, 1–4. Kilgariff, J. L., Lambert, K. M., Henneberg, M., and Townsend, G. (2000). Fluctuating asymmetry of dermatoglyphic patterns in female monozygotic twins. J. Dent. Res. 79, 1053. Klingenberg, C. P., and Mclntyre, G. S. (1998). Geometric morphometrics of developmental instability: Analyzing patterns of fluctuating asymmetry with Procrustes methods. Evolution 52, 1363–1375. Klingenberg, C. P., and Nijhout, H. F. (1999). Genetics of fluctuating asymmetry: A developmental model of developmental instability. Evolution 53, 358–375. Klingenberg, C. P., Badyaev, A. V., Sowry, S. M., and Beckwith, N. J. (2001). Inferring developmental modularity from morphological integration: Analysis of individual variation and asymmetry in bumblebee wings. Am. Nat. 157, 11–23. Leamy, L. (1997). Is developmental stability heritable? J. Evol. Biol. 10, 21–29. Leamy, L., Pomp, D., Eisen, E. J., and Cheverud, J. M. (2000). Quantitative trait loci for directional but not fluctuating asymmetry of mandible characters in mice. Genet. Res. 76, 27–40. Leamy, L. J., Routman, E. J., and Cheverud, J. M. (1997). A search for quantitative trait loci affecting asymmetry of mandibular characters in mice. Evolution 51, 957–969. Leamy, L. J., Routman, E. J., and Cheverud, J. M. (1998). Quantitative trait loci for fluctuating asymmetry of discrete skeletal characters in mice. Heredity 80, 509–518. Lens, L., and Van Dongen, S. (2000). Fluctuating and directional asymmetry in natural bird populations exposed to different levels of habitat disturbance, as revealed by mixture analysis. Ecol. Lett. 3, 516–522. Leung, B. (1998). Correcting for allometry in studies of fluctuating asymmetry and quality within samples. Proc. R. Soc. Lond. Biol. Sci. 265, 1623–1629. Leung, B., and Forbes, M. R. (1996). Fluctuating asymmetry in relation to stress and fitness: Effects of trait type as revealed by meta-analysis. Ecoscience 3, 400–413. Leung, B., Forbes, M. R., and Houle, D. (2000). Fluctuating asymmetry as a bioindicator of stress: Comparing efficacy of analyses involving multiple traits. Am. Nat. 155, 101–115. Loeschcke, V., Bundgaard, J., and Barker, J. S. F. (1999). Reaction norms across and genetic parameters at different temperatures for thorax and wing size traits in Drosophila aldrichi and D. buzzatii. J. Evol. Biol. 12, 605–623. Ludwig, W. (1932). ‘‘Das Rechts-Links Problem in Tierreich und beim Menschen.’’ SpringerVerlag, Berlin. Mardia, K. V., Bookstein, F. L., and Moreton, I. J. (2000). Statistical assessment of bilateral symmetry of shapes. Biometrika 87, 285–300. Markow, T. A. (1987). Genetics and sensory basis of sexual selection in Drosophila. In ‘‘Evolutionary Genetics and Invertebrate Behavior’’ (M. D. Huettel, Ed.), pp. 89–95. Plenum, New York. Markow, T. A. (1995). Evolutionary ecology of developmental instability. Annu. Rev. Entomol. 40, 105–120. Markow, T. A., and Clarke, G. M. (1997). Meta-analysis of the heritability of developmental stability: A giant step backward. J. Evol. Biol. 10, 31–37. Markow, T. A., and Ricker, J. P. (1992). Male size, developmental stability, and mating success in natural populations of three Drosophila species. Heredity 69, 122–127. Markow, T. A., Bustoz, D., and Pitnick, S. (1996). Sexual selection and a secondary sexual character in two Drosophila species. Anim. B Behav. 52, 759–766. Martin, J., and Lopez, P. (2000). Chemoreception, symmetry and mate choice in lizards. Proc. R. Soc. Lond. B Biol. Sci. 267, 1265–1269.


201

Martin, J., and Lopez, P. (2001). Hindlimb asymmetry reduces escape performance in the lizard Psammodromus algirus. Physiol. Biochem. Zool. 74, 619–624. Mather, K. (1953). Genetical control of stability in development. Heredity 7, 297–336. McAdams, H. H., and Arkin, A. (1999). It’s a noisy business! Genetic regulation at the nanomolar scale. Trends Genet. 15, 65–69. McKenzie, J. A., and Batterham, J. A. (1994). The genetic, molecular and phenotypic consequences of selection for insecticide resistance. Trends Ecol. Evol. 9, 166–169. McKenzie, J. A., and Clarke, G. M. (1988). Diazinon resistance, fluctuating asymmetry and fitness in the Australian sheep blowfly, Lucilia cuprina. Genetics 120, 213–220. Mealey, L., Bridgstock, R., and Townsend, G. C. (1999). Symmetry and perceived facial attractiveness: A monozygotic co-twin comparison. J. Pers. Soc. Psychol. 76, 151–158. Merila¨, J., and Bjo¨rklund, M. (1995). Fluctuating asymmetry and measurement error. Syst. Biol. 44, 97–101. Mogie, M., and Cousins, M. (2001). Are sample sizes usually at least an order of magnitude too low for reliable estimates of leaf asymmetry? J. Theor. Biol. 211, 181–185. Møller, A. P. (1990). Fluctuating asymmetry in male sexual ornaments may reliably reveal male quality. Anim. B Behav. 40, 1185–1187. Møller, A. P. (1992a). Parasites differentially increase fluctuating asymmetry in secondary sexual characters. J. Evol. Biol. 5, 691–699. Møller, A. P. (1992b). Female preference for symmetrical male sexual ornaments. Nature 357, 238–240. Møller, A. P. (1993a). Morphology and sexual selection in the barn swallow Hirundo rustica in Chernobyl, Ukraine. Proc. R. Soc. Lond. B Biol. Sci. 252, 51–57. Møller, A. P. (1993b). Female preference for apparently symmetrical male sexual ornaments in the barn swallow Hirundo rustica. Behav. Ecol. Sociobiol. 32, 371–376. Møller, A. P. (1994a). Symmetrical male sexual ornaments, paternal care, and offspring quality. Behav. Ecol. 5, 188–194. Møller, A. P. (1994b). Sexual selection in the barn swallow (Hirundo rustica). IV. Patterns of fluctuating asymmetry and selection against asymmetry. Evolution. 48, 658–670. Møller, A. P. (1994c). ‘‘Sexual Selection and the Barn Swallow.’’ Oxford University Press, Oxford. Møller, A. P. (1997). Developmental stability and fitness: A review. Am. Nat. 149, 916–932. Møller, A. P., and Eriksson, M. (1994). Patterns of fluctuating asymmetry in flowers: Implications for honest signalling for pollinators. J. Evol. Biol. 7, 97–113. Møller, A. P., and Pomiankowski, A. (1993). Fluctuating asymmetry and sexual selection. Genetica 89, 267–279. Møller, A. P., and Sorci, G. (1998). Insect preference for symmetrical artificial flowers. Oecologia 114, 37–42. Møller, A. P., and Swaddle, J. P. (1997). ‘‘Asymmetry, Developmental Stability and Evolution.’’ Oxford University Press, Oxford. Møller, A. P., and Thornhill, R. (1997). A meta-analysis of the heritability of developmental stability. J. Evol. Biol. 10, 1–16. Morris, M. R., and Casey, K. (1998). Female swordtail fish prefer symmetrical sexual signal. Anim. Behav. 55, 33–39. Nachman, G., and Heller, K. E. (1999). Fluctuating asymmetry as an index of fitness: Causality or statistical artifact? Oikos 86, 357–365. Nosil, P., and Reimchen, T. E. (2001). Tarsal asymmetry, nutritional condition, and survival in water boatmen (Callicorixa vulnerata). Evolution 55, 712–720.

202

JOHN P. SWADDLE

Palmer, A. R. (1994). Fluctuating asymmetry analyses: A primer. In ‘‘Developmental Instability: Its Origins and Evolutionary Implications’’ (T. A. Markow, Ed.), pp. 335–364. Kluwer, Dordrecht. Palmer, A. R. (1996a). From symmetry to asymmetry: Phylogenetic patterns of asymmetry variation in animals and their evolutionary significance. Proc. Natl. Acad. Sci. USA 93, 14279–14286. Palmer, A. R. (1996b). Waltzing with asymmetry. Bioscience 46, 518–532. Palmer, A. R. (1999). Detecting publication bias in meta-analyses: A case study of fluctuating asymmetry and sexual selection. Am. Nat. 154, 220–233. Palmer, A. R. (2000). Quasireplication and the contract of error: Lessons from sex ratios, heritabilities and fluctuating asymmetry. Annu. Rev. Ecol. Syst. 31, 441–480. Palmer, A. R., and Strobeck, C. (1986). Fluctuating asymmetry: Measurement, analysis, patterns. Annu. Rev. Ecol. Syst. 17, 391–421. Palmer, A. R., and Strobeck, C. (1992). Fluctuating asymmetry as a measure of developmental stability: Implications of non-normal distributions and power of statistical tests. Acta Zool. Fenn. 191, 57–72. Parsons, P. A. (1992). Fluctuating asymmetry: A biological monitor of environmental and genomic stress. Heredity 68, 361–364. Pechenkina, E. A., Benfer, R. A., Vershoubskaya, G. G., and Kozlov, A. I. (2000). Genetic and environmental influence on the asymmetry of dermatoglyphic traits. Am. J. Phys. Anthropol. 111, 531–534. Pe´labon, C., and Joly, P. (2000). What, if anything, does visual asymmetry in fallow deer antlers reveal? Anim. Behav. 59, 193–199. Pe´labon, C., and van Breukelen, L. (1998). Asymmetry in antler size in roe deer (Capreolus capreolus): An index of individual and population conditions. Oecologia 116, 1–8. Perfectti, F., and Camacho, J. P. M. (1999). Analysis of genotypic differences in developmental stability in Annona cherimola. Evolution 53, 1396–1405. Pertoldi, C., Podesta, M., Loeschcke, V., Schandorff, S., Marsili, L., Mancusi, C., Nicolosi, P., and Randi, E. (2000). Effect of the 1990 die-off in the northern Italian seas on the developmental stability of the striped dolphin Stenella coeruleoalba (Meyen, 1833). Biol. J. Linn. Soc. 71, 61–70. Polak, M. (1997). Ectoparasitism in mothers causes higher FA in their sons: Implications for sexual selection. Am. Nat. 149, 955–974. Polak, M., and Starmer, W. T. (2001). The quantitative genetics of fluctuating asymmetry. Evolution 55, 498–511. Quek, K. C., Sodhi, N. S., and Kara, A. U. (1999). Absence of positive correlation between fluctuating asymmetry and parasitism in the rock pigeon. J. Avian Biol. 30, 225–229. Reeve, E. C. R. (1960). Some genetic tests on asymmetry of sternopleural chaeta number in Drosophila. Genet. Res. 1, 151–172. Rettig, J. E., Fuller, R. C., Corbett, A. L., and Getty, T. (1997). Fluctuating asymmetry indicates levels of competition in an even-aged popular clone. Oikos 80, 123–127. Roux-Rouquie, M. (2000). Genetic and epigenetic regulation schemes: Need for an alternative paradigm. Mol. Genet. Metab. 71, 1–9. Rowe, L., Repasky, R. R., and Palmer, A. R. (1997). Size-dependent asymmetry: Fluctuating asymmetry versus antisymmetry and its relevance to condition-dependent signaling. Evolution 51, 1401–1408. Rutherford, S. L., and Lindquist, S. (1998). Hsp90 as a capacitor for morphological evolution. Nature 396, 336–342. Salonen, V., and Lammi, A. (2001). Effects of root hemiparasitic infection on host performance: Reduced flower size and increased flower asymmetry. Ecoscience 8, 185–190.


203

Sasal, P., and Pampoulie, C. (2000). Asymmetry, reproductive success and parasitism of Pomatoschistus microps in a French lagoon. J. Fish Biol. 57, 382–390. Scheib, J. E., Gangestad, S. W., and Thronhill, R. (1999). Facial attractiveness, symmetry and cues of good genes. Proc. R. Soc. Lond. B Biol. Sci. 266, 1913–1917. Scheiner, S. M., Caplan, R. L., and Lyman, R. F. (1991). The genetics of phenotypic plasticity. III. Genetic correlations and fluctuating asymmetries. J. Evol. Biol. 4, 51–68. Schlaepfer, M. A., and McNeil, J. N. (2000). Are virgin male lepidopterans more successful in mate acquisition than previously mated individuals? A study of the European corn borer, Ostrinia nubilalis (Lepidoptera: Pyralidae). Can. J. Zool. 78, 2045–2050. Shykoff, J. A., and Møller, A. P. (1999). Fitness and asymmetry under different environmental conditions in the barn swallow. Oikos 86, 152–158. Simmons, L. W., Tomkins, J. L., Kotiaho, J. S., and Hunt, J. (1999a). Fluctuating paradigm. Proc. R. Soc. Lond. B Biol. Sci. 266, 593–595. Simmons, L. W., Beesley, L., Lindhjem, P., Newbound, D., Norris, J., and Wayne, A. (1999b). Nuptial feeding by male bushcrickets: An indicator of male quality? Behav. Ecol. 10, 263–269. Sneddon, L. U., and Swaddle, J. P. (1999). Asymmetry and fighting performance in the shore crab, Carcinus maenas. Anim. Behav. 58, 431–435. Soule´, M. E., and Baker, B. (1968). Phenetics of natural populations. IV. The population asymmetry parameter in the butterfly Coenonympha tullia. Heredity 23, 611–614. Swaddle, J. P. (1996). Reproductive success and symmetry in zebra finches. Anim. Behav. 51, 203–210. Swaddle, J. P. (1997a). Within-individual changes in developmental stability affect flight performance. Behav. Ecol. 8, 601–604. Swaddle, J. P. (1997b). Developmental stability and predation success in an insect predator– prey system. Behav. Ecol. 8, 433–436. Swaddle, J. P. (1997c). On the heritability of developmental stability. J. Evol. Biol. 10, 57–61. Swaddle, J. P. (1997d). Experimental design and the signalling properties of fluctuating asymmetry. Anim. Behav. 54, 1034–1037. Swaddle, J. P. (1999a). Visual signalling by asymmetry: A review of perceptual processes. Philos. Trans. R. Soc. Lond. B Biol. Sci. 354, 1383–1393. Swaddle, J. P. (1999b). Is fluctuating asymmetry a visual signal? In ‘‘Animal Signals’’ (Y. Espmark, T. Amundsen, and G. Rosenqvist, Eds.), pp. 155–175. Tapir Academic Press, Trondheim. Swaddle, J. P. (1999c). Limits to length asymmetry detection in starlings: Implications for biological signalling. Proc. R. Soc. Lond. B Biol. Sci. 266, 1299–1303. Swaddle, J. P., and Cuthill, I. C. (1994a). Female zebra finches prefer males with symmetric chest plumage. Proc. R. Soc. Lond. B Biol. Sci. 258, 267–271. Swaddle, J. P., and Cuthill, I. C. (1994b). Preference for symmetric males by female zebra finches. Nature 367, 165–166. Swaddle, J. P., and Pruett-Jones, S. (2001). Experimental investigation of the acquisition of learned symmetry discrimination in starlings. Am. Nat. 158, 300–307. Swaddle, J. P., and Witter, M. S. (1995). Chest plumage, dominance and fluctuating asymmetry in female starlings. Proc. R. Soc. Lond. B Biol. Sci. 260, 219–223. Swaddle, J. P., and Witter, M. S. (1997). On the ontogeny of developmental stability in a stabilized trait. Proc. R. Soc. Lond. B Biol. Sci. 264, 329–334. Swaddle, J. P., and Witter, M. S. (1998). Cluttered habitats reduce wing asymmetry and increase flight performance in European starlings. Behav. Ecol. Sociobiol. 42, 281–287. Swaddle, J. P., Witter, M. S., and Cuthill, I. C. (1994). The analysis of fluctuating asymmetry. Anim. Behav. 48, 986–989.

204

JOHN P. SWADDLE

Swaddle, J. P., Pruett-Jones, S., and Karubian, J. (2000). A novel evolutionary pattern of reversed sexual dimorphism in fairy-wrens: Implications for sexual selection. Behav. Ecol. 11, 345–349. Swaddle, J. P., Arnold, S. J., and Pruett-Jones, S. (2003). Phenotypic variance and evolution of sexual dimorphism. (Submitted.). Thornhill, R. (1992). Female preference for the pheromone of males with low fluctuating asymmetry in the Japanese scorpionfly (Panorpa japonica: Mecoptera). Behav. Ecol. 3, 277–283. Thornhill, R., and Møller, A. P. (1998). The relative importance of size and asymmetry in sexual selection. Behav. Ecol. 9, 546–551. Tomkins, J. L. (1999). The ontogeny of asymmetry in earwig forceps. Evolution 53, 157–163. Tomkins, J. L., and Simmons, L. W. (1998). Female choice and manipulations of forceps size and symmetry in the earwig Forficula auricularia L. Anim. Behav. 56, 347–356. Tomkins, J. L., and Simmons, L. W. (1999). Heritability of size but not symmetry in a sexually selected trait chosen by female earwigs. Heredity 82, 151–157. Van Dongen, S. (1998a). How repeatable is the estimation of developmental stability by fluctuating asymmetry? Proc. R. Soc. Lond. B Biol. Sci. 265, 1423–1427. Van Dongen, S. (1998b). The distribution of individual fluctuating asymmetry: Why are the coefficients of variation of the unsigned FA so high? Ann. Zool. Fenn. 35, 79–85. Van Dongen, S. (1999). Accuracy and power in the statistical analysis of fluctuating asymmetry: Effects of between individual heterogeneity in developmental instability. Ann. Zool. Fenn. 36, 45–52. Van Dongen, S. (2000). The heritability of fluctuating asymmetry: A Bayesian hierarchical model. Ann. Zool. Fenn. 37, 15–23. Van Dongen, S. (2001). Modelling developmental instability in relation to individual fitness: A fully Bayesian latent variable model approach. J. Evol. Biol. 14, 552–563. Van Dongen, S., and Lens, S. (2000). The evolutionary potential of developmental instability. J. Evol. Biol. 13, 326–335. Van Dongen, S., Molenberghs, G., and Matthysen, E. (1999a). The statistical analysis of fluctuating asymmetry: REML estimation of a mixed regression model. J. Evol. Biol. 12, 94–102. Van Dongen, S., Sprengers, E., Lo¨fstedt, C., and Matthysen, E. (1999b). Fitness components of male and female winter moths (Operophtera brumata L.) (Lepidoptera, Geometridae) relative to measures of body size and asymmetry. Behav. Ecol. 10, 659–665. Van Dongen, S., Sprengers, E., and Lo¨fstedt, C. (2001). Lack of evolutionary potential of developmental instability of front tibia length in the Indian meal moth (Plodia interpunctella). Belgian J. Zool. 131, 75–82. Van Valen, L. (1962). A study of fluctuating asymmetry. Evolution 16, 125–162. Vøllestad, L. A., Hindar, K., and Møller, A. P. (1999). A meta-analysis of fluctuating asymmetry in relation to heterozygosity. Heredity 83, 206–218. Waddington, C. H. (1957). ‘‘The Strategy of the Genes.’’ Allen and Unwin, London. Whitlock, M. (1996). The heritability of fluctuating asymmetry and the genetic control of developmental stability. Proc. R. Soc. Lond. B Biol. Sci. 263, 849–853. Whitlock, M. (1998). The repeatability of fluctuating asymmetry: A revision and extension. Proc. R. Soc. Lond. B Biol. Sci. 265, 1429–1431. Whitlock, M. C., and Fowler, K. (1997). The instability of stability studies. J. Evol. Biol. 10, 63–67. Windig, J. J., and Nylin, S. (2000). How to compare fluctuating asymmetry of different traits. J. Evol. Biol. 13, 29–37.


205

Witter, M. S., Cuthill, I. C., and Bonser, R. H. C. (1994). Experimental investigations of massdependent predation risk in the European starling, Sturnus vulgaris. Anim. Behav. 48, 201–222. Woods, R. E., Sgro`, C. M., Hercus, M. J., and Hoffmann, A. A. (1999). The association between fluctuating asymmetry, trait variability, trait heritability, and stress: A multiply replicated experiment on combined stress in Drosophila melanogaster. Evolution 53, 493–505. Zakharov, V. M. (1992). Population phenogenetics: Analysis of developmental stability in natural populations. Acta Zool. Fenn. 191, 7–30. Zakharov, V. M. (2001). Ontogeny and population: Developmental stability and population variation. Russian J. Ecol. 32, 146–150.



From Dwarf Hamster to Daddy: The Intersection of Ecology, Evolution, and Physiology That Produces Paternal Behavior Katherine E. Wynne-Edwards department of biology queen’s university kingston, ontario, k7l 3n6, canada

I. Introduction In early winter of 1981, I was first introduced to the charismatic Djungarian hamster, Phodopus campbelli. The species had just arrived in North America, with an unsubstantiated rumor that they formed pair bonds (Pogosianz and Sokova, 1967; Jordan, 1971), at a time when the ecological conditions necessary for the evolution of monogamy were a clear focus within the young field of behavioral ecology (Kleiman, 1981). Thus began my association with dwarf hamsters in the genus Phodopus that has lasted two decades and continues to delight me on a daily basis. From the steppes of Siberia to the physiology laboratory, dwarf hamsters inform us about endocrine evolution and sociality and point us toward a new understanding of the biological basis for human behavior. In this chapter, I review our current behavioral, ecological, endocrinological, and evolutionary understanding of these dwarf hamsters as a model for the conditions that produce parental care in the hamsters, and in our own species. To cover this range of perspectives, I shall first establish that male Phodopus campbelli are naturally paternal and then explore the natural habitat, natural history, and social ecology of Phodopus. Next, laboratorybased experiments illustrate the extent of maternal dependence on paternal presence and the ecophysiological basis for those constraints. Evolution of biparental care in P. campbelli is seen as the necessary consequence of conflict between adaptations for survival in a cold, dry seasonal habitat and reproductive adaptations appropriate for handling 207 Copyright 2003 Elsevier Science (USA). All rights reserved. 0065-3454/03 $35.00

208

KATHERINE E. WYNNE-EDWARDS

heat load, water stress, and rapid breeding in the same habitat. The following sections then explore the evolution of endocrine physiology and behavioral endocrinology in female, and then male, P. campbelli as a suite of adaptations to obligate biparental care. From there, hormonal correlates of human fatherhood are reviewed as a natural extension of paternal behavior studies and a model for the biological substrates of parental behavior. The chapter concludes with a brief, and speculative, look at the applied insights possible from an evolutionary endocrinology perspective.

II. Natural History Male Djungarian hamsters, Phodopus campbelli, are exceptionally paternal. In the laboratory, males becoming fathers for the first time participate in the birth process in a role functionally equivalent to that of a midwife (Jones and Wynne-Edwards, 2000, 2001). During the pushing phase of the female’s labor, they lick her anogenital region, often resulting in ingestion of amniotic fluid as membranes rupture. They also use their incisors and forepaws to mechanically pull the pup from the birth canal. Perhaps of even greater relevance for pup survival, they handle the neonate more than the female does in the moments after the birth and before the placenta is delivered. During those critical postpartum seconds, they manipulate the pup to access the face and specifically clear membranes around the nostrils and head. Pups that are born purple flush to a bright pink as the behavior opens an airway. Males then continue to clear remaining membranes, to share consumption of the placenta with the female, and to provide primary care for already-born pups (Jones and Wynne-Edwards, 2000). Midwife behavior is not shown by every father, or during the birth of every pup. Nevertheless, the absence of similar behavior in males of other species, including the closely related Siberian hamster, Phodopus sungorus (Jones and Wynne-Edwards, 2000), supports the hypothesis that this behavior is adaptive. In the wild, direct parental care by male P. campbelli has also been documented (Wynne-Edwards, 1995). For example, one female pregnant with a second litter left the burrow containing her first litter for another about 100 m distant. That night, the weanlings first showed their faces at the burrow entrance, but did not venture onto the soil surface. The same night, the male whose home range included that burrow made several foraging expeditions to provision the weanlings. Each time, he left the burrow with empty cheek pouches and returned with them full, then used his forepaws to push his cheek pouch contents into the mouth of the burrow where the weanlings remained. During the next day of sleeping, he was in the same

FROM DWARF HAMSTER TO DADDY

209

Fig. 1. Photograph of an adult male P. campbelli retrieving a 3-day-old pup that was experimentally displaced from the nest area. First-time fathers are as quick (typically less than 35 s) to retrieve a pup experimentally displaced from the nest as the mother. This pup weighs about 2.5 g, which represents approximately 10% of adult body weight. The tail and hind legs of a second pup (that was also displaced from the nest) are visible beneath the male as he walks over the second pup to retrieve the first. Developmentally, this pup will not have its eyes open for another week, will eat its first hard food after that, and will not thermoregulate independently for another 12 days. Nevertheless, if not retrieved by an adult, about 15% of pups of this age will return to the nest unaided within a 10-min trial. (Photo: K. E. WynneEdwards, 2001.)

burrow as the adult female, who delivered a new litter of pups. The following night, he returned to the original burrow, and he continued to provision the weanlings until they dispersed a few days later (WynneEdwards, 1995). As this was the only time when a female P. campbelli weaned a young litter and delivered a second while under observation, we have no means of assessing how widespread this behavior is. First-time P. campbelli fathers are also highly responsive to a displaced pup (Fig. 1). If the mother is removed from the cage and a 3-day-old pup is displaced to a far corner, the father will typically contact that pup within 10 s and have it back in the nest within 1 min (Reburn and WynneEdwards, 1999; Jones and Wynne-Edwards, 2001). In contrast, new P. sungorus fathers typically contact a displaced pup but fail to pick it up and fail to retrieve it to the nest area. Instead, the majority of P. sungorus pups that return to the nest within a 10-min test do so with their own powers of locomotion (Reburn and Wynne-Edwards, 1999).

210


Thus, extensive observations of paternal behavior under laboratory conditions (Wynne-Edwards, 1995) are fully supported by field observation, confirming the expression of an exceptional behavioral repertoire. Male P. campbelli express parental behavior naturally. A satisfying explanation for this behavior will require an adaptive, evolutionary (i.e., ‘‘ultimate’’) explanation, as well as a mechanistic, physiological, and sensory (i.e., ‘‘proximate’’) explanation. A. Phylogenetics, Range, and Habitat The genus Phodopus (hairy-footed hamsters) is one of five extant genera known as the dwarf hamsters that are all native to Central Asia around a latitude of 53 north (Ross, 1995, 1998). Three species are recognized: Phodopus roborovskii, which inhabits shifting sand dunes of the Gobi and adjacent desert areas; Phodopus sungorus, which is the subject of considerable research in photoperiod and circadian biology; and Phodopus campbelli, which is the primary focus of this chapter. Common names for the Phodopus hamsters are not consistent from Europe to North America and are therefore a poor substitute for the Latin designations. In my laboratory, we have consistently referred to P. campbelli as the Djungarian hamster and to P. sungorus as the Siberian hamster. Phodopus have short limbs and digits, a dense coat of hair covering the pads of the hind feet, and a short tail that protrudes only a few millimeters beyond the pelage. Both species weigh from 18 to 45 g in adulthood, are born after a gestation of 18 days, have an average litter size of just less than six, and mate during a postpartum estrus. On the basis of DNA sequence divergence in two nuclear and two mitochondrial genes, we currently estimate that P. sungorus and P. campbelli shared a common ancestor no more than 1 million years ago (Lougheed et al., 2003). Under laboratory conditions, viable hybrid offspring are possible, but suffer from meiotic abnormalities (Sokolov et al., 1998). The natural ranges of the two species do not overlap and are separated by the Altai–Sayan Mountains at their closest approach (Flint, 1966). Phodopus campbelli is found in the arid semidesert in areas of open sand with occasional grasses and Caragana shrubs. In contrast, P. sungorus is found in ungrazed short grass steppe (Artemesia and Potentilla) closely associated with agricultural fields. Both species of Phodopus occur at extremely low population densities of one to six individuals per square kilometer (Flint, 1966), with extensive stretches of unoccupied habitat between local populations. These densities must be determined by ecological variables because each species shares its habitat with another small-bodied rodent that is found at higher density. In


211

the same habitat as P. sungorus, a vole, Lagurus lagurus, is the same body size and occurs at 10 times higher densities. In the same habitat as P. campbelli, another dwarf hamster, Cricetulus barabensis, is likewise much more common (Flint, 1966; Iudin et al., 1979). In the short grass steppe where P. sungorus is found, rainfall amounts to just over 300 mm/year and 70% of that moisture falls in the 5 months including summer. Winter is cold, with mean January temperature just below 20 C and mean summer high near 20 C. Although that habitat is cold, seasonal, and dry, it is much less cold, less seasonal, and less dry than the habitat of P. campbelli (Wynne-Edwards, 1998). For P. campbelli in Tuva, annual rainfall is only 200 mm/year and is reliable only during two summer months. The mean January temperature is 30 C and the mean minimum for the month is below 50 C. In midsummer, the average daily temperature is 15 C yet daily fluctuations from 4 C at night to 35 C at midday are typical. B. Adaptations for Survival

The center of Asia, at the latitude of 53 N, is a harsh environment. To survive, the hamsters have clear adaptations to that habitat. To withstand cold, they are more spherical than a typical rodent of that weight. They have short limbs, a short, hair-covered tail, and external ears than can fold into the pelage to minimize peripheral heat loss. Their metabolism (Klinggenspor et al., 1989) is capable of maintaining core body temperature even when environmental temperatures are 40 C (Heldmaier et al., 1982), they can enter spontaneous daily torpor (Ruf et al., 1991), and their pelage provides excellent insulation (Meschersky, 1993). To survive with little rainfall, the hamsters produce highly concentrated urine (Trojan, 1979; Scribner, 1996) and choose a diet that provides metabolic water (Trojan, 1979). Season is tracked by sensitive photoperiod responsiveness (Heldmaier and Steinlechner, 1981; Yellon and Goldman, 1984; Hoffmann, 1982). During the winter months, P. sungorus molts to a white pelage and remains active (Flint, 1966). Little is known about P. campbelli behavior during the winter months. Neither species is long-lived. In the wild, no reproductive adult has been seen alive in a second year. In the laboratory, female P. campbelli held at a late spring photoperiod of 14 h of light to 10 h of dark (14L:10D) experience declining fertility and fecundity by 5 months of age, with reproduction effectively complete by 8 months of age (Edwards et al., 1998). In the wild, the reproductive trajectory of a pup is determined by the season of its birth. Pups born early in the breeding season mature rapidly

212


and breed in their natal year, whereas pups born later in the season winter as juveniles and breed the following year. Specifically, the melatonin secretion pattern of the mother in late gestation is transferred to the pups and determines their life history strategy (Horton and Stetson, 1990; Stetson et al., 1989). Thus, both growth and reproduction are influenced by photoperiod to respond appropriately to the dramatic seasonal changes in temperature, moisture, and resource availability. C. Predicting Sociality from Space Use Temperature and rainfall differences in the two habitats also affect food availability and population density. Specifically, P. campbelli arouse before or soon after sunset and remain active, traveling more than 1 km each night before returning to the burrow around dawn. In contrast, P. sungorus avoid both dusk and dawn and travel only about 100 m during a few hours of above-ground activity each night (Wynne-Edwards et al., 1999). This extremely harsh habitat with low population densities is the one in which paternal behavior is seen. The distribution of individuals across the landscape does not, however, independently confirm that P. campbelli males will have extensive paternal behavior, whereas P. sungorus males will not. Home ranges, or (the minimum polygon that contains all sightings of that individual), of female P. campbelli do not overlap, whereas male home ranges overlap to a limited extent (Wynne-Edwards et al., 1999) (Fig. 2). However, the home range of an individual P. campbelli male is much larger than the home range of a female and typically includes all or a large portion of the home range of more than one female. This pattern does not immediately suggest biparental territorial defense or mating exclusivity in P. campbelli. Home ranges for P. sungorus are many times smaller than P. campbelli home ranges (Wynne-Edwards et al., 1999) (Fig. 2). Neither the home ranges of P. sungorus females nor the home ranges of P. sungorus males overlap, and both are of similar size. However, each P. sungorus male tends to overlap more than one female and each P. sungorus female tends to overlap more than one male (Wynne-Edwards, 1995). Within a single night P. sungorus males routinely move from the burrow of one female to the burrow of another. The female remains in her own burrow, and receives a visit from a second male for the latter part of the night (Wynne-Edwards, 1995). Therefore, radio-tracking data alone would suggest that male polygyny without paternal care was the most probable social organization in P. campbelli and that promiscuity without paternal care should be predicted for P. sungorus. However, direct behavioral observations of individuals reveal extensive parental care by male P. campbelli.


213

Fig. 2. The spatial distribution of P. campbelli individuals does not immediately suggest biparental care. Shown are minimum polygons enclosing the area visited by individual P. campbelli (main panels) and P. sungorus (inset panels) during 1988, 1989, and 1990. Each polygon in black represents a different adult female. Female polygons do not overlap. Each other polygon is open or shaded as necessary to show the ranges of individual males that overlap substantially. For scale, the blocks comprising the grid are each 250 m 250 m and the scale is the same for main panels and insets. Phodopus campbelli ranges are much larger than P. sungorus ranges and P. campbelli male ranges are much larger than P. campbelli female ranges. With individual male access to more than one female and no evidence of exclusive overlap between one male and one female, this distribution of individual ranges does not independently suggest that biparental care is typical of P. campbelli. Instead, that evidence comes from more direct observation of male behavior.

214


Fig. 3. Contrast between (a) Phodopus (actually P. sungorus) and (b) a laboratory mouse. Note in Phodopus the shorter limbs, which limit stride length; the furry soles of the feet, which deposit scent-marking perfume; and the shorter length of the hind digits, which are adapted for plantigrade rather than digitigrade locomotion. Also shown are comparative body shape and skeleton shape as revealed by two dorsal X-rays of animals of the same body weight. Phodopus are more spherical, and lack the large heat exchange surfaces of the mouse including the long naked tail, the naked soles of the feet, and the large ears. The scale shown is the same for both (a) and (b).

D. Social Behavior in the Wild Although small mammals, particularly rodents, are the cornerstone of physiological and medical research, they have proved extremely difficult to study in a natural context. Most species are nocturnal, cryptic, and faster than a human pursuer, and spend large amounts of time inside burrows or hidden in areas with dense foliage or detritus for cover. Thus, a diligent observer with binoculars has little chance at more than a passing glimpse of individual animals. As a result, investigators have developed techniques ranging from radiotelemetry to trapping and dousing of captured animals with fluorescent powders in an attempt to infer the movements of individuals (Mikesic and Drickamer, 1992; Jike et al., 1988). Although their native habitat is not easy to access from the West, the dwarf hamsters are an extraordinary exception to this general rule.


215

Dwarf hamsters have a flat-footed gait in the hind legs, unlike many rodents that spring from their toes (Fig. 3). Combined with their short limbs, this makes them slower than a strolling human observer. The habitat they occupy is also free of dense cover. These characteristics allowed our team, including Soviet scientists, A. Surov and A. Telitzina, and a diverse group of Earthwatch volunteers, to follow individual hamsters, 24 h/day, from a distance of less than 2 m. We knew when and where they deposited each scent mark, urinated, ate, contacted other individuals, mated, gave birth, and other intimate details of their daily lives. In fact, we typically captured individuals by simply bending over and picking them up. With small local populations, we were also able to collect complete information for every individual in the local population. For example, five female P. campbelli mated while above ground during our observations. Phodopus is the only small-bodied, nocturnal rodent for which direct observations of mating behaviors of freely moving wild animals have been recorded. In four of the five cases, the male with an overlapping home range spent the day before the mating in the burrow with the female and presumably mated before she came above ground. When she came above ground, she did not move more than a few meters from her burrow and continued to allow the male from her burrow to mount her. Other males from adjacent home ranges, including males from more than 1 km away, were attracted to the area, fought with each other, and chased the female, but did not succeed in mating with the receptive female. In the final case, the female was a newly arrived young female that attracted three potential mates and mated to ejaculation with two of them, but did not implant a pregnancy. Four days later (one estrous cycle) she mated again with just one of the males and became pregnant. Thus, although limited in number, the direct observations of mating behavior in P. campbelli were consistent with a single father for each litter and supported laboratory data suggesting that females that mate with more than one male fail to implant and instead mate again 4 days later (WynneEdwards and Lisk, 1984). Alterations to the temporal pattern of vaginocervical stimulation when two males mate might contribute to this block to implantation when more than one male mates (Wynne-Edwards and Lisk, 1988). The last female was also interesting because she had visited the sleeping burrow of each of the three attending males the night before and had left vaginal scent marks on the runway outside each burrow. This peak in vaginal scent marking 24 h before mating is also seen in laboratory enclosures (Wynne-Edwards and Lisk, 1987b) and has been described for Syrian hamsters (Huck et al., 1985). However, the deposition of scent marks and subsequent reaction to them are not easily studied in wild, nocturnal mammals.

