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

Advances in

THE STUDY OF BEHAVIOR VOLUME 29

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 Environmental and Evolutionary 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 29

San Diego San Francisco New York Boston London Sydney Tokyo

This book is printed on acid-free paper.

Copyright © 2000 by ACADEMIC PRESS 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-2000 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. 0065-3454/00 $35.00 Explicit permission from Academic Press is not required to reproduce a maximum of two figures or tables from an Academic Press chapter in another scientific or research publication provided that the material has not been credited to another source and that full credit to the Academic Press chapter is given.

Academic Press A Harcourt Science and Technology Company 525 B Street, Suite 1900, San Diego, California 92101-4495, USA http://www.apnet.com

Academic Press 24-28 Oval Road, London NW1 7DX, UK http://www.hbuk.co.uk/ap/ International Standard Book Number: 0-12-004529-X PRINTED IN THE UNITED STATES OF AMERICA 99 00 01 02 03 04 QW 9 8 7

6

5

4

3

2

1

Contents

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

ix xi

The Hungry Locust STEPHEN J. SIMPSON AND DAVID RAUBENHEIMER I. II. III. IV. V. VI. VII. VIII. IX.

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . The Pattern of Feeding . . . . . . . . . . . . . . . . . . . . . . . . . . . . Complicating the Story: Nutrients . . . . . . . . . . . . . . . . . . . . More Complicated Still: How to Deal with Nutrient Interactions? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Adding Nonnutrient Dimensions: Plant Secondary Compounds . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Adding Spatial Complexity . . . . . . . . . . . . . . . . . . . . . . . . . Using Real Plants and from the Laboratory into the Real World . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . From Locusts to Vertebrates. . . . . . . . . . . . . . . . . . . . . . . . Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1 2 14 21 35 35 36 37 37 38

Sexual Selection and the Evolution of Song and Brain Structure in Acrocephalus Warblers CLIVE K. CATCHPOLE I. II. III. IV. V. VI. VII. VIII.

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Songs of Acrocephalus Warblers . . . . . . . . . . . . . . . . . . . . . Song, Context, and Mating System . . . . . . . . . . . . . . . . . . . Song and Female Choice . . . . . . . . . . . . . . . . . . . . . . . . . . . Song and Reproductive Success . . . . . . . . . . . . . . . . . . . . . Song and Male Quality . . . . . . . . . . . . . . . . . . . . . . . . . . . . Conclusions and Future Research . . . . . . . . . . . . . . . . . . . . Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . v

45 47 51 60 68 73 83 91 92

vi

CONTENTS

Primate Socialization Revisited: Theoretical and Practical Issues in Social Ontogeny BERTRAND L. DEPUTTE I. II. III. IV.

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Methodological and Empirical Issues . . . . . . . . . . . . . . . . . Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

99 111 145 149 150

Ultraviolet Vision in Birds INNES C. CUTHILL, JULIAN C. PARTRIDGE, ANDREW T. D. BENNETT, STUART C. CHURCH, NATHAN S. HART, AND SARAH HUNT I. II. III. IV. V. VI.

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . The Mechanism of Color Vision in Birds . . . . . . . . . . . . . . Studying Color Nonanthropocentrically . . . . . . . . . . . . . . . The Functions of UV Vision . . . . . . . . . . . . . . . . . . . . . . . . Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

159 162 176 177 196 201 202

What Is the Significance of Imitation in Animals? CECILIA M. HEYES AND ELIZABETH D. RAY I. II. III. IV. V. VI. VII. VIII.

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Perceptual Opacity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Theories of Imitation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Associative Sequence Learning Theory . . . . . . . . . . . . . . . Predictions and Theory Testing. . . . . . . . . . . . . . . . . . . . . . Survey of Two-Action Tests . . . . . . . . . . . . . . . . . . . . . . . . Postscript: Imitation and Culture . . . . . . . . . . . . . . . . . . . . Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

215 216 217 222 228 230 240 241 242

Vocal Interactions in Birds: The Use of Song as a Model in Communication DIETMAR TODT AND MARC NAGUIB I. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . II. Dyads: Interactions between Two Individuals . . . . . . . . . .

247 255

CONTENTS

III. IV. V. VI.

vii

Interactions with More Than Two Individuals . . . . . . . . . . Relevance for Other Listeners . . . . . . . . . . . . . . . . . . . . . . Final Remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

281 282 285 287 287

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

297

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

303

This Page Intentionally Left Blank

Contributors

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

ANDREW T. D. BENNETT (159), School of Biological Sciences, University of Bristol, Bristol BS8 1UG, United Kingdom CLIVE K. CATCHPOLE (45), School of Biological Sciences, Royal Holloway, University of London, Egham, Surrey TW0 0EX, United Kingdom STUART C. CHURCH (159), School of Biological Sciences, University of Bristol, Bristol BS8 1UG, United Kingdom INNES C. CUTHILL (159), School of Biological Sciences, University of Bristol, Bristol BS8 1UG, United Kingdom BERTRAND L. DEPUTTE (99), CNRS/UMR 6552 Station Biologique, 35380 Paimpont, France NATHAN S. HART (159), School of Biological Sciences, University of Bristol, Bristol BS8 1UG, United Kingdom CECILIA M. HEYES (215), Department of Psychology, University College London, London WC1E 6BT, United Kingdom SARAH HUNT (159), School of Biological Sciences, University of Bristol, Bristol BS8 1UG, United Kingdom MARC NAGUIB (247), Institut fu¨r Verhaltensbiologie, Freie Universita¨t Berlin, 12163 Berlin, Germany JULIAN C. PARTRIDGE (159), School of Biological Sciences, University of Bristol, Bristol BS8 1UG, United Kingdom DAVID RAUBENHEIMER (1), Department of Zoology and University Museum of Natural History, University of Oxford, Oxford OX1 3PS, United Kingdom ELIZABETH D. RAY (215), Department of Psychology, University College London, London WC1E 6BT, United Kingdom ix

x

CONTRIBUTORS

STEPHEN J. SIMPSON (1), Department of Zoology and University Museum of Natural History, University of Oxford, Oxford OX1 3PS, United Kingdom DIETMAR TODT (247), Institut fu¨r Verhaltensbiologie, Freie Universita¨t Berlin, 12163 Berlin, Germany

Preface

The aim of Advances in the Study of Behavior 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 subjects and those studying human subjects. Scientists studying behavior today range more widely than ever before: from ecologists and evolutionary biologists, to geneticists, endocrinologists, pharmacologists, neurobiologists, and developmental psychobiologists, not forgetting the ethologists and comparative psychologists whose prime domain is the study of behavior. 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 publisher 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 illustrates these aims well, with a good mixture of psychological and biological approaches, as well as laboratory and field studies, on species ranging from insects to primates. All the chapters tackle important and timely topics and come up with insights of wide significance to those interested in the study of behavior from any perspective.

xi

This Page Intentionally Left Blank

ADVANCES IN THE STUDY OF BEHAVIOR, VOL. 29

The Hungry Locust Stephen J. Simpson and David Raubenheimer department of zoology and university museum of natural history university of oxford oxford ox1 3ps, united kingdom

I. Introduction It is a bit of a cheek titling a review on feeding ‘‘The Hungry Locust.’’ Vince Dethier (1976) got there first for the blowfly, and he in turn was preceded by Eric Carle (1969) with his children’s book on an unfeasibly polyphagous caterpillar. Mimicking their titles is a form of flattery, but it does invite potentially embarrassing comparisons. In respect to Carle we fail miserably: too many words and no color pictures. We argue, however, that the locust at least rivals the blowfly as a model system for the study of feeding behavior. In fact, the locust is more typical of the major mammalian model systems for the study of feeding and nutrition—and not just in that locusts chew rather than drink their food. It is probably not too much of a simplification to say that the nutritional world of an adult blowfly is treated by its regulatory systems as being unidimensional, consisting of near pure sources of sugar that supply energetic needs. At times of egg development a second dimension is added when hormonal events lead to proteinaceous foods becoming attractive. In contrast, a locust’s host plants provide complex and variable mixtures of nutrients and other compounds. A growing locust faces a problem that is common to many animals: that of having to balance multiple and changing nutrient needs in a multidimensional and variable nutritional environment. As we will demonstrate, it is in dealing with the interactive, multidimensional aspects of nutritional regulation that the study of locusts has provided its major contribution. Locusts are a group of phylogenetically disparate species of grasshopper within the family Acrididae. They are defined not by any particular taxonomic affiliation but by their ability to respond to population density by changing between a cryptic, solitary form and the notorious gregarious, swarming phase. This feat is of considerable interest and applied importance, and it is the subject of behavioral analysis in our laboratory (Simpson 1

Copyright 䉷 2000 by Academic Press All rights of reproduction in any form reserved. 0065-3454/00 $35.00

2

STEPHEN J. SIMPSON AND DAVID RAUBENHEIMER

et al., 1999) but not the subject at issue here. The pertinent feature of their biology, for which locusts are unfairly renowned, is their voracious appetite. It is true that they eat their own mass of food daily, but this is unsurprising for an animal of their size and only warrants comment when there are several billion locusts all eating their own mass of food daily in the same place. In fact, with regard to eating, we shall show that locusts demonstrate some of the most impressive regulatory abilities of any animal, simultaneously balancing their intake and use of numerous nutrients and nonnutrient compounds. They are also wonderfully convenient as study animals because they are relatively easy to rear en masse, large (for an insect) and hence convenient for physiological study, and have a relatively short life cycle. Running a fully balanced experimental design involving several hundred individual locusts will occupy a bench top in a constant-temperature room, the time of a single researcher, and a few hundred grams of food per day. Contrast that with a similar experiment on any vertebrate. Two species have dominated the work on locust feeding: the African migratory locust, Locusta migratoria, and the desert locust, Schistocerca gregaria. The former specializes in eating grasses, whereas the latter will also accept a wide range of broad-leaved plants. As is usual in such matters, the use of these two species owed as much to contingency as it did to rational decision, although not quite so much as did the choice of the black blowfly, Phormia regina, by Vince Dethier. It was cultured from an egg batch laid during a picnic on the liverwurst sandwich of a colleague (Hanson, 1987). Throughout, we use the term ‘‘locust’’ rather loosely. In the main we mean L. migratoria. Where S. gregaria has been studied, it will be mentioned by name. Our approach to the study of feeding behavior in locusts began with a description of the pattern of feeding under standardized environmental conditions, in which insects were of known age and developmental stage (fifth-instar nymphs) and had constant access to a single type of highly nutritious food (wheat seedlings) (Simpson, 1981, 1982). That work was undertaken in the laboratory of Reg Chapman and Liz Bernays, who before and since then have made substantial contributions to the story of grasshopper feeding behavior. The aim at the time was to infer from analyses of patterns of feeding the mechanisms underlying the control of the initiation and termination of feeding, and then to test such inferences through experimental manipulation. We will show how such early analyses have led to our current level of understanding. II. The Pattern of Feeding Paraphrasing Brobeck (1955), the amount of food (or of an individual nutrient) eaten is the number of meals eaten multiplied by the average

THE HUNGRY LOCUST

3

amount eaten in each meal. The study of feeding thus distils down to two questions: Why do animals start feeding when they do, and why do they stop feeding once they have started? Analyses of ad libitum feeding patterns showed that locusts, like most animals, feed in bouts (meals) separated by more or less extended periods without feeding (intermeals). Defining a meal in locusts is not as problematic as in many animals (Castonguay et al., 1986; Panksepp, 1978; Tolkamp et al., 1999) since they tend to wander a short distance from the feeding site and settle into a characteristic perching position (Simpson, 1981, 1990). Additionally, log-survivor and log-frequency analyses of the distributions of gaps between periods of feeding (Langton et al., 1995; Sibly et al., 1990; Slater and Lester, 1982) indicate the presence of a clear criterion (typically ca. 4 min) for distinguishing intrameal pauses from intermeal intervals (Simpson, 1982, 1990, 1995). Similar analyses for the distribution of episodes of feeding also show a marked distinction between periods of ‘‘committed’’ feeding (true meals) and shorter periods of ingestion which represent sampling behavior and food rejections (Simpson, 1994, 1995). Having established what comprises a meal, the next step was to attempt to make sense of the patterns of feeding exhibited by locust nymphs. Although every effort had been made to control and simplify the environment, the resulting feeding patterns were complex in structure and varied greatly between individuals (Simpson, 1982, 1990; Fig. 1a). Nevertheless, statistical analysis of these patterns, in combination with experiments perturbing putative control systems, allowed us to address the two fundamental questions of why it is that meals start and end. A. What Determines When Meals Begin? Traditionally, log-survivor analysis of intermeal intervals and correlations between meal durations, meal sizes, and preceding and following intermeal intervals have been used to explore the causes of meal initiation (Simpson, 1992). By the early 1980s medical statisticians had developed a more effective set of techniques, namely, the fitting of proportional hazards models to distributions of intervals using generalized linear interactive modelling (GLIM). The medical objective was to quantify the efficacy of drug treatments, but essentially the same issues are involved in investigating meal initiation except that instead of measuring the probability (‘‘hazard’’) that a patient under a given drug regime will die, the aim is to measure the probability that a meal will begin. These techniques were modified for use on the locust data and provided a structure for investigating and quantifying the relative contributions of various factors in the control of meal initiation (Simpson and Ludlow, 1986).

4

STEPHEN J. SIMPSON AND DAVID RAUBENHEIMER

THE HUNGRY LOCUST

5

1. Time Since the Previous Meal The probability that a locust will leave its perch and start a new meal increases with time since the previous meal ended. This relationship is plotted in Figs. 1c and 1d, in which the starting tendency for feeding (the hazard from the GLIM analysis) is directly proportional to the probability of a meal beginning (Simpson and Ludlow, 1986) and is plotted on a logarithmic scale against time since the last meal ended. There was significant variation between individual insects, but this appeared not in the shape of the curves but in the intercept with the y-axis, indicating that some locusts were consistently more likely than others to begin feeding with time since their last meal. 2. Size and Nutritional Quality of the Previous Meal There is considerable variation in the sizes of meals taken by an individual locust given constant access to seedling wheat (Fig. 1b). Although the probability of feeding increases with time since the insect last fed, the rate of this increase is affected by the amount of food eaten during the last meal (Fig. 1c.) In other words, large meals of a given food inhibit feeding more and for a longer time than do small ones. When the curve for feeding tendency against time becomes parallel to the x-axis, it indicates that the probability of a meal beginning is independent of time since the last meal. This occurs about 20 min after a small meal, 1 h after an average-sized meal, and more than 2 h after a large meal. The nutritional quality of food also has a profound influence on the timing of meal initiation, as will be discussed in detail in Section III.

Fig. 1. Individual locusts were observed feeding on seedling wheat under controlled temperature and light conditions (a) and feeding patterns were recorded for each locust over the first 130 h of the fifth stadium (b). Such patterns were analyzed using proportional hazards models (c, d). Figure 1c shows the change in the tendency of a meal beginning with time since the previous meal, plotted on a log scale, as influenced by the size of the previous meal. Small meals inhibit the onset of subsequent feeding to a lesser degree than do large meals. An average-sized meal for the group of locusts studied was 45.8 mg. Figure 1d presents the effects of several causal factors on the tendency to start feeding, all plotted on a log scale. A, the change in starting tendency with time after an average meal, B, the effect of the presence of food nearby on the tendency to commence locomotion (in 90% of cases leading to feeding when food was available). C–E, the effects of other factors on the tendency to start feeding. Factors B–E are all plotted to the same scale as A and multiply the tendency to feed independently of time since the previous meal. As a result, their interactive effects may be derived by summing the curves (after Simpson, 1981, 1982; Simpson and Ludlow, 1986).

6

STEPHEN J. SIMPSON AND DAVID RAUBENHEIMER

3. Physiological Mechanisms Analyses of feeding patterns therefore indicate clear and quantifiable consequences of feeding on the probability that a new meal will begin. This is consistent with feeding inhibiting itself, with such inhibitory consequences declining with time after a meal. There has been considerable work investigating the nature of such feeding-induced inhibition (Bernays and Simpson, 1982; Simpson, 1995; Simpson and Bernays, 1983; Simpson et al., 1995). Briefly, as food enters and distends the gut it stimulates stretch receptors whose activity inhibits feeding (Bernays and Chapman, 1973, Simpson, 1983) and triggers the release of hormones from neurohemal organs. Such hormones have various effects, including stimulating diuresis (Mordue, 1969) and gut emptying (Cazal, 1969), reducing locomotor activity (Bernays, 1980; Bernays and Chapman, 1974c), and reducing chemosensory input by causing the pores at the tips of taste hairs on the mouthparts to close (Bernays et al., 1972; Bernays and Chapman, 1972a; Bernays and Mordue, 1973). Digestion and absorption of food also changes the composition of the blood, providing osmotic and nutrient-specific feedbacks which influence the timing of feeding by acting both centrally and peripherally (Simpson and Raubenheimer, 1993a; see Section III). All these feeding-induced effects decline with time after a meal at a rate that reflects the rising probability of recommencing feeding. 4. Complicating Causal Influences A feeding control system that consists entirely of meals beginning as a direct consequence of the decline in the inhibitory effects of the previous meal will produce highly regular feeding patterns. As discussed, locusts (and indeed virtually all species investigated) have irregular feeding patterns. It transpires that much of the complexity is introduced by endogenous and exogenous causal influences whose timing is irregular and which may not even be obviously feeding related. a. A Short-Term Endogenous Rhythm. Analyses of feeding patterns identified a short-term endogenous rhythm influencing the initiation of feeding (Simpson, 1981). The rhythm has a period of 12–16 min, depending on the individual locust nymph. Feeding does not begin on every cycle, but meals tend to start at the same phase in the rhythm. Fitting proportional hazards models showed that the probability of feeding beginning in the next minute is four times greater during the peak than in the trough half of the cycle, with the size of this effect being independent of time since the previous meal ended (Simpson and Ludlow, 1986). Several other behaviors follow the same rhythm as the initiation of feeding, including walking without subsequent feeding and a range of small movements exhibited

THE HUNGRY LOCUST

7

during intermeal intervals. The thresholds for these latter behaviors are dissociated from that for feeding, as indicated by the fact that, unlike the initiation of feeding, their probability of occurrence does not change with time since the last meal (Simpson, 1981; Simpson and Ludlow, 1986). A similar short-term cycle has recently been reported to influence meal initiation in a caterpillar (Bernays and Singer, 1998). b. Defecation. A surprisingly powerful influence on, or at least predictor of, feeding is defecation. The production of a fecal pellet, which occurs in wheat-fed locusts on average every 30 min but irregularly, is followed by a sevenfold elevation in the tendency to start feeding (Simpson and Ludlow, 1986). The effect lasts for a short time (3 or 4 min) and is neither affected by time since the last meal nor interacts with the short-term oscillation. The influence of defecation is not specific to feeding; there is also an elevated tendency for walking and several other behaviors shown during resting periods to occur soon after defecation (Simpson, 1981, 1990). The evidence suggests that defecation provides an excitatory sensory input, which would appear to occur from stretch receptors associated with the hindgut. These receptors produce a powerful, phasic burst of activity when stimulated in a manner approximating hindgut movements during defecation (Simpson, 1983). The tonic activity of these receptors (indicating the degree of hindgut fullness) plays another role, namely, inhibition of feeding. c. Environmental Stimuli. Two sources of stimuli that have been shown to influence the tendency to feed in locusts are the presence of food nearby (visual and/or olfactory stimuli) and light. A locust is six times more likely to leave its perch and start walking during the next minute of an intermeal interval if food is present nearby than it is if food is absent (Simpson and Ludlow, 1986). In 90% of cases in which food is available insects will then commence a meal. The increase in the probability of starting to walk is independent of time since the last meal. Food stimuli also cause a rise in the probability of the occurrence of several behaviors occurring during resting, as was seen for the rhythm and defecation. Analyses of feeding patterns indicate that average intermeal intervals are longer in the dark than in the light phase of a 12:12 h light:dark photoregime and less is eaten during each meal (Simpson, 1982). The tendency to begin a meal during the next minute is twice as high during light than during dark phases and the effect of light phase is not related to time since the last meal (Simpson and Ludlow, 1986). There are various mechanisms which could contribute to these influences of light, including enabling insects to see the food, acting as an arousing stimulus per se, and serving as a stimulus for entraining circadian rhythms of activity (Simpson and Ludlow, 1986).

8

STEPHEN J. SIMPSON AND DAVID RAUBENHEIMER

5. The Interrelationships among Causal Factors An important advantage of the proportional-hazards approach to analyzing feeding patterns is that, because the contribution of each causal factor is measured in terms of the extent to which it multiplies the probability of a meal beginning, factors can readily be compared (Fig. 1d). Providing that they act independently of time since the previous meal—which seems on statistical grounds to be the case for the factors discussed previously—their interactive effects can also be explored by simply summing their influences on a log scale (Simpson and Ludlow, 1986).

B. What Determines When Meals End? Our initial studies of feeding patterns indicated that the sizes of meals were highly variable, whereas meal durations were much more consistent, leading to the erroneous conclusion that meals were terminated by some time-dependent factor and not by the amount eaten (Simpson, 1982). This conclusion ran counter to other evidence demonstrating that volumetric feedback from the gut controlled meal size, and that severing the relevant nerves caused hyperphagia (Bernays and Chapman, 1973; Simpson, 1983; Roessingh and Simpson, 1984). There is a simple solution to the apparent paradox, however (Simpson and Bernays, 1983). Because meals are of a relatively standard duration but highly variable in size, it must mean that smaller meals were ingested more slowly than larger ones. 1. The Relationship between Meal Size, Meal Duration, and Ingestion Rate An experiment was performed in which several unrelated factors known to influence meal size were combined in a factorial design, and the size, duration, and ingestion rate of a test meal were recorded (Simpson et al., 1988b). The treatment factors included the palatability of the food (wheat seedlings with or without a sugar coating), time of prior food deprivation (2, 5, or 8 h), insect age (1, 4, or 6 days old as fifth-instar nymphs), and the presence of other locusts nearby (alone or in a crowd). The conclusion from the experiment was that various exogenous and endogenous factors are integrated within the central nervous system (CNS) and together provide excitation for feeding. The net level of excitation in CNS feeding circuitry (henceforth termed feeding excitation) determines both the rate of ingestion and the amount of feeding-induced inhibition that must accrue to end the meal. Ingestion rate and meal size covaried positively up to a point, and over this range meal duration was constant. Thereafter, ingestion rate saturated. As a result, meal duration was constant over a wide range

THE HUNGRY LOCUST

9

of meal sizes, which spanned that shown under ad libitum conditions, and only increased for very large meals (Figs. 2a and 2b). Such a model would appear to apply to many animals, with variations in the detailed relationship between meal size and ingestion rate varying according to inter alia species and the physical nature of the food (Simpson, 1995). Two issues remain, however. First, why should ad libitum-fed locusts show such a high level of variation between meals in feeding excitation? Second, what are the sources of feeding-induced negative feedbacks? 2. Variation in Levels of Feeding Excitation Stated simply, a meal is initiated once feeding excitation exceeds a threshold level. The extent by which the threshold is exceeded due to the summed influences of causal factors (both excitatory and inhibitory) then translates into meal size and rate of ingestion. The net level of feeding excitation present as a meal begins is influenced by several factors, including the amount of feeding inhibition remaining from previous meals and excitation generated from food-related stimuli and resulting from approaching, contacting, and actually eating the food. Additionally, there are other excitatory influences discussed previously, such as defecation, the internal rhythm, light intensity, and whether other insects are nearby. As discussed previously, many of the same factors that affect when meals begin also influence meal size by determining the amount of inhibition that is required to end the meal (Simpson, 1990). Because these causal factors interact in their effect on feeding excitation and in some cases are irregular in their timing, meal sizes assume a distribution which in many individuals is indistinguishable from a negative exponential random model (Simpson, 1982). Excitation generated by contacting and ingesting the food is derived mainly from chemoreceptors on the feet and mouthparts responding to nutrients, such as sugars, amino acids, and salts, and from antennal and mouthpart receptors stimulated by olfactory stimuli (Bernays and Simpson, 1982; Chapman, 1995). Such chemosensory inputs act as powerful arousing stimuli, rapidly increasing the level of feeding excitation (Barton Browne et al., 1975a; Bernays and Chapman, 1974c; Dethier et al., 1965). This provides a positive feedback during the early stages of feeding that presumably prevents ‘‘dithering’’ about the feeding threshold (Houston and Sumida, 1985) and ensures that area-concentrated searching occurs if contact with the food is lost during the meal (Bernays and Chapman, 1974c; Dethier, 1957). The size and duration of the excitatory effect of chemostimulation vary with the intensity of chemoreceptor inputs and the time since and nutritional quality of previous meals. The responsiveness of chemoreceptors depends on the chemical composition of the food, but it is also altered by feedbacks

10

STEPHEN J. SIMPSON AND DAVID RAUBENHEIMER

Fig. 2. The relationship between meal size, overall ingestion rate, and meal duration. The overall ingestion rate across a test meal increased with meal size, up to meals of ca. 150 mg in size, after which ingestion rate reached an asymptote (a). Accordingly, meal duration was largely unchanged until meals exceeded 150 mg but increased thereafter (b). Instantaneous ingestion rate declined exponentially throughout a meal. Meals of all sizes began, on average, at closely similar starting rates (c), but ingestion rate then declined more rapidly throughout small meals than in large meals (d) (after Simpson et al., 1988b).

THE HUNGRY LOCUST

11

related to the last meal and the locust’s nutritional state (Simpson and Raubenheimer, 1993a; see Section III). Although certain nutrient chemicals act as phagostimulants, other compounds inhibit feeding by stimulating specialist deterrent receptors and/or blocking responses to phagostimulants and will therefore negate phagostimulatory excitation (Bernays and Chapman, 1977; Chapman, 1995; Chapman et al., 1991). 3. Sources of Negative Feedback during a Meal In addition to longer term inhibitory influences on feeding (see Section II,A,3), there are short-term inhibitory feedbacks acting during a meal which at some point negate the levels of excitation present and thus terminate the meal. These include volumetric feedback from the gut (Bernays and Chapman, 1973; Simpson, 1983; Roessingh and Simpson, 1984) and rapid changes in blood osmolality and nutrient composition (Abisgold and Simpson, 1987; Bernays and Chapman, 1974a,b; Simpson and Raubenheimer, 1993a; see Section III). Adaptation of chemoreceptors may also play a role (Barton Browne et al., 1975b; Blaney and Duckett, 1975), as will input from receptors responding to any chemical deterrents present in the food (Bernays and Chapman, 1972b). The model of locust feeding that we developed above is that a range of exogenous and endogenous factors influence feeding excitation, suprathreshold levels of which cause a meal to begin and then translate into the rate of ingestion and the amount of negative feedback needed to end the meal, with inhibition accumulating as a meal progresses. A logical extension is that the instantaneous rate of ingestion should decline throughout a meal as inhibition accrues. Simpson et al. (1988b) found that this did indeed occur. Ingestion rate in wheat-fed insects was found to be highest at the start of a meal and then declined as a result of both chewing more slowly and taking more intrameal pauses. There was an interesting influence of feeding excitation on the relationship between rate of ingestion and time during a meal. Rather than larger meals (generated by higher levels of excitatory input) starting at a higher initial ingestion rate than that of smaller ones and then following the same rate of decay (as occurs in rats; Davis et al., 1978; McCleery, 1977), locusts fed seeding wheat began meals at the same ingestion rate but then showed different rates of decay, with ingestion rate declining more rapidly during small than large meals (Figs. 2c and 2d). C. An Integrative Simulation Behavioral analyses have provided both an insight into the mechanisms involved in the control of feeding by locusts and, perhaps more important,

12

STEPHEN J. SIMPSON AND DAVID RAUBENHEIMER

parameter estimates for the relative strength and interactive effects of the causal factors involved. We have considered the two separate questions of why meals start and what causes them to stop. However, it is clear that these are not independent issues. The next step must therefore be to combine them into a single model. Just such a simulation is shown in Fig. 3 (Simpson, 1995). Based on the results of Simpson and Ludlow (1986), Simpson et al. (1988b) and other work, the simulation plots the level of feeding excitation in CNS circuits (as inferred from behavioral analyses) on a log scale against time. Thresholds for feeding and locomotion are represented. As discussed earlier, these behaviors share causal factors (although not all causal factors), and a locust typically must leave the perch and walk before it feeds. The level of feeding excitation in the simulation is influenced by the 15-min short-term rhythm (which runs continuously), defecation (which excites for the next 3 or 4 min and is timed to occur in a normal distribution with a mean interval of 30 min), excitation generated during feeding in proportion to the level of inhibition remaining from the previous meal, negative feedbacks accumulating exponentially throughout a meal, and inhibition which persists or begins after the end of the meal and whose size and time course

Fig. 3. Output from a simulation of locust feeding. See text for a detailed description (from Simpson, 1995).

THE HUNGRY LOCUST

13

of decline is a function of meal size and quality. The use of a log scale for feeding excitation enables multiplicative causal effects to be added and straightens exponential relationships. The virtual locust in Fig. 3 begins by sitting on its perch (time ⫽ 0), having fed some time earlier. After 20 min it begins walking as a result of declining inhibition from the earlier meal and a peak in its internal rhythm. The locust walks randomly for 5 min but does not feed, then it settles as its rhythm once again enters a trough. A few minutes later the locomotion threshold is crossed again, followed soon afterwards by crossing of the feeding threshold. Feeding commences and induces central excitation. A meal of virtual wheat (90 mg) is ingested before inhibitory inputs bring excitation below the feeding threshold 6 min later. The locust then wanders around for 2 min, perches, and becomes quiescent. Inhibitory inputs stimulated by the meal then continue to cause feeding excitation to decrease for an additional 17 min, after which the balance between the arrival and removal of feeding inhibition changes such that feeding excitation once again begins to increase. A short time later the insect defecates, but the resulting transient excitation only triggers a brief bout of nondirected walking. After an additional 40 min it again defecates, and now both locomotion and feeding thresholds are exceeded simultaneously. The locust walks straight to the food and eats for 4 min, ingesting 89 mg. During the next 160 min the locust takes two more 6-min meals (79 and 150 mg in size, respectively) and a brief nibble. The simulation reproduces the complex and variable feeding patterns which are a feature of real locusts and generates the same basic relationships between meal sizes, durations, and intermeal intervals, indicating that it captures the fundamental essence of locust feeding. D. The Relationship between Short- and Long-Term Regulation of Intake It is worth briefly discussing an issue that has caused confusion among workers in the field of feeding control. In all animals investigated to date, there is evidence of homeostatic control of intake and relatively little variation between individuals over periods of days or weeks, but at the temporal scale of feeding patterns there is considerable variation in the timing and sizes of meals. What can be seen from the simulation in Fig. 3 is that control of consumption over a period is set by the level of the feeding threshold: The higher the threshold, the lower the amount eaten over a given period. However, whereas a change in the level of the threshold will lead to an overall change in the amount eaten, the actual pattern of meals that will achieve this intake is subject to a high degree of variation because of the

14

STEPHEN J. SIMPSON AND DAVID RAUBENHEIMER

number of interacting causal factors involved and the fact that some of these are temporally irregular and not even specific to feeding behavior.

III. Complicating the Story: Nutrients To this point we have discussed the simple case in which a locust has an ample supply of a single food (seedling wheat). The functional aim of eating is to gain nutrients for structural and energetic purposes, and using a single food source does not allow the effects on feeding control of particular nutrients to be established. To this end, we performed an experimental programme using chemically defined synthetic foods. A. Starting Simple: No-Choice Assays The first experiment involved feeding locust nymphs over a 12-h period one of four foods varying in digestible carbohydrate (14 or 28% by dry weight) and protein (14 or 28%) content and then recording their feeding patterns (Simpson and Abisgold, 1985). At these nutrient levels and under the experimental conditions used, locusts ate more of the low-protein foods, irrespective of digestible carbohydrate content. They did so by eating the same-sized meals more frequently (Figs. 4a–4c). What, then, caused insects on low-protein foods to begin feeding sooner after a meal? Physiological investigations showed that there was no difference in the gut emptying time of locusts fed high- or low-protein foods, thus eliminating volumetric factors as candidates for producing the difference in intermeal interval. There were, however, marked differences in the composition of the blood, with high-protein-fed insects having elevated blood osmolality and levels of various free amino acids (Abisgold and Simpson, 1987; Zanotto et al., 1996). The effects of these were investigated by injecting low-proteinfed locusts with solutions which independently increased the amino acid concentration and osmolality of the blood to that of high-protein-fed insects and measuring the time until the next meal began (Abisgold and Simpson, 1987; C. L. Simpson, 1990). The result was that both osmolality and amino acids per se delayed feeding (Fig. 4d), the former mainly by prolonging the period of quiescence after feeding and the latter mainly by increasing the probability that food would be rejected once contacted. The fact that blood amino acid levels altered the probability of acceptance of food implied some effect on the way in which the insect tasted its food. This could occur either at the level of the taste receptors or within the CNS. A series of experiments in which blood amino acid titers and osmolality were manipulated through injections, in the same way as in the behavioral

THE HUNGRY LOCUST

15

Fig. 4. A summary of the results of experiments on the regulation of protein intake in fifth-instar locusts. Over a 12-h period, nymphs ate more foods containing 14% than 28% protein, irrespective of whether there was 14 or 28% digestible carbohydrate in the food (a). They did so by eating similarly sized meals more frequently (b, c). The high-protein-fed locusts were found to have elevated levels of free amino acids in the blood and to have higher blood osmolality. Figure 4d shows the effect on the time to the next meal of raising the blood levels of amino acids and/or blood osmolality in recently low-protein-fed locusts up to those expected in high-protein-fed insects. Controls received an isotonic saline injection. Increasing either amino acid concentration or osmolality delayed the next meal. The effect of osmolality was shown to be principally on the time to next commence locomotion, whereas the major effect of amino acids per se was to influence the probability of accepting food once it was contacted. Electrophysiological recording from taste receptors on the maxillary palps indicated that amino acids, but not osmolality, specifically reduced gustatory responsiveness to stimulation by amino acids (e). The nature of this blood-borne amino acid feedback was found to be due to local interactions with taste receptors rather than requiring centrifugal feedback from the CNS. This is demonstrated in f. Cutting the sensory nerve from the mouthpart palps is shown not to affect the modulation of gustatory responsiveness to amino acid stimulation following an abdominal injection of amino acids into the blood system. Also, microinjecting amino acids directly into the ligatured tip of the palp resulted in reduced electrophysiological responsiveness to amino acid stimulation (after Abisgold and Simpson, 1987, 1988; Simpson and Abisgold, 1985; Simpson and Simpson, 1992).

16

STEPHEN J. SIMPSON AND DAVID RAUBENHEIMER

experiments, indicated that there was indeed a nutrient-specific influence of blood amino acids on gustatory responsiveness. Elevating titers of blood amino acids but not osmolality caused reduced taste responses to stimulation by amino acids (Abisgold and Simpson, 1988; Fig. 4e). It was also found that elevated levels of all of a suite of eight amino acids were necessary both to provoke the sensory change and to cause the associated compensatory reduction in food consumption (C. L. Simpson et al., 1990b). Therefore, having high levels of amino acids in the blood leads to reduced sensory responsiveness to amino acids in food and hence a lowered probability of commencing (or sustaining) ingestion. Given that the blood provides a repository for those nutrients in transit from the gut to the tissues, and those surplus to current requirements, nutrient titers provide an instantaneous and constantly updated indication of the insect’s nutritional state. The fact that blood nutrients can directly affect the way in which taste receptors respond to the food, which in turn influences feeding behavior, provides a remarkably direct link between the composition of the food, the insect’s nutritional state, and its behavior. In mammals, nutritional feedbacks onto taste-evoked responses appear to occur more centrally in the integrative pathways [in the nucleus of the solitary tract in rats (Giza and Scott, 1983, 1987) and more centrally still in the hypothalamus of monkeys (Rolls, 1989)]. What is the route by which blood amino acids modulate taste responses to amino acids? There are two basic possibilities. First, there is central detection of blood amino acids and then feedback to the periphery via either neural or hormonal pathways. Second, blood nutrients act directly on the sense organs at the periphery, without recourse to centrifugal feedbacks. An experiment in which the nerve was cut to one but not the other maxillary palp on the mouthparts of locusts indicated that neural feedbacks from the CNS were not required to produce the reduction in responsiveness to amino acids following amino acid injection into the blood (Simpson and Simpson, 1992; Fig. 4f ). Furthermore, when a single palp was isolated from the rest of the animal by a ligature, and a tiny quantity of amino acids injected into its tip, there was still a local, amino acid-specific reduction in gustatory responsiveness (Simpson and Simpson, 1992; Fig. 4f ). Hence, the influence of blood amino acids on taste responses occurs directly at the periphery. Perhaps the same molecular receptors that respond to amino acids in the food are also bound by amino acids in the blood, effectively adapting the taste neuron to the background level of the blood. This means that the taste neurons serve as ‘‘difference detectors,’’ measuring the discrepancy between what is in the blood and what is in the food. Since it is necessarily the concentration of nutrients, and not the amounts per se, that provides the feedback in this mechanism, it might be anticipated

THE HUNGRY LOCUST

17

that the levels of water in the blood would play a critical role. It has indeed been demonstrated that manipulating the levels of water in the blood of L. migratoria without altering nutrient levels has a marked effect on consumption. Locusts deprived of food and water for 24 h and then allowed a single meal on dry food took smaller meals than did those allowed to drink distilled water immediately before the meal (Raubenheimer and Ga¨de, 1994). The reduction in food intake was due to a reduction in meal duration, with no detectable effect on ingestion rate. Similar results were obtained for the locust S. gregaria using injections of distilled water directly into the blood (Roessingh et al., 1985). Although the mechanism of this effect has not been determined with certainty, it is likely that reduced nutrient concentrations resulting from the addition of water to the blood, and the associated increased gustatory sensitivity, play a role in enhancing food intake in quenched compared to deprived locusts. There is, also abundant evidence that locusts regulate the intake of water. Progress in this field has been reviewed by Bernays (1990). The major conclusion for L. migratoria is that the initiation of drinking is cued by a reduction in blood volume, whereas the duration of a drink is positively associated with the osmolality of the blood (Bernays, 1977). Subsequent work has demonstrated that water-deprived L. migratoria show a compensatory drinking response for osmotic pressure of saline solution through increasing the volume ingested with NaCl concentration (Raubenheimer and Ga¨de, 1994). This was not a result of phagostimulation due to NaCl since the salt solutions were a significant deterrent, as evidenced by a larger proportion of insects rejecting them than the pure water.

B. Providing a Choice Having established the presence of regulatory systems for protein and water intake, two questions arose. First, are there nutrient-specific mechanisms for other nutrient groups? Second, can the feedback mechanisms found for protein lead to a locust selecting appropriately when provided with a choice of foods? An obvious candidate for exploring whether other nutrients are regulated is carbohydrate since, along with protein, it is the major macronutrient group for a leaf-feeding insect (lipid being relatively scarce). The no-choice experiment of Simpson and Abisgold (1985) found no evidence of a response to dietary carbohydrate content, but only a small number of foods were tested over a short period in a no-choice assay. Other experiments demonstrated that in no-choice assays conducted over several days locusts do show ingestive compensation for both protein and digestible carbohydrate (Raubenheimer and Simpson, 1990; Raubenheimer, 1992).

18

STEPHEN J. SIMPSON AND DAVID RAUBENHEIMER

An experiment that addressed both of the previous questions was one in which locusts were pretreated for a period of 4, 8, or 12 h on one of four synthetic foods (P, which lacked only digestible carbohydrate; C, which lacked only protein; O, which was deficient in both protein and digestible carbohydrate; and PC, which contained a full complement of nutrients) and then given a choice between P and C foods (Simpson et al., 1988a). The results were striking: Locusts showed marked preferences for the food (or foods) containing the nutrient group(s) for which they were deficient, even after only 4 h. An additional study found that the compensatory response to protein deprivation was apparent after a single C meal during ad libitum feeding, whereas it took a couple of hours longer to evoke carbohydrate selection (Simpson et al., 1990). Detailed behavioral observations showed that locusts in the choice assay of Simpson et al. (1991) often rejected foods lacking the required nutrients before ingesting them (Fig. 5a). This would be expected if taste responses were modulated by nutritional experience, as described previously for protein. Accordingly, locusts were pretreated on P, C, or PC foods for 4 h and then their gustatory responsiveness was recorded to stimulation with amino acids and sugars. The taste receptors of insects that were deficient in protein had elevated responsiveness to amino acids but not sugars, whereas the receptors of locusts rendered deficient in carbohydrate exhibited enhanced firing rates to stimulation with sugar but not amino acids (Fig. 5b). A subsequent study (Zanotto et al., 1996) showed that blood sugar levels were affected by diet, in an analogous manner to that for amino acids, leading to the conclusion that both protein and carbohydrate intake are regulated at least in part by direct, blood-borne, nutrient-specific modulation of the taste system. Additional work indicated that similar mechanisms exist in caterpillars (Simpson et al., 1988a; Simmonds et al., 1992). C. Learning to Choose Intriguingly, not only did locusts in a choice test eat or reject foods on the basis of taste inputs but also there was an indication that they actually arrived more frequently at the food containing the nutrient for which they were deficient (Simpson et al., 1988a, 1990). Unless there were also directly modulated responses to other food cues (odor or visual), this suggests that locusts might have learned to associate food-related or positional cues with the nutritional content of food and to respond to such cues according to their nutritional state. Testing such an hypothesis involved training locusts to select from two food dishes, one containing P food and the other C food, each of which was paired with a distinctive and initially repellent odor (spearmint or lemon). Next, the locusts were rendered either protein or

THE HUNGRY LOCUST

19

Fig. 5. A summary of the results of experiments on compensatory selection behavior for protein and carbohydrate. Locusts were pretreated for 4 h on one of four synthetic foods (C, lacking protein; P, lacking digestible carbohydrate; O, lacking both protein and digestible carbohydrate; and PC, a nutritionally complete control food) and then provided with P and C in a choice test. Figure 5a plots the responses of locusts on first contacting the test foods following the pretreatment period and demonstrates compensatory selection for foods containing the nutrient(s) for which a locust was rendered deficient. Figure 1b indicates that these behavioral effects were mirrored by changes in gustatory responsiveness. The responses of maxillary palp receptors of locusts pretreated for 4 h on C food were elevated to stimulation by amino acids, whereas those for P-fed insects were elevated to stimulation by sugar (after Simpson et al., 1991).

carbohydrate deficient for 4 h, and then their subsequent responses to the two odors were tested (Simpson and White, 1990). When this was done, insects that had been deprived of protein moved toward a source of the odor that had previously been paired with the P food but not to the odor previously paired with the C food (Fig. 6). Carbohydrate-deprived locusts, on the other hand, did not demonstrate increased attraction to the C odor. Hence, there was evidence for what vertebrate workers term a ‘‘learned specific appetite’’ for protein but not for carbohydrate. However, a later

20

STEPHEN J. SIMPSON AND DAVID RAUBENHEIMER

Fig. 6. An experiment demonstrating a learned specific appetite for protein in locust nymphs. Locusts were first conditioned for 2 days in an arena with two suboptimal but complementary foods (P, lacking digestible carbohydrate; C, lacking protein). The foods were positioned at the ends of two side arms next to porous wicks, one of which was charged with citral (lemon) and the other carvone (spearmint) (a). After the conditioning period, locusts were rendered protein or carbohydrate deficient by allowing them access to only P or C food for 4 h in an odor-free container. They were then placed back into a test arena of the same form as the conditioning arena (with the odors switched in half of the cases to control and test for positional effects), and their behavioral responses to the odors were recorded. Proteindeficient locusts entered the side arm containing the odor previously paired with the P food significantly more often, and they traveled further down the side arm (scored as 1–4 for protein odor and ⫺1 to ⫺4 for carbohydrate odor) than they did the side arm containing the odor previously paired with the C food (b). Locusts that were deficient in carbohydrate did not show the reciprocal preference for the odor previously paired with the C food (c), although they did so in a later experiment in which visual stimuli rather than odors were used (from Simpson and White, 1990).

experiment designed in the same manner showed that locusts demonstrated both protein- and carbohydrate-specific responses when colors rather than odors were used as the conditioned stimuli (Raubenheimer and Tucker, 1997). Positive learned associations with chemical and visual cues have also been found in locusts for salt (Trumper and Simpson, 1994) and water intake (Raubenheimer and Blackshaw, 1994). In addition to learning positive, nutrient-specific associations, there are examples in grasshoppers of aversion learning to cues associated with the quality of dietary sterols and also

THE HUNGRY LOCUST

21

to nonnutritional toxic consequences of eating a given food (Champagne and Bernays, 1991; Bernays, 1993, 1995). Finally, there are data which suggest the presence of nonassociatively learned responses to dietary composition, such as habituation to toxins (Szentesi and Bernays, 1984) and enhanced neophilia with protein (Bernays and Raubenheimer, 1991) or salt (Trumper and Simpson, 1994) deprivation.

IV. More Complicated Still: How to Deal with Nutrient Interactions? The work described to this point has omitted one critical aspect of the control of feeding behavior, namely, the interactions occurring between the systems that control the intake of different nutrients. It is here that the locust has perhaps made its most important contribution to the understanding of feeding behavior and nutrition. The problem is as follows. Animals, including locusts, must ingest adequate amounts of dozens of different nutrient molecules if they are to survive and prosper. Many of these nutrient molecules are packaged in various ratios as macromolecules (proteins, starches, etc.) and are, in turn, combined in various ratios along with nonnutritive molecules, some of which are dangerous, into foods. These foods are then distributed in various amounts and combinations across space and time, accompanied by biotic and abiotic risks in their location, ingestion, and processing. How do animals deal with the various levels of interaction involved: those occurring between nutrients and nonnutrients within foods, between the consumption and postingestive processing of each of these nutrients, between foods within the environment, and so on? Perhaps more pertinent, how do we deal with studying and interpreting these interactions? Until recently, the most common approach to the issue of nutrient interactions was to ignore them by considering the control of feeding as a univariate phenomenon. For example, in behavioral ecology, it is frequently assumed that a single food property, usually its energy content, is biologically preeminent (Hughes, 1993; Stephens and Krebs, 1986), whereas nutritional ecologists have tended to fix on nitrogen (McNeill and Southwood, 1978; Mattson, 1980; White, 1993). Problems occur to the extent that nutritional regulation involves several nutrient groups; it is becoming increasingly apparent that this is the case in most instances. Such difficulties have required the inclusion of additional nutrients as constraints in univariate models (Belovsky, 1990; Pulliam, 1975). An alternative approach is to allow the animal to indicate to the experimenter the manner in which it prioritizes the ingestion of the various

22

STEPHEN J. SIMPSON AND DAVID RAUBENHEIMER

food components: in other words, to treat the interactions among food components as a primary datum in its own right. This enables the construction of models of nutrition that are firmly based on the biological characteristics of the organism. It also provides an opportunity to detect contextdependent responses of animals to food components, i.e., responses whose expression is contingent on other components of the nutritional milieu. For example, Raubenheimer (1992) performed an experiment on locusts in which protein, digestible carbohydrate (14 vs 28% for both nutrients), and the plant-produced allelochemical tannic acid (0 vs 10%) were varied factorially in synthetic foods. This demonstrated that the regulation of feeding is strongly contingent on the levels of the two nutrients in the food rather than on energy, nitrogen, or any other univariate currency alone. Furthermore, the influence of tannic acid was strongly context dependent, reducing consumption only on the foods containing an excess of carbohydrate relative to protein. An explanatory model was constructed in which nutrients present in surplus (relative to other nutrients) were ingested in excess in an attempt to ingest a sufficient amount of the limiting ones (incidental augmentation of intake). However, the ability of locusts to tolerate excesses of either protein or carbohydrate is limited, resulting in a deficit being ingested of the deficient nutrient (incidental restriction of intake). The need thus arose for a systematic means of implementing this model. Specifically, an approach was required that (i) measures the animal’s simultaneous requirements for two or more nutrients and thus identifies the composition of a nutritionally balanced food and how much of this should optimally be eaten over a stipulated time period; (ii) enables measurements to be made of the trade-offs between incidental augmentation and restriction of nutrients when foods that are nutritionally imbalanced prevent animals from simultaneously satisfying their requirements; and (iii) allows the measurements of intake to be readily integrated across levels of biological analysis, including development, regulatory, and metabolic physiology, ecology, and evolution. Thus was born a class of geometric models of nutrition, an approach that has its closest relative in David McFarland and Richard Sibly’s (1972) state–space models of motivation. The models were first applied to our locust system by Raubenheimer and Simpson (1993) and simultaneously elaborated and extended to comparative analyses of various insect groups by Simpson and Raubenheimer (1993b). Recently, the approach was extended to vertebrate systems (Raubenheimer and Simpson, 1997; Simpson and Raubenheimer, 1997; see Section VIII). In the following sections we provide an introductory outline of our geometrical models; detailed accounts of the structure of the ideas and associated issues can be found in

THE HUNGRY LOCUST

23

Raubenheimer and Simpson (1994, 1997, 1998, 1999) and Simpson and Raubenheimer (1993b, 1995, 1996, 1999b).

A. Targets and Rails in Nutrient Space In our scheme, the animal is depicted as existing within a multidimensional space in which each axis represents a nutrient or, as we discuss in Section V, a nonnutrient compound found in foods. Within this nutrient space lie points which represent the animal’s requirements. Such points represent functional targets, and the expectation is that animals will have evolved mechanisms enabling them to approach these points. Targets may be considered as static points, integrated over a given period in the animal’s life, or as dynamic trajectories. There are several target points within the hierarchy of processes that comprise nutritional regulation. The most fundamental of these is the total nutritional requirement, termed the nutrient target. Part of the nutrient target represents the optimal nutrient allocation to growth for somatic, reproductive, or storage purposes: the growth target. The remaining nonstructural component of the nutrient target comprises the optimal nutrient requirements for supplying energetic needs to support basal metabolism, behavioral activities, etc. Reaching the nutrient target involves finding and eating food. Because it is inevitable that, to some degree, nutrients will be wasted in translating food into tissue needs, intake requirements will exceed the nutrient target in its various nutrient dimensions. We have defined the intake target as the point in nutrient space which, when ingested, will provide an animal with nutrients to its tissues at the optimal rate and balance. Foods are mixtures of various nutrient and nonnutrient compounds and as such are represented as lines radiating out from the origin into nutient space. We have used the term rails for these food lines because if an animal is confined to eating only one food it is forced to ingest the ratio of nutrients that that food contains. Although it can move along the rail into nutrient space by eating more of the food, it cannot get off without switching to another food or differentially utilizing nutrients postingestively. In addition to the requirements of the animal being indicated by points in nutrient space, so too is the animal’s current nutritional state. The vector from the point representing its current state to the intake target determines which food or foods need to be ingested if the animal is to reach its intake target. An optimal food is the one that enables the animal to move directly to the intake target, thus simultaneously achieving its multiple nutrient requirements.

24

STEPHEN J. SIMPSON AND DAVID RAUBENHEIMER

If the animal has only a suboptimal food available, it will (by definition) not be able to reach its intake target. As a result, the animal will have to compromise between eating some nutrients in excess and undereating others, relative to the intake target. However, if two or more suboptimal foods are available whose rails together encompass the intake target in nutrient space, then the intake target can be reached by eating a specific combination of these foods. In such a case the foods would be termed nutritionally complementary. Having briefly introduced the basic structure of the geometric framework, we return to experiments on locust feeding. B. Where Is the Intake Target? The intake target provides an essential reference point for understanding feeding behavior. There are three ways of estimating where it lies: (1) functionally, by adding a fitness axis to nutrient space and deriving a fitness landscape by measuring performance and reproductive variables; (ii) by challenging the animal to defend a point of intake in nutrient space and assuming that such points of homeostasis are aligned with fitness maxima; and (iii) by reconstructing the intake target from measurements of growth, repiration, and wastage. We have used all three for locusts in the context of protein and carbohydrate regulation and found closely similar outcomes (Fig. 7; Chambers et al., 1995; Raubenheimer and Simpson, 1993, 1997; Simpson and Raubenheimer, 1993b). We have found that locusts will adjust the amounts and ratios of foods they eat such that they strongly defend points of protein vs carbohydrate intake. For instance, when nymphs were provided with one of four different pairings of complementary foods, in each case they ate the unique ratio and amounts of the two foods that resulted in their arriving at the same point of intake on a protein–carbohydrate nutrient plane (Chambers et al., 1995). The selected ratio of protein to carbohydrate was 1:1.2. When a near-optimal ratio (1:1) of protein to carbohydrate was provided in the food, but the concentration of these nutrients was diluted across a fivefold range by addition of indigestible cellulose, the locusts adjusted their intake accordingly, eating five times as much of the most dilute food as of the most concentrated (Raubenheimer and Simpson, 1993). Locusts also defended a point of salt vs protein plus carbohydrate intake when given complementary pairs of foods varying in their ratio of salt to protein and carbohydrate (Trumper and Simpson, 1993). C. The Form of Nutritional Compromises As previously discussed, if an animal has only a single suboptimal food or two or more noncomplementary foods available, it cannot reach its

THE HUNGRY LOCUST

25

Fig. 7. Data estimating the position of the intake and growth targets in protein– carbohydrate space for locusts over the fifth stadium. The position of the optimal ratio of protein and carbohydrate in the food (optimal food rail) was derived using performance measures recorded in a study in which locusts were provided with 1 of 25 synthetic foods varying in protein and carbohydrate content. The optimal food rail intersects the estimated position of the intake target, showing congruence between performance measures and points of homeostasis. The position of the intake target was estimated independently in two ways. First, locusts were provided with one of four pairings of foods varying widely in protein and carbohydrate content and allowed to select their diet. In doing so, they exhibited defense of intake in both nutrient dimensions. Second, a physiological estimate of the intake target was derived from measurements of growth and respiration. Growth was tightly defended across a wide range of protein–carbohydrate intakes as a result of differential utilization of ingested protein and carbohydrate. The cluster of points on the graph represents the outcome of 19 different food treatments (after Chambers et al., 1995; Raubenheimer and Simpson, 1993, 1997; Simpson and Raubenheimer, 1993b; Zanotto et al., 1997).

intake target and must therefore balance undereating some nutrients and overeating others. The point of balance (point of nutritional compromise) achieved by the animal in such circumstances is determined by the relative weighting given by its regulatory systems to the nutrients involved. When measured across a range of food rails, these points of compromise form intake arrays that define a more general rule of compromise. As is the case for the estimation of the location of the intake and other targets, the nature of such rules is a matter for experimental measurement and not a priori assumption, although the rule employed should reflect the animal’s ecology and life history (see Section IV,D,3). We began by exploring the interactions between regulatory systems for protein and carbohydrate in juvenile L. migratoria fed 1 of an array of 25

26

STEPHEN J. SIMPSON AND DAVID RAUBENHEIMER

suboptimal foods (Raubenheimer and Simpson, 1993). Collectively, the resulting points of nutrient intake across the array of food rails formed a pattern where locusts on average moved to the geometrically closest point on their food rail to the intake target (Fig. 8a). This is termed the closest distance rule (closest distance optimization in previous papers). By following this rule the animal is minimizing the sum of undereating one nutrient and overeating the other, irrespective of which of the two happens to be in excess or deficit (Raubenheimer and Simpson, 1997). We next repeated the experiment, this time testing L. migratoria concurrently with the desert locust, S. gregaria. Although L. migratoria, as expected, showed the closest distance rule, S. gregaria expressed a different pattern of nutrient intake (Fig. 8b). These locusts moved along their food rail to the point where the sum of the two nutrients was the same as that at the intake target—the equal distance rule. Such an outcome could indicate that the two nutrients are interchangeable and hence treated as a single dimension by regulatory mechanisms controlling their intake (Simpson and Raubenheimer, 1999). However, we know that protein and carbohydrate are regulated separately in this species (hence, the defended intake target in protein–carbohydrate space). Also, protein and carbohydrate are not symmetrically interchangeable to a growing locust. The alternative explanation is that the desert locust maximizes its intake of protein and

Fig. 8. Arrays of intake shown by locusts given one of a range of nutritionally suboptimal foods. The arrays illustrate three different rules of compromise (see text). The closest distance rule (CD) is exhibited by the African migratory locust, Locusta migratoria, for protein and carbohydrate (a), whereas the desert locust, Schistocerca gregaria, demonstrates the equal distance rule (ED) (b). When L. migratoria is fed foods that are suboptimal in salt vs macronutrient content, they regulate macronutrient intake and abandon regulation of salt intake [the no interaction rule (NI)] (c). In each case the defended intake target is indicated, as measured in locusts with choices of complementary foods (after Raubenheimer and Simpson, 1993, 1997; Trumper and Simpson, 1993).

THE HUNGRY LOCUST

27

carbohydrate up to a combined limit (Raubenheimer and Simpson, 1997, 1999). Both Locusta and Schistocerca have similar intake targets for protein and carbohydrate, but the difference in their rules of compromise seems to relate to their respective ecologies (Raubenheimer and Simpson, 1997). The desert locust typically lives in arid environments and will eat a wide range of plant species. It appears to have evolved to take advantage of whatever foods are available, even if protein and carbohydrate are mixed in unbalanced ratios. The African migratory locust, on the other hand, lives in less extreme environments and specializes in eating grasses. Rather than maximize its intake, it balances protein and carbohydrate consumption to minimize errors. In another experiment (Trumper and Simpson, 1993) we forced Locusta to trade off regulation of protein and carbohydrate (combined as a single axis) against salt intake by giving them one of five suboptimal foods. Locusts defended protein and carbohydrate intake and abandoned salt regulation (Fig. 8c), even though they regulate salt intake precisely when allowed to do so independently of the macronutrients (Trumper and Simpson, 1993). We called this the no-interaction rule. The fact that macronutrients should have greater leverage than micronutrients in the control of amounts eaten is to be expected. The same effect is apparent when one macronutrient is required in small amounts relative to another, such as occurs in the sapfeeding aphids in which amino acids are effectively a micronutrient relative to carbohydrate (Abisgold et al., 1994). D. Dynamic Targets The next complexity that must be added is time. So far we have considered targets as points integrated over a given interval in an animal’s life. However, targets move, tracing trajectories at a range of time scales, both within the life of an individual over physiological and developmental time and across generations over evolutionary time. The functional aim of an animal’s regulatory systems is to track these trajectories across each timescale, taking advantage of information provided by feedbacks offering differing degrees of temporal resolution. 1. Physiological Time Tracking the intake target as it moves over minutes and hours poses a fundamental problem, which is neatly illustrated by considering an animal with two nutritionally complementary foods. Tracking a moving target precisely would involve the animal in constant switching between the two foods, with all the attendant costs that such dithering entails (Houston and Sumida, 1985). On the other hand, to switch too infrequently means spend-

28

STEPHEN J. SIMPSON AND DAVID RAUBENHEIMER

ing the majority of time well away from an optimal state along a jagged, zig-zag path through nutrient space. In theory, there should be some optimal compromise between tracking the intake target with extreme precision, perhaps even involving mouthful-by-mouthful switching, and switching very infrequently. The form of this compromise should be embodied in the time course of response of the mechanisms responsible for causing an animal to switch. During a run of meals on a suboptimal food an animal builds up nutrient deficits and excesses. As noted earlier, taste feedbacks and learned specific responses for protein and carbohydrate are apparent in locust nymphs by 4 h of feeding on an unbalanced food. What is the interval at which a locust with constant access to two complementary foods varying in protein and carbohydrate content will switch between the foods? Is it similar to what the mechanisms would suggest? If so, what is the functional significance of this time interval? We explored these issues by providing locusts with one of four complementary food pairings and observing the intervals at which they switched between the foods. Despite the fact that the treatments differed markedly in the protein and carbohydrate content of the paired foods, all treatments yielded the same average intake trajectories through nutrient space. Achieving the same trajectory in nutrient space involved insects adjusting the number of meals taken in a run of one food before switching to the other, according to the nature of the two foods provided. Locusts very seldom switched foods within a meal. Instead, when averaged across treatments, locusts switched every four or five meals, with an average interval of 4 h (Chambers et al., 1995; Fig. 9). Hence, there was congruence between the resolving power of physiological mechanisms controlling intake of protein and carbohydrate and the timing of switching shown by locusts with constant access to complementary foods. We investigated the functional significance of the 4-h switching interval by enforcing both longer and shorter switching intervals on locusts (Chambers et al., 1997). There was no obvious performance benefit to switching at 4-hourly intervals, but there may well be ecological advantages that cannot be measured in the laboratory. Studies of other insects feeding in the wild have shown that moving and feeding can be hazardous activities, attracting the attention of predators and parasites (Bernays, 1997). In addition to feeding on a suboptimal food, another source of perturbation in the relationship between the animal’s state and its intake target that requires a compensatory feeding response is a change in the animal’s behavior. For example, when adult locusts are flown for a period of 2 h they subsequently increase their intake of carbohydrate (used to fuel flight) but not protein (Raubenheimer and Simpson, 1997). Changes in tempera-

THE HUNGRY LOCUST

29

Fig. 9. Cumulative intake of protein and carbohydrate by six locust nymphs over a 16-h period. The nymphs were provided with constant access to two suboptimal but complementary synthetic foods, one containing a 2:1 ratio of protein to digestible carbohydrate and the other the reverse ratio. The average switching interval was 4 h (from Chambers et al., 1995).

ture and other abiotic influences are also likely to induce compensatory feeding responses. 2. Developmental Time Over a longer time scale, targets and rules of compromise change as animals grow, mature, reproduce, and senesce. Tracking developmental changes could involve developmentally ‘‘hardwired’’ switches between different types of intake control mechanisms as occurs, for instance, when a female mosquito becomes responsive to the smell of vertebrate hosts when she needs a bloodmeal for egg development (Bowen, 1991). In such cases the repertoire of mechanisms is employed in sequence in response to signals indicating that the animal has achieved an appropriate developmental state. Alternatively, the same basic mechanism may operate throughout development, responding directly to developmentally changing tissue demands (Barton Browne, 1995). Locusts appear to provide an example of this latter principle. During the first 10 days after molting to adulthood, locusts alter the amounts and ratios of protein and carbohydrate they select and defend

30

STEPHEN J. SIMPSON AND DAVID RAUBENHEIMER

when given pairs of nutritionally complementary foods (Chyb and Simpson, 1990). These changes in intake accompany variation in protein, cuticle, and lipid growth. They are also closely associated with changes in the relative firing rate of mouthpart taste receptors to stimulation with amino acids and sugars, with the responses to sugars and amino acids being greatest during periods when insects both require and ingest high levels of carbohydrate and protein, respectively (C. L. Simpson et al., 1990a). It is likely that changing demands of the tissues for protein and carbohydrate as the insect ages result in variation in levels of amino acids and sugars in the blood, producing altered gustatory responsiveness and hence affecting feeding behavior (C. L. Simpson et al., 1990a). A different kind of developmental influence was found when nymphal locusts were reared in more or less chemically rich environments and their chemoreceptors were counted as adults (Rogers and Simpson, 1997). If kept for the last two nymphal stages with a nutritionally optimal synthetic food, they developed 15% fewer chemoreceptors on their mouthpart palps and antennae than if they had been fed on the more chemically stimulating, but no more nutritious, wheat seedlings. When only the smell of wheat was provided with the artificial food, the number of odor receptors on the antennae increased, whereas mouthpart taste receptors remained impoverished. Increasing the variety of taste stimuli experienced by the mouthparts, either through providing two nutritionally complementary synthetic foods with widely different protein and carbohydrate contents or by addition of nonnutrient secondary plant compounds, resulted in an increase in the number of mouthpart taste hairs developed. The implication is that the number of chemoreceptors developed at the next molt is a function of the richness of the chemical environment experienced by a local receptor population during the previous nymphal stage. Similar plasticity in antennal receptor numbers has been found in another grasshopper (Bernays and Chapman, 1998). 3. Evolutionary Time Tracking nutritional trajectories across generations can involve both genetic and epigenetic transfer of information. Locusts provide a classic example of the latter since phase change between the ‘‘solitarious’’ and ‘‘gregarious’’ forms is transmitted across generations via maternally produced (and paternally influenced) chemical signals (Bouaichi et al., 1995; Islam et al., 1994a,b; McCaffery et al., 1998; Simpson et al., 1999). We have recently explored the nutritional strategies of the two phases and discovered that whereas the gregarious form follows the equal distance (ED) rule of compromise for balancing protein and carbohydrate intake (Fig. 8b), solitarious desert locusts use the closest distance (CD) rule (Simp-

THE HUNGRY LOCUST

31

son and Raubenheimer, 1999a). As discussed in Section IVC, the latter rule of compromise is also demonstrated by the grass specialist, L. migratoria (Fig. 8a). We suggested that the different rules reflect the two species’ diet breadths. Consistent with such an explanation is the fact that the solitarious form of S. gregaria is restricted in its host plant range, both by the paucity and lack of diversity of food plants in its desert habitat, and by the fact that it is considerably less mobile than when in the gregarious phase. Targets and rules of compromise are subject to genetic selection and would be expected to track changes occurring in an animal’s nutritional environment over evolutionary time. It might also be expected that comparative analysis would enable trajectories to be plotted and matched against life-history characteristics. Indeed this is the case, as we have shown in comparative analyses of the protein to carbohydrate requirements of 117 species of insects using phylogenetically independent contrasts (Simpson and Raubenheimer, 1993b). Just as targets and rules of compromise are subject to natural selection, so too must the design features of the regulatory systems that achieve them be explicable, and to a degree predictable, in terms of feedbacks over evolutionary time. We used an evolutionary perspective to derive a unifying model for integrating studies of feeding behavior, chemosensory physiology, and nutrient feedbacks (Simpson and Raubenheimer, 1996).

E. Unifying Feeding Behavior, Sensory Physiology, and Nutrient Feedbacks: the ‘‘Taste Model’’ Regulating nutrient intake requires two sources of information, the first indicating the composition of the food and the second the nutritional state of the animal. The major source of information regarding the composition of the food is food-related sensory cues, preeminent among which is taste. Information about an animal’s nutritional circumstances exists at several levels of temporal resolution. At one extreme, information is hardwired in the genotype of the organism, whereas at the other extreme there are rapid metabolic feedbacks. The former provides a prediction based on the experience of the animal’s ancestors but it lacks the resolution to track changes at shorter time scales, whereas the latter provides an indication of an individual’s current state and offers the opportunity to track requirements as they change in real time. Given that the major source of information regarding the composition of food is taste stimuli, the design features of the gustatory system (including both the peripheral sense organs and the central processing circuits) should show the imprint of feedbacks operating across all timescales.

32

STEPHEN J. SIMPSON AND DAVID RAUBENHEIMER

We begin by making predictions about how evolutionary feedbacks should have affected the taste system. First, it is expected that gustatory sensitivity to key nutrients should have evolved. Second, the taste system should respond to variation in the concentration of these nutrients in the food in a manner that reflects both average ancestral requirements and the spatial and temporal distribution of nutrients in the environment. Third, these dose–responses are expected to be subject to modulation through more current feedback mechanisms, the nature of which should reflect ancestral experience of variation in requirements imposed through development and environmental perturbations. The first of these predictions is probably met by all organisms. Locusts, for instance, have taste receptors that respond to amino acids, sugars, and various salts. The second prediction is more interesting. Comparative analyses of insects show that the intake target tracks changes in the availability of nutrients in the environment: The optimal food and defended intake target for a foliage-feeding locust is quite different from that of a sap-sucking aphid (Raubenheimer and Simpson, 1999; Simpson and Raubenheimer, 1993b). In both cases, the optimal food rail is comfortably within the range of nutrient ratios available in the animal’s natural foods. As a result, animals encounter foods containing both more than and less than the optimal concentration of a given nutrient. If provided with two such complementary foods, the animal must mix them in a particular ratio to achieve its optimal food composition as follows: p1e ⫹ p2 (1 ⫺ e) ⫽ pt

(1)

where p1, p2, and pt are the proportions of a given nutrient in foods 1, 2, and the optimal food, respectively, e is the proportion of total intake from food 1, and (1 ⫺ e) is the proportion of total intake from food 2. The problem facing the animal is to solve Eq. (1) for e. The taste system can provide such a solution if the power of a food to stimulate and maintain feeding (its ‘‘phagostimulatory power’’) is a direct function of e. Thus, the phagostimulatory power (P ) of a food should be maximal for a food containing the optimal concentration of a nutrient and decrease rapidly toward higher and lower concentrations according to Eq. (2): Px ⫽ (( pt /( pt ⫹ 兩pt ⫺ px兩)) ⫺ 0.5) ⫻ 200

(2)

where pt and px are the proportions of a nutrient in the optimal food and food x, respectively. By subtracting 0.5 and multiplying by 200, P is scaled so that the optimal food has a value of 100. The equation can be expanded to include more than one nutrient, and it produces response surfaces that have a peak at the multidimensional optimal concentration (Simpson and Raubenheimer, 1996).

THE HUNGRY LOCUST

33

An animal whose taste system is designed according to such an equation will demonstrate several adaptive behavioral responses. First, it will mix its intake of two or more complementary foods to provide the optimal concentration of nutrients in its diet. Second, it will eat predominantly from the optimal food, if it is available. Third, it will ingest most of the food that is closest to being optimal if all available foods are suboptimal and noncomplementary in composition. We tested this model using locusts in several ways. The first experiment involved regulation of salt intake. Locusts given complementary foods regulate their intake of salt vs protein and carbohydrate to a target concentration of 1.8% dry weight of salt mixture in the food (Trumper and Simpson, 1993). The prediction from Eq. (2) for the phagostimulatory power of foods varying in salt concentration is shown in Fig. 10. Also presented is the actual relationship as measured experimentally (Simpson, 1994). The two are very similar. Another test of the model involved providing locusts with a food containing an optimal ratio of protein to carbohydrate along with a food that differed by varying amounts from the optimum. The taste model predicts the extent to which the insects should misallocate their feeding to the suboptimal food. Again, the predicted and actual values were very similar (Chambers et al., 1997).

Fig. 10. Graphs showing (a) predictions of the taste model (see text) for the relationship between phagostimulatory power of a food and its salt content and (b) data from locusts that support the predictions. Peak phagostimulatory power was found for a food with 1.8% salt, which is the concentration defended by locusts able to select between complementary foods.

34

STEPHEN J. SIMPSON AND DAVID RAUBENHEIMER

The parameters of the taste system as defined by the mathematical model are set in evolutionary time, indicating ancestrally prevailing ecological circumstances and average nutrient requirements. Although a taste system based on the model can enable an animal to locate and defend an optimal food composition (i.e., find the optimal food rail), it cannot determine how much of the optimal food needs to be eaten to reach the intake target (i.e., where on the optimal food rail to stop). Nor can the model cope with nutritional perturbation within the life of an animal. Predictable changes in nutrient requirements with development may become embodied through evolution in the design of the taste system, with response properties changing as part of the default developmental program (see Section IV,D,2). However, such mechanisms lack the resolution to respond to shorter term nutritional perturbations. It is here that short-term feedbacks such as bloodborne nutrient feedbacks onto gustatory sensitivity and learning come into play (Simpson and Raubenheimer, 1996). As has been shown, locusts possess an abundance of such responses. F. The Relationship between Pre- and Postingestive Regulatory Systems Since the focus of this review is on behavior, we largely neglect experimental and theoretical work that deals with postingestive processing of nutrients. It is important to emphasize, however, that ingestive and postingestive mechanisms are interdependent, and dealing with the nature of this relationship has been one of our major aims. Whereas animals are constrained to ingest the particular ratio of nutrients present in available foods, they are nevertheless able to ‘‘jump rails’’ postingestively by differentially utilizing ingested nutrients. Hence, the growth target may be reached from a range of intake points in nutrient space. Locusts show highly efficient defense of a growth target in protein–carbohydrate space (Raubenheimer and Simpson, 1993; Fig. 7). Their major mechanism for defending growth is to assimilate and metabolize protein and carbohydrate with high efficiency up to their growth target levels and then to get rid of excess ingested nutrient by excretion in the case of nitrogen and enhanced respiration in the case of carbohydrate (Zanotto et al., 1993, 1994, 1997). Other experiments have investigated the relationship between diet composition and activity of gut endocrine cells (Zudaire et al., 1998). Discussion of methodological and statistical issues associated with quantifying postingestive regulation may be found in Raubenheimer (1995) and Raubenheimer and Simpson (1992, 1994, 1995). Finally, the dynamic interactions between pre- and postingestive mechanisms regulating intake and utilization of macronutrients have recently been considered in a class of

THE HUNGRY LOCUST

35

models based on the marginal value theorem (Raubenheimer and Simpson, 1996, 1998). These models provide a formal means of dealing with the relationship between rates and efficiencies at various stages of nutrient processing and, most important, the interactions of these rates and efficiencies among different nutrient groups.

V. Adding Nonnutrient Dimensions: Plant Secondary Compounds In addition to the various nutrients combined within a food item there may be a less welcome cocktail of nonnutritive compounds. In plants, such compounds include secondary compounds (allelochemicals), many of which are thought to serve as a defense against herbivory. There is a vast literature dealing with the sensory detection of allelochemicals and their effects on insect (including locust) feeding behavior (Bernays and Chapman, 1994; Schoonhoven et al., 1998). We discuss only one aspect—the interaction between allelochemicals and nutrients. Our first attempts at studying the interaction predated (and stimulated) the development of the geometric models and involved providing locusts (both Schistocerca and Locusta) with artificial foods varying in protein, carbohydrate, and tannic acid content (Raubenheimer, 1992; Raubenheimer and Simpson, 1990). We recently expanded these original experiments in the context of the geometric framework (S. J. Simpson and D. Raubenheimer, unpublished data). The key outcome is that the degree to which addition of tannic acid to the food has detrimental effects on growth and survivorship and reduces food intake is a function of the balance of protein and carbohydrate in the food. When foods contain a near-optimal ratio of the macronutrients, there is no measurable consequence on food intake or performance of adding up to 10% tannic acid. However, as the food becomes more unbalanced in its protein to carbohydrate ratio, tannic acid has greater deleterious effects which are particularly pronounced when foods contain a lower than optimal ratio of protein to carbohydrate. In summary an animal that is able to balance its nutrient intake will be less susceptible to the deterrent and toxic effects of the allelochemical.

VI. Adding Spatial Complexity Most of our experiments on food selection have been carried out in arenas in which the various foods are provided in equal amounts and close together. A critical ecological dimension is distance and spatial relationships between foods. We are currently exploring these effects by distributing

36

STEPHEN J. SIMPSON AND DAVID RAUBENHEIMER

foods at different distances and in differing frequencies (S. Behmer, D. Raubenheimer and S. J. Simpson, unpublished data; Simpson and Raubenheimer, 1999a). The aim is to begin to incorporate ecological complexity into the experimental system in a controlled way. Individual-based models of the interactions between locusts and the spatial structure of their environment have proved extremely useful in understanding the dynamics of phase change (Collett et al., 1998; Simpson et al., 1999).

VII. Using Real Plants and from the Laboratory into the Real World Our research program has been primarily laboratory based, involving the extensive use of chemically defined synthetic foods. As demonstrated previously, the general approach has been to extend this system from the simple toward the increasingly complex, progressively incorporating an increased range of food components and environmental contexts. This approach has provided an understanding of behavioral processes and their physiological control and has given rise to a set of conceptual tools for dealing with the complexities of nutrient and nutrient–allelochemical interactions. However, laboratory experiments necessarily involve a priori decisions about the factors to include and discard in an experimental design, giving rise to the danger that biologically important contexts might inadvertently be discarded. In an attempt to avert (or minimize) this problem, we initiated studies at the other end of the spectrum—in the full complexity of a natural environment. Thus, Raubenheimer and Bernays (1993) performed detailed observations of the feeding behavior of the polyphagous grasshopper Taeniopoda eques in its desert environment in the southwestern United States. Individual insects were observed continuously for 12 h, between sunrise and sunset, and all locomotion events, contacts with potential food items, and the ensuing behavior (rejection vs feeding, meal duration and pauses within meals, and the identity of the food items) were recorded. Detailed analysis of the pattern of feeding demonstrated remarkable similarities to that observed in the laboratory. Most notable was the fact that there was no appreciable difference in the regularity of meal taking and the duration of meals compared with that observed in the relatively simplified laboratory environment. This was encouraging, perhaps indicating that our laboratory model has captured many of the most relevant causal factors underlying locust feeding. Additional studies of this system, concentrating on food selection behavior, have been carried out by Bernays et al. (1992) and Howard et al. (1994).

THE HUNGRY LOCUST

37

Another approach to the general problem of achieving a satisfactory balance between complexity and experimental control is the use of genetically modified food plants (single-gene mutants of Arabidopsis) that differ systematically in the ratio of nutrients in their tissues. Again, it was encouraging to find that results obtained using the relatively simple synthetic foods could be replicated against the full biochemical complexity of living plant tissues (Wright, 1998; G. Wright, D. Raubenheimer, and S. J. Simpson, manuscript in preparation).

VIII. From Locusts to Vertebrates Have locusts helped us to understand feeding behavior in animals other than insects? In recent years we have started to apply our integrative models of feeding and nutrition to published data on vertebrates. Reinterpretation of data on macronutrient selection in rats and hens (Raubenheimer and Simpson, 1997; Simpson and Raubenheimer, 1997, 1999b) has indicated that the insect-based models are of general relevance. Our approach offers insights hitherto unavailable using existing approaches which have largely failed to deal with the interactive, multidimensional nature of nutritional systems. With the extraordinary advances being made at the molecular level in the control of feeding and appetite in vertebrates, there is an urgent need to interpret the effects of particular genetic or pharmacological manipulations within an organismic framework. Unless this framework deals with the complex, interactive influences on feeding behavior, there is a serious risk of misinterpreting the role and action of particular manipulations. Since the ultimate aim of much research on mammals is either to provide clinical solutions to human obesity or to improve the efficiency of animal husbandry, the consequences of such mistakes are potentially serious.

IX. Summary We have reviewed a research program on feeding and nutrition that has used the locust as its principal subject. We began by analyzing the patterns of feeding of locusts provided with ample, nutritious food under controlled environmental conditions. As has been shown for a range of other animals (from rats to humans), even under such simplified conditions feeding patterns are complex in temporal structure and vary considerably between individuals. By employing a novel combination of statistical techniques (including the use of proportional hazards models), physiological proce-

38

STEPHEN J. SIMPSON AND DAVID RAUBENHEIMER

dures, and simulation modeling, we now have a good understanding of the basis of this complexity. The next step was to determine whether locusts have regulatory systems for individual nutrients. Behavioral and electrophysiological investigations indicated that locusts regulate their intake of protein, carbohydrate, salt, and water. The mechanisms involved were determined to include nutrientspecific modulation of taste receptor responsiveness and learning, including learned specific appetites for protein and carbohydrate. We then addressed a key problem in the study of feeding and nutrition, namely, how to deal with the interactions occurring between nutrients. An inability to deal with the multidimensional and interactive nature of nutrition is one of the major failings in the current understanding of nutritional systems. We used experiments on locusts to develop and test a geometric, state–space framework for considering feeding and nutrition. The scheme allows representation of an animal’s multiple nutritional needs and present state, identifies the composition of an optimal food, enables interpretation of the trade-offs reached between overeating some nutrients and undereating others, provides a means of integrating the study of feeding behavior with that of postingestive physiology, and places physiological control mechanisms in an ecological and evolutionary context. Having introduced and illustrated the geometric framework, we briefly outlined ongoing experiments in which we have begun to introduce ecological complexity in a controlled manner by expanding the models to include nonnutrient dimensions (allelochemicals) and to take account of spatial complexity in food distribution. Another approach to introducing ecological reality has been to record feeding patterns of grasshoppers in the field. We have also begun to move from chemically defined synthetic foods to the chemical complexity of plant tissues under laboratory conditions through the use of single-gene mutants of the plant Arabidopsis. Finally, we referred to studies in which we used our geometric models to reinterpret published data on macronutrient selection in vertebrates and in so doing demonstrated the general utility of the approach. Acknowledgments We express our profound thanks for all they have done to Reg Chapman and Liz Bernays and our graduate students, research assistants, and technicians. Among the various granting agencies that have funded the work over the years, particular thanks goes to the Biotechnology and Biological Sciences Research Council (BBSRC) for its support.

References Abisgold, J. D., and Simpson, S. J. (1987). The physiology of compensation by locusts for changes in dietary protein. J. Exp. Biol. 129, 329–346.

THE HUNGRY LOCUST

39

Abisgold, J. D., and Simpson, S. J. (1988). The effect of dietary protein levels and haemolymph composition on the sensitivity of the maxillary palp chemoreceptors of locusts. J. Exp. Biol. 135, 215–229. Abisgold, J. D., Simpson, S. J., and Douglas, A. E. (1994). Responses of the pea aphid (Acyrthosiphon pisum) to simultaneous variations in dietary amino acid and sugar levels. Physiol. Entomol. 19, 95–102. Barton Browne, L. (1995). Ontogenetic changes in feeding behavior. In ‘‘Regulatory Mechanisms of Insect Feeding’’ (R. F. Chapman and J. de Boer, Eds.), pp. 307–342. Chapman & Hall, New York. Barton Browne, L., Moorhouse, J. E., and van Gerwen, A. C. M. (1975a). An excitatory state generated during feeding in the locust, Chortoicetes terminifera. J. Insect Physiol. 21, 1731–1735. Barton Browne, L., Moorhouse, J. E., and van Gerwen, A. C. M. (1975b). Sensory adaptation and the regulation of meal size in the Australian plague locust, Chortoicetes terminifera. J. Insect Physiol. 21, 1633–1639. Belovsky, G. E. (1990). How important are nutrient constraints in optimal foraging models or are spatial/temporal factors more important? In ‘‘Behavioral Mechanisms of Food Selection’’ (R. N. Hughes, Ed.), NATO ASI Series, Vol. 20, pp. 255–278. SpringerVerlag, Berlin. Bernays, E. A. (1977). The physiological control of drinking behaviour in nymphs of Locusta migratoria. Physiol. Entomol. 2, 261–273. Bernays, E. A. (1980). The post-prandial rest in Locusta migratoria and its hormonal regulation. J. Insect Physiol. 26, 119–123. Bernays, E. A. (1990). Water regulation. In ‘‘The Biology of Grasshoppers’’ (R. F. Chapman and A. Joern, Eds.), pp. 129–141. Wiley, New York. Bernays, E. A. (1993). Food aversion learning. In ‘‘Insect Learning’’ (A. C. Lewis and D. Papaj, Eds.), pp. 1–17. Chapman & Hall, New York. Bernays, E. A. (1995). Effects of experience on feeding. In ‘‘Regulatory Mechanisms of Insect Feeding’’ (R. F. Chapman and J. de Boer, Eds.), pp. 279–306. Chapman & Hall, New York. Bernays, E. A. (1997). Feeding by lepidopteran larvae is dangerous. Ecol. Entomol. 22, 121–123. Bernays, E. A., and Chapman, R. F. (1972a). The control of changes in peripheral sensilla associated with feeding in Locusta migratoria (L.). J. Exp. Biol. 57, 755–763. Bernays, E. A., and Chapman, R. F. (1972b). Meal size in nymphs of Locusta migratoria. Entomol. Exp. Appl. 15, 399–410. Bernays, E. A., and Chapman, R. F. (1973). The regulation of feeding in Locusta migratoria. Internal inhibitory mechanisms. Entomol. Exp. Appl. 16, 329–342. Bernays, E. A., and Chapman, E. A. (1974a). Changes in haemolymph osmotic pressure in Locusta migratoria in relation to feeding. J. Entomol. A 48, 149–155. Bernays, E. A., and Chapman, R. F. (1974b). The effects of haemolymph osmotic pressure on the meal size of nymphs of Locusta migratoria L. J. Exp. Biol. 61, 473–480. Bernays, E. A., and Chapman, R. F. (1974c). The regulation of food intake by acridids. In ‘‘Experimental Analysis of Insect Behaviour’’ (L. Barton Browne, Ed.), pp. 48–59. Springer-Verlag, Berlin. Bernays, E. A., and Chapman, R. F. (1977). Deterrent chemicals as a basis of oligophagy in Locusta migratoria. Ecol. Entomol. 2, 1–18. Bernays, E. A., and Chapman, R. F. (1994). ‘‘Host–Plant Selection by Phytophagous Insects.’’ Chapman & Hall, New York. Bernays, E. A., and Chapman, R. F. (1998). Phenotypic plasticity in numbers of antennal chemoreceptors in a grasshopper: Effects of food. J. Comp. Physiol. A 183, 69–76. Bernays, E. A., and Mordue, A. J. (1973). Changes in palp tip sensilla of Locusta migratoria in relation to feeding: The effects of different levels of hormone. Comp. Biochem. Physiol. A 45, 451–454.

40

STEPHEN J. SIMPSON AND DAVID RAUBENHEIMER

Bernays, E. A., and Raubenheimer, D. (1991). Dietary mixing in grasshoppers: Changes in acceptability of different plant secondary compounds associated with low levels of dietary proteins. J. Insect Behav. 4, 545–556. Bernays, E. A., and Simpson, S. J. (1982). Control of food intake. Adv. Insect Physiol. 16, 59–118. Bernays, E. A., and Singer, M. S. (1998). A rhythm underlying feeding behaviour in a highly polyphagous caterpillar. Physiol. Entomol. 23, 295–302. Bernays, E. A., Blaney, W. M., and Chapman, R. F. (1972). Changes in chemoreceptor sensilla on the maxillary palps of Locusta migratoria in relation to feeding. J. Exp. Biol. 57, 745–753. Bernays, E. A., Bright, K., Howard, J. J., Raubenheimer, D., and Champagne, D. (1992). Variety is the spice of life: Frequent switching between foods in the polyphagous grasshopper Taeniopoda eques. Anim. Behav. 44, 721–731. Blaney, W. M., and Duckett, A. M. (1975). The significance of palpation of the maxillary palps of Locusta migratoria (L.): An electrophysiological and behavioural study. J. Exp. Biol. 63, 701–712. Bouaichi, A., Roessingh, P., and Simpson, S. J. (1995). An analysis of the behavioural effects of crowding and re-isolation on solitary-reared adult desert locusts (Schistocerca gregaria, Forska˚l) and their offspring. Physiol. Entomol. 20, 199–208. Bowen, M. F. (1991). The sensory physiology of host-seeking behavior in mosquitoes. Annu. Rev. Entomol. 36, 139–158. Brobeck, J. R. (1955). Neural regulation of food intake. Ann. N.Y. Acad. Sci. 63, 44–55. Carle, E. (1969). ‘‘The Very Hungry Caterpillar.’’ Hamish Hamilton, New York. Castonguay, T. W., Kaiser, L. L., and Stern, J. S. (1986). Meal pattern analysis: Artefacts, assumptions and implications. Brain Res. Bull. 17, 439–443. Cazal, M. (1969). Actions d’extraits de corpora cardiaca sur le peristaltisme intestinal de Locusta migratoria. Arch. Zool. Exp. Gen. 110, 83–89. Chambers, P. G., Raubenheimer, D., and Simpson, S. J. (1997). The selection of nutritionally balanced foods by Locusta migratoria: The interaction between food nutrients and added flavours. Physiol. Entomol. 22, 199–206. Chambers, P. G., Raubenheimer, D., Simpson, S. J. (1998). The functional significance of switching interval in food mixing by Locusta migratoria. J. Insect Physiol. 44, 77–85. Chambers, P. G., Simpson, S. J., and Raubenheimer, D. (1995). Behavioural mechanisms of nutrient balancing in Locusta migratoria nymphs. Anim. Behav. 50, 1513–1523. Champagne, D. E., and Bernays, E. A. (1991). Phytosterol suitability as a factor mediating food aversion learning in the grasshopper Schistocerca americana. Physiol. Entomol. 16, 391–400. Chapman, R. F. (1995). Chemosensory regulation of feeding. In ‘‘Regulatory Mechanisms of Insect Feeding’’ (R. F. Chapman and J. de Boer, Eds.), pp. 101–136. Chapman & Hall, New York. Chapman, R. F., Ascoli-Christensen, A., and White, P. R. (1991). Sensory coding for feeding deterrence in the grasshopper Schistocerca americana. J. Exp. Biol. 158, 241–259. Chyb, S., and Simpson, S. J. (1990). Dietary selection in adult Locusta migratoria L. Entomol. Exp. Appl. 56, 47–60. Collett, M., Despland, E., Simpson, S. J., and Krakauer, D. C. (1998). Spatial scales of locust gregarization. Proc. Natl. Acad. Sci. USA, 95, 13052–13055. Davis, J. D., Collins, B. J., and Levine, M. W. (1978). The interaction between gustatory stimulation and gut feedback in the control of the ingestion of liquid diets. In ‘‘Hunger Models’’ (D. A. Booth, Ed.), pp. 109–143. Academic Press, London. Dethier, V. G. (1957). Communication by insects: Physiology of dancing. Science 125, 331–336.

THE HUNGRY LOCUST

41

Dethier, V. G. (1976). ‘‘The Hungry Fly.’’ Harvard Univ. Press, Cambridge, MA. Dethier, V. G., Solomon, R. L., and Turner, L. H. (1965). Sensory input and central excitation and inhibition in the blowfly. J. Comp. Physiol. Psychol. 60, 303–313. Giza, B. K., and Scott, T. R. (1983). Blood glucose selectively affects taste-evoked activity in rat nucleus tractus solitarius. Physiol. Behav. 31, 643–650. Giza, B. K., and Scott, T. R. (1987). Intravenous insulin infusions in rats decrease gustatory evoked responses to sugars. Am. J. Physiol. 252, R994–R1002. Hanson, F. E. (1987). Chemoreception and the fly: The search for the liverwurst receptor. In ‘‘Perspectives in Chemoreception and Behavior’’ (R. F. Chapman, E. A. Bernays, and J. G. Stoffolano, Jr., Eds.), pp. 99–122. Springer-Verlag, New York. Houston, A. I., and Sumida, B. (1985). A positive feedback model for switching between two activities. Anim. Behav. 33, 315–325. Howard, J. J., Raubenheimer, D., and Bernays, E. A. (1994). Population and individual polyphagy in the grasshopper Taeniopoda eques. Entomol. Exp. Appl. 71, 167–176. Hughes, R. N. (Ed.) (1993). ‘‘Diet Selection: An Interdisciplinary Approach to Foraging Behaviour.’’ Blackwell Sci., Oxford. Islam, M. S., Roessingh, P., Simpson, S. J., and McCaffery, A. R. (1994a). Effects of population density experienced by parents during mating and oviposition on the phase of hatchling desert locusts. Proc. R. Soc. London B 257, 93–98. Islam, M. S., Roessingh, P., Simpson, S. J., and McCaffery, A. R. (1994b). Parental effects on the behavior and coloration of nymphs of the desert locust, Schistocerca gregaria. J. Insect Physiol. 40, 173–181. Langton, S. D., Collett, D. and Sibly, R. M. (1995). Splitting behaviour into bouts. Behaviour 132, 781–799. Mattson, W. J. (1980). Herbivory in relation to nitrogen. Annu. Rev. Ecol. Sys. 11, 119–161. McCaffery, A. R., Simpson, S. J., Islam, M. S., and Roessingh, P. (1998). A gregarizing factor present in egg pod foam of the desert locust Schistocerca gregaria. J. Exp. Biol. 201, 347–363. McCleery, R. H. (1977). On satiation curves. Anim. Behav. 25, 1005–1115. McFarland, D. J., and Sibly, R. (1972). ‘‘Unitary drives’’ revisited. Anim. Behav. 20, 548–563. McNeill, S., and Southwood, T. R. E. (1978). Role of nitrogen in the development of insect– plant relations. In ‘‘Biochemical Aspects of Plant and Animal Coevolution’’ ( J. Harborne, Ed.), pp. 77–98. Academic Press, New York. Mordue, W. (1969). Hormonal control of Malpighian tubule and rectal function in the desert locust, Schistocerca gregaria. J. Insect Physiol. 15, 273–285. Panksepp, J. (1978). Analyses of feeding patterns: Data reduction and theoretical implications. In ‘‘Hunger Models’’ (D. A. Booth, Ed.), pp. 143–146. Academic Press, London. Pulliam, H. R. (1975). Diet optimization with nutrient constraints. Am. Nat. 109, 765–768. Raubenheimer, D. (1992). Tannic acid, protein and digestible carbohydrate: Dietary imbalance and nutritional compensation in the African migratory locust. Ecology 73, 1012–1027. Raubenheimer, D. (1995). Problems with ratio analysis in nutritional studies. Funct. Ecol. 9, 21–29. Raubenheimer, D., and Bernays, E. A. (1993). Patterns of feeding in the polyphagous grasshopper Taeniopoda eques: A field study. Anim. Behav. 45, 153–167. Raubenheimer, D., and Blackshaw, J. (1994). Locusts learn to associate visual stimuli with drinking. J. Insect Behav. 7, 569–575. Raubenheimer, D., and Ga¨de, G. (1993). Compensatory water intake in locusts: Implications for mechanisms regulating drink size. J. Insect Physiol. 39, 275–281. Raubenheimer, D., and Ga¨de, G. (1994). Hunger–thirst interactions in the locust, Locusta migratoria. J. Insect Physiol. 40, 631–639.

42

STEPHEN J. SIMPSON AND DAVID RAUBENHEIMER

Raubenheimer, D., and Simpson, S. J. (1990). The effects of simulataneous variation in protein, digestible carbohydrate and tannic acid on the feeding behavior of larval Locusta migratoria and Schistocerca gregaria—I: Short-term studies. Phsiol. Entomol. 15, 219–233. Raubenheimer, D., and Simpson, S. J. (1992). Analysis of covariance: An alternative to nutritional indices. Entomol. Exp. Appl. 62, 221–231. Raubenheimer, D., and Simpson, S. J. (1993). The geometry of compensatory feeding in the locust. Anim. Behav. 45, 953–964. Raubenheimer, D., and Simpson, S. J. (1994). The analysis of nutrient budgets. Funct. Ecol. 8, 783–791. Raubenheimer, D., and Simpson, S. J. (1995). Constructing nutrient budgets. Entomol. Exp. Appl. 77, 99–104. Raubenheimer, D., and Simpson, S. J. (1996). Meeting nutrient requirements: The roles of power and efficiency. Entomol. Exp. Appl. 80, 65–68. Raubenheimer, D., and Simpson, S. J. (1997). Integrative models of nutrient balancing: Application to insects and vertebrates. Nutr. Res. Rev. 10, 151–179. Raubenheimer, D., and Simpson, S. J. (1998). Nutrient transfer functions: The site of integration between feeding behaviour and nutritional physiology. Chemoecology 8, 61–68. Raubenheimer, D., and Simpson, S. J. (1999) Integrating nutrition: A geometrical approach. Entomol. Exp. Appl., 91, 67–82. Raubenheimer, D., and Tucker, D. (1997). Pairing of visual cues with the separate consumption of protein and carbohydrate. Anim. Behav. 54, 1449–1459. Roessingh, P., and Simpson, S. J. (1984). Volumetric feedback and the control of meal size in Schistocerca gregaria. Entomol. Exp. Appl. 36, 279–286. Roessingh, P., Bernays, E. A., and Lewis, A. C. (1985). Physiological factors influencing preferences for wet and dry food in Schistocerca gregaria. Entomol. Exp. Appl. 37, 89–94. Rogers, S. M., and Simpson, S. J. (1997). Experience-dependent changes in the number of chemosensilla on the mouthparts and antennae of Locusta migratoria. J. Exp. Biol. 200, 2313–2321. Rogers, S. M., and Simpson, S. J. (1999). Chemodiscriminatory neurones in the sub-oesophageal ganglion of Locusta migratoria L. Entomol. Exp. Appl., 91, 19–28. Rolls, E. T. (1989). Information processing in the taste system in primates. J. Exp. Biol. 146, 141–164. Schoonhoven, L. M., Jermy, T., and van Loon, J. J. A. (1998). ‘‘Insect–Plant Biology: From Physiology to Evolution.’’ Chapman & Hall, London. Sibly, R. M. Nott, H. M. R., and Fletcher, D. J. (1990). Splitting behaviour into bouts. Anim. Behav. 39, 63–69. Simmonds, M. S. J., Simpson, S. J., and Blaney, W. M. (1992). Dietary selection behaviour in Spodoptera littoralis: The effects of conditioning diet and conditioning period on neural responsiveness and selection behaviour. J. Exp. Biol. 162, 73–90. Simpson, C. L. (1990). Dietary compensation by Locusta migratoria: Aspects of physiology and behaviour D.Phil. thesis, University of Oxford, Oxford. Simpson, C. L., Chyb, S., and Simpson, S. J. (1990a). Changes in chemoreceptor sensitivity in relation to dietary selection by adult Locusta migratoria L. Entomol. Exp. Appl. 56, 259–268. Simpson, C. L., Simpson, S. J., and Abisgold, J. D. (1990b). The role of various amino acids in the protein compensatory response of Locusta migratoria. Symp. Biol. Hungar. 39, 39–46. Simpson, S. J. (1981). An oscillation underlying feeding and a number of other behaviours in fifth instar Locusta migratoria nymphs. Physiol. Entomol. 6, 315–324.

THE HUNGRY LOCUST

43

Simpson, S. J. (1982). Patterns of feeding: A behavioural analysis using Locusta migratoria nymphs. Physiol. Entomol. 7, 325–336. Simpson, S. J. (1983). The role of volumetric feedback from the hindgut in the regulation of meal size in fifth instar Locusta migratoria nymphs. Physiol. Entomol. 8, 451–467. Simpson, S. J. (1990). The pattern of feeding. In ‘‘A Biology of Grasshoppers’’ (R. F. Chapman and T. Joern, Eds.), pp. 73–103. Wiley, New York. Simpson, S. J. (1992). Mechanoresponsive neurones in the suboesophageal ganglion of the locust. Physiol. Entomol. 17, 351–369. Simpson, S. J. (1994). Experimental support for a model in which innate taste responses contribute to regulation of salt intake by nymphs of Locusta migratoria. J. Insect Physiol. 40, 555–559. Simpson, S. J. (1995). The control of meals in chewing insects. In ‘‘Regulatory Mechanisms of Insect Feeding’’ (R. F. Chapman and J. de Boer, Eds.), pp. 137–156. Chapman & Hall, New York. Simpson, S. J., and Abisgold, J. D. (1985). Compensation by locusts for changes in dietary nutrients: Behavioural mechanisms. Physiol. Entomol. 10, 443–452. Simpson, S. J., and Bernays, E. A. (1983). The regulation of feeding: Locusts and blowflies are not so different from mammals. Appetite 4, 313–346. Simpson, S. J., and Ludlow, A. R. (1986). Why locusts start to feed: An analysis of causal factors. Anim. Behav. 34, 480–496. Simpson, S. J., and Raubenheimer, D. (1993a). The central role of the haemolymph in the regulation of nutrient intake in insects. Physiol. Entomol. 18, 395–403. Simpson, S. J., and Raubenheimer, D. (1993b). A multi-level analysis of feeding behaviour: The geometry of nutritional decisions. Philos. Trans. R. Soc. London B 342, 381–402. Simpson, S. J., and Raubenheimer, D. (1995). The geometric analysis of feeding and nutrition: A user’s guide. J. Insect Physiol. 41, 545–553. Simpson, S. J., and Raubenheimer, D. (1996). Feeding behaviour, sensory physiology and nutrient feedback: A unifying model. Entomol. Exp. Appl. 80, 55–64. Simpson, S. J., and Raubenheimer, D. (1997). The geometric analysis of feeding and nutrition in the rat. Appetite 28, 201–213. Simpson, S. J., and Raubenheimer, D. (1999a). Assuaging nutritional complecity: a geometrical approach. Proc. Nutrit. Soc. 58, 1–11. Simpson, S. J., and Raubenheimer, D. (1999b). Geometric models of macronutrient selection. In ‘‘Neural Control of Macronutrient Selection’’ (H.-R. Berthoud and R. J. Seeley, Eds.). CRC Press, Boca Raton, FL. Simpson, S. J., and Simpson, C. L. (1990). The mechanisms of nutritional compensation by phytophagous insects. In ‘‘Insect–Plant Interactions’’ (E. A. Bernays, Ed.), Vol. 2, pp. 111–160. CRC Press, Boca Raton, FL. Simpson, S. J., and Simpson, C. L. (1992). Mechanisms controlling modulation by haemolymph amino acids of gustatory responsiveness in the locust. J. Exp. Biol. 168, 269–287. Simpson, S. J., and White, P. R. (1990). Associative learning and locust feeding: Evidence for a ‘‘learned hunger’’ for protein. Anim. Behav. 40, 506–513. Simpson, S. J., Simmonds, M. S. J., and Blaney, W. M. (1988a). A comparison of dietary selection behaviour in larval Locusta migratoria and Spodoptera littoralis. Physiol. Entomol. 13, 225–238. Simpson, S. J., Simmonds, M. S. J., Wheatley, A. R., and Bernays, E. A. (1988b). The control of meal termination in the locust. Anim. Behav. 36, 1216–1227. Simpson, S. J., Simmonds, M. S. J., Blaney, W. M., and Jones, J. P. (1990). Compensatory dietary selection occurs in larval Locusta migratoria but not Spodoptera littoralis after a single deficient meal during ad libitum feeding. Physiol. Entomol. 15, 235–242.