216


E. Chemical Communication At least six discrete sources of chemical secretions are involved in dwarf hamster scent-marking behavior in the wild (Wynne-Edwards et al., 1992). Adults have a ventral sebaceous gland (Reasner and Johnston, 1987), a musk gland in the corner of the mouth that marks items carried in the cheek pouches, Harderian glands, and skin glands behind the ears that secrete substances spread through the fur by grooming. Urine and feces are important sources of chemical signals, and in the females, vaginal secretions change in composition and vaginal scent marks change in frequency throughout the reproductive cycle (Wynne-Edwards and Lisk, 1987b). Laboratory studies have considered the context of these scent marks and demonstrated pheromonal acceleration and delay of puberty as well as individual recognition in P. campbelli (Reasner and Johnston, 1988; Lai and Johnston, 1994; Lai et al., 1996). Scent marks are an important component of social communication in Phodopus (Wynne-Edwards et al., 1992). In addition to direct scent marking, grooming combines the secretions of all scent glands and transfers that complex signal to the ground via the densely haired feet and through rolling. In the wild, individual Phodopus trapline from one location to the next along specific paths, with regular stops to groom. Later, different hamsters use the same ‘‘highways,’’ passing within a few centimeters of the flags that originally denoted the route. The home ranges of individual male P. campbelli, as denoted by the minimum polygon enclosing all of their known scent marks, are nonoverlapping (Wynne-Edwards et al., 1992) (Fig. 4). Instead, when a male traveled outside the home range delineated by his scent marks, he did not stop to forage and his locomotion was directional rather than meandering. Thus, the location of chemical signals suggests that females use vaginal scent marks to alert neighboring males to an impending estrus whereas males use their urine and ventral scent glands to defend their home ranges. Additional, compelling evidence for the importance of chemical communication in Phodopus was collected opportunistically. Specifically, patterns of movement suggested that chemical signals provided effective directional information over long distances, but not over short distances. When a male was approaching a receptive female, for example, the headlamps of the following observers traveled as directly as local topography would allow. However, when the individual hamsters were within a few meters of each other, both fell into back-and-forth searching patterns with repeated rearing on the hind legs and sniffing. From this we


217

Fig. 4. The overlapping home ranges of three adult P. campbelli males (as defined by the minimum polygon enclosing all of the known locations visited; polygons are equivalent to those shown Fig. 2) and the nonoverlapping revealed when the home ranges of the same three males are defined by the location of scent marks (symbols). Only a single symbol is shown for any scent mark location, and scent mark locations are distributed for ease of viewing. There was considerable variation in the total number of scent marks at each flag location over time (range, 1–29). Also shown are the locations of burrows occupied for at least one sleeping day by a female or by a male. Note that scent marks by individual males that are not close to the focus of their scent marks are usually associated with a known female sleeping burrow. Scent marks were from either ventral gland rubbing or urine dragging, both of which are behaviorally distinct and confirmed by laboratory studies. [Redrawn from Wynne-Edwards et al. (1992) with permission.]

inferred that some combination of local air flow around vegetation and overwhelming of sensory neurons interfered with short-range directional communication. A failure of short-distance chemical communication is the probable explanation for another behavioral observation. One P. campbelli male was held for 12 h downwind from an array of cages holding other individuals of both sexes and two species that included one conspecific female that would be receptive that evening. The intent was to

218


demonstrate that the P. campbelli male would locate the unfamiliar, but sexually receptive, conspecific female on his release. Instead, he raced past her cage (coming within 10 cm) on his way to a free-ranging estrous female, more than 250 m distant, which was not part of the experimental design. Another incident demonstrated that long-range chemical information about the presence and precise location of a stranger was available to, and acted on by, resident individuals of both sexes. During documentary filming (‘‘The Trials of Life,’’ D. Attenborough, BBC Natural History Unit, Bristol, UK), a P. campbelli male was captured more than 5 km away so that he could be handled on camera without disturbing the local population under study. The male arrived and left in a coat pocket and was placed repeatedly into a scooped-out depression in the soil. Within minutes of turning off the generator, the lights, and driving a truck away, the male and female that normally ‘‘controlled’’ that area each independently made multiple excursions to the disturbed soil, scent marking extensively each time. Behavioral responses to chemical information were also seen with P. sungorus. For example, in 1989 two females mutually respected a shared boundary for more than 1 week of continuous observations. Each night they traveled along the boundary and placed their scent marks adjacent to but not on top of each other’s mark. Neither female encroached on the area that was scentmarked by the other. Then a battery failed in the radiotransmitter of one female, and she was captured and taken off-site for a battery change. Within 20 min, the remaining female had crossed the shared boundary and was within 1 m of the current sleeping burrow of the female that had been removed. This behavioral response to the sudden absence of a familiar individual is consistent with laboratory experiments. Pregnancy block in Phodopus occurs readily in response to the absence (removal) of the male mate but not in response to the addition of chemical signals from a male stranger (Wynne-Edwards et al., 1987a; Johnson, 1992). These anecdotal observations, although far from rigorously controlled experimental results, support a picture of Phodopus natural history that involves extensive local social knowledge through chemical signals. Specifically, individuals appear to act on information about the absence, as well as the presence, of other conspecific individuals and appear to acquire that information over distances of hundreds of meters. Male P. campbelli also clearly seek opportunities to mate with more than one female, but there is no evidence from the field data to suggest that a female will routinely mate with more than one male or that she will carry a pregnancy if she does.


219

III. Evolution of Biparental Care Mammalian paternal behavior has long been of special interest to behavioral and evolutionary biologists because such a large proportion of mammalian parental care must come from the mother. Fertilization and gestation are internal, leaving the female parent in literal control of the fetus until birth. After birth, the sole source of water and nutrition for the young remains milk from the female parent. This unavoidable maternal investment involves complex physiological changes in the female parent and requires the expression of appropriate maternal behavior before offspring survival is possible. In contrast to this obligatory and extensive maternal investment and behavior, the contributions of the male parent are smaller, and more variable. In the majority of mammalian species, the male leaves more offspring in the next generation if he pursues many females than if he chooses a single mate and helps her to rear their offspring. As a result of these sex-specific roles, sole parental care from the male is unknown in mammals and shared parental effort involving extensive direct paternal care is uncommon. It is largely restricted to well-characterized examples in the carnivores (e.g., wolves) and nonhuman primates (e.g., callitrichids), with rare examples among the rodents that include the prairie vole (Microtus ochrogaster; Carter et al., 1995), the California mouse (Peromyscus californicus; Gubernick and Alberts, 1987a), and the Mongolian gerbil (Meriones unguiculatus; Agren, 1984) as well as P. campbelli. A. Origins of the Laboratory Population The original P. campbelli that founded my laboratory population were a gift to R. D. Lisk of Princeton University from M. R. Murphy in 1981 and were imported from the United Kingdom. Since then, our population of P. campbelli has been outbred against laboratory stock from H. Pogosianz (Moscow, 1984) and wild-caught individuals from the field research site (Erzin, Tuva Autonomous Region, Siberia, 50.16 N, 95.14 E; 1988; 1989, 1990). Likewise, the P. sungorus population is derived from wild individuals caught near Karasuk, Novosibirsk Province (53.45 N, 78.01 E; 1984; 1988) and Beya, Hakasskaya Region (52.58 N, 90.45 E; 1989, 1990). During the past two decades, inbreeding has been strictly limited by preventing all pairings closer than with first cousins (although it is still probable that significant inbreeding has occurred). In spite of dozens of captive generations under identical husbandry, patterns of spontaneous activity in the two species retain the essential characteristics of each species in the wild (Wynne-Edwards et al., 1999).

220


B. Obligate Biparental Care in Phodopus campbelli Social monogamy with biparental care can be classified as obligate or facultative (Clutton-Brock, 1991). In facultative social monogamy a pair forms under some circumstances but not all. For example, a pair might be the typical social unit if there are few other opportunities to find mates because the species is sparsely distributed. In obligate social monogamy, the costs of abandoning the mate are so high that biparental care is essential for either parent to have reproductive success (Gubernick and Teferi, 2000). On the basis of laboratory experiments performed on a population derived from wild-caught individuals, paternal care in P. campbelli is obligate, rather than facultative (Wynne-Edwards, 1987). Laboratory experiments support the hypothesis that biparental care in P. campbelli is obligate, whereas in P. sungorus it is facultative and seasonal (Wynne-Edwards, 1987, 1995). In the laboratory, solitary P. campbelli females successfully raise less than half of their young, whereas mated pairs raise every litter and 95% of pups (Wynne-Edwards, 1987). In contrast, a solitary P. sungorus female will successfully raise more of her young than a paired female will (Wynne-Edwards and Lisk, 1989). When environmental conditions are altered to increase the physiological stress of the mother, these differences intensify. Under a mild heat stress of 23 C, P. sungorus females are not affected and raise litters successfully, whereas P. campbelli females rapidly lose condition and dependent pups (Wynne-Edwards and Lisk, 1989; Walton and Wynne-Edwards, 1998). Solitary P. campbelli females lose more than half of their litter before weaning although, when temperature is held at 18 C (deep burrow temperature in summer), all pups survive (Walton and Wynne-Edwards, 1998) (Fig. 5). Solitary females of both species lose pups when water access is restricted, and this effect is greater in P. campbelli than in P. sungorus (Scribner and Wynne-Edwards, 1994b). These maternal stresses are reflected in patterns of pup growth and the interval before she delivers a second litter (Newkirk et al., 1997, 1998). Preimplantation pregnancy block in P. sungorus has not been reported in response to manipulations of her social environment. However, in P. campbelli, patterns of pregnancy block support an important role for the male parent. If the male is removed between 24 and 48 h after mating, or if the female is introduced to the male less than 2 days before mating (Wynne-Edwards and Lisk, 1987a), the probability of pregnancy block more than doubles. Preimplantation pregnancy block also occurs in P. campbelli females when they mate with more than one male and this block can happen repeatedly, with mating every 4 days, until one of the males is injured or removed (Wynne-Edwards and Lisk, 1984). The first mating

221


Fig. 5. Detrimental effect of 23 C ambient room temperature relative to 18 C on pup growth and development in P. campbelli. Pup ages and ambient temperatures during rearing are shown. Note that the pup reared at 23 C (normal laboratory rodent holding temperature) is delayed by at least 3 days in its growth relative to the pups reared at the optimal laboratory temperature (18 C), corresponding to the burrow temperature in the wild during the breeding season. [Reproduced with permission from Walton and Wynne-Edwards (1998).]

after the second male is removed then results in a pregnancy (WynneEdwards and Lisk, 1984). Other aspects of social behavior also support a higher degree of social isolation in P. campbelli than in P. sungorus. Female–female patterns of aggression suggest that unrelated P. sungorus females will establish a stable dominance hierarchy, whereas unrelated P. campbelli females will show levels of aggression that lead to injury (Wynne-Edwards and Lisk, 1987a). The male–female dominance dynamic also differs. In P. sungorus, males show ‘‘on-back’’ behavior to females as often as females show that submissive behavior to males (Wynne-Edwards and Lisk, 1987b). In contrast, aggression during initial encounters between P. campbelli males and females is more limited and includes higher levels of scent marking by both sexes and more time spent in mutual sniffing (Wynne-Edwards and Lisk, 1987b). Proceptive scent-marking behavior in both species peaks 24 h before mating. During mating, females place vaginal scent marks immediately in front of the male and then drag a trail forward with a stiff-legged gait that has the dorsal flexion characteristic of lordosis (Wynne-Edwards and Lisk, 1987b). When more than one male is present on the day a female is receptive, a dominant P. sungorus male succeeds in excluding the subordinate from mating (Wynne-Edwards and Lisk, 1988). In contrast, more than half of P. campbelli subordinate males mate with the female, and those vaginal intromissions are typically alternated between the two males within each ejaculatory series (Wynne-Edwards and Lisk, 1988). Of course, P. campbelli females do not become pregnant after these multiple male matings, whereas

222


P. sungorus females do (Wynne-Edwards and Lisk, 1984, 1988). Thus, differences in social behavior are also characteristic of these two species and are consistent with female reproductive decisions that ensure paternity certainty and mating exclusivity in P. campbelli but not in P. sungorus. Both species of Phodopus also respond to changes in their social housing with hormonal changes consistent with stress. Pair-bond disruption is an acute stress causing neuropharmacological symptoms of depression in both species of Phodopus (Crawley, 1984; Castro and Matt, 1997b). Likewise, social housing decreases cortisol, whereas social separation both increases cortisol and decreases prolactin in both species of Phodopus (Castro and Matt, 1997a; Reburn and Wynne-Edwards, 2000). This stress response to social isolation, and/or to social crowding, is expected in a small mammal that has not been aggressively domesticated for laboratory rearing and that normally uses longer distance communication modalities for social contact. Thus, it is clear from both the wild and the laboratory that P. campbelli require biparental care to reproduce successfully and that the requirement is not relieved by optimizing the environment in a laboratory setting (18 1 C, 14L:10D). At the same time, it is clear that P. sungorus are not limited in the same way by the same conditions. Thus, we sought a functional explanation for the species difference in social structure within the physiological constraints imposed by the two habitats. C. Physiological Constraints on Reproduction in Phodopus campbelli In P. campbelli, the obligate requirement for a male parent has its basis in physiological constraints on the female imposed by the natural habitat (Wynne-Edwards, 1998). The cold, arid, seasonal environment demands superb insulation, metabolically efficient thermogeneration, and an ability to use water so efficiently that metabolic water does not need to be supplemented by drinking (Schierwater and Klingel, 1985). These phenotypic and physiological traits of Phodopus are essential components of their survival in their environment. However, a successful life history strategy demands that successful reproduction, as well as survival, be achieved in the habitat. The physiology of reproductive challenges places additional burdens on the female, all of which are affected by her small body size (Wynne-Edwards, 1998). The first reproductive challenge is limited time. The reproductive life of a female dwarf hamster is short—only a single Siberian summer. Within that time frame the physiological demands of follicular development, ovulation, implantation, pregnancy, and lactation must be met. Estrous cycles in Phodopus are only 4 days long and culminate in a spontaneous ovulation (Wynne-Edwards et al., 1987b; McMillan and Wynne-Edwards,


223

1999), although regular cyclicity can also be influenced by signals from the male (Erb et al., 1993). A luteal phase is not part of the estrous cycle but is induced by the vaginal stimulation of mating and lasts 6–8 days (Edwards and Wynne-Edwards, 1994). Beyond that interval, if the female is pregnant the corpora lutea are sustained by luteotropic factors from the placenta. Pregnancy lasts only 18 days (Edwards et al., 1994; McMillan and WynneEdwards, 1999; Parkening and Collins, 1991) and is followed by a postpartum estrus that usually results in the birth of a second litter 18 days later, as the first litter is weaned (Roy and Wynne-Edwards, 1995); although delayed implantation can also occur (Newkirk et al., 1997). Thus, on day 36 after mating, a female can wean her first, deliver her second, and conceive her third litter. To become independent before the next litter is born, as well as to reach sexual maturity in their birth year, pups also grow extremely quickly. Pups are born naked, blind, and poikilothermic at a weight of less than 2 g. By 6 days of age thermogenesis begins and thermoregulation, with the activation of brown adipose tissue, is possible at 12 days of age (Newkirk et al., 1995, 1998). Nutritional independence and pelage completion occur at 15 days, and Phodopus pups reach complete independence at a weight of about 14 g on day 18 after birth (Newkirk et al., 1995, 1998; Stulberg and Wynne-Edwards, 1998). As few as 32 days after birth, at a weight of 18 g, a female can mate and become pregnant. This is the most compact mammalian life history known (Newkirk et al., 1997). The second reproductive challenge is heat. Cold-tolerant adaptations of Phodopus leave little heat exchange surface and render individuals acutely vulnerable to hyperthermia (Weiner and Heldmaier, 1987). During late gestation and throughout lactation, a small-bodied female mammal must provide all of the energy for herself plus her offspring. To achieve this, she experiences progressive increases in basal metabolic rate that typically reach five times higher than that of a nonreproductive female of the same original body weight (Thompson, 1992; Schierwater and Klingel, 1986). This energy flux increases metabolic heat production and causes a maternal hyperthermia (Jans and Leon, 1983; Leon et al., 1983; Schierwater and Klingel, 1986). During late gestation and lactation the maternal hyperthermia in Phodopus also eliminates the 3 C circadian rhythm in core body temperature (Scribner and Wynne-Edwards, 1994a). Over the same interval, Phodopus pups develop from poikilothermic heat sinks for the mother into insulated, homeothermic pups that are a heat source (Scribner and Wynne-Edwards, 1994c). These thermal interactions with the pups drive maternal absences from the nest (Woodside et al., 1980). Female P. campbelli leave when near their upper thermal tolerance and return only when they have cooled (Scribner and Wynne-Edwards, 1994c).

224


In fact, the frequency distribution of core body temperatures in breeding females, measured each minute by telemetry, is the same in females held at an ambient temperature of 18 C and females held at an ambient temperature of 23 C (Walton and Wynne-Edwards, 1998). In each case, almost 30% of each day is spent near the upper thermal tolerance limit (more than 1 C higher than the nonreproductive core temperature) (Walton and Wynne-Edwards, 1998). However, at the higher ambient temperature, maternal thermoregulatory absences from the nest are longer (Walton and Wynne-Edwards, 1998; Wynne-Edwards, 1995). During these absences the pups cool rapidly and are unable to grow (Newkirk et al., 1995, 1998). Thus, reproductive success is low because maternal hyperthermia is not manageable without prolonged absences from the nest (Walton and WynneEdwards, 1998). Prolonged absences also lead to the loss of precious water. The third simultaneous reproductive challenge is water availability. When faced with hyperthermia, evaporative water loss, including saliva spreading, is one of the most effective means of heat dissipation (Wilson and Stricker, 1979). During lactation, the mother is also the only source of water for the litter and must provide that water in the form of milk. When possible, this water is recycled to the female through licking pup urine (Gubernick and Alberts, 1987b). However, it is not possible to recover all of the water from young pups because much of the water is lost through skin and respiratory surfaces. Huddling over the pups in an underground burrow is an excellent way to limit cutaneous and respiratory water loss by creating a local environment of high humidity. However, P. campbelli does not have access to sufficient water for these simultaneous challenges. Unable to waste saliva water for evaporative cooling, she must leave her pups to control her hyperthermia, and they immediately begin to lose water to the surrounding environment. Without environmental water, this cycle is destructive and leads to poor maternal condition and low pup viability. The final reproductive challenge is energy availability. Even here, a solitary P. campbelli is not able to meet the day-to-day energy requirements of reproduction. Like most mammals, P. sungorus females gain about 10% of their body weight during pregnancy and return that reserve to the growing litter so that the pups are weaned when the mother is close to her original body weight (Stulberg and Wynne-Edwards, 1998; Scribner and Wynne-Edwards, 1994b; McInroy et al., 2000). In contrast, a solitary P. campbelli females loses about 10% of her body weight during pregnancy and an additional 10% of her body weight during lactation so that she weans a litter at a body weight more than 20% lower than her original weight (Stulberg and Wynne-Edwards, 1998; Scribner and WynneEdwards, 1994b; McInroy et al., 2000). This occurs under stable laboratory conditions of ad libitum food and water availability. Therefore, it is likely


225

that the metabolic heat of catabolism imposes a limit on energy intake (Koteja, 1996). Under controlled laboratory conditions, these constraints on female ability to raise a litter alone are clear in P. campbelli. If pups are crossfostered to a mother of their own or a different species, then a P. sungorus mother rears large, plump P. campbelli pups, and a P. campbelli mother rears small P. sungorus pups that are variable in developmental stage and weight (Stulberg and Wynne-Edwards, 1998). Thus, for P. sungorus, seasonal reproduction synchronized with the peak in available rainfall and resources allows a female to rear her litter without biparental care. In contrast, conflict between adaptations for survival and adaptations for rapid reproduction creates a simultaneous time, water, energy, and heat challenge that constrains reproductive success in solitary P. campbelli (WynneEdwards, 1998). Successful reproduction is achieved through biparental care synchronized with the peak in available rainfall and resources.

D. Indirect Paternal Care in Phodopus campbelli Both in the wild, and in the laboratory, male P. campbelli provide extensive parental care to pups. They participate in the birth, are attentive to displaced pups, and continue to provision older pups as their mother moves on to invest in her next litter (Reburn and Wynne-Edwards, 1999; Jones and Wynne-Edwards, 2000, 2001; Wynne-Edwards, 1995). While in a cage with the female they also share time huddled over the pups so that the pups are rarely uncovered during the first week of development (WynneEdwards, 1995). However, these direct forms of paternal care are secondary to the indirect effects of the male on the physiology of the female. With the continued presence of the male, pups have higher survival, less variable weights, and higher weaning weights than when he is absent (Wynne-Edwards, 1987; Wynne-Edwards and Lisk, 1989; Scribner and Wynne-Edwards, 1994b). Maternal physiology is also profoundly altered. When the male remains in the cage with a pregnant female, the extent of her maternal hyperthermia is greatly reduced (Scribner and WynneEdwards, 1994a), and her core body temperature distribution remains similar to a nonreproductive pattern (Scribner and Wynne-Edwards, 1994c; Walton and Wynne-Edwards, 1998) (Fig. 6). She also avoids the 10% body weight loss of pregnancy and the additional 10% body weight loss of lactation and weans her pups at her original weight (McInroy et al., 2000). Instead, when water availability is limited, paired P. sungorus females lose weight and dependent pups although the male retains his body weight, whereas biparental care protects a P. campbelli female under the same

226


Fig. 6. Frequency distribution of female core body temperatures in P. campbelli on days 3, 9, and 15 of lactation. In P. campbelli, lactation in solitary females involves substantial hyperthermia. Conditions were as follows: (1) a solitary female held at the optimized ambient temperature of 18 C, at which pup survival and growth are excellent; (2) a solitary female held at the ambient temperature of 23 C recommended for laboratory rodent housing but known to produce poor pup survival and growth in P. campbelli, and (3) a female held at the stressful, 23 C, temperature in the continued presence of her male mate so that pup survival and growth will be excellent. Core body temperature was recorded by telemetry (0.1 C) each minute for 24 h. Values were normalized to the core body temperature of each individual after her pups were weaned and then expressed as a deviation from that 24-h mean. Frequency distributions are shown separately for body temperatures during the 10-h dark phase (thick line, activity) and the 14-h light phase (thin line, inactivity). The portion of each frequency distribution falling in the shaded area represents core body temperatures more than 1 C higher than the mean for that female. Temperature distributions in the two solitary treatments are similar, suggesting that they represent the maximum tolerable hyperthermia for an individual female during lactation. However, when that lactational hyperthermia is being managed by a female in a cool ambient environment, she successfully raises her pups, whereas the same temperature profile at the warmer ambient temperature is achieved only through failures in pup growth and survival. If the male remains in the cage with the female—even at this stressfully high ambient temperature—the female does not experience the hyperthermia and spends less than 6% of her day in a more than 1 C hyperthermic state as compared with 20% of her day when solitary. [Reproduced with permission from Walton and Wynne-Edwards (1998).]


227

conditions. She maintains her body weight and raises her litter while her male mate loses body weight (Scribner and Wynne-Edwards, 1994b). No satisfactory explanation for these changes in maternal physiology is known. Counterintuitively, male presence does not simply allow the female to spend more time away from the litter so that she can manage her core body temperature (Walton and Wynne-Edwards, 1998). Instead, the female spends more time in direct physical contact with her young pups when the male is present than when the male is absent (Walton and Wynne-Edwards, 1998; Wynne-Edwards, 1995). Her absences are also shorter, rather than rare and longer (Walton and Wynne-Edwards, 1998). This change in maternal behavior creates a stable thermal environment for rapid pup growth (Newkirk et al., 1995, 1998) and minimizes evaporative water loss from young pups. In contrast, in P. sungorus the continued presence of the male does not improve pup survival, alter maternal core body temperature, or change maternal activity patterns (Wynne-Edwards and Lisk, 1989; Scribner and Wynne-Edwards, 1994a). Pups are expected to benefit from the direct paternal care they receive. However, the indirect effects of paternal presence on maternal physiology appear to be the essential component of her improvement in reproductive success. In other words, male presence allows the female to be a ‘‘better’’ mother. Her physiological disruption is minimized, her attentiveness to pups increases, and her pups show the benefits of those improvements by having higher survival, faster growth, and heavier body weight at weaning. In fact, she also invests more in a subsequent litter and invests in a second litter sooner when the male remains in the cage (McInroy et al., 2000). Thus, the extensive behavioral repertoire of P. campbelli males toward their pups might be secondary to the physiological relief for the mother that arises from their presence in the nest. The obvious candidate for the mechanism for this improvement in maternal physiology is a combination of improvements in her water balance and water use efficiency and changes in her endocrine status that alter her behavioral motivation and or metabolic status. However, these hypotheses have not yet been subjected to experimental test. This integrated view of P. campbelli social organization, physiology, and evolution is summarized in Fig. 7. It predicts the evolutionary divergence of hormone signaling in both female and male P. campbelli that is described next.

IV. Endocrine Evolution in Phodopus campbelli Physiological adaptations to survive water stress and heat load have their primary impact on the survival of the individual. However,

228


Fig. 7. Outline of the hypothesis that the harsh ecological niche of P. campbelli resulted in conflict between adaptations for survival and adaptations for reproduction. In P. sungorus, seasonal breeding resolves the conflict. In P. campbelli, biparental care resolves the conflict. [Figure redrawn from Wynne-Edwards (1998) with permission.]

physiological constraints on reproduction also extend to the mechanisms through which the reproductive attempt is realized. Specifically, the neuroendocrine brain must interact with the social and physical environment to ‘‘manage’’ pregnancy, parental behavior, and the eventual sexual maturation of the next generation. In P. campbelli, there is evidence that the basic endocrinology of pregnancy and the behavioral endocrinology of fatherhood have also diverged from P. sungorus and their shared ancestors. A. Repeated Hormone Sampling from Freely Moving Animals Over the past 6 years, we have devoted considerable energy to developing methodology to obtain repeated measurements of hormone concentrations from individuals without causing stress or handling-related changes in those hormones. The result is a method in which individuals or


229

groups of animals of either species are housed in modified home cages with a continuous flow of compressed air that can be switched to isoflurane in O2 vehicle without approaching the cages (Reburn and Wynne-Edwards, 2000). Animals all succumb to the anesthesia within 90 s, most without ever leaving the nest area. With repeated anesthesia as often as every 2 h, we have used this method to obtain more than 12 successive samples in 24 h or samples every few days over periods of weeks. While the animals are under anesthesia, isoflurane flow is stopped and each animal is removed from the cage for retro-orbital sinus sampling of a small (75 ml) or a ‘‘one-time only’’ large (1200 ml) volume of blood. On return to the home cage, the animals return to a deeply anesthetized state because the isoflurane is denser than air and remains in the cage. Reestablishing the forced air supply clears the isoflurane and immediately revives the animals. Thus, animals can live in social or family groups for intensive, short-term studies or less intense, long-term studies. In this method development we have found that Phodopus respond to manual restraint with behavioral distress, an increase in the concentration of their dominant glucocorticoid, cortisol, and a decrease in their prolactin concentration (Reburn and Wynne-Edwards, 2000). Both effects are evident within 1 min and are potentially serious confounding variables in other sampling and anesthesia methods. In particular, hamsters that were in the same room as another hamster that was restrained had an ‘‘audience effect’’ change in their hormone concentration equivalent to being held themselves (Reburn and Wynne-Edwards, 2000). Suppression of prolactin was graded in response to the degree of stress and equaled the pharmacological prolactin reduction caused by bromocriptine mesylate (50 mg of CB154 for 3 days). In contrast, the home-cage isoflurane method alters neither cortisol nor prolactin in response to repeated sampling, prolonged isoflurane exposure, or substantial blood volume reduction (1200 ml of blood is approximately 30% of the total blood volume in an adult Phodopus). With this technique, we have been able to reduce both the number of animals used for our research and the stress that each experiences as a result of invasive procedures (Reburn and Wynne-Edwards, 2000). The method has also allowed us to study the evolution of endocrinology in hormones that are released in a pulsatile fashion, like prolactin, as well as to study the behavioral endocrinology of freely interacting animals in social and family groups. B. Endocrine Evolution in Female Phodopus campbelli Short estrous cycle mammalian species such as P. campbelli have only a follicular ovarian phase, in which the ova are prepared for ovulation

230


(Wynne-Edwards et al., 1987b). As for other short estrous cycles (Gunnet and Freeman, 1983), the signal to initiate a luteal phase (or pseudopregnancy) in P. campbelli is the vaginal stimulation during mating that initiates a neuroendocrine reflex involving pituitary prolactin release (Erb and Wynne-Edwards, 1994; Edwards and Wynne-Edwards, 1994). Prolactin surges then prevent the degeneration of the corpora lutea formed after ovulation. Corpora lutea are the primary source of the progesterone (Erb and Wynne-Edwards, 1993) that is necessary to maintain the uterus in readiness for implantation and early fetal growth (Edwards et al., 1994). After 9 days, that pseudopregnant reflex extinguishes (Edwards and Wynne-Edwards, 1994) but the corpora lutea are maintained through other luteotropic factors (Soares et al., 1998). These traits are shared by the rat, the mouse, and the golden hamster and have been assumed to be both ancestral and essential to completing a mammalian pregnancy within 3 weeks (Bronson, 1989). In all respects, P. sungorus is similar to other traditional laboratory rodents but P. campbelli is not (McMillan and Wynne-Edwards, 1998, 1999). Using our sampling method that allows repeated sampling from freely moving individuals without causing changes in hormone concentrations (Reburn and Wynne-Edwards, 2000), we have identified several features of P. campbelli maternal endocrinology that diverge from those in other domesticated laboratory rodent species. Three major differences are related to progesterone, and three are related to prolactin. Progesterone is the dominant steroid of pregnancy and plays important roles in behavioral receptivity, in uterine preparation for implantation, and in suppressing the contractility of the uterine myometrium. Thus, evolutionary changes in progesterone secretion patterns are fundamental alterations of hormonal mechanisms in reproduction. Three distinct changes in progesterone secretion from P. sungorus to P. campbelli are known (Fig. 8). First, the hormonal requirements for behavioral receptivity (lordosis) in P. campbelli do not include exposure to progesterone (Wynne-Edwards et al., 1987b; McMillan and Wynne-Edwards, 1998) and that distinction is reflected in marked attenuation of the proestrus progesterone surge during estrous cycles or before a postpartum mating (Wynne-Edwards et al., 1987b; McMillan and Wynne-Edwards, 1998) (Fig. 8, 1a vs. 1b). Second, peripheral serum contains a higher concentration of progesterone in the presence of the ephemeral corpus luteum of an estrous cycle than in the presence of a corpus luteum 2 days after mating (Erb and Wynne-Edwards, 1993; Edwards et al., 1994; 1995; Reburn and WynneEdwards, 1996) (Fig. 8, 2a vs. 2b). Third, serum progesterone concentrations during P. campbelli pregnancy do not exceed those seen in a nonpregnant estrous cycle until the last 6 days of development (Edwards


231

et al., 1994, 1995; Reburn and Wynne-Edwards, 1996; Roy and WynneEdwards, 1995; McMillan and Wynne-Edwards, 1999) (Fig. 8, 3). In effect, progesterone concentrations during P. campbelli pregnancy remain within the normal range of an estrous cycle until the last days before parturition. Prolactin is secreted in pulses during the establishment of the corpora lutea of pregnancy in species such as Phodopus that induce a luteal phase after mating. Those prolactin surges are essential for successful pregnancy and changes to the pulsatility of prolactin are therefore essential changes to the hormonal establishment and maintenance of pregnancy. Evolutionary divergences between P. campbelli and P. sungorus in prolactin and progesterone secretion patterns occur at similar stages of reproduction. First, the ephemeral life span of the nonpregnant corpus luteum is the only stage of the estrous cycle without prolactin secretion (Erb and WynneEdwards, 1994; Reburn and Wynne-Edwards, 1996; McMillan and Wynne-Edwards, 1999) in contrast to the surges seen 2 days after mating (Edwards et al., 1995; McMillan and Wynne-Edwards, 1999) (Fig. 8, 4a vs. 4b). This interval is the most sensitive for pregnancy block in P. campbelli (Wynne-Edwards et al., 1987b). Second, mating on proestrus is not accompanied by prolactin release except before a postpartum mating (McMillan and Wynne- Edwards, 1998, 1999) (Fig. 8, 5a vs. 5b). Prolactin during the last days of pregnancy is still seen. At that time, it is essential for maternal behavior in P. campbelli. Pharmacological suppression of prolactin in late pregnancy does not alter birth weight or the presence of milk in neonatal stomachs, but does reduce pup survival over the first week of life (Edwards et al., 1995). Third, both P. sungorus and P. campbelli lose maternal pituitary prolactin surges by day 12 of pregnancy, at the time that placental lactogens are expected to take over (Edwards et al., 1995; McMillan and Wynne-Edwards, 1999) (Fig. 8, 6a). However, maternal pituitary prolactin surges resume by day 15 in P. campbelli (Fig. 8, 6b) but not in any other domesticated rodent pregnancies (Edwards et al., 1995). The temporal pattern of prolactin surges in P. campbelli pregnancy suggests (1) that the absence of surges is the key indicator that a female is not pregnant, rather than the presence of surges being the indicator that a female is pregnant, and (2) that the maternal pituitary is involved in late pregnancy. C. Adaptive Value of Endocrine Evolution in Female Phodopus campbelli The adaptive value of these evolutionary changes in reproductive endocrinology is not known. From a theoretical perspective, it could be argued that the hormonal mechanisms supporting pregnancy should be highly conserved across mammalian species. Pregnancy is an ancestral trait

Fig. 8. Concentrations of progesterone (ng/ml; top) and prolactin (ng/ml; bottom) during an estrous cycle (E, estrus; D1, diestrus 1; D2, diestrus 2; P, proestrus) and pregnancy (days 2, 3, 6, 9, 12, 15, and 18 of the 18-day pregnancy) for P. campbelli and P. sungorus females. Within any 24-h period, points connected by a line represent repeated measurements of a cohort of females sampled by the method detailed in Reburn and Wynne-Edwards (2000). Annotations note key developmental shifts as landmarks for the hormonal patterns shown. The pattern shown for P. sungorus is broadly similar to familiar patterns in rats, mice, and golden hamsters and is thus considered both ancestral and conserved. Large numbers in circles identify the key differences between P. campbelli and this ancestral pattern that are discussed in text. Three of those differences are in progesterone secretion. They relate to (1) a reduced role for progesterone in behavioral receptivity, (2) a reduction, rather than


233

of the class Mammalia and is essential for reproductive success. Any mutation that alters regulatory pathways involved in pregnancy would therefore carry a high risk of lost fertility and would be unlikely to persist. On the other hand, any mutational change in regulatory pathways of pregnancy that improved the ability of a female to reproduce in her ecological niche would be so advantageous that strong selection for the trait would help it to spread through the population. The similarities between P. sungorus and other laboratory rodent species that are distant relatives suggest that the hormonal pathways regulating pregnancy have been conserved over evolutionary time (McMillan and Wynne-Edwards, 1998, 1999). At the same time, differences between the two Phodopus species suggest that strong selection has acted on P. campbelli since it shared a common ancestor with P. sungorus. In Fig. 7, the cold, dry, seasonal habitat of P. campbelli is hypothesized to be the constraint that interacted with small body size to yield extensive biparental care in P. campbelli. Endocrine evolution in P. campbelli since it shared an ancestor with P. sungorus should also be considered an adaptive response to that ecological niche and social organization. In that context, these examples of endocrine evolution are adaptations that might allow a female to adjust her reproductive investment. In contrast to P. sungorus, P. campbelli females typically rear litters with a high variance in developmental status and body weight (Stulberg and Wynne-Edwards, 1998). That is, they tend to raise one or more pups to an excellent body size and to allow other pups to grow poorly (WynneEdwards, 1987; Wynne-Edwards and Lisk, 1989; Newkirk et al., 1998). The relative roles of pup–pup competition versus maternal discrimination in this differential growth are not known. As an index of the variance, we have typically compared the number of pups weaned at a weight of 11.5 g or more on day 18 after birth (Wynne-Edwards, 1987; Wynne-Edwards and Lisk, 1989; Newkirk et al., 1995, 1998; Stulberg and Wynne-Edwards, 1998; McInroy et al., 2000). This is conservatively determined, on the basis of the weight of dispersing juveniles in the wild, as the minimum viable body an increase, in progesterone from diestrus 1 of an estrous cycle to day 2 of pregnancy, and (3) the absence of an increase in progesterone as the placenta growth is completed. The remaining three key differences occur in prolactin secretion patterns. Those relate to (4) the absence of prolactin surges on diestrus 1 compared with the other 3 days of an estrous cycle and compared with day 2 of pregnancy, (5) the absence of prolactin surges on the day of mating (proestrus) in contrast to the presence of surges immediately before the postpartum mating (with a probable role in the onset of lactation), and (6) the expected loss of prolactin surges after placenta development (day 12) as compared with the unexpected resumption of those surges during the phase of rapid embryonic growth (day 15). [Redrawn from McMillan and Wynne-Edwards (1999) with permission.]

234


weight at independence (Wynne-Edwards, 1987). The threshold acknowledges the laboratory artifact that very small pups of the same age ( 0.05), species with polygynous mating systems (Birdsall and Nash, 1973; Xia and Millar, 1991) and low levels of paternal behavior

PATERNAL BEHAVIOR AND AGGRESSION

277

(Wolff and Cicirello, 1990; Schug et al., 1992; Bester-Meredith et al., 1999). The variation in T levels across Peromyscus species may be more likely due to phylogenetic or ecological constraints, as the white-footed mouse, deer mouse, and oldfield mouse are more closely related to each other than to the other three species examined (Avise et al., 1974; Stangl and Baker, 1984; Kass et al., 1992), and all three species with lower T levels live in desert habitats in the southwestern United States and Central America (Baker, 1968; Sullivan et al., 1997). Our results indicate that high baseline T levels are not incompatible with a monogamous, highly paternal mating system. In addition, the variation in T levels across Peromyscus species is more likely due to phylogenetic or ecological constraints, rather than to the species typical mating system. These results highlight the complexity of the relationships between hormone levels, phylogeny, ecology, and behavior and, as in some of the bird studies, suggest that habitat may affect the function of T. Future T manipulations in Peromyscus species other than the California mouse (as discussed below) may prove to be illuminating. 2. Hormone Manipulation Studies Testosterone manipulations have been performed in the California mouse, the prairie vole, and the Mongolian gerbil. We have investigated experimentally how baseline T levels affect paternal behavior in the California mouse. The birth of pups in California mice is associated with a significant decrease in male T levels. Male California mice show high levels of paternal behavior starting on the day of parturition (Gubernick and Alberts, 1987b), and continue to care for pups throughout their development until weaning (Bester-Meredith et al., 1999). Interestingly, the onset of expression of paternal behavior coincides with the postpartum estrus (Gubernick, 1988). The California mouse therefore may be similar to the fish species referenced earlier because high aggression levels associated with mate-guarding behavior may occur during the same time frame that paternal care is expressed. Most males undergo a transition from attacking or ignoring pups to exhibiting paternal behavior immediately after the birth of their own pups (Gubernick and Nelson, 1989). We manipulated baseline T levels in the California mouse to examine the effects on paternal behavior (Trainor and Marler, 2001, 2002; K. Cravens, B. C. Trainor, and C. A. Marler, unpublished data). We observed castrated males with their own pups from their second litter; castrations or sham manipulations were performed after the postpartum mating period following the birth of their first litter. Castration decreased paternal behavior whether males were observed with their own pups or with foster pups. Castration combined with T replacement therapy restored paternal behavior in males observed interacting with foster pups (Trainor and

278

CATHERINE A. MARLER et al.

Fig. 2. Huddling and pup-grooming behavior of castrated (solid columns) and intact male California mice (open columns) at different stages in the development of their pups. Castration reduced both huddling and pup grooming on day 3 (modified from Trainor and Marler, 2001), but only pup grooming on days 15–20 (K. Cravens, B. C. Trainor, and C. A. Marler, unpublished data). **p < 0.001; *p < 0.05.

Marler, 2001). The effect of T on paternal behavior can, however, vary with different stages of pup development. In Fig. 2, we summarize data from two studies. In the first study, males were observed for 30 min with their mate and pups 3 days after parturition (Trainor and Marler, 2001). In a second study, males were observed for eight 10-min observation periods spaced throughout the dark phase between days 15 and 20 after parturition (K. Cravens, B. C. Trainor, and C. A. Marler, unpublished data). To facilitate comparisons between the two studies, analyses were conducted on rates of behavior during observations. Castration reduced pup-grooming rates both on day 3 and during days 15–20 after birth, but only huddling on day 3. However, examination of huddling rates across the studies shows that on days 15–20 both castrated and intact males huddled less frequently than did castrated males on day 3. This is not surprising as male huddling behavior steadily decreases as pups mature (Fig. 3; Bester-Meredith et al., 1999) and are able to thermoregulate more efficiently. Although pup grooming responded to T across different stages of pup development, the functional basis of this behavior may change. Even though the overall rate of pup grooming changes little over time (Fig. 3; Bester-Meredith et al., 1999), males decrease the amount of time spent grooming the anogenital region of pups 2 weeks after


279

Fig. 3. Amount of time spent huddling with pups, outside of the nest, licking/grooming pups, retrieving pups, and nest building by male white-footed mice (P. leucopus) and California mice (P. californicus) during 1200-s observation periods beginning on day 3 and ending on day 23 after the birth of pups (from Bester-Meredith et al., 1999). Data are presented as means standard errors. *p < 0.05.

parturition (Gubernick and Alberts, 1987a). As the pups mature, the function of pup grooming may change from stimulating the development of the anogenital tract (Moore, 1984, 1992) to other functions such as coat maintenance or possibly social functions. Testosterone also mediates the effect of exposure to pups on paternal behavior. When castrated males with T implants were tested with foster pups on consecutive days, males showed significantly higher levels of pup grooming (paired t12 ¼ 2.24, p ¼ 0.04) and tended to show higher levels of huddling (paired t12 ¼ 1.99, p ¼ 0.06) on the second day compared with the first (Trainor and Marler, 2002). This effect was absent for pup grooming (paired t12 ¼ 0.86, p ¼ 0.4) and huddling (paired t12 ¼ 1.02, p ¼ 0.32) in

280


castrated males without T replacement. Thus, higher T levels were associated with higher levels of paternal care. The reported variability in the effects of T on male parental behavior in other rodent species may reflect species differences or, as we discuss below, variation in methodology (i.e., social experience). When virgin adult male prairie voles were castrated and exposed to pups, they displayed either a decrease in paternal behavior (Wang and De Vries, 1993), which was restored with T implants, or no change (Lonstein and De Vries, 1999). However, T can act early during postnatal development in prairie voles to promote the expression of parental behavior after maturation (Lonstein et al., 2002), although apparently not prenatally (Roberts et al., 1996; Lonstein et al., 2002). This effect of T during early development may be related to social structure because the prairie vole is a cooperative breeder (Getz et al., 1990) and can express parental behaviors before acquiring reproductive experience (Lonstein and De Vries, 1999). In such species, the positive influence of T on paternal behavior may occur early during development, before expression of paternal behavior. The role of T may also be related to territorial aggression levels because male prairie voles display a high degree of home range overlap, in contrast to the fairly exclusive territories maintained by California mice (review by Goodson and Bass, 2001). Overall, T is not negatively related to paternal behavior in prairie voles and, in fact, there is evidence that T may be positively associated with paternal behavior. It also demonstrates, however, how variation in the timing of paternal behavior and aggression might affect how T influences these behaviors. In contrast to California mice and prairie voles, an increase in paternal behavior followed castration in Mongolian gerbils (Clark and Galef, 1999). Methodological differences may explain why these results differ. The male gerbils were sexually inexperienced, but had cohabitated for 10 days with a female already inseminated by a male. Males therefore lacked the normal sequence of stimuli typically experienced before having pups (i.e., mating, cohabitation, birth of pups). These additional stimuli may initiate paternal behavior, whereas T may maintain paternal behavior and the effect of T on parental behavior may change after the onset of paternal behavior. The initiation of paternal behavior may be more strongly associated with other steroid and peptide hormones. For example, correlational studies have documented that steroid and peptide hormones such as prolactin change before parturition (Brown et al., 1995; Gubernick and Nelson, 1989; Reburn and Wynne-Edwards, 1999). These changes could play an important role in priming males to exhibit paternal behaviors in response to changes in T. It will therefore be important to examine the role of T in maintaining paternal behavior in male gerbils.


281

Summary In rodents, there does not seem to be a simple relationship between T levels and paternal behavior. Across species of Peromyscus, low T levels are not a prerequisite for paternal behavior as seen with high baseline T levels measured in the monogamous P. polionotus. Within rodent species, in almost every case, male T levels peak at parturition and decline shortly afterward. There is substantial variation in the timing of the onset of paternal behavior across species, although the onset of paternal behavior usually appears after a period of increased T levels. In Djungarian hamsters, Mongolian gerbils, and California mice, the onset of paternal behavior follows the high T levels preceding parturition while in the prairie vole the onset of male parental behavior occurs postnatally. This raises the possibility that the increased T levels observed before the onset of paternal behavior are more important than the decreased T levels that have been observed after parturition. Again, high T levels are not incompatible with some aspects of paternal behavior. The potential interpretations of the correlational and manipulative California mouse studies also demonstrate the difficulties of extrapolating a relationship between T and paternal behavior from only correlational studies, as described in the set of amphibian studies. While the correlational studies indicated a negative relationship between T and paternal behavior in California mice, the manipulative studies demonstrated a positive relationship between T and paternal care. While T levels were lower in fathers, it was these low T levels that maintained paternal care. Two of the three biparental rodent species displayed a positive relationship between T and paternal behavior, although the timing of the effect may differ. Because of differences among the studies and because so few species have been studied, it is not yet possible to determine whether a positive association is more likely to be found in highly paternal species with a postpartum estrus. However, these studies, combined with the previously discussed fish studies, suggest that this is still a viable hypothesis. Future studies may find that there are common actions of T on paternal behavior across rodents, although the social system of each species is likely to play an important role in explaining interspecific variation in this relationship.

E. Testosterone and Paternal Care: Nonhuman Primates Most studies examining hormones and paternal behavior in primates have documented the changes in male peripheral hormone levels across a

282


mate’s pregnancy or in response to infants. Early studies on common marmosets (Callithrix jacchus) detected no changes in T levels (Dixson and George, 1982). However, more recent studies have found somewhat contrasting results. In cotton-top tamarins (Saguinus oedipus), gradual increases in male urinary T levels were observed across a mate’s pregnancy regardless of previous reproductive experience (Ziegler and Snowdon, 2000). Testosterone levels remained high after parturition when males began infant-carrying behavior (Ziegler et al., 2000). In contrast, urinary T levels in the related black tuft-eared marmosets (Callithrix kuhlii) were negatively correlated with paternal behavior. Male black tuft-eared marmosets with more parental experience, and fathers that spent more time carrying infants, had lower T levels (Nunes et al., 2000, 2001). Currently it is unclear why T and paternal behaviors would be positively correlated in one tamarin species, but negatively correlated in a closely related marmoset. It is possible that methodological differences may have contributed to the contrasting results across species. Studies on the cottontop tamarins reported T levels averaged across months whereas studies on the black tuft-eared marmosets reported T levels averaged across weeks. There is also variation in the onset of paternal behavior. Black tuft-eared marmoset fathers express their highest levels of infant-carrying behavior 3–4 weeks after parturition (Nunes et al., 2000), whereas cotton-top tamarin fathers express their highest levels of carrying behavior immediately after parturition (Snowdon, 1996). The timing of the paternal behaviors, as well as the methodological differences, could contribute to the varying results between these related species. Summary Again, these studies indicate that high T is not incompatible with paternal care. As in the rodents, there is also a hint that the timing of the expression of paternal behavior may be important. The species that displayed the highest levels of paternal care immediately after parturition, cotton-top tamarins, also maintained high T levels during this time period. As in the rodents, there was also an elevation of T in males immediately before parturition. F. Testosterone and Paternal Care: Humans Studies on human participants found differences in testosterone levels between fathers after parturition and expectant fathers during pregnancy that are qualitatively similar to patterns observed in other paternal mammals (e.g., Wynne-Edwards and Reburn, 2000). Testosterone levels were lower in males sampled during the 3-week period after parturition, but


283

were not different from preparturition levels in a separate group of males sampled 4–7 weeks after parturition (Storey et al., 2000). Cortisol levels were increased during the 4 weeks before parturition and did not fully return to baseline until 4–8 weeks later. Especially interesting were behavioral observations, in which men were asked to hold a doll on their shoulder during a 30-min interview period. Men who held the doll on their shoulders for the full 30 min had significantly lower T levels and a higher prolactin response to a crying baby than did men who put the doll down before the 30-min interview was finished (Storey et al., 2000). In another experiment men responded to tapes of crying infants with an increase in T during the 3-week period after parturition, but not at other times. Although these findings do not involve actual paternal behaviors, such comparisons between hormone changes and behavior from the same individuals are relatively rare, even in the animal literature. Despite the small sample sizes, the results suggest that male hormonal changes may be associated with paternal behavior. In a more recent study, on average, male T levels declined on becoming fathers, but there was extensive variation in the pattern of T level changes surrounding parturition (Berg and Wynne-Edwards, 2001). Summary As in many other species, it is not clear whether human paternal behavior alters hormones levels or whether hormones are more important in regulating behavior. However, T levels appear to be associated with paternal behavior in human fathers. G. Summary for Testosterone and Paternal Care Across taxa there appears to be considerable variation in the association between androgens and paternal care (Table I), despite a general tendency to assume a negative interaction. For example, T was high, or at least above baseline, when paternal behavior was expressed in a variety of species. In several species, artificial increases in T did not always result in a large decrease in paternal behavior. Finally, artificial increases in T raised the level of paternal behavior in several fish and rodent species. One factor that may dictate some of the general effects of androgens on paternal behavior is the temporal association between aggression, mating, and paternal behavior. In temperate zone, seasonally breeding birds, mating and paternal behaviors are temporally dissociated and T inhibits paternal behavior. In contrast, in species in which mating behavior, paternal behavior, and aggression coincide, such as in a number of fish species and California mice, T can promote paternal behavior. This effect of temporal association or dissociation of behaviors and hormones on

284


behavior–hormone relationships may parallel the variation found in androgen control of mating behaviors mentioned earlier (e.g., Crews, 1984). It was proposed that testosterone control of mating behavior depends on the temporal association of mating behavior, gametogenesis, and high levels of T (Crews, 1984). More species using androgen manipulations must be further studied to determine the predictive power of the hypothesized relationship. At a minimum, however, this review demonstrates that there are a number of species that do not fit the generally held view that there is a negative association between paternal care and T.

III. Arginine Vasopressin: Functionally Similar to Testosterone? The neuropeptide AVP and T are related both at a behavioral and a cellular level. Significant parallels can be drawn between the behavioral correlates of AVP and of T, including the diversity of social behaviors that they can influence. Arginine vasopressin and T may act in concert to shape sexually dimorphic behaviors such as aggression and may also function to make the two sexes more similar with respect to parental behaviors (reviewed by De Vries and Boyle, 1998). In this section, we describe the positive effects of AVP on aggression. We also describe how AVP can positively influence paternal care, despite evidence of a positive relationship between AVP and aggression. Thus, as in the relationship between T and social behaviors, we also find that AVP can have positive effects on both paternal behavior, and aggression. A physiological relationship between T and AVP also has been demonstrated. For example, castration in rodents can cause a decrease in AVP immunoreactivity and receptors in the bed nucleus of the stria terminalis and amygdala (reviewed by De Vries and Boyle, 1998; Viau et al., 1999; Young et al., 2000). In addition, T can modify the behavioral response to central injections of AVP (e.g., Albers et al., 1988) and can accomplish this by altering densities of V1 AVP receptors (Delville et al., 1996). Thus, levels of the two hormones (AVP is referred to as a hormone instead of a neurochemical in this review) are at least partially associated in a positive manner in rodents, and again both have the potential to positively influence aggressive as well as paternal behavior in some species. A. Arginine Vasopressin: An Aggression Hormone, a Nurturing Hormone, or Both? In this section, we discuss evidence that AVP has positive associations with both aggression and the nurturing behaviors found in paternal care,


285

thus further illustrating that the two categories of behavior can be positively associated with each other through linkage with the same underlying physiological mechanisms. Because AVP is less well studied than T, we discuss the relationship between AVP and aggression in more detail than in the previous discussion of the relationship between T and aggression. Arginine vasopressin has been associated with aggression in mammals (reviewed by De Vries and Boyle, 1998; Koolhaas et al., 1998), while the nonmammalian homolog, arginine vasotocin (AVT), has similar effects in fish (e.g., Semsar et al., 2001), birds (reviewed by Goodson and Bass, 2001), and amphibians (e.g., Semsar et al., 1998; Klomberg and Marler, 2000). Within mammals, AVP has been associated with aggression in a number of species including house mice (e.g., Bluthe et al., 1993; Compaan et al., 1993), Rattus spp. (e.g., Everts et al., 1997), prairie voles (e.g., Winslow et al., 1993), golden hamsters (Mesocricetus auratus; reviewed by Ferris, 1992), California mice (e.g., Bester-Meredith and Marler, 2001), and humans (Coccaro et al., 1998). Variation in aggressive behavior has been correlated with AVP-immunoreactive cell and fiber characteristics, peptide and receptor mRNA, and AVP receptor binding (reviewed by Goodson and Bass, 2001). In mammals, attention has been focused primarily on the behavioral function of a sexually dimorphic AVP pathway originating in the bed nucleus of the stria terminalis and medial amygdala and projecting to the lateral septum (reviewed by De Vries and Miller, 1998). The density of AVP-immunoreactive staining in this sexually dimorphic pathway differs between strains of house mice bred for long and short attack latencies (Compaan et al., 1993). The medial amygdala, which provides input into this pathway and has extensive connections to the olfactory system and to other brain areas associated with social behavior, has been associated with aggression in rats (Koolhaas et al., 1990). In addition, evidence from the golden hamster suggests a behavioral role for AVP located within the hypothalamus, particularly in the anterior hypothalamus, a brain area associated with aggression (Delville et al., 1998, 2000). In the golden hamster, neuroanatomical tracing studies and comparisons of fos expression reveal connections between the anterior hypothalamus and the medial amygdala and the bed nucleus of the stria terminalis that are associated with offensive aggression (Delville et al., 2000). Interactions between the vasopressinergic neurons of this pathway and serotonergic neurons may be critical in the regulation of aggressive behavior in hamsters (Ferris and Delville, 1994; Ferris et al., 1997). Overall there is strong evidence for an association between AVP/AVT and aggression in these brain areas across a wide variety of species; we emphasize here

286


associations within three mammalian genera: Mesocricetus, Microtus, and Peromyscus. For the relationship between AVP and paternal behavior we focus on Microtus and Peromyscus. Fewer studies have examined the relationship between AVP and paternal behavior. Nonetheless, an emerging picture from these two genera is that there are close links among aggression, paternal behavior, and, in some cases, pair bonding: three behavioral categories that appear to be positively associated in these species with each other, as well as AVP. Within brain areas where limbic system fos expression has been associated with paternal behavior in the prairie vole, including several areas connected to the medial amygdala such as the lateral septum and the bed nucleus of the stria terminalis (Kirkpatrick et al., 1994a), AVP has been identified as a possible substrate of paternal behavior. Disruption of this system via removal of input from the olfactory bulb disrupts the expression of paternal care in prairie voles (Kirpatrick et al., 1994b). The possible role of AVP in the lateral septum of prairie voles and in the bed nucleus of the stria terminalis in California mice is discussed in more detail later. We also review the potential plasticity of the AVP neurochemical system in response to social conditions provided by conspecifics, with subsequent effects on aggressive and parental behaviors. The plasticity of the AVP neurochemical system becomes crucial when we later discuss the cross-generational effects of parental behavior and its interactions with aggression and AVP in California mice. B. Arginine Vasopressin and Aggression: Golden Hamsters (Mesocricetus auratus) The most detailed studies that relate AVP to aggression have been performed with adult golden hamsters. In golden hamsters, rank and dominance behavior, as well as specific aggressive behaviors, have been linked to AVP in both group-housed and isolated individuals. Subordinate males have significantly lower levels of AVP immunoreactivity in the anterior hypothalamus as compared with dominant males (Ferris et al., 1989). Manipulations of AVP also alter dominance and aggressive behaviors toward an intruder: AVP microinjections increase dominance behaviors, whereas AVP blockers decrease them (Ferris et al., 1984, 1986, 1988, 1993, 1997; Ferris and Potegal, 1988). Arginine vasopressin receptor blockers also decreased the number of attacks and increased attack latency toward a conspecific male intruder (Ferris and Potegal, 1988; Potegal and Ferris, 1990). However, the social status of a male golden hamster can modify the effectiveness of AVP manipulations on aggression.


287

A subordinate male injected with AVP will increase flank-marking behavior and will become the dominant individual in a pair (Ferris et al., 1986). Conversely, a dominant male treated with the AVP blocker dPTyr(me)AVP becomes the subordinate individual in a pair (Ferris et al., 1986). This is accompanied by a decrease in flank-marking behavior. However, a dominant male will not increase flank-marking behavior in response to AVP injections, possibly because he has reached a maximum level of AVP-induced flank-marking behavior. In summary, the evidence of a contribution by AVP to the control of dominance and aggressive behaviors is extensive in golden hamsters. As suggested indirectly by the studies described above, there is plasticity in the AVP neurochemical system in response to social conditions in adults (e.g., Ferris et al., 1989), but plasticity can also occur during development in golden hamsters. Male golden hamsters exposed to aggressive adults shortly after weaning were less likely than control males to attack individuals of a similar size to themselves. In these socially subjugated males, AVP levels were also 50% lower in the anterior hypothalamus (Delville et al., 1998; although it should be noted that these males were also more likely to attack intruders weighing 35–40% less). By altering the social conditions of these males during late development, adult neurotransmitter levels and aggression levels were also altered. Overall, these studies on golden hamsters demonstrate that within a species, AVP can influence aggression and that changes in social conditions can alter both behavior and the AVP neurochemical system. Summary The studies with golden hamsters elegantly demonstrate a positive association between aggression and AVP in a mammal. In addition, these studies provide evidence that alterations in AVP can cause plasticity in competitive aggression and that changes in social condition can also alter AVP levels. C. Arginine Vasopressin, Aggression, Paternal Behavior, and Pair Bonding: Voles (Microtus) The formation of pair bonds in prairie voles coincides with an increase in aggression that may be associated with mate-guarding aggression or possibly with increased defense of pups against infanticide. In this section, we discuss how AVP is involved in these behaviors, as well as with paternal behavior, another behavior associated with mating. For example, AVP manipulations reveal that AVP can regulate both affiliative behaviors and the formation of pair bonds (Winslow et al., 1993; Cho et al., 1999; Young

288


et al., 1999; Pitkow et al., 2001; Liu et al., 2001). In addition, at the time of mating there is a significant change in AVP-immunoreactive staining in the lateral septum (Bamshad et al., 1994) and an increase in AVP gene expression in the bed nucleus of the stria terminalis (Wang et al., 1994b), suggesting that mating may trigger an AVP-induced increase in mateguarding aggression. This onset of mate-guarding aggression can also be blocked through administration of an AVP antagonist (Winslow et al., 1993). AVP injections also caused an increase in resident–intruder aggression in sexually experienced male prairie voles, although not in the promiscuous montane vole (Young et al., 1997). Interestingly, AVP administered during development also increased adult aggression (Stribley and Carter, 1999), suggesting that the formation of AVP pathways earlier in development may shape adult levels of mate-guarding aggression. Therefore, in male Microtus, it has been hypothesized that AVP may regulate a variety of behaviors associated with pair bond formation that tie in with the increased aggression. In addition to the role of AVP in mate-guarding aggression and the formation of pair bonds, it is also important for the regulation of paternal care, another behavior associated with pair bond formation in prairie voles. After young are born, male prairie voles display high levels of paternal care (reviewed by Wang and Insel, 1996). The onset of paternal behavior coincides with an increase in AVP gene expression in the bed nucleus of the stria terminalis and a decrease in AVP-immunoreactive staining in males, suggesting an increase in production and release of AVP as paternal behavior is initiated (Bamshad et al., 1993, 1994; Wang et al., 1994b, 2000). Castration eliminated AVP-immunoreactive staining in the lateral septum and caused a decrease in the expression of paternal behavior (Wang and De Vries, 1993). Furthermore, intracerebroventricular injections of AVP increased paternal behavior, whereas an AVP antagonist decreased paternal behavior (Wang et al., 1994a). The effects of AVP on paternal behavior also are supported by the finding that AVP induced paternal behavior (an aggregate measure of pup grooming, huddling, and time spent in contact with the pup) in the facultatively paternal meadow vole (Microtus pennsylvanicus) (Parker and Lee, 2001). In this same species, AVP receptor-binding patterns are also altered by the experience of mating and raising their own pups (Parker et al., 2001). Overall, there is considerable evidence that AVP promotes paternal care, suggesting that AVP is involved in the regulation of a suite of behaviors associated with pair bond formation, including mate-guarding aggression and paternal care. These studies also provide evidence of plasticity in AVP within individuals. As previously mentioned, within Microtus, changes in AVP immunoreactivity and AVP receptor-binding patterns were induced by the


289

experience of mating and raising pups. This demonstrates that AVP may change in response to events, which in turn leads to a potential change in paternal behavior. Summary The Microtus studies do not demonstrate the association between aggression and AVP as strongly as in the golden hamsters; however, the positive association between AVP and paternal behavior emerges in these vole species. In addition, these studies provide further evidence of plasticity in the AVP neurochemical system in response to experience. D. Arginine Vasopressin, Aggression, and Paternal Behavior: California Mouse (Peromyscus californicus) and White-Footed Mouse (Peromyscus leucopus) There is evidence of an association between AVP and both paternal and aggressive behaviors in Peromyscus, although the role of AVP in regulating affiliative behavior has not yet been examined. AVP may be associated with resident–intruder aggression in both unmated and mated individuals, similar to both prairie voles and golden hamsters. In this section, we discuss how two species of Peromyscus differ in several types of behavior and then describe the links between these behaviors and AVP. 1. Aggression and Paternal Behavior California mice are monogamous, and perhaps even more strictly monogamous than prairie voles (Ribble and Salvioni, 1990; Ribble, 1991), and males display high levels of parental behavior (Fig. 3) equivalent to levels found in females (Dudley, 1974a,b; Gubernick and Alberts, 1987b; Gubernick and Teferi, 2000). The white-footed mouse provides a contrast to the California mouse because males are polygynous and display lower levels of facultative paternal care (Fig. 3) (Xia and Millar, 1988; Schug et al., 1992). For example, male white-footed mice rarely retrieve pups during the third week after the birth of the pups and also show less huddling and licking behavior toward the pups (Bester-Meredith et al., 1999). In addition to these species differences in mating system and in paternal investment, the monogamous California mouse and polygynous whitefooted mouse also differ in levels of aggression. We investigated aggression in both Peromyscus species, using two types of tests, and, as is discussed later, these tests revealed two different types of aggression. In a resident– intruder aggression test, a male is placed in an observation cage 48–60 h before an encounter and, at the onset of the test, an intruder male is placed

290


Fig. 4. Attack latencies in resident–intruder and neutral aggression tests for male whitefooted mice (P. leucopus) and male California mice (P. californicus) (from Bester-Meredith et al., 1999). Data are presented as means standard errors. *p < 0.01.

into the smaller chamber of the observation cage. This testing paradigm gives males a resident advantage such that they are more likely to win an aggressive encounter (Archer, 1988) and is similar to territorial aggression because it involves defense of a familiar site. During a neutral aggression test, both animals are placed simultaneously in separate compartments of a novel observation cage. All tests were terminated as soon as an attack occurred and if no attack occurred, mice were separated after 10 min. Using these two types of aggression tests, we confirmed field studies by other researchers that male California mice are highly territorial and aggressive (Fig. 4) and maintain exclusive territories (Ribble and Salvioni, 1990; Bester-Meredith et al., 1999; Bester-Meredith and Marler, 2001), unlike male prairie voles, which display more overlap in their home ranges (Getz et al., 1981). In contrast, white-footed mice are less aggressive than California mice and display territoriality only when population densities are high (Fig. 4) (Metzgar, 1971; Wolff and Cicirello, 1991; BesterMeredith et al., 1999; Bester-Meredith and Marler, 2001). Therefore, we found that these two species display different patterns of paternal care and aggression, two behaviors that have been linked with AVP. These two species also differ in AVP-immunoreactive staining, with California mice showing more extensive AVP immunoreactivity in the bed nucleus of the stria terminalis in addition to higher levels of aggression and paternal care (Bester-Meredith et al., 1999). 2. Arginine Vasopressin, Aggression, and Paternal Behavior As with golden hamsters and voles, the AVP neurochemical system may underlie the species-typical patterns of social behavior described above.


291

We examined the associations between neurochemical and behavioral patterns by using both between and within-genera comparisons. We first compared AVP and behavioral patterns between Microtus and Peromyscus species. Specifically, we have hypothesized that species patterns of aggression in sexually inexperienced males are better predictors than paternal care of AVP-immunoreactive staining and receptor distribution at a species level (Bester-Meredith et al., 1999). We rejected paternal behavior as a predictor of AVP-immunoreactive staining patterns in these broad species comparisons because of differences in patterns of AVP-immunoreactive staining between paternal and nonpaternal Peromyscus and Microtus species. Whereas the paternal Microtus species, the prairie vole, had lower levels of AVP-immunoreactive staining in the bed nucleus of the stria terminalis than the less paternal species, the polygamous and less paternal meadow vole (M. pennsylvanicus) (Wang, 1995), and fewer AVP receptors in the lateral septum than the polygamous and less paternal montane vole (M. montanus) (Insel et al., 1994; Wang et al., 1997; Young et al., 1997), the opposite was true for Peromyscus. The more paternal California mouse had higher levels of AVP-immunoreactive staining in the bed nucleus of the stria nucleus and amygdala and higher receptor densities in the lateral septum as compared with the less paternal white-footed mouse (Bester-Meredith et al., 1999). There is evidence, however, that the AVP-neurochemical patterns in these brain areas may be better explained by aggression and, as proposed by Goodson and Bass (2001), by social spacing patterns. Within these Peromyscus and Microtus species, those with higher levels of AVPimmunoreactive staining and AVP receptor density (California mice and meadow voles) may be more aggressive (Microtus: Hofmann et al., 1982; Getz, 1962; Dewsbury, 1983; but see Colvin, 1973; Peromyscus: see above), and also maintain more exclusive territories (reviewed by Goodson and Bass, 2001). These results do not exclude AVP control of paternal behavior at a finer level of analysis (as described below) or address the question of plasticity in AVP response to experiences such as mating or raising pups, but do provide suggestive evidence that AVP-immunoreactive staining patterns in inexperienced male rodents could potentially be explained better by differences in aggression levels between species. Within Peromyscus species there is further evidence linking AVP and aggression. When we cross-fostered between species of Peromyscus such that the more aggressive and paternal California mouse pups were raised by the less aggressive and less paternal white-footed mouse parents, and vice versa, we found changes in both aggression and AVP-immunoreactive

292


Fig. 5. Attack latencies in the resident–intruder test and test of aggression in a neutral arena for control white footed mice (resident–intruder, n ¼ 12; neutral arena, n ¼ 13), crossfostered white-footed mice (resident–intruder, n ¼ 14; neutral, n ¼ 13), cross-fostered California mice (resident–intruder, n ¼ 10; neutral, n ¼ 9), and control California mice (resident–intruder, n ¼ 24; neutral, n ¼ 23) (from Bester-Meredith and Marler, 2001). A shorter attack latency indicates a higher level of aggression. Data are presented as means standard errors. *p < 0.05, **p ¼ 0.01.

staining in the bed nucleus of the stria terminalis (Table II, Fig. 5, and Fig. 6). However, there is an interesting interaction between the species and the two forms of aggression tests. While both species became more similar to their foster parents with respect to level of aggression, the two species differed in the type of aggression that changed (Fig. 5 and Table II). California mice raised by the less aggressive white-footed mice displayed a decrease in resident–intruder aggression, but no change in neutral aggression. In contrast, white-footed mice raised by the more aggressive California mice displayed an increase in neutral aggression, but no change in resident–intruder aggression. It is important to note that changes in AVP in the bed nucleus of the stria terminalis appeared to be associated with resident–intruder aggression but not neutral aggression because changes in AVP immunoreactivity were observed only in California mice, the species displaying a change in resident–intruder aggression (Table II and Fig. 6). Further evidence from manipulations and correlations strengthens the proposed association between AVP and resident–intruder aggression. In the case of resident–intruder aggression, an intracerebroventricular injection of 1 ng of an AVP V1a receptor antagonist blocked resident–intruder

293


TABLE II Effects of Cross-Fostering on Aggression and Arginine Vasopressin-Immunoreactive Staining in California Mice and White-Footed Micea Effect of cross-fostering on: Species crossfostered

R–I aggression

Neutral arena aggression

California mice

Control > crossfostered

No effect

White-footed mice

No effect

Control < crossfostered

Paternal behavior

AVP-IR staining in the BNST

Control > crossfostered No effect

Control > crossfostered No effect

Abbreviations: AVP, Arginine vasopressin; IR, immunoreactive. Effects of cross-fostering on aggression in the resident–intruder (R–I) test, aggression in the neutral arena, paternal behavior (retrievals), and AVP-IR staining in the bed nucleus of the stria terminalis (BNST) in male California mice (more paternal and aggressive species) raised by white-footed mice (less aggressive and paternal species) and male white-footed mice raised by California mice. The cross-fostered mice were compared with control mice. Summarized from Bester-Meredith and Marler (2001, 2003). a

aggression, but not neutral arena aggression in California mice (J. K. Bester-Meredith, P. Martin, and C. A. Marler, unpublished data). In addition, in sexually experienced males, there is a direct association between levels of aggression and AVP-immunoreactive staining: resident– intruder aggression was positively correlated with optical density in the bed nucleus of stria terminalis (J. K. Bester-Meredith and C. A. Marler, unpublished data). Therefore, the association between AVP and aggression that is found in sexually inexperienced individuals remains present and perhaps even more potent in sexually experienced males. A potential link between AVP and paternal behavior is also evident (Table II). California mice that were raised by the less parental and aggressive white-footed mice showed not only less aggression in the resident–intruder test, but also a decrease in paternal care when raising their own offspring (Table II). As adults, these cross-fostered male California mice displayed a decrease in pup retrievals, although no changes in nest building or in a composite score of time spent huddling, grooming, and in the nest (HGI score) (J. K. Bester-Meredith and C. A. Marler, 2003). Thus the decrease in AVP-immunoreactive staining in the bed nucleus of the stria terminalis found in the male California mice raised by white-footed mice is accompanied by both a decrease in resident–intruder aggression and a decrease in paternal behavior in the form of pup retrievals. In contrast, white-footed mice raised by California mice did not exhibit any change in resident–intruder aggression, paternal behavior, or

294


Fig. 6. Representative photomicrographs of AVP-immunoreactive staining in the bed nucleus of the stria terminalis, comparing sexually naive, male (a) control white-footed mice, (b) control California mice, (c) cross-fostered white-footed mice, and (d) cross-fostered California mice. Cross-fostered California mice had significantly less AVP-immunoreactive staining than the control California mice, but cross-fostering had no effect on AVPimmunoreactive staining in the white-footed mice (from Bester-Meredith and Marler, 2001). Bar: 100 mm.

AVP immunoreactivity in the bed nucleus of the stria terminalis. Therefore, only the species displaying a change in both paternal behavior and resident–intruder aggression after cross-fostering also exhibited a change in AVP immunoreactivity in the bed nucleus of the stria terminalis. Similar to our findings with aggressive behavior, additional evidence from correlations strengthens the proposed association between AVP and paternal behavior. In the case of paternal behavior, examination of AVP levels in the bed nucleus of the stria terminalis in sexually experienced males showed that the degree of AVP-immunoreactive staining in the bed nucleus of the stria terminalis was positively correlated with a composite score consisting of huddling, grooming and time spent inside the nest when cross-fostered and control males were combined (Fig. 7; J. K. BesterMeredith and C. A. Marler, 2003).


295

Fig. 7. Statistically significant correlation between maximum percent AVP-immunoreactive staining in the bed nucleus of the stria terminalis and Z-score composite scores of the average time spent huddling, grooming, and in the nest (HGI score) by in-fostered and crossfostered white-footed (P. leucopus) and California mice (P. californicus) (J. K. BesterMeredith and C. A. Marler, 2003).

In addition, AVP-immunoreactive staining was positively associated with retrievals: males that expressed retrieval behaviors during the observation periods had significantly higher levels of AVP-immunoreactive staining than did males that did not express retrieval behavior (Fig. 8; J. K. Bester-Meredith and C. A. Marler, 2003). Because our results rely on immunocytochemistry measurements, we cannot identify the direction of the relationship (i.e., amount of AVP produced or released) until we examine the effect of AVP manipulations on paternal behavior. Our results suggesting an association between paternal behavior and AVP are consistent with the findings in Microtus that AVP can influence behaviors such as pup grooming, huddling, and time spent in contact with pups (Wang et al., 1994a,b; Parker and Lee, 2001). However, because there was no effect of AVP on retrievals in the sexually inexperienced male M. ochrogaster and retrievals were not measured in the facultatively paternal M. pennsylvanicus (Parker and Lee, 2001), a critical future step is to examine the effects of AVP on retrievals in sexually experienced male California mice. Nevertheless, the correlations we found between AVP and the HGI score, along with retrievals, are generally consistent with the association found between AVP and paternal behavior in Microtus (see above). In addition, our cross-fostering studies again suggest that there is plasticity in both AVP and paternal behavior: this social manipulation

296


Fig. 8. Maximum percent AVP-immunoreactive staining in the bed nucleus of the stria terminalis of males that did not retrieve pups (n ¼ 20) and males that did retrieve pups (n ¼ 4) during the observation periods (J. K. Bester-Meredith and C. A. Marler, 2003). Data are presented as means standard errors. *p < 0.05.

modified both the intensity of AVP-immunoreactive staining in the bed nucleus of the stria terminalis and the degree of paternal behavior expressed by cross-fostered California mice. Summary As in golden hamsters and prairie voles, we also found in Peromyscus that AVP pathways can be associated with both aggression and paternal care and that there is plasticity in these AVP pathways. California mice and white-footed mice differ in both paternal care and aggression, with California mice showing more extensive paternal care, more aggression toward conspecifics, and higher AVP-immunoreactive staining in the bed nucleus of the stria terminalis. Cross-fostering studies indicate that AVP plays a role in aggression in sexually inexperienced California mice and it is possible that AVP modulates aggression even more strongly in mated individuals as indicated by the direct association between levels of AVP immunoreactivity and levels of aggression. In these two Peromyscus species, we found that modifications may be made in the behavior and neurochemistry of future generations by exposing them to paternal behavior atypical for their own species. Crossfostering between these two species produced a suite of behavioral changes that led fostered pups to adopt behavioral patterns typical of their foster parents, including a decrease in paternal care and resident– intruder aggression in cross-fostered California mice and an increase in neutral aggression in cross-fostered white-footed mice. Along with these behavioral changes, a reorganization of AVP pathways as typified by a


297

change in AVP-immunoreactive staining in the bed nucleus of the stria terminalis also occurred in California mice. The association between AVP and aggression is supported by other data in Peromyscus, including a decrease in resident–intruder aggression in California mice after intracerebroventricular injections of an AVP antagonist (Bester-Meredith and Marler, unpublished data). Although differences in paternal care cannot explain patterns of AVP-immunoreactive staining between rodent genera, plasticity in AVP pathways may be associated with changes in paternal care within individual rodent species. In support of this idea, we found a correlation between AVP immunoreactivity and two measures of paternal care when both species are combined. Together with the previously discussed findings in golden hamsters and voles, these results indicate that AVP may be associated with both aggression and paternal care in rodents. In monogamous species, AVP also functions to modify affiliative behaviors, as in the prairie vole; however, this remains to be tested in Peromyscus. Finally, plasticity in AVP pathways may underlie plasticity in the expression of social behaviors in these species. E. Comparison of Functions of Vasopressin and Testosterone Parallels can be drawn between the functions of AVP and T across species. The previous discussion indicates that both AVP/AVT and T have the potential to influence positively both aggression and paternal behavior. These hormones clearly have a significant impact on social behaviors. The association between each hormone and aggression can, however, vary between species (reviewed by Goodson and Bass, 2001; Canoine and Gwinner, 2002). It is not yet clear how the two hormones vary together across species and whether this relationship could be linked to the expression of paternal or other behaviors. For example, in the monogamous prairie vole, there is evidence that AVP can increase paternal behavior and aggression, as described earlier, but castration, which decreases AVP levels (Lonstein and De Vries, 1999; although note that time of the behavioral testing after castration may be important), does not decrease aggression (Demas et al., 1999). Similar patterns occur in the monogamous California mouse: as described earlier, there is some evidence that AVP is associated with both paternal behavior and aggression and, in addition, castration also does not decrease aggression (Trainor and Marler, 2001). However, in California mice, T increases in response to winning an aggressive encounter (T. Oyegbile and C. A. Marler, unpublished data), and castration appears to inhibit the typical increase in aggression that occurs with increasing numbers of aggressive encounters (Trainor and

298


Marler, 2001). Therefore, while baseline levels of T may not influence aggression, transient T increases caused by social experience may have a significant impact on aggression. Within California mice, we are continuing to test whether T is positively associated with aggression and paternal care and have also initiated studies examining the role of progesterone (E. Davis and C. A. Marler, in press). We cannot rule out the possibility that T is less important in influencing aggression in monogamous species with high levels of paternal care compared with other species. For the time being, then, it is interesting to note that AVP can influence resident–intruder aggression in both prairie voles (Young et al., 1997) and California mice (J. K. Bester-Meredith and C. A. Marler, unpublished data), two species that display paternal care; but current evidence suggests that AVP may not influence aggression in either montane voles (Young et al., 1997) or white-footed mice (J. K. Bester-Meredith and C. A. Marler, unpublished data), two species that do not display high levels of paternal care during early development. So far, these comparisons of closely related species reveal that aggression of the more paternal species is characterized by being less influenced by baseline levels of T and perhaps more by variation in AVP, although changes in T in response to social experience may still prove to be important. Currently, however, we propose that these two neurochemicals can act in concert to regulate levels of aggression and paternal care within the same individuals under varying social conditions.

IV. Cross-Generational Transmission of Aggression through Behavioral Mechanisms and the Role of Arginine Vasopressin Up to this point, we have established that AVP and T can be positively associated with both aggression and paternal behavior in some species. As we describe in this section, these characteristics make AVP and T ideal candidates for involvement in the behavioral transmission of aggression and paternal behavior across generations that we have identified in Peromyscus. We describe how a change in paternal retrievals in one generation can be transmitted to influence future levels of paternal retrievals in later generations. Furthermore, we suggest that this plasticity in paternal retrievals is a behavioral mechanism for transmitting resident– intruder aggression across generations in Peromyscus. A positive relationship between paternal retrievals and resident–intruder aggression is an integral component of this cascade of behavioral traits across generations.