44

STEPHEN J. SIMPSON AND DAVID RAUBENHEIMER

Simpson, S. J., Simmonds, M. S. J. Blaney, W. M., and James, S. (1991). Variation in chemosensitivity and the control of dietary selection behaviour in the locust. Appetite 17, 141–154. Simpson, S. J., Raubenheimer, D., and Chambers, P. G. (1995). Nutritional homeostasis. In ‘‘Regulatory Mechanisms of Insect Feeding’’ (R. F. Chapman and J. de Boer, Eds.), pp. 251–278. Chapman & Hall, New York. Simpson, S. J., McCaffery, A. R., and Ha¨gele, B. (1999). The behavioural analysis of phase change in the desert locust. Biol. Rev., in press. Slater, P. J. B., and Lester, N. P. (1982). Minimising errors in splitting behaviour into bouts. Behaviour 79, 153–161. Stephens, D. W., and Krebs, J. R. (1986). ‘‘Foraging Theory.’’ Princeton Univ. Press, Princeton, NJ. Szentesi, A., and Bernays, E. A. (1984). A study of behavioural habituation to a feeding deterrent in nymphs of Schistocerca gregaria. Physiol. Entomol. 9, 329–340. Tolkamp, B. J., Allcroft, D. J., Austin, E. J., Nielsen, B. L., and Kyriazakis, I. (1999). Satiety splits feeding behaviour into bouts. J. Theor. Biol., in press. Trumper, S., and Simpson, S. J. (1993). Regulation of salt intake by nymphs of Locusta migratoria. J. Insect Physiol. 39, 857–864. Trumper, S., and Simpson, S. J. (1994). Mechanisms regulating salt intake in fifth instar nymphs of Locusta migratoria. Physiol. Entomol. 19, 203–215. White, T. C. R. (1993). ‘‘The Inadequate Environment: Nitrogen and the Abundance of Animals.’’ Springer, Berlin. Wright, G. A. (1998). The effects of variation in plant chemistry on the pattern of feeding of two generalist insect herbivores. D.Phil. thesis, University of Oxford, Oxford. Zanotto, F. P., Simpson, S. J., and Raubenheimer, D. (1993). The regulation of growth by locusts through post-ingestive compensation for variation in the levels of dietary protein and carbohydrate. Physiol. Entomol. 18, 425–434. Zanotto, F. P., Raubenheimer, D., and Simpson, S. J. (1994). Selective egestion of lysine by locusts fed nutritionally unbalanced foods. J. Insect Physiol. 40, 259–265. Zanotto, F. P., Raubenheimer, D., and Simpson, S. J. (1996). Haemolymph amino acid and sugar levels in locusts fed nutritionally unbalanced diets. J. Comp. Physiol. B 166, 223–229. Zanotto, F. P., Gouveia, S. M., Simpson, S. J., Raubenheimer, D., and Calder, P. C. (1997). Nutritional homeostasis in locusts: Is there mechanism for increased energy expenditure during carbohydrate overfeeding? J. Exp. Biol. 200, 2437–2448. Zudaire, E., Simpson, S. J., and Montuenga, L. M. (1998). Effects of food nutrient content, insect age and stage in the feeding cycle on the FMRFamide immunoreactivity of diffuse endocrine cells in the locust gut. J. Exp. Biol. 201, 2971–2979.

ADVANCES IN THE STUDY OF BEHAVIOR, VOL. 29

Sexual Selection and the Evolution of Song and Brain Structure in Acrocephalus Warblers Clive K. Catchpole school of biological sciences royal holloway, university of london egham, surrey tw0 0ex, united kingdom

I. Introduction In 1871 Charles Darwin published The Descent of Man and Selection in Relation to Sex, which set out in detail his theory of sexual selection. He suggested that sexual selection was the driving force behind the evolution of puzzling and apparently maladaptive male traits such as the peacock’s tail. He also noted that only male nightingales sang and suggested that their beautiful and elaborate songs may somehow serve to charm the females of the species. Darwin (1871) recognized that there were two possible ways in which such male characters conferred an evolutionary advantage in reproduction. First, males could compete among themselves, with the victors acquiring females for reproduction. This he called intrasexual selection, which would result in selection for traits such as size to increase success in male–male competition. Second, males could influence females directly by producing more attractive courtship displays. This he called intersexual selection, which would result in selection for more elaborate and conspicuous male displays. Darwin’s theory of intrasexual selection was readily accepted because aggressive males could be seen fighting for possession of females, and the concept fitted comfortably with Victorian notions of strong, superior males. Intersexual selection by female choice was much less obvious, and the idea of evolution driven by more discriminating females was hotly disputed. Indeed, sexual selection by female choice remained a discredited and untested aspect of sexual selection theory until quite recent times. A theoretical paper by Trivers (1972) reignited the debate by suggesting that greater female investment would lead to females becoming the more choosy and discriminating sex. This led to a renewed interest in sexual selection and to a dramatic growth in studies on female choice. A host of theoretical and 45

Copyright 䉷 2000 by Academic Press All rights of reproduction in any form reserved. 0065-3454/00 $35.00

46

CLIVE K. CATCHPOLE

empirical studies flourished in the 1980s, culminating in the publication of a major review by Andersson (1994). As Andersson’s (1994) review shows, many experimental studies have confirmed that Darwin was correct about the importance of female choice. It is ubiquitous and widespread across taxa, there no longer seems to be any reasonable doubt about its existence, and it has been demonstrated particularly well in field and laboratory studies of birds. There is also no doubt about the obvious reproductive advantage which displaying males gain from attracting females. However, one central question still remains the subject of great debate: What benefits, if any, do females obtain from selecting a particular male? There are two main possibilities which are usually classified into direct and indirect benefits (Kirkpatrick and Ryan, 1991). Direct benefits include obtaining superior resources from the male, which in birds usually means a larger territory containing more food, or help in the demanding task of feeding the young. If male song is used as a cue in female choice, then the question arises as to how song might serve as an honest indicator of either male or territory quality. Zahavi (1975, 1977) proposed that such signals might indeed reflect male quality as long as the signal entailed some cost to the signaler. This prevents individuals of low quality from deceiving the female because they cannot afford to pay the costs involved. During evolution females will have been selected to respond to songs which reveal some aspects of male quality. In doing so, however, they will also have to pay the increased costs of energy spent during the process of sampling and assessing different males. These arguments also apply to the other class of benefits that females might obtain—indirect genetic benefits. Such benefits do not accrue to the female herself but are passed on indirectly via her offspring. This can occur by two different mechanisms—either Fisher’s (1930) ‘‘runaway’’ process or by ‘‘good genes.’’ In Fisherian models, the male trait is quite arbitrary and only develops as a genetic correlation because of the female preference for it. The trait and the preference then become linked in a coevolutionary process, but the female only benefits when her ‘‘sexy sons’’ inherit the attractive trait and so increase their fitness. In good genes models, the genetically correlated male trait is not arbitrary but acts as an ‘‘honest indicator’’ of male quality. Females benefit indirectly because their offspring will acquire good genes for increased viability. Finally, there remains the possibility that females obtain no benefits at all from selecting a particular male. Therefore, why are they attracted by his song? One explanation is the ‘‘sensory exploitation hypothesis.’’ Ryan et al. (1990), working on frog acoustics, suggested that singing males simply tap into the female nervous system and exploit some preexisting sensory

SEXUAL SELECTION AND EVOLUTION IN Acrocephalus WARBLERS

47

bias for a sound signal. This requires no benefit to the female: She is simply exploited by singing males. However, how does all this theory apply to bird songs, and how can we test any predtions? Testing evolutionary hypotheses using field studies and laboratory experiments has obvious difficulties and dangers. Current utility may not reflect past evolutionary advantage, and even captive zebra finches are less convenient than Drosophila for studies on genetics. However, as recent reviews have pointed out (Catchpole and Slater, 1995; Searcy and Yasukawa, 1996), the study of bird songs has been a productive and innovative field, contributing significantly to the explosive growth of studies on sexual selection and female choice in recent years. Nowhere has this been more apparent than in the many studies on European Acrocephalus warblers, a group whose complex songs I have often described as ‘‘acoustic peacock’s tails.’’ In the following review, I describe how a series of studies throughout Europe—descriptive, comparative, and from both laboratory and field—have combined to increase our understanding of sexual selection by female choice.

II. Songs of Acrocephalus Warblers When I first heard the song of an Acrocephalus warbler, more than 30 years ago, I was a young PhD student searching for a project for my thesis at Nottingham. I can vividly remember my first encounter with a singing male. It was a still, summer morning, and a faint mist hung over the riverbank and the marshes. The silence was broken by a loud, harsh, chattering song, and a pale, tiny bird took to the air from a nearby bush. The song was complicated and continuous, with a rhythm that seemed to gradually increase in tempo as the bird ascended. As the male reached the top of what I now know to be its song flight, it seemed to reach a crescendo and produced a dazzling variety of notes in quick succession. It opened its wings and glided slowly back to its perch, with the rhythm and complexity of the song declining during the descent. It was an impressive display and I was struck by this unexpected virtuosity from such a small, brown bird which apparently no one had studied. At this stage, I had no access to analytical equipment, and it was several years before I was to discover just how variable and complicated the songs really were. The following account is based on my first sonagraphic analysis of both temporal and sequential organization of sedge warbler song (Catchpole, 1976). Although many other papers have since dealt with the analysis of Acrocephalus songs, this one established the methodology for estimating

48

CLIVE K. CATCHPOLE

what has become known as syllable repertoire size, and it remains the most detailed and revealing study of their extraordinary song complexity. The detailed structure of a typical song is illustrated in Fig. 1. Songs are composed from small units called syllables. Syllables can be defined (Catchpole, 1976) as either ‘‘a continuous trace on a sonagram’’ or ‘‘the smallest unit in a song which can occur independently.’’ They are easily recognized and classified, and the reader will have no difficulty in recognizing, for example, syllables 23 and 7 which are used to finish song 1 and start song 2 (Fig. 1). This illustrates one of the basic rules of composition for a sedge warbler song. Two syllables (in this case 23 and 7) are alternated at the end of one song and then used to start the next song. Although superficially this seems quite a simple rule it has enormous scope for variation, particularly in the number of syllable repetitions used. In song 2, syllable 23 is initially repeated eight times, then five, then two, then three, then four, and so on in an unpredictable pattern. Its alternating partner, syllable 7, is repeated less, in bouts of between one and four, but again in a quite unpredictable way. There is also a gradual increase in tempo as the

Fig. 1. A complete song recorded from a male sedge warbler, together with the end of the song that preceded it and the start of that which followed. New syllable types are classified and numbered when they first appear (after Catchpole, 1976).

SEXUAL SELECTION AND EVOLUTION IN Acrocephalus WARBLERS

49

song builds up to the central section. In this central section a sudden switch occurs, and 5 new syllables (27, 16, 40, 34, and 20) are then injected in sequence. This is the climax of the song, and in a song flight there may be up to 10 new syllables injected here. Two of these syllables (in this case 40 and 20) are then used to form the alternating pattern which will end song 2 and start song 3. In the middle of song 3, a new batch of syllables will be injected, and so on. The important point to note is that these few simple rules of composition result in a stream of long songs (some can be more than 1 min long) which are all different from each other. The sedge warbler is unusual in this respect because most songbirds tend to repeat a short, fixed song of 2 or 3 sec several times before switching to another (Catchpole and Slater, 1995). We have examined thousands of sedge warbler songs over the years but never found any evidence that a particular song is ever exactly repeated. Even when the same two syllables occur together in the first section of another song, their patterns of repetition and alternation are quite different. To date, our attempts to determine more precise rules in formal sequence analysis (Catchpole, 1976) have failed. The sedge warbler seems to be a composer of songs and rather like a jazz musician in performance. Describing the sedge warbler song as an acoustic equivalent of the peacock’s tail is certainly no exaggeration. Even in all its undoubted glory, the peacock’s tail is in reality a much more stereotyped signal than the constantly changing song of the sedge warbler male. At first, the infinite variety of sedge warbler songs appears to pose problems for measuring and comparing complexity between individual males. This potential problem has been overcome by using the most basic building block, the syllable, as the unit of measurement. As seen in Fig. 1, syllables can be recognized and classified, and the total number a male has in a sample of songs is known as repertoire size. Although we have also used other measures, such as the mean number of syllables per song (Buchanan and Catchpole, 1997), syllable repertoire size remains the most meaningful and robust index of song complexity. As discussed later, this measure has also proved useful in our later studies on the song control pathway in the brain. How many songs are needed to give a reasonably accurate comparison between males? If the number of new syllable types is plotted against the number of songs analyzed (Fig. 2), individual differences in repertoire size quickly become apparent, even after 10 songs. This is because the sudden switch in the middle of a song, which produces 5–10 new syllables each time, means that most of the repertoire is quickly cycled through. It seems designed to reveal the male’s repertoire as quickly as possible, and as discussed later this may well be important regarding the function of his

50

CLIVE K. CATCHPOLE

Fig. 2. Cumulative plots of new syllable types produced by two male sedge warblers A and B. After 20 consecutive songs the repertoire size of male A is estimated as 73 and that of male B as 37 (data and figure provided by K. L. Buchanan).

song. In practice, we take a sample of 20 consecutive songs because by this stage we calculate from even longer runs that we will have between 95 and 100% of the real repertoire. Our measure of syllable repertoire size is therefore an estimate, but a fairly good estimate, which enables us to make comparisons within and between species. Using this technique, I first estimated the repertoire range of the sedge warbler in a tiny sample of three males to be between 35 and 55 syllables (Catchpole, 1976). In a recent study on a larger population of 20–30 males over 3 years, we estimated the real range as 35–75 (Buchanan and Catchpole, 1997). To date, the repertoire ranges of five European Acrocephalus species have been estimated, and the total variation is 25–100 (Szekely et al., 1996). The significance of this variation within the genus will be discussed at length later, but here the species with the largest repertoire and most complex songs is worth special mention. The marsh warbler A. palustris has been studied by Dowsett-Lemaire (1979) and has the largest repertoire size in the group (approximately 100). How the marsh warbler male acquires his large repertoire is particularly interesting because it is the result of extensive mimicry. Dowsett-Lemaire first studied the marsh warbler on its breeding grounds in Belgium. She found that about half of the repertoire was acquired by interspecific mimicry

SEXUAL SELECTION AND EVOLUTION IN Acrocephalus WARBLERS

51

of species in the local avifauna. Populations breeding near the coast imitated marine species not found in the repertoires of individuals further inland. Where did the other half come from? Like all European Acrocephalus species, the marsh warbler is a trans-Saharan migrant spending the European winter in central Africa. By careful detective work Dowsett-Lemaire found that most of the other remaining syllables were acquired from the African avifauna. Some of the African endemic species are extremely local in their distribution, and it is even possible to map part of the migratory route by studying the contents of the repertoire. On average, an individual male repertoire consists of 76 imitations of which 31 are European and 45 African. It is curious that whenever an ornithologist (or a marsh warbler) hears the song there is no confusion—it is definitely a marsh warbler. Therefore, how is species recognition maintained? The answer is in the rules of composition that the marsh warbler uses to construct his songs. The different syllables from different species are never repeated for very long; instead, they are rapidly sequenced in a dazzling display of vocal virtuosity. This remarkable study reveals that one particular Acrocephalus warbler has gone to unusual lengths in order to acquire a vast repertoire. The marsh warbler is able to compose even more elaborate songs than the sedge warbler. What is behind this extraordinary drive toward increasingly elaborate songs?

III. Song, Context, and Mating System A. Song and Context It is likely that the first indication of the function of a behavior pattern will come from quite simple, contextual observations, and this has certainly been true in the case of song in Acrocephalus warblers. In the examples that follow, quantitative observations and measurements have first been taken in the field, and these have led to the formation of hypotheses concerning the functions of song. Many predictions have then been tested directly in the field using playback experiments. Three species in particular have been subjected to this approach. 1. The Sedge Warbler I have already described how impressed I was when observing and listening to my first singing sedge warbler male. However, this did not to last long because a few days after he was caught and ringed he became silent and disappeared before I had time to record him. I was not only frustrated

52

CLIVE K. CATCHPOLE

but also extremely puzzled. According to the established dogma of the day, male birds sang to defend their territories against rival males. I was expecting my star male to continue his singing for most of the breeding season and not become silent after only a week. A day later, I managed to locate the elusive male and determine the reason for his silence. He was following a female very closely, and she was carrying nest material. When she left him and flew some distance in search of more material, he uttered a few quieter songs until her return. Apart from that, he never sang again for the rest of the breeding season, but he was still quite aggressive and chased off intruding males using a short threat call. Clearly, he was perfectly capable of maintaining his territory without the use of song. This pattern was repeated with the other males as they became paired, and the conclusion seemed obvious. The few days of intense and elaborate song were for female attraction. When I told my story to colleagues they were skeptical. How thorough were my observations? Was I there at dawn to check the males? I decided to mount a series of complete 24-h timebudget studies on males before and after pairing. The results for a typical sedge warbler male are shown in Fig. 3. The unpaired male has a characteristic diurnal rhythm, starting just before dawn and reaching a peak soon after. Singing then declines throughout the day as feeding activity gradually increases. There is a smaller evening peak of singing activity at dusk and then a few hours of silence, although a few males continued to sing throughout the 24 h. As the histogram shows, there is a massive overall investment in singing activity, presumably at considerable cost. Equally impressive is the empty histogram obtained on any day after pairing. It was true that mated males never sang even for short periods of time at dawn or dusk. Clearly, the sedge warbler is an extreme case because most songbirds show only a decline after pairing and no total cessation. A parallel study on the sympatric reed warbler A. scirpaceus (Catchpole, 1973) revealed a similar pattern of massive song output before pairing, but paired male reed warblers did produce short bursts of song at dawn and dusk. Reed warblers breed much closer together than sedge warblers in close-packed ‘‘colonies’’ of tiny territories. It seemed to me that there was a greater threat of territorial incursions and also of a mate being stolen. Today, we would also predict a greater threat of extra-pair copulations, and so all these factors may help to explain the retention of some singing activity after pairing. Although the evidence for song as a mate attractant seems strong, the hypothesis clearly needed further testing—but how? A playback experiment to test it directly seemed impractical in the dense marshes, but a test to eliminate the alternative hypothesis that song functions in territorial defense was possible. I carried out a simple playback experiment (Catch-

SEXUAL SELECTION AND EVOLUTION IN Acrocephalus WARBLERS

53

Fig. 3. Diurnal rhythms of song in a male sedge and a male reed warbler (a) before pairing and (b) after pairing. Arrows indicate approximate times of sunrise and sunset (after Catchpole, 1973).

pole, 1973) on paired and unpaired males of both species using recorded songs of the other species as controls. Although I used only a simple response/no-response score, the trend was quite clear. All the unpaired males used some song in their response, although this was more of an interruption of their spontaneous song as they silently approached the speaker. Paired male reed warblers gave some song, and although paired male sedge warblers responded by approach they never sang at all. Later, a more detailed series of experiments was carried out on another population of sedge warblers (Catchpole, 1977). This time many different categories of response were measured, but they confirmed the same pattern of response. Although paired male sedge warblers responded aggressively, they never used song in territorial defense. 2. The Great Reed Warbler The great reed warbler A. arundinaceus, as its name suggests, is the largest of the European species. Like its smaller congeners, the reed and

54

CLIVE K. CATCHPOLE

sedge warbler, the unpaired male is highly vocal and produces long, loud songs containing many different syllables until a female is attracted. As in the other Acrocephalus species, there are also changes in the song when the male becomes paired. In a field study in Germany (Catchpole, 1983) I noticed that the paired male does not become completely silent, but instead produces much shorter, simpler, and quieter songs containing only two or three different syllable types (Fig. 4). The two types of songs are so different in structure that I named them ‘‘short’’ and ‘‘long’’ songs, respectively. In terms of actual length, long songs are about 4 s in duration compared to only 1 s for short songs. Although 4 s is comparatively short compared to the length of sedge warbler songs, the great reed warbler has the same Acrocephalus habit of composing each song from a repertoire of syllables, so there is still enormous variation and complexity compared to songs of many other songbirds. Further observations revealed that long and short songs were given in quite different contexts giving valuable clues regarding their probable functions. As in the sedge warbler, the long songs were given in a typical diurnal rhythm with peaks at dawn and dusk. In a later Swedish study Hasselquist et al. (1993) found that 7000–8000 songs were produced each day, but as in my study these stopped when the male became paired. In contrast, short songs were produced in no regular pattern but rather at intervals throughout the day. The main stimulus seemed to be an intruding male, who was then chased off. These observations suggest that the function of the long song is in female attraction, and that the short song functions in territorial

Fig. 4. A long song and two short songs recorded from a male great reed warbler (from Catchpole, 1983).

SEXUAL SELECTION AND EVOLUTION IN Acrocephalus WARBLERS

55

defense. The great reed warbler also has another complication: Males may attract a second or even a third female and eventually become polygynous. In another study on a German population (Catchpole et al., 1985), we noticed that quiet, paired males would move to the other end of their large reed bed territory and start singing long songs to attract another female. A prospecting female would find it difficult to detect that such a male was already paired, and we suggested that such males were effectively deceiving their mates. In Sweden, Hasselquist and Bensch (1991) found that 40% of their males were polygynous, and that they also left their first female and resumed long songs when attempting to attract another. They found the same pattern of use in long and short songs and found that short songs were used particularly to guard females from rival males. In an earlier study (Catchpole, 1983), I also designed a series of playback experiments in the field to test the hypothesis that short songs were aggressive signals used between males. In the first experiment, 10 territorial males were played 4 min of spontaneous song recorded from an unpaired male. Four were already paired and silent, but all approached the speaker singing short songs. The other six were unpaired and singing long songs; they all switched to short songs and approached the speaker. The second experiment was designed to further test this idea. If short songs really are threat signals, then males should be more likely to avoid them than long songs during an approach. An additional 10 males were exposed to tapes of another male which switched from short to long or long to short. Those which heard long songs first soon approached close to the speaker, but when the tape changed to the same male singing short songs they moved several meters away. Those which heard short songs first were slower to approach and stayed further away. When the tape switched to long songs they approached much closer to the speaker. Taken together, these results offer strong support for the hypothesis that short songs are perceived by rival males as aggressive signals, used in territorial defense and mate guarding. They also suggest, by default, that long songs may well have another function. 3. The Aquatic Warbler The aquatic warbler A. paludicola is closely related to the sedge warbler. It inhabits the marshes of eastern Europe and, as will be shown, is very different in many aspects of its behavior, particularly its mating system. In a population from the Biebrza marshes in Poland, Catchpole and Leisler (1989) found that its song is also quite different in structure from that of sedge warbler song (Fig. 5) and probably also different in function. Instead of long songs, the aquatic warbler tends to produce short, repetitive songs which we classified in terms of length and complexity. A songs are the shortest and simplest, consisting of a single rattle or buzz. B songs consist

56

CLIVE K. CATCHPOLE

Fig. 5. Different types of song (A–C) recorded from a male aquatic warbler (from Catchpole and Leisler, 1989).

of two phrases—a buzz being followed by a slower phrase containing more complex syllables. This seems to be the most common and characteristic song of the aquatic warbler. Finally, C songs consist of many syllables and are the longest, most complex, and most like those of other Acrocephalus species. Observations showed that when single males were advertising, they sang mostly in B–C mode, but in interactions with other males A–B was more common. This interpretation, and the song classification, was supported by a study by Dyrcz and Zdunek (1993). They also noted that the diurnal rhythm was slightly different, with the main peak of activity at dusk and a smaller one at dawn. Perhaps, as for the great reed warbler, the shorter songs are used in aggressive interaction with other males. We decided to test this hypothesis with a similar series of playback experiments in the field (Catchpole and Leisler, 1989). As before, the first experiment was a simple one, with a tape of a male singing in B–C mode played back to 10 males. Before playback the 10 males were singing spontaneously in B–C mode, but during playback this changed dramatically. All 10 males approached the speaker, and as they did so they switched from B–C to A–B. Close to the speaker they produced

SEXUAL SELECTION AND EVOLUTION IN Acrocephalus WARBLERS

57

only A songs. The reduction in C songs suggests that these are not important in male interaction, whereas the sudden appearance of A songs suggests that they signal aggression toward males. This was tested further by a second, more complex experiment. We prepared a tape which had another male singing exclusively C songs before he was approached by a rival male and then switched entirely to mode A–B. We also produced another tape in which this order was reversed. Ten additional males were exposed to both tapes in random order on different days. The results confirmed the findings of our first experiment but also showed that males did not approach as closely when A–B songs were played. There was also a significant inverse correlation between the distance to the speaker and the number of A songs that responding males produced. As they came nearer and became more aggressive, more A songs were produced. As in the great reed warbler, there is evidence that shorter, simpler songs are used in aggressive interactions with rival males. The aquatic warbler in particular seems to reserve a special song (the A song) for close aggressive interactions. This may be related to its unusual sexual behavior; the aquatic warbler has a highly promiscuous mating system. DNA fingerprinting has now shown that multiple paternity occurs in more than half of the broods, and some have three or four different fathers (Schulze-Hagen et al., 1993). This has led to intense sperm competition, and male aquatic warblers spend the longest time copulating, have the largest testes, and produce more sperm than the other Acrocephalus species (Schulze-Hagen et al., 1995). Intense sperm competition may well explain why male aquatic warblers have developed a special aggressive song which they use at close quarters against rival males. 4. Other European and Tropical Studies Fessl and Hoi (1996) studied the moustached warbler A. melanopogon in Austria. As with reed and marsh warblers, the song is long and complex and used primarily for mate attraction. It is usually started by a few introductory whistles which precede large numbers of later syllables delivered in complex warbling Acrocephalus fashion. Fessl and Hoi discovered what seems to be the origin of the dual-song structure found in great reed and aquatic warblers. When a female has been attracted, the warbling part decreases and the short whistles are used against intruding males. They confirmed this pattern in a series of playback experiments. Kelsey (1989) carried out an unusual study on the marsh warbler A. palustris in which he studied the context and function of song both in the breeding grounds in southern England and in the wintering grounds in Africa. In Europe he found the typical diurnal rhythm of the unmated, territorial Acrocephalus male, in which song decreased dramatically on

58

CLIVE K. CATCHPOLE

pairing (Dowsett-Lemaire, 1979). He then studied a wintering population in Zambia and found that the birds occupied and defended winter feeding territories. They produced very little song, and when they did it was shorter and much simpler than that in England. He also challenged the males with a playback experiment. Although they approached the speaker, none sang in response. In the Seychelles Islands in the Indian Ocean lives one of the rarest warblers—the Seychelles warbler A. sechellensis. Unlike the other species mentioned so far, the Seychelles warbler is a tropical resident and stays all year on its island home. Jan Komdeur and I visited Cousin Island to record and study its song (Catchpole and Komdeur, 1993). I had been told that the song was unusually short and simple for an Acrocephalus warbler, and it certainly seemed to be upon hearing it for the first time. A similar pattern had been found in the song of another island Acrocephalus, the Christmas Island warbler A. aequinoctialis, by Milder and Schreiber (1989). Sonagraphic analysis confirmed that the songs of the Seychelles warbler were only between 1 and 3 s long, and they did sound similar in the field. However, the analysis also revealed that each song was composed from different syllables, so the Seychelles warbler also had the Acrocephalus habit of producing song complexity. Being resident in a tropical environment, usually paired throughout the year, there was no obvious seasonal burst of song that is typical of the migrants in Europe. Instead, song was heard only rarely, usually when a neighbor encroached. We confirmed this by a simple playback experiment in which all the males responded with sudden bursts of their short songs. Territories were at a premium on Cousin: All were occupied and young males had to wait for a vacancy to occur. However, breeding activity has a rhythm of sorts and peaks twice a year. From counts of singing males taken by Komdeur over several years, we were able to determine that singing activity also peaked just before breeding activity. Although song is important in retaining a territory on Cousin, it may be equally important in attracting a mate, keeping her, and guarding the paternity of the single chick produced. Although these various studies are diverse, they add further evidence to support our developing hypothesis that, in Acrocephalus warblers, shorter, simpler songs have an aggressive function in male–male interactions and longer, more complex songs function in mate attraction. B. Song and Mating System Quite early in the study it became clear that there was also a relationship between song complexity and mating system in Acrocephalus warblers (Catchpole, 1980). In other words, it seemed that marsh, reed, and sedge

SEXUAL SELECTION AND EVOLUTION IN Acrocephalus WARBLERS

59

warblers had the longest and most complex songs, and these species had long been classified as monogamous in the literature. It was the polygamous species, the great reed and aquatic warblers, which had shorter, simpler songs. At first this was puzzling because, in theory, increased sexual selection pressure in polygamous species should lead to more elaborate songs and not simpler ones. Before I continue, I discuss in depth the mating systems of Acrocephalus warblers. In recent years, the increased interest in sexual selection (Andersson, 1994) has led to more detailed studies on the mating system of birds. With the advent of DNA fingerprinting it has also become clear that in many species previously classified as socially monogamous, the female may have offspring fathered by more than one male (Birkhead and Møller, 1992). Even in species classified as socially polygamous, a greater degree of multiple paternity than expected may occur. Extensive multiple paternity in an Acrocephalus species, the aquatic warbler, is one of the more spectacular discoveries of recent years (Schulze-Hagen et al., 1993). The mating systems of Acrocephalus warblers have been extensively reviewed by Leisler and Catchpole (1992) and in relation to the evolution of their songs by Catchpole (1995). I consider them in order of departure from social monogamy. First is the aquatic warbler, previously classified as polygamous and now known to be highly promiscuous, with females mating with several males leading to extensive multiple paternity. Next is the great reed warbler, for which several studies report varying degrees of social polygyny as high as 80%, and the proportion of extra-pair young approaches 3%. These two species were regarded as polygamous in my earlier study (Catchpole, 1980), and this crude classification is still useful for comparative purposes; the other four species (sedge, marsh, reed, and moustached) are still generally believed to be socially monogamous. However, the sedge warbler shows the most tendency to depart from this, with up to 20% of males becoming polygynous and extra-pair paternity up to 12% (Buchanan and Catchpole, 1999; Hasselquist and Langefors, 1998; Langefors et al., 1998). The marsh warbler is next, with up to 7% of males becoming polygynous and up to 3% of young extra-pair (Leisler and Wink, 1999). In the reed warbler polygyny approaches 4%, but the moustached warbler has never been recorded as departing from social monogamy. The last two species have yet to be DNA fingerprinted. In my earlier paper (Catchpole, 1980) I showed that the two polygamous species (aquatic and great reed) had shorter and simpler songs than the monogamous species. Just as our knowledge of mating systems has improved since then, so have the repertoire size measurements from several other studies, but the relationship still holds. However, this relationship seems counter to that predicted by sexual selection theory, which suggests

60

CLIVE K. CATCHPOLE

that increased selection pressure on polygamous males should lead to more elaborate traits. However, there may be an explanation for this, and the argument is as follows. The key point is the relationship between female choice and investment in polygamous and monogamous birds. Males in a polygamous system invest much less in incubating and feeding their offspring, and indeed the female is often deserted and left to raise the young on her own. This is always the case for the aquatic warbler, where the male invests only with sperm and in many different females. When selecting a male in this system, a female pays less attention to direct male quality and much more to territory quality. Great reed and aquatic warblers defend large, resourcebased territories, so a female is less likely to select a male directly by song quality and more likely to select him indirectly by the size and quality of his territory. Male song thus evolves though male–male intrasexual selection, producing the shorter, simpler songs used in territorial defense. This is in contrast with a female selecting a male in a monogamous system. Here male quality is important, and the male is needed to feed the young. Territories are smaller in all these species and long flights are often made to collect food outside. Here the female needs to select by male quality rather than territory quality. As discussed later, somehow male song reflects aspects of male quality, and males are selected directly on the quality of their songs by discriminating females. Intersexual selection by female choice leads to the evolution of longer and more complex songs—the evolution of the acoustic equivalent of the peacock’s tail. The trend in male parental investment in the six species fits very well with the measurements of song length and complexity and not only at the extremes. The sedge warbler seems to be intermediate in many respects because it departs most from social monogamy, the males invest less, and the song is of intermediate length and complexity. The comparative approach has derived a complex and convincing hypothesis for the evolution of different song structures, but many key predictions need to be tested. For example, do females really prefer males with more elaborate songs?

IV. Song and Female Choice It is a basic assumption of modern sexual selection theory (Andersson, 1994) that females choose their mates. However, female birds are less easy to observe than singing males, and before the 1980s there was little evidence for active female choice. One exception was the pioneering study of Howard (1974) on mockingbirds, Mimus polyglottos, famous for the complexity of their songs. Howard studied a wild population, estimated their repertoire

SEXUAL SELECTION AND EVOLUTION IN Acrocephalus WARBLERS

61

sizes, and found a significant inverse correlation between repertoire size and pairing date. In other words, males with the most complex songs attracted females before their rivals with simpler songs attracted them. This was a pivotal study, and one which inspired me to examine more closely another sedge warbler breeding population. A. Female Choice in the Field 1. The Sedge Warbler The sedge warbler population I studied (Catchpole, 1980) was in the Lea Valley just north of London. I caught and ringed all the males in one area as they arrived and then recorded their songs. When a male suddenly became silent, I knew he had been chosen by a female. As in Howard’s (1974) study, pairing date was taken as the point of female choice. Even in this small sample I obtained a highly significant inverse correlation between repertoire size and pairing date. It could have been that earlier arriving males or those with larger territories were also preferred by females, but neither settling pattern nor territory size were correlated with pairing date. Partial correlation analysis confirmed that repertoire size alone exerted a significant effect, and that females seemed to prefer males with more complex songs. This study did not investigate other aspects of singing behavior, such as the length or quality of individual songs, nor song flighting, which is an obvious feature of the sedge warbler singing performance. There was also an increasingly popular opinion in the sexual selection literature ( Johnstone, 1996) that females should be basing their choice on ‘‘multiple cues’’ rather than only on one. It was with these considerations that we designed our most recent field study on the sedge warbler (Buchanan and Catchpole, 1997). Using multivariate statistics we investigated a whole suite of possible cues from male and territory quality in a 3-year study. Only three cues were found to exert a significant effect on pairing date, and two of these were song cues. Both repertoire size and time spent song flighting were significantly inversely correlated with pairing date (Fig. 6). The other cue was territory size. This pattern was also remarkably consistent because the effects of both repertoire size and territory size were significant in all 3 years and song flighting in 2. Again, we were able to eliminate male settling pattern as a confounding variable. None of the three cues were correlated with each other in any year, and each had independent effects on pairing date. This suggests that females are indeed selecting males using multiple cues, but cues which reflect different aspects of male and territory quality. Song flighting must incur considerable energetic costs and may well be a good

62

CLIVE K. CATCHPOLE

Fig. 6. The relationships between repertoire size (left), song flighting time (right) and pairing date for 1994–1996 in a sedge warbler population. Lines have been fitted where the correlations are significant (p ⬍ 0.05) (after Buchanan and Catchpole, 1997).

SEXUAL SELECTION AND EVOLUTION IN Acrocephalus WARBLERS

63

indicator of male quality in terms of immediate condition and health. Territory size may give a female information regarding the direct benefits in terms of food resources she will obtain. What about repertoire size? In both our field studies (Catchpole, 1980; Buchanan and Catchpole, 1997) comprising 4 separate years, we find that male repertoire size correlates significantly with our measure of female choice. This suggests that the relationship between repertoire size and female choice is robust and important. Why repertoire size has been so favored in female choice remains a central question in this study. At this stage there are few clues from the field. There is a tendency for older males to have larger repertoires, but our sample of males of known age is still very low. However, in another recent field study in Hungary, we did find a significant correlation between male age and repertoire size (Birkhead et al., 1997). This suggests that females who select males with larger repertoires could also obtain good genes for viability for their offspring, and this hypothesis is considered further in Section V. 2. The Great Reed Warbler A population of the great reed warbler was studied for several years in southern Germany (Catchpole, 1986). The timing of pair formation was more difficult to determine with accuracy in this species. Because it is polygynous our criterion for female choice was the number of females that a male obtained. A detailed multivariate analysis revealed no significant correlations between repertoire size and various measures of male quality. There was also no correlation between repertoire size and territory size, but there was a significant positive correlation between repertoire size and the amount of reeds at the edge of a territory. This measure is known as edge length, and it may well be ecologically significant because it measures the length of the reed–water interface known to be important for foraging. We found a significant positive correlation between repertoire size and the number of females that a male obtained. However, when edge length was held constant in a partial correlation analysis, the correlation became nonsignificant. In a more detailed, long-term study in Sweden, Hasselquist et al. (1996) found repertoire size to be the most important predictor of male pairing success, and they also found a correlation between repertoire size and male age. As in the sedge warbler, it may be that repertoire size is an indicator of good genes for viability. During the same Swedish study, Bensch and Hasselquist (1992) carried out one of the most important studies on how female choice occurs in a wild population. Do females really choose between males, or do they simply settle with the first male to which they are attracted? The answers to such fundamental questions, frequently posed by

64

CLIVE K. CATCHPOLE

Fig. 7. The search pattern of a radio-tracked unpaired female great reed warbler in a reed bed over 2 days. The dots indicate known radio contact and male signs indicate singing territorial males. The track shows that the female visited several territories before finally returning to pair with her selected male (final choice point) (after Bensch and Hasselquist, 1992).

theoreticians developing rival models of female choice, had previously eluded those with the difficult job of actually studying female choice in the wild. Bensch and Hasselquist (1992) solved the problem of following elusive females in dense marshland by fitting radio transmitters to their females. Their results showed that females did search their reed bed thoroughly before choosing a particular male with which to breed (Fig. 7). They systematically visited all the territories in the reed bed where males were singing their long, complicated songs. The females visited an average of six territories before eventually selecting one particular male and took 2 or 3 days to make their choice. These results are important because they demonstrate that females are not just passively attracted; they actively sample several males and apparently reject some in favor of their chosen male. Theories of sexual selection by female choice (Andersson, 1994) have relied crucially on this assumption for many years, and an Acrocephalus warbler has now shown it to be valid. B. Female Choice in the Laboratory Although field studies are important, there are far too many potentially confounding variables beyond the control of the researcher. Although our

SEXUAL SELECTION AND EVOLUTION IN Acrocephalus WARBLERS

65

females appeared to be choosing males with more complex songs, perhaps the male had other qualities such as size, aggression, or age. One solution is to eliminate or control the other qualities in a laboratory experiment. After all, a tape-recorded song completely eliminates the male; therefore, any response can only be due to the song. Initially, this was a problem because captive females seemed unresponsive to playback of tape-recorded songs. This changed when Searcy and Marler (1981) developed a new technique. By implanting captive females with estradiol they found that females became highly receptive and manifested their interest by sexual (copulation solicitation) displays. These could be quantified by counting or timing, providing at last a sensitive behavioral measure of female choice. 1. The Sedge Warbler We were now ready to directly test whether female sedge warblers preferred more complicated songs. We hormone-implanted a group of captive females and prepared a range of experimental tapes to present under standardized conditions in a soundproof experimental chamber (Catchpole et al., 1984). In our first experiment, we wanted to check that the new technique really worked, so we exposed each female once to tapes of three different species; the sedge warbler, the closely related reed warbler, and a blackbird Turdus merula as an unrelated control. Each tape contained 4 min of continuous song from one individual male, amplitude was held constant, and the presentations were made on different days and in random order to control for habituation. The repertoire size of the sedge warbler was only 20, the reed warbler 80, and the blackbird 115. Although the females moved around in response to all the playbacks, they only gave full displays to sedge warbler song. The response was clearly species specific, even though the other songs were more complex. Having tested our technique and scoring system, we could now pose more interesting questions about the sexual preferences of the females. In my earlier field study (Catchpole, 1980) I had shown indirectly that females appeared to prefer males with more complex songs. To test this directly we decided to use tapes of the same males from the field study. The recordings we used were from the first male to pair (repertoire size 41), the last male to pair (repertoire size 14), and a male from the middle of the range (repertoire size 24). By doing this our experimental tapes embraced the whole range of natural variation in our small population. The results were very clear (Catchpole et al., 1984): The females displayed most to the male with the largest repertoire, an intermediate amount to the middle male, and least to the male with the smallest repertoire. The females in captivity showed exactly the same preference pattern as that of those that paired naturally in the wild.

66

CLIVE K. CATCHPOLE

Fig. 8. The relationship between repertoire size and the amount of sexual displays given to playback of experimental songs by captive female sedge warblers. Means and standard errors are shown for the group (from Catchpole et al., 1994).

Although this close agreement of field and laboratory results from the songs of the same males was particularly convincing, there may have been some confounding variables in the system. For example, although we used 4 min of continuous song, our males may have differed in other aspects of singing behavior besides repertoire size, such as singing rate or tonal quality. To control for this we designed a final experiment in this particular series to eliminate individual variation (Catchpole et al., 1984). The obvious way to do this was to use a recording from just one male. Because sedge warbler songs are long, complicated compositions, it is possible to construct several different tapes from the same male. We made seven different tapes with approximately equal numbers of syllables, the only variation being in syllable repertoire size. Over 7 days our captive females were exposed to one different tape each day in random order. We obtained a significant positive correlation between repertoire size and the amount of female display (Fig. 8). This experiment provides much stronger support for our hypothesis that female sedge warblers select males by their repertoire size. 2. The Great Reed Warbler In the great reed warbler, field studies suggested that the short songs functioned in aggressive interactions between rival males (Catchpole, 1983) and that long, complex songs with large repertoires functioned in female attraction (Catchpole, 1986). The hormone implant technique also allowed

SEXUAL SELECTION AND EVOLUTION IN Acrocephalus WARBLERS

67

us to test this hypothesis directly. The copulation solicitation displays were much stronger in this species and their duration could be timed to give a more sensitive measure of female choice. In the first experiment, three different test tapes were prepared, each containing 36 songs delivered in 4 min. As in the sedge warbler experiments, the males were selected from the population in which their success with females in the field was already known. The first tape consisted of short songs taken from a polygynous male with a repertoire size of 23, and the second contained long songs from the same male. A third tape contained long songs from a monogamous male with a much smaller repertoire size of only 13. The females were exposed to all the tapes only once each on different days and in random order. The most obvious result was that females showed no response to playback of short songs, lending indirect support to the hypothesis that these function in male–male aggression. All the females responded to the long songs of the same male, supporting the hypothesis that these are important in female attraction. Finally, the females responded significantly more to the tape with the larger repertoire size recorded from our successful polygynous male. This supports the finding from the field that males with larger song repertoires are more successful in attracting females. 3. The Aquatic Warbler Our findings regarding this species from the field suggested that, like the great reed warbler, shorter, simpler A–B songs functioned in male–male aggression and longer, more complex C songs in female attraction (Catchpole and Leisler, 1989). The hormone implant technique was less successful in the aquatic warbler: Our hormone-implanted females never produced clear solicitation displays in captivity. However, they did show phonotaxis in response to playback by orientating and then approaching the speaker. For our next experiments we therefore designed a large choice-chamber aviary 11 m long with speakers positioned at either end (Catchpole and Leisler, 1996). Females were placed in a central holding cage and tapes were played from one end selected at random. Staying in the central holding cage when it was opened was regarded as neutral behavior. Moving to the side in which the speaker was playing was regarded as attraction, and moving to the silent side was regarded as avoidance. The experimental tapes were made from the same individual used in the earlier field experiment and were edited to produce contrasting bouts of either A–B or C songs from the same individual. To control for side effects, each female was exposed to both tapes played from both sides, a total of four experiments carried out over 4 different days in random order. Females were more attracted

68

CLIVE K. CATCHPOLE

Fig. 9. Attraction to and avoidance of different song structures by captive female aquatic warblers. Means and standard errors are shown for the group (from Catchpole and Leisler, 1996).

by C songs than by A–B songs of the same male (Fig. 9). In fact, there was evidence that females avoided A–B songs. These results support our field experiments, which showed that shorter, simpler songs were given in aggressive interactions, and provide much–needed direct evidence that longer and more complex songs do function in female attraction.

V. Song and Reproductive Success The hypothesis that females are attracted by more complex songs has received considerable support from both laboratory and field studies. In the field, we have shown that males in a population with more complex songs have a reproductive advantage by attracting females earlier than do their rivals with less complex songs. In the more controlled environment of the laboratory, we confirmed that tapes of the same males with more complex songs caused females to display more. However, for sexual selection to work, we need to return to the field to determine whether such males really enhance their reproductive success by creating more offspring. There are obvious difficulties in such an approach, not only in following individuals in the field but also in estimating their true genetic reproductive success. Despite this, recent studies of song in Acrocephalus warblers have

SEXUAL SELECTION AND EVOLUTION IN Acrocephalus WARBLERS

69

produced important findings in the wider field of sexual selection and reproductive success. The study of reproductive success in birds has been revolutionized by the advent of DNA fingerprinting (Birkhead and Møller, 1992), and I already discussed the incidence of extra-pair paternity in Acrocephalus warblers. Not only is the estimation of true, genetic reproductive success important but also it may be that patterns of female choice vary when selecting either a social mate or an extra-pair male. This is illustrated by two species in which song complexity and its relationship to reproductive success have recently been studied. A. The Sedge Warbler It has been shown that in the sedge warbler males with larger repertoire size attracted and paired with females earlier (Catchpole, 1980). The study by Buchanan and Catchpole (1997) confirmed this result but was unable to follow through with later measures of breeding success. However, a few of the earlier males became socially polygynous, and they went on to produce more young than did monogamous males (Buchanan and Catchpole, 1999). In a study in Poland, Bell et al. (1997) found that polygynous males had larger repertoires and produced twice as many young as did monogamous males. From these studies there is evidence to suggest that males with larger repertoires not only attract females earlier but also may attract a second female and therefore double their reproductive success. As in many species which are primarily socially monogamous, sedge warbler males may also increase their reproductive success by obtaining extra-pair copulations (EPCs) and producing extra-pair young (EPY). The sedge warbler has been the subject of several DNA fingerprinting studies, which show that EPY are produced in appreciable numbers. This will obviously increase the reproductive success of males, but why should females accept or solicit EPCs? One possibility is that a female may be paired to an inferior-quality social mate and takes the chance to moderate her choice by obtaining EPY sired by a superior quality neighbor. Langefors et al. (1998) studied a Swedish population using multilocus DNA fingerprinting and found that 7.5% were EPY. In 11 of 15 cases, they were able to identify the father, which in every case was a close neighbor. Langefors et al., measured various aspects of male quality, but no obvious physical trends were found. They found that males which sired EPY tended to arrive and pair earlier and were more likely to be polygynous. They concluded that females were probably selecting more ‘‘attractive’’ neighbors for EPCs, but a male trait not measured was song complexity. However, on the days the EPY females settled, their EPY males were already paired and silent.

70

CLIVE K. CATCHPOLE

It seems unlikely from this study that song complexity could be a cue which females use in selecting an extra-pair male. Our parallel study in England also used multilocus DNA fingerprinting and found a remarkably similar level of 7.7% EPY (Buchanan and Catchpole, 1999). In addition to various aspects of male and territory quality, we also measured song characteristics including repertoire size. At this stage we had not determined paternity and simply compared the characteristics of cuckolded males that had lost paternity with those that had not. We paid particular attention to song measures such as song flighting and repertoire size. However, we found no differences in song and other behavioral, physical, or territory measures; thus, there was no evidence to support the view that females with EPY were socially paired to inferior males. Currently, we are determining paternity in our population using microsatellites, but to date we have identified only seven males that obtained extra-pair paternity. The comparison between these males and the males that they cuckolded is a much more rigorous test of the hypothesis that females are selecting superior males for EPCs. As before, we compared as many measures of male and territory quality as we could, but we placed particular emphasis on the three cues known to be important in social female choice; song flighting, repertoire size, and territory size. The results were striking (Marshall et al., 1998); in six of the seven cases the female had mated with an extra-pair male who sang inferior songs in terms of both repertoire size and the amount of song flighting. Even with territory size, the result was the same. Searching for other trends revealed one other behavioral result which also fitted this pattern, namely, that the extra-pair males arrived in the study area later. It was almost as though the males were inferior rather than superior and that in some way were managing to visit paired females and sneak copulations. Our observations on intruding males offer some support for this view. Although our sample size is very small, we will soon have as many as 20 pairs to compare and only then can we be sure that this result is robust. Currently, these three studies provide no direct support to the idea that females are selecting higher quality males for EPCs. B. The Great Reed Warbler As already noted, in the German field study (Catchpole, 1986) repertoire size in this species was an important predictor of pairing success. However, territory quality was also involved, and therefore a detailed multivariate analysis was needed to separate the effects of song and territory quality. Edge length, the length of the reed–water interface, was the most important measure and was significantly correlated with repertoire size. Therefore,

SEXUAL SELECTION AND EVOLUTION IN Acrocephalus WARBLERS

71

we studied the affects of both repertoire size and edge length on various measures of reproductive success. As Fig. 10 shows, polygynous males had the largest repertoire size, followed by monogamous males and unpaired males. Although a significant correlation was found between repertoire size and the number of females a male obtained, an even stronger relationship was found between edge length and the number of females obtained by a male. Partial correlation analysis confirmed that edge length was indeed the more important predictor of pairing success. However, the number of females obtained is only an initial measure of male reproductive success, and when the number of young produced was considered a different pattern emerged. We found no correlation between edge length and number of young, but there was a highly significant correlation between repertoire size and the number of young produced. The overriding importance of repertoire size in predicting eventual reproductive success was confirmed by partial correlation analysis. Although we were not able to DNA fingerprint this population to determine whether our measures reflected the true genetic reproductive success of males, this has been done in a more long-term population study on the great reed warbler by Hasselquist et al. (1996).

Fig. 10. The relationship between repertoire size and the number of females obtained by male great reed warblers. Mean and standard errors are shown for each group and sample sizes are in brackets (from Catchpole, 1986).

72

CLIVE K. CATCHPOLE

Hasselquist et al. (1996) studied their population in Sweden for nearly 10 years and included multilocus DNA fingerprinting. In one of the most thorough field studies ever completed, they were able to estimate not only genetic but also lifetime reproductive success. They were even able to check the viability of the young who survived and subsequently returned to breed. Using sophisticated multivariate analysis, they were able to isolate the main predictors of reproductive success while controlling for other potentially confounding measures of male and territory quality. Fortunately, they also measured aspects of song quality including repertoire size. Their main finding was a significant correlation between repertoire size and the survival of true, genetic offspring produced by each male. They established that repertoire size was indeed the sole predictor of lifetime reproductive success in their great reed warbler population. These results also offer an important clue as to why females select males on repertoire size, and they suggest that a large repertoire indicates that a male carries good genes for offspring viability. A female basing her choice on repertoire size obtains indirect, genetic benefits through her more viable offspring. What about any direct benefits from the male or his territory? Hasselquist et al. (1996) showed that although paternal care and territory quality were correlated to some extent with the annual production of young, they were not correlated with the more crucial measure of offspring survival. Indeed, any effects of direct benefits could also be ruled out by using the other major finding of this study—the males who also sired EPY. The DNA fingerprinting results showed that 3% of all young were EPY, and their 10 genetic fathers were duly traced and found to be neighboring males. Because all attributes of the males and their territories had been measured, they were able to test the hypothesis that females were seeking good genes from superior males. Again, the one clear male trait was repertoire size. As Fig. 11 shows, in each case the true genetic father of the EPY had a larger repertoire size than that of the cuckolded male. This study provides not only good evidence that females are selecting males on repertoire size but also some of the best evidence in support of female choice for good genes. In this case the alternative hypothesis that the male provides direct benefits can be ruled out. The female receives nothing directly from the male because he does not feed her young, and she receives nothing from his territory which she does not occupy. Overall, the evidence that females use repertoire size as a cue in female choice is increasing, and so is evidence that by doing so they obtain good genes. How repertoire size in Acrocephalus warblers became an honest indicator of male quality is not obvious, but I attempt to provide some clues in the final sections of this review.

SEXUAL SELECTION AND EVOLUTION IN Acrocephalus WARBLERS

73

Fig. 11. Repertoire sizes of social and genetic fathers of 10 extra-pair chicks in the great reed warbler (from data in Hasselquist et al., 1996).