299

While there is extensive literature examining the relationship between parental behavior and offspring aggression in humans, there is surprisingly little nonhuman research, despite the limitations inherent in research using humans. We therefore review existing animal studies (rodents and nonhuman primates) and a subset of the human studies. The focus is expanded to encompass different types of parental behaviors that might influence offspring aggression and two types of parental behavior that begin to emerge from some of these studies. Finally, in Section V, we examine whether behavioral traits could pass beyond two generations in Peromyscus and address the role of AVP in this behavioral transmission of paternal and aggressive behaviors across generations. We have not yet extensively examined the role of T in this process, although our studies of AVP indicate that AVP is a likely mechanism for the observed behavioral plasticity. A. Parent–Offspring Interactions: Nongenomic Transmission of Aggression across Generations Thus far, we have focused primarily on hormone–behavior interactions, specifically the relationships between aggressive and paternal behaviors and the endocrine compounds T and AVP. A further complexity can be added to the interactions by focusing on how the two categories of behavior can interact. We have already described how an increase in adult aggression does not necessarily exclude an increase in paternal behaviors. Interactions between these two behavioral categories can also occur between parents and offspring during development. These interactions are potentially important because they provide a nongenomic mechanism for transmission of behaviors across generations. The effect of parental behavior on offspring aggression is likely to vary significantly depending on both the quantity and characteristics of parental care, but may also vary depending on the type of aggression that is measured. In this section, we review interactions between parental behavior and offspring aggression. In the following section we examine how parental behavior of the parents might influence parental behavior of the offspring when they become adults. 1. Parent–Offspring Interactions: Rodents More manipulations have been performed with rodents than with other taxa, although research in this area is still limited, and focuses especially on the effects of father removal and cross-fostering. When fathers are removed early during development in rodents, there is typically a decrease in offspring adult aggression (house mice: Mugford and Nowell, 1972;

300


Wuensch and Cooper, 1981; southern grasshopper mouse [Onychomys torridus]: McCarty and Southwick, 1977; California mouse: Wallace et al., 1998). Unfortunately, the effects of these studies on offspring aggression could be the result of either direct effects of the father on the pups, or indirect effects via changes in the mother. Cross-fostering studies represent a more sophisticated manipulation, but do not directly manipulate single behaviors. In one set of studies, a low-aggression line of house mice (Mus musculus) was cross-fostered with a high-aggression line that also displayed more maternal huddling, grooming, nursing, and retrieving (Mendl and Paul, 1990; Benus and Ro¨ndigs, 1997). Despite the many behavioral differences between these two lines, there was no effect of this manipulation on resident–intruder aggression. It is important to note that the role of the father was not tested in these studies because fathers provided little paternal care in either strain. We have examined correlations between parental behaviors and offspring aggression in Peromyscus studies and found different associations between parental behavior and offspring aggression, depending on the sex of the parent and the type of aggression tested in the offspring. Our crossfostering studies indicated that high maternal HNGI scores (indicating high levels of huddling, nursing, grooming, and time spent in the nest) were associated with lower levels of aggression expressed by male offspring in a neutral arena (J. K. Bester-Meredith and C. A. Marler, unpublished data). Aggression in a neutral arena was not associated with paternal HGI scores (huddling, grooming, and time spent in the nest) or pup retrieval behavior displayed by mothers or fathers. The suggested association between maternal HNGI scores and offspring aggression was somewhat unexpected because, as described above, house mice that are cross-fostered between highly aggressive and less aggressive strains that also differ in maternal care do not show changes in aggression. However, these previous studies concerned house mice, not Peromyscus, and males were tested in resident– intruder aggression tests and not neutral aggression tests (Mendl and Paul, 1990; Benus and Ro¨ndigs, 1996). Therefore, it is possible that an association between HNGI scores and offspring aggression may be revealed in house mice under other testing conditions. In addition to this role for maternal behavior in regulating offspring aggression in the neutral arena, paternal behavior in the form of pup retrievals may serve a critical function in the development of resident– intruder aggression. We found that, in offspring raised by foster parents of the same or another species, the duration of paternal retrievals was positively associated with their aggression as adults in the resident–intruder test (J. K. Bester-Meredith and C. A. Marler, unpublished data). This positive association between male retrievals and offspring aggression in the


301

resident–intruder test seems to be independent of other factors including weight gain during development. In addition, the association between male retrievals and offspring aggressiveness was not driven by a single species or fostering group: all four groups (in-fostered California mice, cross-fostered California mice, in-fostered white-footed mice, cross-fostered white-footed mice) showed a similar pattern of association between retrievals and resident–intruder aggression. Thus some aspect of paternal retrieval behavior may influence pups and cause long-term changes that are revealed through resident–intruder aggression tests as adults. Retrieving behavior and the more ‘‘nurturing’’ parental behaviors are associated differently with aggression in Peromyscus. Retrieving behavior in California mice may be fundamentally different from huddling and grooming. Huddling and grooming typically occur within the nest and can be considered ‘‘nurturing’’ behaviors. In contrast to the retrieving behavior most often observed in female rodents (Felton et al., 1998; Gammie and Nelson, 1999), retrieving behavior in California mice occurs most frequently when pups are old enough to be active and locomote independently and at a time when huddling behavior occurs less frequently (Fig. 3; Bester-Meredith et al., 1999). Retrieving behavior is sometimes preceded by a variety of grabbing and pulling behaviors by the parent, in which pups appear to resist retrieval. When the parent retrieves the pup, it grabs the pup just posterior to the forelegs. When the parent lifts the pup up, the pup is motionless and turned sideways. The parent usually places the pup back in the nest, after which the pup usually remains motionless for a short period of time. In contrast to pup-grooming behavior, castration did not affect pup retrieval behavior (t7 ¼ 0.83, p ¼ 0.44; K. Cravens, B. C. Trainor, and C. A. Marler, unpublished data), suggesting another difference between the two types of behavior. Finally, although huddling, nursing, pup grooming, and time spent in the nest were significantly correlated with each other, retrieval behavior was statistically independent from the other parental behaviors in both males and females (J. K. BesterMeredith and C. A. Marler, unpublished data). Thus, pup retrieval behavior may represent a different style of parental behavior. It is also possible that the different types of parental behavior may be controlled by slightly different mechanisms. The effect of cross-fostering on resident–intruder aggression in California mouse pups is consistent with the proposed positive association between paternal retrievals and resident–intruder aggression. These pups displayed a decrease in resident–intruder aggression when raised by whitefooted mice. To be consistent with the proposed relationship between retrievals and resident–intruder aggression, one would also have predicted an increase in resident–intruder aggression in the white-footed mouse pups

302


raised by the more paternal California mice. There was, however, no change in resident–intruder aggression in these cross-fostered white-footed mice. The answer to this seeming inconsistency may lie in a change in the retrieval behavior of California mouse fathers when raising white-footed mouse pups: the fathers decreased their rate of retrievals to a level similar to those typically displayed by white-footed mouse fathers. The characteristics of the white-footed mouse pups that caused this change in retrievals is unknown, but nonetheless resulted in a change in male California mouse retrieval behavior. Within Peromyscus, we have identified two classes of parental behavior that are associated with two different types of aggression. Higher levels of the more nurturing maternal behaviors composing the HNGI score were associated with a decrease in neutral arena aggression, whereas higher levels of the ‘‘rougher’’ retrieval behavior by fathers was associated with increased resident–intruder aggression. We are currently in the process of manipulating paternal huddling and pup-grooming behaviors by castrating males. Preliminary evidence indicates that retrieval behavior is not influenced by castration (t8 ¼ 0.51, p ¼ 0.62), so we are independently manipulating pup retrievals to examine the effects on resident–intruder aggression (K. Cravens, B. C. Trainor, and C. A. Marler, unpublished data). In rodents, it appears that paternal behavior can have a significant impact on the behavior of male offspring, although further manipulations are needed. As more studies are performed with paternal rodents, we hope to identify more clearly the differential effects of maternal and paternal behavior and to continue to isolate specific paternal behaviors that influence offspring aggression. 2. Parent–Offspring Interactions: Nonhuman Primates Few studies have examined the potential nongenomic effects of parental behavior on offspring aggression toward conspecifics in nonhuman primates. As a first approach, nonhuman primates have been removed from their mothers and raised in different environments, such as with peers present or absent. This manipulation in rhesus macaques, thought to provide a model of maternal neglect, resulted in increased aggression, accompanied by lower serotonin metabolites in a study by Higley et al. (1996b). In another study involving social separation, serotonin metabolite levels were higher in the peer-reared individuals than in mother-reared individuals (Higley et al., 1991). It is difficult to ascertain what component of the altered environment caused the changes in aggression and neurochemicals. However, it is possible that they are due to the presence or absence of the mother. When bonnet macaque (Macaca radiata)


303

mothers were exposed to a more unpredictable feeding regimen, there were lower levels of social interaction of any type between mothers and their offspring. These lower interaction levels were accompanied by decreased aggression and higher levels of serotonin metabolites (Andrews and Rosenblum, 1994; Rosenblum and Andrews, 1994; Coplan et al., 1998). In a third species, the Japanese macaque (Macaca fuscata), neither maternal protectiveness nor maternal rejection was correlated with intensity or frequency of aggressive interactions (Schino et al., 2001). The variability in the results obtained with these nonhuman primate models of maternal neglect suggests that the relationship between maternal behavior and offspring aggression needs to be further investigated. It is not clear whether the variation in results reflects species differences, parental effects, social interactions eliciting aggression, or other factors such as nutrition. There is also a need to examine whether offspring aggression is correlated with the different maternal styles of parenting that have been identified (e.g., Altmann, 1980; Fairbanks, 1996). 3. Parent–Offspring Interactions: Humans The most extensive research on associations between parental and offspring behavior is found in humans, with the strongest emphasis on maternal behavior. There are some fairly consistent findings regarding the relationship between both child abuse and neglect with offspring aggression. Both child abuse and neglect are generally found to be associated with higher levels of aggression (reviewed by Widom, 2000). A number of researchers have reported a positive association between aggressive behavior in children and adults, and parental behaviors such as physical abuse, harsh discipline, restrictive discipline, and lack of warmth/ rejection (e.g., Pettit and Bates, 1989; Weiss et al., 1992; Travillion and Snyder, 1993; Dodge et al., 1995; Scerbo and Kolko, 1995; Pettit et al., 1996; Raine et al., 1997; Schwartz et al., 1997; Widom, 2000; Barnow et al., 2001). In comparison, a more responsive or ‘‘warm’’ style of mothering can be associated with lower levels of aggression/misbehavior toward peers in boys (e.g., Chen and Rubin, 1994; Mize and Pettit, 1997). In the studies above, maternal and paternal effects appear similar in that warmth/ affection and positive involvement with the child are negatively associated with misbehavior (including aggression). Studies have, however, revealed differences in the associations for fathers and mothers. For example, fathers, who responded to their children’s negative affect, such as anger or pouting and whining during a physical play paradigm, with a similar negative affect response of their own, were more likely to have children that were physically and verbally aggressive toward peers, but this was not true of mothers (Carson and Parke, 1996). Other differences in

304


associations between the parenting styles of mothers and fathers and offspring behavior have been found, although not related to aggression (Kahen et al., 1994). Overall, negative associations are typically found between aggression and positive affect, whereas positive associations are found between aggression and harsh or restrictive discipline. B. Summary of Parent–Offspring Interactions The potential nongenomic effects of parental behavior on offspring aggression toward conspecifics have been the focus of considerable attention in humans and to a lesser extent in nonhuman primates and rodents. Thus far, it appears that Peromyscus may provide a rodent model system for examining parent and offspring interactions that are similar to humans. Unlike many other mammalian species but like humans, male California mice provide extensive paternal care toward offspring that may make critical contributions to the development of normal patterns of adult social behavior. Our Peromyscus studies suggest that, while human studies sometimes focus more on extreme parental behaviors such as child abuse or harsh discipline, more subtle parental behaviors may also influence offspring aggression. Retrievals, in particular, appear to represent a rougher type of parental care that may cause adaptive changes in aggression, and there may be equivalent behaviors with potent effects in humans that are less extreme than overt physical abuse. The advantage of this animal model system is that we can use an experimental approach to both behavior and the associated endocrine mechanisms. In Peromyscus, the rearing environment and neurochemicals, including AVP and T, can be manipulated under controlled conditions to pinpoint how specific factors influence offspring development. In both humans and Peromyscus, nurturing parental behaviors may produce effects on offspring aggression that are different from those produced by rougher forms of parental behavior. Manipulations in Peromyscus of parental behavior and other relevant variables, including AVP and T, may provide valuable insights into how human behavior is shaped by the conditions under which a child is raised.

V. Role of Plasticity in Paternal Behavior and Arginine Vasopressin in the Nongenomic Transmission of Aggression across Multiple Generations in Peromyscus We have described how Peromyscus fathers can potentially influence the aggressive behavior of their offspring, with a negative association between


305

paternal retrievals and resident–intruder aggression in the offspring. We have not yet asked, however, whether paternal behavior of the parents influences paternal behavior of their adult offspring. For aggression to pass on to future generations via nongenomic mechanisms (by behavioral processes), offspring displaying a change in resident–intruder aggression as adults must also alter their interactions with their own pups, that is, there is likely to be plasticity in their parental behavior. For example, crossfostering from the less paternal meadow vole to the more paternal prairie vole increased paternal huddling behavior with pups (McGuire, 1988), although no changes in time spent in the nest occurred when white-footed mice and the more parental deer mice (P. maniculatus) were cross-fostered (Hawkins and Cranford, 1992). In this section we discuss evidence in Peromyscus that specific paternal traits may link generations through nongenomic effects on behavior. An additional important question that we will be addressing concerns the physiological bases for the maintenance of these interactions across generations. Evidence for the effect of maternal traits on physiological traits of successive generations of offspring is more extensive than for paternal traits. Studies with female rats have found suites of neurochemical and behavioral traits linking generations through maternal effects (e.g., Francis et al., 1999; Gonzalez et al., 2001; Lovic et al., 2001). In an interesting parallel to our studies, higher amounts of maternal licking/ grooming behaviors alter the maternal behaviors of future generations (Caldji et al., 1998; Francis et al., 1999; Liu et al., 1997; Boccia and Pederson, 2001). The hypothalamic–pituitary–adrenal (HPA) axis is a critical component in these cross-generational effects because individuals exposed to a higher level of maternal licking/grooming behavior are less fearful and have weaker HPA responses to stress (Liu et al., 1997; Caldji et al., 1998). Boccia and Pederson (2001) have also proposed plasticity in the density of oxytocin receptors as a potential mechanism. It will be valuable to examine whether huddling and grooming behaviors could also result in a decrease in neutral arena and/or resident–intruder aggression. In this section we discuss how a different physiological mechanism, AVP, may underlie the transfer of resident–intruder aggression and paternal behavior across generations in Peromyscus mice. In California mice, paternal behavior seems to produce nongenomic effects on the behavior of future generations. We investigated whether cross-fostered California mice displayed any changes in paternal behavior in addition to the previously described aggressive behaviors. As adults, these male California mice raised by the less parental white-footed mice parents displayed a decrease in pup retrievals, although we found no changes in nest building or a composite score of time spent huddling,

306


Fig. 9. Average time spent retrieving pups by male in-fostered and cross-fostered whitefooted (P. leucopus) and California mice (P. californicus) (modified from J. K. BesterMeredith and C. A. Marler, 2003). Data are presented as adjusted means standard errors. *p < 0.05. n ¼ 9 per group for cross-fostered California mice, in-fostered California mice, and cross-fostered white-footed mice; n ¼ 4 for in-fostered white-footed mice.

grooming, and in the nest (Fig. 9; J. K. Bester-Meredith and C. A. Marler, unpublished data). Therefore, paternal retrievals appear to be especially important for transferring both paternal behavior and aggression traits across generations. Maternal retrievals may also prove to be important in influencing aggression, but we have yet to find a consistent effect (J. K. BesterMeredith and C. A. Marler, unpublished data). Within house mice, a relationship between high levels of maternal aggression and maternal pup retrieval behavior has been found within a single generation, indicating that the two could be linked (Meek et al., 2001), as they both serve to protect offspring. Our results provide new evidence that, at least in a highly biparental species, mothers may not be the only individuals that have nongenomic effects on the behavioral traits of their offspring. It appears that fathers play a larger role in offspring aggression than previously thought. The link between retrievals and aggression may prove to have interesting parallels with the positive relationship between harsher discipline/physical abuse and aggression described for humans. In rodents, the effect of natural variation in paternal retrievals on aggression may result in adaptive changes in aggression. It is also possible that this parallels the ‘‘less warm’’ parenting styles described earlier for humans. There are numerous potential mechanisms through which aggression may be transmitted nongenetically from parents to offspring, but all, alone or together, must satisfy three critical components for a mechanism or


307

mechanisms involved in this transfer: the ability to influence paternal behavior, the ability to influence aggression, and, last, the potential for plasticity in response to the social environment. Currently, we have no evidence that T is involved in these cross-generational effects. Radioimmunoassay of plasma samples collected from California mice and whitefooted mice that were not exposed to aggressive or sexual encounters revealed no differences in T levels between cross-fostered and control animals (Bester-Meredith and Marler, 2001). However, variation in these adult T levels may have varied depending on other factors such as social experience. The absence of differences between cross-fostered and control animals in adult T levels also does not exclude the possibility that T levels could have been critical during development in shaping future aggressiveness. Because T can significantly influence paternal huddling and pupgrooming behaviors (Trainor and Marler, 2001), and we have evidence that these paternal behaviors are associated with variation in offspring neutral aggression, it is worth studying further the possibility that T is involved in some cross-generational effects. Arginine vasopressin currently appears to be the best candidate for satisfying the requirements mentioned above. As described earlier, AVP is known to be positively associated with aggression and paternal care in rodents and is plastic in response to changing social conditions. In California mice, AVP was associated with variation in paternal care and was positively associated with resident–intruder aggression. In addition, also as described earlier, exposure to variation in paternal care also altered both AVP in the bed nucleus of the stria terminalis and resident–intruder aggression. We propose that AVP provides a link between aggression and paternal behavior so that these behaviors can be changed in concert and also be plastic in response to social interactions. Thus, AVP is currently a primary candidate for mediating this transfer of resident–intruder aggression and paternal care across generations.

VI. Summary of Nongenomic Transmission of Aggression and Paternal Behavior across Generations and the Role of Arginine Vasopressin The last two sections suggest that both paternal and maternal effects may play a role in the nongenomic transfer of behavior. Evidence is building that parents can have a significant impact on the social behaviors of their offspring. Paternal effects may be particularly important in highly paternal species and may also play a significant role in the transfer of resident–intruder aggression across generations. As suggested by the data

308


in Fig. 9, paternal effects can potentially influence several future generations given the dramatic effect of cross-fostering on the subsequent retrieval behavior of offspring. In rodents and other species there may be a greater degree of responsiveness to the social environment during development than previously believed. Arginine vasopressin is a likely candidate for part of the mechanism underlying this transfer of behaviors between generations because it has been positively linked with both aggression and paternal behavior and appears to be plastic in response to the social environment. The studies described in the last two sections provide additional support for a positive association between paternal behavior and resident–intruder aggression, as found in some of the T studies described earlier. The behavioral and neurochemical links further suggest that aggression and paternal care may interact across several generations.

VII. Conclusions Our review reveals a number of important points about paternal behavior with respect to both behavior and physiology. First, there is a tendency to overestimate the negative effect of male aggression on the ability to provide paternal care and the negative relationship between T and paternal behavior. While it is often assumed that high levels of male aggression are incompatible with paternal behavior, we found several lines of evidence suggesting otherwise, including the research on the California mouse. There is considerable plasticity in the relationship between aggression and paternal behavior across species and, as in maternal aggression and maternal behavior, high levels of aggression can be expressed while individuals are also caring for young. It appears that in some species (and sexes), mating behavior, parental behavior, and aggression may coincide, in response to a variety of ecological pressures. For example, in some rodent species postpartum estrus may allow females to produce a larger number of litters during their lifetimes and allow these females to be pregnant with one litter while still nursing another litter. If a male of such a species provides parental care and defends a territory, as in the case of the California mouse and some fish species, then males must be able to express both aggression and paternal behavior in the same general time frame. This behavioral overlap could also occur in species in which there is a shorter breeding season and the behaviors need to be expressed in a shorter time frame, such as in Arctic species. This contrasts with the majority of avian studies that have involved species with longer breeding seasons. In these species, aggression and courtship may be temporally


309

separated from the expression of paternal behavior. Thus there may be environmental and social conditions that change the probability that aggression and paternal behavior will overlap and be positively associated. Consistent with this plasticity in behavioral relationships, the relationship between T and paternal behavior can also vary across species. While there are several bird studies indicating a negative relationship between T and paternal care, there are also studies across taxa suggesting that T can also be positively associated with paternal behavior. In the example described above with Arctic avian species, where paternal care, aggression, and courtship may be more temporally compressed, these species must be capable of showing paternal behaviors in spite of high T levels. In the case of the California mice and some fish species among which the behaviors also overlap, T manipulations have demonstrated a positive association between T and paternal care. As more manipulative studies are conducted across species, this positive relationship between T and paternal care may prove to be important in more species, depending on a variety of costs and benefits of higher T levels that vary from species to species. However, as in the case of the behavioral associations, there may be general conditions that increase the probability that high T levels are compatible with paternal behavior, as well as cases where increased T can actually increase levels of paternal care in species with higher levels of temporal overlap between aggression, paternal care and courtship behavior. We also described examples where T levels decreased, but nonetheless there was a positive causal relationship between T and paternal care in some species of fish and mammals. We have evidence that one mechanism that may allow this to occur, at least in the California mouse, is the conversion of T to estradiol via the enzyme aromatase (Trainor and Marler, 2002). Thus, plasticity in the relationship between T and paternal behavior could be partially controlled by levels of the aromatase enzyme in specific brain areas associated with paternal care. A drop in T during paternal care may therefore still be compatible with a positive relationship between T and paternal behavior because of changing levels of aromatase. The positive relationship between AVP and paternal behavior is also potentially compatible with this scenario. Aromatase may modulate effects of T on AVP, as T could increase AVP levels via conversion of T to estrogen (De Vries et al., 1994). A second important point revealed by this review is that behavioral variation across species could also be influenced by the responsiveness or plasticity of the species to the social environment provided by the parents (and likely other individuals in the environment), and that includes both

310


maternal and paternal effects. The nongenomic effects across generations may be important only for some mammalian species. We found that crossfostering between California mice and white-footed mice changed only resident–intruder aggression in California mice and only aggression in a neutral arena in white-footed mice. Further studies will be necessary to determine whether white-footed mice failed to display changes in resident– intruder aggression because they were not exposed to significant variation in paternal retrieval behavior or whether they are less responsive to variation in paternal behavior. Because male white-footed mice typically display less paternal care early during development, it is possible that offspring are less responsive to variation in paternal care. A point not mentioned earlier is that, while an AVP blocker significantly decreased resident–intruder aggression in California mice, it had no effect on resident–intruder aggression in white-footed mice (J. K. BesterMeredith, P. Martin, and C. A. Marler, unpublished data). Thus, there may be variation across species regarding plasticity in offspring aggression in response to paternal behavior, as well as in the physiological mechanisms underlying the plasticity in behavior. There are so few studies examining how parental care in general can influence offspring aggression through nongenomic mechanisms that it is difficult to draw generalities from the studies presented. Nonetheless, studies in these Peromyscus species provide a starting point for examining how paternal behavior can influence aggression and paternal behavior of offspring. The power of the effect of paternal behavior on the aggression and paternal behavior of offspring may have been significantly underestimated. We have evidence that this effect could potentially continue on to future generations. It is also worth noting that the type of paternal behavior studies we have described may be crucial for revealing paternal effects on the aggression of offspring. Retrievals are more rarely expressed than other parental behaviors, but may have a significant impact on resident–intruder aggression. In contrast, other paternal behaviors such as huddling and grooming may act to reduce aggression, although the type of aggression may differ. Studies into the physiological bases of these different types of paternal behavior (as well as different types of aggression) may help to reveal potential differences in the functions of these behaviors. This review, combined with others such as that of Goodson and Bass (2001), continues to support the idea that AVP can be plastic in response to social conditions, both in developing individuals and in adults. Variation in certain types of aggression, such as resident–intruder aggression, may be strongly influenced by AVP. Plasticity in T levels during development could also contribute to variation seen in aggressive behavior, but more studies need to be performed in which the roles of the two hormones are


311

compared. It may be that AVP and T perform in concert to influence plasticity in aggressive behavior. Finally, this review is a reminder of the plasticity in the relationship between behavior and hormonal mechanisms both within and across species, as discussed at the beginning of the chapter. Both the AVP and T studies provided examples of the dynamic variation that can exist in these relationships. The AVP and behavior studies demonstrated that the social environment can modify individual hormone levels and this can, in turn, modify that individual’s behavior. The degree of plasticity for specific behaviors can vary even among closely related species, perhaps as a reflection of their behavioral ecology, including the degree to which they display territorial behavior and the timing of paternal behavior. Hormones may have evolved, in part, to link together certain behaviors in a variety of ways, influenced by such traits as the timing of different social behaviors. Our studies support the generality of concepts that relate the evolution of plasticity in behavior to hormone–behavior relationships across species.

VIII. Summary Aggression and paternal behavior can be linked in a variety of ways, and the relationship between hormones and these social behaviors may be equally variable. We have illustrated how high levels of aggression can be compatible with high levels of paternal care. Under such conditions, T may be positively associated with paternal care and aggression, even in species in which a decrease in T occurs with the onset of paternal care. Similar to T, AVP also can be positively associated with both aggression and paternal care. Individual variation in paternal care and aggression may be mediated by variation in AVP and T levels and receptors. This physiological variation could in turn be important for survival of offspring and also for shaping variation in paternal behavior and aggression of those offspring. Behavioral and endocrine changes may be passed on to multiple generations. The degree of plasticity in these relationships remains to be elucidated, but our results suggest that variation in AVP and T may be important for altering paternal and aggressive behaviors in response to the social environment. Acknowledgments Many thanks to those individuals who read and commented on the manuscript, including L. Berkowitz, E. Davis, P. Marler, and C. Snowdon. We also thank P. Martin and K. Rouse for hours spent photocopying, helping with figures, and organizing references, J. P. Crossland for providing Peromyscus blood samples, and C. J. Clark, C. J. Cravens, and D. Wittwer for

312


technical help. Bill Feeny kindly provided us with the drawings in Fig. 2. This research was supported by NRSA Predoctoral Fellowship F31 MH12287 and a Sigma XI Grant-in-Aid Fellowship to J. K. B., NIH NRSA F31 MH64328-01 and a Sigma Xi Grant-in-Aid to B. C. T., and NSF IBN-9703309 and IBN-0110625 to C. A. M. This is publication number 42-001 of the WRPRC.

References Albers, H. E., Liou, S. Y., and Ferris, C. F. (1988). Testosterone alters the behavioral response of the medial preoptic–anterior hypothalamus to microinjection of arginine vasopressin in the hamster. Brain Res. 456, 382–386. Altmann, J. (1980). ‘‘Baboon Mothers and Infants.’’ University of Chicago Press, Chicago, IL. Andrews, M. W., and Rosenblum, L. A. (1994). The development of affiliative and agonistic social patterns in differentially reared monkeys. Child. Dev. 65, 1398–1404. Archer, J. (1988). ‘‘The Behavioral Biology of Aggression.’’ Cambridge University Press, Cambridge. Avise, J. C., Smith, M. H., and Selander, R. K. (1974). Biochemical polymorphism and systematics in the genus Peromyscus. VI. The boylii species group. J. Mammal. 55, 751–763. Baker, R. H. (1968). Habitats and distribution. In ‘‘Biology of Peromyscus (Rodentia)’’ (J. A. King, Ed.), pp. 98–126. American Society of Mammalogists, Stillwater, OK. Ball, G. F. (1992). Endocrine mechanisms and the evolution of avian parental care. In ‘‘Proceeding of the XXth International Ornithology Congress,’’ pp. 984–991. Bamshad, M., Novak, M. A., and De Vries, G. J. (1993). Sex and species differences in the vasopressin innervation of sexually naı¨ve and parental prairie voles, Microtus ochrogaster and meadow voles, Microtus pennsylvanicus. J. Neuroendocrinol. 5, 247–255. Bamshad, M., Novak, M. A., and De Vries, G. J. (1994). Cohabitation alters vasopressin innervation and paternal behavior in prairie voles (Microtus ochrogaster). Physiol. Behav. 56, 751–758. Barnow, S., Lucht, M., and Freyberger, H. J. (2001). Influence of punishment, emotional rejection, child abuse, and broken home on aggression in adolescence: An examination of aggressive adolescents in Germany. Psychopathology 34, 167–173. Beletsky, L. D., Gori, D. F., Freeman, S., and Wingfield, J. C. (1995). Testosterone and polygyny in birds. In ‘‘Current Ornithology’’ (D. M. Power, Ed.), Vol. 12, pp. 1–41. Plenum Press, New York. Benus, R. F., and Ro¨ndigs, M. (1996). Patterns of maternal effort in mouse lines bidirectionally selected for aggression. Anim. Behav. 51, 67–75. Benus, R. F., and Ro¨ndigs, M. (1997). The influence of the postnatal maternal environment in accounting for differences in aggression and behavioural strategies in Mus domesticus. Behaviour 134, 623–641. Berg, S. J., and Wynne-Edwards, K. E. (2001). Changes in testosterone, cortisol, and estradiol levels in men becoming fathers. Mayo Clin. Proc. 76, 582–592. Bester-Meredith, J. K., and Marler, C. A. (2001). Vasopressin and aggression in cross-fostered California mice (Peromyscus californicus) and white-footed mice (Peromyscus leucopus). Horm. Behav. 40, 51–64. Bester-Meredith, J. K., and Marler, C. A. (2003). Vasopressin and the transmission of paternal behavior across generations in mated cross-fostered Peromyscus mice. Behav. Neurosci. (In press.)


313

Bester-Meredith, J. K., Young, L. J., and Marler, C. A. (1999). Species differences in paternal behavior and aggression in Peromyscus and their associations with vasopressin immunoreactivity and receptors. Horm. Behav. 36, 25–38. Birdsall, D. A., and Nash, D. (1973). Occurrence of multiple insemination of females in natural populations of deer mice (Peromyscus maniculatus). Evolution 27, 106–110. Bluthe, R. M., Gheusi, G., and Dantzer, R. (1993). Gonadal steroids influence the involvement of arginine vasopressin in social recognition in mice. Psychoneuroendocrinology 18, 323–335. Boccia, M. L., and Pederson, C. A. (2001). Brief vs. long maternal separations in infancy: Contrasting relationships with adult maternal behavior and lactation levels of aggression and anxiety. Psychoneuroendocrinology 26, 657–672. Borg, B. (1987). Stimulation of reproductive behaviour by aromatizable and nonaromatizable androgens in the male three-spined stickleback, Gasterosteus aculeatus L. In ‘‘Proceedings of the 5th Congress of European Ichthyologists’’ (S. O. K. Kullander and B. Fernholm, Eds.), pp. 269–271. Swedish Museum of Natural History, Stockholm, Sweden. Borg, B., Schoonen, W. G. E. J., and Lambert, J. G. D. (1989). Steroid metabolism in the testes of the breeding and nonbreeding three-spined stickleback, Gasterosteus aculeatus. Gen. Comp. Endocrinol. 73, 40–45. Boyd, S. K. (1997). Brain vasotocin pathways and the control of sexual behaviors in the bullfrog. Brain Res. Bull. 44, 345–350. Brown, R. E. (1985). Hormones and paternal behavior in vertebrates. Am. Zool. 25, 895–910. Brown, R. E., Murdoch, T., Murphy, P. R., and Mojer, W. H. (1995). Hormonal responses of male gerbils to stimuli from their mate and pups. Horm. Behav. 29, 474–491. Buntin, J. D. (1996). Neural and hormonal control of parental behavior in birds. In ‘‘Advances in the Study of Behavior’’ (C. T. Snowdon and J. S. Rosenblatt, Eds.), Vol. 25, pp. 161–213. Academic Press, San Diego, CA. Burmeister, S. S., and Wilczynski, W. (2001). Social context influences androgenic effects on calling in the green treefrog (Hyla cinerea). Horm. Behav. 40, 550–558. Caldji, C., Tannenbaum, B., Sharma, S., Francis, D., Plotsky, P. M., and Meaney, M. J. (1998). Maternal care during infancy regulates the development of neural systems mediating the expression of fearfulness in the rat. Proc. Natl. Acad. Sci. USA 95, 5335–5340. Caldwell, G. S., Glickman, S. E., and Smith, E. R. (1984). Seasonal aggression independent of seasonal testosterone in wood rats. Proc. Natl. Acad. Sci. USA 81, 5255–5257. Camazine, B., Garstka, W., Tokarz, R., and Crews, D. (1980). Effects of castration and androgen replacement on male courtship behavior in the red-sided garter snake (Thamnophis sirtalis parietalis). Horm. Behav. 14, 358–372. Canoine, V., and Gwinner, E. (2002). Seasonal differences in the hormonal control of territorial aggression in free-living European stonechats. Horm. Behav. 41, 1–8. Cantoni, D., and Brown, R. E. (1997). Paternal investment and reproductive success in the California mouse, Peromyscus californicus. Anim. Behav. 54, 377–386. Carson, J. L., and Parke, R. D. (1996). Reciprocal negative affect in parent–child interactions and children’s peer competency. Child Dev. 67, 2217–2226. Chen, X., and Rubin, K. H. (1994). Family conditions, parental acceptance, and social competence and aggression in Chinese children. Soc. Dev. 3, 269–290. Cho, M. M., De Vries, A. C., Williams, J. R., and Carter, C. S. (1999). The effects of oxytocin and vasopressin on partner preferences in male and female prairie voles (Microtus ochrogaster). Behav. Neurosci. 113, 1071–1079. Clark, M. M., and Galef, B. G., Jr. (1999). A testosterone-mediated trade-off between parental and sexual effort in male Mongolian gerbils (Meriones unguiculatus). J. Comp. Psychol. 113, 388–395.

314


Clark, M. M., and Galef, B. G., Jr. (2000). Effects of experience on the parental responses of male Mongolian gerbils. Dev. Psychobiol. 36, 177–185. Clemens, L. G., Wee, B. E. F., Weaver, D. R., Roy, E. J., Goldman, B. D., and Rakerd, B. (1988). Retention of masculine sexual behavior following castration in male B6D2F1 mice. Physiol. Behav. 42, 69–76. Coccaro, E. F., Kavoussi, R. J., Hauger, R. L., Cooper, T. B., and Ferris, C. F. (1998). Cerebrospinal fluid vasopressin levels: Correlates with aggression and serotonin function in personality-disordered subjects. Arch. Gen. Psychiatry 55, 708–714. Collias, N. E. (1950). Hormones and behavior with special reference to birds and the mechanisms of hormone action. In ‘‘A Symposium on Steroid Hormones’’ (E. S. Gordon, Ed.), pp. 277–329. University of Wisconsin Press (Business Press), Lancaster, PA. Colvin, D. V. (1973). Agonistic behaviour in males of five species of voles Microtus. Anim. Behav. 21, 471–480. Compaan, J. C., Buijs, R. M., Pool, C. W., de Ruiter, A. J. H., and Koolhaas, J. M. (1993). Differential lateral septal vasopressin innervation in aggressive and nonaggressive male mice. Brain Res. Bull. 30, 1–6. Coplan, J. D., Trost, R. C., Owens, M. J., Cooper, T. B., Gorman, J. M., Nemeroff, C. B., and Rosenblum, L. A. (1998). Cerebrospinal fluid concentrations of somatostatin and biogenic amines in grown primates reared by mothers exposed to manipulated foraging conditions. Arch. Gen. Psychiatry 55, 473–477. Crews, D. (1983). Diversity of hormone-behavior controlling mechanisms. BioScience. 33, 545–566. Crews, D. (1984). Gamete production, sex hormone secretion, and mating behavior uncoupled. Horm. Behav. 18, 22–28. Crews, D., and Moore, M. C. (1986). Evolution of mechanisms controlling mating behavior. Science 231, 121–125. Crews, D., Camazine, B., Diamond, M., Mason, R., Tokarz, R. R., and Garstka, W. R. (1984). Hormonal independence of courtship behavior in the male garter snake (Thamnophis sirtalis parietalis). Horm. Behav. 18, 29–41. Davis, E. S., and Marler, C. A. (2003). Progesterone challenge in female Pero myscus mice. Horm. Behav. (In press.) Delville, Y., Mansour, K. M., and Ferris, C. F. (1996). Testosterone facilitates aggression by modulating vasopressin receptors in the hypothalamus. Physiol. Behav. 60, 25–29. Delville, Y., Melloni, R. H., Jr., and Ferris, C. F. (1998). Behavioral and neurobiological consequences of social subjugation during puberty in golden hamsters. J. Neurosci. 18, 2667–2672. Delville, Y., De Vries, G. J., and Ferris, C. F. (2000). Neural connections of the anterior hypothalamus and agonistic behavior in golden hamsters. Brain Behav. Evol. 55, 53–76. Demas, G. E., and Nelson, R. J. (1998). Short-day enhancement of immune function is independent of steroid hormones in deer mice (Peromyscus maniculatus). J. Comp. Physiol. B 168, 419–426. Demas, G. E., Moffatt, C. A., Drazen, D. L., and Nelson, R. J. (1999). Castration does not inhibit aggressive behavior in adult male prairie voles (Microtus ochrogaster). Physiol. Behav. 66, 59–62. De Ridder, E., Pinxten, R., and Eens, M. (2000). Experimental evidence of a testosteroneinduced shift from paternal to mating behaviour in a facultatively polygynous songbird. Behavi. Ecol. Sociobiol. 49, 24–30. De Vries, G. J., and Boyle, P. A. (1998). Double duty for sex differences in the brain. Behav. Brain Res. 92, 205–213.


315

De Vries, G. J., and Miller, M. A. (1998). Anatomy and function of extrahypothalamic vasopressin systems in the brain. Prog. Brain Res. 119, 3–20. De Vries, G. J., Wang, Z., Bullock, N. A., and Numan, S. (1994). Sex differences in the effects of testosterone and its metabolites on vasopressin messenger RNA levels in the bed nucleus of the stria terminalis of rats. J. Neurosci. 14, 1789–1794. Dewsbury, D. A. (1983). A comparative study of rodent social behavior in a seminatural enclosure. Aggress. Behav. 9, 207–215. Dixson, A. F., and George, L. (1982). Prolactin and parental behaviour in a male new world primate. Nature 299, 551–553. Dodge, K. A., Pettit, G. S., Bates, J. E., and Valente, E. (1995). Social information-processing patterns partially mediate the effect of early physical abuse on later conduct problems. J. Abnorm. Psychol. 104, 632–643. Dudley, D. (1974a). Contributions of paternal care to the growth and development of the young in Peromyscus californicus. Behav. Biol. 11, 155–166. Dudley, D. (1974b). Paternal behavior in the California mouse, Peromyscus californicus. Behav. Biol. 11, 247–252. Emerson, S. B., and Hess, D. L. (1996). The role of androgens in opportunistic breeding, tropical frogs. Gen. Comp. Endocrinol. 103, 220–230. Emerson, S. B., Rowsemitt, C. N., and Hess, D. L. (1993). Androgen levels in a Bornean voiceless frog, Rana blythii. Can. J. Zool. 71, 196–203. Everts, H. G. J., De Ruiter, A. J. H., and Koolhaas, J. M. (1997). Differential lateral septal vasopressin in wild-type rats: Correlation with aggression. Horm. Behav. 31, 136–144. Fairbanks, L. A. (1996). Individual differences in maternal style: Causes and consequences for mothers and offspring. In ‘‘Advances in the Study of Behavior,’’ Vol. 25 (C. T. Snowdon and J. S. Rosenblatt, Eds.), pp. 579–611. Academic Press, San Diego, CA. Felton, T. M., Linton, L., Rosenblatt, J. S., and Morrell, J. I. (1998). Intact neurons of the lateral habenular nucleus are necessary for the nonhormonal, pup-mediated display of maternal behavior in sensitized virgin female rats. Behav. Neurosci. 112, 1458–1465. Ferris, C. (1992). Role of vasopressin in aggressive and dominant/subordinate behaviors. Ann. N. Y. Acad. Sci. 652, 212–226. Ferris, C. F., and Delville, Y. (1994). Vasopressin and serotonin interactions in the control of agonistic behavior. Psychoneuroendocrinology 19, 593–601. Ferris, C. F., and Potegal, M. (1988). Vasopressin receptor blockade in the anterior hypothalamus suppresses aggression in hamsters. Physiol. Behav. 44, 235–239. Ferris, C. F., Albers, H. E., Wesolowski, S. M., Goldman, B. D., and Luman, S. E. (1984). Vasopressin injected into the hypothalamus triggers a stereotypic behavior in golden hamsters. Science 224, 521–523. Ferris, C. F., Meenan, D. M., Axelson, J. F., and Albers, H. E. (1986). A vasopressin antagonist can reverse dominant/subordinate behavior in hamsters. Physiol. Behav. 38, 135–138. Ferris, C. F., Singer, E. A., Meenan, D. M., and Albers, H. E. (1988). Inhibition of vasopressin-stimulated flank marking behavior by V1-receptor antagonists. Eur. J. Pharmacol. 154, 153–159. Ferris, C. F., Axelson, J. F., Martin, A. M., and Roberge, L. F. (1989). Vasopressin immunoreactivity in the anterior hypothalamus is altered during the establishment of dominant/subordinate relationships between hamsters. Neuroscience 29, 675–683. Ferris, C. F., Delville, Y., Grozonka, Z., Luber-Narod, J., and Insel, T. R. (1993). An iodinated vasopressin (V1) antagonist blocks flank marking and selectively labels neural binding sites in golden hamsters. Physiol. Behav. 54, 737–747.