VI. Song and Male Quality So far I have reviewed the evidence for song as an important cue in female choice and the evidence that females that select males on the basis of song enhance their fitness. Now I consider the most difficult problem—one which continues to generate debate in the general area of sexual selection theory (Andersson, 1994). How does song indicate to the prospecting female that a particular male is of superior quality to his neighbors? Such questions have been discussed by other reviewers of the relationship between song and sexual selection (Catchpole, 1982; Searcy and Andersson, 1986; Catchpole, 1987; Searcy and Yasukawa, 1996), and I attempt to show how recent studies on Acrocephalus warblers have provided a few more clues and also opened up some new avenues of inquiry. A. Song and Parasites Hamilton and Zuk (1982) presented a hypothesis which has had a major impact on sexual selection theory. They suggested that male signals, such as plumage and songs, indicate genetic resistance to parasites so that females selecting brighter plumage or more complex songs will acquire good genes for heritable resistance for their offspring. In support of their hypothesis, Hamilton and Zuk produced a comparative survey which suggested that species with more complex songs were more likely to suffer infection from blood parasites. However, Read and Weary (1990) found that this correla-

74

CLIVE K. CATCHPOLE

tion was confounded by phylogeny, and when this was controlled for the correlation disappeared. A more crucial test of the Hamilton–Zuk hypothesis would be to show that, within a single species, the expression of the male trait (singing) is dependent on parasite load. So far, the study of the relationships between song and parasites has lagged far behind that of plumage and parasites in birds (Hillgarth and Wingfield, 1997). The only intraspecific study on song and parasites in birds has shown that song rate in the swallow Hirundo rustica decreases in the presence of ectoparasites (Møller, 1991a; Saino et al., 1997). However, there is no evidence that song rate in the swallow is a sexually selected male trait. Because we have established that song complexity in Acrocephalus warblers is a sexually selected male trait, we decided to investigate its relationship to parasite infection in our sedge warbler population (Buchanan et al., 1999). First, we searched for the ectoparasites that were successfully studied in the swallow (Møller, 1991a; Saino et al., 1997) but found almost none on our sedge warblers. We then took blood smears and examined them under the microscope for hematozoan parasites. Three separate genera of parasitic blood protozoan were found to be quite common, namely, Plasmodium, Trypanosoma, and Haemoproteus. These were counted using standard procedures on 100 fields per blood sample. During the 2 years of the study, we found high overall prevalence values; 19.5% of adults were infected in 1995 and 37.5% in 1996. Although counts of red blood cells remained constant, counts of white blood cells were significantly elevated in the parasitized birds. Parasitized males are clearly paying the costs of mounting an immune response, but how does this affect their singing behavior? Overall, and in both years, we found that parasitized males had significantly smaller repertoire size than did uninfected males (Fig. 12), but in

Fig. 12. Repertoire sizes of healthy and parasitized males in the sedge warbler. Means and standard errors are shown, and data are combined for both years of the study (from data in Buchanan et al., 1999).

SEXUAL SELECTION AND EVOLUTION IN Acrocephalus WARBLERS

75

only 1 year did they also song flight less. Comparison of other physical and behavioral aspects of male quality revealed no other significant differences between the two groups. It seems that repertoire size is the only reliable indicator of an important aspect of male quality, in this case the ability to resist infection by blood parasites. Females that select males using repertoire size as a cue could therefore avoid parasitized males and perhaps obtain good genes for parasite resistance for their offspring as originally suggested by Hamilton and Zuk (1982). They may also obtain more obvious direct benefits from being mated to a healthy male partner. For example, feeding the young is a demanding task requiring effort from both parents, and an infected partner might not be able to perform as well. Repertoire size may also signal the quality of male parental care which a female might expect. We tested this possibility by examining male provisioning rates, standardized for both brood size and time of the breeding cycle. When combining all our data we found that parasitized males did provision their young at significantly lower rate than did uninfected males. B. Song and Brain Structure The development of standard techniques to locate and measure the volume of individual brain nuclei in the avian song system has provided new opportunities to link studies in neurobiology to behavior and evolution. The brain pathway controlling song production and learning in birds is one of the best studied and most influential models in vertebrate neurobiology (Nottebohm, 1993; DeVoogd, 1994; Catchpole and Slater, 1995; Brenowitz and Kroodsma, 1996). We now know that a discrete pathway of forebrain nuclei controls both song production and learning. Although many nuclei are involved, the most important for song production is the higher vocal center (HVC), which projects to the robustus archistriatalis (RA), whereas two other nuclei (1-MAN and area X) are essential for song learning. One of the most striking features of the system is its sexual dimorphism, which is directly related to singing behavior. In most species, song is produced exclusively or primarily by the male, and this is reflected in the size of the song pathway, which is much larger in the male. For example, Nottebohm and Arnold (1976) found that HVC and RA in male canaries and zebra finches were four or five times larger than those in females. This is the first clue that sexual selection may be involved not only in the evolution of song complexity but also in the evolution of the size and structure of the underlying neural pathway (Catchpole, 1996; Jacobs, 1996). One way to test the song, sexual selection, and brain size hypothesis is by using the comparative approach. If sexual selection has driven the evolution of song complexity and the associated areas of the brain, then we

76

CLIVE K. CATCHPOLE

might expect to find that species with larger repertoires also have larger song nuclei. DeVoogd et al. (1993) tested this prediction on 41 species of songbird using the modern method of independent contrasts to control for the confounding effects of phylogeny, and they also controlled for variations in brain size. They used multivariate analysis and confirmed that there was a highly significant correlation between relative HVC volume and repertoire size. They used area X as a control and found no correlation between its volume and repertoire size. Neither was HVC volume correlated with the volume of other brain regions such as the hippocampus. This suggests that, during evolution, there was a repeated association between the development of song complexity and the relative volume of HVC. Although this was an important finding, there are problems with such a broad comparative study, particularly with obtaining standard measurements of song complexity. DeVoogd et al. (1993) took their measurements indirectly from the literature, and they used estimates of song-type repertoires from a wide range of different species and different studies. A better test would be to use a more closely related taxonomic group with a known phylogeny and use standard song complexity measures. Such a group is the Acrocephalus warblers, for which there is a robust molecular phylogeny (Leisler et al., 1997). There is also a standard measure of song complexity and syllable repertoire size, and, most important, good evidence that the evolution of repertoire size has been driven by sexual selection pressure. Taking the study of DeVoogd et al. (1993) into consideration, we set out to test the prediction that HVC has become associated with the development of song complexity in Acrocephalus warblers (Sze´kely et al., 1996). We also included another genus of warblers (Locustella) because these are closely related and similar in every other way except one important difference—their songs. Acrocephalus warblers are famed for their song complexity, whereas Locustella warblers are known for the extreme simplicity and monotony of their songs. For example, the grasshopper warbler L. naevia is so named because its song sounds like the stridulation of a grasshopper and it has only one or two syllables in its repertoire. For an unknown reason, sexual selection on song complexity has increased in the genus Acrocephalus but has remained a simple but highly convenient control in the genus Locustella. It also gives our data set more variation (repertoire sizes from 1 to 100) to add to its existing validity. As shown in Fig. 13, there was a significant correlation between repertoire size and the volume of HVC (Sze´kely et al., 1996). We used the same methods as those used by DeVoogd et al. (1993), but we measured all four of the main song control nuclei (HVC, RA, area X, and 1-MAN). Together with our smaller taxonomic unit, molecular phylogeny, and standard measure of repertoire size, this provides a robust confirmation of the earlier

SEXUAL SELECTION AND EVOLUTION IN Acrocephalus WARBLERS

77

Fig. 13. The relationship between repertoire size and the residual volume of HVC in Acrocephalus warblers. The method of independent contrasts was used, and the number of nodes used in the analysis are shown numbered (from Sze´kely et al., 1996).

findings. It seems that when song complexity evolves in a group, it is associated with an increase in the relevant brain space in the song control pathway. This may have important implications for our understanding of sexual selection by female choice. Selection may be acting not only on observable behavior but also on the underlying structure of the brain. What about the female brain? The only species in which females have a well-developed song pathway are duetting species in which both sexes sing (Brenowitz and Kroodsma, 1996). However, even in species in which females do not sing, such as the zebra finch, the female retains a smaller version of the song pathway (DeVoogd, 1994). The most likely explanation is that the song pathway in both sexes is also used for reception and discrimination of song, and there is evidence to support this theory (Catchpole and Slater, 1995). Following our study on males (Sze´kely et al., 1996), we can hypothesize that if female choice is driving the evolution of male repertoire size, then females of species with larger repertoires should have larger brain nuclei to enable them to discriminate between more complex songs. This is essentially the same prediction as that which we tested with males, but it has not been tested using female brains. We also measured song nuclei in our more

78

CLIVE K. CATCHPOLE

limited collection of female Acrocephalus and Locustella brains, and an interesting relationship was revealed. Females in species with larger repertoires have one particular part of the song pathway, nucleus 1-MAN, which is also larger in volume (DeVoogd et al., 1996). Until we confirm this finding using a larger sample and the same rigorous phylogenetic analysis (Sze´kely et al., 1996), we remain cautious. However, this is potentially an exciting new finding because it suggests that the song control pathway contains a perceptual neural mechanism which may be important in female choice. Although comparative studies provide valuable clues, intraspecific tests are needed to substantiate our hypotheses concerning the neural basis of song complexity and female choice. Do individual males with larger repertoires also have larger song nuclei, and does the same apply to females who are more selective? So far, there have been few such studies and the results are mixed. In female brown-headed cowbirds, Molothrus ater, Hamilton et al. (1997) found that females which were more selective in response to male songs had a larger volume and more neurons in 1-MAN. In males, Nottebohm et al. (1981) reported a correlation between repertoire size in canaries and the volume of both HVC and RA. However, this was not confirmed in a study on red-winged blackbirds by Kirn et al. (1989). Bernard et al. (1996) studied the song control pathway in the European starling, and they found no correlations with repertoire size. However, they also measured other aspects of song structure, such as song length. Like sedge warblers, starlings sing songs in long sections, and Bernard et al. (1996) found positive correlations between the length of song bouts and the volumes of HVC and RA. We recently completed a study of sedge warbler brains and found a similar pattern (Airey et al., 2000). We investigated three measures of song complexity: song length, the number of syllable types per song, and repertoire size. All three showed positive correlations with the size of HVC, but the strongest relationship was the correlation between repertoire size and the size of HVC. C. Other Aspects of Male Quality So far I have concentrated on the relationships between song as an indicator of two aspects of male quality: resistance to blood parasites and the size of the underlying neural pathway controlling song production and learning. This is because we are currently investigating both of these aspects and have started to derive some interesting results. There is also the intriguing possibility of trade-offs between immunocompetence and brain space during development, and I discuss this later. However, there are many other aspects of male quality which song could potentially indicate

SEXUAL SELECTION AND EVOLUTION IN Acrocephalus WARBLERS

79

to females, and several of these have recently come to light as a result of our long-term field studies. 1. Song and Territory Quality One of the most likely male attributes to affect female fitness is the quality of his territory because territory provides both the nest site and much of the food needed to provision the young. It may also provide the food needed for the male to be able to produce his singing display. Does song indicate the quality of a territory rather than the quality of the male. Several studies [reviewed by Catchpole and Slater (1995) and Searcy and Yasukawa (1996)] have demonstrated that increasing food supply can boost song output and that females prefer males with higher song rates. Our studies on the sedge warbler (Buchanan and Catchpole, 1997) showed that in each of 3 years there was a positive correlation between territory size and female choice as indicated by pairing date. This suggests that females may well be choosing their males indirectly by the size of their territories. However, the study found that repertoire size and song flighting were also correlated with pairing date, so females are selecting males on multiple cues. A key point is that there were no correlations between song and territory measures, so song does not appear to indicate territory quality to females. In our great reed warbler study (Catchpole, 1996) repertoire size but not territory size was correlated with the number of females attracted. Another measure of territory quality (length of the reed–water interface) was correlated with the number of females attracted and also with repertoire size. However, partial correlation analysis revealed that repertoire size was the most important indicator of reproductive success. This was confirmed by Hasselquist et al. (1996) in their long-term study of another great reed warbler population in Sweden. So far, the evidence suggests that repertoire size is unlikely to be influenced by territory quality, but that singing rate may well fluctuate in relation to food supply. Variations in singing rate could thus reflect either the quality of a territory or the ability of a male to find sufficient food. The amount of song produced is likely to be condition dependent and is best viewed as a trade-off between singing and feeding behavior, thus providing valuable additional information to prospecting females. In the case of the sedge warbler (Buchanan and Catchpole, 1997) the amount of song flighting and repertoire size are not correlated and are used as multiple cues in female choice ( Johnstone, 1996, 1997). 2. Song and Parental Effort Besides territory, another way in which a male can contribute toward female fitness is in feeding her young. In many passerines, successful fledging

80

CLIVE K. CATCHPOLE

depends crucially on biparental care, and this is particularly the case in socially monogamous Acrocephalus species (Catchpole, 1995). Duckworth (1992) removed male reed warblers from feeding pairs and found that females had to increase their feeding rate to prevent starvation of young in the nest. Clearly, females would benefit from selecting males on any trait which could signal that a male was capable of providing high-quality parental care. Studies linking male traits to parental care are few, but Hill (1991) showed that in the house finch Carpodacus mexicanus females selected the most colorful males who also fed their young at a higher rate. Only one study has revealed a similar link with male song. Greig-Smith (1982) studied the stonechat Saxicola torquata and found that males with higher song rates helped more with feeding young and defending the nest. Having established that female sedge warblers choose males on the basis of song and territory cues (Buchanan and Catchpole, 1997), we decided to test the hypothesis that one or more of these cues might signal the quality of future male feeding effort. To obtain a measure of male feeding effort, we counted the number of male feeding visits to the nest. Because feeding effort increases with age of the chicks and brood size, we standardized the day of observation to Day 7 after hatching and the number of visits to number per chick. We obtained a strong positive correlation between male feeding effort and repertoire size but not with amount of song flighting or territory size (Buchanan, 1997). It seems that repertoire size may also provide females with information on the future quality of male parental effort. 3. Song, Sperm, and Fertility We have become used to the idea that females may be selecting males on traits which somehow reflect their superior quality. One aspect of male quality is fertility, and it has been suggested that females who seek extrapair copulations may do so to obtain direct fertility benefits in case their own males become sperm depleted. This is known as the fertility-insurance hypothesis (Birkhead and Moller, 1992). Sheldon (1994) extended this idea in his phenotype-linked fertility hypothesis, which posits that male fertility covaries with male phenotype. In other words, a male trait such as song advertises male fertility so that a male with a more complex song would produce more or better quality sperm and be more likely to fertilize a female. Although often used to explain why females mate with several males, it could equally explain why they prefer to pair with one particular male. In the sedge warbler both types of mating occur (Buchanan and Catchpole, 1999), and because song is used in mate choice we decided to investigate the relationship between song and sperm quality.

SEXUAL SELECTION AND EVOLUTION IN Acrocephalus WARBLERS

81

To do this, we recorded a population breeding in Hungary, analyzed their songs, and then analyzed their sperm for both quantity and quality (Birkhead et al., 1997). Sperm were counted and measured for length, and the proportion of abnormal sperm was also estimated. We found that the more successful males which had paired at the time of sampling had larger repertoires than those which were single and still trying to attract a female. However, when we compared their sperm, we found no differences in quality or quantity between the two groups. Nor did we find any correlation between repertoire size (or other song measures) and our measures of sperm quantity and quality. There was no evidence from our study to support hypotheses suggesting that song can indicate male fertility. Another hypothesis relating song to fertility was proposed by Møller (1991b). He noticed that in some species a peak of male song activity coincides with the ‘‘fertile window’’ of the female and suggested that a prime function of male song is in mate guarding fertile females against EPCs. However, in many species, including Acrocephalus warblers, male song stops, declines, or simplifies after pairing, and in general there is little support for Møller’s hypothesis (Catchpole and Slater, 1995). Indeed, currently there is very little evidence to support the idea of direct links between song and either male or female fertility. 4. Song, Age, and Viability If females are selecting males for indirect, genetic benefits the male trait should somehow reflect the ability to survive and reproduce in future years. By selecting such males the female will obtain good genes for viability for her offspring (Andersson, 1994). Because female Acrocephalus warblers are selecting males by song, it may be that song is an indicator of male viability. Testing this particular prediction is clearly difficult, but considerable progress has been made recently. Our Hungarian field study of the sedge warbler (Birkhead et al., 1997) contained a good sample of males of known age, ranging from 1 to 5 years. We found a significant correlation between male age and repertoire size (Fig. 14). This correlation could occur in two different ways, either by males which survive longer adding to their repertoire each year or by differential mortality of males related to repertoire size. Either way, females selecting males on repertoire size could still be obtaining males with good genes for viability. In our latest English study (Buchanan, 1997) we were able to study changes in repertoire size for returning males. We found a significant increase in repertoire size between years. In a smaller sample of males who returned for 3 years, we found slight, incremental increases each year. We were also able to follow the other two traits known to be important in female choice (Buchanan and Catchpole, 1997), the amount of song flighting

82

CLIVE K. CATCHPOLE

Fig. 14. The relationship between repertoire size and age in a population of male sedge warblers (from data in Birkhead et al., 1997).

and territory size, but there were no significant trends in either. Several other studies have provided evidence for female choice based on age-related changes in repertoire size, for example, red-winged blackbirds (Searcy and Yasukawa, 1995) and European starlings (Eens, 1997). In the great reed warbler, I initially found no correlation between repertoire size and age in our German population (Catchpole, 1986). However, in a more detailed long-term study in Sweden, Hasselquist et al. (1996) did find a positive correlation. They then tested a most important prediction. If repertoire size is indeed an indicator of good genes for viability, then the offspring of females who choose males with large repertoires should be more likely to survive and return to breed. They used multivariate statistics on their 10-year data set to reveal a significant correlation between male repertoire size and postfledging survival of their offspring. Further support for the hypothesis that female great reed warblers are selecting males by repertoire size for good genes is derived from their DNA fingerprinting analysis (Hasselquist et al., 1996). They found 10 cases of extrapair paternity, and they compared the characteristics of the social and genetic fathers. In each case, the genetic father of the extra-pair chick had a larger repertoire than the male whom he cuckolded. The case of extrapair young is particularly informative because female obtains no direct benefits, such as feeding or a territory, from the extra-pair father. His only

SEXUAL SELECTION AND EVOLUTION IN Acrocephalus WARBLERS

83

contribution is an indirect, genetic one—his sperm and the genes they contain. This important study offers considerable support to the hypothesis that females select males on the basis of song because repertoire size is an honest indicator of good genes for viability.

VII. Conclusions and Future Research This study, like many before it, began with a simple observation which triggered an obvious question. In this case, why does a male bird sing such a complicated song? The initial answer came quite quickly—to attract a female. However, this was only the beginning, when simple hypotheses were formed and tested by observations and early experiments. The results led to more refined hypotheses and their predictions became more difficult to test, needing long-term field studies linked to laboratory experiments. Linking male song with female attraction landed me in the then unfashionable area of sexual selection. The hypothesis that male song functions mainly in female attraction, rather than in territorial defense, was, like Darwin’s theory of female choice, widely disbelieved and waiting to be tested. What was then the most extensive work on bird song, Armstrong (1973) unhesitatingly used the term ‘‘territorial song,’’ alluding to female attraction only briefly. Even though Trivers (1972) had predicted that female choice should be widespread, it was not until a decade later that Andersson’s (1982) dramatic experiment confirmed it in the field. A. Female Choice and Song Searcy and Yasukawa (1996) reviewed the evidence for female choice based on male song traits. As they point out, some evidence is indirect and weak. For example, my early studies on the cessation of song after pairing in sedge warblers and the seasonal correlation between singing and later breeding activity (Catchpole, 1973) provide indirect evidence that song functions in female attraction. Such studies are useful pointers toward function but need to be tested. 1. The Experimental Approach One way of testing this is by removing the female and observing whether song production resumes. In all species in which this has been done (Searcy and Yasukawa, 1996) males resumed or increased singing, but the evidence is still indirect and open to alternative explanations. For example, the presence of females and sexual behavior might simply interfere with the normal pattern of male song production, and the song could be directed

84

CLIVE K. CATCHPOLE

at other males. Møller (1991b) used a similar argument to explain the correlation between increased song and the female fertile period in some species. In this case, he claims that song repels rival males seeking EPCs, but it could also be claimed that it is directed at keeping the female close to her social mate. Arguably, a more convincing test would be to present a singing male with a female and study the effects on his song. Searcy and Yasukawa (1990) presented taxidermic mounts of females to red-winged blackbirds in their territories and found that the males greatly increased the rate of switching between song types. Eens et al. (1993) found that male starlings quadrupled their song output when presented with a live female. Although these results are consistent with the hypothesis that male song functions to attract females, they still provide weak evidence because they are all indirect tests. Another approach has been to search for correlations between song and some measures of reproductive success. Howard’s (1974) pioneering study on mockingbirds revealed that males with larger repertoires attracted females first. This was followed by my study on sedge warblers (Catchpole, 1980), and there have since been several more studies demonstrating correlations between song quality or quantity and some measure of reproductive success (reviewed by Searcy and Yasukawa, 1996). Although these constitute a stronger body of evidence, there remains the possibility of confounding variables from some aspects of either male or territory quality. What was needed was a stronger experimental test in which confounding variables could be either eliminated or controlled. The answer is the playback experiment, in which only tape-recorded song is used, thus eliminating territory and other aspects of male quality. However, female songbirds do not respond as strongly as males to playback, and therefore special techniques had to be developed to measure their preferences (Searcy, 1992a). The real breakthrough came when Searcy and Marler (1981) implanted captive females with estradiol, which made them more receptive. When songs were played back, the females responded with copulation solicitation, a sexual display given prior to copulation with a male. This can be quantified and used as a sensitive behavioral assay of female preference for a particular male song. Searcy (1984) then used this to demonstrate that females preferred larger repertoires in the song sparrow, and we followed this study by using this technique to confirm our field results in the sedge warbler (Catchpole et al., 1984). There have since been many more studies utilizing this technique and the findings show the same overall trend. It seems that females really do prefer larger repertoires and more complex songs (Searcy, 1992b; Searcy and Yasukawa, 1996). A criticism leveled at these experiments is that they often used too few exem-

SEXUAL SELECTION AND EVOLUTION IN Acrocephalus WARBLERS

85

plars (Kroodsma, 1986, 1990; McGregor, 1992), but taken together they provide a much stronger and more direct test. An even more direct test using playback is phonotaxis, in which females are attracted to approach the speaker playing recorded song (Searcy, 1992a). The first experiments were performed by Eriksson and Wallin (1986), who studied pied and collared flycatchers (Ficedula hypoleuca and F. albicollis) breeding in nest boxes. Half the nest boxes were equipped with speakers playing male song, and these attracted females who ignored the nonsinging control boxes. A similar result was obtained by Mountjoy and Lemon (1991) with starlings and by Johnson and Searcy (1996), who studied the house wren Troglodytes aedon. Lampe and Saetre (1995) also studied pied flycatchers but gave females the choice of boxes playing large or small repertoires. The females all preferred the boxes playing the larger repertoire size. This overall pattern of results provides the most direct and strongest test of the hypothesis that females are attracted by male song, and that they prefer larger repertoires. 2. The Integrated Approach Although some tests are stronger than others, perhaps the most convincing are the cases in which a whole series of tests have been carried out on individual species, what I previously called an integrated approach (Catchpole, 1992). Searcy and Yasukawa (1996) agreed and chose four species in which such an extensive program had been carried out. One was the starling, studied in both Europe and the United States and recently the subject of an extensive review by Eens (1997). In the starling, song is extremely complex and variable, and in structure, function, and breadth of study it parallels the Acrocephalus song. Song production decreases after pairing, and removal of the female leads to resumption. Males also increase song output and complexity when presented with a female. Females in breeding colonies prefer males with larger repertoires, and these attract more females and produce more young. As already discussed, females are attracted to nest boxes playing recordings of song. Another species highlighted by Searcy and Yasukawa (1996) was the red-winged blackbird, which shows a similar pattern in many field and experimental studies. The other species were the two Acrocephalus warblers I have concentrated on in this review—the sedge warbler and the great reed warbler. In the case of the sedge warbler, it has been shown how the dramatic cessation of song upon pairing provided the first evidence implicating mate attraction and how this led to the correlation between repertoire size and pairing date and recently the number of females attracted. Tapes were taken from the same males and used to confirm our findings in a laboratory experiment, and then songs were synthesized on one male to demonstrate

86

CLIVE K. CATCHPOLE

how increasing repertoire size correlated with an increased female sexual response. In the great reed warbler we found a similar pattern from a combination of field and laboratory studies, including a demonstration of active female choice in the field. The great reed warbler study also found clear evidence for a separation of song structure and function; short, simple songs for male–male interactions and longer, more complex songs for female attraction. This separation has proceeded even further in the aquatic warbler, in which there seems to be a special short song related to mate guarding and sperm competition. This led me to hypothesize that short, simpler songs evolved by intrasexual selection and longer, more complex ones by intersexual selection through female choice (Catchpole, 1980, 1982). These ideas are related to the ‘‘dual-function hypothesis’’ frequently applied to bird song in general and discussed at length in our major review (Catchpole and Slater, 1995). In its simplest form this suggests that the main functions of male song are to repel rival males and to attract females, accepting that in some species there is more emphasis on one than the other and that in most cases the same song may serve both functions. Acrocephalus warblers have provided a rare opportunity to observe a range of species, including some in which special song structures have developed for male repulsion and female attraction. Clearly, we can only speculate about the origins, pathways, and selective forces which have led to the evolution of the range of song structures and complexity we now observe. After all, what we are now detecting is current utility and function and this may not be the same as what has been important during evolution. Kroodsma and Byers (1991) urged caution in accepting the dual-function hypothesis, but Catchpole and Slater (1995) found the evidence convincing. Regarding the role of song in female choice, Searcy and Yasukawa (1996) concluded, ‘‘we feel confident in taking the existence of female preferences as given.’’ I hope that this review has also demonstrated that the evidence for song and female choice is compelling. B. Song and the Benefits of Female Choice Having demonstrated that females in some species use song as a cue in female choice, we must finally turn our attention to why they do so. In particular, what possible benefits accrue to females who select males with more complex songs and larger repertoires? In doing so, we will follow recent convention by considering such benefits as either direct and phenotypic or indirect and genetic (Kirkpatrick and Ryan, 1991; Searcy and Yasukawa, 1996). 1. Direct Benefits Such preferences as we have discussed could clearly exert direct effects on female fitness. For example, the female must maximize her chances of

SEXUAL SELECTION AND EVOLUTION IN Acrocephalus WARBLERS

87

mating with the right species, but she should also minimize the costs of searching for the right mate. However, in birds with biparental care, female fitness is most likely to be directly affected by the resources provided by the male’s territory and the parental care provided by the male. How can a male reliably signal such information in his song? As Zahavi (1975, 1977) pointed out, a song can be an honest indicator of male quality as long as it entails a cost to the signaler. Poor-quality males cannot bluff because only better quality males can afford to pay the costs of singing. The most obvious cost of song is the amount of singing a male can afford to do, particularly in cases such as the sedge warbler in which song flighting also occurs. Whatever the energetic costs of song, the male also has to allocate valuable time to this activity as opposed to time which could be spent feeding. Many studies [reviewed by Catchpole and Slater (1995) and Searcy and Yasukawa (1996)] demonstrated that song production is dependent on food supply, that it can be increased by adding extra food, and that song rate is an important cue in female choice. The amount of male song could well give an indication of food supply within a territory. In our studies on the sedge warbler we did not measure food supply but did find that both the amount of song flighting and territory size were cues in female choice. Only two studies have examined the relationship between song production and male parental quality. Greig-Smith (1982) found that male stonechats with higher song rates participated more in feeding their young and helped more in nest defense. However, Hoi-Leitner et al. (1993) found the opposite trend in a study on blackcaps Sylvia atricapilla. Although it seems intuitive that the amount of song might signal phenotypic quality, it seems less obvious in the case of repertoire size. There is also less evidence linking repertoire size to any direct benefits the female might obtain. The only case mentioned in the review by Searcy and Yasukawa (1996) was our study on the great reed warbler (Catchpole, 1986) in which repertoire size was correlated with one aspect of territory quality— the length of the reed–water interface. This is where much of the insect food is found, and so there may be a link with food supply. In the case of the sedge warbler there does seem to be a link between repertoire size and male parental effort because we also found a correlation between repertoire size and male feeding effort (Buchanan, 1997). There was no correlation between either the amount of song flighting or territory size and male feeding effort. We have already discussed Hamilton and Zuk’s (1982) theories on the relationship between parasites and song. Although usually considered in the context of indirect benefits for good genes, there remains the possibility that females gain direct benefits by avoiding parasitized males. In the swallow, parasitized males show a decrease in song rate (Møller, 1991a; Saino et al., 1997), although there is no evidence to link song rate with female

88

CLIVE K. CATCHPOLE

choice in swallows. In our study on the sedge warbler (Buchanan et al., 1999) we found that parasitized males had smaller repertoires than did uninfected males. Because repertoire size is an important cue for female choice in the sedge warbler, it follows that males could avoid pairing with parasitized males. We have also been able to show that parasitized males feed their young at a lower rate. It follows that females who select males with larger repertoires obtain direct benefits from increased male effort in feeding the young. Why repertoire size and not song flighting appears to reflect male quality is still not obvious, but I discuss possible reasons later. 2. Indirect Benefits The possibility that females obtain indirect, genetic benefits from female choice has been an important and central argument in modern sexual selection theory (Andersson, 1994), although the validity of such models has also been questioned (Kirkpatrick and Ryan, 1991). Studies on bird song are important in this respect because they can provide much needed evidence for good genes models of female choice (Catchpole and Slater, 1995; Searcy and Yasukawa, 1996). Indeed, arguably the most convincing studies have been those on song and female choice in Acrocephalus warblers (Catchpole, 1987, 1996). Indirect benefits occur when the fitness benefits accrue not to the female but to her offspring. In Fisherian runaway models, a male trait such as song is not related to fitness benefits for the female, other than the ability her sons will inherit to attract more females. There is simply a genetic correlation between the male trait and the female preference for it. In studies of bird song, obtaining evidence for such a correlation has proved to be an intractable problem, and there is no body of evidence to support the Fisher effect. As Searcy and Yasukawa (1996) pointed out, several difficult steps would be needed. In addition to evidence for the genetic correlation, we would also need to demonstrate that song complexity is heritable, and no study has attempted to do so. The only study on song is that of Houtman (1992), who claimed that in captive zebra finches song rate (not complexity) was heritable. Currently, there is no direct evidence to support the Fisher effect, but neither can we rule it out as a contributory factor in the evolution of song complexity. This leaves us with the possibility that females select males with more complex songs not because they are simply more attractive but because they somehow indicate good genes for viability. To test this we would need to show that females who select males with larger repertoires really do produce offspring of higher viability. This has been extremely difficult to test because it requires long-term field studies linking song quality to reproductive success and survival. Several studies (reviewed by Searcy and Yasukawa, 1996) have shown correlations between repertoire size and

SEXUAL SELECTION AND EVOLUTION IN Acrocephalus WARBLERS

89

measures of male lifetime reproductive success or age. Although important, these studies do not provide strong evidence that the fitness components are heritable and that offspring survival is related to repertoire size. However, one study of an Acrocephalus warbler does provide such evidence. Hasselquist et al. (1996) studied a population of the great reed warbler in Sweden for many years and were able to demonstrate a significant positive correlation between male repertoire size and postfledging survival. Young from females who selected males with large repertoires were more likely to survive and return to breed. Because Hasselquist and colleagues DNA fingerprinted their birds, they were able to determine the paternity of any extra-pair young. In each case, they found that the extra-pair chick was fathered by a male who had a larger repertoire than the cuckolded male. This also enabled them to exclude the possibility of any direct benefits to females because extra-pair males do not feed or defend their extra-pair young. The only benefits the females could obtain were indirect genetic ones—in this case good genes for the viability of their offspring. 3. Repertoire Size as an Honest Indicator The detailed study by Hasselquist et al. (1996) provides the most powerful and convincing test of the hypothesis that repertoire size is an honest indicator of good genes for viability. How does this come about and, in particular, how is it maintained? Why has selection favored the complexity index of repertoire size and not the intuitively obvious measures of amount of song or song flighting? Part of the answer may be in the way song is learned, stored, and then produced (Catchpole and Slater, 1995), as well as in how the female is able to perceive and sample male song. As discussed previously, the amount of singing or singing rate varies with season and time of day. It also varies with temperature and other environmental conditions as well as the health and condition of the singing male. Therefore, the amount of song or song rate is probably a sensitive and honest indicator of immediate, phenotypic condition. When we recorded the amount of singing and song flighting in sedge warblers (Buchanan and Catchpole, 1997), we took great care to standardize the stage of the breeding cycle, time of day, and weather as far as possible. When we plotted daily samples they varied considerably from day to day, building up to a peak before cessation at pairing. In contrast, for repertoire size in the sedge warbler, sonagraphic analysis gave the same measure, irrespective of the day or the time it was recorded. Repertoire size remained as a constant, repeatable measure throughout the breeding season. It is therefore more likely to be used by females, and favored by selection, as a reliable, honest indicator of long-term male quality.

90

CLIVE K. CATCHPOLE

How is repertoire size maintained as an honest indicator of male quality, whether phenotypic or genetic? In other words, what are the costs of repertoire size? Our studies on repertoire size and brain space in Acrocephalus warblers led me to suggest that there are neural costs involved (Catchpole, 1996). We have established that, in Acrocephalus warblers, males of species with large repertoires have a larger volume in HVC (Sze´kely et al., 1996), and our current intraspecific study on the sedge warbler is revealing a similar story. Complex songs and large repertoires may well need more brain space for learning, storage, and production, an idea suggested in several studies (Nottebohm et al., 1981; Canady et al., 1984; Kroodsma and Canady, 1985). Nowicki et al. (1998) linked the relationship between sexual selection and repertoire size more closely to song learning. They point out that the nature and timing of song learning provides an opportunity for repertoire size to be influenced by nutritional stress during development. This may affect the development of the areas of the song control pathway, such as HVC, which in turn control the expression of repertoire size. Thus, the learning process may reflect the early developmental history of the young male, and repertoire size later provides an indicator of some aspects of male quality. However, nutritional stress is unlikely to be the only factor influencing the process of song development. We have shown that males with hematozoan parasites also have smaller repertoire sizes (Buchanan et al., 1999), allowing females to select males with good genes for resistance, as suggested by Hamilton and Zuk (1982). We have also found that male sedge warblers can become infected as nestlings. In this first year, song develops as a result of a complex neuroendocrine pathway (DeVoogd, 1994), and this would provide the opportunity for repertoire size to develop as an immunocompetence handicap (Folstad and Karter, 1992). Only males of superior quality would be able to afford the extra costs of mounting an immune response and still develop a large repertoire. The precise nature of the relationships between the immune response, neural space, song development, and repertoire size seems a promising line for future research. 4. Final Thoughts Our current studies on the neurobiology of females suggest that an area for reception and discrimination of male song may be located in their smaller song pathway. This indicates the possibility of correlations in the song control pathway between the neural bases of the male trait and the female response. With our new phylogeny of Acrocephalus species, we may soon be able to test the sensory exploitation hypothesis (Ryan et al., 1990). To do this we will have to demonstrate that, in our phylogeny, enlargement of the female sensory area predated the evolution of the increase in male

SEXUAL SELECTION AND EVOLUTION IN Acrocephalus WARBLERS

91

HVC. Searcy (1992b) found that although common grackles have only one song type, females show a clear preference for playback of a repertoire of four song types. Future research will also need to address the question of heritability in traits such as repertoire size while controlling for or taking account of the complications of song learning. Comparing repertoires of fathers and sons, both in the wild and in controlled breeding experiments, would be a start. Linking these to parallel studies on the song control pathway in the brain is also a promising area for future research. Recent studies on North American marsh wrens Cistothorus palustris (Canady et al., 1984; Kroodsma and Canady, 1985; Brenowitz et al., 1995) are particularly interesting because they demonstrated that eastern populations have smaller repertoire sizes than western ones, and that this is reflected in the size of HVC and RA. Experiments on song learning in captivity showed that, no matter how many songs were played to them, the eastern males could develop only a smaller repertoire size. This suggests a tight genetic constraint on repertoire size and the brain space needed to store it, an idea also suggested by the work of Nottebohm et al. (1981) on canaries. I started this review by showing how a simple observation led me to hypothesize that the complex song of the sedge warbler functioned in female attraction. Although controversial at the time, it led me to design a series of field and laboratory studies to test the resulting predictions. The results were certainly interesting and have been used to formulate more general views on the functions of bird song (Catchpole and Slater, 1995) and also in the wider context of sexual selection and female choice (Andersson, 1994). The early studies led to testing more difficult and detailed predictions, which require more complex and sophisticated techniques. Further research integrating behavioral and evolutionary studies with mechanisms and neurobiology is needed before we can fully understand why birds such as Acrocephalus warblers sing their complicated and beautiful songs.

VIII. Summary Acrocephalus warblers and their elaborate songs have proved to be rich and fertile ground on which to test functional and evolutionary hypotheses. In particular, they have contributed a great deal to our understanding of the functions of song and to the importance of song as a cue in female choice. In socially monogamous species, such as the sedge warbler, the female needs a good quality male to help raise the young successfully. This has placed the emphasis on selecting the male directly, and the long, complex songs function as a cue in female attraction. In polygamous species,

92

CLIVE K. CATCHPOLE

such as the great reed warbler, resources are also important because the male may not help with feeding the young. Although females still use more complex male songs as a cue in mate choice, males have also developed short, simple songs for territorial defense. In the aquatic warbler, in which there is intense sperm competition, a special type of short song is used to repel rival males. Intrasexual selection has led to the evolution of short, simple songs used in aggressive interactions with rival males, and intersexual selection by female choice has led to the longer, more complex songs used to attract females. A considerable body of field and experimental evidence supports the hypothesis that female choice is the driving force behind the evolution of song complexity in Acrocephalus warblers. Repertoire size emerges as the most important cue from most studies and correlates with measures of reproductive success, parental care, age, and even offspring viability. In general, these findings support the view that females that select males with large repertoires also acquire good genes for offspring viability. Additional support is derived from new findings that males infected with hematozoan parasites have smaller repertoires. Recent research has also shown that there is a relationship between repertoire size and the size of the neural pathway controlling song learning in the brain. It is suggested that the extra space for learning and storing larger repertoires involves additional neural costs. During development the neural pathway may be influenced by various factors, such as nutrition and disease, and these may also affect neural space. Only better quality males may be able to afford the extra costs of mounting an immune response and producing more neurons to store complex songs. Such a hypothesis remains to be tested, but if true it would help to explain why repertoire size has evolved as an honest indicator of male quality. Acknowledgments I am grateful to many people who have helped and collaborated with me over the years, particularly Bernd Leisler who is coauthor on many important papers based on our field and experimental projects in Europe. In England, I have been ably supported by Kate Buchanan and Rupert Marshall, who have carried out most of the later fieldwork and DNA fingerprinting on the sedge warbler project. In recent years, Tim DeVoogd and Dave Airey have collaborated on brain studies, Terry Burke on DNA fingerprinting, Tim Birkhead on sperm, Tamas Sze´kely on phylogeny, and John Lewis on parasites.

References Airey, D. C., Buchanan, K. L., Catchpole, C. K., and DeVoogd, T. J. (2000). Song complexity, sexual selection and a song control nucleus (HVC) in the brains of European sedge warblers. Manuscript in preparation.

SEXUAL SELECTION AND EVOLUTION IN Acrocephalus WARBLERS

93

Andersson, M. (1982). Female choice selects for extreme tail length in a widowbird. Nature 299, 818–820. Andersson, M. (1994). ‘‘Sexual Selection.’’ Princeton Univ. Press, Princeton, NJ. Armstrong, E. A. (1973). ‘‘A Study of Bird Song.’’ Dover, New York. Bell, B. D., Boroweic, M., McKonkey, K., and Ranoszek, E. (1997). Settlement, breeding success and song repertoires of monogamous and polygynous sedge warblers Acrocephalus schoenobaenus. Die Vogelwarte 39, 87–94. Bensch, S., and Hasselquist, D. (1992). Evidence for female choice in a polygynous warbler. Anim. Behav. 44, 301–311. Bernard, D. J., Eens, M., and Ball, G. F. (1996). Age and behaviour related variation in the volumes of song control nuclei in male European starlings. J. Neurobiol. 30, 329–339. Birkhead, T. R., and Møller, A. P. (1992). ‘‘Sperm Competition in Birds. Evolutionary Causes and Consequences.’’ Academic Press, London. Birkhead, T. R., Buchanan, K. L., DeVoogd, T. J., Pellat, E. J., Sze´kely, T., and Catchpole, C. K. (1997). Song, sperm quality and testes asymmetry in the sedge warbler. Anim. Behav. 53, 965–971. Brenowitz, E. A., and Kroodsma, D. E. (1996). The neuroethology of birdsong. In ‘‘Ecology and Evolution of Acoustic Communication in Birds’’ (D. E. Kroodsma and E. H. Miller, Eds.), Cornell Univ. Press, Ithaca, NY. Brenowitz, E. A., Lent, K., and Kroodsma, D. E. (1995). Brain space for learned song in birds develops independently of song learning. J. Neurosci. 15, 6281–6286. Buchanan, K. L. (1997). Song and sexual selection in the sedge warbler Acrocephalus schoenobaenus. Unpublished PhD thesis, University of London, London. Buchanan, K. L., and Catchpole, C. K. (1997). Female choice in the sedge warbler Acrocephalus schoenobaenus: Multiple cues from song and territory quality. Proc. R. Soc. London Ser. B 264, 521–526. Buchanan, K. L., and Catchpole, C. K. (1999). Extra-pair paternity in the socially monogamous sedge warbler Acrocephalus schoenobaenus as revealed by multilocus DNA fingerprinting. Ibis 141, in press. Buchanan, K. L., Catchpole, C. K., Lewis, J. L., and Lodge, A. (1999). Song as an indicator of parasitism in the sedge warbler. Anim. Behav. 57, 307–314. Canady, R. A., Kroodsma, D. E., and Nottebohm, F. (1984). Population differences in complexity of a learned skill are correlated with the brain space involved. Proc. Natl. Acad. Sci. USA 81, 6232–6234. Catchpole, C. K. (1973). The functions of advertising song in the sedge warbler Acrocephalus schoenobaenus and the reed warbler A. scirpaceus. Behaviour 46, 300–320. Catchpole, C. K. (1976). Temporal and sequential organisation of song in the sedge warbler (Acrocephalus schoenobaenus). Behaviour 59, 226–246. Catchpole, C. K. (1977). Aggressive responses of male sedge warblers (Acrocephalus schoenobaenus) to playback of species song and sympatric species song, before and after pairing. Anim. Behav. 25, 489–496. Catchpole, C. K. (1979). ‘‘Vocal Communication in Birds.’’ Edward Arnold, London. Catchpole, C. K. (1980). Sexual selection and the evolution of complex songs among warblers of the genus Acrocephalus. Behaviour 74, 149–166. Catchpole, C. K. (1982). The evolution of bird sounds in relation to mating and spacing behaviour. In ‘‘Acoustic Communication in Birds’’ (D. Kroodsma and E. H. Miller, Eds.), pp. 297–319. Academic Press, New York. Catchpole, C. K. (1983). Variation in the song of the great reed warbler Acrocephalus arundinaceus in relation to mate attraction and territorial defence. Anim. Behav. 31, 1217–1225.

94

CLIVE K. CATCHPOLE

Catchpole, C. K. (1986). Song repertoires and reproductive success in the great reed warbler Acrocephalus arundinaceus. Behav. Ecol. Sociobiol. 19, 439–445. Catchpole, C. K. (1987). Bird song, sexual selection and female choice. Trends Ecol. Evol. 2, 94–97. Catchpole, C. K. (1989). Responses of male sedge warblers to playback of different repertoire sizes. Anim. Behav. 37, 1046–1047. Catchpole, C. K. (1992). Integrating playback: A wider context. In ‘‘Playback and Studies of Animal Communication’’ (P. K. McGregor, Ed.), pp. 35–46. Plenum, New York. Catchpole, C. K. (1995). The evolution of mating systems in Acrocephalus warblers. Jpn. J. Ornithol. 44, 195–207. Catchpole, C. K. (1996). Song and female choice: Good genes and big brains? Trends Ecol. Evol. 11, 358–360. Catchpole, C. K., and Komdeur, J. (1993). The song of the Seychelles warbler Acrocephalus sechellensis an island endemic. Ibis 135, 190–195. Catchpole, C. K., and Leisler, B. (1989) Variation in the song of the aquatic warbler Acrocephalus paludicola in response to playback of different song structures. Behaviour 108, 125–138. Catchpole, C. K., and Leisler, B. (1996) Female aquatic warblers Acrocephalus paludicola are attracted by playback of longer and more complicated songs. Behaviour 133, 1153–1164. Catchpole, C. K., Dittami, J., and Leisler, B. (1984). Differential responses to male song repertoires in female songbirds implanted with oestradiol. Nature 312, 563–564. Catchpole, C. K., and Slater, P. J. B. (1995). ‘‘Bird Song: Biological Themes and Variations.’’ Cambridge Univ. Press, Cambridge, UK. Catchpole, C. K., Leisler, B., and Winkler, H. (1985) Polygyny in the great reed warbler, Acrocephalus arundinaceus: A possible case of deception. Behav Ecol Sociobiol 16, 285–291. Catchpole, C. K., Leisler, B., and Dittami, J. (1986) Sexual differences in the responses of captive great reed warblers (Acrocephalus arundinaceus) to variation in song structure and size. Ethology 73, 69–77. Darwin, C. (1871) ‘‘The Descent of Man and Selection in Relation to Sex.’’ John Murray, London. DeVoogd, T. J. (1994) The neural basis for the acquisition and production of bird song. In ‘‘Causal Mechanisms of Behavioural Development’’ ( J. A. Hogan and J. J. Bolhuis, Eds.) pp. 49–81. Cambridge Univ. Press, Cambridge, UK. DeVoogd, T. J., Krebs, J. R., Healy, S. D., and Purvis, A. (1993). Relations between song repertoire size and the volume of brain nuclei related to song: Comparative evolutionary analyses amongst oscine birds. Proc. R. Soc. London Ser. B 254, 75–82. DeVoogd, T. J., Cardin, J. A., Sze´kely, T., Buki, J., and Newman, S. W. (1996) Relative volume of 1MAN in female warbler species varies with number of songs produced by conspecific males Soc. Neurosci. Abstr. 22, 1401. Dowsett-Lemaire, F. (1979). The imitative range of the song of the marsh warbler Acrocephalus palustris, with special reference to imitations of African birds. Ibis 121, 453–468. Duckworth, J. W. (1992). Effects of mate removal on the behaviour and reproductive success of reed warblers Acrocephalus scirpaceus. Ibis 34, 164–170. Dyrcz, A., and Zdunek, W. (1993). Breeding ecology of the aquatic warbler Acrocephalus paludicola on the Biebrza marshes, northeast Poland. Ibis 135, 181–189. Eens, M. (1997). Understanding the complex song of the European starling: An integrated ethological approach. Adv. Study Behav. 26, 355–434. Eens, M., Pinxten, R., and Verheyen, R. F. (1993). Function of the song and song repertoire in the European starling (Sturnus vulgaris): An aviary experiment. Behaviour 125, 51–66.

SEXUAL SELECTION AND EVOLUTION IN Acrocephalus WARBLERS

95

Eriksson, D., and Wallin, L. (1986). Male bird song attracts females—A field experiment. Behav. Ecol. Sociobiol. 19, 297–299. Fessl, B., and Hoi, H. (1996). The significance of a two part song in the moustached warbler (Acrocephalus melanopogon). Ethol. Ecol. Evol. 8, 265–278. Fisher, R. A. (1930). ‘‘The Genetical Theory of Natural Selection.’’ Clarendon, Oxford. Folstad, I., and Karter, A. J. (1992). Parasites, bright males, and the immunocompetence handicap. Am. Nat. 139, 603–622. Greig-Smith, P. W. (1982). Song rates and parental care by individual male stonechats Saxicola torquata. Anim. Behav. 30, 245–252. Hamilton, K. S., King, A. P., Sengelaub, D. R., and West, M. (1997). A brain of her own: A neural correlate of song assessment in a female songbird. Neurobiol. Learning Memory 68, 325–332. Hamilton, W. D., and Zuk, M. (1982). Heritable true fitness and bright birds: A role for parasites? Science 218, 384–387. Hasselquist, D., and Bensch, S. (1991). Trade-off between mate guarding and mate attraction in the polygynous great reed warbler. Behav. Ecol. Sociobiol. 28, 187–193. Hasslequist, D., and Langefors, A. (1998). Variable social mating system in the sedge warbler Acrocephalus schoenobaenus. Ethology 104, 759–769. Hasselquist, D., Bensch, S., and Ottoson, U. (1993). Diurnal song pattern in the great reed warbler, Acrocephalus arundinaceus. Ornis Svecica 3, 43–60. Hasselquist, D., Bensch, S., and Schantz, T. (1996). Correlation between male song repertoire, extra-pair paternity and offspring survival in the great reed warbler. Nature 381, 229–232. Hill, G. (1991). Plumage colouration is a sexually selected indicator of male quality. Nature 350, 337–339. Hillgarth, N., and Wingfield, J. C. (1997). Parasite-mediated sexual selection; Endocrine aspects. In ‘‘Host–Parasite Evolution: General Principles and Avian Models’’ (D. H. Clayton and J. Moore, Eds.), pp. 78–104. Oxford Univ. Press, Oxford. Hoi-Leitner, M., Nechtelberger, H., and Dittami, J. (1993). The relationship between individual differences in male song frequency and parental care in blackcaps. Behaviour 126, 1–12. Houtman, A. M. (1992). Female zebra finches choose extra-pair copulations with genetically attractive males. Proc. R. Soc. London Ser. B 249, 3–6. Howard, R. D. (1974). The influence of sexual selection and interspecific competition on mocking bird song (Mimus polyglottos). Evolution 28, 428–438. Jacobs, L. F. (1996). Sexual selection and the brain. Trends Ecol. Evol. 11, 82–86. Johnson, L. S., and Searcy, W. A. (1996). Female attraction to male song in house wrens (Troglodytes aedon). Behaviour 133, 357–366. Johnstone, R. A. (1996). Multiple displays in animal communication: Backup signals and multiple messages. Philos. Trans. R. Soc. London Ser. B 351, 329–338. Johnstone, R. A. (1997). The evolution of animal signals. In ‘‘Behavioural Ecology: An Evolutionary Approach’’ ( J. R. Krebs and N. B. Davies, Eds.), pp. 155–178. Blackwell, Oxford. Kelsey, M. G. (1989). A comparison of the song and territorial behaviour of a long-distance migrant, the marsh warbler Acrocephalus palustris, in summer and winter. Ibis 131, 403–414. Kirkpatrick, M., and Ryan, M. (1991). The evolution of mating preferences and the paradox of the lek. Nature 350, 33–39. Kim, J. R., Clower, R. P., Kroodsma, D. E., and Voogd, T. J. D. (1989). Song-related brain regions in the red-winged blackbird are affected by sex and season but not repertoire size. J. Neurobiol. 20, 139–163. Kroodsma, D. E. (1986). Design of song playback experiments. Auk 103, 640–642.