316


Ferris, C. F., Melloni, R. H., Koppel, G., Perry, K. W., Fuller, R. W., and Delville, Y. (1997). Vasopressin/serotonin interactions in the anterior hypothalamus control aggressive behavior in golden hamsters. J. Neurosci. 17, 4331–4340. Foltz, D. W. (1981). Genetic evidence for long-term monogamy in a small rodent, Peromyscus polionotus. Am. Nat. 117, 665–675. Francis, D., Diorio, J., Liu, D., and Meaney, M. J. (1999). Nongenomic transmission across generations of maternal behavior and stress responses in the rat. Science 286, 1155–1158. Gammie, S. C., and Nelson, R. J. (1999). Maternal aggression is reduced in neuronal nitric oxide synthase-deficient mice. J. Neurosci. 19, 8027–8035. Garland, M., and Svare, B. (1988). Suckling stimulation modulates the maintenance of postpartum aggression in mice. Physiol. Behav. 44, 301–305. Garstka, W. R., Camazine, B., and Crews, D. (1982). Interaction of behavior and physiology during the annual reproductive cycle of the red-sided garter snake (Thamnophis sirtalis parietalis). Herpetology 38, 104–123. Getz, L. L. (1962). Aggressive behavior in meadow and prairie voles. J. Mammal. 43, 351–358. Getz, L. L., Carter, C. S., and Gavish, L. (1981). The mating system of the prairie vole, Microtus ochrogaster: Field and laboratory evidence for pair-bonding. Behav. Ecol. Sociobiol. 8, 189–194. Getz, L. L., McGuire, B., Hofmann, J., Pizzuto, T., and Frase, B. (1990). Social organization and mating system of the prairie vole, Microtus ochrogaster. In ‘‘Social Systems and Population Cycles in Voles’’ (R. H. Tamarin, R. S. Ostfeld, S. R. Pugh, and G. Bujalska, Eds.), pp. 69–80. Birkhauser Verlag, Basel, Switzerland. Gleason, P. E., Michael, S. D., and Christian, J. J. (1981). Prolactin-induced aggression in female Peromyscus leucopus. Behav. Neural Biol. 33, 243–248. Gonzalez, A., Lovic, V., Ward, G. R., Wainwright, P. E., and Fleming, A. S. (2001). Intergenerational effects of complete maternal deprivation and replacement stimulation on maternal behavior and emotionality in female rats. Dev. Psychobiol. 38, 11–32. Goodson, J. L., and Bass, A. H. (2001). Social behavior functions and related anatomical characteristics of vasotocin/vasopressin systems in vertebrates. Brain Res. Brain Res. Rev. 35, 246–265. Gross, M. R. (1984). Sunfish, salmon, and the evolution of alternative reproductive strategies and tactics in fishes. In ‘‘Fish Reproduction: Strategies and Tactics’’ (G. Potts and R. Wootton, Eds.), pp. 55–75. Academic Press, New York. Gross, M. R., and MacMillan, A. M. (1981). Predation and the evolution of colonial nesting in bluegill sunfish (Lepomis macrochirus). Behav. Ecol. Sociobiol. 8, 163–174. Gubernick, D. J. (1988). Reproduction in the California mouse, Peromyscus californicus. J. Mammal. 69, 857–860. Gubernick, D. J., and Alberts, J. R. (1987a). ‘‘Resource’’ exchange in the biparental California mouse (Peromyscus californicus): Water transfer from pups to parents. J. Comp. Psychol. 101, 328–334. Gubernick, D. J., and Alberts, J. R. (1987b). The biparental care system of the California mouse, Peromyscus californicus. J. Comp. Psychol. 101, 169–177. Gubernick, D. J., and Nelson, R. J. (1989). Prolactin and paternal behavior in the biparental California mouse, Peromyscus californicus. Horm. Behav. 23, 203–210. Gubernick, D. J., and Teferi, T. (2000). Adaptive significance of male parental care in a monogamous mammal. Proc. R. Soc. Lond. B Biol. Sci. 267, 147–150. Gubernick, D. J., Wright, S. L., and Brown, R. E. (1993). The significance of father’s presence for offspring survival in the monogamous California mouse, Peromyscus californicus. Anim. Behav. 46, 539–546.


317

Hau, M., Wikelski, M., Soma, K. K., and Wingfield, J. C. (2000). Testosterone and year-round territorial aggression in a tropical bird. Gen. Comp. Endocrinol. 117, 20–33. Hawkins, L. K., and Cranford, J. A. (1992). Long-term effects of intraspecific and interspecific cross-fostering on two species of Peromyscus. J. Mammal. 73, 802–807. Hegner, R. E., and Wingfield, J. C. (1987). Effects of experimental manipulation of testosterone levels on parental investment and breeding success in male house sparrows. Auk 104, 462–469. Higley, J. D., Suomi, S. J., and Linnoila, M. (1991). CSF monoamine metabolite concentrations vary according to age, rearing, and sex, and are influenced by the stressor of social separation in rhesus monkeys. Psychopharmacology. 103, 551–556. Higley, J. D., Mehlman, P. T., Poland, R. E., Taub, D. M., Vickers, J., Suomi, S. J., and Linnoila, M. (1996a). CSF testosterone and 5-HIAA correlate with different types of aggressive behaviors. Biol. Psychol. 40, 1067–1082. Higley, J. D., Suomi, S. J., and Linnoila, M. A. (1996b). Nonhuman primate model of type II alcoholism? 2. Diminished social competence and excessive aggression. Alcohol. Clin. Exp. Res. 20, 643–650. Hofmann, J. E., Getz, L. L., and Klatt, B. J. (1982). Levels of male aggressiveness in fluctuating populations of Microtus ochrogaster and M. pennsylvanicus. Can. J. Zool. 60, 898–912. Houck, L. D., and Woodley, S. K. (1995). Field studies of steroid hormones and male reproductive behavior in amphibians. In ‘‘Amphibian Biology’’ (H. Heatwole and B. K. Sullivan, Eds.), pp. 677–703. Surrey Beatty & Sons, Chipping Norton, NSW, Australia. Hunt, K. E., Hahn, T. P., and Wingfield, J. C. (1997). Testosterone implants increase song but not aggression in male Lapland longspurs. Anim. Behav. 54, 1177–1192. Hunt, K. E., Hahn, T. P., and Wingfield, J. C. (1999). Endocrine influences on parental care during a short breeding season: Testosterone and male parental care in Lapland longspurs (Calcarius lapponicus). Behav. Ecol. Sociobiol. 45, 360–369. Insel, T. R., Wang, Z. X., and Ferris, C. F. (1994). Patterns of brain vasopressin receptor distribution associated with social organization in microtine rodents. J. Neurosci. 14, 5381–5392. Johns, L. S., and Liley, N. R. (1970). The effects of gonadectomy and testosterone treatment on the reproductive behavior of the male blue gourami Trichogaster trichopterus. Can. J. Zool. 48, 977–987. Jones, J. S., and Wynne-Edwards, K. E. (2000). Paternal hamsters mechanically assist the delivery, consume amniotic fluid and placenta, remove fetal membranes, and provide parental care during the birth process. Horm. Behav. 37, 116–125. Kahen, V., Katz, L. F., and Gottman, J. M. (1994). Linkages between parent–child interaction and conversations of friends. Soc. Dev. 3, 238–254. Kass, D. H., Berger, F. G., and Dawson, W. D. (1992). The evolution of coexisting highly divergent LINE-1 subfamilies within the rodent genus Peromyscus. J. Mol. Evol. 35, 472–485. Ketterson, E. D., and Nolan, V., Jr. (1994). Male parental behavior in birds. Annu. Rev. Ecol. Syst. 25, 601–628. Ketterson, E. D., Nolan, V., Jr., Wolf, L., and Ziegenfus, C. (1992). Testosterone and avian life histories: Effects of experimentally elevated testosterone on behavior and correlates of fitness in the dark-eyed junco (Junco hyemalis). Am. Nat. 140, 981–999. Kindler, P. M., Philipp, D. P., Gross, M. R., and Bahr, J. M. (1989). Serum 11-ketotestosterone and testosterone concentrations associated with reproduction in male bluegill (Lepomis macrochirus: Centrarchidae). Gen. Comp. Endocrinol. 75, 446–453.

318


Kindler, P. M., Bahr, J. M., and Philipp, D. P. (1991). The effects of exogenous 11-ketotestosterone, testosterone, and cyproterone acetate on prespawning and parental care behaviors of male bluegill. Horm. Behav. 25, 410–423. Kirkpatrick, B., Kim, J. W., and Insel, T. R. (1994a). Limbic system fos expression associated with paternal behavior. Brain Res. 658, 112–118. Kirkpatrick, B., Williams, J. R., Slotnick, B. M., and Carter, C. S. (1994b). Olfactory bulbectomy decreases social behavior in male prairie voles (M. ochrogaster). Physiol. Behav. 55, 885–889. Kishida, M., and Specker, J. L. (2000). Paternal mouthbrooding in the black-chinned tilapia, Sarotherodon melanotheron (Pisces: Cichlidae): Changes in gonadal steroids and potential for vitellogenin transfer to larvae. Horm. Behav. 37, 40–48. Klein, S. L., and Nelson, R. J. (1997). Sex differences in immunocompetence differ between two Peromyscus species. Am. J. Physiol. 273, R655–R660. Klomberg, K. F., and Marler, C. A. (2000). The neuropeptide arginine vasotocin alters male call characteristics involved in social interactions in the grey treefrog, Hyla versicolor. Anim. Behav. 59, 807–812. Knapp, R., Wingfield, J. C., and Bass, A. H. (1999). Steroid hormones and paternal care in the plainfin midshipman fish (Porichthys notatus). Horm. Behav. 35, 81–89. Koolhaas, J. M., van den Brink, T. H. C., Roozendaal, B., and Boorsma, F. (1990). Medial amygdala and aggressive behavior: Interaction between testosterone and vasopressin. Aggress. Behav. 16, 223–229. Koolhaas, J. M., Everts, H., de Ruiter, A. J. H., de Boer, S. F., and Bohus, B. (1998). Coping with stress in rats and mice: Differential peptidergic modulation of the amygdala–lateral septum complex. Prog. Brain Res. 119, 437–448. Liu, D., Diorio, J., Tannenbaum, B., Caldji, C., Francis, D., Freedman, A., Sharma, S., Pearson, D., Plotsky, P. M., and Meaney, M. J. (1997). Maternal care, hippocampal glucocorticoid receptors, and hypothalamic–pituitary–adrenal responses to stress. Science 277, 1659–1662. Liu, Y., Curtis, J. T., and Wang, Z. X. (2001). Vasopressin in the lateral septum regulates pair bond formation in male prairie voles. (Microtus ochrogaster). Behav. Neurosci. 115, 910–919. Lonstein, J. S., and De Vries, G. J. (1999). Sex differences in the parental behavior of adult virgin prairie voles: Independence from gonadal hormones and vasopressin. J. Neuroendocrinol. 11, 441–449. Lonstein, J. S., Rood, B. D., and De Vries, G. J. (2002). Parental responsiveness is feminized after neonatal castration in virgin male prairie voles, but is not masculinized by perinatal testosterone in virgin females. Horm. Behav. 41, 80–87. Lovic, V., Gonzalez, A., and Fleming, A. S. (2001). Maternally separated rats show deficits in maternal care in adulthood. Dev. Psychobiol. 39, 19–33. Lynn, S. E., Hayward, L. S., Benowitz-Fredericks, Z. M., and Wingfield, J. C. (2002). Behavioural insensitivity to supplementary testosterone during the parental phase in the chestnut-collared longspur, Calcarius ornatus. Anim. Behav. 63, 795–803. Margulis, S. W. (1998). Relationships among parental inbreeding, parental behaviour and offspring viability in old field mice. Anim. Behav. 55, 427–438. Marler, C. A., and Ryan, M. (1996). Energetic constraints and steroid hormone correlates of male calling behaviour in the tungara frog. J. Zool. Lond. 240, 397–409. Mayer, A. D., Monroy, M. A., and Rosenblatt, J. S. (1990). Prolonged estrogen–progesterone treatment of nonpregnant ovariectomized rats: Factors stimulating home-cage and maternal aggression and short-latency maternal behavior. Horm. Behav. 24, 342–364. Mazur, A., and Booth, A. (1998). Testosterone and dominance in men. Behav. Brain Sci. 21, 353–363 [discussion, pp. 363–397].


319

Mazur, A., and Lamb, T. A. (1980). Testosterone, status and mood in human males. Horm. Behav. 14, 236–246. McCarty, R., and Southwick, C. H. (1977). Paternal care and the development of behaviour in the southern grasshopper mouse, Onychomys torridus. Behav. Biol. 19, 476–490. McGuire, B. (1988). Effects of cross-fostering on parental behavior of meadow voles (Microtus pennsylvanicus). J. Mammal. 69, 332–341. McInroy, J. K. E., Brousmiche, D. G., and Wynne-Edwards, K. E. (2000). Fathers, fat, and maternal energetics in a biparental hamster: Paternal presence determines the outcome of a current reproductive effort and adipose tissue limits subsequent reproductive effort. Horm. Behav. 37, 399–409. Meek, L. R., Dittel, P. L., Sheehan, M. C., Chan, J. Y., and Kjolhaug, S. R. (2001). Effects of stress during pregnancy on maternal behavior in mice. Physiol. Behav. 72, 473–479. Mendl, M., and Paul, E. S. (1990). Parental care, sibling relationships and the development of aggressive behaviour in two lines of wild house mice. Behaviour 116, 10–41. Mendonça, M. T., Chernetsky, S. D., Nester, K. E., and Gardner, G. L. (1996). Effects of gonadal sex steroids on sexual behavior in the big brown bat, Eptesicus fuscus, upon arousal from hibernation. Horm. Behav. 30, 153–161. Metzgar, L. H. (1971). Behavioral population regulation in the woodmouse, Peromyscus leucopus. Am. Midl. Nat. 86, 434–448. Mize, J., and Pettit, G. S. (1997). Mothers’ social coaching, mother–child relationship style, and children’s peer competence: Is the medium the message? Child Dev. 68, 312–332. Moore, C. L. (1984). Maternal contributions to the development of masculine sexual behavior in laboratory rats. Dev. Psychobiol. 17, 347–356. Moore, C. L. (1992). The role of maternal stimulation in the development of sexual behavior and its neural basis. Ann. N. Y. Acad. Sci. 662, 160–177. Moreno, J., Veiga, J. P., Cordero, P. J., and Minguez, E. (1999). Effects of paternal care on reproductive success in the polygynous spotless starling Sturnus unicolor. Behav. Ecol. Sociobiol. 47, 47–53. Mugford, R. A., and Nowell, N. W. (1972). Paternal stimulation during infancy: Effects upon aggression and open-field performance of mice. J. Comp. Physiol. Psychol. 79, 30–36. Nelson, R. J. (2000). ‘‘An Introduction to Behavioral Neuroendocrinology.’’ Sinauer Associates, Sunderland, MA. Nunes, S., Fite, J. E., and French, J. A. (2000). Variation in steroid hormones associated with infant care behaviour and experience in male marmosets (Callithrix kuhlii). Anim. Behav. 60, 857–865. Nunes, S., Fite, J. E., Patera, K. J., and French, J. A. (2001). Interactions among paternal behavior, steroid hormones, and parental experience in male marmosets (Callithrix kuhlii). Horm. Behav. 39, 70–82. Oliveira, R. F., Almada, V. C., Goncalves, E. J., Forsgren, E., and Canario, A. V. M. (2001). Androgen levels and social interactions in breeding males of the peacock blenny. J. Fish Biol. 58, 897–908. Parkening, T. A., and Collins, T. J. (1991). Reproduction, pregnancy, and circulating reproductive hormones in the lactating Siberian hamster (Phodopus sungorus). J. Exp. Zool. 257, 401–407. Parker, K. J., and Lee, T. M. (2001). Central vasopressin administration regulates the onset of facultative paternal behavior in Microtus pennsylvanicus (meadow voles). Horm. Behav. 39, 285–294. Parker, K. J., Kinney, L. F., Phillips, K. M., and Lee, T. M. (2001). Paternal behavior is associated with central neurohormone receptor binding patterns in meadow voles (Microtus pennsylvanicus). Behav. Neurosci. 115, 1341–1348.

320


Pettit, G. S., and Bates, J. E. (1989). Family interaction patterns and children’s behavior problems from infancy to four years. Dev. Psychol. 55, 729–739. Pettit, G. S., Clawson, M. A., Dodge, K. A., and Bates, J. E. (1996). Stability and change in peer-rejected status: The role of child behavior, parenting, and family ecology. MerrillPalmer Q 42, 267–294. Pitkow, L. J., Sharer, C. A., Ren, X., Insel, T. R., Terwilliger, E. F., and Young, L. J. (2001). Facilitation of affiliation and pair-bond formation by vasopressin receptor gene transfer into the ventral forebrain of a monogamous vole. J. Neurosci. 21, 7392–7396. Potegal, M., and Ferris, C. F. (1990). Intraspecific aggression in male hamsters is inhibited by intrahypothalamic vasopressin-receptor antagonist. Aggress. Behav. 15, 311–320. Raine, A., Brennan, P., and Mednick, S. A. (1997). Interaction between birth complications and early maternal rejection in predisposing individuals to adult violence: Specificity to serious, early-onset violence. Am. J. Psychiatry. 154, 1265–1271. Reburn, C. J., and Wynne-Edwards, K. E. (1999). Hormonal changes in males of a naturally biparental and a uniparental mammal. Horm. Behav. 35, 163–176. Ribble, D. O. (1991). The monogamous mating system of Peromyscus californicus as revealed by DNA fingerprinting. Behav. Ecol. Sociobiol. 29, 161–166. Ribble, D. O., and Salvioni, M. (1990). Social organization and nest co-occupancy in Peromyscus californicus, a monogamous rodent. Behav. Ecol. Sociobiol. 26, 9–16. Richmond, M., and Conaway, C. H. (1969). Management, breeding, and reproductive performance of the vole, Microtus ochrogaster, in a laboratory colony. Lab. Anim. Care 19, 80–87. Roberts, R. L., Zullo, A., Gustafson, E. A., and Carter, C. S. (1996). Perinatal steroid treatments alter alloparental and affiliative behavior in prairie voles. Horm. Behav. 30, 576–582. Rosenblatt, J. S., Factor, E. M., and Mayer, A. D. (1994). Relationship between maternal aggression and paternal care in the rat. Aggress. Behav. 20, 243–255. Rosenblum, L. A., and Andrews, M. W. (1994). Influences of environmental demand on maternal behavior and infant development. Acta Paediatr. Suppl. 397, 57–63. Saino, N., and Møller, A. P. (1995). Testosterone-induced depression of male parental behavior in the barn swallow: Female compensation and effects on seasonal fitness. Behav. Ecol. Sociobiol. 36, 151–157. Scerbo, A. S., and Kolko, D. J. (1995). Child physical abuse and aggression: Preliminary findings on the role of internalizing problems. J. Am. Acad. Child Adolesc. Psychol. 34, 1060–1066. Schino, G., Speranza, L., and Troisi, A. (2001). Early maternal rejection and later social anxiety in juvenile and adult Japanese macaques. Dev. Psychobiol. 38, 186–190. Schug, M. D., Vessey, S. H., and Underwood, E. M. (1992). Paternal behavior in a natural population of white-footed mice (Peromyscus leucopus). Am. Midl. Nat. 127, 373–380. Schwartz, D., Dodge, K. A., Pettit, G. S., and Bates, J. E. (1997). The early socialization of aggressive victims of bullying. Child Dev. 68, 665–675. Semsar, K., Klomberg, K., and Marler, C. A. (1998). Arginine vasotocin increases calling-site acquisition by nonresident male grey treefrogs. Anim. Behav. 56, 983–987. Semsar, K., Kandel, F. L., and Godwin, J. (2001). Manipulations of the AVT system shift social status and related courtship and aggressive behavior in the bluehead wrasse. Horm. Behav. 40, 21–31. Sikkel, P. C. (1993). Changes in plasma androgen levels associated with changes in male reproductive behavior in a brood cycling marine fish. Gen. Comp. Endocrinol. 89, 229–237. Smith, C. J., and Haley, S. R. (1988). Steroid profiles of the female tilapia, Oreochromis mossambicus, and correlation with oocyte growth and mouthbrooding behavior. Gen. Comp. Endocrinol. 69, 88–98.


321

Smith, R. J. F. (1969). Control of prespawning behavior of sunfish (Lepomis gibbosus and L. megalotis). I. Gonadal androgen. Anim. Behav. 18, 575–587. Smith, R. J. F., and Hoar, W. S. (1967). The effects of prolactin and testosterone on the parental behaviour of the male stickleback Gasterosteus aculeatus. Anim. Behav. 15, 342–352. Snowdon, C. T. (1996). Infant care in cooperatively breeding species. In ‘‘Advances in the Study of Behavior,’’ Vol. 25 (C. T. Snowdon and J. S. Rosenblatt, Eds.), pp. 643–689. Academic Press, San Diego, CA. Solis, R., and Penna, M. (1997). Testosterone levels and evoked vocal responses in a natural population of the frog Batrachyla taeniata. Horm. Behav. 31, 101–109. Specker, J. L., and Kishida, M. (2000). Mouthbrooding in the black-chinned tilapia, Sarotherodon melanotheron (Pisces: Cichlidae): The presence of eggs reduces androgen and estradiol levels during paternal and maternal parental behavior. Horm. Behav. 38, 44–51. Stangl, F. B., Jr., and Baker, R. J. (1984). Evolutionary relationships in Peromyscus: Congruence in chromosomal, genetic and classical data sets. J. Mammal. 65, 643–654. Stern, J. M., and Lehrman, D. S. (1969). Role of testosterone in progesterone-induced incubation behaviour in male ring doves (Streptopelia risoria). J. Endocrinol. 44, 13–22. Storey, A. E., Walsh, C. J., Quinton, R. L., and Wynne-Edwards, K. E. (2000). Hormonal correlates of paternal responsiveness in new and expectant fathers. Evol. Hum. Behav. 21, 79–95. Stribley, J. M., and Carter, C. S. (1999). Developmental exposure to vasopressin increases aggression in adult prairie voles. Proc. Natl. Acad. Sci. USA 96, 12601–12604. Sullivan, J., Markert, J. A., and Kilpatrick, C. W. (1997). Phylogeography and molecular systematics of the Peromyscus aztecus species group (Rodentia: Muridae) inferred using parsimony and likelihood. Syst. Biol. 46, 426–440. Svare, B., and Gandelman, R. (1976). Postpartum aggression in mice: The influence of suckling stimulation. Horm. Behav. 7, 407–416. Tito, M. B., Hoover, M. A., Mingo, A. M., and Boyd, S. K. (1999). Vasotocin maintains multiple call types in the gray treefrog, Hyla versicolor. Horm. Behav. 36, 166–175. Townsend, D. S., and Moger, W. H. (1987). Plasma androgen levels during male parental care in a tropical frog (Eleutherodactylus). Horm. Behav. 21, 93–99. Townsend, D. S., Stewart, M. M., and Pough, F. H. (1984). Male parental care and its adaptive significance in a neotropical frog. Anim. Behav. 32, 421–431. Townsend, D. S., Palmer, B., and Guillette, L. J., Jr. (1991). The lack of influence of exogenous testosterone on male parental behavior in a neotropical frog (Eleutherodactylus): A field experiment. Horm. Behav. 25, 313–322. Trainor, B. C., and Marler, C. A. (2001). Testosterone, paternal behavior, and aggression in the monogamous California mouse (Peromsycus californicus). Horm. Behav. 40, 32–42. Trainor, B. C., and Marler, C. A. (2002). Testosterone promotes paternal behaviour in a monogamous mammal via conversion to oestrogen. Proc. R. Soc. Lond. B Biol. Sci. 269, 823–829. Travillion, K., and Snyder, J. (1993). The role of maternal discipline and involvement in peer rejection and neglect. J. Appl. Dev. Psychol. 14, 37–57. Trewavas, E. (1983). ‘‘Tilapine fishes of the genera Sardotherodon, Oreochromis and Danakilia.’’ Cornell University Press, Ithaca, NY. Tyron, P. R., and MacLean, S. F. (1980). Use of space by Lapland longspurs breeding in arctic Alaska. Auk 97, 509–520. Van Duyse, E., Pinxten, R., and Eens, M. (2000). Does testosterone affect the trade-off between investment in sexual/territorial behaviour and parental care in male great tits? Behaviour 137, 1503–1515.

322


Viau, V., Chu, A., Soriano, L., and Dallman, M. F. (1999). Independent and overlapping effects of corticosterone and testosterone on corticotropin-releasing hormone and arginine vasopressin mRNA expression in the paraventricular nucleus of the hypothalamus and stress-induced adrenocorticotropic hormone release. J. Neurosci. 19, 6684–6693. Wada, M., and Gorbman, A. (1977). Mate calling induced by electrical stimulation in freely moving leopard frogs, Rana pipiens. Horm. Behav. 9, 141–149. Wallace, H., Meredith, C., and Marler, C. A. (1998). Effect of father’s presence on offspring’s aggression and vasopressin levels in the mouse Peromyscus californicus. Soc. Neurosci. Abstr. 24, 1440. Wang, Z. X. (1995). Species differences in the vasopressin-immunoreactive pathways in the bed nucleus of the stria terminalis and medial amygdaloid nucleus in prairie voles (Microtus ochrogaster) and meadow voles (Microtus pennsylvanicus). Behav. Neurosci. 109, 305–311. Wang, Z. X., and De Vries, G. J. (1993). Testosterone effects on paternal behavior and vasopressin immunoreactive projections in prairie voles (Microtus ochrogaster). Brain Res. 631, 156–160. Wang, Z. X., and Insel, T. R. (1996). Parental behavior in voles. In ‘‘Advances in the Study of Behavior,’’ Vol. 25 (C. T. Snowdon and J. S. Rosenblatt, Eds.), pp. 643–689. Academic Press, San Diego, CA. Wang, Z. X., Ferris, C. F., and De Vries, G. J. (1994a). Role of septal vasopressin innervation in paternal behavior in prairie voles (Microtus ochrogaster). Proc. Natl. Acad. Sci. USA 91, 400–404. Wang, Z. X., Smith, W., Major, D. E., and De Vries, G. J. (1994b). Sex and species differences in the effects of cohabitation on vasopressin messenger RNA expression in the bed nucleus of the stria terminalis in prairie voles (Microtus ochrogaster) and meadow voles (Microtus pennsylvanicus). Brain Res. 650, 212–218. Wang, Z. X., Young, L. J., Liu, Y., and Insel, T. R. (1997). Species differences in vasopressin receptor binding are evident in early development: Comparative anatomic studies in prairie and montane voles. J. Comp. Neurol. 378, 535–546. Wang, Z. X., Liu, Y., Young, L. J., and Insel, T. R. (2000). Hypothalamic vasopressin gene expression increases in both males and females postpartum in a biparental rodent. J. Neuroendocrinol. 12, 111–120. Weiss, B., Dodge, K. A., Bates, J. E., and Pettit, G. S. (1992). Some consequences of early harsh discipline: Child aggression and a maladaptive social information processing style. Child Dev. 63, 1321–1335. Wiegmann, D. D., and Baylis, J. R. (1995). Male body size and paternal behaviour in smallmouth bass, Micropterus dolomieui (Pisces: Centrarchidae). Anim. Behav. 50, 1543–1555. Wetzel, D. M., and Kelley, D. B. (1983). Androgen and gonadotropin effects on male mate calls in South African clawed frogs, Xenopus laevis. Horm. Behav. 17, 388–404. Widom, C. S. (2000). Motivation and mechanisms in the ‘‘cycle of violence.’’ Nebr. Symp. Motiv. 46, 1–37. Wikelski, M., Hau, M., Robinson, W. D., and Wingfield, J. C. (1999a). Seasonal endocrinology of tropical passerines: A comparative approach. In ‘‘Proceedings of the 22nd International Ornithological Congress, Durban’’ (N. J. Adams and R. H. Slotow, Eds.), pp. 1224–1241. BirdLife, Johannesburg, South Africa. Wikelski, M., Hau, M., and Wingfield, J. C. (1999b). Social instability increases plasma testosterone in a year-round territorial neotropical bird. Proc. R. Soc. Lond. B Biol. Sci. 266, 551–556.


323

Williamson, F. S. L. (1968). Common Lapland longspur. In ‘‘Life Histories of North American Cardinals, Grosbeaks, Buntings, Towhees, Finches, Sparrows, and Allies’’ (O. L. Austin, Jr., Ed.), US National Museum Bulletin 237, Part 3, Smithsonian Institute, Washington, D. C. pp. 1597–1608. . Wingfield, J. C. (1994a). Regulation of territorial behavior in the sedentary song sparrow, Melospiza melodia morphna. Horm. Behav. 28, 1–15. Wingfield, J. C. (1994b). Hormone–behavior interactions and mating systems in male and female birds. In ‘‘The Differences between the Sexes’’ (R. V. Short and E. Balaban, Eds.), pp. 303–330. Cambridge University Press, Cambridge. Wingfield, J. C., Hegner, R. E., Dufty, A. M., Jr., and Ball, G. F. (1990). The ‘‘challenge hypothesis’’: Theoretical implications for patterns of testosterone secretion, mating systems, and breeding strategies. Am. Nat. 136, 829–846. Wingfield, J. C., Lynn, S. E., and Soma, K. K. (2001). Avoiding the ‘‘costs’’ of testosterone: Ecological bases of hormone–behavior interactions. Brain Behav. Evol. 57, 239–251. Winslow, J. T., Hastings, N., Carter, C. S., Harbaugh, C. R., and Insel, T. R. (1993). A role for central vasopressin in pair bonding in monogamous prairie voles. Nature 365, 545–548. Wolff, J. O., and Cicirello, D. M. (1990). Mobility versus territoriality: Alternative reproductive strategies in white-footed mice. Anim. Behav. 39, 1222–1224. Wolff, J. O., and Cicirello, D. M. (1991). Comparative paternal and infanticidal behavior of sympatric white-footed mice (Peromyscus leucopus noveboracensis) and deermice (P. maniculatus nubiterrae). Behav. Ecol. 2, 38–45. Wuensch, K. L., and Cooper, A. J. (1981). Preweaning paternal presence and later aggressiveness in male, Mus musculus. Behav. Neurol. Biol. 32, 510–515. Wynne-Edwards, K. E. (2001). Hormonal changes in mammalian fathers. Horm. Behav. 40, 139–145. Wynne-Edwards, K. E. (2003). From dwarf hamster to daddy. In ‘‘Advances in the study of Behavior’’ (P. J. B. Slater, Ed.), pp. 207–261. Academic Press, San Diego, CA. Wynne-Edwards, K. E., and Reburn, C. J. (2000). Behavioral endocrinology of mammalian fatherhood. Trends Ecol. Evol. 15, 464–468. Xia, X., and Millar, J. S. (1988). Paternal behavior by Peromyscus leucopus in enclosures. Can. J. Zool. 66, 1184–1187. Xia, X., and Millar, J. S. (1991). Genetic evidence of promiscuity in Peromyscus leucopus. Behav. Ecol. Sociobiol. 28, 171–178. Young, L. J., Winslow, J. T., Nilsen, R., and Insel, T. R. (1997). Species differences in V1a receptor gene expression in monogamous and nonmonogamous voles: Behavioral consequences. Behav. Neurosci. 111, 599–605. Young, L. J., Nilsen, R., Waymire, K. G., MacGregor, G. R., and Insel, T. R. (1999). Increased affiliative response to vasopressin in mice expressing the V1a receptor from a monogamous vole. Nature 400, 766–768. Young, L. J., Wang, Z., Cooper, T. T., and Albers, H. E. (2000). Vasopressin (V1a) receptor binding, mRNA expression and transcriptional regulation by androgen in the Syrian hamster brain. J. Neuroendocrinol. 12, 1179–1185. Young, L. J., Lim, M. L., Gingrich, B., and Insel, T. R. (2001). Cellular mechanisms of social attachment. Horm. Behav. 40, 133–138. Ziegler, T. E., and Snowdon, C. T. (2000). Preparental hormone levels and parenting experience in male cotton-top tamarins, Saguinus oedipus. Horm. Behav. 38, 159–167. Ziegler, T. E., Wegner, F. H., Carlson, A. A., Lazaro-Perea, C., and Snowdon, C. T. (2000). Prolactin levels during the periparturitional period in the biparental cotton-top tamarin (Saguinus oedipus): Interactions with gender, androgen levels, and parenting. Horm. Behav. 38, 111–122.



Cognitive Ecology: Foraging in Hummingbirds as a Model System Susan D. Healy1 and T. Andrew Hurly2 1

institute of cell, animal, and population biology university of edinburgh edinburgh eh9 3jt, united kingdom 2 department of biological sciences university of lethbridge lethbridge, alberta, t1k 3m4, canada

Cognitive abilities, especially learning and memory, are nearly always studied in the laboratory. This is because the laboratory environment allows for strict control over the testing conditions, any one of which may affect performance. These potential extraneous variables range from those emanating from the test environment, for example, lighting (both the level and the light–dark schedule), temperature, presence or absence of conspecifics and availability of cues (such as visual landmarks, directional noises or smells), through the animal’s motivational state (hunger level, sexual arousal), to those from the animal’s previous experience, for example, familiarity with the test procedure (pulling levers, pecking at lights). Investigations of learning and memory performance under such strictly controlled conditions have been useful in demonstrating what animals are capable of learning and remembering. For example, pigeons have been found to remember at least 400 photographic slides for longer than 12 months (Vaughan and Greene, 1984). However, while this memory capacity is impressive, it is still not clear to what use the pigeon puts this ability in its daily life. In other instances, assessments of learning and memory performances of animals in the laboratory vastly underestimate their real abilities. For example, tits tested with touch screen tasks can remember locations of images only for a few seconds or minutes (Biegler et al., 2001; Hampton and Shettleworth, 1996a,b; Healy, 1995), whereas when retrieving food caches in the laboratory or in the field, they can remember locations of caches for hours, days, weeks, even months (Haftorn, 1956; Healy and Suhonen, 1996; Stevens and Krebs, 1986). One of the fundamental assumptions underlying the validity of testing cognitive abilities under laboratory conditions (and, thus, the almost 325 Copyright 2003 Elsevier Science (USA). All rights reserved. 0065-3454/03 $35.00

326

SUSAN D. HEALY AND T. ANDREW HURLY

universal testing of these abilities in either pigeons or rats) is that there is a commonality of such abilities due to evolutionary descent (see Macphail and Bolhuis, 2001, for an exposition on this). However, there is a different, and not incompatible, view that evolution by natural selection produces diversity in design through local adaptation. Although Macphail and Bolhuis (see also Bolhuis and Macphail, 2001) argue that natural selection has acted only on perceptual systems such as the sensory processing areas, but not on central, cognitive processing regions (e.g., the hippocampus), this is not the view of others (e.g., Dwyer and Clayton, 2002; Flombaum et al., 2002; Hampton et al., 2002; MacDougall-Shackleton and Ball, 2002). At the least it would seem sensible to look for evidence that selection has or has not had a hand in shaping cognitive processing. A more ethological view has driven much of comparative psychology and, more recently, cognitive ecology [such as work on learning and memory in food-storing birds (e.g., Hilton and Krebs, 1990; Bednekoff et al. 1997), on cue use of sticklebacks living in different environments (e.g., Girvan and Braithwaite, 1998), and on shore crab prey handling (e.g., Hughes and O’Brien, 2001)]. The cognitive abilities of a wide range of species have been investigated although these tests, too, have all been carried out in the laboratory. It is rare, however, for results gained in the laboratory to be compared with results from the field. And yet, the strict control allowed by laboratory investigations requires that some caution should be exercised in generalizing to evolutionary or ecological contexts. This is, in part, because such control requires many variables to be kept constant, made irrelevant or removed, often without knowing whether they are important to the animal. There are, potentially, at least two advantages to investigating cognitive abilities in the field environment in which the animal lives. The first is that of gaining an understanding of the ecological or social demands that may act as selective forces on the cognitive abilities of the animal. The other is that tests of animals in the field, under ‘‘natural’’ conditions, offer the potential for insight into what it is that animals actually do rather than what it is they can do. This is important as it is in this situation that natural selection acts. Although we emphasize this second advantage, the rationale underpinning all our experiments is based on the first, as we believe that, to understand fully the workings of the brains of other animals, the mechanistic underpinnings elucidated in the laboratory (and perhaps in the field) will need to be married to the predictions produced from an understanding of the selective pressures under which the behavior of the animal has evolved and is maintained. We have used foraging in free-living rufous hummingbirds, Selasphorus rufus, as a model system in which to assess cognitive abilities, particularly

COGNITIVE ECOLOGY: FORAGING IN HUMMINGBIRDS

327

learning and memory, in animals in the field [see also Gass and Montgomerie (1981) and Gass and Sutherland (1985)]. The value of this system is twofold: first, and possibly most important, are the logistic advantages. These are multiple: the birds (1) will feed from artificial feeders in preference to real flowers; (2) establish and vigorously and successfully defend a territory around the feeder; (3) can be individually marked by application of ink onto breast feathers (rings on the legs would not be visible); (4) can be watched from within a few meters without the use of a hide or other stealth, or binoculars; (5) come to feed regularly throughout the day (approximately every 10 min); and (6) learn rapidly (i.e., within 1–2 h) to feed from a wide variety of ‘‘feeders’’ and ‘‘flowers.’’ Furthermore, (7) once initially trained, the number, color, size, shape, location, and contents of these flowers can be varied without apparently disturbing the bird. Ironically, the combination of all of these features results in a system that is close to an open-field version of a laboratory system. It is for this reason that we consider the hummingbird system useful and thus perhaps a model system for addressing questions concerning learning and memory in the field. Classic field experiments typically involve manipulating one feature or another in the environment of some animals and comparing the outcome of that manipulation with the behavior of animals that were not manipulated. A major assumption is that all other things are equal aside from the manipulated feature. Such an assumption is different from the approach used to assess learning and memory in the laboratory, in which each variable (sometimes even implausible ones) is kept either constant or, at the least, measured. Venturing into the field to test learning and memory abilities, then, means a compromise between the opposite desires of wanting to test animals under natural conditions and of wanting to be sure that the results are not due to one of many possible extraneous variables. The logistics of testing foraging in territorial rufous hummingbirds are such that this behavior appears to offer a compromise that will irritate both field zoologists and laboratory psychologists but may also continue the integration between the two groups that has been seen in song learning, imprinting, and food storing. The second advantage to using foraging in hummingbirds for investigating cognition in the field is that the birds feed on flowers that would seem to require learning and memory in at least two ways: (1) because the flowers of different species vary in color, shape, size, location and content, the birds may learn their species-specific features; and (2) some of the flowers refill once emptied, but do so over the course of several hours. Foraging could thus be more efficient if they could avoid flowers emptied recently.