96

CLIVE K. CATCHPOLE

Kroodsma, D. E. (1990). Using appropriate experimental designs for intended hypotheses in song playbacks, with examples for testing effects of song repertoire sizes. Anim. Behav. 40, 1138–1150. Kroodsma, D. E., and Byers, B. E. (1991). The function(s) of bird song. Am. Zool. 31, 318–328. Kroodsma, D. E., and Canady, R. A. (1985). Differences in repertoire size, singing behaviour, and associated neuroanatomy among Marsh Wren populations have a genetic basis. Auk 102, 439–446. Lampe, H. M., and Saetre, G.-P. (1995). Female pied flycatchers prefer males with larger song repertoires. Proc. R. Soc. London Ser. B 262, 163–167. Langefors, A., Hasselquist, D., and Schantz, T. V. (1998). Extra-pair fertilizations in the sedge warbler. J. Avian Biol. 29, 134–144. Leisler, B., and Catchpole, C. K. (1992). The evolution of polygamy in European reed warblers of the genus Acrocephalus: A comparative approach. Ethol. Ecol. Evol. 4, 225–243. Leisler, B. and Wink, M. (1999). Frequencies of multiple paternity in three Acrocephalus species (Aves: Syviidae) with different mating systems (A. palustris, A. arundinaceus, A. paludicola). Behaviour, in press. Leisler, B., Heidrich, P., Schulze-Hagen, K., and Wink, M. (1997). Taxonomy and phylogeny of reed warblers (genus Acrocephalus) based on mtDNA sequences and morphology, J. Ornithol. 138, 469–496. Marshall, R. C., Buchanan, K. L., and Catchpole, C. K. (1998). Song and female choice for extra-pair copulations in the sedge warbler. In ‘‘Abstracts of the 7th International Behavioral Ecology Congress,’’ p. 278. McGregor, P. K. (Ed.) (1992). ‘‘Playback and Studies of Animal Communication.’’ Plenum, New York. Milder, S. C., and Schreiber, R. W. (1989). The vocalizations of the Christmas Island Warbler Acrocephalus aequinoctialis, an island endemic. Ibis 131, 99–111. Møller, A. P. (1991a). Parasite load reduces song output in a passerine bird. Anim. Behav. 41, 723–730. Møller A. P. (1991b). Why mated songbirds sing so much: Mate guarding and male announcement of mate fertility status. Am. Nat. 138, 994–1014. Mountjoy, D. J., and Lemon, R. E. (1991). Song as an attractant for male and female European starlings, and the influence of song complexity on their response. Behav. Ecol. Sociobiol. 28, 97–100. Nottebohm, F. (1993). The search for neural mechanisms that define the sensitive period for song learning in birds. Netherlands J. Zool. 43, 193–234. Nottebohm, F., and Arnold, A. P. (1976). Sexual dimorphism in vocal control areas of the songbird brain. Science 194, 211–213. Nottebohm F., Kasparian, S., and Pandazis, C. (1981). Brain space for a learned task. Brain Res. 213, 99–109. Nowicki, S., Peters, S., and Podos, J. (1998). Song learning, early nutrition and sexual selection in songbirds. Am. Zool. 38, 179–190. Read, A. F., and Weary, D. M. (1990). Sexual selection and the evolution of bird song: A test of the Hamilton–Zuk hypothesis. Behav. Ecol. Sociobiol. 26, 47–56. Ryan, M. J., Fox, J. H., Wilczynski, W., and Rand, A. S. (1990). Sexual selection for sensory exploitation in the frog Physalaemus pustulosus, Nature 343, 66–67. Saino, N., Galeotti, P., Sacchi, R., and Moller, A. P. (1997). Song and immunological condition in male barn swallows Hirundo rustica. Behav. Ecol. 8, 364–371. Schulze-Hagen, K., Swatschek, I., Dyrcz, A. and Wink, M. (1993). Multiple paternity in broods of aquatic warblers Acrocephalus paludicola, first results of a DNA fingerprinting study. J. Ornithol. 134, 145–154.

SEXUAL SELECTION AND EVOLUTION IN Acrocephalus WARBLERS

97

Schulze-Hagen, K., Leisler B., Birkhead, T. R., and Durcz, A. (1995). Prolonged copulation, sperm reserves and sperm competition in the aquatic warbler Acrocephalus paludicola. Ibis 137, 85–91. Searcy, W. A., (1984). Song repertoire size and female preferences in song sparrows. Behav. Ecol. Sociobiol. 14, 281–286. Searcy, W. A. (1990). Species recognition of song by female red-winged blackbirds. Anim. Behav. 40, 1119–1127. Searcy, W. A. (1992a). Measuring responses of female birds to male song. In ‘‘Playback and Studies of Animal Communication (P. K. McGregor, Ed.) pp. 175–189. Plenum, New York. Searcy, W. A. (1992b). Song repertoire and mate choice in birds. Am. Zool. 32, 71–80. Searcy, W. A., and Andersson, M. (1986). Sexual selection and the evolution of song. Annu. Rev. Ecol. Syst. 17, 507–533. Searcy, W. A. and Marler, P. (1981). A test for responsiveness to song structure and programming in female sparrows. Science 213, 926–928. Searcy, W. A., and Yasukawa, K. (1990). Use of the song repertoire in intersexual and intrasexual contexts by male red-winged blackbirds. Behav. Ecol. Sociobiol. 27, 123–128. Searcy, W. A., and Yasukawa, K. (1995). ‘‘Polygyny and Sexual Selection in Red-Winged Blackbirds.’’ Princeton University Press, Princeton, NJ. Searcy, W. A., and Yasukawa, K. (1996). Song and female choice. In ‘‘Ecology and Evolution of Acoustic Communication in Birds’’ (D. E. Kroodsma and E. H. Miller, Eds.), pp. 454–473. Cornell Univ. Press, Ithaca, NY. Sheldon, B. (1994). Male phenotype, fertility, and the pursuit of extra-pair copulations by female birds. Proc. R. Soc. London Ser. B 257, 25–30. Sze´kely, T., Catchpole, C. K., DeVoogd, A., Marchl, Z., and DeVoogd, T. J. (1996). Evolutionary changes in a song control area of the brain (HVC) are associated with evolutionary changes in song repertoire among European warblers (Sylviidae). Proc. R. Soc. London Ser. B. 263, 607–610. 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. Zahavi, A. (1975). Mate selection—A selection for a handicap. J. Theor. Biol. 53, 205–214. Zahavi, A. (1977). The cost of honesty—Further remarks on the handicap principle. J. Theor. Biol. 67, 603–605.

This Page Intentionally Left Blank

ADVANCES IN THE STUDY OF BEHAVIOR, VOL. 29

Primate Socialization Revisited: Theoretical and Practical Issues in Social Ontogeny Bertrand L. Deputte cnrs/umr 6552 station biologique, 35380 paimpont, france I. Introduction At birth, a newborn mammal enters a new world composed of a huge variety of stimuli and will continue to develop within this complex environment, progressively responding or learning to respond to a much reduced set of significant features. In nonhuman primates the environmental complexity includes physical objects, plants, animals, and conspecifics. These conspecifics form a stable but dynamic environment which remains constant during most of the initial stages of the newborn’s development. Reflexively grasping its mother, a newborn monkey will progressively enter the real world of conspecifics in which its mother lives. Most of its first contacts with its physical environment will be accompanied by the presence of one or several conspecifics. Later, the infant’s behavior will be more constrained by, oriented to, or simply influenced by the conspecifics remaining with it day after day. Thus, for a part of its life, during which its nervous structures (which process, store, and analyze information) develop, the infant has to learn how to become a member of the group, to be a partner for the other conspecifics in order to remain in the group while fulfilling other fundamental biological needs. Although in this description I did not use the word ‘‘social,’’ it summarizes what authors have published under the titles of ‘‘socialization in monkeys’’ or ‘‘social development in monkeys.’’ One major consequence of the inherent complexity of social development is that concepts developed much before empirical evidence was derived. In other words, in social development the conceptual level has remained mainly one of discussion and was not put into operation. Do the necessary analytical tools now exist to make operational the most elaborate concepts in primate social development? In this chapter, after defining some important terms, I review these concepts showing that they actually constitute different approaches to the 99

Copyright 䉷 2000 by Academic Press All rights of reproduction in any form reserved. 0065-3454/00 $35.00

100

BERTRAND L. DEPUTTE

same question with variable emphasis on differents features. In a second section I show how quantitative analytical methods have progressively become more suitable to provide synthetic views of the interweaving developmental variables and to untangle their influences. A. Definition of Terms To clarify the issues and to avoid potential misunderstanding, I follow the advice of Etkin (1964) and Fedigan (1982) in making explicit what I mean by social. Social is derived from the latin word Socius (companion); therefore, ‘‘social’’ is used where there is more than one individual involved (Espinas, 1878; Crook, 1970a; Tinbergen, 1967). Thus, if being socialized means being able to live in a group, then studying infant socialization would be essential for the understanding of the proximate causes of group life (Fragaszy and Mitchell, 1974). The term social is linked to that of ‘‘society,’’ which refers to a special kind of grouping of individuals of the same species. Consequently, a social interaction should only mean interaction between individuals of the same species in the context of a group they have formed. Sociality, the essence of being social, is more often defined by its consequences (different kinds of groupings—Sailer and Gaulin, 1984; Smuts et al., 1986; recurrent interactions—Altmann, 1965) than by its unique features that differentiate social groups from mere aggregations of animals. The unique feature of sociality is an ‘‘interattraction’’ between members of the same species (Wheeler, 1923; Rabaud, 1937; Maier and Schneirla, 1964). Thus, the term society, in a strict sense, stresses ‘‘the complexity of the individual relationships in the [social] group’’ (Maier and Schneirla, 1964, p. 165). It is difficult to envisage the development of a young mammal, in general, and of a young primate, in particular, outside the presence of its mother. In mammals the initial infant–mother mandatory bond, arising from nutritional dependency and the general helplessness of the young, might be an excellent candidate for the emergence of sociality. The primate infant– mother bond has been characterized as an ‘‘attachment,’’ a process that underlies the tendency of an organism to show positive behavior toward another organism (reminiscent of the intraspecific ‘‘interattraction’’ that makes a species social). Attachment is measured by maintenance of spatial proximity and disruption of behavior on separation (Swartz, 1982a; Kraemer, 1992). In addition, attachment has cognitive implications involving some kind of internal representations of the ‘‘attachment figure.’’ Beyond the early stages of development in which mothers are unique and specific attachment figures for infants, attachment may become incorporated within a more general approach–withdrawal process (Schneirla, 1959).

PRIMATE SOCIALIZATION REVISITED

101

This attachment process might also be considered as including the reconciliation process defined by de Waal and van Roosmalen (1979) which leads to the maintenance of social cohesion despite aggression, received or given. To reconcile with an aggressor, or a ‘‘victim,’’ could be seen as analogous to the reunion phase in a separation–reunion process from and with an attachment figure. The social features transcend the individuality of group members. Most primates spent their entire lives as members of an organized group despite intragroup conflicts and turnover in membership (Mason, 1976). Thus, sociality must be primarily defined by two basic features: an interattraction between individuals of the same species. Bernstein and Williams (1986) contend that social animals differ from solitary ones ‘‘only in a quantitative sense.’’ A new gibbon group without offspring is like any other mated pair of mammals except that it lasts for an extended period of time, quantitatively much longer than that of mated pairs in solitary animals. However, this quantitative difference is only superficial. The gibbon pair is maintained because the species has acquired during evolution what Mason (1978a) called a ‘‘social disposition’’ which allows same species individuals to stay together by adjusting their behavior to each other. What differentiates a social behavior from any other kind of behavior? Definitions insist on the contingent and/or interactive character of behavior within a group of animals. Etkin (1964), instead of defining a social behavior, defined a social response, which is social because it is restricted to group members. Crook (1970a) and Bernstein and Williams (1986) introduced either implicitly or explicitly the concept of communication within their definition of a social response, whereas Altmann (1962) considers that all social behavior is communicative. One might consider that communicative behaviors constitute a subset of social behaviors as, in social groups, the concept of communicative acts goes beyond the traditional notion of signals. Bernstein and Williams (1986) and Deputte and Vauclair (1998) stress the fact that nonhuman primate communication is basically ‘‘inferential.’’ This means that, in a social context, the mere proximity of a group member, its movement in whatever direction, and its posture provide information that might influence and shape the behavior of some other group members. The quantitative level of interactions is a criterion Altmann (1967) used to define a nonhuman primate society which ‘‘consists of conspecific intercommunicating individuals that are bounded by frontiers.’’ Bernstein and Williams (1986) considered four basic features to define a society: temporal stability, spatial cohesion, communication and coordination of activities, and recognition and differentiation of group members from nonmembers. All these features could be considered as a consequence of mutual attraction.

102

BERTRAND L. DEPUTTE

Defining a social group by the consequences of ‘‘mutual attraction’’ involves a certain degree of circularity: Social groups differ from mere ‘‘aggregations’’ when the cohesion results from ‘‘social responses to one another rather than by responses in their environment’’ (Etkin, 1964, p. 4). A high level of interaction is a likely consequence of proximity (Altmann, 1965; Kummer, 1978; Sailer and Gaulin, 1984), and that proximity is probably achieved after a certain degree of interactive process yielding a certain degree of organization. From the concept of social groups, two new concepts, social structure and social organization, emerge. Because they are often used interchangeably, they need to be clarified. Hinde (1975) defined a social structure as the upper level of a conceptual framework which relates interactions, relationships, and social structure; the latter is seen at the data stage as the content, quality, and patterning of relationships. Other authors call this level social organization. Rowell (1972) and van Schaik and van Hooff (1983) defined social organization (in the strict sense) as ‘‘the processes of social interaction and their patterns of distribution over group members’’ (van Schaik and van Hooff, 1983) and social structure as ‘‘the composition of the group and the spatial patterns of individuals’’ (van Schaik and van Hooff, 1983, p. 92). The social structure is essentially the demographic composition of the group, and social organization is what Hinde (1975) called the ‘‘social structure’’ and resembles the social organization defined by Crook (1970a) as ‘‘described mainly in terms of interactions between 2 individuals.’’ The distinction made here between social structure and organization allows one to describe different social groupings with the same organization and, symmetrically, different social organization in groups of the same structure. This distinction also helps to disentangle the influences of the environment, most likely on social structure, and that of the social dispositions and/or differential attachments of the group members, most likely on social organization. B. Socialization and Social Development in Primates Although these two terms are not synonymous, they have been used more or less interchangeably to describe how a young primate develops. ‘‘Socialization’’ is generally defined differently by almost every author who deals with this issue (Fedigan, 1982). Definitions most often assign a normative feature to the process: Poirier (1973) defined socialization as the process linking an ongoing society to a new individual and then enabling the ways of life and social traditions of a group to be passed to succeeding generations. Crook (1970b), Fragaszy and Mitchell (1974), Rosenblum and Coe (1977), and Fedigan (1982) all agreed on this conformity feature and/or on

PRIMATE SOCIALIZATION REVISITED

103

the transmission of group lifestyle, a kind of ‘‘group assimilation of the individual’’ (Schneirla, 1952). Socialization is also defined more explicitly by its processes, keeping in mind that it is not a single process but rather an interplay of processes (McKenna, 1979): The processes referred to are the interactions between a new individual and the established group which it joins (Chance and Jolly, 1970; Hinde, 1971; Lewis and Cherry, 1977; Berman, 1982a,b; Deputte, 1986b). Wilson (1975, p. 159) went further in asserting that socialization is ‘‘the sum total of all social experiences that alter the development of an individual.’’ The other related process is learning (Fragaszy and Mitchell, 1974; McKenna, 1979; Fedigan, 1982; Poirier, 1982), making socialization ‘‘the sum total of an individual’s past social experience’’ (Poirier, 1972, 1982). The normative feature of socialization could be viewed as a consequence of interactive learning. The likely consequences of both the normative and the interactive features are also expressed by several authors who assumed that different social settings should produce different socialization trajectories (Maier and Schneirla, 1964; Mason, 1965; Baldwin and Baldwin, 1979; Poirier, 1982; Deputte, 1986a; Deputte and Quris, 1996). Actually, socialization in primates is a life-long process that includes what happens to an individual who enters an established group by means of either birth or immigration (Rosenblum and Coe, 1977; Bernstein and Williams, 1986). Therefore, it is worth stressing that socialization means ‘‘the process of being socialized’’, that is, to adjust behaviorally in order to achieve an integration into an organized group. An example of ‘‘being socialized’’ in learning the ‘‘group lifestyle’’ is provided by the introduction of anubis baboons into hamadryas groups and the converse introduction of hamadryas females into anubis groups in the baboon hybrid zone in Ethiopia (Kummer et al., 1970). Although this experiment was concerned with interspecific interactions, it showed how an individual could become a group member by virtue of learning (or forgetting) species typical behaviors, specifically the hamadryas herding behavior (Kummer et al., 1970). Infant socialization could be equated with social development. The latter term does not necessarily possess the normative connotation of socialization, although Schneirla contended that ‘‘development environments . . . tend to be species-standardized’’ making experiences ‘‘more or less prestructured at all stages’’ (Maier and Schneirla, 1964, p. 573). Social development refers to a domain of behavioral development for which a set of determinants, the members of the group, receives a special emphasis after it has been postulated that each newborn monkey, still developing, is already fit to function within a particular social niche (Mason, 1976).

104

BERTRAND L. DEPUTTE

C. Searching for a Unifying Concept for Social Development and Socialization This preamble is intended to make clear that a nonhuman primate infant is born with ‘‘social dispositions’’ which are basically a compelling attraction to conspecifics and a readiness to learn and to accumulate experiences. All group members represent the ‘‘socializing agents’’ for a newborn (Poirier, 1977). Harlow and Harlow (1965) considered how the essence of sociality, ‘‘binding disposition,’’ develops in a newborn primate; they described five ‘‘affectional systems,’’ four of which included the infant (infant–mother, infant–infant, mother–infant, or maternal, paternal, and the sexual and heterosexual affectional system which applied to adolescents). These systems all involve different kinds of interactions and each system includes different stages that might be common to the other systems (e.g., the reflex stage common to infant–infant and infant–mother systems). These systems all referred to a dyadic system without making explicit that each system might influence or might be influenced by another one. The thoughtful idea that a social group functions as a social nexus was developed by Hinde (1971): Each group member is embedded in a unique social nexus which is represented by the nature of the relationships this group member has with all other group partners [the ‘‘social nexus’’ formulation rephrased from that of Maier and Schneirla (1964), who underscored the ‘‘complexity of the individual relationships in the social group’’]. Each dyadic interaction involves, directly or indirectly, the social nexus (or network) of each participant. At a group level, it is also possible that different classes of individuals or different individuals display a behavioral differentiation which could be class specific or status specific. This behavioral differentiation led to the concept of social role (Bernstein, 1966; Bernstein and Sharpe, 1966) that was defined later by Reynolds (1970, p. 450) as ‘‘statistically probable behaviors in a given interactional and ecological setting.’’ Several authors have made this definition operational using different multivariate statistical techniques (Bramblett, 1973; Fedigan, 1976; Fairbanks et al., 1978; Fairbanks and McGuire, 1979). The Harlows’ affectional systems and the classes of partners determined by Hinde (1971) and later relabeled ‘‘socializing agents’’ by Poirier (1977) all refer to differential roles that group members play in shaping the social behavior of developing infants. The consequence of these differential roles might be an infant’s behavioral differentiation, which is partner based (Berman, 1982a; Swartz, 1982b). An individual dimension could be added as a first level to the three other levels Hinde (1976b) described. The interaction level and the individual level would be the only two observable levels. When Hinde (1971) and

PRIMATE SOCIALIZATION REVISITED

105

Fragaszy and Mitchell (1974) referred to physical development while describing social development or socialization, respectively, they made an implicit reference to the individual. Fragaszy and Mitchell distinguished two levels in the socialization process—the group level and the individual level. The individual is the ‘‘basic unit in a social system’’ (Mason, 1976) and, from the beginning, the developing individual is a part of an elaborate network of social relations (Mason, 1965). This consequently stressed the importance of relationships, the ‘‘qualitative unique element’’ of a society (Kummer, 1978). D. Social Ontogeny as the Unifying Concept An individual has a certain internal anatomy, a certain morphology, and a certain physiology. It processes information, behaves in a certain way, and can acquire a certain experience. All these dimensions, from anatomy to behavior, are subjected to development. All these processes, the outcome of which is one individual, participate in the comprehensive process of ontogeny, ‘‘the genesis of an individual.’’ Ontogeny refers explicitly and only to an individual. Through all the developmental processes ontogeny leads to a unique individual (Mason, 1976). Ontogeny should not be used as a mere synonym of ‘‘development’’ because it encompasses many developmental processes. One then could add a qualifier to restrict one’s interest to specific domains of the genesis of an individual, such as ‘‘physiological ontogeny’’ or ‘‘hormonal ontogeny.’’ Here, we use ‘‘behavioral ontogeny’’ and ‘‘social ontogeny’’ to refer to individual development of behavior and to this development within a social setting while keeping in mind that these ontogenetic domains are overlapping (Fig. 1). Behavioral ontogeny is a subset of physical ontogeny because it refers specifically to the development of the control function that the brain exerts on the two-way relation between the organism and its environment (Fig. 1). Psychological ontogeny is another subset that refers to the development of cognitive capacities (perception, processing, and storage of information; Fig. 1). Eventually, social ontogeny is at the intersection of all these domains (Fig. 1). Behavioral and social ontogeny could be considered as epigenetic processes, functions of both genetic–maturational and environmental factors (Denenberg, 1979), and are a clear example of what Gottlieb (1970) labeled ‘‘probabilistic epigenesis.’’ Hailman (1982) defined a ‘‘phenotypic control function’’ which determines, in a probabilistic way in a two-way process, a certain behavioral distribution from a combination of motivational state and stimulus input. Therefore, Hailman (1982) emphasized the epigenetic feature of (behavioral) ontogeny by considering that ontogeny consists of the permanent

106

BERTRAND L. DEPUTTE

Fig. 1. Ontogeny and some ‘‘convenient’’ subdivisions.

changes of the control function with time. He formalizes this assessment with an ontogenetic function (O): O: (Pt⫺1,G, E ) 씮 Pt which maps the ontogenetic changes of one phenotype, P, its genotype G, and the present environment E to a new phenotype during some small increment of time (from t⫺1 to t). This formulation is in line with the statement that behavioral development consists of ‘‘a program of progressive changing relationships between organisms and environment in which the contributions of growth are always inseparably interrelated with those of the effects of energy changes in the environs’’ (Schneirla, 1959). The Hailman framework enables us to schematize the ontogeny of a social being such as a primate (Fig. 2). According to Maier and Schneirla (1964, p. 268), development ‘‘must be considered as an interaction between undifferentiated germ tissue and an environment.’’ Therefore, during fetal life, embryogenic processes, in interaction with the maternal uterine envi-

PRIMATE SOCIALIZATION REVISITED

107

Fig. 2. Ontogeny in primates: processes in time. Genotype controls the development of an ‘‘individual phenotype’’ through embryogeny. Immediately before and after birth behavioral ontogeny and its subset, social ontogeny, occut within and are influenced by a social environment. Thus, both behavioral and social ontogeny are primarily epigenetic processes.

ronment, yield a unique phenotype labeled ‘‘individual phenotype.’’ During the last stages of gestation, behavioral ontogeny starts with the development of motor and nervous systems and brain structures (sensorimotor development; Fig. 2). During its intrauterine development, a primate fetus is already within a social group because of its mother and is subjected to social influences through its mother’s reactions. The new primate really joins the group at birth, but it will still be under social influences through its mother and her previous social experience. From that particular moment, its ontogeny will occur within a social setting; therefore, social ontogeny starts as the component of behavioral ontogeny that is most influenced by the group members, an environmental and nongenetic variable (even if kin might take a large role as socializing agents). However, social ontogeny remains dependent on some fundamental features of primate development, especially the need for extensive parental care or parental-like care and a slow rate of maturation (involving motor coordination and, to a much greater extent, sexual maturation). Through epigenetic processes, social ontogeny yields a new phenotype—a social phenotype defined as the individual phenotype (a genotypic outcome) plus the social relationships comprising the social network (Fig. 2). The model described in Fig. 2 is clearly an ‘‘organis-

108

BERTRAND L. DEPUTTE

Fig. 3. The ‘‘Russian dolls phenotype’’ paradigm in primates. The different phenotypes, each characterized by some attributes, are piled up on top of genotype. Social phenotype, here represented as larger than the behavioral phenotype, is actually an additonal attribute of the latter, allowing an individual (here a primate) to establish lasting relationships with conspecifics.

mic model’’ which views development as a dynamic process in which changes depend on complex interactions among genetic, internal, and external environmental factors (Maier and Schneirla, 1964; Sackett et al., 1981). The phenotype approach to ontogeny is a ‘‘Russian dolls’’ one (Fig. 3). An

Fig. 4. Social ontogeny and the social ontogenetic function, Os. Os maps the changes of an individual phenotype at birth, IPt 0, into a social phenotype, SP, through interactions with already established social phenotypes, both males and females, in the context of a social group. (Right) The social group is represented by the different individualized social phenotypes of both sexes which interact (arrows to and from) with the developing social phenotype (italic SP). The numbers associated with the social phenotypes to which the infant is exposed indicate the age–sex class to which they belong.

PRIMATE SOCIALIZATION REVISITED

109

individual is seen as an accumulation of several ‘‘types’’ from the genotype to the social phenotype. Actually, the social phenotype is not a new element. It is an ‘‘added’’ behavioral feature which allows an individual to adjust behaviorally to its partners. This also means that this additional behavioral (and motivational) feature is not restricted to sexual or parental interactions but includes all the other different behaviors involved in the constant regulation of proximity to other individuals. To rephrase Hailman’s (1982) statement, a social ontogenetic function, Os, maps the ontogenetic changes of one phenotype, its genotype, and the present social environment, Es, to a new kind of phenotype—a social phenotype (Fig. 4). This social ontogenetic function is a component of the general ontogenetic function and so is controlled by it. E. Social Ontogeny Processes The infant primate is exposed to many social phenotypes that are gender differentiated morphologically and possibly behaviorally. The social networks of these social phenotypes are already differentiated through the

Fig. 5. Russian dolls social phenotypes: the specific attributes of the different types of social phenotypes. During its ontogeny a primate infant discriminated, identified, and individualized its social phenotypes partners through the different attributes that required more or less elaborate cognitive skills to be perceived by the infant; from the simplest on the right (perception of a spatial proximity) to the most elaborate on the left (perception of individual uniqueness).

110

BERTRAND L. DEPUTTE

constraints of the social organization of the group. Any group member could be considered as an accumulation of features that are more or less accessible to new infants (Fig. 5): The ‘‘group member’’ feature is the easiest to detect and is exemplified by spatial proximities (Fig. 5). Age, sex, and individuality are also features which are observable through morphology (size and other features) and behaviors (Fig. 5). Age and sex are class features, whereas individuality is characterized by its uniqueness. Therefore, a social phenotype is both unique along one dimension (individual) and class specific along others (Fig. 5). The features of group social phenotypes are accessible to an infant in relation to its developing cognitive abilities (Poirier, 1977). Infants learned

Fig. 6. Building a social network: a social ontogenetic process leading to the integration of an infant primate in a group through learning of the different attributes of the social phenotypes composing the group. During its ontogeny, an infant (a gray-cheeked mangabey) becomes aware of the several dimensions of a partner (individual, social, and class related). He eventually sets up a unique social network. Interactions are indicated with arrows, the lengths of which indicate their intensity. The numbers within the circles individualize each partner and show the genealogical relationships. The different perceptual, cognitive, and social processes involved are mentioned under each stage. (Top, from left to right) First month, 1–3 months, 3–8 months. (Bottom) two different individuals more than 1 year old. Some relationships between their two networks are indicated.

PRIMATE SOCIALIZATION REVISITED

111

partners’ group membership, age, and sex classes through simple observation and/or comparison (Rosenblum and Pauly, 1980; Figs. 5 and 6). The perception of individuality requires higher cognitive abilities, discrimination, and categorization, which include age and sex discriminating features (Rosenblum and Pauly, 1980; Fig. 6). Age, gender, and individuality might also be detected through behavioral diversity displayed during interactions (Figs. 5 and 6). This detection requires higher forms of learning and increased memory abilities. Being confronted with different behavioral sets from those of partners will lead to the development of the infant’s social repertoire which will progressively fit most of the variety of social situations in which it is involved. During interactions an infant will learn that all its group members do not behave the same way toward him or her: This differentiation is the consequence of the social phenotypes shaped by the social organization. The reccurrence of differentiated interactions with identified partners leads to the establishment of a unique individual social network (Fig. 6; Papousek and Papousek, 1979). The two fundamental processes of social ontogeny are building a social behavioral repertoire and building a social network. The first process is tightly connected to the more general process of behavioral and cognitive development. The second process imposes a constant reshaping of the behavioral repertoire as a consequence of trial-and-error learning from social actions.

II. Methodological and Empirical Issues How should the complexity of social ontogeny be described and analyzed? How should the influences of maturational and social variables be disentagled? How should the concept of social network be put into practice? Only longitudinal studies are relevant to the issue of social ontogeny, a point stressed by Sackett et al. (1981). The same individuals should be observed from birth during periods of time ranging from a few months to several years. In the following sections I present different methodological approaches, from the simplest to the more integrative, to different levels of complexity concerning the related issues of behavioral development, social development, socialization, and social ontogeny. The latter term refers to the most complex and integrative approach to primate development. Examples will be drawn from ontogenetic studies of gray-cheeked mangabeys (Deputte, 1983, 1985, 1986a,b; Deputte and Quris, 1996) and rhesus macaques (Deputte and Goy, 1991; Deputte and Quris, 1997).

112

BERTRAND L. DEPUTTE

The study on gray-cheeked mangabeys (Cercocebus albigena albigena) was a 6-year study of social ontogeny on nine subjects (five females and four males). Subjects were observed longitudinally from birth to various ages: 4–6 years (five subjects, including three females) and more than 6 months and up to 1 year (four subjects, including two females). They were born in three social groups, each housed in indoor heated cages the volume of which varied from 36 to 75 m3. These wire-mesh cages were connected to wire-mesh outdoor cages of the same volume, to which monkeys had access during at least 8 months per year. The cement floor of the indoor cages was covered with a litter of straw. Both types of cage were equipped with dead trees and branches. The groups varied in size from three to eight individuals and in their composition (1 or 2 adult males and 1–4 adults females). Compared to captive macaques, the fertility rate of mangabeys was low and births could occur any time during the year (Deputte, 1991); the nine subjects were born over 6 years with a peak of four births during the next to last year of the study. In the macaque 7-month longitudinal study, four groups of rhesus macaques (Macaca mulatta) were observed at the Wisconsin Regional Primate Research Center. Each group contained five mother–infant pairs. Observations started when infants were from 1 to 1 months old and continued until they were at least 7 months old. In each group, the five infants included two males, two females, and one prenatally hormone-treated female, a ‘‘DES female’’ (diethylstilbestrol) whose mother had been injected daily with 100 애g DES from Day 115 to Day 139 of gestation. The protocol for the study was approved by the Animal Care Committee of the University of Wisconsin. Each group lived in a 2.2 ⫻ 3.0 m and 2.5-m-high pen with aluminum ladders, swings, and shelves (Goy and Robinson, 1982). Temperature, light, and humidity were automatically controlled in animal housing. Five-minute focal-animal samples were used (Altmann, 1974). The mangabey study represented 33,456 focal-animal samples (2788 h) and the macaque study consisted of 3624 samples (302 h). Within groups, individuals were observed in a random order. In both studies observation sessions occurred daily, five times a week, and were randomly distributed between 8:00 a.m. and 6:00 p.m. to cover a daily cycle. For the mangabey longitudinal study, yearly breaks in observation were no longer than 2 weeks. During a 5-min focal-animal sample, all the behaviors given and received by the focal subject were hand-written while preserving the sequential aspects of the interactions. A focal sample was analyzed as a single sentence composed of a certain number of syntactically organized triplets, for example, ‘‘Initiator I—gives the Behavior B—to Receiver R’’ (the focal subject being either I or R). The recorded behavioral units were considered as

PRIMATE SOCIALIZATION REVISITED

113

events (Altmann, 1974). Emphasis was on initiation and frequencies of behaviors; hence, long-lasting units such as ‘‘Grooms someone’’ were scored every 10 s when they were uninterrupted. Frequency per 5-min focal sample was computed for each behavior. The operational behavioral repertoire consisted of 156 and 160 mutually exclusive behavioral units for the mangabey and the rhesus macaque studies, respectively (Deputte, 1983, 1986a; Deputte and Goy, 1991; Deputte and Quris, 1996). This behavioral repertoire was large in order to cover the behavioral expression of individuals from birth to adulthood and to permit a fine-grain analysis which is ‘‘the cornerstone of much of ethology, [which] can reveal the plans that underlie hierarchically organized behavior’’ (Altmann, 1995, p. 8). A. The ‘‘Univariate’’ Approach There is quite a large literature on primate infant development both in the field and in captivity. Qualitative approaches describe when behaviors appear, how they change, and when some behaviors disappear (ChevalierSkolnikoff, 1974; Deputte, 1986a; Fragaszy, 1989; Adams-Curtis and Fragaszy, 1994). Most quantitative studies address the issue of behavioral development and present changes in frequency across time for selected behaviors, what Wohlwill (1973) called ‘‘developmental functions.’’ To emphasize the individual aspects of ontogeny, Hinde and collaborators presented medians and quartile ranges instead of means and standard deviations which have a more normative intent (Hinde and Spencer-Booth, 1967; Berman, 1980, 1982a,b). Both qualitative and quantitative approaches in behavioral development are used to define stages and eventually age classes. However, as Hinde (1971) pointed out, when setting criteria to define stages physical and social development should be dissociated because they are controlled by different factors that might change independently. In our gray-cheeked mangabey study we followed this advice and age classes, ‘‘stages,’’ were based only on physical criteria (Deputte, 1992). Each behavior was selected to represent a set of ontogenetic processes. The univariate approach mainly refers to the fact that these developmental functions are analyzed independently. The relations between different developmental functions were reconstructed a posteriori in a discussion section. Developmental functions describe gross activity, sleep–awake cycle lengths, postural development, development of perceptual capabilities (Deputte, 1985, 1986a; Fig. 7) or motor, gestural, and/or locomotor capabilities (Chalmers, 1972; Burton, 1972; Fragaszy, 1990a), and behaviors that might be involved in interactions, especially with the mother (Rhine and HendyNeely, 1978; Dolhinow and Murphy, 1982; Fragaszy, 1990b). In graycheeked mangabeys, the development of social looking behavior (looking

114

BERTRAND L. DEPUTTE

Fig. 7. Development of social looking behavior in nine gray-cheeked mangabey infants from birth until 42 months (social looking behavior ⫽ visual fixation on group members). Arrows indicate the end of the two developmental periods, from 0 to 4 months and from 4 to 19 months (see text). Developmental markers and a morphological developmental scale relate maturation and growth to behavioral/perceptual development (0–1 month, 1–3 months, 3–8 months, 8–12 months, and more than 1 year).

at a partner) showed three main periods. The first, which lasted from birth until Month 4, was characterized by a sharp increase in looking behavior. The second one, from Months 4 to 19, was characterized by a plateau of frequency of looking behaviors, followed by a fluctuating high level of visual social awareness (Fig. 7). These two phases were likely related to the neurophysiological development of vision, with the initial phase corresponding to the rapid development of perceptual capabilities such as visual acuity and contrast sensitivity (Teller et al., 1978; Boothe et al., 1980). The second phase likely corresponded to the later slow development of the neocortical structures involved in the cognitive processing of information (Gibson, 1981). When mother–infant interactions were considered, emphasis was placed on the important process of the infant’s increasing independence from the mother and the changes in related behavior such as contact behavior. Therefore, most often, the outcomes of locomotion (i.e., infant–mother distances) are analyzed instead of locomotor abilities per se. At 3 months

PRIMATE SOCIALIZATION REVISITED

115

captive infant mangabeys and pig-tailed macaques spend most of their time out of reach of their mother (Deputte, 1986a; Jensen et al., 1968) (Fig. 8), and in the wild infant baboons are at least 5 m from their mother at 5 months of age (Altmann, 1978). Increased infant–mother distance is likely correlated with modifications in other behaviors: In gray-cheeked mangabeys ventroventral contacts, huddling behaviors and embraces, disappear in male infants when they are at a distance from their mother that is greater than that from other partners (Fig. 9). These behaviors showed a dramatic decrease in female infants, but did not disappear as they had in male infants (Fig. 9). Incidentally, female infants did not remain at a distance from their mother greater than that from other partners (Fig. 8). In primates, independence from the mother means it is possible to interact with a variety of other kinds of partners (Burton, 1972; Berman, 1982a,b; Fairbanks, 1988). Infant development becomes mostly social, and social ontogeny proceeds through infant–partner interactions.

Fig. 8. Infant–mother distances during the first 2 years in nine gray-cheeked mangabey infants. The dashed lines indicate the average infant–partner (other than the mother) distance for female and male infants (for physical development, see legend to Fig. 7).

116

BERTRAND L. DEPUTTE

Fig. 9. Changes in huddling behavior in male and female gray-cheeked mangabey infants during the first 2 years. Huddling behavior includes embraces, ventroventral contacts, huddles, inverse huddles, and different types of embrace, antiparallel standing contacts. Arrows indicate either peaks or changes in huddling frequency. The changes are more or less synchronous between males and females but huddling disappears in males after month 15 (for physical development, see legend to Fig. 7).

The interactive aspect of an infant primate’s behavioral development is exemplified by the behaviors that it receives from selected partners including the mother (Hinde and Spencer-Booth, 1967; Hinde, 1971; Fairbanks, 1988). In the early stages of social ontogeny, interactions between infants and other partners are complementary (Hinde, 1976a; Wade, 1977): Partners and infants show different sets of behavior in their interactions. Along with physical development, interactions tend to become more reciprocal at least in certain dyads such as infant–other immatures. Actually, most of the behaviors selected to describe social development are involved in the regulation of interindividual distances, different forms of contact, approaches, and leaves. B. The Index Approach Although the interactive feature of social development is claimed, using the univariate approach one could not document such interactions, which could only be inferred from indirect evidence. The ‘‘index approach’’ was intended to take into account the interactive features of some aspects of social development. 1. Assessing the Nature of Social Participation in a Dyad Hinde and collaborators attempted to overcome the unsatisfactory univariate approach. They focused on mother–infant interactions and analyzed

PRIMATE SOCIALIZATION REVISITED

117

data to assess the respective roles of the mother and her infant in the dynamic of their relationships. They defined ‘‘indices’’ which aimed at representing the contributions of both partners to regulation of mother– infant distances, e.g., maintaining contact or proximity. Both an approach– leave index, the ‘‘proximity index’’ (Fig. 10; Hinde and Spencer-Booth, 1967; Hinde and Atkinson, 1970), and a making–breaking contact index, the ‘‘ventro-ventral contact index’’ (Hinde and White, 1974; also called ‘‘ventral contact index’’ in Berman, 1980), are based on the balance of behaviors having opposite consequences on infant–mother distance. These behaviors are initiated either by the infant or by the mother and the relative balance was assessed for each partner (Ai% ⫺ Li% and %Mki ⫺ %Bki, where, for instance, Ai% ⫽ Ai/Ai ⫹ Am, where Ai are the approaches initiated by the infant and Am are those initiated by the mother). Hence, the responsibility of each partner in achieving a certain outcome could be assessed. Hinde and White (1974) also assumed that, given the importance of interactive processes in the dynamics of the infant–mother relationship, it is the balance between reinforcement by ventro-ventral contact (%Mki) and the punishing effect by mother (%Bkm) that are crucial. To emphasize the interactive feature of the relationship, they mixed two components— one belonging to each of the partners. With the same reasoning, Berman (1980) logically devised a ‘‘body contact index’’ which measures the infant’s relative responsibility for maintaining body contact of any kind with the mother. All these indices focused on only one, although essential, aspect of a relationship. In addition, the fact that several indices seem to be needed to detail the developmental dynamic of the relationship suggests that more comprehensive approaches are necessary.

Fig. 10. Proximity index in rhesus monkeys, A%–L%. Circles indicate the median value and the shaded area indicates the range of index values from the set of eight infants (main group) (adapted from Hinde and Spencer-Booth, 1967).

118

BERTRAND L. DEPUTTE

Consequently, to proceed a step further, in line with Hinde’s indices, I proposed a more integrated index, the social investment index (SII), which aimed at assessing the role of each partner of a dyad in determining the overall nature of a relationship and not just one aspect of it (Deputte, 1983). This index helped make inferences from interactions to relationships. It is based on a ‘‘3 A’s’’ concept (affiliation, aggression, and avoidance) which constitutes the essence of social behavior and all relationships. The SII was the difference between the relative (positive or negative) amount of affiliative behaviors given by a focal subject, G%, and the same ratio from a given partner, R% (Fig. 11). In comparison to Hinde’s indices, the SII includes a much larger number of different behaviors (69), grouped in the 3 A categories, because it was intended to provide an integrated picture of an entire relationship (Deputte, 1983). Using this large number of behaviors actually emphasized the multidimensionality of social behavior, with many different behaviors having possibly the same meaning in terms of the nature of a relationship. Moreover, the large number of behaviors that is used makes it possible to apply this index to any kind of dyad in a group and not just the infant–mother dyad. The display of negative behaviors results in an increased distance between the dyad members. However, this increase could result from two different but complementary behaviors: Some relate to aggression (a subject intends to make its partner move away) and others to ‘‘spontaneous’’ avoidance (not preceded by aggressive behaviors). The SII ranged from 100% to ⫺100% (Fig. 11; Deputte, 1983). Within the 0–50% range, both partners invested positively in the relationship with the focal subject initiating more positive interactions than did its partner (Fig. 11; Deputte, 1983). The 50–100% range can be called the ‘‘overinvestment range’’ because the subject kept on giving positive behaviors while receiving a majority of negative behaviors, either aggressive or avoidance ones, from its partner (Fig. 11; Deputte, 1983). The SII yielded a single numeric value, but addi-

Fig. 11. The social investment index (SII) and the 3 A’s approach: (A) Computation, (B) interpretation, and (C) types of relationships. In C, the matrix is symmetrical. Along the diagonal, the cells with thick borders include totally symmetrical relationships. The first component of the relationships, O, A, AE, or E, in the lower half-matrix, is that of the subject. The shaded cells represent types of relationships that are symmetrical to those in the part below the diagonal. (In these cells the first component is that of the partner. The types in italics and lowercase letters, along the diagonal, are the most unstable ones, a-/a- and e-/e-; see text) (adapted from Deputte, 1995a).

PRIMATE SOCIALIZATION REVISITED

119

120

BERTRAND L. DEPUTTE

tional information related to this value was needed to provide an accurate interpretation (Figs. 11A and 11B). An important datum was the balance between aggressive and avoidance behaviors for each member of the dyad (Fig. 11A). This balance, plus the sign of G% and R%, yields 28 different types of social relationships (Fig. 11C). As with Hinde’s indices, the SII is intended to describe the dynamics of a relationship. Some types of relationships describe dominance/subordination relationships which are stable, e.g., A⫹/E⫹ (for both partners affilitative behaviors are more frequent than negative ones; when negative behaviors occur the subject gives mostly aggressive behaviors and his partner mostly avoiding ones). Other types of relationships are very unstable, such as A⫺/A⫺, where interactions between the two partners are characterized by a dominant frequency of aggressive behaviors on both sides, and E⫺ / E⫺, where the two partners avoid each other constantly. Some types of unstable relationships might constitute ‘‘transitory’’ relationships which eventually develop into more stable types such as A⫺ /E⫺, possibly merging into a stable A⫹ /E⫹ relationship (Fig. 11). The dynamics of infant–partner relationships are clearly exemplified by changes in the SII in two mother–infant and adult female–infant graycheeked mangabey dyads (the two female infants belong to two different groups; Fig. 12). Until the third month, mothers were responsible for the positive nature of the relationships with their infants (Fig. 12). From Month 6 or 7 on, the relationships changed as the mothers became more rejecting and negative behaviors surpassed the affiliative ones. Infants then ‘‘overinvested’’ in their relationships with their mothers, still showing more seeking of contact or contact-maintaining behaviors despite receiving rejecting behaviors from the mothers (Fig. 12), a situation Hinde and White (1974) described in macaque infant–mother dyads. Infant–mother relationships were clearly different from infant–adult female relationships. Although the general time courses were similar, adult females invested more positively in relationships than did the mothers (Fig. 12). Figure 12 presents two contrasting examples: For one infant, its relationship with the adult female suggested that this female acted as a substitute mother; the group contained only two adult females. The second infant was born in a group with three adult females, including its grandmother. The relationship, displayed in Fig. 12, concerned that between the infant and the lowest ranking adult female; the investment of the female in the positive relationship is important. As with all the possible interpretations of the approaches and leaves index (Hinde, 1977), the three dimensions of an integrating index, such as the SII (numeric value, sign, and nature of agonistic component), illustrate the difficulty in rendering a dynamic interactive process with only one

PRIMATE SOCIALIZATION REVISITED

121

Fig. 12. Ontogenetic changes in the social investment index (SII) in two infant–mother and infant–adult female dyads in gray-cheeked mangabeys. The type of relationship is associated with each index value (see text and Fig. 8). Physical and morphological changes are indicated with infants’ profiles above the time axis.

number or only one behavior or one individual. In addition, the SII remained at the dyadic level and did not allow one to consider the influences of other individuals on the infant–mother relationship. 2. Assessing the Comprehensive Nature of Behavioral Expression: Diversity Indices The indices previously presented address the issue of changes of dyadic relationships during ontogeny. They do not address specifically the two other processes of social ontogeny: development of a social repertoire and building a social network. Diversity indices can be used to address these processes. Altmann (1965) derived diversity or uncertainty indices from Shannon’s information theory. A diversity of behavior index actually describes, with one comprehensive number, the ‘‘rank–frequency’’ curves or ‘‘behavioral abundance distributions’’ (Altmann, 1965; Fagen and Gold-

122

BERTRAND L. DEPUTTE

man, 1977). The index Altmann (1965) computed was intended to measure with one comprehensive number how much the relative frequencies of the behavioral patterns of a social repertoire depart from equiprobability. The index summarizes the rank–frequency curves or behavioral abundance distributions (Altmann, 1965; Fagen and Goldman, 1977). From this we designed indices for the diversity of behaviors, the diversity of interaction, and the diversity of attention (Deputte, 1985, 1986a; Deputte and Quris, 1996, 1997). In practice, these indices are computed from behavior ⫻ partner matrices (Fig. 13). The diversity of behaviors is intended to deal with the ontogenetic process of building a social repertoire. It describes either how the social behavioral repertoire of an infant develops in reference to an a priori defined operational repertoire or what this infant receives from its partner (i.e., its social experience). The diversity of behaviors approach requires the use of a large behavioral repertoire, allowing fine-grain analysis of behaviors at every stage of the infant’s development from birth to adulthood (Altmann, 1962, 1995; Bernstein and Williams, 1986). A high relative diversity of behaviors means that there is a rich and complex individual repertoire that includes

Fig. 13. Computation of diversity indices from behavior ⫻ partner matrices.

PRIMATE SOCIALIZATION REVISITED

123

most defined behavioral units, and that a high proportion of these units were observed with a comparable high frequency. In contrast, a low relative diversity index means a rich but simple repertoire which may include a large number of units, but a few units, recurring very often, dominate the behavioral expression. One might expect that the number of behavior patterns displayed by an infant gradually increases with physical and cognitive maturation and that a bulk of highly recurrent behaviors will emerge as infants mature. Translated into diversity of behavior values, these assumptions will be supported by a progressive increase of the diversity of behavior index. In the development of the behavioral expression in gray-cheeked mangabeys, three different phases occur (Fig. 14): During the first 3 months, there is a rapid increase in behavioral diversity corresponding to a gradual increase in the variety of behaviors displayed, with many behaviors, expressed at a low frequency (Deputte, 1986a). Then, up to 1 year, there is a plateau at which the number of different behavioral units expressed progressively stopped increasing and only the frequency of the behaviors changed, with many behaviors being expressed with the same frequency (Deputte, 1986a). The last phase corresponded to a differentiation of the behavioral expression reflecting the uniqueness of each subject’s interactions. The diversity of behaviors received by the infants actually described the nature of the infant’s experience. The behavioral expression of the infant proved to be more or less independent of this diversity because for some individuals the diversity of behaviors given was higher than the diversity of behaviors they received; the opposite trend existing for other subjects (Fig. 14). Behavioral expression is not restricted to interaction of a single dyad, such as the mother–infant one, but rather spread over an entire social network (Capitanio, 1985). The complexity of interaction patterns can be assessed using a diversity of interactions index computed in the same way as the diversity of behaviors index. The diversity of interactions index is intended to deal with the ontogenetic process of building a social network: A complex network of interactions will be the consequence of a balanced level of interactions with all available partners, exemplified by a high relative diversity of interactions. In contrast, a simple network will be assessed if an individual interacts most of the time with only one or a few partners exemplified by a low relative diversity of interactions. The diversity of interactions index is independent of the number of available partners but quite sensitive to the number of partners contacted: The value of the index increases with the number of partners enlisted by the subject even though the subject interacts mostly with only one or two partners (increase of the minimal value; Fig. 15).

124

BERTRAND L. DEPUTTE

Fig. 14. Ontogenetic changes in diversity of behaviors in five gray-cheeked mangabey infants whether they initiate interactions (givers; top) or they are targets of interactions (receivers; bottom). Arrows indicate different phases in ontogenetic changes of this index.

PRIMATE SOCIALIZATION REVISITED

125

Fig. 15. Theoretical model of changes in values of diversity of interactions according to the number of partners with which the subjects interact and depending on the group size (which includes the subject). The different theoretical values are obtained from different distributions of the same amount of interactions between different numbers of partners (x axis). Each point represents a diversity of interaction obtained for a certain distribution of the overall behavioral frequency between a certain number of partners within a group of a certain size. For each combination ‘‘number of partners ⫹ group size,’’ the highest values are obtained for the more balanced distribution (50% for each partner when two partners are involved and 25% each when four partners are involved; in these two cases if two or four partners are all the partners the subject could contact, the relative diversity of interaction will be 100), and the lowest values are obtained for the most unbalanced distributions (90 and 10% for each partner, respectively, when only two partners are involved; 70, 10, 10, and 10% when four partners are involved). For each group size, the diversity index increases with the number of partners involved.