328


In the heyday of optimal foraging studies, the notion that these birds, or indeed any nectarivore, might use memory of flowers during foraging was not initially entertained. Rather, descriptions of decision making began with attempts to match the behavior of the animal to simple rules in the hope that once one understood how the animal made a single choice then subsequent choices could be accurately predicted. To a considerable degree such simple decision rules have been applied successfully to foraging in the hymenopterans, for example, the upward spiral visiting of flowers on an inflorescence by bees or the downward movement by some other species (Waddington and Heinrich, 1979). Sharing as they do with the hymenopterans many physical and behavioral characteristics, it was, perhaps, surprising that hummingbird foraging could not be described in the same, relatively simple ways (Hainsworth et al., 1983; Pyke, 1981, 1984; Wolf and Hainsworth, 1990; for a possible exception see discussion of traplining by hummingbirds, e.g., Feinsinger, 1976; Gill, 1988; also see below). However, unlike the hymenopterans, at least some species of hummingbird are territorial (Feinsinger and Chaplin, 1975, Kodric-Brown and Brown, 1978). Territoriality has been correlated with systematic foraging in other bird species [amakihis, Loxops virens (Kamil, 1978); pied wagtails, Motacilla alba (Davies and Houston, 1981)] and, in the sunbird Leonotis nepetifolia, such systematic foraging was estimated to enhance intake by as much as 25% (Gill and Wolf, 1977). One of the ways in which territory owners might use systematic foraging to enhance intake is to feed on the edges of their territories in the morning. As the edges are the most susceptible to conspecific intrusion, this early edge foraging followed by moving to the center of the territory as the day goes on should result in lowered nectar loss to intruders and a smaller area to defend later in the day (Paton and Carpenter, 1984). This ‘‘defense by exploitation’’ might well be energetically much less costly than suffering the loss of nectar (intruders can be responsible for removing significant amounts; Yeaton and Laughrin, 1976) or of chasing intruders from within their territory (Gass and Montgomerie, 1981). Rufous hummingbirds en route to overwintering grounds appear to carry out such defense by exploitation, with foraging bouts during the early morning at one or a few plants near to the territory edge (Gass and Montgomerie, 1981; see also Sutherland et al., 1982). Both this behavior and that of returning to feed at plants from which they have just chased off intruders (see Paton and Carpenter, 1984) would seem to require a far more flexible foraging technique than those typically described for the hymenopterans. Combining this foraging behavior with the observation that a territorial male spends the majority of his day sitting high up in a tall


329

tree from which he can survey all, or most, of his territory, led us to propose that foraging efficiency might be increased if the male could remember which flowers he has recently emptied in order to avoid revisiting them too soon. It is perhaps pertinent to point out here that defense by exploitation would not be successful if territorial hummingbirds chose which flowers to visit solely on the basis of their color, as has sometimes been suggested (see below).

I. Learning and Memory Systematic foraging by hummingbirds was first discussed with reference to traplining, a behavior in which the birds revisit frequently renewing flowers or plants apparently in some regular, routelike fashion, as fur trappers and bees do (Feinsinger and Colwell, 1978; Stiles and Wolf, 1979; see also Garrison and Gass, 1999; Gill, 1988). There has been some attempt at differentiating between traplining and territoriality in hummingbirds in terms of the morphological characteristics of birds (e.g., body size; bill length; wing disk loading, which affects hovering energetics; see Feinsinger and Colwell, 1978) and flower availability, as well as the spectrum of potential competitors, but it seems plausible that both foraging techniques would require at least a modicum of memory. The maintenance of a trapline route may entail visiting a small number of flowers (usually dispersed ‘‘rich’’ flowers containing 70–120 l of nectar per flower), whereas territorial hummingbirds (those for which there are quantitative data) defend territories containing at least several hundred flowers (Armstrong et al., 1987; Gass, 1979). Nonetheless, unless it is supposed that trapliners can see the next flower to visit, or can remember the appropriate direction in which to fly from the one they are currently feeding on, these birds must remember the locations of flowers and possibly associate those locations with the time at which they were last visited. Unfortunately there are no substantial, quantitative data on traplining in hummingbirds, so we do not know the number of flowers, or the time intervals between visits, that characterize this kind of foraging, which is found particularly in the hermit hummingbirds. Territorial birds, on the other hand, may well be able to see all, or most, of the flowers in their territory from their guard post, although this is not very helpful with respect to foraging systematically. They, too, must remember which flowers or plants they have fed on and how recently. Some of the first descriptions of learning and memory in hummingbirds come from Bene´ (1941, 1945), who showed that preferences for the color of a flower could be learned, irrespective of the common anecdotes

330


associating hummingbirds with an innate preference for red objects. Bene´ also described the rapid learning by a female black-chinned hummingbird (Archilochus alexandri) of an association between a new artificial feeder and its contents of dilute honey. After three successive trips to the feeder, the visual features of the feeder were dramatically changed, and yet on her fourth visit, the bird flew straight to the feeder and probed it. Bene´ changed the visual characteristics of the feeder several times, seemingly without affecting the foraging of the bird, before moving the feeder to a number of different locations in succession. Bene´ reported that, for up to four locations, the bird visited the previous locations in the reverse order to that in which they had been learned. Later, when as many as nine locations had been used, the bird no longer visited them in any apparent order but would take shortcuts between these previously rewarded locations. Although not totally compelling, these early data, along with an experiment by Lyerly et al. (1950), which tested color and position preferences in a Mexican violet-eared hummingbird (Colibri thalassinus), suggest that these birds pay attention to, and learn, the locations of food rewards, irrespective of their visual features. A. Color and Its Role in Flower Choice More substantial subsequent experiments have shown that the anecdotal reports of hummingbirds visiting red flowers (or red objects of many kinds) are unlikely to be due to an innate preference for red but rather that birds learn an association between reward and the color of the rewarded location. So why is this preference for red still a common assertion? Grant (1966) put forward a hypothesis that seems plausible although it is still untested. Grant noted that the prevalence of red-colored flowers that are exclusively hummingbird pollinated was true for the Californian flora and perhaps, after Pickens (1930), also for the hummingbird-pollinated flora of eastern North America. However, this is not the case for tropical and subtropical America, which have many species of hummingbird apparently feeding from and pollinating flowers of colors other than red. Grant also noted that, of the hummingbird species found in California and eastern North America, six were migratory while the sole resident species (Anna’s hummingbird, Calypte anna) typically makes postbreeding movements to higher elevations. Thus the flora that these birds are feeding on is frequently exposed to birds that are either simply passing through or making only a temporary stop. Grant supposed that it would be advantageous to both the plants and the birds for the flowers to be highly conspicuous (and most agree that for diurnal birds foraging in a somewhat open environment, red on a green background offers the highest contrast;


331

e.g., Ridley, 1930; Siitari et al., 1999). Once the birds had learned that the conspicuous red flowers signalled reward, it would be beneficial for them to generalize from one red flower to another. It would also be advantageous for both parties (plants and hummingbirds) for the hummingbird-pollinated Californian flora to converge on the use of red rather than other colors. Support for this idea comes from Thomson et al. (2000), who found that the North American flowers that are visited by hummingbirds are more likely to be red or orange than are their close relatives that are visited by bees. An additional possibility is that by using red the plants are also decreasing the detectability of their flowers by bees (Raven, 1972). Chittka et al. (2001) found that bumblebees took longer to find red flowers than to find blue or yellow flowers. The morphology of these hummingbirdpollinated flowers is also different from those typically pollinated by insects or bats, with differing flower structure and nectar content (both amount and concentration). Insects and bats are therefore inappropriate for pollinating these flowers, and plants would benefit by dissuading them. In spite of this some bees do rob hummingbird flowers. Other birds, such as songbirds, that are equally able to detect hummingbird flowers, are likely to learn to avoid them as a food source as they are unable to digest sucrose, the major component of the nectar of these flowers (Stiles and Freeman, 1993). However, experimental evidence, of the kind we have for the role of visual information in foraging and on the evolution of color vision in the hymenopterans (e.g., Chittka and Briscoe, 2001; Chittka et al., 2001), is still sadly lacking for hummingbirds. Because of the ever recurring discussion of the role of red coloration in hummingbird foraging (see also a parallel literature on coloration of fruits; e.g., Willson and Whelan, 1990), the most substantial investigations into learning and memory in these birds have involved presentations of differently colored feeders, often containing the same level of reward. All these studies have found both that hummingbirds will readily learn any given color, if rewarded for doing so, and that they quickly develop preferences for feeders in particular locations. For example, a number of the experiments have involved presenting birds with feeders in a line, and found that the birds always develop preferences for the feeders at the ends, even though they will also readily visit the other, intervening feeders. However, there is a problem with assuming that spatial cues are playing a more substantial role than color when considering the data collected in these experiments. Several articles describe this kind of experiment, and each apparently tested the same individual(s) on a series of occasions (e.g., Bene´, 1941, 1945; Collias and Collias, 1968; Grant, 1966; Lyerly et al., 1950; Miller and Miller, 1971). However, it is not possible to determine how

332


many birds were really involved in these experiments, as they were not individually marked. It was presumed that the birds were making return visits because the feeder visited on one trip was more likely to receive a subsequent visit. Visiting the feeder on the end of a row, however, may also be a result of that feeder being more conspicuous, or nearer to the place the bird(s) approach from. Although it is likely to be correct that the birds in question were revisiting, one of the prerequisites for carrying out work on learning and memory, whether in field or laboratory, is that it must be possible to identify animals individually. Sometimes in the laboratory, and nearly always in the field, this means that animals must be marked, particularly animals like the hummingbirds, among whom individual differences in morphology are insufficient to discriminate between them. Making a comment at all about the need for individual recognition may seem patronizing, as it is a fundamental feature of most experiments on behavior in the field; however, it is a recurring feature of most of the fieldwork on learning and memory in hummingbirds that it has been carried out on unmarked individuals. Reading this literature might give the impression, inaccurately, that individual marking was not necessary. The problem of determining whether the same birds visit the test flowers might explain the discrepancy between two studies examining the relative preference of hummingbirds for flower color and morphology (see below for a description). The intention of these studies was to examine the role that flower color plays in ‘‘pollination syndromes,’’ the elements that tie together a specific plant–pollinator coevolutionary relationship. Fixed (usually assumed to be innate) color preferences have long been assumed to be an essential component in this but, more recently, not least because of the consistent finding that pollinators (not just hummingbirds) will rapidly learn color–reward associations, it has been suggested that color merely plays a role in increasing the conspicuousness of the flower. Morphological features, such as the size and shape of the flower, have been proposed as being more closely associated with the quality of reward offered and thus these elements are more likely to have been selected for in close association with pollinator morphology. Meleńdez-Ackerman et al. (1997) allowed 6 captive hummingbirds to choose between 66 experimental flowers (Ipomopsis aggregata and closely related species), whose color, morphology, and nectar reward either varied naturally or, in some tests, were manipulated. When color was varied, but reward held constant, the initial preference for red disappeared in all birds (see also Elam and Linhart, 1988, who found no discrimination by hummingbirds among differently colored I. aggregata flowers in the field). When color was held constant, birds preferred I. aggregata flowers to Ipomopsis tenuituba, and it


333

was assumed that birds were making this discrimination on the basis of variation in morphological features such as corolla width. Grant and Temeles (1992) showed that these birds extract more nectar from flowers with wider corollas (see also Lange et al., 2000). These data, then, offered support for the notion that coevolution between hummingbirds and their flowers may be focused on flower morphology and content rather than color. However, field experiments in which the same manipulations were carried out (Meleńdez-Ackerman and Campbell, 1998) did not yield the same result. Birds preferred to visit red flowers throughout these tests, and when color was held constant, they did not discriminate among flowers on the basis of their morphological differences. The authors themselves raise the possibility that there is a ‘‘learning and memory’’ explanation for the difference between the results. Unlike the earlier study, in which individuals were tracked and birds were allowed to make more than 20 visits to each flower type, in this later experiment, not only were birds unmarked (although there were at least 2 of them as data were collected from 2 species) but only 11–18 visits were recorded in total. The suggested explanations for this difference were that (1) the birds in the first study had not had prior experience of one of the flower types used in the test, as that plant species did not occur in the area from which the birds were caught, or (2) the birds in the second experiment did not have enough experience with the test flowers to learn to discriminate between the flower types on the basis of their different morphological features. Neither of these possibilities can be excluded. In spite of these plausible explanations for the difference in outcome between the experiments, the authors effectively dismiss them by concluding that the color–nectar reward relationship is more important than morphological variation among flowers of different species. This conclusion has potentially important implications for on-going discussions of plant–pollinator coevolutionary relationships, about the specificity of which there is a dispute (Aigner, 2001; Johnson and Steiner, 2000). If pollinators pay more attention to the color of flowers and/or learn this feature better or more quickly than morphological features, then one might also expect to see selection acting more strongly on color, a rather nonspecific floral feature. However, to determine whether the discrepancy between the experiments is a biological or a logistical one requires marking birds and following each individual’s choices, coupled with providing birds with equal experience of the two experimental set-ups. Whatever the interpretation put on the data of Meleńdez-Ackerman et al., they are consistent with the suggestion by Grant (1966) that birds with brief access to flowers, such as when they are simply passing through

334


an area on migration, will use the most conspicuous cue to make flower choices. With increasing experience, however, birds will both use additional cues and learn to ignore the color information if it is incompatible with reward outcome. Ironically, one of the studies with the largest sample of individually marked birds (22 wild black-chinned hummingbirds Archilochus alexandri) did not take individual performances into account when assessing discrimination abilities (Goldsmith et al., 1981; Goldsmith and Goldsmith, 1979). In spite of this, the authors discuss whether the birds learned associations with reward, and whether red and green were learned at the same rate, as well as whether birds can simultaneously learn two conflicting color associations. They, too, seem to have found that birds developed strong location preferences and that the position of the feeder easily overrode the value of the color of the feeder to the birds. The authors claim that the birds learned red–reward associations at the same rate as green–reward association and that brightness was not used as a major cue. However, because the birds have a striking position bias and because individual data were not analyzed separately, these conclusions cannot be substantiated. Learning rates cannot be summed across individuals in this way—group data may mask wide variation in individual performance such that a majority of birds that learn rapidly, for example, may obscure the fact that some of the birds never learned or made the discrimination at all. At this stage, then, wavelength discrimination by black-chinned hummingbirds (or, indeed, any hummingbirds) has yet to be adequately investigated. Another study that also used marked birds and assessed learning of cues was that carried out by Sandlin (2000a; see also Sandlin, 2000b) to investigate the relationship between cue use and competitor density in foraging decisions made by males and females of three species of hummingbird (black-chinned hummingbirds, Archilochus alexandri; bluethroated hummingbirds, Lampornis clemenciae; and magnificent hummingbirds, Eugenes fulgens). However, just as in the studies by Goldsmith and co-authors, surprisingly, effort was put into marking the birds and then these marks were not used to distinguish the data within and between birds. It is not clear whether the data presented are fair representations of the behavior of the birds, or how many birds participated in the experiments, even though the experiment purports to investigate the role of reliable versus ambiguous cues in foraging decisions, yet another question that is reliably addressed only by collecting data from individuals. The reliable cues were given by blue symbols on feeders providing a rich (30%) sucrose source and yellow symbols on a poorer (15%) sucrose source while the ambiguous cues were symbols with quarters in blue and


335

yellow. The birds were presented with arrays of 10 feeders for several 30min periods each day. The arrays had either 5 yellow and 5 blue feeders or 10 ‘‘ambiguous’’ feeders. The colors were kept constant both across days (six trials in a day) and through the field season. Locations, however, were changed at the beginning of each trial, so birds were trained, effectively, to pay more attention to the color rather than to spatial cues. Nonetheless, although the birds were less likely to feed from the rich feeders when the visual cues were ambiguous, when competition was low their performance was conspicuously better than by chance. One explanation for this result is that, rather than differentiating between reliability and ambiguity, as was the intention, the birds were being offered feeders that, by their color, required no additional learning of the color–reward association after the first trial or feeders that were differentiated only by their locations, which had to be learned anew on each trial. Even though the responses of the birds were sorted only by species, they appear to have learned the color–reward association rapidly and well, as they nearly always chose the blue feeders when these were offered, a result that is consistent across trials and days and one to be expected if the cue is reliable across time. However, in the ambiguous color trials the birds also seemed to have learned the location–reward associations, information about which was available only during the halfhour of testing. This experiment, therefore, seems to have added at least as much to the literature on the rapid rate at which these birds learn information about reward locations, irrespective of the location’s color, as it has to the role of reliability in information acquisition during foraging (see also Biegler and Morris, 1996; Roberts and Pearce, 1998). Color is, undisputedly, a cue that these birds use when foraging. However, from these rather uncontrolled experiments, as well as from experiments we have carried out (detailed below), color seems to be used in a way that fits well with the suggestion by Grant (1966) that the use of this cue by these birds is context dependent. When birds are in unfamiliar territory, such as on migration, they depend heavily on flower color when making decisions about which flowers to visit. However, territorial birds, such as the rufous hummingbirds resident throughout the breeding season in the foothills of the Rocky Mountains in Canada, pay little attention to color, at least under some circumstances. Some of the evidence we have to support this comes from a series of experiments we carried out with a 10-flower array. In some testing situations, there was a single rewarded flower (containing 20% sucrose solution) and nine flowers that were either empty or contained water (which the birds find somewhat aversive and will spit out), while in other experiments there were three rewarded and seven unrewarded flowers (see

336


Fig. 1. Schematic array of 10 flowers used to determine the role of color pattern in rufous hummingbird memory for flowers. In (a) 10 flowers were presented, each with a different color pattern (designated by the different letters). One of the flowers contained sucrose (designated by the boldface circle) while the other flowers contained water. Once the bird had learned which was the rewarded flower (phase 1), the array was moved 2 m (edge to edge) and the bird observed until he had learned which flower contained reward (phase 2). In this example, the flowers are the same color patterns and in the same array positions in phase 2 as they were in phase 1. In (b) the same procedure was followed, but three nearest neighbor flowers contained reward and seven contained water. (From Hurly and Healy, 2002.)

Fig. 1; Hurly and Healy, 2002). Each bird tested was an individually marked territorial male. During testing only the focal male visited the array of flowers and he received each set of trials consecutively. We found that when flowers each bore a unique color pattern (two colors in geometric patterns), the birds learned with fewer errors which was\were the rewarded flower(s). Once the bird had learned which were the rewarding flowers we shifted the array 2 m from the original location and presented him with either exactly the same array (as depicted in Fig. 1) or an array with either the flowers’ color patterns, the position of the rewarded flowers, or both, altered. We reasoned that, if the bird remembered both what the flower looked like and where it was in the array, then if neither had changed (apart from the location of the entire array) he would learn which were the rewarded flowers with fewer errors in this second phase of a trial than in the first phase. In some tests birds did learn with fewer errors in phase 2 but only when either position alone or both color and position were unaltered. We tried a number of different array manipulations (e.g., leaving rewarded positions the same, changing flower colors, rearranging the flowers within the array) in order to show that the birds did remember flower color pattern, but


337

Fig. 2. (a) Schematic of two arrays designed to prevent any spatial information used in phase 1 from being used when learning which were the rewarded flowers (three, designated by boldface circles) in phase 2. In the first array, flowers retained the same color patterns between the phases while in the second array, the flower color patterns were unique in the second phase. (b) Mean number of errors (SE) made before reaching criterion in the two phases for each array. Phase 1 is represented by the shaded columns and phase 2 by the solid columns (n = 10 males). The difference between the phases was significant for 3-Change-Shape (p = 0.001) but not for 3-Change-Color and Shape (p = 0.10). From Hurly and Healy (2002).

these were all to little avail if there was any kind of spatial cue available. To show that birds do remember color information in these kinds of tests, we had to present the birds with arrays containing the same flowers but in differing array shapes (see Fig. 2). We had found a similar result in a previous test, in which the bird had to learn the positions of 8 rewarded flowers in an array of 16. Arrays contained flowers that were either all the same color or were all of different colors. In half the arrays the flowers were 10 cm apart, and in the other half they were 80 cm apart. When the flowers were all different colors the birds learned which contained rewards with fewer errors than when the flowers were all of the same color. The distance between the flowers had no effect on the speed of acquisition. Once a bird had learned which were the rewarded flowers, the arrays were moved as shown in Fig. 3, such that half the flowers were in new locations but the remainder were in previously occupied locations. When the bird returned to the array his visits were recorded. There was an effect of having moved the array in all four array types but the major effect was due to the distance between the flowers. Birds did not appear to use information pertaining to flower color

338


Fig. 3. (a) Schematic of an array of 16 flowers, in which 8 flowers contained sucrose (shaded circles) while 8 flowers contained water (open circles). Once a bird had learned which flowers contained reward to a criterion of 80% or greater, the array was shifted such that half of it occupied half of the locations of the preshift array (designated by the circles overlying half of the plus symbols and half of the minus symbols). The plus symbols represent locations (relative to landmarks outside the array) of flowers that had been previously rewarded and the


339

pattern once the rewarded locations were known (see Fig. 3 and discussion later). This clear use of color information to learn locations more rapidly but not to improve, or to be required for, subsequent recall of those locations, is somewhat akin to the phenomenon of potentiation (Batsell et al., 2001). Potentiation has also been invoked to explain the coupling of nonaversive odorants with bright coloration in aposematic insects (Guilford et al., 1987; also see Rowe, 2002). Naive predators encountering a novel, brightly colored unpalatable prey that also smells (not necessarily unpleasantly) learn to avoid the new prey faster than if the prey is odorless. Once the association is learned the odor does not have to be present for prey avoidance to be observed. Investigation of potentiation in the psychological literature has also typically been with odor as the potentiating stimulus (e.g., Durlach and Rescorla, 1980; Rusiniak et al., 1979; Slotnick et al., 1997). The color augmentation of location learning seen in our hummingbird experiments is, to our knowledge, the first example of a visual cue potentiating a spatial location–reward association. The question as to whether hummingbirds might have innate color preferences has not yet been properly addressed. The evidence that, like butterflies (Weiss, 1997, 2001), hummingbirds will learn associations between color and reward readily (Hurly and Healy, 1996) does not rule out the possibility that there are innate preferences as well. Given the greater likelihood of coevolution between hummingbirds and plants in the tropics, perhaps species from this region or further south would be more profitable to test. Given the requirement for captive raising in order to properly address this question, however, it is likely to remain unanswered for the foreseeable future.

II. Spatial Learning and Memory It seems odd that the role that color plays in both hummingbird foraging and, by extension, plant–hummingbird pollination syndromes continues to minus symbols represent locations of flowers that previously contained water. (b) The proportion of visits to correct flowers (means SE) by birds before the shift (lighter columns) and following the shift (darker columns). There were four array types: two in which all the flowers bore the same color pattern (Same 10 and Same 80) and two in which each flower bore a unique color pattern (Different 10 and Different 80); two arrays had flowers 10 cm apart while the other two arrays had flowers 80 cm apart (n = 4 males). Postshift performance was significantly more affected (p = 0.01) when flowers were 80 cm apart than at a spacing of 10 cm. (From Healy and Hurly, 1998.)

340


attract attention, given that the most consistent finding in hummingbird foraging is that they pay much more attention to the location of a flower. Miller et al. (1985), testing rufous hummingbirds in a laboratory, found that even striking changes to the visual aspects of feeders did not affect which ones the birds preferred to feed from. A common anecdote is also that migrant hummingbirds appear at locations where feeders had been supplied the previous year, thus providing the stimulus for the feeder to be retrieved from the cupboard in which it has spent the winter, filled, and made available to the birds. One possible explanation for the emphasis on the use of color cues, to the neglect of memory for spatial locations, may be that optimal foraging theory (OFT) has played an important and influential role in setting the questions asked about foraging decision making in a range of pollinators, so that the kinds of information use incorporated even into more recent models has been rather simplistic. In addition, much of the modeling has focused on decision rules that birds might use to decide when to leave a patch rather than which patch to visit next [see, e.g., Gass and Montgomerie (1981) and Armstrong et al. (1987)]. It may also seem unparsimonious to assume that animals may remember many locations, rather than use some computationally simpler technique such as movement rules or to visit all conspicuous objects. Some models have also assumed that birds retained memories for flowers in a patch only until they began foraging in the next patch (e.g., Armstrong et al., 1987). Color is simply a much more obvious cue to investigators than is memory of the location of a flower. However, all the attempts to show that the hummingbirds use the kinds of rules that have featured in OFT models have failed, and the consistent outcome is always that the birds use information gained on previous visits (Pyke, 1981; Valone, 1992; Wolf and Hainsworth, 1990). While the incorporation of learning and memory into these models has taken some considerable time, the idea that an ecological approach might provide insight into learning and memory has been around and productive for at least two decades. Garcia et al. (1966) found that rats could learn to avoid a novel food that had been associated with an aversive outcome (long delay taste aversion learning). This association was a precise pairing between the novel taste and smell of the food with the ensuing illness. Psychologists subsequently termed such specific associations ‘‘constraints on learning.’’ This label was used in recognition that the general principles of learning assumed to underlie all learning, did not always apply. It has since been developed into the notion of ‘‘adaptive specializations’’ of learning and memory, that is, the evolution of ways of processing, storing, and using information in a functionally appropriate manner (Krebs, 1990; Shettleworth, 1998).


341

The most thoroughly worked example of an adaptive specialization to date is that centered around the hypothesis that food storing is associated both with an enlarged hippocampal region (the area of the brain subserving spatial memory) and with enhanced spatial memory (Krebs et al., 1989; Olson et al., 1995; Shettleworth, 1990). But an ecological adaptation approach was first applied to hummingbird cognition by Cole et al. (1982). Given that flowers do not immediately replace their nectar once emptied, they suggested that hummingbirds might use their memory for recently visited flowers to avoid them in the near future. They predicted that these birds might find it easier to learn to visit a flower in a new location (win–shift) than to return to a flower in the same location (win–stay). Cole et al. tested three species and all the birds learned to win– shift much faster than to win–stay. Even on the first day of training the birds were more ready to shift than to stay and the subsequent rate of improvement on shift trials was also faster. In a field version of equivalent tasks, including both a delayed nonmatching-to-sample task and a delayed matching-to-sample task, we found similar results (Healy and Hurly, 2001). This ability to acquire a win–shift rule more readily than a win–stay rule is not, however, confined to hummingbirds. Rats, pigeons, pigs, and tits are also much faster to learn a nonmatching-to-sample rule than a matching-to-sample rule (e.g., McGregor and Healy, 1999; Mendl et al., 1997). It has been argued that animals should respond flexibly to whether or not a food patch depletes and to the scale over which resource depletion occurs. Cole et al. (1982) suggested, for example, that, as birds might view patches of inflorescences as a resource that was effectively not depleting, they may more readily learn to win–stay in such situations. The birds do, clearly, learn to return to profitable locations both over shorter time periods than a normal flower would take to refill and also after intervals that are considerably longer (witness their fidelity to artificial feeders, both within a season and even after a winter spent in Mexico). They also pay attention to spatial scale. In 2 field experiments using arrays of 5 or 16 flowers, we found that birds used a different spatial scale to remember which flower(s) contained reward depending on how close these flowers were to each other (Healy and Hurly, 1998; Hurly, 1996; see also experiments on scale in relation to spatial associative learning: Brown and Gass, 1993; Brown, 1994). Birds were presented with arrays in which there were either 1 of 5 flowers containing sucrose or 8 of 16. In different trials, flowers were presented at different distances from each of the nearest neighbors. Once the bird had learned which were the rewarded flower(s), the array was shifted such that half of it overlapped previous flower locations. When the flowers were separated by 40 cm or

342


Fig. 4. Proportion of choices (means SE) to visit flowers in either the location (darker columns) or in the position (lighter columns) they occupied in phase 1. When flowers were spaced 10 cm apart, birds were more likely to visit flowers in the appropriate position and when flowers were 80 cm apart, birds were more likely (p < 0.0001) to visit flowers in the appropriate location (as defined by landmarks outside the array; n = 4 males). (From Healy and Hurly, 1998.)

less, the birds visited flowers in the shifted array that were in the appropriate position in the array (i.e., using other flowers in the array to make flower choices), but when the flowers were more distant, they visited the flowers in the appropriate location (i.e., using landmarks outside the array to make flower choices; see Fig. 4 and Hurly and Healy, 1996). The fact that the birds appear to pay attention to, and/or remember, cues based on their spatial scale may have an ecological basis: some of the flowers they feed on are close together, such as inflorescences on a single stalk, whereas others are somewhat further apart, such as flowers found on bushes, and still others are single flowers separated by tens or hundreds of centimeters. It would seem ‘‘sensible’’ for the birds to use a hierarchy of spatial scale if the flowers are close together, such as ‘‘encode and remember where the plant is in the field, then remember roughly where on the plant flowers have been visited.’’ In one experiment we presented birds with arrays in which the rewarded flowers either were nearest neighbors or were dispersed around the array with at least one intervening flower between them (Hurly and Healy, 2002). The birds learned more quickly which were the rewarded flowers when the flowers were nearest neighbors than when the rewarded flowers were dispersed around the array (see Sutherland and Gass, 1995, for a laboratory demonstration of this effect). It is not clear whether this is because the birds learn to fly a simple pattern from one flower to another or whether it is due to them ‘‘chunking,’’ or


343

grouping, the rewarded locations together in memory, perhaps as a particular pattern or simply a general region of the array (Dallal and Meck, 1990; Macuda and Roberts, 1995). When the flowers are in close proximity, the costs of revisiting a flower that has not yet refilled or missing a flower that has not been visited is probably minimal. When the flowers are far apart, however, the cost of making a mistake is likely to be higher. We suggested this as an explanation for a counterintuitive result we found when we tested birds in the following way: they were presented with a red or a yellow flower (half the birds got red, the others got yellow) that contained too much sucrose to empty on a single visit (Hurly and Healy, 1996). When each bird had left the flower, we refilled it and added a second. The second flower was either 3, 40, or 80 cm from the original flower, and it was either the same color or the alternative color. This second flower contained water. The counterintuitive result was that, when the flowers were of the same color, birds were more likely to choose correctly the original flower the closer the flowers were together (Fig. 5). This was particularly striking given that the closest distance was 3 cm, which meant the flowers were actually touching each other, implying high levels of spatial accuracy. When the flowers were 80 cm apart the birds tended to be more likely to choose the new flower, even though they were never rewarded for doing so (see Fig. 5). We suggested that the birds knew which was the flower that definitely contained a reward, but it was to their advantage en route to the known source of food to check out new, distant flowers (sampling) when they appeared. Plants may use the spacing of flowers as a way of manipulating their hummingbird pollinators. When flowers are close together, the plant could present empty as well as rewarding flowers, which would not dissuade the bird from continuing to forage on that plant because of the small price paid by moving to another nearby flower. On the other hand, widely spaced flowers may be more noticeable (in the sense of being surprising rather than literally more conspicuous) and thus more likely to be visited than flowers seen on previous occasions. We tested whether male rufous hummingbirds treat new and seen but previously unvisited flowers differently by presenting them with three different types of array: ‘‘free,’’ ‘‘forced,’’ and ‘‘mixed’’ (Henderson et al., 2001). This experimental design was a variation on an open-field radial maze we had first used to show that rufous hummingbirds could remember, and avoid, flowers (Healy and Hurly, 1995). In free arrays, the bird was faced with eight flowers arranged in a circle and each containing 15 l of sucrose (an amount that requires the bird to visit several flowers before his crop is filled), and he was allowed to visit four flowers before being chased

344


Fig. 5. Mean number of choices (SE) made to visit the focal flower when the alternative flower was either of the same (lighter columns) or different (darker columns) color and either 3, 40, or 80 cm from the focal flower (n = 6 males). There was a significant interaction between color and distance (p = 0.0001). When the distracter flower was a different color, there was a significant negative relationship between distance and performance (p < 0.05). (From Hurly and Healy, 1996.)

from the array. In forced arrays, there were only four flowers containing 15 l of sucrose, so he was forced to visit the flowers we chose for him. In mixed arrays, he was presented with six flowers all containing 15 l of sucrose, but allowed to visit only four. In all array types he was prevented from returning for at least 5 min (this was the minimum retention interval; he could return at any time after 5 min), and his choice of flowers on this return visit was recorded. When he did return he was presented with eight flowers, four containing sucrose and four empty. In the free trials, the sucrose-containing flowers were the four flowers he had seen previously but not visited; in the forced trials, the rewarded flowers were all new; in mixed trials, two of the rewarded flowers were ones he had seen and not visited while two were new. There were no differences in the performances of the birds on the different trial types, but on the mixed trials birds were much more likely to visit new flowers than flowers they had seen but not visited on their earlier trip to the array. It appears as if birds are not only remembering and avoiding recently emptied flowers, they are also taking


345

note of flowers in the vicinity that have yet to be visited. It may be that not only are new flowers more noticeable but they, by their very recency, are less likely to have had their contents reduced or lost to dehydration or to other nectarivores. Once the location of the flowers is learned, radical visual changes can be made and the bird ignores them (e.g., Miller et al., 1985; Healy and Hurly, 2001; Hurly and Healy, 1996). Although we have not quantified it, we saw a dramatic example of how major the visual changes could be: when arrays with the flowers spaced at 80-cm intervals were moved in the experiments described in Healy and Hurly (1998), on a number of occasions birds flew to, and hovered, at apparently the appropriate (in x, y, and z coordinates) place for a rewarded flower that was now some 160 cm distant. After failing to gain reward at this location, the bird often flew off, apparently oblivious to the presence of the shifted array, even though there were no visual obstacles to prevent him seeing it. This observation led us to investigate the use of a spatial component that has rarely been examined in any animal: the use of the z coordinate when encoding or remembering a spatial location. Wolf and Hainsworth (1990) and Blem et al. (1997) both found that hummingbirds preferred higher rather than lower flowers. Blem et al. favored a predation avoidance interpretation, but we suggested that higher flowers might be more conspicuous as they were less likely to be visually impeded by undergrowth. We also suggested that height variation among flowers might make each more discriminable as well as memorable. As a preliminary investigation into the role that flower height plays in hummingbird foraging, we presented birds with either four or eight flowers, in forced or free trial types as described above, in four x–y locations at one of two heights, in another variation of the open-field radial maze (Henderson et al., 2001). In both trial types the expectation was that, if the bird could remember which flowers he had recently emptied, he would avoid those four and visit the four remaining flowers (in the free trials) or the four new ones (in the forced trials). In support of the finding by Blem et al., we found a slight preference for visiting the higher flowers when the bird returned to the array in free trials, but this was not the case in forced trials. Birds were, however, significantly better than random chance at avoiding the flowers they had recently emptied, a result that would have been obscured if the birds had strong height preferences. Although they tended to avoid emptied flowers slightly more on free trials than on forced, this difference was not significant. As the bird can choose when to return to the array (in this case, the range was between 5 and 32 min), we also looked at whether or not performance was affected by the length of retention interval. For this range of durations at least, performance did not decline with increasing time since the bird’s first

346


visit. If the birds do remember which flowers have been recently emptied, it would be advantageous to avoid them at least until the flower has refilled to a significant level. In this experiment we did not have control trials in which height was not varied, so we cannot determine whether birds are better able to remember a three-dimensional (3-D) location than a twodimensional (2-D) one. A direct comparison of performance on 2-D and 3-D mazes has not been carried out in any species as yet, although Brown (1994) and Brown and Gass (1993) presented birds with vertical arrays of feeders and we have presented birds with horizontal arrays of flowers. This may be, in part at least, because the classic animal used to investigate spatial memory, the rat, lives a largely two-dimensional life. This 2-D life notwithstanding, when Grobe´ty and Schenk(1992) tested rats in a 3-D maze and also compared performance in 2-D mazes that were either horizontal or vertical, they found that in both the maze types with a vertical component, the rats appeared to pay more attention (measured as making fewer errors) to the vertical dimension than to either of the horizontal dimensions. This they interpreted as being due to an increase in the energetic demands of climbing up or down for these animals, which it would pay them to minimize relative to moving in the horizontal plane. This problem of energetic cost is also one faced by flying birds. In species such as hummingbirds it is enormously costly to hover (Chai et al., 1999; Chai and Millard, 1997), while for other avian species in flight it is costly to climb (Piersma et al., 1997; Thomas and Hedenstrom, 1998) as it may well also be for terrestrial arboreal animals. In fish, except for bottom dwellers such as catfish, however, moving up and down in the water column entails no more energetic effort than moving in the horizontal plane. Alternatively, it may be that moving vertically is novel for the rats and they pay more attention to novel information than do animals that regularly traverse a 3-D environment. It remains to be seen, then, whether there is a difference between aerial/ terrestrial animals and fish in the amount of attention they pay to the vertical dimension relative to the horizontal dimensions and whether this is related to novelty or to energy expenditure. Ecological and morphological variation aside, it seems plausible to us that locations that are differentiable in three dimensions rather than just two should be more memorable. However, all these possibilities require experimental testing. It seems plausible that, like the horizontal components, the vertical component to a location may be encoded with regard to spatial scale, that is, flowers that are close together are remembered relative to other flowers on the plant while flowers that are vertically far apart may be remembered relative to landmarks beyond the plant. Wiegmann et al. (2000)


347

investigated relational and absolute height learning in bumble bees (Bombus impatiens), by training two groups of bees to forage from a flower at the same height. For the bees in one of these groups the rewarded flower was presented alongside a higher, empty flower. The test was the presentation of two unrewarded flowers, one at the trained height and the other at a lower height. The bees with experience of flowers at only one height chose the higher flower, that is, the one at the training height. Of the bees in the group exposed to flowers of two different heights, those that had never visited the unrewarded flower during training also visited the higher flower in the test, but those bees that had visited both during training chose the lower flower in the test. Some of the bees, then, appeared to have encoded the correct height of the flower in relation to its immediate surroundings. This is somewhat akin to the way in which the hummingbirds seem to encode locations that are less than 40 cm apart horizontally (as were these bee flowers vertically). We have yet to test whether the hummingbirds treat the vertical component to a flower’s location in the way these bees did.

III. Cognitive Maps One of the aspects of spatial cognition we have not dealt with explicitly in our experiments on hummingbirds is that of cognitive mapping. There are a number of methods by which animals may navigate from one location to another. The methods for which there is the most convincing evidence are those that might be generally labeled as ‘‘route retracing,’’ for example, following a chemical trail, or path integration, which does not require literal retracing but does entail use of distances and directions from the outward journey to calculate the current position. Occupying something of a superior position in the hierarchy of navigation methods is cognitive mapping (see Healy et al., 2003, for a review). Many aspects of cognitive mapping have been debated, not least its definition, but most agree that a cognitive map is a neural representation of the spatial aspects of an animal’s familiar environment. The especial advantage of possessing such a map of one’s environment, over other navigation techniques, is that an animal is able to plan optimal routes, and, more especially, to take shortcuts and detours. Whether nonhuman animals have, or use, cognitive maps has been hotly disputed, particularly in the hymenopterans, sparked off by the claim that bees used cognitive maps during foraging (Gould, 1986; and see, e.g., Dyer, 1991; Kirchner and Braun, 1994; Menzel et al., 1998). Thus far we claim that rufous hummingbirds use spatial memory. We might also claim that

348


the ability to take short cuts and to plan routes around his territory, before foraging flights, would also be advantageous to a male whose days are consumed by watching out for females and in intense territorial defense. The result we discussed earlier, that birds more readily learn rewarded locations when these are nearest neighbors, however, is consistent with an alternative possibility, which has been proposed for navigation in the hymenopterans. This is that these insects use ‘‘snapshots’’ or visual images of their environment, either alone or in a series, to return to rewarded locations or home (Cartwright and Collett, 1983). Each snapshot is paired with a compass direction and therefore when the current view matches a snapshot in the album, the animal knows which way to go, either to reach home or the next snapshot. We have explored this issue little to date, due, in part, to the difficulty of presenting birds with hidden goals, the most common method of determining the mechanism by which the animal reaches a goal. Some clever testing on the possible use of cognitive maps by animals has been done using Clark’s nutcrakers, Nucifraga columbiana (Gibson and Kamil, 2001; see also review in Healy et al., 2003). In these experiments, the birds were trained to locate a goal hidden within a large (diameter, 132 cm; height, 154 cm) steel cylinder that had an opening along 90 of its wall to allow the bird to enter. The cylinder was placed in a larger experimental room (4.4 3.0 m), such that the birds, when at the goal in the cylinder, could see only some of the landmarks placed around the room. The test involved turning the cylinder such that some, or none, of the landmarks previously seen at the goal, could be seen during the test. The supposition was that if the birds were using a cognitive mapping strategy they would use their memory of room/landmark panorama to localize the goal, even though at test the panorama only partly matched (or did not match at all) the training view. Under most of the test conditions the birds were accurate at locating the goal, but Gibson and Kamil are cautious about concluding that the birds use a cognitive map mechanism, as the birds’ searching patterns are also consistent with calculating distance (from the wall of the cylinder) and direction (from the landmarks outside the cylinder, before entering) independently. The interpretation of the results of this study, beautifully designed as it is, demonstrates the difficulty of determining conclusively whether animals use a cognitive map. We used a similar kind of apparatus to examine whether our birds used a simpler technique than a mapping strategy to locate the reward flowers in an array, that of a learned route. Birds were trained to fly into a large cage of semirigid netting that enveloped an array of 16 flowers, 40 cm apart, 3 of which contained sucrose solution. In one treatment there was only one entrance to the cage and this remained the same throughout training, while


349

in the second treatment there was an entrance on each of the four sides of the cage and the birds could choose through which they wanted to enter and leave. Once the birds had reached a criterion level of performance, they were tested with only one of the cage doors open. For the birds trained with only one door open, the test door was a different one. In all cases, the birds could view the surrounding landscape through the walls and roof of the cage. Although these birds do not normally forage in a set movement pattern, it may be that when a location is frequently visited, a similar approach flight path is taken. The hymenopterans are well known for using stereotyped approach directions when returning to rewarded locations, and we thought it possible that the hummingbirds were doing something similar, even though they do not use the ‘‘turn back and look’’ strategy used by bees when learning both about the location of the have and of rewarding flowers (e.g., Capaldi et al., 2000; Collett and Zeil, 1998). In this experiment we assumed that the birds trained with a single door open would be more disrupted by the test condition (i.e., where they are forced to use a different entrance and therefore final approach direction) than those trained with all the doors open. In fact, all the birds were no better than chance at locating the reward flowers on the test. Whatever the explanation for their behavior, they showed no hint of using a cognitive map strategy (i.e., in this case, using the landmarks surrounding the cage to localize the rewarded flowers).

IV. Timing Although we have focused much of our attention on the ability of rufous hummingbirds to learn and to remember visual or spatial information about the flowers they feed on, an integral component of their ability to avoid flowers they have emptied recently is the capacity to remember when it was that they visited that flower. The question of whether animals measure time now encompasses a huge literature, much of which is especially concerned with seasonal changes such as those involved in breeding and migration (for reviews see Aschoff, 1989; Gallistel, 1990; Gwinner, 1986). While hummingbirds undoubtedly carry out some kind of endogenous tracking of circannual changes, it is their ability to time shorter durations in which we are specifically interested. There are thought to be two different mechanisms for timing such shorter durations: phase sense or timing, by which the animal can learn the time of day (as in a 24-h period) at which an event regularly occurs, and interval sense or timing, by which an animal can learn a fixed duration between two events that themselves can occur at any time during the day.

350


The ability to associate time of day with a specific location (time–place learning) has been seen in foraging both in the field (e.g., oystercatchers, Haematopus ostralegus, foraging on tidal mudflats; Daan Koene, 1981) and in the laboratory (e.g., garden warblers, Sylvia borin; Biebach et al., 1989). It is possible that this is what hummingbirds are doing if they are defending their territories by exploitation as we discussed earlier, that is, they go to the edges of their territories each morning in order to deplete the flowers there. The question as to whether they might be able to remember when they emptied flowers, especially those that are not on their territory boundaries, seems, however, to be an interval timing problem. Timing of intervals is thought to be used when there is no fixed time of day associated with a significant event, as with, for example, emptying a flower. Rather, when an appropriate interval has elapsed (e.g., 3–4 h), the bird could then return to a refilled flower. Each flower, or plant, would be associated with a timer that is started once the bird leaves the flower. There are at least two issues that arise here with regard to proposing that hummingbirds use interval timing in deciding which flowers to avoid: (1) a bird doing this accurately would have to keep track of at least several dozen, and possibly as many as several hundred, timers all running concurrently; and (2) as the precision of interval timing is proportional to the interval being timed, the ability of the bird to time flowers that were emptied several hours previously (the time thought to elapse before total refilling; Armstrong et al., 1987) is likely to be quite poor. That they are able to measure time at all was shown by Gill (1988) in long-tailed hermit hummingbirds (Phaethornis superciliosus), which were at least capable of learning 10- and 15-min fixed interval reinforcement schedules. Gill presented birds in the field with up to three feeders, which were refilled at fixed intervals (FI, either 10 or 15 min) after the last successful visit. In addition, Gill doubled the amount of nectar if the FI passed without the feeder being emptied (e.g., at 15 min on a 10-min FI schedule) and emptied the feeder if the bird did not return within 20 min after the FI. One bird that dominated one of the feeders returned after the feeder had been refilled and so obtained nectar on 88% of visits. The return intervals of this bird also increased to longer than the FI schedule, which increased the amount of nectar obtained on each visit. At one of the feeders the dominant male had to compete at reasonably regular intervals with another male. The visits of the dominant male were characterized by losses of two kinds, either losing to his competitor or returning too early so that competitive losses were negatively correlated with timing losses. These hummingbirds, then, under conditions of no competition such as a territorial bird might face, can time intervals of 10 and 15 min, and they learned to do so within a morning.


351

Fig. 6. The time intervals between visits to eight flowers in an array by one bird. Four of the flowers were refilled 10 min (open columns) after the bird had emptied them and four were refilled 20 min afterward (solid columns). The data for this bird were collected for approximately 8–10 h/day for 11 days (number of visits to each type of flower: 10 min, n = 1937; 20 min, n = 1343).

On the basis of this experiment by Gill, we presented rufous hummingbirds with a similar but slightly more extensive array of eight flowers each containing 20 l of sucrose, four of which were refilled on a 10-min FI schedule and four on a 20-min FI schedule. Depending on the time intervals between visits, a bird can be expected to take at least 100 l on average per visit. The territorial male was allowed to visit the array as and when he wished throughout the day (approximately 8–10 h/day). The experiment continued for 10 days, and three males were tested. All three appeared to have learned the relevant FI schedule (see Fig. 6 for an example of data from one bird; Henderson et al., in preparation) such that, although there is considerable scatter in the data, the peak of return to the 10-min flowers was around or just after 10 min, and for the 20-min flowers around or just after 20 min. As predicted theoretically, the precision of return to the 20-min flowers is much less than for the 10-min flowers. This degree of precision in proportion to the duration of the interval being timed is a feature of interval timing, termed the ‘‘scalar property’’ (Gibbon et al., 1997). It is the case that the birds often return to our experimental arrays (not just the one presented in this experiment) after intervals of approximately 10 min, and this may have made it particularly easy for the birds to learn and to use the shorter time interval. Nonetheless, they were required to keep track concurrently of four flowers with a 10-min FI and four flowers with a 20-min FI. This requires the use of more than one or two timers, although not necessarily as many as eight as birds may have

352


coupled together flowers visited on the same bout. Keeping track of more than two intervals concurrently has not, to our knowledge, been demonstrated before this experiment although the ability to do so has been implicitly assumed in the literature on episodic memory. Episodic memory, the memory for distinct episodes in one’s life (often described as ‘‘what,’’ ‘‘when,’’ and ‘‘where’’ memory), has largely been discussed in the context of human learning and memory, not least because it is easier to test for recall of specific episodes in humans than it is in nonhuman animals (e.g., Aggleton and Pearce, 2001; Morris, 2001; Tulving, 2002). However, it has plausibly been argued that food-storing birds require such multifaceted memories in order to recall locations of caches accurately, and there is some evidence that they can, indeed, remember what they have stored, where they have stored it, and when they stored it (Clayton et al., 2001a,b). Although we have not yet explicitly tested the ability of hummingbirds to remember what, where, and when in the same experiment, the evidence from the timing experiment is that they seem to be able to remember ‘‘when’’ (at least to some degree). As other experiments by us and a number of others have shown they can remember what and where rewards are located, these hummingbirds seem, in principle, to demonstrate the three components of episodic memory. Although memory for each flower hummingbirds have fed on in the recent past is not the same kind of episode that is typically inferred in the human context, nor would it be efficient for the birds to remember this information over the long term, it would seem advantageous for birds to remember all three components with respect to each of their flower visits. Further testing of episodic or episodic-like memory in animals other than food storers would be valuable.

IV. Conclusion The major advantage of testing these hummingbirds in the field is that we may eventually gain some insight into their abilities as they use them in their daily lives. We acknowledge that much we have shown to date might have been shown equally clearly in a laboratory setting. However, in a laboratory setting we cannot be sure of the relevance of the results to animals in the field because of the constraints in spatial scale (although not true for testing many invertebrates), in temporal scale (e.g., tests of spatial memory in tits on touch screens), and in other quantitative features (e.g., number of items to be remembered). By testing the hummingbirds in the field we have managed, thus far, to assess their learning and memory abilities by using arrays of flowers on a spatial scale that is comparable to


353

that they are faced with in everyday foraging. We think, therefore, we now have a better understanding of the way in which a male rufous hummingbird learns and remembers the flowers in his territory. The next steps will involve attempting to test these birds with much greater numbers of flowers so as to assess the extent of their memory capacity, and to test them over a temporal scale that more closely matches real flower-refilling rates. It would be useful in future to attempt to match our results with comparable data from the laboratory to gain an idea of how well matched the relatively uncontrolled field testing situation and the tightly controlled laboratory situation one. We have little idea as yet of how the quantitative changes in scale (temporal, spatial, or numerical) between the two settings might affect conclusions drawn about learning and memory in any animal from laboratory testing. The rufous hummingbird may be a useful species in which to make such a comparison as they are fairly readily kept and tested in captivity.

V. Summary Their territory holding, coupled with a number of other important features of the rufous hummingbird, has allowed us to investigate the cognitive abilities of these animals in the field. These birds can remember, so as to subsequently avoid, flowers they have recently emptied. We have also found that, although these birds can learn color–reward associations, when revisiting flowers they pay more attention to the spatial location of the flowers. When flowers are close together, the birds learn flower locations with respect to other nearby flowers. They also learn faster which flowers are rewarded if these flowers are next to each other. When flowers are further apart (>40 cm), however, their locations are encoded relative to landmarks outside of the array. Once the location of a flower has been learned, birds are apparently oblivious to changes in its color pattern. These birds also pay attention to whether they have seen flowers previously, preferring to visit new flowers before those they have seen and not visited. Not only can these birds remember what they have fed from and where, they may also be able to remember when they visited. We do not know yet how well matched their timing abilities are to relevant flower-refilling rates. Acknowledgments We thank Tim Roper, Peter Slater, and Charles Snowdon for helpful comments on earlier drafts of the manuscript; NATO, NERC, and NSERC for funding the work described here; and all the research assistants who helped to collect the data.

354


References Aggleton, J. P., and Pearce, J. M. (2001). Neural systems underlying episodic memory: Insights from animal research. Philos. Trans. R. Soc. Lond. B Biol. Sci. 356, 1467–1482. Aigner, P. A. (2001). Optimality modelling and fitness trade-offs: When should plants becomes pollinator-specialists? Oikos 95, 177–184. Armstrong, D. P., Gass, C. L., and Sutherland, G. D. (1987). Should foragers remember where they’ve been? Explorations of a simulation model based on the behavior and energetics of territorial hummingbirds. In ‘‘Foraging Behavior’’ (A. C. Kamil, J. R. Krebs, and H. R. Pulliam, Eds.), pp. 563–586. Plenum Press, New York. Aschoff, J. (1989). Temporal orientation: Circadian clocks in animals and humans. Anim. Behav. 37, 881–896. Batsell, W. R., Paschall, G. Y., Gleason, D. I., and Batson, J. D. (2001). Taste preconditioning augments odor-aversion learning. J. Exp. Psychol. Anim. Behav. Process 27, 30–47. Bednekoff, P. A., Balda, R. P., Kamil, A. C., and Hile, A. G. (1997). Long-term spatial memory in four seed-caching corvids. Anim. Behav. 53, 335–341. Bene´, F. (1941). Experiments on the color preference of the black-chinned hummingbird. Condor 43, 237–242. Bene´, F. (1945). The role of learning in the feeding behavior of black-chinned hummingbirds. Condor 47, 3–21. Biebach, H., Gordijn, M., and Krebs, J. R. (1989). Time–place learning by garden warblers, Sylvia borin. Anim. Behav. 37, 353–360. Biegler, R., and Morris, R. G. M. (1996). Landmark stability: Studies exploring whether the perceived stability of the environment influences spatial representation. J. Exp. Biol. 199, 187–193. Biegler, R., McGregor, A., Krebs, J. R., and Healy, S. D. (2001). A larger hippocampus is associated with longer-lasting spatial memory. Proc. Natl. Acad. Sci. USA 98, 6941–6944. Blem, C. R., Blem, L. B., and Cosgrove, C. C. (1997). Field studies of rufous hummingbird sucrose preference: Does source height affect test results? J. Field Ornithol. 68, 245–252. Bolhuis, J. J., and Macphail, E. M. (2001). A critique of the neuroecology of learning and memory. Trends Cogn. Sci. 5, 426–433. Brown, G. S. (1994). Spatial association learning by rufous hummingbirds (Selasphorus rufus): Effects of relative spacing among stimuli. J. Comp. Psychol. 108, 29–35. Brown, G. S., and Gass, C. L. (1993). Spatial association learning by hummingbirds. Anim. Behav. 46, 487–497. Capaldi, E. A., Smith, A. D., Osborne, J. L., Fahrbach, S. E., Farris, S. M., Reynolds, D. R., Edwards, A. S., Martin, A., Robinson, G. E., Poppy, G. M., and Riley, J. R. (2000). Ontogeny of orientation flight in the honeybee revealed by harmonic radar. Nature 403, 537–540. Cartwright, B. A., and Collett, T. S. (1983). Landmark learning in bees. J. Comp. Physiol. A 151, 521–543. Chai, P., and Millard, D. (1997). Flight and size constraints: Hovering performance of large hummingbirds under maximal loading. J. Exp. Biol. 200, 2757–2763. Chai, P., Altshuler, D. L., Stephens, D. B., and Dillon, M. E. (1999). Maximal horizontal flight performance of hummingbirds: Effects of body mass and molt. Physiol. Biochem. Zool. 72, 145–155. Chittka, L., and Briscoe, A. (2001). Why sensory ecology needs to become more evolutionary—insect color vision as a case in point. In ‘‘The Ecology of Sensing’’ (F. G. Barth, Ed.). Springer-Verlag, Berlin.


355

Chittka, L., Spaethe, J., Schmidt, A., and Hicklesburger, A. (2001). Adaptation, constraint, and chance in the evolution of flower color and pollinator color vision. In ‘‘Cognitive Ecology of Pollination’’ (L. Chittka and J. D. Thomson, Eds.), pp. 106–126. Cambridge University Press, Cambridge. Clayton, N. S., and Krebs, J. R. (1994). One-trial associative memory—comparison of foodstoring and nonstoring species of birds. Anim. Learn. Behav. 22, 366–372. Clayton, N. S., Yu, K. S., and Dickenson, A. (2001a). Scrub jays (Aphelocoma coerulescensi) form integrated memories of the multiple features of caching episodes. J. Exp. Psychol. Anim. Behav. Process 27, 17–29. Clayton, N. S., Griffiths, D. P., Emery, N. J., and Dickenson, A. (2001b). Elements of episodic-like memory in animals. Philos. Trans. R. Soc. Lond. B Biol. Sci. 356, 1483–1491. Cole, S., Hainsworth, F. R., Kamil, A. C., Mercier, T., and Wolf, L. L. (1982). Spatial learning as an adaptation in hummingbirds. Science 217, 655–657. Collett, T. S., and Zeil, J. (1998). Places and landmarks: an arthropod perspective. In ‘‘Spatial Representation in Animals’’ (S. D. Healy, Ed.), pp. 18–53. Oxford University Press, Oxford. Collias, N. E., and Collias, E. C. (1968). Anna’s hummingbirds trained to select different colors in feeding. Condor 70, 273–274. Daan, S., and Koene, P. (1981). On the timing of foraging flights by oystercatchers, Haematopus ostalegus on tidal mud flats. Neth. J. Sea Res. 15, 1–22. Dallal, N. L., and Meck, W. H. (1990). Hierarchical structures: Chunking by food type facilitates spatial memory. J. Exp. Psychol. Anim. Behav. Process 16, 69–84. Davies, N. B., and Houston, A. I. (1981). Owners and satellites: The economics of territory defence in the pied wagtail, Motacilla alba. J. Anim. Ecol. 50, 157–180. Durlach, P. J., and Rescorla, R. A. (1980). Potentiation rather than overshadowing in flavoraversion learning: An analysis in terms of within-compound associations. J. Exp. Psychol. Anim. Behav. Process 6, 175–187. Dwyer, D. M., and Clayton, N. S. (2002). A reply to the defenders of the faith. Trends Cogn. Sci. 6, 109–111. Dyer, F. (1991). Bees acquire route-based memories but not cognitive maps in a familiar landscape. Anim. Behav. 41, 239–246. Elam, D. R., and Linhart, Y. B. (1988). Pollination and seed production in Ipomopsis aggregata differences among and within flower color morphs. Am. J. Bot. 75, 1262–1274. Feinsinger, P. (1976). Organization of a tropical guild of nectarivorous birds. Ecol. Monogr. 46, 257–291. Feinsinger, P., and Chaplin, S. B. (1975). On the relationship between wing disc loading and foraging strategy in hummingbirds. Am. Nat. 109, 217–224. Feinsinger, P., and Colwell, R. K. (1978). Community organization among neotropical nectarfeeding birds. Am. Zool. 18, 779–795. Flombaum, J. I., Santos, L. R., and Hauser, M. D. (2002). Neuroecology and psychological modularity. Trends Cogn. Sci. 6, 106–108. Gallistel, C. R. (1990). ‘‘The Organization of Learning.’’ MIT Press, Cambridge, MA. Garcia, J., Ervin, F. R., and Koelling, R. A. (1966). Learning with prolonged delay of reinforcement. Psychonomic Sci. 5, 121–122. Garrison, J. S. E., and Gass, C. L. (1999). Response of a traplining hummingbird to changes in nectar availability. Behav. Ecol. 10, 714–725. Gass, C. L. (1979). Territory regulation, tenure, and migration in rufous hummingbirds. Can. J. Zool. 57, 914–923. Gass, C. L., and Montgomerie, R. D. (1981). Hummingbird foraging behavior: Decision making and energy regulation. In ‘‘Foraging Behavior: Ecological, Ethological and

356


Psychological Approaches’’ (A. C. Kamil and T. D. Sargent, Eds.), pp. 159–195. Garland STPM Press, New York. Gass, C. L., and Sutherland, G. D. (1985). Specialization by territorial hummingbirds on experimentally enriched patches of flowers: Energetic profitability and learning. Can. J. Zool. 63, 2125–2133. Gibbon, J., Malapani, C., Dale, C., and Gallistel, R. (1997). Towards a neurobiology of temporal cognition: Advances and challenges. Curr. Opin. Neurobiol. 7, 170–184. Gibson, B. M., and Kamil, A. C. (2001). Test for cognitive mapping in Clark’s nutcrackers (Nucifraga columbiana). J. Comp. Psychol. 115, 403–417. Gill, F. B. (1988). Trapline foraging by hermit hummingbirds: Competition for an undefended renewable resource. Ecology 69, 1933–1942. Gill, F. B., and Wolf, L. L. (1977). Non-random foraging by sunbirds in a patchy environment. Ecology 58, 1284–1296. Girvan, J. R., and Braithwaite, V. A. (1998). Population differences in spatial learning in three-spined sticklebacks. Proc. R. Soc. Lond. B Biol. Sci. 265, 913–918. Goldsmith, T. H., and Goldsmith, K. M. (1979). Discrimination of colors by the black-chinned hummingbird, Archilocus alexandri. J. Comp. Physiol. 130, 209–220. Goldsmith, T. H., Collins, J. S., and Periman, D. L. (1981). A wavelength discrimination function for the hummingbird Archilochus alexandri. J. Comp. Physiol. 143, 103–110. Gould, J. L. (1986). The locale of honey bees: Do insects have cognitive maps? Science 232, 861–863. Grant, K. A. (1966). A hypothesis concerning the prevalence of red coloration in California hummingbird flowers. Am. Nat. 100, 85–97. Grant, V., and Temeles, E. J. (1992). Foraging ability of rufous hummingbirds on hummingbird flowers and hawkmoth flowers. Proc. Natl. Acad. Sci. USA 89, 9400–9404. Grobe´ty, M.-C., and Schenk, F. (1992). Spatial learning in a three-dimensional maze, Anim. Behav. 43, 1011–1020. Guilford, T., Nicol, C., Rothschild, M., and Moore, B. P. (1987). The biological roles of pyrazines: Evidence for a warning odor function. Biol. J. Linnean Soc. 31, 113–128. Gwinner, E. (1986). ‘‘Circannual Rhythms.’’ Springer-Verlag, Berlin. Haftorn, S. (1956). Contribution to the food biology of tits especially about storing of surplus food. III. The willow-tit (Parus atricapillus L.). Det Kgl. Norske Videnskabers Selskab 3. Hainsworth, F. R., Mercier, T., and Wolf, L. L. (1983). Floral arrangements and hummingbird feeding. Oecologia 58, 225–229. Hampton, R. R., Healy, S. D., Shettleworth, S. J., and Kamil, A. C. (2002). Neuroecologists are not made of straw. Trends Cogn. Sci. 6, 6–7. Hampton, R. R., and Shettleworth, S. J. (1996a). Hippocampus and memory in a food-storing and in a nonstoring bird species. Behav. Neurosci. 110, 946–964. Hampton, R. R., and Shettleworth, S. J. (1996b). Hippocampal lesions impair memory for location but not color in passerine birds. Behav. Neurosci. 110, 831–835. Healy, S. D. (1995). Memory for objects and positions: Delayed non-matching to sample in storing and nonstoring tits. Q. J. Exp. Psychol. B 48, 179–191. Healy, S. D., and Hurly, T. A. (1995). Spatial memory in rufous hummingbirds (Selasphorus rufus): A field test. Anim. Learn. Behav. 23, 63–68. Healy, S. D., and Hurly, T. A. (1998). Hummingbirds’ memory for flowers: Patterns or actual spatial locations? J. Exp. Psychol. Anim. Behav. Process 24, 1–9. Healy, S. D., and Hurly, T. A. (2001). Foraging and spatial learning in hummingbirds. In ‘‘Cognitive Ecology of Pollination’’ (L. Chittka and J. Thomson, Eds.), pp. 127–147. Cambridge University Press, Cambridge.


357

Healy, S. D., and Suhonen, J. S. (1996). Food retrieval by marsh tits and willow tits. Behaviour 133, 71–80. Healy, S. D., Hodgson, Z., and Braithwaite, V. A. (2003). Do animals use maps? In ‘‘Biological Basis of Navigation’’ (K. Jeffery, Ed.). Oxford University Press, Oxford (In Press.). Henderson, J., Hurly, T. A., and Healy, S. D. (2001). Rufous hummingbirds’ memory for flower features. Anim. Behav. 61, 98–106. Hilton, S. C., and Krebs, J. R. (1990). Spatial memory of four species of Parus: Performance in an open-field analogue of a radial maze. Q. J. Exp. Psychol. B 4, 345–368. Hughes, R. N., and O’Brien, N. (2001). Shore crabs are able to transfer learned handling skills to novel prey. Anim. Behav. 61, 711–714. Hurly, T. A. (1996). Spatial memory in rufous hummingbirds: Memory for rewarded and nonrewarded sites. Anim. Behav. 51, 177–183. Hurly, T. A., and Healy, S. D. (1996). Location or local visual cues? Memory for flowers in rufous hummingbirds. Anim. Behav. 51, 1149–1157. Hurly, T. A., and Healy, S. D. (2002). Spatial pattern learning by rufous hummingbirds. J. Exp. Psychol. Anim. Behav. Process 28, 209–223. Johnson, S. D., and Steiner, K. E. (2000). Generalization versus specialization in plant pollination systems. Trends Ecol. Evol. 15, 140–143. Kamil, A. C. (1978). Systematic foraging by a nectar-feeding bird, the Amakihi (Loxops virens). J. Comp. Physiol. Psychol. 3, 388–396. Kirchner, W. H., and Braun, U. (1994). Dancing honey-bees indicate the location of food sources using path integration rather than cognitive maps. Anim. Behav. 48, 1437–1441. Kodric-Brown, A., and Brown, J. H. (1978). Influence of economics, interspecific competition, and sexual dimorphism on territoriality of migrant rufous hummingbirds. Ecology 59, 285–296. Krebs, J. R. (1990). Food-storing birds: Adaptive specialization in brain and behavior? Philos. Trans. R. Soc. Lond. B Biol. Sci. 329, 153–160. Krebs, J. R., Sherry, D. F., Healy, S. D., Perry, V. H., and Vaccarino, A. L. (1989). Hippocampal specialization of food-storing birds. Proc. Natl. Acad. Sci. USA 86, 1388–1392. Lange, R. S., Scobell, S. A., and Scott, P. E. (2000). Hummingbird-syndrome traits, breeding system, and pollinator effectiveness in two syntopic Penstemon species. Int. J. Plant Sci. 161, 253–263. Lyerly, S. B., Riess, B. F., and Ross, S. (1950). Color preference in the Mexican violet-eared hummingbird, Calibri T. thalassinus (Swainson). Behaviour 2, 237–248. MacDougall-Shackleton, S. A., and Ball, G. F. (2002). Revising hypotheses does not indicate a flawed approach: Reply to Bolhuis and Macphail. Trends Cogn. Sci. 6, 68–69. Macphail, E. M., and Bolhuis, J. J. (2001). The evolution of intelligence: Adaptive specializations versus general process. Biol. Rev. 76, 341–364. Macuda, T., and Roberts, W. A. (1995). Further evidence for hierarchical chunking in rat spatial memory. J. Exp. Psychol. Anim. Behav. Process 21, 20–32. McGregor, A., and Healy, S. D. (1999). Spatial accuracy in food-storing and nonstoring birds. Anim. Behav. 58, 727–734. Meleńdez-Ackerman, E., and Campbell, D. R. (1998). Adaptive significance of flower color and inter-trait correlations in an Ipomopsis hybrid zone. Evolution 52, 1293–1303. Meleńdez-Ackerman, E., Campbell, D. R., and Waser, N. M. (1997). Hummingbird behavior and mechanisms of selection on flower color in Ipomopsis. Ecology 78, 2532–2541.

358


Mendl, M., Laughlin, K., and Hitchcock, D. (1997). Pigs in space: Spatial memory and its susceptibility to interference. Anim. Behav. 54, 1491–1508. Menzel, R., Geiger, K., Joerges, J., Muller, U., and Chittka, L. (1998). Bees travel novel homeward routes by integrating separately acquired vector memories. Anim. Behav. 55, 139–152. Miller, R. S., and Miller, R. E. (1971). Feeding activity and color preference of ruby-throated hummingbirds. Condor 73, 309–313. Miller, R. S., Tamm, S., Sutherland, G. D., and Gass, C. L. (1985). Cues for orientation in hummingbird foraging: Color and position. Can. J. Zool. 63, 18–21. Morris, R. G. M. (2001). Episodic-like memory in animals: Psychological criteria, neural mechanisms and the value of episodic-like tasks to investigate animal models of neurodegenerative disease. Philos. Trans. R. Soc. Lond. B Biol. Sci. 356, 1453–1465. Olson, D. J., Kamil, A. C., Balda, R. P., and Nims, P. J. (1995). Performance of four seedcaching corvid species in operant tests of nonspatial and spatial memory. J. Comp. Psychol. 109, 173–181. Paton, D. C., and Carpenter, F. L. (1984). Peripheral foraging by territorial rufous hummingbirds: Defense by exploitation. Ecology 65, 1808–1819. Pickens, A. L. (1930). Favorite colors of hummingbirds. Auk 47, 346–352. Piersma, T., Hedenstrom, A., and Bruggemann, J. H. (1997). Climb and flight speeds of shorebirds embarking on an intercontinental flight: Do they achieve the predicted optimal behavior? Ibis 139, 299–304. Pyke, G. H. (1981). Optimal foraging in hummingbirds: Rule of movement between inflorescences. Anim. Behav. 29, 889–896. Pyke, G. H. (1984). Optimal foraging theory: A critical review. Annu. Rev. Ecol. Syst. 15, 523–575. Raven, P. H. (1972). Why are bird-visited flowers predominantly red? Evolution 26, 274. Ridley, H. N. (1930). ‘‘The Dispersal of Plants throughout the World.’’ L. Reeve and Co, Ashford, UK. Roberts, A. D. L., and Pearce, J. M. (1998). Control of spatial behavior by an unstable landmark. J. Exp. Psychol. Anim. Behav. Process 24, 172–184. Rowe, C. (2002). Sound improves visual discrimination learning in avian predators. Proc. R. Soc. Lond. B Biol. Sci. 269, 1353–1357. Rusiniak, K. W., Hankins, W. G., Garcia, J., and Brett, L. P. (1979). Flavor-illness versions: Potentiation of odour by taste in rats. Behav. Neural Biol. 25, 1–17. Sandlin, E. A. (2000a). Foraging information affects the nature of competitive interactions. Oikos 91, 18–28. Sandlin, E. A. (2000b). Cue use affects resource subdivision among three coexisting hummingbird species. Behav. Ecol. 11, 550–559. Shettleworth, S. J. (1990). Spatial memory in food-storing birds. Philos. Trans. R. Soc. Lond. B Biol. Sci. 329, 143–151. Shettleworth, S. J. (1998). ‘‘Cognition, Evolution and Behavior.’’ Oxford University Press, Oxford. Siitari, H., Honkavaara, J., and Viitala, J. (1999). Ultraviolet reflection of berries attracts foraging birds. A laboratory study with redwings (Turdus iliacus) and billberries (Vaccinium myrtillus). Proc. R. Soc. Lond. B Biol. Sci. 266, 2125–2129. Slotnick, B. M., Westbrook, F., and Darling, F. M. C. (1997). What the rat’s nose tells the rat’s mouth: Long delay aversion conditioning with aqueous odors and potentiation of taste by odors. Anim. Learn. Behav. 25, 357–369. Stevens, T. A., and Krebs, J. R. (1986). Retrieval of stored seeds by marsh tits Parus palustris in the field. Ibis 128, 513–525.


359

Stiles, F. C., and Freeman, C. E. (1993). Patterns in floral nectar characteristics of some birdvisited plant species from Costa Rica. Biotropica 25, 191–205. Stiles, F. G., and Wolf, L. L. (1979). Ecology and evolution of lek mating behavior in the longtailed hermit hummingbird. Ornithol. Monogr. 27, 1–78. Sutherland, G. D., and Gass, C. L. (1995). Learning and remembering of spatial patterns by hummingbirds. Anim. Behav. 50, 1273–1286. Sutherland, G. D., Gass, C. L., Thompson, P. A., and Lartzman, K. P. (1982). Feeding territoriality in migrant rufous hummingbirds: Defense of yellow-bellied sapsucker (Sphyrapicus varius) feeding sites. Can. J. Zool. 60, 2046–2050. Thomas, A. L. R., and Hedenstrom, A. (1998). The optimum flight speeds of flying animals. J. Avian Biol. 29, 469–477. Thomson, J. D., Wilson, P., Valenzuela, M., and Malzone, M. (2000). Pollen presentation and pollination syndromes, with special reference to Penstemon. Plant Sp. Biol. 15, 11–29. Tulving, E. (2002). Episodic memory and common sense: How far apart? In ‘‘Episodic Memory: New Directions in Research’’ (A. Baddeley, M. Conway, and J. Aggleton, Eds.), pp. 269–287. Oxford University Press, Oxford. Valone, T. J. (1992). Information for patch assessment: A field investigation with blackchinned hummingbirds. Behav. Ecol. 3, 211–222. Vaughan, W., and Greene, S. L. (1984). Pigeon visual memory capacity. J. Exp. Psychol. 10, 256–271. Waddington, K. D., and Heinrich, B. (1979). The foraging movements of bumblebees on vertical ‘‘inflorescences’’: An experimental analysis. J. Comp. Psychol. 134, 113–117. Weiss, M. R. (1997). Innate color preferences and flexible color learning in the pipevine swallowtail. Anim. Behav. 53, 1043–1052. Weiss, M. R. (2001). Vision and learning in some neglected pollinators: Beetles, flies, moths, and butterflies. In ‘‘Cognitive Ecology of Pollination’’ (L. Chittka and J. D. Thomson, Eds.), pp. 171–190. University of Cambridge, Cambridge. Wiegmann, D. D., Wiegmann, D. A., MacNeal, J., and Gafford, J. (2000). Transposition of flower height by bumble bee foragers (Bombus impatiens). Anim. Cogn. 3, 85–89. Willson, M. F., and Whelan, C. J. (1990). The evolution of fruit color in fleshy-fruited plants. Am. Nat. 136, 790–809. Wolf, L. L., and Hainsworth, F. R. (1990). Non-random foraging by hummingbirds: Patterns of movement between Ipomopsis aggregata (Pursch) V. Grant inflorescences. Funct. Ecol. 4, 149–157. Yeaton, R. L., and Laughrin, L. (1976). Fall resource division in Santa Cruz hummingbirds. Wilson Bull. 88, 272–279.