The development of a social network is evidenced by the increase in the diversity of interactions during the first 5 months of ontogeny (Fig. 16). Values reached at the plateau suggest that infants interact with most of their available partners no matter the size and complexity of the social environment (Fig. 16). The diversity of interactions initiated by partners toward the infant indicated a high initial attraction toward the infant (high

126

BERTRAND L. DEPUTTE

Fig. 16. Ontogenetic changes in diversity of interactions in five gray-cheeked mangabey infants whether they initiated interactions (givers; top) or were targets of interactions (receivers; bottom). Arrows indicate the beginning of a plateau after a rapid initial increase, especially when infants initiated interactions (top).

diversity) followed by marked preferences after the fifth month (decreased diversity; Fig. 16). In general, infants interacted with a large variety of partners, whereas fewer partners initiated interactions with infants. A diversity of attention index can be computed by taking into account only one class of behaviors—the visual behaviors (Fig. 13). It is intended

PRIMATE SOCIALIZATION REVISITED

127

to describe how an infant distributes his looks toward his partners and conversely how much an infant is the focus of attention from the group members during ontogeny. In contrast to the diversity of interactions index, the diversity of attention index was high for visual attention both given and received by the infants without showing any clear developmental trends (Deputte, 1986a); thus, in captivity, visual attention is a two-way process that generally includes all partners and that is much less susceptible to focalization or preferences than other kinds of ‘‘true’’ interactions. In addition, the diversity of attention index showed much more individual variation than the diversity of interactions index. Diversity indices by themselves do not thoroughly describe all the processes they are intended to describe. Neither diversity of behaviors nor diversity of interactions on their own say anything about the nature of the behaviors exchanged during the interactions and how they change during ontogeny. Diversity of interactions does say determine the privileged partners. Also, other indices failed to describe the global complexity of the process. Although the index approach is an improvement to describe specific processes in social ontogeny, it is not sufficient to disentangle the intricacies of all the processes. C. The Multivariate Approach Given the multidimensionality and complexity of social ontogeny, one might assume than only multivariate statistics might be able to disentangle the ‘‘complex and dynamic interplay of forces and events that go into any interaction’’ (Thoman et al., 1979, p. 305). Even in the dyadic approach a single factor proved to be too simplistic to deal with the interactive features of social ontogeny. The multivariate approach actually encompasses all the univariate approaches while preserving the variety of behavioral expression and clarifng the degrees of dependence between all the behavioral variables. Accordingly, several authors have claimed that only a multivariate approach can provide an answer to the multidimensional nature of social behaviors and to the meta-interactive character of social ontogeny: Even in 1970, Crook (1970b, p. xxii) asserted that ‘‘Multifactorial studies of processes . . . make clear the complexity . . . of the phenomenon with which the ethologist deals.’’ The multivariate approach is the tool to put into practice according to Hinde (1966), who notes that once the complexity is recognized, it remains necessary to abstract and simplify in order to analyze the process. Berman (1982a) contended that in analyzing the development of social networks and differentiation, multivariate analyses are ideal compared to separate tests. The data structure of complex dynamical systems often prevented the use of the multivariate techniques that were available at a

128

BERTRAND L. DEPUTTE

given time (Bateson, 1991). In addition, multivariate approaches to social behaviors most often concerned the role concept (Bramblett, 1973; Fedigan, 1972, 1976; Fairbanks et al., 1978; Fairbanks and McGuire, 1979) and rarely addressed the issue of social ontogeny (Capitanio and Reite, 1984). The following discussion is intended to prove that efficient, diverse multivariate techniques are available and helpful to reduce the intrinsic complexity of social ontogeny processes to manageable hypotheses. Multivariate approaches allow us to take advantage of fine-grained analyses which require the definition of a large behavioral repertoire. These fine-grained analyses, applied to social ontogeny, are especially suitable to describe individual experiences because the early experiences might have important consequences on later social interactions. Defining a large behavioral repertoire represents what Bernstein and Williams (1986) described as ‘‘the synthetic approach’’ to the study of social organization. They stressed that this procedure, which they called ‘‘a simple quantification,’’ may prove very powerful. They contrast the synthetic approach with the analytic approach, which started with the description of a set of predictable patterns of interactions (roles) ‘‘enforced by the society as a whole.’’ These authors gave the example of an infant who ‘‘should direct a different set of responses to its own mother than to other adult females who may also be mothers’’ (Bernstein and Williams, 1986, p. 202). This is exactly the outcome of processing the data included in the individuals ⫻ behaviors matrix by multivariate factorial analysis. Therefore, the multivariate approach, applied to the study of social ontogeny, is where synthetic and analytic approaches applied to social organization meet. The existence of a meeting point emphasizes the influence of social factors, through interactions, on behavioral development during social ontogeny. The multivariate approach to social ontogeny investigates whether ontogenetic behavioral differentiation is controlled mainly by maturational factors or is shaped by partner–infant interactions. However, in the latter case the differentiation, though partner related, might change during infants’ ontogeny in interaction with maturational parameters. Although behavioral differentiation in infant–partner interactions was more or less explicitly assessed by previous students of infant primate development (Harlow and Harlow, 1965; Hinde and Spencer-Booth, 1967 Hinde, 1969, 1971), each kind of infant–partner interaction was treated separately and has not been illustrated using a behavioral profile approach. Previous multivariate analyses used different techniques such as principal components factor analysis (Fairbanks et al., 1978), multiple discriminant analysis (Fedigan, 1976), and stepwise multiple regression (Capitanio and Reite, 1984). The following examples present two other multivariate techniques: correspondence factorial analysis (CFA) and principal component analysis with instrumental variables (PCAIV) (Deputte, 1986a; Deputte

PRIMATE SOCIALIZATION REVISITED

129

and Goy, 1991; Deputte and Quris, 1996, 1997). Neither technique requires special assumptions about the data sets. They both analyze observations (individuals) ⫻ variables (behaviors) matrices. One requirement of the analysis (a commonsense requirement) is that the number of columns (variables) does not exceed that of rows (observations). In CFA, data are frequencies of occurence (counts), and they are measurements, counts, or ratios in PCAIV. In that case, relevant transformations, according to the nature of the data, are performed in order to normalize the data set (Sokal and Rohlf, 1969). 1. The Multivariate Descriptive Approach The multivariate descriptive technique used is CFA, which basically provides a description of the latent structure underlying the data (Benzecri et al., 1973; Ter Braak, 1987). CFA explored, in a descriptive way, the contingencies between the two sets of variables, for example, behaviors and subject–partner–age (Fig. 17). These contingencies were computed in a multidimensional space (the number of the dimensions of the space is

Fig. 17. Completion of a behaviors ⫻ partners matrix to be submitted to a multivariate technique (e.g, canonical correspondence analysis and factorial correspondence analysis). BU, behavioral unit; Inf1-M-P1; as a row identifier, indicates interactions of Infant No. 1 (Inf1) with its mother (M) during the period 1 (P1) of its ontogeny. In the different cells of this row (different columns) are indicated the frequencies per sample (F/s) of the corresponding behavioral units. M, mother; F, father; A.F., adult female; Im, immature partner. ⌺C and ⌺R mean respectively the sum per column and the sum per row of the different frequencies. The behavioral repertoire includes 156 mutually exclusive behavioral units.

PRIMATE SOCIALIZATION REVISITED

131

given by the number of columns). CFA extracted several factors which best explained the underlying structure of the data and determined the best fit representation of the variables (the columns) and the observations (the rows). In contrast with the multivariate techniques mentioned previously, CFA is based on geometrical principles yielding a graphical representation of the results instead of tables of numbers. The factors are actually represented by the axes of the graphical representations which maximized the variance extracted from the data set. The results were presented as scatterplots of both the variables and the observations on a plane defined by the factors, generally the most important ones—factors 1 and 2 (Fig. 18; Deputte and Goy, 1991; Deputte and Quris, 1996, 1997). Studies of social ontogeny in rhesus macaques and gray-cheeked mangabeys will serve to exemplify the multivariate descriptive approach. In the basic matrix of each study, the columns are the mutually exclusive behavioral units and the rows are the combination of infant–partner dyads with infant’s age or period of ontogeny (Fig. 17). The cells of the data matrix contain the frequencies per unit of time (e.g., frequencies per 5-min focal sample) of the different behaviors observed during interactions between infants and the different classes of partners during defined periods of ontogeny (Fig. 17). In the study of the social ontogeny of rhesus macaques, CFA discriminated three behavioral clusters which were characteristic of interactions between infants and the three different classes of partners (mother, other females, and

Fig. 18. Behavioral profiles of infant–partner interactions in rhesus macaques (20 infant– mother pairs): A, infant-to-partner (infants initiate); B, partner-to-infant (partners initiate). Graphical representation of the results of correspondence factorial analysis is shown; projection on the plane of the first two factors: the principal clusters are identified within boxes. Behaviors are indicated in italics. Partners are identified by a letter in a circle at the center of gravity of the points representing each month-dyads (not represented here). M, mother; F, adult female; i, infants (modified from Deputte and Goy, 1991). In both A and B, factor 1 indicates the duration of contacts, and factor 2 indicates the quality of the relationships. How to read the graph: As a rule of thumb, the further from the center of the graph, the more important the variables. The axes of the graph (a maximum of three can be represented simultaneously) feature the factors that explain the greatest amount of variation of the data sets. The names given to the factors are given after the data points (variables) which generally are close to one axis and further from the center. The correlation coefficients between the variables and the factors that might be presented in tables are represented here as proportional to the cosine of the projection angle of the variable on the plane. This explains why the most important variables are not near the center of the graph (a high correlation coefficient results in an acute angle). In addition, the proximity between two points (from one set of data or from both sets) indicates an association, leading to the individualization of clusters of points. Here, the proximity between behaviors and partners indicates that these behaviors characterize the infant–partner interactions.

132

BERTRAND L. DEPUTTE

other infants) whether the infants initiated interactions or were the recipients (Fig. 18). The results showed that relationships are multiplex (Hinde, 1976a)—either symmetrical, as in infant–infant interactions, or complementary, such as in mother–infant interactions: Infants looked for close contact, for example, and in return, received protection and possibly contact-breaking behaviors. Infant–other female interactions were characterized by deference by infants (presenting and mere proximity) and coercion (punishing and

Fig. 19. A. Ontogenetic changes in behavioral profiles of infant-to-partners interactions in nine gray-cheeked mangabey infants; graphical representation of the results of correspondence factorial analysis. Projection on the plane of the first two factors: the principal behavioral clusters are mentioned in capital italics. The different age–sex classes of the partner are indicated in capital letters. M, mother; AF, adult female; F, father; AM, adult male; O, older immature; Y, younger immature. The numbers associated with the classes of partner correspond to the period of infant’s ontogeny: 1, 0 to 1 month; 2, 2 or 3 months; 3, 4– 8 months; 4, 8–12 months; 5, 12–18 months; 6, 18–24 months; 7, 24–30 months. Subject–younger immatures interactions only start when subjects are 4 months old (period 3). The arrows connect the same partner–infant dyads for successive periods of ontogeny. When infants initiate interactions with partners the main factor is the duration of contacts (horizontal axis, axis 1; see the legend to Fig. 18). (B) Ontogenetic changes in behavioral profiles of partners-toinfant interactions in nine gray-cheeked mangabey infants. When partners initiated interactions with infants, the main factor (horizontal axis) was the quality of interactions, either dynamic or static (for details see the legend to Fig. 18).

PRIMATE SOCIALIZATION REVISITED

133

threats) by adult females while both partners kept an eye, or both, on each other (Fig. 18). This example described relationships over a 7 month time span. However, these relationships were essentially dynamic. In gray-cheeked mangabeys, infant–partner relationships were shown to change over a 3-year period (Fig. 19). After the behavioral clusters were determined, as in the rhesus study, infant–partner interactions for each age period were placed within this behavioral space (Fig. 19). Infant 씮 mother interactions changed much more dramatically and rapidly than did any other kind of interaction. Infant 씮 mother interactions changed from early search for close contact to regulation of proximity. To some extent, infant 씮 adult female interactions followed a similar trend. However, the trajectory of these interactions indicated that adult females might serve as mother substitutes during early periods of ontogeny. Infant–adult fe-

Fig. 19. (Continued).

134

BERTRAND L. DEPUTTE

male interactions were later characterized by infants challenging adult females. Infant 씮 adult male and infant 씮 father interaction profiles were mainly characterized by visual contacts suggesting that infants were cautious in interacting with adult males. During ontogeny, behavioral diversity seemed to be restricted to infant–infant relationships and these infant– infant interactions became the major direct source of not yet determined social experiences. However, as Rosenblum and Coe (1977) pointed out, infrequent interactions might have as important, or even more important, social consequences as frequent ones. This might be the case for infant– adult male (including father) relationships; infants’ visual contacts and rare approaches could be considered as consistent responses to rare punishment and threats (Fig. 19). Ontogenetic changes in both infant 씮 partner and partner 씮 infant interaction profiles illustrate the concepts of ‘‘multiplex’’ relationships, symmetry, complementarity, and compensation formulated by Hinde (1975, 1976a; see also Wade, 1977). The graphs presented so far are all global and synthetic. It should be remembered that the original graphs included individualized dyads for the different periods of infants’ ontogeny, therefore yielding true ontogenetic trajectories. The graphical representation of the behavioral profiles visualizes the properties of the relationships and helps in defining the roles of ‘‘socializing agents’’ played as recipients of infants’ social trials and as actors in pointing out infants’ errors or successes through punishments or social rewards. Therefore, applied to social ontogeny, the representation that the CFA technique produced permits one to tackle more specifically the issue of a partner-related behavioral differentiation. It also illustrates that behavioral profiles are the consequences of predictable behavioral contents of interactions. This predictability is a consequence of established social networks and social organization which generally tend to reduce the behavioral variability of the group members. However, this descriptive approach does not allow us to demonstrate how different environments (both social and nonsocial) or an infant’s intrinsic variables (such as its sex) might influence the two ontogenetic processes, building a social repertoire and building a network of relationships, which correspond to what Capitanio (1985) defined as acquiring ‘‘a behavioral competence’’ and ‘‘a social competence,’’ respectively. 2. The Multivariate Analytic Approach a. Influence of the Social Milieus on Socialization Processes ‘‘I mean,’’ [Alice] said, ‘‘that one can’t help growing older.’’ ‘‘One can’t, perhaps,’’ said Humpty Dumpty, ‘‘but two can.’’ —Lewis Carroll (Through the Looking Glass)

The influence of the physical environment on infant–mother relationships or other aspects of infant social development has been demonstrated in

PRIMATE SOCIALIZATION REVISITED

135

both field and captive settings (Lee, 1984; Johnson and Southwick, 1987; Rosenblum, 1982). In addition, supporting Hinde’s social network concept, it often has been suggested that group size, group structure, and group organization might influence the infant’s social ontogeny directly or through the constraints that the infant’s mother endures (Mason, 1965; Hinde, 1971; Baldwin and Baldwin, 1979; Altmann and Altmann, 1979; Poirier, 1982). Altmann (1984, p. 15) suggested that infants received either social stress or social support depending mainly on social organization and that relationships among adult females ‘‘are important determinants of a mother’s social experience.’’ These assessments were less often supported by empirical data. Experimental demonstrations came from manipulations of the social environment (Suomi, 1985) or by comparing species with contrasting social organizations and social structures (Rosenblum and Alpert, 1977; Rosenblum and Coe, 1977; Rosenblum and Plimpton, 1979). However, as was the case for the study of social development, these studies most often addressed questions of influences of social milieus on selected aspects of development rather than the influence of social milieu variables on the social ontogeny of the developing individuals. It should be remembered that, according to the social ontogeny framework, demographic influences mean availability of a diversity of social phenotypes, with their individual and social features. The social environment variable actually includes several dimensions, such as the size of the group, its demographic structure (mating structure and the presence of a certain number of immatures—individual phenotypes), its social organization (social networks and nature of social relationships—social phenotypes), and the ‘‘personal history’’ of each group member as well as individual behavioral flexibility (both individual and social phenotypes). Therefore, it seems justified to refer to a social environment variable as a ‘‘social complexity’’ variable. In a given species, it can be hypothesized that large and complex groups, including a large diversity of social phenotypes, would provide a young primate with greater opportunities to acquire a large behavioral repertoire rapidly or, possibly, a larger repertoire than could be acquired by young in smaller and less diverse groups. However, an infant will acquire a greater and/or more complex repertoire only if it actually interacts with a large variety of partners having different behavioral profiles. In large groups, constraints of social organization could shrink the number of available partners to a small number similar to that in small, less diverse groups. The question of the influence of social milieu variables on social ontogeny was posed in an experimental study in gray-cheeked mangabeys and was addressed using a multivariate analytical technique called the PCAIV (Ter Braak, 1987; Sabatier et al., 1989; Lebreton et al., 1991; Deputte and Quris, 1996). This multivariate technique combined the properties of the classical MANOVA and those of PCA. It used the technique of regression. Thus,

136

BERTRAND L. DEPUTTE

PCAIV allowed us to consider simultaneously several dependent variables, to analyze their correlations, and to compute their expected values in relation to a model which includes several independent variables (Deputte and Quris, 1996). The PCAIV technique required two types of matrix: (i) the classical dependent variable matrix (subject ⫻ behaviors or subject ⫻ other variables; matrix [D] in Fig. 20) and (ii) an independent variable matrix (matrix [M] in Fig. 20). In the latter matrix, the columns represented the several modalities that each independent variable could take. For each subject (row), the independent variables were transformed into ‘‘dummy variables’’ (zeros or ones) according to the subject (Fig. 20). The principle of the analysis consists in computing the expected values of the data matrix [D] in relation to the model as expressed in the columns of the [M] (model) matrix (Fig. 20; Deputte and Quris, 1996). In addition, in the PCAIV technique, a multivariate Fisher’s F (explained variance/nonexplained variance) can be computed and the value of this multivariate F can be tested using a Monte Carlo technique (Ter Braak, 1992). Therefore, PCAIV added the power of a statistical test to the visual representation of the interrelations between and within the two types of variables represented as vectors (see Fig. 21). The experimental study of gray-cheeked mangabeys reported on the first 18 months of subjects’ social ontogeny. Behavior–partner matrices were constructed for each month and each subject (85 months–subjects; Fig. 20). Because we were dealing with the two essential processes of social ontogeny, only the following six dependent variables were extracted from each month–subject matrix instead of dealing with the whole behavioral repertoire as in the descriptive approach (matrix [D] in Fig. 20): 1. The size of the social repertoire: The number of different behavioral units observed, at least once, during a given subject–month, which reflects

Fig. 20. Completion of behaviors ⫻ partners matrices to be submitted to multivariate analyses such as principal component analysis with instrumental variables. [B], behavior matrix; [V], visual behavior matrix; [T], total behavior matrix irrespective of the distinction between visual behavior and other behaviors; [D], dependent variable matrix; [M], independent variable matrix. Cell contents are the frequencies per 5-min sample over a given period of time. The dependent variables in [D] are extracted from the [B], [V], or [T] matrices. SR, size of repertoire: number of different behavioral units observed at least once during a given sampling period; FB, total frequency of behaviors; B, from the [B] matrix; Db, diversity of behaviors; FL, total frequency of looks; V, from the [V] matrix; Di, diversity of interactions; Da, diversity of attention. The independent variables are noted as 0 or 1 in the independent variable matrix depending on the subject and its related features: for example on the first line, the subject is a male No. 1 aged from 1 to 6 months, P1, and living in the social environment 1 (see Fig. 13 for details on computation of the diversity indices) [modified from Deputte (1995) and Deputte and Quris (1996)].

PRIMATE SOCIALIZATION REVISITED

137

138

BERTRAND L. DEPUTTE

PRIMATE SOCIALIZATION REVISITED

139

the overall variety of an infant’s behavioral expression (or the variety of the social stimulation received by an infant; Deputte and Quris, 1996). 2. The total frequency of behaviors: This is an expression of the density of social interactions. 3. The diversity of behavior: This refers to the structure of the observed behavioral repertoire (as discussed previously). 4. The total frequency of looks: This refers to the global visual attention. It provides a subject with information on its social environment and is used in interindividual distance regulation (Rowell, and Olson, 1983; C. Blois-Heulin and B. Deputte, manuscript in preparation).

Fig. 21. Influence of social environments on gray-cheeked mangabey infants’ social ontogeny: graphical display of the results of the three-variable principle components analysis with instrumental variables (PCAIV). A, Infants as initiators; B, infants as receivers. How to read this graphical representation: Interpretation of the graphical displays used the same rules of thumb as those used in the CFA (see Fig. 18). However, the dependent variables here are represented as vectors. A dashed circle indicates the ‘‘correlation circle’’ of radius ⫽ 1. Within this circle, the length of the vectors indicates how much a dependent variable is explained by the model: The longer the vector, the better the corresponding dependent variable is explained by the model. The correlations between the dependent variables are represented by the cosine of the angle formed by the corresponding vectors: The more acute the angle, the more the dependent variables are positively correlated. Points, or cluster of points, in the direction of a vector, the lengths of which are near 1, are positively correlated with this vector variable. Influence of social environments in mangabey infants’ social ontogeny: The Figure provides a representation of the projections on the plane of the first two factors of the PCAIV, of the six dependent variables, shown as vectors with the name of the dependent variable provided, and of the centers of gravity of the clusters of points featuring the individuals grouped following the model variables (these centers of gravity are represented by a female or male symbol preceded by a roman number, indicating the age period: I, 0–6; II, 7–12; and III, 13–18 months). For the sake of clarity, clusters of points, representing an independent variable, are generally represented by their center of gravity. The centers of gravity for individuals belonging to the same social structure are enclosed in an ellipse with a distinctive shading. The three social environments are partly described within the box attached to the ellipses. On each axis the percentage of extracted variance is indicated. The age arrow on the left indicates the influence of the age variable. Age is positively correlated with the size of the repertoire. The social environment I is positively correlated with the diversity of behaviors. In addition, among the dependent variables, the size of the repertoire, the diversity of interactions, and the frequency of looks are highly positively correlated. (B) The dashed lines, identified by sex symbols within boxes, isolate points featuring male infants (top) from those featuring female infants (bottom) (modified from Deputte and Quris, 1996). Partnersto-female infant interactions are characterized by the high values of at least three dependent variables (frequency of looks, diversity of interactions, and size of the repertoire), whereas partners-to-male interactions are characterized by low values of these variables. Therefore, most of the partners in a group interacted with female infants in displaying a large variety of behaviors and having many visual contacts, prior to or during interactions.

140

BERTRAND L. DEPUTTE

5. The diversity of interaction as discussed previously. 6. The diversity of attention as discussed previously. The first three variables described more specifically the acquisition of a social behavioral repertoire and the last three the development of a social network. The subjects were eight of the nine infants born within three different social environments (I–III; groups differed in size and composition; as discussed previously). Three developmental periods, P1–P3, were considered: from 1 to 6 months, from 7 to 12 months, and from 13 to 18 months, respectively. Therefore, the model matrix, [M], included four independent variables—the social environment, the infants’ age, the infants’ sex, and the infants’ individuality—used to explain the six dependent variables (Fig. 20; for further details, see Deputte and Quris, 1996). When infants initiated interactions, age had the strongest effect. Infants did not begin to develop their own social network and to differentiate their social behavior according to their partners before 6 months. The highly positive correlation between a high level of visual awareness and an evenly distributed amount of interactions suggested the importance of processing visual information before and during interactions alongside social development (Fig. 21A). The three social environments were clearly discriminated (Fig. 21A). The simplest social environment included a mated pair, a young subadult male, and one and, later, two male infants (social environment I; Fig. 21A). This social milieu nevertheless provided the highest diversity of behaviors associated with a low diversity of interactions, suggesting that most interactions occurred between immature males and were playful. In social environment II, which was intermediate in size and complexity, infants were characterized by a large repertoire, interactions were evenly distributed between partners, and there was a high level of visual attention, suggesting that social organization was not highly constraining (high diversity of interactions) and likely confirming that infants’ behavioral differentiation was partner related (large repertoire and high diversity of interactions; Fig. 20). The processes of acquiring a behavioral repertoire and building up a social network seemed to be similar in female and male infants. When partners initiated interactions with infants, the social environment had the primary influence on the two socializing processes regardless of the infants’ age. Social environment I is even more clearly discriminated than when infants initiated interactions; the larger the group, the more diverse the interactions in which infants are involved and the greater the visual awareness of the partners toward the infants. The high selectivity of interactions in the smallest social group is likely a consequence of the playful nature of interactions between the immature males, which involved a high degree of repetition and mutual attraction (for captive and free-

PRIMATE SOCIALIZATION REVISITED

141

ranging rhesus macaques, see Goldfoot et al., 1984; Goy et al., 1988a; Deputte and Goy, 1991; Berman, 1982b). Partners seem to treat female and male infants differently. They provide a richer repertoire of social stimulation (size of repertoire) to female infants than to male infants, and within a group more partners contact infant females than infant males (diversity of interactions). b. Influence of Infant’s Sex in Social Ontogeny i. Defining behavioral ‘‘diposotism.’’ The design of the mangabey study, constrained by natural mangabey reproductive parameters (Deputte, 1991), did not permit us to study the issue of the influence of infant’s sex on its ontogenetic behavioral differentiation. Many behavioral differences between male and female primates have been described (Fedigan, 1982; Deputte, 1995a). However, documenting differences in male and female behaviors is not sufficient to disentangle the relative contribution of environmental and genetic inputs that could be identified only by means of experiments (Bernstein, 1978). Observed behavioral differences between sexes are often referred to as sex behavioral dimorphism in reference to sexual dimorphism applied to weight, size, color, or other physical features. However, concerning behaviors, two sets of phenomena could be observed: The homologous behavioral patterns displayed by females are different from those of the males (e.g., vocal behavior). In this case, the term sexual dimorphism is appropriate (Deputte and Leclerc-Cassan, 1981). In contrast, many other differences between the sexes are related to frequencies of behaviors, the best example being grooming behavior, which is much more frequent in females than in males (Goy et al., 1988b). In this case, the patterns of behavior are the same, only their frequency is dependent on the sex of the performer. Thus, to be more specific about the nature of these differences, it is more appropriate to use the term ‘‘diposotism’’ (from the Greek Posos, ‘‘in what quantity,’’ and Di, ‘‘two’’). Diposotism thus refers specifically to differences, between males and females, in frequencies of behaviors, the patterns of which are otherwise homologous. During ontogeny the genome determines the hormonal environment, yielding a sexual individual phenotype (Fig. 4). However, after birth, the social environment has a paramount influence on the social ontogeny. Thus, is the social phenotype sexually organized? In other words, does the sexual behavioral diposotism fit the sexual dimorphism? The alternative hypothesis is that, in contrast to the morphological phenotype, the social phenotype is nonsexually differentiated since social behavior develops through interactions with partners from both sexes; the social phenotype would have a bipotentiality (Deputte, 1995a). This would partly explain the dominance of a reversible, context-dependent, diposotism over a fixed behavioral dimorphism.

142

BERTRAND L. DEPUTTE

ii. Development of behavioral sex differentiation in rhesus macaque infants. The development of a possible behavioral sex differentiation has been tested in rhesus monkey infants (Deputte and Goy 1991; Goy and Deputte, 1996). The prenatal influences of hormones on later behavioral development has been demonstrated in many studies (Goldfoot et al., 1984; Goy and Phoenix, 1972; Goy and Resko, 1972; Goy and Robinson, 1982; Goy et al., 1988a,b). For example, virilized females display some behaviors, such as play and mounting, with the same frequency as those displayed by normal males or at least much more (high diposotism) than males castrated at birth and normal females. Isosexual rearing studies have emphasized the importance of an adequate companion for the development of some diposotic behaviors: Females reared in an isosexual environment mounted more and presented less than females reared in an environment including peers of both sexes, whereas, symmetrically, isosexual males mounted less and presented more than males reared with both females and males (Goldfoot et al., 1984). A study that investigated the development of behavioral sex differentiation was completed using 20 rhesus infants (8 females, 8 males, and 4 virilized females whose mothers has been injected with a low total dose of a virilizing hormone, DES) (Goy and Deputte, 1996). Infants lived in four groups of similar composition (five mother–infant pairs). For this analysis the development of infants followed from Months 1 to 6 was presented. As in the mangabey study, PCAIV was used, with the same dependent variables. Partners–behaviors data matrices were constructed (see Fig. 17). The rows of both dependent- and independent-variable matrices identified a combination of the individual infant, its age, its sex, and the partner with which it interacted Deputte and Goy, 1991; Deputte and Quris, 1997). The model included four independent variables infant’s sex, infant’s age, social group, and rank of the subject’s mother (Fig. 22). The first two variables

Fig. 22. Influence of sex on social ontogeny of rhesus macaque infants: graphical display of the results of a four variable principle components analysis with instrumental variables (PCAIV). (A) Infants as initiators. Representation of the projections on the plane of the first two factors of the PCAIV of the six dependent variables. The dependent variables are represented as vectors (see legend to Fig. 21 for details). The name of the dependent variable is mentioned in a box at the extremity of the corresponding vector. The independent variables are represented at the corresponding centers of gravity (bold capitals for ‘‘group’’; Female, with a DES (D) female; or male symbol for the sex and age periods within ellipses; Rk, ‘‘mother’s rank’’). On each axis, the percentage of explained variation is indicated, and the name of the corresponding factor is indicated in capital italics. (B) Infants as receivers: (modified from Deputte and Quris, 1997). The names of the factors are indicated (principal horizontal axis, factor 1; A, age and social environment; B, social environment only; vertical axis, factor 2; A, sex; B, social organization.

PRIMATE SOCIALIZATION REVISITED

143

144

BERTRAND L. DEPUTTE

were intrinsic variables, and the last two were variables of social organization. Maternal rank is actually one attribute of the social phenotype—a component of its relationships. When infants were initiators, the four variables of the model significantly explained the variation of the variables that describe the two socialization processes, building a social repertoire and setting a social network. The joint influence of the model variables highlights the fact that social ontogeny is dependent on complex interactions between the genome and the environment, with the nature of partners being the most influential variable. The fact that, as in mangabeys, there existed a high positive correlation between a large behavioral repertoire, a high level of visual attention, and evenly distributed interactions (high diversity of interactions) emphasizes the epigenetic nature of social ontogeny in primates (Fig. 22A). Age and sex, intrinsic-genetic variables, have confounding effects: The oldest individuals, especially males and DES females, tended to use their behavioral repertoire in a more evenly distributed way than did females and/or younger individuals (Fig. 22A; Deputte and Quris, 1997). In contrast, when partners were receivers only the social organization variables ‘‘social group’’ and ‘‘mother’s rank’’ had a significant effect (Deputte and Quris, 1997), with social group having the strongest effect [social environment as the principal factor (horizontal axis) and social organization featured by the rank variable, RK, as the second factor (vertical axis); Fig. 22B]. These social organization variables mainly influenced both the diversity of interactions and the diversity of attention, which were greater in interactions initiated by lower ranking individuals with infants. This has been described as mothering styles, ‘‘restrictive,’’ or ‘‘laissez-faire’’ (Altmann, 1980; Fairbanks and McGuire, 1987). Fewer partners interacted with infants of low-ranking mothers than with infants of high-ranking mothers, a phenomenon described in other species (Rowell et al., 1968; White and Hinde, 1975; Berman, 1978; Johnson et al., 1979; Altmann, 1980). When partners were receivers, neither infants’ age nor infants’ sex influenced the way partners contacted infants. However, there was a positive correlation between a large behavioral variety, a high level of visual awareness, and an evenly distributed amount of interactions (high diversity). This suggests that when many different partners contacted infants, they did so by displaying many different behaviors, thus providing infants with a large variety of experiences. The fact that infants’ sex was only influential when infants were behavior donors is nevertheless in agreement with other studies showing behavioral sex diposotism, or even behavioral sex dimorphism in two sets of behaviors (play behaviors and mounting behaviors; (Deputte and Goy, 1991; Goy and Deputte, 1996). Therefore, existing behavioral sex diposotism seems not to be directly influenced by interactions with partners.

PRIMATE SOCIALIZATION REVISITED

145

Previous multivariate analysis (CFA; see Fig. 18) on data from the same study indicated a sex segregation in play and a general same-sex preference in infants (Deputte and Goy, 1991), a phenomenon previously described in rhesus macaques (Suomi, 1985). Thus, it could be hypothetized that behavioral sex diposotism in rhesus macaques is dependent on two processes, a ‘‘match’’ process and a ‘‘fit’’ process. The match process refers to the early same-sex attractiveness: Female infants are more attracted to quiet social targets such as male and female adults, whereas male infants are more attracted to fast-moving targets, generally other male infants or juveniles (Baldwin, 1986). Therefore, the match process, same-sex attractiveness, would lead to reinforcement of basic shared characteristics by one given sex that might be different from those of the other sex (Goy, 1968). Rough-and-tumble play in males and, later, face-to-face male aggressive interactions, on the one hand, and female–female cohesion, despite sometimes strong dominance relationships, on the other hand, are examples of the match process. The fit process refers to adjustment of complementary behaviors between partners within an interaction, such as for sexual behaviors and in chase–flight play interactions (Goy, 1979). During early periods in ontogeny, male infants performed mounts indiscriminately with male or female partners. Later, in males, a preference for mounting females emerged because the responses of females, standing quietly, led to a better adjustment than a response from a male who tried to escape from the partner’s mounting posture and by twisting away tried to initiate a wrestle or a rough-and-tumble play interaction. The fit process leads, for example, to the development of adaptive copulatory behaviors between opposite-sex partners. The fit process actually includes the development of the capacity to deal with behavioral contingencies stressed by Mason (1978b) and to complementarity in social interactions and relationships. The social organization of squirrel monkeys, showing a sexual segregation outside the breeding season and a mating season (Coe and Rosenblum, 1974; Leger et al., 1981), is an illustration of both match and fit processes, respectively, in play. The match and fit processes explain how slight initial differences between the sexes could increase during development, being reinforced by the same-sex attractiveness, and why, at the same time, complex behavioral adjustments remain possible because overall behavioral similarities exist and are preserved in the two sexes (Burton, 1977).

III. Conclusion The complexity of the social ontogeny can be addressed with appropriate multivariate techniques, from MANOVAs to more geometrically oriented techniques, such as CPAIV, CFAIV, and CFA. Thus, behavioral develop-

146

BERTRAND L. DEPUTTE

ment, and especially social ontogeny, might cease to be the Cinderella that, according to Bateson (1991, p. 19), ‘‘ethologists [considered the] subject for many years—interesting and beautiful, yet widely ignored.’’ A. What Multivariate Techniques Bring to the Study of Social Ontogeny Multivariate techniques provide a more comprehensive view of the entangled processes of social ontogeny than do approaches using univariate or even bivariate indices. The multivariate techniques presented in this chapter visually display the reduction of the complexity of the structure of the data set that is analyzed. Contrary to other related techniques, the complex and multiple interrelations between the variables are presented at once on a single graph instead of having to be mentally reconstructed from the reading of tables. The use of multivariate techniques for dealing with the intrinsic complexity of the data from the beginning of the analysis without having to reduce the complexity a priori yields comprehensive figures showing how a complex set of data is structured along only few dimensions. The analytical approach allows us to demonstrate and statistically assert what was previously assumed. The differentiation of infants’ behavioral repertoire is partner related, confirming the importance of the diversity of the social milieu or the flexibility of individual behavioral repertoire in the development of the infant’s social competence. It has been established that in both species infants that acquired a large behavioral repertoire also showed a high level of visual attention and contacted evenly most, if not all, of the available partners. This could probably be a more general phenomenon which emphasizes the process of learning behavioral contingencies during interactions. These multivariate techniques also help us demonstrate that behavioral sex differences develop early through same-sex preference. However, these differences mostly affect competitive and sexual behaviors, whereas these remains a potential for each sex to perform the same basic repertoire as that of the other sex. In behavioral sex differences diposotism is much more developed than dimorphism. All these results are obtained only with multivariate techniques that deal with complexity and that present all the interrelations of the variables. B. Final Considerations Social ontogeny is thus demonstrated to be a behavioral ontogeny shaped and triggered by social interactions which are dependent on infants’ matura-

PRIMATE SOCIALIZATION REVISITED

147

tion and growth, on the one hand, and social organization variables, on the other hand. This is another formulation of the statement by Fragaszy and Mitchell (1974, p. 567): ‘‘Group members act on [that] original material to develop its expression, but they can not create new responses or force expression of existing ones.’’ Social interactions constitute the bulk of individual social experience which has an obvious cognitive component (Swartz, 1982b). Social experience could be seen as the unique way an individual processes and stores the outcomes of the social interactions in which it was involved. The contention by Chevalier-Skolnikoff (1977, p. 182) that ‘‘preliminary analysis of primate potential for socialization in terms of cognitive abilities suggests that the socialization process is qualitatively different at different age levels’’ emphasizes the intricacy between maturational processes and social environmental influences (synonymous with social experience). The features of maturation and social experience emphasize the ‘‘uniqueness’’ of social ontogeny (Mason, 1976). Maturation and social environment are as precisely interwoven in social ontogeny as the two horsehide parts of a baseball or a cricket ball (Fig. 23). Multivariate techniques have helped us explore how the two skins are tightly sewed together and the strength of the thread that keeps them together. Then the ball is ready to be thrown and hit in a game, just as an individual is ready to take its place within a social group for its life or at least for a part of it. Social ontogeny establishes a social phenotype through epigenetic processes. Infants develop their social network through an identification process with the social phenotypes that surround them and through behavioral differentiation that is partner related (Swartz, 1982b). Although dyads are, and were used as, the basic subsystem, it should be remembered that a dyad should actually be considered only as the interface of two social networks. Regarding Hailman’s formula, which determines the complexity of ontogeny, it should be pointed out that es (the social environment), a component of E (the environment), is fundamentally dynamic:

Fig. 23. Baseball or cricket ball metaphor of social ontogeny.

148

BERTRAND L. DEPUTTE

E: (es , ep , C, F, A)

(1)

where ep is the population, C is the living community that includes predators and competitors, F is the food items, and A is the abiotic factors. This dynamics feature could be formalized in defining a ‘‘systemic function,’’ S (Eq. 2) representing the dynamic interactions between all the social phenotypes in a group (through their social networks): S: (es t⫺2, SP1, SP2, SP3, . . . , SPn, E⬘) 兩씮 est⫺1

(2)

where the function S controls the changes of the social environment, es, during a time span from t-2 to t-1, under the influence of the social phenotypes which composed the group and the rest of the environmental variables, E⬘. All the different studies presented in this chapter have stressed the integrated influence of the social phenotypes on the development of new ones through the existence of social networks. Thus, the original Hailman formula could be modified to account for the multilevel dialectic feature of primate social ontogeny. The outputs of the systemic function (Eq. 2) could be introduced within the ontogenetic function, which can be renamed a ‘‘social ontogenetic function’’: O: (SP0 t⫺1, G, est⫺1, E⬘) 兩씮 SP0t

(3)

which maps the ontogenetic changes of one social phenotype, SP0, its genotype G, the initial social environment, and other environmental components, E⬘, to a new social phenotype, during an increment of time (from t-1 to t). In the same way, the social group, the social networks, and consequently the social systemic function are influenced by the presence and the development of a new phenotype. To account for this, the social systemic function could be modified: SP0 is the new social phenotype and n has consequently increased: S: (est⫺1, SP0, SP1, SP2, SP3, . . . , SPn, E⬘) 兩씮 est

(4)

The social systemic function can be viewed as the composition of several ontogenetic functions. If is worth remembering that the ontogenetic function acts during the whole life on an individual. Hence, the social phenotypes, in Eq. (4) are not fixed but dynamic entities of the social systemic function. As Mason (1976) asserted, ‘‘Individuals are both the products and the producers of societies.’’ Studies in social ontogeny might elucidate how this occurs (Bernstein and Williams, 1986). Application of multivariate techniques to social ontogeny issues might be considered as a successful empirical methodology to match the theoretical expectations of many authors, especially Robert Hinde, whose general approach, according to Bateson (1991), is to overemphasize complexity.

PRIMATE SOCIALIZATION REVISITED

149

Many other students of primate development have also highlighted this complexity: Rosenblum and Coe (1977, p. 498) lamented that ‘‘If males and females mature at different rates, sex and maturational level will be confounded during early development. . . . If in addition, environmental factors influence the respective rates of maturation of the sexes . . . the potential for even further confounding is evident.’’ Multivariate techniques, in providing comprehensive visual representations of the inherent complexity of interactional processes, help to clarify the issue of social ontogeny. In the social phenotype framework, understanding social ontogeny, and therefore social development at an individual level, might help to answer questions about the ultimate causes of sociality (Fragaszy and Mitchell, 1974; Bernstein and Williams, 1986), relating in a metadialectical view attachment theory to ecological constraints.

IV. Summary Social ontogeny could be described as one component of the general process of ontogeny which refers explicitly and only to the development of an individual. Social ontogeny deals more specifically with the development of behaviors and cognitive abilities necessary for an individual to live in an organized social milieu. Because these aspects of individual development occur within a social group, social ontogeny results in a complex interplay between individual maturational factors, other genetic factors, and environmental influences, especially those of social partners. Everyone in a group participates in a certain way in the social ontogeny of a new member. Many papers have reviewed the different processes involved in social ontogeny and described the role the partners played in the development of an infant’s social behavior. However, the description of the development of social behavior has far from matched the complexity of the hypotheses or theoretical assumptions. This chapter reviewed the different approaches that were and are still used to deal with the issue of social development. Some of these approaches are purely descriptive, as in the univariate approach, whereas others have attempted to determine the interactive aspects of social development such as in the index approach. Finally, different new multivariate approaches are presented as attempts to determine the influences of some of the variables, the importance of which in social ontogeny has been previously assumed although not demonstrated. The multivariate approaches were exemplified with studies on social ontogeny of gray-cheeked mangabeys and rhesus macaques. They mainly considered two fundamental processes in social ontogeny: building a social behavioral repertoire and building a network of social relationships. The results

150

BERTRAND L. DEPUTTE

of these studies provided a visual description of the complex interactions between many variables involved in these two ontogenetic processes. These analyses and the results they yielded were considered as useful analytical tools which meet the theoretical expectations of students of social ontogeny.

Acknowledgments This chapter is dedicated to Robert Hinde, William Mason, Irwin Bernstein, and Robert Goy, who were for me the most influential ‘‘agents’’ in my scientific ontogeny, especially in my interest for social behavior and social ontogeny. A special thanks to Georges Cancela da Fonseca, who introduced me to multivariate analyses when these were in the very early stages of development, and to Gaston Richard, who made ethology so attractive to me. I am very much indebeted to Annie Gautier-Hion and Jean-Pierre Gautier, who gave me the opportunity to develop my interests in social behavior. Annie Gautier-Hion spent a considerable number of hours trying to reduce the complexities of my thoughts about complexity into a manageable range. Professor Charles T. Snowdon was my guide who inspired me with strength to accept many scientific challenges and provided me with generous, continuous, and outstanding support. My debt to him is immeasurable. My interest in social ontogeny would have been fruitless without Andre´ Baudas, who took care of all the mangabey subjects I have been observing for more than 10 years, and to Steve Eisele and Mike Hempel, who took care of the rhesus subjects. This paper could not have been written without the friendly and expert collaboration of Rene´ Quris, who helped me in my challenge to explain complex phenomena without having to oversimplify them. Alain Bellido, another expert in teaching multivariate techniques, provided me with invaluable assistance in trying to make these very useful techniques attractive for the students of behavior. I thank Dr. Vaclav Vancata for inviting me to ‘‘Primate Ontogeny,’’ an international symposium he organized in Trest, Czech Republic, in 1995 at which I first presented the ideas that are developed in this chapter. Some of these ideas were published in Anthropologie. Financial support was provided by a CNRS/ UMR 6552 base grant and in part by National Institutes of Health Grant RR00167 to Wisconsin Regional Primate Research Center and a postdoctoral NSF-CNRS grant.

References Adams-Curtis, L. E., and Fragaszy, D. M. (1994). Development of manipulation in capuchin monkeys during the first 6 months. Dev. Psychobiol. 27, 123–136. Altmann, J. (1974). Observational study of behavior: Sampling methods. Behaviour 49, 227–267. Altmann, J. (1978). Infant independence in yellow baboons. In ‘‘The Development of Behavior: Comparative and Evolutionary Aspects’’ (G. M. Burghardt and M. Bekoff, Eds.), pp. 253–277. Garland, New York. Altmann, J. (1980). ‘‘Baboon Mothers and Infants.’’ Harvard Univ. Press, Cambridge, MA. Altmann, J. (1984). Sociobiological perspectives on parenthood. In ‘‘Parenthood: A Psychodynamic Perspective’’ (R. Cohen, B. Cohler, and S. Weissman, Eds.), pp. 9–23. Guilford, New York. Altmann, S. A. (1962). Social behavior of anthropoid primates: Analysis of recent concepts. In ‘‘Roots of Behavior’’ (E. L. Bliss, Ed.), pp. 277–285. Harper, New York.

PRIMATE SOCIALIZATION REVISITED

151

Altmann, S. A. (1965). Sociobiology of rhesus monkeys. II: Stochastics of social communication. J. Theor. Biol. 8, 490–522. Altmann, S. A. (1967). The structure of primate social communication. In ‘‘Social Communication among Primates’’ (S. A. Altmann, Ed.), pp. 325–362. Univ. of Chicago Press, Chicago. Altmann, S. A. (1995). Consciousness and animal welfare. Newslett. Anim. Behav. Soc. 40, 6–8. Altmann, S. A., and Altmann, J. (1979). Demographic constraints on behavior and social organization. In ‘‘Ecological Influences on Social Organization’’ (E. O. Smith and I. S. Bernstein, Eds.), pp. 47–63. Garland, New York. Baldwin, J. D. (1986). Behavior in infancy: Exploration and play. In ‘‘Comparative Primate Biology, Vol. 2, Part A: Behavior, Conservation and Ecology’’ (G. Mitchell and J. Erwin, Eds.), pp. 295–326. A. R. Liss, New York. Baldwin, J. D., and Baldwin, J. I. (1979). The phylogenetic and ontogenetic variables that shape behavior and social organization. In ‘‘Ecological Influences on Social Organization’’ (E. O. Smith and I. S. Bernstein, Eds.), pp. 89–116. Garland, New York. Bateson, P. (1991). Are there principles of behavioural development? In ‘‘The Development and Integration of Behaviour: Essays in Honour of Robert Hinde’’ (P. Bateson, Ed.), pp. 19–39. Cambridge Univ. Press, Cambridge, UK. Benzecri, J. P., et al. (1973). ‘‘L’analyse des donne´es II: L’analyse des correspondances.’’ Dunod, Paris. Berman, C. M. (1978). The analysis of mother–infant interactions in groups: Possible influence of yearling siblings. In ‘‘Recent Advances in Primatology. Vol. 1, Behaviour’’ (D. J. Chivers and J. Herbert, Eds.), pp. 111–113. Academic Press, London. Berman, C. M. (1980). Mother–infant relationships among free-ranging rhesus monkeys on Cayo Santiago: A comparison with captive pairs. Anim. Behav. 28, 860–873. Berman, C. M. (1982a). The ontogeny of social relationships with group companions among free-ranging infant rhesus monkeys. I. Social networks and differentiation. Anim. Behav. 30, 149–162. Berman, C. M. (1982b). The ontogeny of social relationships with group companions among free-ranging infant rhesus monkeys. II. Differentiation and attractiveness. Anim. Behav. 30, 163–170. Bernstein, I. S. (1966). Analysis of a key role in a capuchin (Cebus albifrons) group. Tulane Stud. Zool. 13, 49–54. Bernstein, I. S. (1978). Sex differences in the behavior of nonhuman primates. Soc. Sci. Med. 12B, 151–154. Berstein, I. S., and Sharpe, L. G. (1996). Social roles in a rhesus monkey group. Behaviour 26, 91–104. Bernstein, I. S., and Williams, L. E. (1986). The study of social organization. In ‘‘Comparative Primate Biology, Vol. 2A: Behavior, Conservation & Ecology’’ (G. Mitchell and J. Erwin, Eds.), pp. 195–213. A. R. Liss, New York. Boothe, R. G., Williams, R. A., Kiorpes, L., and Teller, D. Y. (1980). Development of contrast sensitivity in infant Macaca nemestrina monkeys. Science 208, 1290–1291. Bramblett, C. A. (1973). Social organization as an expression of role behavior among Old World monkeys. Primates 14, 101–112. Burton, F. D. (1972). The integration of biology and behavior in the socialization of Macaca sylvana of Gibraltar. In ‘‘Primate Socialization’’ (F. E. Poirier, Ed.), pp. 29–62. Random House, New York. Burton, F. D. (1977). Ethology and the development of sex and gender identity in non-human primates. Acta Biotheor. 26, 1–18. Capitanio, J. P. (1985). Early experience and social processes in rhesus macaques (Macaca mulatta): II. Complex social interactions. J. Comp. Psychol. 99, 133–144.

152

BERTRAND L. DEPUTTE

Capitanio, J. P., and Reite, M. (1984). The roles of early separation experience and prior familiarity in the social relations of pigtail macaques: A descriptive multivariate study. Primates 25, 474–484. Chalmers, N. R. (1972). Comparative aspects of early infant development in some captive cercopithecines. In ‘‘Primate Socialization’’ (F. E. Poirier, Ed.), pp. 63–82. Random House, New York. Chance, M. R. A., and Jolly, C. J. (1970). ‘‘Social Groups of Monkeys, Apes and Men.’’ Jonathan Cape, London. Chevalier-Skolnikoff, S. (1974). The ontogeny of communication in the stumptail macaque (Macaca arctoides). Contrib. Primatol. 2. Chevalier-Skolnikoff, S. (1977). A Piagetian model for describing and comparing socialization in monkeys, apes and human infants. In ‘‘Primate Bio-social Development: Biological, Social, and Ecological Determinants’’ (S. Chevalier-Skolnikoff and F. E. Poirier, Eds.), pp. 159–187. Garland, New York. Coe, C. L., and Rosenblum, L. A. (1974). Sexual segregation and its ontogeny in squirrel monkey social structure. J. Hum. Evol. 3, 551–561. Crook, J. (1970a). Introduction—Social behaviour and ethology. In ‘‘Social Behaviour in Birds and Mammals’’ ( J. Crook, Ed.), pp. xxi–xl. Academic Press, New York. Crook, J. (1970b). The socio-ecology of primates. In ‘‘Social Behaviour in Birds and Mammals’’ ( J. Crook, Ed.), pp. 103–169. Academic Press, New York. Denenberg, V. H. (1979). Paradigms and paradoxes in the study of behavioral development. In ‘‘Origins of the Infant’s Social Responsiveness’’ (E. B. Thoman, Ed.), pp. 251–289. Erlbaum, Hillsdale, NJ. Deputte, B. L. (1983). Ontogenetic development of dyadic social relationships: Assessing individual roles. Am. J. Primatol. 4, 309–318. Deputte, B. L. (1985). De´veloppement du comportement visuel social chez une jeune femelle cercoce`be (Cercocebus albigena albigena): E´tude pre´liminaire sur un groupe captif. Biol. Behav. 10, 113–150. Deputte, B. L. (1986a). Ontogene`se du cercoce`be a` joues blanches (Lophocebus albigena albigena) en captivite´: de´veloppement des comportements de communication et des relations sociales. Unpublished doctoral dissertation, Universite´ de Rennes I. Deputte, B. L. (1986b). Parner’s roles in the socialization of young captive mangabeys (Lophocebus albigena albigena). In ‘‘The Individual and Society’’ (L. Passera and J. P. Lachaud, Eds.), pp. 109–117. I. E. C. Privat, Toulouse, France. Deputte, B. L. (1991). Reproductive parameters of captive grey-cheeked mangabeys. Folia Primatol. 57, 57–69. Deputte, B. L. (1992). Life-history of captive grey-cheeked mangabeys: Physical and sexual development. Int. J. Primatol. 13, 509–531. Deputte, B. L. (1995a). Structuration des comportements en fonction du sexe chez les primates. In ‘‘La Frontie`re des Sexes’’ (A. Ducros and M. Panoff, Eds.), pp. 29–71. P. U. F., Paris. Deputte, B. L. (1995b). Social ontogeny in primates: Semantics, concepts, facts and processes. Anthropologie 33, 83–98. Deputte, B. L., and Goy, R. W. (1991). Development of behavioral sexual differences: Experimental study of prenatal diethylstilbestrol (DES) female rhesus macaque infants. In ‘‘Primatology Today’’ (A. Ehara, T. Kimura, O. Takenaka, and M. Iwamoto, Eds.), pp. 465–468. Elsevier, Amsterdam. Deputte, B. L., and Leclerc-Cassan, M. (1981). Sex determination and age estimation in the white-cheeked gibbon (Hylobates concolor leucogenys): Anatomical and behavioural features. Int. Zoo Yearbook 21, 187–193.

PRIMATE SOCIALIZATION REVISITED

153

Deputte, B. L., and Quris, R. (1996). Socialization processes in primates: Use of multivariate analyses. I. Application to social development of captive mangabeys. Behav. Processes 36, 135–149. Deputte, B. L., and Quris, R. (1997). Socialization processes in primates: Use of multivariate analyses. II. Influence of sex on social development of captive rhesus monkeys. Behav. Processes 40, 85–96. Deputte B. L., and Vauclair, J. (1998). Connaıˆtre le monde, connaıˆtre les siens: Les grands singes [n’] ont [pas] la parole. In ‘‘Grands Singes: La Fascination du Double’’ (S. Bobbe´, Ed.), pp. 12–76. Editions Autrement, Paris. de Waal, F. B. M., and van Roosmalen, A. (1979). Reconciliation and consolation among chimpanzees. Behav. Ecol. Sociobiol. 5, 55–66. Dolhinow, P., and Murphy, G. (1982). Langur monkey (Presbytis entellus) development: The first 3 months of life. Folia Primatol. 39, 305–331. Etkin, W. (1964). Co-operation and competition in social behavior. In ‘‘Social Behavior and Organization among Vertebrates’’ (W. Etkin, Ed.), pp. 1–34. Univ. of Chicago Press, Chicago. Fagen, R. M., and Goldman, R. N. (1977). Behavioural catalogue analysis methods. Anim. Behav. 25, 261–274. Fairbanks, L. A. (1988). Vervet monkey grandmothers: Interactions with infant grandoffspring. Int. J. Primatol. 9, 425–441. Fairbanks, L. A., and McGuire, M. T. (1979). Inhibition of control role behaviors in captive vervet monkeys (Cercopithecus aethiops sabaeus). Behav. Processes 4, 145–153. Fairbanks, L. A., and McGuire, M. T. (1987). Mother–infant relationships in vervet monkeys: Responses to new adult males. Int. J. Primatol. 8, 351–366. Fairbanks, L. A., McGuire, M. T., and Page, N. (1978). Social roles in captive vervet monkeys (Cercopithecus aethiops sabaeus). Behav. Processes 3, 335–352. Fedigan, L. M. (1972). Roles and activities of male geladas (Theropithecus gelada). Behaviour 41, 82–90. Fedigan, L. M. (1976). ‘‘A Study of Roles in the Arashiyama West Troop of Japanese Macaques (Macaca fuscata),’’ Contrib. Primat. No. 9. Karger, Basel. Fedigan, L. M. (1982). ‘‘Primate Paradigms: Sex Roles and Social Bonds.’’ Eden Press, Montreal. Fragaszy, D. M. (1989). Activity states and motor activity in an infant capuchin monkey (Cebus apella) from birth through eleven weeks. Dev. Psychobiol. 22, 141–157. Fragaszy, D. M. (1990a). Early behavioral development in capuchins (Cebus). Folia Primatol. 54, 119–128. Fragaszy, D. M. (1990b). Sensorimotor development in hand-reared and mother-reared tufted capuchins: A systems perspective on the contrasts. In ‘‘‘Language’ and Intelligence in Monkeys and Apes: Comparative Developmental Perspectives’’ (S. T. Taylor and K. R. Gibson, Eds.), pp. 172–204. Cambridge Univ. Press, Cambridge, UK. Fragaszy, D. M., and Mitchell, G. (1974). Infant socialization in primates. J. Hum. Evol. 3, 563–574. Gibson, K. R. (1981). Comparative neuro-ontogeny, its implications for the development of human intelligence. In ‘‘Infancy and Epistemology’’ (G. E. Butterworth, Ed.), pp. 52–82. Harvester Press, Brighton, UK. Goldfoot, D. A., Wallen, K., Neff, D. A., McBrair, M. C., and Goy, R. W. (1984). Social influences on the display of sexually dimorphic behavior in rhesus monkeys: Isosexual rearing. Arch. Sexual Behav. 13, 395–412. Gottlieb, G. (1970). Conceptions of prenatal behavior. In ‘‘Development and Evolution of Behavior—Essays in Memory of T. C. Schneirla’’ (L. R. Aronson, D. S. Lehrman, J. S. Rosenblatt, and E. Tobach, Eds.), pp. 111–137. Freeman, San Francisco.

154

BERTRAND L. DEPUTTE

Goy, R. W. (1968). Organizing effects of androgen on the behavior of rhesus macaques. In ‘‘Endocrinology and Human Behavior’’ (R. P. Michael, Ed.), pp. 12–31. Oxford Univ. Press, London. Goy, R. W. (1979). Sexual compatibility in rhesus monkeys: Predicting sexual performance of oppositely sexed pairs of adults. Sex Horm. Behav. 62, 227–255. Goy, R. W., and Deputte, B. L. (1996). The effects of diethylstilbestrol (DES) before birth on the development of masculine behavior in juvenile female rhesus monkeys. Horm. Behav. 30, 379–386. Goy, R. W., and Phoenix, C. H. (1972). The effects of testosterone propionate administered before birth on the development of behavior in genetic female rhesus monkeys. In ‘‘Steroid Hormones and Brain Functions’’ (C. Sawyer and R. Gorski, Eds.), pp. 193–201. Univ. of California Press, Berkeley. Goy, R. W., and Resko, J. A. (1972). Gonadal hormones and behavior of normal and pseudohermaphroditic nonhuman female primates. Rec. Prog. Horm. Res. 28, 707–733. Goy, R. W., and Robinson, J. A. (1982). Prenatal exposure of rhesus monkeys to patent androgens: Morphological, behavioral and physiological consequences. Banbury Rep. 11, 355–378. Goy, R. W., Bercovitch, F. B., and McBrair, M. C. (1988a). Behavioral masculinization is independent of genital masculinization in prenatally androgenized female rhesus macaques. Horm. Behav. 22, 552–571. Goy, R. W., Kraemer, G., and Goldfoot, D. A. (1988b). Biological influences on grooming in nonhuman primates. Ann. N. Y. Acad. Sci. 525, 56–68. Hailman, J. P. (1982). Ontogeny: Towards a general theoretical framework for ethology. In ‘‘Perspectives in Ethology, Vol. 5. Ontogeny’’ (P. P. G. Bateson and P. H. Klopfer, Eds.), pp. 133–189. Plenum, New York. Harlow, H. F., and Harlow, M. K. (1965). The affectional systems. In ‘‘Behavior of Nonhuman Primates: Modern Research Trends’’ (A. M. Schrier, H. F. Harlow, and F. Stollnitz, Eds.), pp. 287–334. Academic Press, London. Hinde, R. A. (1966). ‘‘Animal Behaviour.’’ McGraw-Hill, London. Hinde, R. A. (1969). Influence of social companions and temporary separation on mother– infant relations in rhesus monkeys. In ‘‘Determinants of Infant Behaviour’’ (B. M. Foss, Ed.), Vol. 4, pp. 37–40. Methuen, London. Hinde, R. A. (1971). Development of social behavior. In ‘‘Behavior of Nonhuman Primates: Modern Research Trends’’ (A. M. Schrier, H. F. Harlow, and F. Stollnitz, Eds.), Vol. 3, pp. 1–60. Academic Press, New York. Hinde, R. A. (1975). Interactions, relationships and social structure in nonhuman primates. Symp. 5th In. Congr. Primatol. Soc., 13–24. Hinde, R. A. (1976a). On describing relationships. J. Child. Psychol. Psychiatr. 17, 1–19. Hinde, R. A. (1976b). Interactions, relationships and social structure. Man 11, 1–17. Hinde, R. A. (1977). On assessing the bases of partner preferences. Behaviour 42, 1–9. Hinde, R. A., and Atkinson, S. (1970). Assessing the roles of social partners in maintaining social proximity as exemplified by mother–infant relations in rhesus monkeys. Anim. Behav. 18, 169–176. Hinde, R. A., and Spencer-Booth, Y. (1967). The behaviour of socially living rhesus monkeys in their first two and a half years. Anim. Behav. 15, 169–196. Hinde, R. A., and White, L. (1974). The dynamics of a relationship: Rhesus monkey ventroventral contact. J. Comp. Physiol. Psychol. 86, 8–23. Jensen, G. D., Bobbitt, R. A., and Gordon, B. N. (1968). Sex differences in the development of independence of infant monkeys. Behaviour 30, 1–14.

PRIMATE SOCIALIZATION REVISITED

155

Johnson, C. K., Gilbert, M. D., and Herdt, G. H. (1979). Implications for adult roles from differential styles of mother–infant bonding: An ethological study. J. Nervous Mental Dis. 167, 29–37. Johnson, R. L., and Southwick, C. H. (1987). Ecological constraints on the development of infant independence in rhesus. Am. J. Primatol. 13, 103–118. Kraemer, G. W. (1992). A psychobiological theory of attachment. Behav. Brain Sci. 15, 493–541. Kummer, H. (1978). On the value of social relationships to nonhuman primates: A heuristic scheme. Social Sci. Information 17, 687–705. Kummer, H., Goetz, W., and Angst, W. (1970). Cross species modifications of social behavior in baboons. In ‘‘Old World Monkeys: Evolution, Systematics, and Behavior’’ ( J. R. Napier and P. H. Napier, Eds.), pp. 351–363. Academic Press, New York. Lebreton, J. D., Sabatier, R., Banco, J. L., and Bacou, A. M. (1991). Principal component and correspondence analysis with respect to instrumental variables: An overview of their role in studies of structure–activity and species–environment relationships. In ‘‘Applied Multivariate Analysis in SAR and Environmental Studies’’ ( J. Devillers and W. Karcher, Eds.), pp. 85–114. ECSC, EEC, EAEC, Brussels. Lee, P. C. (1984). Ecological constraints on the social development of vervet monkeys. Behaviour 91, 245–262. Leger, D. W., Mason, W. A., and Fragaszy, D. M. (1981). Sexual segregation, cliques, and social power in squirrel monkey (Saimiri) groups. Behaviour 76, 163–181. Lewis, M., and Cherry, L. (1977). Social behavior and language acquisition. In ‘‘Interaction, Conversation and the Development of Language’’ (M. Lewis and L. A. Rosenblum, Eds.), pp. 227–245. Wiley, New York. Maier, N. R. F., and Schneirla, (1964). ‘‘Principles of Animal Psychology.’’ Dover, London. Mason, W. A. (1965). The social development of monkeys and apes. In ‘‘Primate Behavior: Field Studies of Monkeys and Apes’’ (I. DeVore, Ed.), pp. 514–543. Holt, Rinehart & Winston, New York. Mason, W. A. (1976). Primate social behavior: Pattern and process. In ‘‘Evolution of Brain and Behavior in Vertebrates’’ (R. B. ‘‘Masterton et al., Eds.), pp. 425–455. Erlbaum, Hillsdale, NJ. Mason, W. A. (1978a). Ontogeny of social systems. In ‘‘Recent Advances in Primatology. Vol. 1. Behaviour’’ (D. J. Chivers and J. Herbert, Eds.), pp. 5–14. Academic Press, London. Mason, W. A. (1978b). Social experience and primate cognitive development. In ‘‘The Development of Behavior: Comparative and Evolutionary Aspects’’ (G. M. Burghardt and M. Herbert, Eds.), pp. 233–251. Garland, New York. McKenna, J. J. (1979). Aspects of infant socialization, attachment, and maternal caregiving patterns among primates: A cross-disciplinary review. Yearbook Phys. Anthropol. 22, 250–286. Papousek, H., and Papousek, M. (1979). The infant’s fundamental response system in social interaction. In ‘‘Origins of the Infant’s Social Responsiveness’’ (E. B. Thoman, Ed.), pp. 175–208. Erlbaum, Hillsdale, NJ. Poirier, F. E. (1972). Introduction. In ‘‘Primate Socialization’’ (F. E. Poirier, Ed.), pp. 3–28. Random House, New York. Poirier, F. E. (1973). Primate socialization and learning. In ‘‘Culture and Learning’’ (S. Kimball and J. Burnett, Eds.), pp. 3–41. Univ. Washington Press, Seattle. Poirier, F. E. (1977). Introduction. In ‘‘Primate Bio-social Development: Biological, Social, and Ecological Determinants’’ (S. Chevalier-Skolnikoff and F. E. Poirier, Eds.), pp. 1–39. Garland, New York. Poirier, F. E. (1982). Primate socialization variables. Ann. Ist. Super. Sanita 18, 251–266.

156

BERTRAND L. DEPUTTE

Rabaud, E. (1937). ‘‘Phe´nome`ne Social et Socie´te´s Animales.’’ Alcan, Paris. Reynolds, V. (1970). Roles and role change in monkey society: The consort relationship of rhesus monkeys. Man 5, 449–465. Rhine, R. J., and Hendy-Neely, H. (1978). Social development of stumptail macaques (Macaca arctoides): Synchrony of changes in mother–infant interactions and individual behaviors during the first 60 days of life. Primates 19, 618–692. Rosenblum, L. A. (1982). The influence of the social and physical environment on mother– infant relations. Ann. Ist. Super. Sanita 18, 215–222. Rosenblum, L. A., and Alpert, S. (1977). Response to mother and stranger: A first step in socialization. In ‘‘Primate Bio-social Development: Biological, Social, and Ecological Determinants’’ (S. Chevalier-Skolnikoff and F. E. Poirier, Eds.), pp. 463–477. Garland, New York. Rosenblum, L. A., and Coe, C. L. (1977). The influence of social structure on squirrel monkey socialization. In ‘‘Primate Bio-social Development: Biological, Social, and Ecological Determinants’’ (S. Chevalier-Skolnikoff and F. E. Poirier, Eds.), pp. 479–499. Garland, New York. Rosenblum, L. A., and Pauly, P. W., Jr. (1980). The social milieu of the developing monkey: Studies of the development of social perception. Reprod. Nutr. De´v. 20, 827–841. Rosenblum, L. A., and Plimpton, E. H. (1979). The effects of adults on peer interactions. In ‘‘The Child and Its Family’’ (M. Lewis and L. A. Rosenblum, Eds.), pp. 195–217. Plenum, New York. Rowell, T. E. (1972). ‘‘The Social Behaviour of Monkeys.’’ Penguin, London. Rowell, T. E., and Olson, D. K. (1983). Alternative mechanisms of social organization in monkeys. Behaviour 86, 31–54. Rowell, T. E., Din, N. A., and Omar, A. (1968). The social development of baboons in their first three months. J. Zool. London 155, 159–167. Sabatier, R., Lebreton, J. D., and Chessel, D. (1989). Principal component analysis with instrumental variables as a tool for modelling composition data. In ‘‘Multiway Data Analysis’’ (R. Coppi and S. Bolasco, Eds.), pp. 341–352. Elsevier North Holland, Amsterdam. Sackett, G. P., Sameroff, A. J., Cairns, R. B., and Suomi, S. J. (1981). Continuity in behavioral development: Theoretical and empirical issues. In ‘‘Behavioral Development: The Bielefeld Interdisciplinary Project’’ (K. Immelmann, G. W. Barlow, L. Petrinovitch, and M. Main, Eds.), pp. 23–57. Cambridge Univ. Press, New York. Sailer, L. D., and Gaulin, S. J. C. (1984). Proximity, sociality and observation: The definition of social groups. Amer. Anthrop. 86, 91–98. Schneirla, T. C. (1952). Basic correlations and coordinations in insect societies with special reference to ants. In ‘‘Structure et Physiologie des Socie´te´s Animales’’ (P. P. Grasse´, Ed.), pp. 247–269. CNRS, Paris. Schneirla, T. C. (1959). An evolutionary and developmental theory of biphasic process underlying approach and withdrawal. In ‘‘Nebraska Symposium on Motivation’’ (M. R. Jones, Ed.), pp. 1–42, Univ. of Nebraska Press, Lincoln. Smuts, B. B., Cheney, D. L., Seyfarth, R. M., Wrangham, R. W., and Struhsaker, T. T. (1986). ‘‘Primate Societies,’’ Univ. of Chicago Press, Chicago. Sokal, R. R., and Rohlf, F. J. (1969). ‘‘Biometry: The Principles and Practices of Statistics in Biological Research.’’ Freeman, San Francisco. Suomi, S. J. (1985). Response styles in monkeys: Experiential effects. In ‘‘Biologic Responses Styles: Clinical Implications’’ (H. Klar and L. J. Siever, Eds.), pp. 1–17 American Psychiatric Press, Washington, DC.

PRIMATE SOCIALIZATION REVISITED

157

Swartz, K. B. (1982a). Issues in the measurement of attachment in non-human primates. J. Hum. Evol. 11, 237–245. Swartz, K. B. (1982b). A comparative perspective on perceptual, cognitive, and social development. J. Hum. Evol. 11, 315–320. Teller, D. Y., Regal, D. M., Videen, T. O., and Pulos, E. (1978). Development of visual acuity in infant monkeys (Macaca nemestrina) during the early postnatal weeks. Vision Res. 18, 561–566. Ter Braak, C. J. F. (1987). Ordination. In ‘‘Data Analysis in Community and Landscape Ecology’’ (R. H. G. Jongman, C. J. F. ter Braak, and O. F. R. van. Tongeren, Eds.), pp. 91–173. PUDOC, Wageningen, The Netherlands. Ter Braak, C. J. F. (1992). Permutation versus bootstrap significance tests in multiple regression and ANOVA. In ‘‘Bootstrapping and Related Techniques’’ (K.-H. Jo¨ckel, G. Rothe, and W. Sendler, Eds.), pp. 79–86. Springer-Verlag, Berlin. Thoman, E. B., Acebo, C., Dreyer, C. A., Becker, P. T., and Freese, M. P. (1979). Individuality in the interactive process. In ‘‘Origins of the Infant’s Social Responsiveness’’ (E. B. Thoman, Ed.), pp. 305–338. Erlbaum, Hillsdale, NJ. Tinbergen, N. (1967). ‘‘La vie Sociale des Animaux.’’ Petite Bibliothe`que Payot, Paris. Van Schaik, C. P., and van Hooff, J. A. R. A. M. (1983). On the ultimate causes of primate social systems. Behaviour 85, 91–117. Wade, T. D. (1977). Complementarity and symmetry in social relationships of nonhuman primates. Primates 18, 835–847. Wheeler, W. M. (1923). ‘‘Social Life among the Insects.’’ Harcourt Brace, New York. White, L. E., and Hinde, R. A. (1975). Some factors affecting mother–infant relations in rhesus monkeys. Anim. Behav. 23, 527–542. Wilson, E. O. (1975). ‘‘Sociobiology: The New Synthesis.’’ Harvard Univ. Press, Cambridge, MA. Wohlwill, J. (1973). ‘‘The Study of Behavioral Development.’’ Academic Press, New York.

This Page Intentionally Left Blank

ADVANCES IN THE STUDY OF BEHAVIOR, VOL. 29

Ultraviolet Vision in Birds Innes C. Cuthill, Julian C. Partridge, Andrew T. D. Bennett, Stuart C. Church, Nathan S. Hart, and Sarah Hunt school of biological sciences university of bristol bristol bs8 1ug, united kingdom

I. Introduction All animals depend on their senses for information about the environment around them but, of all taxa, birds are arguably the animals most dependent on vision. In part this is a consequence of flight, which demands a visual system with both high spatial resolution and fast temporal responses, but many other avian behaviors, including foraging, predator avoidance, and diverse intraspecific interactions, are also, in the main, visually mediated. An obvious example is the involvement of vision in avian mate choice; many birds have spectacularly colored plumage, mating rituals involving elaborate display, and assessment of potential partners involving close visual inspection. As a consequence, much research on bird behavior has, either tacitly or overtly, involved the study of avian vision. In order to correctly interpret any visually mediated behavior, however, it is necessary to appreciate the way in which visual systems work and, in particular, to recognize that the vision of many animals differs considerably from our own. Ignoring this fact can have unforeseen consequences for the behavioral ecologist and, at worst, experiments can be inappropriately designed, and painstaking measurements may be irrelevant or fundamentally flawed. It is thus only by attempting to ‘‘see the world through the eye of a bird’’ that we may understand much of their behavior (Burkhardt, 1989; Burkhardt and Finger, 1991; Bennett et al., 1992, 1994). The differences between human and bird vision are nowhere more important than in the case of color vision. Although we tend to be somewhat self-satisfied with our own color vision, it is not particularly well developed when compared with that of most vertebrates. For color vision, the eye must have several types of photoreceptor that differ from each other in their spectral sensitivities, and the color vision of most humans relies on three types of retinal cone photoreceptors. Because all three cones are 159

Copyright 䉷 2000 by Academic Press All rights of reproduction in any form reserved. 0065-3454/00 $35.00

160

INNES C. CUTHILL et al.

neurally integrated in the assessment of spectral radiances and thus in the perception of color, our colors are mapped into three-dimensional (‘‘trichromatic’’) color space (see Section II). In this we are somewhat better endowed than most mammals, which have but two types of retinal cone and consequently only dichromatic color vision, attributes that are shared with red–green ‘‘color-blind’’ humans (Jacobs, 1981). However, human trichromacy is the result of a relatively recent evolutionary event which occurred in the ancestors of all Old World primates only approximately 45 million years ago (Hunt et al., 1998; Bowmaker and Hunt, 1999). This event involved the derivation of a third cone type by the duplication and subsequent mutation of a gene determining cone spectral sensitivity (see Section II), but it also required that existing neural circuits be coopted for purposes for which they could not originally have evolved. For these reasons it is most unlikely that human color vision is at some global optimum for terrestrial vision; it has emerged only relatively recently via the evolutionary bottleneck of mammalian dichromacy. In contrast, many vertebrates have retained visual systems that, at least at the retinal level, are more complex than ours and are presumably similar to that found in the fish-like ancestor of the subphylum (see Section II). Molecular studies suggest that the ancestral vertebrate condition, as well as that existing in many extant vertebrates, involves four types of retinal cone (Bowmaker and Hunt, 1999). Since psychophysical research on fish (Neumeyer and Arnold, 1989; Neumeyer, 1992) has shown that all four cones can be integrated into fish color vision, it is possible that ancestral vertebrates had a similar visual system. It is thus likely that, far from being advanced, our trichromatic vision is depauperate compared with the ancestral vertebrate condition. We can also anticipate that many extant vertebrates will be shown to have retained the neural apparatus for fourdimensional (tetrachromatic) color vision and, as discussed later, there is substantial evidence that most birds both have four types of cone involved in their color vision and are likely to be tetrachromatic. The consequence of four cone pigments, and tetrachromacy in particular, is that birds see the world differently from humans and in a way for which it is hard to compensate because we simply lack the neural machinery. Color vision does not provide a spectroscopic measure of the spectral reflectance of objects in the environment; instead, the neurons of the visual system map spectral radiance into a perceptual color space that depends on neural processing in the central nervous system and on the spectral sensitivities of the retinal receptors (see Section II,E). Thus, color, as perceived or assessed by an animal, does not provide an unambiguous measure of spectral radiance but is an abstraction that will often be species specific (Lythgoe, 1979; Jacobs, 1983; Endler, 1990). For this reason, it is not possible

UV VISION IN BIRDS

161

to extrapolate from human color experience to that of birds. Specifically, bird color vision will divide the visual environment in different ways compared to human color vision; objects will be categorized differently on the basis of color, contrasts will be different, and color-based assessments will differ from those made by human observers. The nature of these differences is described in detail in Section II. Although tetrachromacy creates differences between human and avian vision, there are other additional physiological differences that limit our appreciation of a bird’s view of the world. First, most of the retinal cones of birds contain oil droplets that, by virtue of their high carotenoid content, act as spectral filters which modify the spectral sensitivities of the cones. By narrowing the spectral sensitivities in this way, many natural radiances will stimulate one cone type far more than all others, resulting in more saturated colors than a human would perceive (Govardovskii, 1983; Goldsmith, 1990; Vorobyev et al., 1998). Second, birds are sensitive to ultraviolet (UV) wavelengths, whereas humans are not. Birds are sensitive to UV wavelengths primarily because their ocular media (particularly lens and cornea) are UV transparent. All vertebrates have the potential for UV vision unless short wavelengths are specifically absorbed before the retina (aphakic humans can see UV; Stark et al., 1994). Indeed, UV vision is widespread in many vertebrate taxa, including fish, reptiles, and mammals (Jacobs, 1992). In addition, many vertebrates have cone photoreceptors with peak sensitivities at very short (UV or ‘‘violet’’) wavelengths and molecular evidence strongly indicates that one of the four plesiomorphic vertebrate cones is principally a channel for UV vision (Kawamura and Yokoyama, 1996; Yokoyama and Yokoyama, 1996; Wilkie et al., 1998; Yokoyama et al., 1998; Bowmaker and Hunt, 1999). The presence of such a cone, in addition to UV-transparent ocular media, confers considerable UV sensitivity. This sensitivity was in fact first demonstrated in birds about a quarter of a century ago by operant discrimination experiments with hummingbirds (Huth and Burkhardt, 1972) and pigeons (Wright, 1972a,b). Later, electrophysiological and behavioral experiments provided mounting evidence that UV vision in birds was widespread (Emmerton and Delius, 1980; Chen et al., 1984; Parrish et al., 1984; Chen and Goldsmith, 1986; Burkhardt and Maier, 1989; Maier, 1992), although it was not until 1993 that microspectrophotometry provided direct measurements of an avian cone with peak absorption in the UV (Maier and Bowmaker, 1993). To date, there is positive evidence for UV vision in at least 35 species of birds (see Section II). The differences between human and avian vision mean that, for many purposes, human vision, or standards derived from human psychophysics, are inappropriate for studying bird visual behavior. The importance of this

162

INNES C. CUTHILL et al.

point is most clearly demonstrated by considering UV vision in birds and its role in their behavior, the subject of this review.

II. The Mechanism of Color Vision in Birds The retina functions as a two-dimensional detector array, extracting visual information from the images produced by the eye’s dioptric apparatus (Martin, 1985). Avian retinae, like those of most other vertebrates, are duplex (Schultze, 1866, 1867). In other words, their visual photoreceptors, which are distinguished morphologically from other retinal neurons by their proliferation of specialized membranes, can be divided into two distinct subtypes: rods, which subserve scotopic vision under low light levels, and cones, which dominate the retinae of diurnal species and mediate photopic vision during daylight (Walls, 1963). Cones are approximately 25–100 times less sensitive than rods, but their electrical responses are several times faster (Yau, 1994). These differences in rod and cone physiology are thought to occur because of quantitative differences in the transduction cascade, although the nature of these differences is unclear (Yau, 1994). Avian photoreceptors consist of a cell body which synapses with the neural retina, a central ‘‘inner’’ segment, and an ‘‘outer’’ segment (Cajal, 1893) which contains the photosensitive visual pigment molecules. Rods have relatively large, cylindrical outer segments, whereas cones have a much thinner, tapering outer segment and usually display a large (1–4 애m in diameter) oil droplet in their inner segment (Morris and Shorey, 1967; Braekevelt, 1990, 1993, 1994). Avian retinae usually contain a single type of rod photoreceptor, four classes of single cone, and one type of double cone consisting of a closely juxtaposed pair of cones of different sizes (Cajal, 1893; Morris and Shorey, 1967; Morris, 1970; Bowmaker, 1977; Bowmaker and Knowles, 1977; Mariani and Leure-DuPree, 1978; Cserha´ti et al., 1989). A. Visual Pigments In the rods and cones, visual pigments mediate the acquisition of visual information by transducing and amplifying photon energy into synaptic activity and the movement of ions in the central nervous system (Rodieck, 1998). Visual pigments are colored compounds concentrated in the extensive lipid membranes of the outer segments. All vertebrate visual pigments consist of a chromophore molecule which is embedded in, and covalently bound to, a large protein moiety called an opsin (Bownds, 1967; Nakanishi, 1991). With the exception of some lizards (Provencio et al., 1992), the chromophore of most terrestrial vertebrates, including all birds studied to

UV VISION IN BIRDS

163

date, is 11-cis retinal, the aldehyde of vitamin A1. Visual pigments containing 11-cis retinal are called rhodopsins to distinguish them from visual pigments based on 11-cis 3-dehydroretinal, the aldehyde of vitamin A2 , which are known as porphyropsins (Knowles and Dartnall, 1977). Porphyropsins are mainly restricted to some teleost fish, amphibians, and some aquatic reptiles (Bowmaker, 1991a,b). They tend to have spectral sensitivities shifted to longer wavelengths than equivalent rhodopsins, but in birds the spectral sensitivities of visual pigments are determined only by the opsin and not by use of a different chromophore. Opsins are a member of a large family of cell membrane receptor proteins collectively known as G-protein-linked receptors (Rodieck, 1998), which all have similarities in structure and transduction pathways. Opsin molecules are single polypeptide chains, containing approximately 350 amino acid residues, embedded in the lamellar membranes of the photoreceptor outer segment (Applebury and Hargrave, 1986). The three-dimensional structure of opsin consists of a palisade of seven 움-helices traversing the lipid bilayer, each of which is composed of 24–28 largely nonpolar amino acids connected by short, nonhelical segments rich in polar amino acids (Applebury and Hargrave, 1986; Stryer, 1987). Opsin absorbs maximally below 300 nm, whereas 11-cis retinal has a maximum absorption at about 375 nm (Knowles and Dartnall, 1977; Nakanishi, 1991). The broad, asymmetrical, bell-shaped absorption spectrum characteristic of visual pigments is formed when the chromophore binds with opsin. In birds, the spectral location of the peak sensitivity of the visual pigment (␭max ) is due solely to differences in the genetically determined amino acid complement of the opsin and their electrostatic interactions with the embedded chromophore (Bowmaker and Hunt, 1999). In particular, charged amino acids buried within the opsin’s chromophore-binding cavity are thought to affect the conformational structure of 11-cis retinal (Applebury and Hargrave, 1986; Nakanishi, 1991). By comparing the amino acid sequences of different opsins, it is possible to identify particular residues which are responsible for the spectral tuning of different visual pigments (Takao et al., 1988; Okano et al., 1992; Wang et al., 1992, 1993; Wilkie et al., 1998; Bowmaker et al., 1999; Bowmaker and Hunt, 1999). Phylogenetic analyses of opsin sequences have revealed that vertebrates have four classes of cone visual pigment—a long wavelength-sensitive (LWS), a medium wavelength-sensitive (MWS), a short wavelengthsensitive (SWS), and an extreme short wavelength-sensitive (UVS/VS) ‘‘UV’’ or ‘‘violet’’ pigment—in addition to a medium wavelength-sensitive rod visual pigment which is thought to have evolved from the MWS cone opsin (Okano et al., 1992; Yokoyama and Yokoyama, 1996; Heath et al., 1997; Yokoyama et al., 1998; Bowmaker et al., 1999; Bowmaker and Hunt,

164

INNES C. CUTHILL et al.

1999). All four cone visual pigment types are retained in birds (Bowmaker et al., 1999), but mammals have lost two of the ancestral vertebrate visual pigments—those based on the SWS and MWS opsins. It is only in the Old World, and some New World, primates that the mammalian complement of cone visual pigments has been modified to give distinct ‘‘green’’ and ‘‘red’’ sensitivity in addition to ‘‘blue,’’ thus establishing trichromacy. The current evidence is that the UVS/VS visual pigments of birds, reptiles, and fish cluster together with the shortwave sensitive blue cones of mammals in phylogenetic analyses of opsin gene sequences (Yokoyama and Yokoyama, 1996; Wilkie et al., 1998; Yokoyama et al., 1998; Bowmaker and Hunt, 1999). The mammalian ‘‘blue’’ cone is thus in fact a member of the UVS/ VS family of cone pigments. Interestingly, just as in birds, this cone in mammals shows a high degree of ␭max variability; for example, whereas the human ‘‘blue’’ cone has a ␭max of 앑420 nm (Dartnall et al., 1983), that of rats and mice is nearer 360 nm (Jacobs et al., 1991; Deegan and Jacobs, 1993; Yokoyama et al., 1998), firmly in the UV. The avian and human ‘‘blue’’ pigments are not homologous, nor are the avian and human ‘‘green’’ pigments. The primate ‘‘green’’ cone pigment is derived from the ancestral mammalian longwave (‘‘red’’) cone pigment, the sequence of which is similar to that of the avian ‘‘red’’ visual pigment gene (Wilkie et al., 1998). Visual pigment absorption spectra can be measured in situ, in the outer segments of single cells, using the technique of microspectrophotometry. Although the relatively small size of avian cone outer segments (typically 1.5 애m in diameter and 10 애m long), which imposes severe technical difficulties, has restricted the number of species studied (Table I), some generalizations can be made. Avian rods contain a medium wavelength-sensitive rhodopsin, with a ␭max between 501 and 509 nm, and the cells contain no oil droplet. The four classes of single cone contain, respectively, a LWS (␭max , 543–571 nm), a MWS (␭max , 497–509 nm), a SWS (␭max , 430– 463 nm), and either a VS (␭max , 402–426 nm) or UVS (␭max , 355– 376 nm) rhodopsin. Both members of the double-cone pair contain the LWS rhodopsin; the difference between the members is in their size and the oil droplet they possess. The photoreceptors of the starling Sturnus vulgaris (Hart et al., 1998) appear typical of passerines (Bowmaker et al., 1997; Hart, 1998). Absorption spectra of the cone visual pigments of this species are displayed in Fig. 1A. The domestic turkey Meleagris gallopavo (Fig. 1B; Hart et al., 1999), on the other hand, is typical of the galliform and anseriform species studied to date (Bowmaker et al., 1997; Hart, 1998). The main difference between these two groups is the spectral location of the ␭max of the extreme short wavelength-sensitive cone visual pigment, with the turkey having a VS cone pigment with a ␭max of 420 nm compared to a UVS pigment of ␭max

UV VISION IN BIRDS

165

362 nm in the starling. In each case, and for clarity only, mathematical functions known as visual pigment templates (Stavenga et al., 1993) have been used to model the mean absorptance spectra of each of the pigments. Raw data from spectrophotometric measurements can be fitted with such templates to determine the ␭max of the visual pigment (Dartnall, 1953; Knowles and Dartnall, 1977). For comparison, the visual pigments measured microspectrophotometrically in human photoreceptors (Dartnall et al., 1983) are displayed in Fig. 1C. Although the shortwave-sensitive pigment in the human retina would confer a similar degree of UV sensitivity to the violet-sensitive pigment of the turkey, such sensitivity is prevented largely by UV-absorbing pigments in the lens of the human eye (Goldsmith, 1991; Griswold and Stark, 1992). The ocular media of the turkey (cornea, aqueous humor, lens, and vitreous humor), and indeed the majority of avian species studied to date (Govardovskii and Zeuva, 1977; Emmerton et al., 1980; Hart, 1998; Hart et al., 1998, 1999), with the apparent exception of the mallard (Anas platyrhynchos; Jane and Bowmaker, 1988), have a relatively high transmission of short wavelengths and thus permit the VS (or UVS) pigment to confer considerable UV sensitivity. Nevertheless, the short wavelength limit to vision in birds is ultimately determined by the spectral transmission of the ocular media. Aromatic amino acids in lenticular proteins absorb strongly at short wavelengths, with 0.5 absorptance occurring between 320 and 350 nm (depending on lens diameter), and thus no bird is likely to be able to detect wavelengths shorter than approximately 310 nm (Douglas and Marshall, 1999). B. Oil Droplets All visual pigments, even those with a long wavelength ␭max , have considerable absorption at short wavelengths due to a secondary (웁) peak of the visual pigment absorption spectrum (Jacobs, 1992; Palacios et al., 1996), and current evidence shows that photons absorbed by the 웁 peak lead to phototransduction (Stark et al., 1994). In other words, an animal without intraocular UV-blocking filters would see UV by virtue of absorption by the 웁 peak of their longer wavelength-sensitive visual pigments. However, such a mechanism would be poor for color vision involving the ultraviolet. Because a visual pigment is ‘‘blind’’ to the wavelength stimulating it, UV stimulation could not be deconfounded from longer waveband stimulation and thus UV information could not constitute a separate dimension to color vision. As has been shown, in humans UV is filtered out by the lens; in birds this is achieved via colored oil droplets which lie within the cone cells, between the incident light and the visual pigment.

TABLE I Summary of Avian Visual Pigments Determined from Microspectrophotometry a

166

Visual pigment ␭max (nm) Subclass Neognathae

Order Anseriformes

Ciconiiformes

Columbiformes Galliformes

Specific name Anas platyrhynchos Anas platyrhynchos domesticus Anas platyrhynchos domesticus Puffinus puffinus Larus atricilla Buteo jamaicensis Spheniscus humboldti Columba livia Gallus gallus domesticus Coturnix coturnix japonica Meleagris gallopavo Pavo cristatus

English name

UVS/VS

SWS

MWS

LWS

Rod

Mallard duck1 Aylesbury duck1 Khaki Campbell duck1 Manx shearwater2 Laughing gull3 Red-tailed Hawk4 Humboldt penguin5 Feral pigeon2 Domestic chicken2 Japanese quail6 Domestic turkey7 Peacock7

415 415 426 402

452 449 456 452

506 501 501

567 570 570

505 504 505 505 508 501 504 506 506 505 504 504

560–575 403 409 419 418 420 421

450 453 455 450 460 457

507 508 505 505 505

543 567 570 567 564 566

Passeriformes

Neognathae Palaeognathae

Psittaciformes Strigiformes Struthioniformes

Tinamiformes

Leiothrix lutea Taeniopygia guttata Serinus canaria Corvus frugilegus Sturnus vulgaris Turdus merula Parus caeruleus Melopsittacus undulatus Strix aluco Dromaius novae-hollandiae Struthio camelus Rhea americana Nothoprocta cinerascens cinerascens Nothoprocta perdicaria sanborni

Pekin robin8 Zebra finch2 Canary9 Rook10 European starling11 Blackbird7 Blue tit7 Budgerigar2 Tawny owl12 Emu4 Ostrich13 Rhea13 Brushland tinamou4 Chilean tinamou4

앑355 360–380 369

453 430 444

앑362 376 374 371

449 454 449 444 463 445 445

501 503 500 497 504 504 503 509 503 510 510 ?498

567 567 571 565 563 557 563 564 555 567 570 570 564 566

501 507 506 504 503 505 503 509 503 502 505 505 504 501

167

a ␭max, wavelength of maximum absorbance. The system of classification of avian species is that given by Sibley and Monroe (1990) and is based on recent results from DNA–DNA hybridization. 1, Jane and Bowmaker (1988); 2, Bowmaker et al. (1997); 3, Liebman (1972); 4, Sillman et al. (1981); 5, Bowmaker and Martin (1985); 6, Bowmaker et al. (1993); 7, Hart (1998); 8, Maier and Bowmaker (1993); 9, Das (1997); 10, Bowmaker (1979); 11, Hart et al. (1998); 12, Bowmaker and Martin (1978); 13, Wright and Bowmaker (1998). LWS, MWS, SWS, and UVS/VS refer to the long, medium, short, and extreme short wavelength-sensitive cone visual pigments, respectively.

Fig. 1. Rhodopsin absorbance spectrum templates (Stavenga et al., 1993) representing visual pigments measured microspectrophotometrically in the single-cone photoreceptors of (A) European starlings (Sturnus vulgaris), (B) domestic turkeys (Meleagris gallopavo), and (C) humans (Homo sapiens sapiens). Mean values for the wavelengths of maximum absorbance (␭max) for the different visual pigment types are as follows: (A) European starling, UVS ⫽ 362 nm, SWS ⫽ 449 nm, MWS ⫽ 504 nm, and LWS ⫽ 563 nm (Hart et al., 1998); (B) domestic turkey, VS ⫽ 420 nm, SWS ⫽ 460 nm, MWS ⫽ 505 nm, and LWS ⫽ 564 nm (Hart et al., 1999); and (C) human, VS ⫽ 419 nm, MWS ⫽ 531 nm, and LWS ⫽ 558 nm (Dartnall et al., 1983).

UV VISION IN BIRDS

169

Oil droplets are highly refractile spherical organelles, which often appear brightly colored due to the incorporation of carotenoid pigments (Wald and Zussman, 1937; Meyer et al., 1965; Goldsmith et al., 1984). Photoreceptor oil droplets occur in many vertebrates, including a few fish, amphibians, reptiles, and even marsupials (Walls, 1963; Goldsmith, 1990), but in marsupials and nocturnal birds they are not densely pigmented. In groups with densely pigmented oil droplets, such as birds, their function appears to be to filter the light entering the cones (Bowmaker, 1980; Goldsmith et al., 1984; Partridge, 1989; Kawamuro et al., 1997; Vorobyev et al., 1998). Each visual pigment is twinned with a particular oil droplet type which cuts out wavelengths somewhat less than the ␭max of the visual pigment (Goldsmith et al., 1984; Bowmaker, 1991b; Bowmaker et al., 1997; Hart, 1998; Hart et al., 1998). Thus, in single cones, the LWS pigment is paired with a red droplet (R type), the MWS with a yellow (Y type), the SWS with a UV-blocking clear droplet (C type), and the UVS/VS pigment with a fully transparent droplet (T type) which shows no significant absorption over the avian visible spectrum (Goldsmith et al., 1984). The net effects of the pigmented oil droplets are to increase the effective ␭max of the longer wavelength cones, to narrow the waveband to which each cone is sensitive, and to reduce the overlap in their spectral sensitivities. Figure 2A shows the absorption spectra of the oil droplets found in the single cones of the European starling (Hart et al., 1998). Their effects on the spectral sensitivity of the starling’s cone cells are illustrated in Fig. 2B, as compared with Fig. 1A. The peak effective spectral sensitivity of the MWS and LWS cones is increased by approximately 40 nm, and there is a striking difference in overlap compared to the human ‘‘red’’ and ‘‘green’’ cone spectral sensitivities (Fig. 1C). The suggested advantages of photoreceptors with narrowed spectral sensitivity and reduced overlap in waveband, as in birds, is increased color saturation and enhanced discrimination of certain classes of spectra (Govardovskii, 1983; Goldsmith, 1990; Vorobyev et al., 1998) as well as improved color constancy (Osorio et al., 1997; Vorobyev et al., 1998). Color constancy is the phenomenon whereby objects are perceived as the same color despite changes in the illuminant. This is highly desirable when, for example, the spectral quality of incident light changes as the sun moves behind a cloud or the animal moves into shade. These advantages of oil droplets are obtained at the cost of reduced quantal catch (particularly in the longwave cones, in which the densely pigmented oil droplets block a high proportion of the incident light), so the system is most appropriate in animals active at high light intensities. Reduced overlap in spectral sensitivities of photoreceptors may reduce discriminability of the sort of monochromatic lights used in psychophysical tests, but narrowed spectral tuning is well suited to discrimination of natural objects, many of

170

INNES C. CUTHILL et al.

Fig. 2. (A) Mean absorptance spectra of oil droplets found in the single-cone photoreceptors of the European starling, Sturnus vulgaris. T-type (transparent), C-type (colorless), Ytype (yellow), and R-type (red) oil droplets are found in the inner segments of the UVS/VS, SWS, MWS, and LWS single cones, respectively. Mean values for the cutoff wavelengths (␭cut) of the C-type, Y-type, and R-type droplet types, as defined by Lipetz (1984), are 399, 514, and 571 nm, respectively. The T-type droplet displays no significant absorptance over the range of wavelengths measured. (B) Predicted effective spectral sensitivities (expressed as absorptance) of the four types of single-cone photoreceptors found in the retina of the starling. Spectral sensitivity is calculated by multiplying the absorptance of the visual pigment with the transmission (1 - absorptance) of the oil droplet with which it is associated and the transmission of the ocular media. It is evident by comparison with Fig. 1A that the effective spectral sensitivity function of a given cone is much narrower than that of the visual pigment it contains and, at least for the MWS and LWS cones, the peak is shifted by approximately 40 nm toward longer wavelengths. Wavelengths of peak sensitivity for the UVS, SWS, MWS, and LWS single cones are 371, 453, 543, and 605 nm, respectively.

UV VISION IN BIRDS

171

which have step-function changes, or wide peaks and troughs, of reflectance (see Vorobyev et al., 1998). Modeling the separation, in avian color space, of natural reflectance spectra from a diversity of avian plumage colors, Vorobyev et al. (1998) concluded that tetrachromacy and possession of oil droplets each improved the discriminability of these colors, as compared to trichromatic systems. C. Interspecific Variation The cone visual pigment ␭max values measured in the species studied to date are displayed in Fig. 3. Interestingly, the largest variation in recorded ␭max value within a visual pigment class occurs in the UVS/VS type. To some extent, this is an artifact of the increased difficulty in measuring this visual pigment type microspectrophotometrically, with UVS/VS cone outer segments being generally the smallest of all the photoreceptor types. Furthermore, most microspectrophotometers utilize a light source with relatively low emission at short wavelengths which reduces the signal to noise ratio of the recordings and, therefore, the accuracy with which the ␭max can be estimated. Nevertheless, it appears that there are at least two, and

Fig. 3. Distribution of avian cone visual pigment ␭max values for all species measured microspectrophotometrically. Values approximately 565, 505, 450, 420, and 370 nm represent LWS, MWS, SWS, VS, and UVS single-cone visual pigments, respectively. References are the same as in Table I.

172

INNES C. CUTHILL et al.

probably three, possible spectral locations for this class of cone visual pigment (see Table I and references therein): in the UV spectral region (␭max , 앑355–376 nm), in the most short wavelength region of the violet spectral range (␭max , 400–410 nm), and at slightly longer wavelengths (␭max , 415–421 nm). The wider range of avian taxa possessing a violet pigment, and the fact that two paleognathus species also possess the violet receptor (Wright and Bowmaker, 1998), suggests that this is the ancestral state for birds. However, due to the paucity of data, even this conclusion is premature. There is variation in the ␭max of violet receptors of nonpasserines, and recent data suggest that the pigeon actually has a near-UV receptor, with ␭max 393 nm (Yokoyama et al., 1998). Furthermore, the ␭max of other vertebrates’ UVS/VS receptors can also be in either the UV or the violet waveband (e.g., rat, 358 nm; chameleon, 358 nm; Xenopus, 425 nm; data based on regeneration of visual pigments from cDNA; Kawamura and Yokoyama, 1996; Starace and Knox, 1998; Yokoyama et al., 1998). Thus, although these different UVS/VS opsins are thought to have evolved from the same ancestral opsin type (Yokoyama et al., 1998), it is not clear which evolved first, either in birds or in vertebrates as a whole. The apparent absence of a UVS/VS cone class in the tawny owl, Strix aluco (Bowmaker and Martin, 1978), may reflect an adaptation to nocturnal/ crepuscular activity (Martin, 1990). It may, however, be an artifact of the relatively low abundance of all cone types in the retinae of nocturnal birds compared to strongly diurnal species (Bowmaker and Martin, 1978; Braekevelt, 1993; Braekevelt et al., 1996) and the difficulties in exhaustively sampling the retinal mosaic by microspectrophotometry. The UVS/VS cone class is often the least abundant in the retina (Goldsmith et al., 1984; Hart, 1998; Hart et al., 1998; Wilkie et al., 1998) and we have to consider that it may have been overlooked in microspectrophotometric examination. This may also explain the apparent absence of the MWS and LWS rhodopsins of the Manx shearwater, Puffinus puffinus, and the MWS rhodopsin of the Humboldt penguin, Spheniscus humboldti (Table I). With the exception of the UVS/VS cone class, there appears to be little variation in the ␭max of the different visual pigment types and in the spectral transmission characteristics of the oil droplets with which they are associated. This is surprising given the apparent variety in the visual ecology of the bird species studied to date, but insect visual pigments also appear to be remarkably conservative across a range of lifestyles (Peitsch et al., 1992; Chittka, 1996, 1997). There are, however, significant variations in the relative abundance of the different cone types topographically within the retinae of a given species (Goldsmith et al., 1984; Partridge; 1989; Hart, 1998; Hart et al., 1998) and between the retinae of different species (Peiponen, 1964; Muntz, 1972; Partridge, 1989). In several species, such as the feral pigeon

UV VISION IN BIRDS

173

Columba livia (Waelchli, 1883), starling (Hart et al., 1998), peacock Pavo cristatus, and blackbird Turdus merula (Hart, 1998), all types of single cones are generally more abundant in the downwards and forwards-looking [posterior–dorsal (PD)] region of the retina relative to double cones. More subtle differences may exist, however, and in the starling a relatively higher proportion of UVS cones occurs in the PD retina, which has a reduced proportion of LWS cones (Hart et al., 1998). Intraretinal variations in relative cone abundance are probably correlated with both visual ecology and optical aspects of the eye. Because of the position of the eyes in these species, for instance, only the PD region of the retina can be used for binocular vision (Martin, 1986). Interspecific variations in relative cone abundance are thought to reflect visual ecology to a greater extent than phylogeny (Muntz, 1972; Partridge, 1989). For example, bird species that fly above water and need to look through its surface from a distance have a much higher proportion of MWS and LWS single cones than species which live on the water but do not need to look through it (Muntz, 1972). Such adaptations may increase visual contrast in some spectral regions but, as shown by recent mathematical models (Vorobyev and Osorio, 1998), photopic spectral sensitivity is only predicted when single-cone spectral sensitivities are integrated in models that consider their relative abundance in the retina. It is thus possible that changes in the relative abundance of the different cone types are potentially as important an adaptation for determining photopic spectral sensitivities in birds as variation in cone spectral absorption characteristics. This also implies that intraretinal topographical variations in relative cone abundance will result in differences in the photopic spectral sensitivity between retinal regions. D. Double Cones One of the most intriguing features of the avian retina is the abundance of double cones, a cone class also found in fish and turtles (Liebman and Granda, 1971; Ohtsuka, 1985) but lacking in mammals. Each pair occupies four times the cross-sectional area of a single cone and the double cones constitute approximately half of the photoreceptor population in diurnal species (Waelchli, 1883; Meyer, 1977; Goldsmith et al., 1984; Jane and Bowmaker, 1988; Hart, 1998; Hart et al., 1998) and yet their function is unclear. The larger, principal member contains an oil droplet (P type) with a variable cutoff wavelength, between approximately 410 and 500 nm depending on retinal location (Goldsmith et al., 1984; Bowmaker et al., 1997; Hart, 1998; Hart et al., 1998). The accessory member occasionally displays a tiny oil droplet (A type) or simply a low concentration of short

174

INNES C. CUTHILL et al.

wavelength-absorbing, carotenoid-like pigment at the distal tip of the inner segment (Bowmaker et al., 1997; Hart et al., 1998). However, the function of double cones is unclear. A behavioral measure of photopic spectral sensitivity in the Pekin robin, Leiothrix lutea, appeared to show no contribution by the double cones, with peaks in sensitivity corresponding to the corrected spectral sensitivities of the four single-cone types (Maier and Bowmaker, 1993). Nevertheless, electroretinographically determined photopic spectral sensitivity functions are dominated by a broad peak at approximately 570 nm (Blough et al., 1972; Chen and Goldsmith, 1986) which corresponds to the peak effective spectral sensitivity of the double cones. This mismatch suggests that the neural signal from the double cones is not involved in color discrimination, at least under the conditions used for the behavioral test of photopic spectral sensitivity (Maier and Bowmaker, 1993). This conclusion is in agreement with the models of spectral sensitivity of Vorobyev and Osorio (1998), but not all data are consistent with this view (Palacios and Varela, 1992). It has been suggested that double cones may be involved in the detection of polarized light (Young and Martin, 1984; Cameron and Pugh, 1991), which is used by both vertebrates and invertebrates for orientation and navigation (Brines and Gould, 1982), or adapted for movement detection (Campenhausen and Kirschfeld, 1998), but strong empirical evidence for their function is lacking.