Index

A Aggression arginine vasopressin and in golden hamsters, 286–287 nongenomic transmission of, 305–308 overview of, 284–285 in Peromyscus, 290–296, 305–307 resident–intruder, 292–293, 298 in rodents, 290–296, 305–307 in voles, 287–289 cross-generational transmission of, through behavioral mechanisms description of, 298–299 parent–offspring interactions in humans, 303–304 male retrievals, 301 maternal role, 300 in primates, 302–303 in rodents, 299–302, 304 summary overview of, 304 medial amygdala’s role in, 285 paternal behavior effects on, 301–302, 310 in Peromyscus arginine vasopressin’s role, 291–293 description of, 289 parental behavior, 302 parent–offspring interactions, 299–302, 304 paternal behavior, 289–290 resident–intruder aggression, 301–302 retrieving behavior and, 301 territoriality, 290

territorial, 272 Animal motion, individual-based models of, 4 Ant trails decision making considerations, 13–14 feedback mechanisms, 13–14 organization of, 11–12 pheromones for, 11 reinforcement of, 11 Antisymmetry, 174 Arginine vasopressin (AVP) aggression and in golden hamsters, 286–287 nongenomic transmission of, 305–308 overview of, 284–285 in Peromyscus, 290–296, 305–307 resident–intruder, 292–293, 298 in rodents, 290–296, 305–307 in voles, 287–289 paternal behavior and in Microtus, 295 overview of, 284–285 in Peromyscus, 293–295, 297 pup retrievals, 295 in voles, 287 plasticity of, 307, 310 testosterone vs. function comparisons, 297–298 parental behavior, 284 Arginine vasotocin, 285 Asocial taxis, 33 Asymmetry absolute, 177, 185 detection of, 191–192 directional, 174 environmental factors, 189–190 361

362

INDEX

Asymmetry (continued ) fluctuating. See Fluctuating asymmetry heritability of, 195 radial, 180 relative, 177 symmetry differentiation from, 192 translational, 180 Avoidance behaviors, 29–30

B Bateman gradient definition of, 132 operational sex ratio effects on, 148 Syngnathus typhle females, 143–144 Behavior collective. See Collective behaviors environmental factors that affect, 4–5 fluctuating asymmetry influences on, 194–196 inappropriate sex steroid hormones effect on, 251–252 paternal. See Paternal behavior Behavioral rules, 5 Biological realism, 22 Biparental care in Phodopus campbelli evolution of, 207–208, 219–227 juveniles, 235–236 laboratory population for studying, 219 obligate, 220–222, 242 reasons for, 207–208 spatial distribution effects, 213 in Phodopus sungorus facultative, 220 laboratory population for studying, 218 Bird flocks collective behaviors in avoidance, 29–30 description of, 21–22 collective memory in, 27–29 color–reward association in foraging patterns of, 334

computer models for, 38–39 group shape description of, 21–22 external stimuli effects, 29–34 individual behavior effects on, 24–27 information transfer in, 30–32 internal structure of, 37–39 models of, 22–24 position shifting in, 37 Reynolds model, 23 testosterone effects on paternal behavior in, 267, 270, 309 thought transference in, 22 turning in, 21–22

C Cognitive abilities field environment studies of, 326–327 hummingbird foraging as model system for studying advantages of, 327 cognitive mapping, 347–349 color of flowers innate preferences, 339 learned preferences, 329–331 red, 330–331 reward associated with, 332–333, 335–339 ‘‘defense by exploitation’’ approach, 328 emptied flowers costs of revisiting, 343 timing considerations, 349–350 learning and memory adaptive specializations of, 340–341 color–nectar reward relationship, 332–333, 335–339 description of, 329–330 episodic memory, 352 spatial, 339–347 location of flowers cognitive mapping of, 347–349

363

INDEX

conspicuous, 345 costs of revisiting flowers already pollinated, 343 description of, 340 ecological basis for remembering, 342 energetic costs associated with, 346 higher vs. lower flowers, 345 learning and memory used in, 339–347 plant spacing of flowers, 343 studies of, 341–342 optimal foraging theory, 340 plant–pollinator coevolutionary relationship, 332–333 potentiation, 339 reasons for choosing, 326–327 red object preferences, 330–331 stereotyped approach to flowers, 349 territoriality, 328 time remembrance interval timing, 349–350 phase timing, 349 traplining, 329 laboratory studies of, 325–326 natural selection influences, 326 Cognitive ecology, 326 Cognitive mapping, 347–349 Collective behaviors avoidance, 29–30 in crowds attraction to other pedestrians, 18 banding patterns, 18 collision avoidance, 16–18 emergency situations, 20–21 environment geometry effects, 18–19 dynamics of, 3–5 environmental factors, 4–5 examples of, 1–2 home ranges, 15–16 migrating wildebeest herds as example of, 6–9 stochastic events, 5

summary of, 66 trail use. See Trails Collective memory, 27–29 Conspecifics, 59, 120 Correlational model, 46 Cortisol, 238, 244–245 Crowd behavior attraction to other pedestrians, 18 banding patterns, 18 collision avoidance, 16–18 emergency situations, 20–21 environment geometry effects, 18–19

D Developmental instability description of, 169–170 genetic influences, 183 heritability of, 185 individual asymmetry parameters, 181 studies of, 178–181 Developmental stability definition of, 169 fluctuating asymmetry as indicator of, 169–170 as organism-wide phenomenon, 181 ‘‘Dilution effect,’’ 61 Dominance double-reinforcement, 45–46 hierarchies of, 45 self-organized structuring effects, 44–47 Dwarf hamster. See Phodopus campbelli; Phodopus sungorus

E Emergency situations, crowd behavior in, 20–21 Endocrine system, in Phodopus campbelli females adaptive value of, 231–235 description of, 229–230 hormonal shifts, 243

364

INDEX

Endocrine system (continued ) Phodopus sungorus females vs., 233–234 pituitary gland, 234 progesterone, 230–231 prolactin, 231, 234, 244 species differences, 230–231 testosterone levels, 245–246 Environment asymmetry created by, 189–190 behavioral effects of, 4–5 spatial heterogeneity of, 5 trails as modification of. See Trails Episodic memory, 352 Estradiol, 243 Ethological isolation, 118 Eulerian models, 3–4, 15

F Fathers. See also Paternal behavior; Paternal care human cortisol levels, 247 estradiol levels, 247 hormone dynamics in, 246–250 testosterone levels, 247–250 Phodopus campbelli cortisol levels, 244–245 description of, 243–244 estrogen levels, 251 hormonal changes, 244–246 prolactin levels, 244 Females Nerophis ophidion. See Nerophis ophidion, females ornamentation, in Syngnathus typhle amplifier function of, 152–154 attractiveness function of, 154–155 benefits of, 158–159 body size, 133 competitive displays of, 156–157 costs of, 158–159 description of, 133, 151–152

dietary pigment signals, 134 fecundity and, 154 female quality and, 154 female–female competition uses of, 154–159 predation risks, 158 social costs of, 134 summary overview of, 161–162 parental investment by, 132–133 Phodopus campbelli. See Phodopus campbelli, females Phodopus sungorus. See Phodopus sungorus, females reproductive success in, 132 Fish. See also Pipefish estradiol levels in, 273 paternal care by, 144 predatory, 34–35 testosterone effects on paternal behavior in, 268, 272–274 Fish schools collective behaviors in avoidance, 29–30 description of, 21–22 collective memory in, 27–29 computer models for, 38–39 decision making in, 23–24 group shape description of, 21–22 external stimuli effects, 29–34 individual behavior effects on, 24–27 parasite effects, 34 in predatory fish, 34–35 information transfer in, 30–32 internal structure of, 35–37 klinotaxis use by, 33–34 members differences in, 39–40 spatial positioning of, 39–40 models of, 22–24, 40–41 position shifting in, 37, 39–40 self-organized structuring in description of, 42, 44 social dominance and, 44–47 social dominance in, 44–47 spatial positions in

365

INDEX

behavior heterogeneity effects on, 40–42 description of, 35–39 parasite effects, 44 phenotypes, 44 self-organized mechanisms for determining, 42, 44 stimulus–response behaviors, 23 ‘‘waves of excitation’’ in, 22 Fluctuating asymmetry adaptive significance of, 171, 186–187 animal behavior and, 194–196 antisymmetry vs., 174 causes of, 169 characteristics of, 191 definition of, 169 developmental instability and description of, 169–170 genetic influences, 183 heritability of, 185 individual asymmetry parameters, 181 studies of, 178–181 developmental noise and, 187 directional asymmetry vs., 174 environmental effects, 189 evolutionary effects, 194–195 fitness of populations and description of, 170–171 relationship between examples of, 172 inconsistency in, 172 indirect associations, 173 studies of. See Fluctuating asymmetry, studies of genetic origins of, 182–184 genetically related, 179 heritability of, 184–186, 195–196 identification of, 174–175 interest in, 170 Klingenberg and Nijhout model of, 183 low adaptive significance of, 186–187 nongenetic benefits of, 188–190 preference for, 191

neutral character of, 187, 194 origins of, 182–184 quantitative trait locus mapping of, 182, 184 repeatability of, 180 sexual selection and nongenetic benefits of low fluctuating asymmetry, 188–190 overview of, 187–188 perceptual processes, 191–193 studies of description of, 188 recommendations for, 193–194 visual role, 192 stressor effects, 169 studies of absolute vs. relative asymmetry, 177 in barn swallows, 195 controversies regarding, 171–172 identification issues, 174–175 measurement error, 176 methodology issues, 173–178 sampling, 176–177 statistical models, 177–178 statistical power analysis, 177 trait selection, 175–176 summary overview of, 196–197 as trait, 181–182 Whitlock model, 178–179 Frontal bias, 48

G Genetic relatedness, in rodents definition of, 100 description of, 91–92 differential responses to kin and nonkin based on, 95–96 odor similarities as basis for, 94–95 Golden hamsters, arginine vasopressin effects on aggression in, 286–287 G-ratios, for establishing kin adaptive responses, 112, 115, 119 definition of, 99, 103

366

INDEX

G-ratios (continued ) description of, 79 evidence to support, 105 future studies of, 119–123 individual recognition by association vs., 105–106 neurophysiological basis of, 123 as premating isolating mechanism, 117 prenatal learning effects on, 121–123 self-referent matching vs., 100–101 species-specific sex attractants or excitants vs., 120–121 Group benefits of, 61 cohesion of, 4, 55 correlational model, 46 cutoff size of, 51–53 ‘‘dilution effect,’’ 61 disadvantages of, 61 fragmentation of, 56 frontal bias in, 48 global level dynamics of, 4 heterogeneity of, 57, 59 individuals sorted based on size, 60–61 information transfer in, 30–32 internal structure of, 37–39, 66 leadership of, 47–48 motion of, 6–9 phenotypic sorting of, 44, 56–61 probability of being in, 64 range of interaction, 64 self-organization in, 60, 65 self-sorting in, 59, 67 shape of description of, 6–9 fish schools description of, 21–22 external stimuli effects, 29–34 individual behavior effects on, 24–27 parasite effects, 34 in predatory fish, 34–35 individual behavior effects, 24–27 in wildebeest herds, 6–9 size of context-dependent, 62–64

cutoff, 51–53 decreases in, 50–51 ecological influences, 61–62 fission–fusion processes effect on, 49–56 habitat structure effects on, 53–56 long-tailed distributions, 51 ‘‘maximum entropy" principle, 51 optimal, 61–66 seasonal variations in, 65 self-regulation of, 64 truncated power law, 51 space use by, 53–55 spatial positions in behavior heterogeneity effects on, 40–42 description of, 35–39 self-organized mechanisms for determining, 42, 44 speed of individual and, 42 splitting of, 50 turbulence effects, 55

H Habitat changes in, 55 group size and, 53–56 Habituation–discrimination technique, 81–82 Habituation–generalization technique, 82–85 Hamsters arginine vasopressin effects on aggression in, 286–287 Phodopus campbelli. See Phodopus campbelli Phodopus sungorus. See Phodopus sungorus ‘‘Herding effect,’’ 21 Herding model, 40–41 Hierarchy dominance and, 45 formation of, 46 ‘‘steepness’’ of, 46–47

367

INDEX

Human(s) aggression in, parent–offspring interactions effect on, 303 collision avoidance by, 16–18 couples, 250–251 crowd behavior attraction to other pedestrians, 18 banding patterns, 18 collision avoidance, 16–18 emergency situations, 20–21 environment geometry effects, 18–19 expectant fathers cortisol levels, 247 estradiol levels, 247 hormone dynamics in, 246–250 testosterone levels, 247–250 parent–offspring interactions in, 303–304 self-organization approach, 2–3 symmetry preference by, 190, 193 testosterone effects on paternal behavior in, 282–283 trails created by description of, 9–11 reinforcement of, 11 Hummingbird foraging, for studying cognitive abilities advantages of, 327 cognitive mapping, 347–349 color of flowers innate preferences, 339 learned preferences, 329–331 red, 330–331 reward associated with, 332–333, 335–339 ‘‘defense by exploitation’’ approach, 328 emptied flowers costs of revisiting, 343 timing considerations, 349–350 learning and memory adaptive specializations of, 340–341 color–nectar reward relationship, 332–333, 335–339 description of, 329–330

episodic memory, 352 spatial, 339–347 location of flowers cognitive mapping of, 347–349 conspicuous, 345 costs of revisiting flowers already pollinated, 343 description of, 340 ecological basis for remembering, 342 energetic costs associated with, 346 higher vs. lower flowers, 345 learning and memory used in, 339–347 plant spacing of flowers, 343 studies of, 341–342 optimal foraging theory, 340 plant–pollinator coevolutionary relationship, 332–333 potentiation, 339 reasons for choosing, 326–327 red object preferences, 330–331 stereotyped approach to flowers, 349 territoriality, 328 time remembrance interval timing, 349–350 phase timing, 349 traplining, 329 Hypothalamic-pituitary-adrenal axis, 305

I Information transfer, 30–32 Interindividual interaction, 1 Interval timing, 349–350

K Kin definition of, 77 discriminative mechanisms for identifying importance of, 77

368

INDEX

Kin (continued ) odors. See Kin, odor-based discrimination recognition alleles, 78–79 spatial distribution, 78 nonkin vs., 77–78 odor-based discrimination of description of, 95–96 G-ratios. See G-ratios individual recognition by association description of, 92–94, 103 G-ratios vs., 105–106 phenotype matching vs., 103–104 self-referent matching, 96–102 Klinotaxis, 33

L Laboratory setting cognitive ability testing in, 325–326 learning and memory testing in, 325 Lagrangian model, 4 Leadership of group, 47–48 Learning, hummingbird foraging studies of adaptive specializations of, 340–341 color–nectar reward relationship, 332–333, 335–339 description of, 329–330 episodic memory, 352 spatial, 339–347

M Major histocompatibility complex, 107 Males energy expenditure in offspring, 133 Nerophis ophidion. See Nerophis ophidion, males nonpaternal, 132–133 operational sex ratio in, 131

Phodopus campbelli. See Phodopus campbelli, males sexual selection in, 131 surplus of, 132 Mathematical modeling, 3 Mating competition description of, 131–132 operational sex ratio as measure of, 131 ‘‘Maximum entropy’’ principle, 51 Memory episodic, 352 hummingbird foraging studies of adaptive specializations of, 340–341 color–nectar reward relationship, 332–333, 335–339 description of, 329–330 episodic memory, 352 spatial, 339–347 Modeling mathematical, 3 stochastic, 5

N Nerophis ophidion characteristics of, 136–137 females body size preferences, 148 competition among, 146 energy provided to offspring, 141 reproductive rates, 144 males body size preferences, 148 choosiness of, 148 energy provided to offspring, 141 ornamented female preference of, 148 reproductive rates, 144 mating in, 137, 139–140 parental investment, 141 reproductive rates, 141–142 sexual dimorphism, 137

INDEX

O Odor(s) biological states that affect, 108–109, 120 coloration and, role of potentiation in, 339 cues, 109–110 familiarity with, 112–114 genotype in description of, 90–91, 109 metabolic processes of, 91 phenotypic variations expressed in, 115–116 individual types of, 107–109 markers, 109–110, 118 phenotype matching using, 103–104 recognition of, 111–112 signals, 109–110 species-specific variations in, 107–108 Odor maps, 102 Odor–based discrimination of kin vs. nonkin cues, 109–110 description of, 95–96 G-ratios adaptive responses, 112, 115, 119 as premating isolating mechanism, 117 definition of, 99, 103 description of, 79, 99, 103, 105 evidence to support, 105 future studies of, 119–123 individual recognition by association vs., 105–106 neurophysiological basis of, 123 preferential responses based on, 114–119 as premating isolating mechanism, 117 prenatal learning effects on, 121–123 self-referent matching vs., 100–101 species-specific sex attractants or excitants vs., 120–121 individual odors, 107–109 individual recognition by association description of, 92–94, 103

369

G-ratios vs., 105–106 phenotype matching vs., 103–104 markers, 109–110 self-referent matching, 96–102 signals, 109–110 terminology associated with, 106–114 Odor–genes covariance definition of, 91, 124 description of, 80 differential responses to kin and nonkin based on odor description of, 95–96 self-referent matching, 96–102 genetic relatedness definition of, 100 description of, 91–92 differential responses to kin and nonkin based on, 95–96 odor similarities as basis for, 94–95 kin recognition by chemosensory cues in urine, 80–81 description of, 80 mole rat studies habituation–discrimination technique, 81–82 habituation–generalization technique, 82–85, 90 kinship established by odor similarities, 85–88 odor similarities across species, 89–90 overview of, 80–81 odor similarities across species, 89–90 genetic relatedness studied using, 88–89 in populations and species, 88–89 odorant, factors that affect, 80–81 principles of, 90–91 Offspring parental interactions, effect on cross-generational transmission of aggression in humans, 303–304 male retrievals, 301 maternal role, 300

370

INDEX

Offering (continued ) in primates, 302–303 in rodents, 299–302, 304 summary overview of, 304 paternal behavior effects on aggression in, 301–302 Operational sex ratio Bateman gradient affected by, 148 definition of, 131 Ornamentation, female amplifier function of, 152–154 attractiveness function of, 154–155 benefits of, 158–159 body size, 133 competitive displays of, 156–157 costs of, 158–159 description of, 133, 151–152 dietary pigment signals, 134 fecundity and, 154 female quality and, 154 female–female competition uses of, 154–159 predation risks, 158 social costs of, 134 summary overview of, 161–162

P Parasites, sexual selection effects of, 149–150 Parental behavior and care. See also Biparental care arginine vasopressin vs. testosterone, 284 parent–offspring interactions effect on cross-generational transmission of aggression in humans, 303–304 male retrievals, 301 maternal role, 300 in primates, 302–303 in rodents, 299–302, 304 summary overview of, 304 in Peromyscus, 302 Paternal behavior. See also Fathers

aggression and, 301–302, 310 arginine vasopressin effects description of, 263, 284 in Microtus, 295 overview of, 284–285 in Peromyscus, 293–295, 297 pup retrievals, 295 testosterone vs., 284 in voles, 287 description of, 263 offspring survival benefits, 264 in Peromyscus arginine vasopressin effects on, 293–295, 297 plasticity role in, 304–305 testosterone effects in amphibians, 268, 271–272 arginine vasopressin vs., 284 in birds, 267, 270 description of, 263, 266 in fish, 268, 272–274 in humans, 282–283 in primates, 269, 281–282 in rodents castration effects, 277–278, 280 correlational studies, 275–277 description of, 268–269, 274–275 hormone manipulation studies, 277–280 summary of, 281 summary overview of, 283–284 Paternal care, in Phodopus campbelli birthing, 208, 225 cortisol levels, 238 environmental stimuli, 240–241 estradiol secretion, 243 estrogen levels and, 237–238 hormonal influences, 237–240 indirect, 225, 227 maternal benefits of, 225 prolactin levels, 238 pup benefits, 227 testosterone decreases associated with, 240 in wild, 208–209 Pattern formation, 2

INDEX

Pedestrians crowd behavior attraction to other pedestrians, 18 banding patterns, 18 collision avoidance, 16–18 emergency situations, 20–21 environment geometry effects, 18–19 traffic patterns of, 16 Peromyscus aggression in arginine vasopressin’s role, 291–293, 305–307 description of, 289 maternal retrieval effects on, 306 nongenomic transmission of, 305–307, 305–308 parental behavior, 302 parent–offspring interactions, 299–302, 304 paternal behavior, 289–290 paternal retrieval effects on, 306 resident–intruderaggression,301–302 retrieving behavior and, 301 territoriality, 290 paternal behavior in arginine vasopressin effects on, 293–295, 297 plasticity role in, 304–305 Phase timing, 349 Phenotype asymmetry detection of, 191–192 fitness and, 189 genotypic variations expressed in, 115–116 group dynamics and, 44, 56–61 kin recognition by matching odor with, 103–104 Phodopus campbelli aggression in, 221 ancestry of, 210 biparental care evolution of, 207–208, 219–227 juveniles, 235–236 laboratory population for studying, 219

371 obligate, 220–222, 242 reasons for, 207–208 spatial distribution effects, 213 chemical communication by, 216–218 endocrine evolution in hormone sampling, 228–229 ‘‘opposite sex’’ hormones on behavior, 251–252 overview of, 227–228 sex specificity, 251–252 evolutionary history of, 241–243 females adaptive changes, 241, 243 birthing delay by, 234 developmental delay by, 234–235 endocrine evolution in adaptive value of, 231–235 description of, 229–230 hormonal shifts, 243 Phodopus sungorus females vs., 233–234 pituitary gland, 234 progesterone, 230–231 prolactin, 231, 234, 244 species differences, 230–231 testosterone levels, 245–246 energy availability effects, 224–225 estrous cycle in, 229–230 maternal transition for, 236–237 water availability for, 224 juveniles alloparental care by, 235 infanticide by, 235–236 males home range of, 216 juveniles caregiver transition of, 236 hormonal changes associated with paternity, 236–240 infanticide by, 235–236 mating effects on, 241 new and expectant cortisol levels, 244–245 description of, 243–244 estrogen levels, 251

372

INDEX

Phodopus campbelli (continued ) hormonal changes, 244–246 prolactin levels, 244 paternal care by birthing, 208, 225 cortisol levels, 238 environmental stimuli, 240–241 estradiol secretion, 243 estrogen levels and, 237–238 hormonal influences, 237–240 indirect, 225, 227 maternal benefits of, 225 prolactin levels, 238 pup benefits, 227 testosterone decreases associated with, 240 in wild, 208–209 retrieval of pups by, 209 natural habitat of description of, 210–211 social behavior in, 214–215 natural history of, 208–218 overview of, 207–208 pair-bond disruption in, 222 phylogenetics of, 210 physiological adaptations of, 227–228 pregnancy block, 220–221 pups growth of, 223 male retrieval of, 209 range of, 210, 212 reproduction in adaptations, 243 energy availability effects, 224–225 heat effects on, 223–224 physiological constraints, 222–225 social behaviors, 222 time limitations, 222–223 water availability effects, 224 scent marks by, 216–218, 221 social behaviors description of, 222 in wild, 214–215 social isolation in, 221 sociality of, space use effects on, 212–213

survival adaptations, 211–212 in wild capture methods, 215 male paternal care, 208–209 social behavior in, 214–215 Phodopus roborovskii, 210 Phodopus sungorus ancestry of, 210 biparental care facultative, 220 laboratory population for studying, 218 chemical information effect on behavioral responses in, 218 females aggression patterns in, 221 progesterone secretion in, 230–231 prolactin secretion in, 231 habitat of, 210–211 paternal care by birthing, 208 pup retrieval, 209 phylogenetics of, 210 range of, 210, 212 sociality of, space use effects on, 212–213 survival adaptations, 211–212 Pipefish egg brooding in, 135 Nerophis ophidion. See Nerophis ophidion Syngnathus typhle. See Syngnathus typhle Population-level processes, modeling of, 3–4 Potentiation, 339 Predatory fish, group shape in, 34–35 Prenatal learning, effect on G-ratios, 121–123 Primates parent–offspring interactions in, 302–303 testosterone effects on paternal behavior in, 281–282 Progesterone, 230–231 Prolactin, 231, 234

INDEX

Q Quantitative trait locus mapping, of fluctuating asymmetry, 182, 184

R Radial asymmetry, 180 Rats odor-discriminatory ability of, 14–15 trail formation by, 14–15 Recognition alleles, 78–79 Repulsion description of, 23 zone of, 36, 42, 56–57 Reynolds model, 23 Rodents. See also Phodopus campbelli; Phodopus sungorus genetic relatedness in definition of, 100 description of, 91–92 differential responses to kin and nonkin based on, 95–96 odor similarities as basis for, 94–95 odor-based discrimination of kin by. See Odor-based discrimination of kin vs. nonkin testosterone effects on paternal behavior in castration effects, 277–278, 280 correlational studies, 275–277 description of, 268–269, 274–275 hormone manipulation studies, 277–280 summary of, 281

S Scents home range generation using, 15–16 Phodopus campbelli marking using, 216–218, 221 Seahorses, sex role reversal in, 136

373

Self-organization definition of, 1–2 dominance effects, 45 group, 60, 65 vertebrate application of, 2–3, 67 Semistable solutions, 7–8 Sex role-reversed species Nerophis ophidion. See Nerophis ophidion seahorses, 136 signals of body size, 133 dietary pigment signals, 134 ornamentation. See Ornamentation Syngnathus typhle. See Syngnathus typhle Sex roles, 131–132 Sexual selection description of, 131 fluctuating asymmetry and nongenetic benefits of low fluctuating asymmetry, 188–190 overview of, 187–188 perceptual processes, 191–193 studies of description of, 188 recommendations for, 193–194 visual role, 192 parasite effects on, 149–150 trade-offs in life history traits, 159–160 Social discrimination, 78 Social taxis, 33 Spatial heterogeneity, 5 Stochastic events, 5 Symmetry. See also Asymmetry asymmetry differentiation from, 192 human preference for, 190, 193 preference for, 191 Syngnathids description of, 134–135 females, 145–146 male egg nursing by, 135 parental care by, 135

374

INDEX

Syngnathids (continued ) Nerophis ophidion. See Nerophis ophidion parental care provided by, 135 sex role reversal in, 135–136 species of, 135 Syngnathus typhle. See Syngnathus typhle Syngnathus typhle body size, 146–147 characteristics of, 136–137 color patterns of, 138–139 females Bateman gradient, 143–144, 148 body size preferences, 146–147, 152, 161 characteristics of, 140–141 competition among, 146 offspring quality based on mate choice by, 151 ornamentation amplifier function of, 152–154 attractiveness function of, 154–155 benefits of, 158–159 competitive displays of, 156–157 costs of, 158–159 description of, 151–152 fecundity and, 154 female quality and, 154 female–female competition uses of, 154–159 predation risks, 158 summary overview of, 161–162 parasitic, 149–150 reproductive rates, 142–143 striped, 152–153 summary overview of, 161 trade-offs in life history traits, 159–160 males Bateman gradient, 148 body size preferences, 146–147, 152, 158, 161 characteristics of, 140

choosiness of benefits, 150–151 description of, 147–148 dominant females, 156 offspring quality and, 150–151 operational sex ratio effects on, 147–148 ornamentation, 154 parasitic females, 149–150 reproductive rates, 142 summary overview of, 161 mating in, 137–139 operational sex ratios description of, 142–143 male choosiness affected by, 147–148 parental investment, 141 pregnancy in, 138 sexual dimorphism, 137

T Taxis behavior, 33 Testosterone aggression and competitive, 264 negative association between, 264–265 positive association between, 266 studies of, 264 in amphibians, 268, 271–272 arginine vasopressin vs. function comparisons, 297–298 parental behavior, 284 in birds, 267, 270 in human fathers, 247–250 paternal behavior and in amphibians, 268, 271–272 arginine vasopressin vs., 284 in birds, 267, 270, 309 description of, 266 in fish, 268, 272–274, 309 in humans, 282–283 in primates, 269, 281–282 in rodents

375

INDEX

castration effects, 277–278, 280 correlational studies, 275–277 description of, 268–269, 274–275 hormone manipulation studies, 277–280 summary of, 281 summary overview of, 283–284 in Phodopus campbelli fathers, 240 plasticity of, 310 social challenge effects, 264 Thought transference, in bird flocks, 22 Trails ant decision making considerations, 13–14 feedback mechanisms, 13–14 organization of, 11–12 pheromones for, 11 reinforcement of, 11 description of, 9 human, 9–11 ungulate, 11–12 vertebrate chemical deposition, 14–15 information transfer, 14

system interactions, 14 Translational asymmetry, 180 Truncated power law, of group size, 51

V Voles, arginine vasopressin effects on aggression, 287–289 pair bond formation, 287 paternal behavior, 287

W Whitlock model, 178–179 Wildebeest herds collective behaviors of, 6–9 wavelike front of, 6–9

Z Zone of repulsion, 36, 42, 56–57


Contents of Previous Volumes

The Evolution of Courtship Behavior in Newts and Salamanders T. R. HALLIDAY

Volume 18 Song Learning in Zebra Finches (Taeniopygia guttata): Progress and Prospects PETER J. B. SLATER, LUCY A. EALES, AND N. S. CLAYTON

Ethopharmacology: A Biological Approach to the Study of Drug-Induced Changes in Behavior A. K. DIXON, H. U. FISCH, AND K. H. MCALLISTER

Behavioral Aspects of Sperm Competition in Birds T. R. BIRKHEAD

Additive and Interactive Effects of Genotype and Maternal Environment PIERRE L. ROUBERTOUX, MARIKA NOSTEN-BERTRAND, AND MICHELE CARLIER

Neural Mechanisms of Perception and Motor Control in a Weakly Electric Fish WALTER HEILIGENBERG

Mode Selection and Mode Switching in Foraging Animals GENE S. HELFMAN

Behavioral Adaptations of Aquatic Life in Insects: An Example ANN CLOAREC

Cricket Neuroethology: Neuronal Basis of Intraspecific Acoustic Communication FRANZ HUBER

The Cicadian Organization of Behavior: Timekeeping in the Tsetse Fly, A Model System JOHN BRADY

Some Cognitive Capacities of an African Grey Parrot (Psittacus erithacus) IRENE MAXINE PEPPERBERG

Volume 19 Volume 20 Polyterritorial Polygyny in the Pied Flycatcher P. V. ALATALO AND A. LUNDBERG

Social Behavior and Organization in the Macropodoidea PETER J. JARMAN

Kin Recognition: Problems, Prospects, and the Evolution of Discrimination Systems C. J. BARNARD

The t Complex: A Story of Genes, Behavior, and Population SARAH LENINGTON

Maternal Responsiveness in Humans: Emotional, Cognitive, and Biological Factors CARL M. CORTER AND ALISON S. FLEMING 377

The Ergonomics of Worker Behavior in Social Hymenoptera PAUL SCHMID-HEMPEL

378

CONTENTS OF PREVIOUS VOLUMES

‘‘Microsmatic Humans’’ Revisited: The Generation and Perception of Chemical Signals BENOIST SCHAAL AND RICHARD H. PORTER

Parasites and the Evolution of Host Social Behavior ANDERS PAPE MØLLER, REIJA DUFVA, AND KLAS ALLANDER

Lekking in Birds and Mammals: Behavioral and Evolutionary Issues R. HAVEN WILEY

The Evolution of Behavioral Phenotypes: Lessons Learned from Divergent Spider Populations SUSAN E. RIECHERT

Volume 21

Proximate and Developmental Aspects of Antipredator Behavior E. CURIO

Primate Social Relationships: Their Determinants and Consequences ERIC B. KEVERNE The Role of Parasites in Sexual Selection: Current Evidence and Future Directions MARLENE ZUK Conceptual Issues in Cognitive Ethology COLIN BEER Response in Warning Coloration in Avian Predators W. SCHULER AND T. J. ROPER Analysis and Interpretation of Orb Spider Exploration and Web-Building Behavior FRITZ VOLLRATH Motor Aspects of Masculine Sexual Behavior in Rats and Rabbits GABRIELA MORALI AND CARLOS BEYER On the Nature and Evolution of Imitation in the Animal Kingdom: Reappraisal of a Century of Research A. WHITEN AND R. HAM

Volume 22 Male Aggression and Sexual Coercion of Females in Nonhuman Primates and Other Mammals: Evidence and Theoretical Implications BARBARA B. SMUTS AND ROBERT W. SMUTS

Newborn Lambs and Their Dams: The Interaction That Leads to Sucking MARGARET A. VINCE The Ontogeny of Social Displays: Form Development, Form Fixation, and Change in Context T. G. GROOTHUIS

Volume 23 Sneakers, Satellites, and Helpers: Parasitic and Cooperative Behavior in Fish Reproduction MICHAEL TABORSKY Behavioral Ecology and Levels of Selection: Dissolving the Group Selection Controversy LEE ALAN DUGATKIN AND HUDSON KERN REEVE Genetic Correlations and the Control of Behavior, Exemplified by Aggressiveness in Sticklebacks THEO C. M. BAKKER Territorial Behavior: Testing the Assumptions JUDY STAMPS Communication Behavior and Sensory Mechanisms in Weakly Electric Fishes BERND KRAMER


Volume 24 Is the Information Center Hypothesis a Flop? HEINZ RICHNER AND PHILIPP HEEB Maternal Contributions to Mammalian Reproductive Development and the Divergence of Males and Females CELIA L. MOORE

379

An Overview of Parental Care among the Reptilia CARL GANS Neural and Hormonal Control of Parental Behavior in Birds JOHN D. BUNTIN Biochemical Basis of Parental Behavior in the Rat ROBERT S. BRIDGES

Cultural Transmission in the Black Rat: Pine Cone Feeding JOSEPH TERKEL

Somatosensation and Maternal Care in Norway Rats JUDITH M. STERN

The Behavioral Diversity and Evolution of Guppy, Poecilia reticulata, Populations in Trinidad A. E. MAGURRAN, B. H. SEGHERS, P. W. SHAW, AND G. R. CARVALHO

Experiential Factors in Postpartum Regulation of Maternal Care ALISON S. FLEMING, HYWEL D. MORGAN, AND CAROLYN WALSH

Sociality, Group Size, and Reproductive Suppression among Carnivores SCOTT CREEL AND DAVID MACDONALD Development and Relationships: A Dynamic Model of Communication ALAN FOGEL Why Do Females Mate with Multiple Males? The Sexually Selected Sperm Hypothesis LAURENT KELLER AND HUDSON K. REEVE

Maternal Behavior in Rabbits: A Historical and Multidisciplinary Perspective ´ LEZ-MARISCAL GABRIELA GONZA AND JAY S. ROSENBLATT Parental Behavior in Voles ZUOXIN WANG AND THOMAS R. INSEL Physiological, Sensory, and Experiential Factors of Parental Care in Sheep ´ VY, K. M. KENDRICK F. LE E. B. KEVERNE, R. H. PORTER, AND A. ROMEYER

Cognition in Cephalopods JENNIFER A. MATHER

Socialization, Hormones, and the Regulation of Maternal Behavior in Nonhuman Simian Primates CHRISTOPHER R. PRYCE

Volume 25

Field Studies of Parental Care in Birds: New Data Focus Questions on Variation among Females PATRICIA ADAIR GOWATY

Parental Care in Invertebrates STEPHEN T. TRUMBO Cause and Effect of Parental Care in Fishes: An Epigenetic Perspective STEPHEN S. CRAWFORD AND EUGENE K. BALON Parental Care among the Amphibia MARTHA L. CRUMP

Parental Investment in Pinnipeds FRITZ TRILLMICHx Individual Differences in Maternal Style: Causes and Consequences of Mothers and Offspring LYNN A. FAIRBANKS

380


Mother–Infant Communication in Primates DARIO MAESTRIPIERI AND JOSEP CALL Infant Care in Cooperatively Breeding Species CHARLES T. SNOWDON Volume 26 Sexual Selection in Seawood Flies THOMAS H. DAY AND ´ S. GILBURN ANDRE Vocal Learning in Mammals VINCENT M. JANIK AND PETER J. B. SLATER

Volume 27 The Concept of Stress and Its Relevance for Animal Behavior DIETRICH VON HOLST Stress and Immune Response VICTOR APANIUS Behavioral Variability and Limits to Evolutionary Adaptation P. A. PARSONS Developmental Instability as a General Measure of Stress ANDERS PAPE MØLLER

Behavioral Ecology and Conservation Biology of Primates and Other Animals KAREN B. STRIER

Stress and Decision-Making under the Risk of Predation: Recent Developments from Behavioral, Reproductive, and Ecological Perspectives STEVEN L. LIMA

How to Avoid Seven Deadly Sins in the Study of Behavior MANFRED MILINSKI

Parasitic Stress and Self-Medication in Wild Animals G. A. LOZANO

Sexually Dimorphic Dispersal in Mammals: Patterns, Causes, and Consequences LAURA SMALE, SCOTT NUNES, AND KAY E. HOLEKAMP

Stress and Human Behavior: Attractiveness, Women’s Sexual Development, Postpartum Depression, and Baby’s Cry RANDY THORNHILL AND F. BRYANT FURLOW

Infantile Amnesia: Using Animal Models to Understand Forgetting MOORE H. ARNOLD AND NORMAN E. SPEAR Regulation of Age Polyethism in Bees and Wasps by Juvenile Hormone SUSAN E. FAHRBACH Acoustic Signals and Speciation: The Roles of Natural and Sexual Selection in the Evolution of Cryptic Species GARETH JONES

Welfare, Stress, and the Evolution of Feelings DONALD M. BROOM Biological Conservation and Stress HERIBERT HOFER AND MARION L. EAST

Volume 28

Understanding the Complex Song of the European Starling: An Integrated Ethiological Approach MARCEL EENS

Sexual Imprinting and Evolutionary Processes in Birds: A Reassessment CAREL TEN CATE AND DAVE R. VOS

Representation of Quantities by Apes SARAH T. BOYSEN

Techniques for Analyzing Vertebrate Social Structure Using Identified


Individuals: Review and Recommendations HAL WHITEHEAD AND SUSAN DUFAULT Socially Induced Infertility, Incest Avoidance, and the Monopoly of Reproduction in Cooperatively Breeding African Mole-Rats, Family Bathyergidae NIGEL C. BENNETT, CHRIS G. FAULKES, AND JENNIFER U. M. JARVIS Memory in Avian Food Caching and Song Learning: A General Mechanism or Different Processes? NICOLA S. CLAYTON AND JILL A. SOHA Long-Term Memory in Human Infants: Lessons in Psychobiology CAROLYN ROVEE-COLLIER AND KRISTIN HARTSHORN Olfaction in Birds TIMOTHY J. ROPER Intraspecific Variation in Ungulate Mating Strategies: The Case of the Flexible Fallow Deer SIMON THIRGOOD, JOCHEN LANGBEIN, AND RORY J. PUTMAN

Volume 29 The Hungry Locust STEPHEN J. SIMPSON AND DAVID RAUBENHEIMER Sexual Selection and the Evolution of Song and Brain Structure in Acrocephalus Warblers CLIVE K. CATCHPOLE Primate Socialization Revisited: Theoretical and Practical Issues in Social Ontogeny BERTRAND L. DEPUTTE

381

Ultraviolet Vision in Birds INNES C. CUTHILL, JULIAN C. PARTRIDGE, ANDREW T. D. BENNETT, STUART C. CHURCH, NATHAN S. HART, AND SARAH HUNT What Is the Significance of Imitation in Animals? CECILIA M. HEYES AND ELIZABETH D. RAY Vocal Interactions in Birds: The Use of Song as a Model in Communication DIETMAR TODT AND MARC NAGUIB

Volume 30 The Evolution of Alternative Strategies and Tactics H. JANE BROCKMANN Information Gathering and Communication during Agonistic Encounters: A Case Study of Hermit Crabs ROBERT W. ELWOOD AND MARK BRIFFA Acoustic Communication in Two Groups of Closely Related Treefrogs H. CARL GERHARDT Scent-Marking by Male Mammals: Cheat-Proof Signals to Competitors and Mates L. M. GOSLING AND S. C. ROBERTS Male Facial Attractiveness: Perceived Personality and Shifting Female Preferences for Male Traits across the Menstrual Cycle IAN S. PENTON-VOAK AND DAVID I. PERRETT The Control and Function of Agonism in Avian Broodmates HUGH DRUMMOND

382


Volume 31 Conflict and Cooperation in a Female-Dominated Society: A Reassessment of the ‘‘Hyperaggressive’’ Image of Spotted Hyenas MARION L. EAST AND HERIBERT HOFER Birdsong and Male–Male Competition: Causes and Consequences of Vocal Variability in the Collared Dove (Streptopelia decaocto) CAREL TEN CATE, HANS SLABBEKOORN, AND MECHTELD R. BALLINTIJN Imitation of Novel Complex Actions: What Does the Evidence from Animals Mean? RICHARD W. BYRNE

Lateralization in Vertebrates: Its Early Evolution, General Pattern, and Development LESLEY J. ROGERS Auditory Scene Analysis in Animal Communication STEWART H. HULSE Electric Signals: Predation, Sex, and Environmental Constraints PHILIP K. STODDARD How to Vocally Identify Kin in a Crowd: The Penguin Model THIERRY AUBIN AND PIERRE JOUVENTIN

Advances in the Study of Behavior, Volume 32 (Advances in the Study of Behavior)

Advances in the Study of Behavior, Volume 34 (Advances in the Study of Behavior)

Advances in the Study of Behavior, Volume 27 (Advances in the Study of Behavior)

Advances in the Study of Behavior, Volume 38 (Advances in the Study of Behavior)

Advances in the Study of Behavior, Volume 30 (Advances in the Study of Behavior)

Advances in the Study of Behavior, Volume 35 (Advances in the Study of Behavior)

Advances in the Study of Behavior, Volume 36 (Advances in the Study of Behavior)

Advances in the Study of Behavior, Volume 33 (Advances in the Study of Behavior)

Advances in the Study of Behavior, Volume 29 (Advances in the Study of Behavior)

Advances in the Study of Behavior, Volume 37 (Advances in the Study of Behavior)

Advances in the Study of Behavior, Volume 23 (Advances in the Study of Behavior)

Advances in the Study of Behavior, Volume 25: Parental Care

Advances in the Study of Behavior, Volume 43

Advances in The Study of Behavior, Volume 18

Advances in the Study of Behavior, Volume 14

Advances in the Study of Behavior, Volume 6

Advances in the Study of Behavior, Volume 17

Advances in the Study of Behavior, Volume 24

Advances in The Study of Behavior, Volume 15

Advances in the Study of Behavior, Volume 21