E. The Dimensionality of Avian Color Space The notion of a ‘‘color space’’ relates to a graphical or geometric representation of cone interactions (Burkhardt, 1989; Goldsmith, 1990; Chittka, 1992; Neumeyer, 1992; Thompson et al., 1992). The human sense of color derives from comparison of the activity of three cone types; therefore, one can represent the color of an object as a point in three-dimensional space, with each axis corresponding to the photon catch (and thus output) of the three receptor types. Because hue relates to the relative rather than absolute cone output, the hue of an object can be represented as a point in a twodimensional plane cutting the three-dimensional receptor space, a ‘‘color triangle’’ with the vertices representing the relative blue, green, and red cone stimulation (Goldsmith, 1990). Perceived color differences do not correspond directly to distances between points in such a color space, but the latter provide a first approximation if psychophysical data are lacking (Goldsmith, 1990; Thompson et al., 1992). A tetrachromatic receptor space would have four dimensions, and therefore tetrachromatic hues can be represented as points in a three-dimensional space within this hypervolume, with a tetrahedron being the direct equivalent of the human color triangle

UV VISION IN BIRDS

175

(Burkhardt, 1989; Goldsmith, 1990; Neumeyer, 1992; Thompson et al., 1992). Birds can make accurate visual discriminations on the basis of the spectral radiance of observed objects independent of brightness and therefore have color vision (Osorio et al., 1999). The presence of six types of retinal cones confers the potential for hexachromatic (six-dimensional) color space but, as previously shown, if only the single cones are involved then tetrachromacy is more likely. However, it is possible that bird color space is less than tetrachromatic. For example, if the only neural comparisons were between the UVS and SWS cones, and the MWS and LWS cones, then many theoretical tetrachromatic hues would not be realized. To demonstrate the dimensionality of a color vision system requires experiments in which the equivalence of monochromatic light mixtures is compared (e.g., can yellow be simulated by an appropriate mix of monochromatic green and red lights?). Using such techniques, Neumeyer and colleagues demonstrated tetrachromacy in goldfish (Neumeyer and Arnold, 1989; Neumeyer, 1992), and parts of the pigeon’s color space have been established (Palacios and Varela, 1992); therefore, avian tetrachromacy remains the most likely hypothesis. Interestingly, the dimensionality of goldfish color space shifts with changes in illumination (Neumeyer and Arnold, 1989), being tetrachromatic at high light levels and trichromatic at reduced intensities. In birds, this is another level of complexity of avian vision which remains to be explored. An interesting consequence of higher than two-dimensional color vision is the existence of so-called nonspectral colors. These are colors which are not part of the rainbow and cannot be simulated by any monochromatic light source; they arise from stimulation of two or more photoreceptors which are sensitive to nonadjacent wavebands. Purple is the only human nonspectral color, perceived when red and blue cones are stimulated (i.e., purple objects are those which reflect long- and short-, but not medium-, wave light). A tetrachromat, with four receptor types being compared, might have five nonspectral colors. These would correspond to the three possible two-way combinations (UVS/VS ⫹ MWS, UVS/VS ⫹ LWS, and SWS ⫹ LWS) and two possible three-way combinations (UVS/VS ⫹ SWS ⫹ LWS and UVS/VS ⫹ MWS ⫹ LWS) of receptors sensitive to nonadjacent wavebands of light (Burkhardt, 1989; Goldsmith, 1990, 1994; Thompson et al., 1992). If birds are truly tetrachromatic then the nonspectral colors are one of the main reasons why it is particularly hard to achieve a subjective impression of a bird’s view of objects’ hues. We can gain insight into a bee’s color world by modifying a video camera to encode UV, blue, and green wavelengths and outputting to a regular television monitor (Eisner et al., 1969; Loew and Lythgoe, 1985). In this way the color world of the bee (encoded as the relative amounts of UV, blue, and green wave-

176

INNES C. CUTHILL et al.

bands) can be mapped directly to the color world of the human (encoded as the relative amounts of blue, green, and red wavebands). ‘‘Bee purple’’ (a mixture of UV and green; Chittka et al., 1992, 1994) would be represented as human purple (Eisner et al., 1969). However, there is no simple way to represent a tetrachromatic color space to a trichromat because the dimensionalities differ. We can, however, describe both the reflectance spectra and their likely processing by the bird’s visual system in ways that can be analyzed quantitatively.

III. Studying Color Nonanthropocentrically Reflectance spectra are the raw signals which, in combination with light environment and receiver sensory processing, give rise to the receiver’s color perception. Reflectance spectra are the invariant features which we expect perceptual systems to be designed to extract. Therefore, a logical approach to the objective analysis of ‘‘color’’ is to analyze reflectance spectra. The quotation marks are a reminder that color per se can never be described objectively because it is a perceptual construct. To understand the evolution of colors which are under selective pressures from visual systems, and vice versa, analyses of reflectance spectra alone will never be enough. A major aim must be to investigate the color perception of the receiver and map the relevant spectra into the color space of the animal concerned. This has been done for very few species, most notably bees (Menzel et al., 1991; Chittka and Menzel, 1992; Chittka et al., 1994; Lunau and Maier, 1995; Chittka, 1996, 1997; Lunau et al., 1996; Chittka and Waser, 1997; Vorobyev et al., 1997). Bees are an attractive subject for this approach because not only is their color perception very well characterized but also, in sterile workers at least, the major visual tasks for which their eye is designed are clear: flower location and celestial navigation. Birds are sexual and longer lived, so they probably have a wider range of important visual tasks to perform. This makes it harder to specify a priori what their eyes may be designed to do. Furthermore, the colors of a particular species’ plumage are likely to have evolved under selective pressures from more than one type of color vision system: for example, conspecific tetrachromats and dichromatic mammalian predators. Thus, there will always be value in describing spectral reflections in objective terms as well as with reference to the color space(s) of potential receivers. Features of reflectance spectra, such as the position and steepness of peaks and cutoffs, seem to reliably correlate with psychometric measures of human color perception (Endler, 1990); therefore, the ␭max values of peaks, or cutoff positions of step functions, can be used to describe and

UV VISION IN BIRDS

177

categorize spectra. Likewise, relatively simple spectra can be classified according to the relative amounts of light within broad wavebands (Endler’s segment method; Endler, 1990; Zuk and Decruyenaere, 1994). However, an increasingly common approach is to use principal component analysis (Chatfield and Collins, 1995), which can be applied to spectra of any complexity. This can reduce the large multivariate data set that a set of measured spectra comprises (e.g., each spectrum could be 400 dependent variables, with the reflectance at 1-nm intervals from 300 to 700 nm) to a few orthogonal variables. More than 90% of the variation in shape of even quite complex spectra can often be reduced to two or three principal components (Hurlbert, 1986; Endler, 1990; Endler and The´ry, 1996; Bennett et al., 1997; Hunt et al., 1998; Ruderman et al., 1998; Cuthill et al., 1999; Langmore and Bennett, 1999). These summary variables can then be used in further analyses to test for differences in spectral shape between, for example, the sexes or other groups of interest (Bennett et al., 1997; Hunt et al., 1998; Langmore and Bennett, 1999). Cuthill et al. (1999) describe a protocol for the measurement of avian reflectance spectra and for the analysis of such data. In subsequent sections on the role of UV in foraging and signaling, examples of reflectance spectra of both prey and plumage are presented.

IV. The Functions of UV Vision Avian photoreceptors and their retinal organization superficially resemble that found in the retinae of some diurnal reptiles, such as freshwater turtles (Liebman and Granda, 1971, 1975; Bowmaker, 1991a,b). Consequently, caution must be exercised in ascribing any theory regarding the specific characteristics of the photoreceptors of the avian retina exclusively to birds. It is also important to remember the distinction between origins and maintenance in discussions of the adaptive significance of avian UV vision (Bennett and Cuthill, 1994). As shown, it is probable that possession of a UV-sensitive cone class and tetrachromacy are the ancestral states for tetrapods and perhaps for vertebrates (Bowmaker, 1991a; Jacobs, 1992; Yokoyama and Yokoyama, 1996; Yokoyama et al., 1998; Bowmaker and Hunt, 1999). Thus, we should not look to factors peculiar to avian ecology, or even the terrestrial environment, for the selective pressures favoring the initial evolution of a UV photoreceptor. Certainly, UV cones are found in some fish (Avery et al., 1983; Ha´rosi and Hashimoto, 1983; Chen and Stark, 1994; Fratzer et al., 1994). An explanation of the ultimate origins of the UV receptor in some aquatic ancestral vertebrate is beyond the scope of this review, but a consideration of the role which UV vision serves in modern birds may shed light on its function in other groups, and this

178

INNES C. CUTHILL et al.

role is of central importance to avian behavioral ecology. Two extreme possibilities can be considered. The UV cone may function in isolation rather than being part of a ‘‘true color’’ system because the cone output is not being compared at the neural level with the output of any other receptor type (Jacobs, 1981). Such a role need not even involve resolution of a visual image, for example, if, as has been suggested, UV wavelengths are involved in the entrainment of circadian rhythms (Pohl, 1992). However, as previously discussed, available evidence indicates that the UV cone is part of a true color vision system. The other extreme is the possibility that UV vision serves no special role but is simply part of a general-purpose color vision system that happens to be tetrachromatic rather than trichromatic. In between these extremes, there is the possibility that the UV cone is involved in a color channel used for specific visual tasks (see discussion of wavelength-dependent behaviors in Goldsmith, 1990). Wavelength-dependent properties of light may predispose the ultraviolet for use in some visual tasks (Bennett and Cuthill, 1994). These may influence the uses to which UV light can be put by an animal. First, UV light is scattered more than light of longer wavelengths. Similar effects at the short wavelength end of the human-visible spectrum make the sky look blue (Lythgoe, 1979), and this is why UV filters on cameras are effective at reducing ‘‘haze’’ at a distance. For similar reasons, short wavelengths are prone to scattering by imperfections in the animal’s optical media (Lythgoe, 1979). Thus, UV wavelengths may be less useful for object detection/recognition/assessment at a distance or particularly good for signals which are intended for nearby receivers while remaining indistinct for distant receivers (Bennett and Cuthill, 1994). Second, scattering by small particles in the atmosphere also induces plane polarization, the degree of which is, like scattering, inversely proportional to the fourth power of the wavelength (Rayleigh’s law; Lythgoe, 1979). Polarization of UV is therefore greater than that for longer wavelengths, a fact that insects exploit by using their UV receptors in polarization detection (Brines and Gould, 1982; Wehner, 1989). Whether birds utilize a similar mechanism is more contentious (see Section IV,A). Another consequence of wavelength-dependent scattering is that near dawn and dusk a high proportion of the light available to animals is of short wavelengths (Lythgoe, 1979; Endler, 1993). UV wavelengths may therefore be particularly useful for visually mediated tasks at these times of day. Therefore, there are good reasons why UV may be predisposed for certain sorts of visual tasks. However, it is probably no more meaningful to talk of ‘‘UV vision’’ in birds than to single out ‘‘blue vision’’ in humans. Of the plausible functions, attention has focused on orientation, foraging, (prey detection), and signaling (e.g., in mate choice).

UV VISION IN BIRDS

179

A. UV Vision and Orientation Perhaps surprisingly, given the research effort directed at avian navigation and orientation, there have been few experiments investigating the relative importance of different wavelengths of light. In bees, UV and polarization vision are linked because it is the intricate arrangement of UV receptors in the eye that allows detection of the pattern of polarized light in the sky (Wehner 1989). In birds, most experimental effort has been devoted to determining whether polarization patterns per se in the sky are used in orientation and how these interact with other compass information (Kreithen and Keeton, 1974; Phillips and Moore, 1992; Able and Able, 1993, 1995; Coemans et al., 1994a; Munro and Wiltschko, 1995; Vos Hzn et al., 1995). However, polarization vision need not be linked to UV receptors (Hawryshyn and McFarland, 1987; Cameron and Pugh, 1991) and the UV cones could be involved in orientation, irrespective of polarization sensitivity, through detection of color gradients in the sky which relate to the sun’s position (Coemans et al., 1994b). In support of this, it has been noted that there are more short-wavelength receptors in the ventral (skyward-looking) surface of the pigeon’s retina and the cones here are relatively dispersed (Bowmaker, 1977; Vos Hzn et al., 1994), which suggests panoramic vision rather than fine spatial resolution (Vos Hzn et al., 1994). However, not all birds show this pattern; indeed, starlings have more UV cones in the posterior–dorsal (down and forward-looking) region of the retina (Hart et al., 1998). Light is also implicated in magnetic orientation and its effects seem wavelength dependent in a variety of taxa, both vertebrate and invertebrate (Phillips and Sayeed, 1993; Wiltschko et al., 1993; Phillips and Borland, 1994). Although magnetite has long been invoked in magnetoreception (for a general review, see Wiltschko and Wiltschko, 1993), there is also evidence of a light-dependent magnetic compass, presumably involving the visual system. Amongst vertebrates, this was first shown to be wavelength dependent in newts (Phillips and Borland, 1992, 1994), in which orientation follows the normal pattern under blue, but not red, light. Red light has also recently been shown to interfere with the magnetic compass in birds (Wiltschko et al., 1993; Wiltschko and Wiltschko, 1995; R. Wiltschko and W. Wiltschko, 1998; Munro et al., 1997). Whether UV is as effective or more effective in enabling correct magnetic orientation than blue light, and whether there are differences between species with violet or UV visual pigments, remains to be investigated. In the following sections, we concentrate on those hypothesized functions of UV vision which have been more thoroughly investigated experimentally, namely, foraging and signaling.

180

INNES C. CUTHILL et al.

B. UV Vision and Foraging Some of the earliest demonstrations of UV vision in birds were in hummingbirds (Huth and Burkhardt, 1972; Goldsmith, 1980; Goldsmith et al., 1981). This was perhaps a natural choice given the importance of UV vision for bee pollinators (Chittka and Menzel, 1992; Chittka et al., 1994; Giurfa et al., 1995; Kevan et al., 1996; Lunau et al., 1996; Vorobyev et al., 1997). Surprisingly, the role of UV cues in natural foraging by hummingbirds has not been investigated. This may be because red coloration is commonly found in bird-pollinated flowers and, due to bees’ relative lack of sensitivity at long wavelengths, it has been postulated (Raven, 1972) that red represents a ‘‘private channel’’ for flowers to signal to birds but not bees [see also Porch (1931) cited in Lunau and Maier (1995)]. UV patterns also appear to be more common among insect-pollinated than bird-pollinated flowers (Silberglied, 1979). However, bees are not truly ‘‘red blind’’ (Chittka and Waser, 1997); many bird-pollinated flowers are not red, and some reflect UV (Lunau and Maier, 1995). Indeed, some red flowers also reflect strongly in the UV (Burkhardt, 1982). Thus, although there may be no innate preferences for particular flower colors (see references in Lunau and Maier, 1995), the role of UV in flower visitation by birds needs to be reassessed. In the first study to demonstrate that wild birds use UV wavelengths while foraging, Viitala et al. (1995) showed that kestrels (Falco tinnunculus) used UV cues to detect active vole trails. Using reflectance spectrophotometry, they demonstrated that the scent marks of male voles contrasted strongly with their background in the UV but not in the human-visible spectrum. Field experiments showed that raptors tended to hunt in areas containing artificially created vole trails (urine- and feces-soaked straw) rather than in areas in which water-soaked straw trails had been created or in which no trails were present (Viitala et al., 1995). In a lab experiment, the kestrels spent more time scanning and more time in the vicinity of arenas with added vole scent marks, but only under UV illumination and not when illumination was restricted to human-visible light (Fig. 4). The overall implication of the work of Viitala et al. (1995) is that vole trails, and hence areas of high vole abundance, are more detectable by kestrels when UV information is present. In similar laboratory experiments on adults and juveniles of another major predator of voles, Tengmalm’s owl (Aegolius funereus), Koivula et al. (1997) found no preference for arenas containing vole scent marks under UV illumination. This is not surprising since A. funereus is a nocturnal predator and owls may lack UV cones (Bowmaker and Martin, 1978). Despite the absence of microspectrophotometric or electrophysiological studies on UV sensitivity in birds of prey, the behavioral evidence from

UV VISION IN BIRDS

181

Fig. 4. (A) Mean number of scans (⫾SD) of different arenas and (B) mean time (min, ⫾SD) spent by kestrels above arenas differing in illumination and the presence or absence of vole scent marks. The four treatments were dry vole trails in UV light, clean arena in UV light, dry vole trails in visible (VL) light, and clean arena in visible light (data replotted with permission from Viitala et al., 1995. Copyright 䉷 1995 MacMillan Magazines Ltd.).

the kestrel (Viitala et al., 1995) strongly suggests that diurnal raptors can detect UV wavelengths. Since many bird species possess plumage which reflects in the UV (see Section IV,C), and many common backgrounds such as leaves, bark, and soil do not (Chittka et al., 1994), it is likely that raptors will use both UV and human-visible cues to detect their avian prey. Furthermore, there may be costs to birds which have conspicuous plumage in terms of increased predation risk (Johnson, 1991; Go¨tmark and Hohlfalt, 1995), although bright colors can also potentially reduce predation risk (Go¨tmark, 1992, 1996, 1997; Go¨tmark and Unger, 1994; Go¨tmark and Olsson, 1997) and it has been suggested that brightly colored birds may actually be advertising unpalatability or unprofitability to avian predators (Baker and Parker, 1979; Hasson, 1991; Dumbacher et al., 1992; Go¨tmark, 1994).

182

INNES C. CUTHILL et al.

Unfortunately, the vast majority of studies of avian conspicuousness do not consider the UV (Bennett et al., 1994), instead relying on humansubjective assessments (for exceptions, see Go¨tmark and Hohlfalt, 1995; Endler and The´ry, 1996; Go¨tmark, 1996, 1997; Go¨tmark and Olsson, 1997). This argument can apply to birds with apparently drab plumage: Some species which look almost black to humans in fact reflect very strongly in the UV (see Section IV,C). Many bird-dispersed fruits are red or black (Wheelwright and Janson, 1985; Snow and Snow, 1988; Willson et al., 1990; Willson and Whelan, 1990). However, Burkhardt (1982) noted that the waxy bloom that develops on some fruits, such as sloes (Prunus spinosa; Fig. 5A), enhances UV

Fig. 5. (A) Reflectance from 300 to 700 nm of sloe berries (Prunus spinosa) demonstrating high UV reflectance with waxy bloom intact (1) but not with bloom removed (2). (B) Reflectance changes in the UV and human-visible spectrum of ripening yew berries (Taxus baccata) as the berries change from ‘‘green’’ (1) to ‘‘orange’’ (2) and then ‘‘red’’ (3) (S. Hunt, unpublished data).

UV VISION IN BIRDS

183

reflectance. He postulated that this might increase their contrast against foliage and make them more detectable to birds. Although this may be true to some extent, such blooms tend to enhance reflectance at all wavelengths, making the fruit brighter rather than producing a specific UV-rich hue (Willson and Whelan, 1989; S. Hunt, unpublished data). Indeed, because the bloom adds ‘‘white’’ to the underlying hue of the fruit, the resulting colors are less saturated than in the fruit without its bloom (Willson and Whelan, 1989). In the only published experimental studies to investigate whether the bloom enhances detection by, or affects the preferences of, wild birds, there was no obvious effect (Willson and Whelan, 1989; Allen and Lee, 1992). However, as the authors pointed out, the fruit in these trials were presented as artificial clusters and not always against a leaf background. Although UV reflectance appears to be relatively low in species which lack blooms (Burkhardt, 1982; S. Hunt, unpublished data), blooms are not necessarily a prerequisite for UV reflectance. For example, yew berries (Taxus baccata; Fig. 5B) have a reflectance peak in the UV despite lacking any obvious bloom. As the berry matures from green to red, UV reflectance declines. Thus, both UV and human-visible cues may provide avian frugivores with information regarding fruit quality or ripeness. Despite the widespread interest in the role of color in avian frugivory (Darwin, 1859; Ridley, 1930; Osche, 1983; Mason et al., 1984; Wheelwright and Janson, 1985; Owens and Prokopy, 1986; Willson et al., 1990; Willson and Whelan, 1990; Allen and Lee, 1992; Fischer and Chapman, 1993; Murray et al., 1993; Sallabanks, 1993; Lee et al., 1994; Whelan and Willson, 1994; Avery et al., 1995; Sanders et al., 1997), very few studies have systematically investigated the influence of hue, brightness, saturation, or background contrast in fruit selection (Puckey et al., 1996), and far less have studied the role of UV. Invertebrates are important prey items for birds, and many insects, particularly Lepidoptera, reflect in the UV (Eisner et al., 1969; Silberglied, 1979, 1984; Eguchi and Meyer-Rochow, 1983; Meyer-Rochow and Eguchi, 1983; Meyer-Rochow 1991; Meyer-Rochow and Ja¨rvilheto, 1997). However, whether birds actually utilize these UV patterns in prey detection or recognition is virtually unknown. In the first study to experimentally investigate the importance of UV for birds hunting cryptic insect prey, Church et al. (1998a) examined the behavior of blue tits (Parus caeruleus) searching for green cabbage moth (Mamestra brassicae) or winter moth (Operophtera brumata) caterpillars in a lab arena. The latency to find the first prey item increased in trials in which UV wavelengths were removed from the illuminating light. This effect was most pronounced when the UV contrast between prey and background was greatest, namely, M. brassicae against

184

INNES C. CUTHILL et al.

a cabbage leaf (Fig. 6). However, the reduction in performance was transient because the birds soon appeared to switch to other foraging cues besides color mismatch between prey and background (Church et al., 1998a). It is perhaps unwise to interpret this experiment as demonstrating that blue tits locate prey using UV cues per se. The detrimental effect of removing UV could be via alteration of the hue of both prey and background, and it is the combined color change which renders the task more difficult. Although there is evidence that birds are able to discriminate between different wavelengths in the UV (Emmerton and Delius, 1980), prey detectability might also be lowered due to a reduction in perceived brightness in the absence of UV (Church et al., 1998b). Since short wavelengths account for a high proportion of the spectral composition of ambient light at dawn and dusk (Endler, 1993), it is possible that UV sensitivity allows diurnal birds to extend the period during which they may forage for prey. The use of UV cues for detecting insect prey by birds has important implications for our understanding of protective coloration (i.e., crypsis, aposematism, and mimicry). In order to be cryptic to birds, and not just to the human eye, insects should match their visual backgrounds throughout the avian-visible spectrum. Where reflectance spectra of cryptic caterpillars have been measured, this is often, but not always, the case. Stalker (unpublished data, cited in Majerus, 1998) demonstrated that the degree of color background matching of the melanic and nonmelanic forms of the peppered moth Biston betularia depends to a large extent on the amount of UV reflected by different types of lichens on tree bark. Church et al. (1998a)

Fig. 6. Difference in latencies (⌬ log (s), mean ⫾ SE) to find first prey item under UV⫹ (human-visible plus UV) and UV⫺ (human-visible only) illumination by blue tits searching for cabbage moth (Mamestra brassicae) caterpillars on cabbage and black paint backgrounds. Positive values indicate that the blue tits find prey faster when UV cues are present (data derived from Church et al., 1998b).

UV VISION IN BIRDS

185

showed that the spectral characteristics of lepidopteran larvae of oak tend to match those of either leaves or twigs in the UV (Fig. 7). However, in one species, the grey shoulder knot (Lithophane ornitopus), maximum reflectance was in the UV (Fig. 8). It is therefore hypothesized that the gray shoulder knot is aposematic rather than cryptic and is advertising distastefulness to birds via a UV communication channel (Church et al., 1998a). These data also have implications for our understanding of higher order perceptual processes in birds. For example, experiments designed to test whether birds form search images or rely on search rate modification to detect cryptic prey (Gendron, 1984, 1986; Lawrence, 1985a,b; Guilford and Dawkins, 1989; Reid and Shettleworth, 1992; Plaisted and Mackintosh, 1995; Langley et al., 1996) depend largely on producing accurate and ecologically relevant measurements of prey crypsis. For avian prey detection tasks involving color (c.f. Alatalo and Mappes, 1996), this is impossible to achieve if UV wavelengths are not considered. Despite the large interest in the design features of aposematic colors (Guilford, 1986; Guilford and Dawkins, 1991; Marples and Roper, 1996; Rowe and Guilford, 1996; Roper and Marples, 1997a,b), these color patterns have likewise only been assessed from a human perspective. It is likely that aposematic patterns that are conspicuous and highly contrasting to us (e.g., red or yellow on black) will be equally or more conspicuous to birds because the avian eye is well designed for long-wavelength discriminations (Emmer-

Fig. 7. Reflectance spectra in the wavelength range 300–700 nm of (1) the grey shoulder knot (Lithophane ornitopus), (2) the winter moth (Operophtera brumata), and (3) the underside of a common oak (Quercus robur). Although the reflectance characteristics of the winter moth, like most ‘‘green’’ caterpillars of oak, are very similar to those of its oak leaf background, the gray shoulder knot contrasts strongly, particularly in the UV (data derived from Church et al., 1998a).

186

INNES C. CUTHILL et al.

Fig. 8. Still video images of the grey shoulder knot (A) under UV illumination (300– 400 nm) and (B) in the human-visible spectrum (400–700 nm), illustrating increased contrast against an oak leaf in the UV (S. C. Church, unpublished data).

ton and Delius, 1980; Goldsmith et al., 1981; Vorobyev and Osorio, 1998; Vorobyev et al., 1998). In addition to the possibility of UV-aposematic patterns in species that are cryptic to humans (Church et al., 1998a), some aposematic butterflies which are seen as red by humans often have a strong UV component (Crane, 1954; Eguchi and Meyer-Rochow, 1983; MeyerRochow and Eguchi, 1983). Although UV was once thought to represent a ‘‘private channel’’ for intraspecific communication among insects (Silberglied, 1979, 1984; Meyer-Rochow and Eguchi, 1983), under the assumption that birds were UV blind, avian predators are likely to see UV-red aposematic coloration as a nonspectral color which humans cannot perceive. In contrast, other species, such as Danaus chrysippus and its mimics (Hypolimnas missippus, Mimacraea marshalli, Acraea encedon, Acraea encedana, and Pseudacraea poggei) in southern Africa, have virtually no UV reflectance in the red elements of their dorsal wing patterns (S. C. Church, unpublished data). Studies based on UV photography suggest that the color match in lepidopteran (Remington, 1973) and coleopteran (Hinton, 1973) mimicry systems may be quite variable in the UV. Since the color match between models and mimic species is likely to differ for avian and human viewers (Cuthill and Bennett, 1993; Dittrich et al., 1993), it is unwise to assume that

UV VISION IN BIRDS

187

mimicry always extends into the UV simply because there is apparent similarity in the human-visible spectrum. The small proportion of UV-sensitive cones found in the avian retina (앑5%; Bowmaker et al., 1997; Hart et al., 1998, 1999), coupled with the fact that UV is scattered more than other wavelengths in air (Lythgoe, 1979), suggests that UV may only allow relatively coarse spatial discrimination (Bennett and Cuthill, 1994; Tove´e, 1995). Thus, achieving a greater understanding of (i) how birds use visual cues in foraging and (ii) how avian vision imposes selection on the design and evolution of protective coloration requires further investigation of the manner in which spatial (textural) and chromatic information is processed by birds over the entire wavelength range to which they are sensitive (i.e., including the UV) and under natural conditions of illumination (Endler, 1978, 1990, 1993; Vorobyev et al., 1998). C. UV Vision and Signaling The beauty, diversity, and complexity of bird coloration have, since the time of Darwin, led to hypotheses regarding its function, particularly the role of plumage in signaling and sexual selection. The fact that most birds can see UV wavelengths, and may perceive colors in fundamentally different ways from humans, implies that in many cases a reassessment of plumage colors is required (Bennett et al., 1994). Ultraviolet plumage reflection has been documented in a relatively large number of bird species (Burkhardt, 1989; Burkhardt and Finger, 1991; Maier, 1993; Bennett and Cuthill, 1994; Finger and Burkhardt, 1994; Andersson, 1996; Bennett et al., 1996, 1997; Andersson and Amundsen, 1997; Andersson et al., 1998; Hunt et al., 1998; Cuthill et al., 1999). Feathers of a whole range of human-perceived colors can be found both with and without UV reflection, implying that the degree of UV reflection cannot be predicted accurately from a feather’s human-visible appearance (Burkhardt, 1989, 1996; Finger and Burkhardt, 1994). Ultraviolet reflection has been measured using UV photography (Radwan, 1993) and/or reflectance spectrophotometry (Burkhardt, 1989). Early expectations of UV photographs revealing hidden UV plumage patterns, similar to hidden floral honey guides and lepidopteran wing patterns (Silberglied, 1979, 1984), have not been fulfilled. This may be due in part to the nature of the photographs. A UV black-and-white photograph is typically juxtaposed with a humanvisible black-and-white photograph (Silberglied, 1979, 1984; Maier, 1993). In both cases, patches that differ in hue but not brightness will go undetected. Hence, UV patterns may be overlooked. Similarly, even the supposedly ‘‘hidden’’ patterns of flowers and butterflies, which contrast strikingly only in the UV, can be correlated with differences in hue in the human-

188

INNES C. CUTHILL et al.

visible spectrum that are not apparent in the human-visible black-and-white photographs (A. T. D. Bennett, I. C. Cuthill, and S. C. Church, personal observation). Correlated UV and human-visible differences may well be common among avian plumage but, as in insects, marked UV contrast can still be visually relevant (Giurfa et al., 1995; Lunau and Maier, 1995; Chittka, 1997). Correlated human-visible reflection therefore does not devalue the potential relevance of the UV component of signals. Unlike photographs, reflectance spectra give a more complete description of plumage coloration and have the advantage of being quantitative measures which allow statistical analysis of between-spectra variation. Feathers that yield reflectance spectra with a single peak of reflection in the UV can appear dark to the human eye, as in the black lory, Chalcopsitta atra, and the Asian whistling thrushes, Myiophonus spp. (Burkhardt and Finger, 1991; Finger et al., 1992; Andersson, 1996, 1999). Alternatively, where the tail of reflection extends into the human-visible spectrum, feathers often appear blue or violet (Burkhardt, 1989; Finger and Burkhardt, 1994; Andersson and Amundsen, 1997; Hunt et al., 1998). Figure 9A shows maximum reflectance at 앑350 nm for crest feathers of a British population of blue tits. The blue tit is the first example of a species with plumage regions showing highest sexual dimorphism in the UV (Andersson et al., 1998; Hunt et al., 1998). Both Hunt et al., with English birds, and Andersson et al., with Swedish birds, found that male blue tit crests were brighter than those of females and exhibited small but significant between-sex variation in spectral shape and the wavelength of peak reflection. This crest may also play a role in mate choice. In mate choice trials, females chose males with the brightest crest (Hunt et al., 1998), whereas blue tits in the field showed assortative mating based on the estimated UV chroma (saturation) of this region of plumage (Andersson et al., 1998). It has been suggested that the spectral location of the peak varies geographically since maximum crest reflectance was at 430 nm in the Swedish population, but differences in measurement protocol may also account for this variation (S. Hunt, unpublished data). The crest is not the only region of the blue tit to show a peak of reflectance in the UV. The yellow chest and olive green back (Figs. 9B and 9C) both show a prominent UV peak and an increase in reflectance at approximately 500 nm which yields their human-perceived coloration (Hunt et al., 1998). The similarity in reflectance between the chest and back regions suggests that their structure and pigmentation are closely related. Long wavelength reflection, perceived, for example, as red, orange, or yellow, typically results from the selective absorption of light by pigments, commonly carotenoids or melanin (Fox, 1976). However, no vertebrate pigments have been discovered which reflect maximally in the UV region of the spectrum, and short

UV VISION IN BIRDS

189

Fig. 9. Reflectance spectra showing peaks of reflection in the UV region of the spectra from (A) the ‘‘blue’’ crest, (B) the ‘‘yellow’’ chest, and (C) the ‘‘olive green’’ back of male and female blue tits. Measurements are the means of six measurements from nine individuals of each sex (details as in Hunt et al., 1998), with mean male spectra as the thicker lines and female spectra as the thinner lines. Blue tits are most sexually dimorphic for the crest.

190

INNES C. CUTHILL et al.

wavelength feather reflection (including most blues and violets as well as UV reflection) appears to be structural in origin (Dyck, 1971a,b, 1976; Finger et al., 1992; Finger, 1995; Andersson, 1996, 1999; Prum et al., 1998). Structural colors are not exclusively blue, violet, or UV, however; some of the best understood structural colors are iridescent, as in the hummingbirds or starling (see Fig. 12B; Greenewelt et al., 1960; Durrer and Villiger, 1970; Bennett et al., 1997; Cuthill et al., 1999). Iridescence results from interference of light reflected from the interfaces of regular stacks of higher and lower optical density material, which in the case of feathers often consists of rows of regularly orientated keratin rods (Land, 1972; Fox, 1976). There is controversy regarding the precise feather structures and optical mechanisms responsible for ‘‘noniridescent’’ structural colors. Transmission electron microscopy suggests they are the result of light scattering by ‘‘spongy structure’’ (rods of keratin separated by air vacuoles) beneath the cortex of feather rami (Auber, 1957; Dyck, 1971a,b, 1976; Finger et al., 1992; Finger, 1995; Andersson, 1999; Prum et al., 1998). Originally termed Tyndall blues, such colors were thought to be produced by Raleigh scattering (Fox, 1976), but the diameters of the scattering particles in feathers are generally larger than those to which the Raleigh equation applies (Finger et al., 1992; Finger, 1995). Instead, Finger (1995) proposed a Mie scattering model which he found predicts more closely the reflection characteristics of a range of feather colors (UV, violet, blue, blue-green, and even some noncarotenoid yellows) that differ in the dimensions of the spongy structure they contain. In this model, keratin rods are believed to be randomly orientated, yielding incoherent scattering of light (Finger, 1995; Prum et al., 1998). As with Raleigh scattering, the proportion of backscattered light decreases exponentially as wavelength increases. Long wavelength light that is not scattered penetrates further into the rami of feathers in which it is probably absorbed by melanin granules. It is suggested that light of very short wavelengths is also absorbed by denser keratin in the ramus cortex, resulting in peaks of reflectance at short wavelengths similar to those illustrated in Fig. 9A (Finger, 1995; Andersson, 1999). In contrast, Dyck (1971a, 1976) and Prum et al. (1998) ascribe structurally produced reflectance at short wavelengths to interference effects. Prum et al.’s Fourier analysis revealed previously undetected spatial periodicity in the keratin scatterers, which theory predicts should lead to the constructive and destructive interference of coherently scattered light. A model of these interference effects predicted spectral reflectance in reasonable agreement with that measured directly from feathers (Prum et al., 1998). This model is also supported by measurements of reflectance spectra from the noniridescent feathers of the blue tit crest, which vary significantly in spectral shape and brightness according to the angle between incident and reflected

UV VISION IN BIRDS

191

light (S. Hunt, I. C. Cuthill, and A. T. D. Bennett, unpublished data; Andersson et al., 1998). Both integuments and pigments can exhibit reflection (or absorption) of UV wavelengths as a by-product of their chemical or physical composition (Andersson, 1996). Keratin, for example, is widely used as a strengthening compound; therefore, the fact that it reflects UV may be purely incidental. UV reflection may also be correlated with reflection at longer wavelengths, with the latter being the signal of behavioral significance. In other words, the fact that a species’ plumage reflects in the UV does not guarantee a role for UV in visual signaling. Pure UV reflectance (Andersson, 1996) or hidden sexual dimorphism in the UV (Andersson et al., 1998; Hunt et al., 1998; Cuthill et al., 1999) may be more suggestive of a signaling role, but behavioral experimentation involving manipulation of short wavelength reflectance is the only conclusive method of determining whether UV cues are an adaptive part of avian signal design. Studies to date have employed two techniques for manipulating UV cues: use of UV-blocking filters between test and stimulus birds (Maier, 1993; Bennett et al., 1996, 1997; Hunt et al., 1997) and direct application of UVblocking chemicals (sunblock) to the plumage (Andersson and Amundsen, 1997; Johnsen et al., 1998). The filter method has the advantage that filters have precise, homogeneous, and stable optical qualities, and the technique does not interfere directly with the bird. However, it does change the appearance of both stimuli and the background against which stimuli are viewed. The sunblock method allows more selective manipulation of specific plumage areas, but in turn may alter the physical structure and appearance of the plumage and possibly the birds’ behavior (e.g., time spent preening). Nevertheless, both methods have yielded the same pattern of results, namely, a reduction in preference for stimulus birds with reduced reflection in the UV. Maier (1993) was the first to use behavioral choice tests to show that Pekin robins (Leiothrix lutea) preferred to associate with members of the same or opposite sex that could be viewed through a UV-transmitting filter (UV⫹ ) over those visible through UV-blocking filters (UV⫺ ). Bennett et al. (1996) demonstrated a similar effect in zebra finches (Taeniopygia guttata) (Fig. 10) but also examined whether the effect resulted from changes in hue or brightness. Simple removal of UV reduces the total amount of light transmitted (and hence the perceived brightness of the stimulus) and changes its spectral composition (and hence the likely hue). In an attempt to control for the former effect, one can attempt to balance quantal transmittance between treatments (Andersson and Amundsen, 1997; Bennett et al., 1997). For example, Andersson and Amundsen (1997) compared pairwise preferences of female bluethroats (Luscinia svecia svecia) for males whose

192

INNES C. CUTHILL et al.

Fig. 10. (A) Filter types used in the experiment of Bennett et al. (1996) to manipulate both spectral composition of light (i.e., blocking UV) and overall quantal flux. The latter is correlated with perceived brightness and the former with hue; therefore, examining the influence of both types of manipulation on the attractiveness of birds viewed through these filters can potentially identify whether UV is contributing (mainly) to hue or brightness perception. The filters were UV transmitting (UV⫹ ) or UV blocking with varying degrees of total light transmission (UV⫺ ⬎ ND1 ⬎ ND2; ND, neutral density). (B) The mean number of hops (⫾SE) by female zebra finches facing stimulus cages fronted by filters of four types in A. There were three phases to the experiment: ‘‘mate choice,’’ during which randomly assigned males were in the stimulus cages, and two control phases, during which no males were present. These controls tested for a preference for a particular ‘‘colored view’’ irrespective of mate choice (details in Bennett et al., 1996). Females preferred males behind UV⫹ filters and showed no preferences when the males were absent. (C) Mean hops (⫾SE) by females choosing between males behind filters differing only in the total quantity of light transmitted. Filters were UV⫹ (as in A) and three UV-transmitting neutral density filters which reduced light transmission (UV⫹ brighter than ND3 ⬎ ND4 ⬎ ND5; details in Bennett et al., 1996). Females showed no significant preferences, indicating that the preference in B is due to the effect of UV removal on hue rather than brightness. (D) Mean hops (⫾SE) by females facing stimulus cages containing males with UV-symmetric or UV-asymmetric leg bands (Bennett et al., 1996). Two of the four leg band arrangements were asymmetrical (ASYM1 and ASYM2) and two were symmetrical (SYM1 and SYM2). Females preferred symmetrically banded males, even though the ornaments’ symmetry was only visible in the UV. Replotted with permission from Nature Bennett et al., 1996, 䉷 1996 MacMillan Magazines Ltd.

UV VISION IN BIRDS

193

UV-blue throat patches had been treated with either preen oil ⫹ sunblock (no UV) or preen oil ⫹ black pigment (UV present but reduced overall reflectance). The two treatments were matched for total amount of reflected light but test birds still preferred UV-reflecting stimulus birds (c.f. Bennett et al., 1997). However, even this may not produce stimuli of equal perceived brightness if, as is likely, the avian eye is more sensitive to certain wavelengths than others (Burkhardt and Maier, 1989; Vorobyev and Osorio, 1998). Consequently, a more powerful approach is to compare directly the effect of changing spectral composition with reduction of total quantal flux by using neutral-density filters (Fig. 10). Using such treatments, Bennett et al. (1996) found no preference for ‘‘bright’’ compared to ‘‘dull’’ zebra finches (Fig. 10). This implies that it is the change in perceived hue of stimuli, rather than any change in brightness, that produces a reduction in association preference for UV⫺ birds. To UV-sensitive birds, removal of all reflection below 400 nm might produce an ‘‘odd’’-colored bird, in extreme cases perhaps unrecognizable as a conspecific. Hunt et al. (1997) showed that the mate-choice preference for particular color bands shown by zebra finches (Burley et al., 1982) disappeared in the absence of all UV cues (Fig. 11). This was shown to be due to the effect of UV-blocking filters on the overall appearance of stimulus birds rather than to any change in the color of the leg bands (Hunt et al., 1997). Such effects may be important in explaining inconsistent results in experiments involving artificial illumination and one-way glass (which often absorbs most UV) or in which plumage has been manipulated using paints or dyes that match to human eyes but not necessarily to avian eyes (Collins and Ten Cate, 1996; Hunt et al., 1997). However, more subtle effects of UV removal on mate choice have also been demonstrated. Bennett et al. (1996) found that, as in the later bluethroat studies (Andersson and Amundsen, 1997; Johnsen et al., 1998), UV can be important in the assessment of specific sexual ornaments, when the overall appearance of the bird has not been changed. Zebra finches show preferences not only for particular colors of leg bands but also for symmetrical arrangements of bands (Swaddle and Cuthill, 1994). By producing sets of UV-reflecting and UV-blocking leg bands, identical in appearance to human eyes, Bennett et al. (1996) showed that females preferred males wearing UV-symmetrical compared to UV-asymmetrical bands (Fig. 10). Note that all stimulus birds had normal, natural-colored plumage and all birds wore the same number of each type of band so they were likely to be equally ‘‘odd looking.’’ Such effects are therefore very unlikely to be due to the abolition of species recognition. Bennett et al. (1997) also compared patterns of mate choice in the presence and absence of UV cues. Under illumination which included UV

194

INNES C. CUTHILL et al.

Fig. 11. (A) Mean number of hops (⫾SE) made by female zebra finches (n ⫽ 8) in front of red (R), orange (O), light green banded (LG), or unbanded (N) males. Females viewed males through UV-transmitting (clear bars) or UV-blocking filters (solid bars). A band color preference was observed under the UV⫹ treatment, but not when UV wavelengths were removed. (B) Mean number of hops (⫾SE) made by females (n ⫽ 8) in front of red (R), orange (O), light green banded (LG), or unbanded (N) males. Color bands were overlain with either UV-transmitting (clear bars) or UV-blocking filters (solid bars). In both cases, filters separating males and females were UV transmitting. For both band treatments, a band color preference was observed, indicating that the effect of blocking UV in A was due to changing the overall bird’s appearance rather than altering the colors of the bands.

wavelengths, female starlings ranked groups of males in a consistent manner that correlated with the reflectance of iridescent feathers of the throat and coverts (Fig. 12, regions displayed during mate attraction; Feare, 1984; Eens et al., 1990). Under UV-deficient conditions, females agreed on the relative attractiveness of stimulus males, strongly suggesting that species recognition

UV VISION IN BIRDS

195

Fig. 12. (A) Correlations in preferences of pairs of female starlings independently viewing quartets of males through filters which either transmitted (UV⫹ ) or blocked (UV⫺) ultraviolet wavelengths (details in Bennett et al., 1997). Boxplots represent the median and interquartile range of correlations for eight sets of females/males, with correlations being calculated from the number of hops facing each male. Under UV⫹ conditions, preferences of pairs of females were highly significantly correlated. They were also highly correlated under UV⫺ conditions, although less so than under UV⫹ conditions. Preferences across UV⫹ and UV⫺ treatments were not correlated, indicating that changes in ambient lighting can radically alter mate choice criteria. (B) Reflectance spectra of the iridescent throat feathers of the most preferred and least preferred males in the UV⫹ treatment of Bennett et al. (1997). Each spectrum is the mean of 10 measurements from eight males, relative to a Spectralon 99% white standard. Preferred males have spectra with more sharply defined peaks (and thus probably more saturated colors; Endler, 1990), and the location of the peaks suggests they are purplish (red ⫹ blue), whereas nonpreferred males are ‘‘UV ⫹ green.’’ Replotted with permission from Proceedings of the National Academy of Sciences U.S.A. Bennett et al., 1997, 䉷 1997, PNAS.

and motivation for choice remained intact. However, the preference rankings under the two filter conditions were significantly different. Under UV⫺ conditions, there was no correlation between male rank and plumage reflectance. We can conclude that natural variation in UV reflectance is important in starling mate assessment and that ambient lighting can have

196

INNES C. CUTHILL et al.

profound effects on observed mating preferences and the criteria on which mate choice is based. The latter result reinforces the conclusions of Hunt et al. (1997): Behavioral experiments conducted under UV-deficient conditions can produce repeatable results which are nonetheless unrelated to the patterns of choice under other lighting conditions, including natural daylight. Recent work has shown that preferences for naturally UV-reflecting partners can be very robust, observable not only in laboratory and aviary experiments but also in a field population of bluethroats in which the sunblock-type treatment affected both social mate choice and extra-pair success (Johnsen et al., 1998). Currently, it is unknown whether the UV is a ‘‘special’’ waveband for avian sexual signaling (Bennett and Cuthill, 1994) or whether removal of other wavebands (e.g., green wavelengths) will produce equivalent effects. We strongly suspect that for many species the latter will be the case. However, if there is something special about signals with a UV component, the nature of that information is unclear. There is evidence that visual signals in different regions of the spectrum are correlated with different aspects of reproductive behavior: Structural colors, including the UV, may predict high levels of extra-pair paternity, whereas melanin-based traits may vary with aspects of parental care (Owens and Hartley, 1998). Although the costs of producing some red pigment-based signals are relatively well understood (Hill, 1993a,b,c; Hill and Montgomerie, 1994; Hill et al., 1994), we know far less about structural colors. It is plausible that condition at the time of molt is important for correct deposition of the structures producing the color, but there are no data on what the relationship might be. Meanwhile, we should be aware of the possible contribution of the UV waveband when investigating avian plumage coloration, mate choice, or display behavior or when designing experiments in which the light environment might influence results.

V. Conclusions The most important message in this chapter is not concerned with UV vision or even birds. It is that there must be closer ties between sensory ecology and behavioral ecology if we are to understand signals and other objects under selective pressure from sensory systems (e.g., cryptic and aposematic prey). Visual ecologists have long recognized the importance of differences between human color vision and that of other animals [e.g., in 1882, Lubbock (cited in Goldsmith, 1994) noted the effect of blocking UV on ant behavior] but until recently this message has gone largely

UV VISION IN BIRDS

197

unnoticed by most behavioral ecologists working on birds (Cuthill and Bennett, 1993; Bennett et al., 1994). This is despite the fact that birds have long been known to see UV wavelengths (Huth and Burkhardt, 1972; Wright, 1972a), the implications of this for their behavioral ecology have been explicit in the work of Burkhardt and colleagues (Burkhardt, 1982, 1989; Burkhardt and Finger, 1991), and the parallels with studies of the visual ecology of other vertebrates have been clear (Hailman, 1977; Endler, 1978, 1983, 1987; Lythgoe, 1979). Many animals can see UV light, so the fact that birds possess this visual capability is by no means remarkable (Jacobs, 1992; Tove´e, 1995). The most spectacular examples are found in the mantis shrimps (Crustacea and Stomatopoda), some of which have four distinct UV receptors among a total of up to 12 spectrally distinct photoreceptor types in their compound eyes (Cronin et al., 1994a,b, 1996). Such diversity is unusual, and no vertebrate has near this number of photoreceptor types, but mammals (including humans) are notable for the depauperate nature of their photopigment complement. We have tried to stress in this review that the inability of humans to see UV light is just one, not necessarily even the most important, difference between the way birds and humans see color. Humans, like all Old World and some New World primates, have a (UV-blind) trichromatic system based on three retinal cone types. Birds have more cone classes and these ‘‘divide up’’ the visible spectrum in a different way from the cones of trichromatic primates. In this review, we have concentrated on the mechanism and function of avian color vision, which is mainly, perhaps exclusively, the function of the single cones, but the reader should bear in mind that the differences between avian and human vision extend beyond color perception (Zeigler and Bischof, 1993). The ability of birds to see UV arises from UV-transparent ocular media and a very shortwave visual pigment, the wavelength of maximum sensitivity of which lies in either the near-UV or violet, depending on species (Figs. 1 and 3). Evidence from psychophysical (Palacios and Varela, 1992) and behavioral (Bennett et al., 1996) experiments, coupled with color space modeling (Vorobyev and Osorio, 1998; Vorobyev et al., 1998), suggests that the violet/UV cones are part of one or more color opponent mechanisms; therefore, UV seems to be involved in hue perception. The available evidence suggests that it forms one axis of a tetrachromatic color space, as in goldfish (Neumeyer and Arnold, 1989; Neumeyer, 1992). All these factors indicate that color perception will differ between birds and humans, indeed between species of birds with cones of differing spectral sensitivities (e.g., turkey vs starling; Fig. 1). For example, by modeling the color space of bird species with either violet-type or UV-type cones, Vorobyev et al. (1998) found that a ‘‘passerine’’-type eye outperformed a ‘‘pigeon’’-type eye in

198

INNES C. CUTHILL et al.

terms of the number of natural spectra that were discriminable. However, this conclusion is limited to the data set of spectra that were used (a sample of European and Australasian birds chosen for their color diversity), and the pigeon system may well outperform the passerine system in other circumstances (indeed, the adaptationist would expect this to occur). Nevertheless, we would expect birds with the same number and type of visual pigments to see broadly similar classes of colors (primary UV/violets, blues, greens, and reds), differing only in the details of their spectral sensitivities and color discrimination. This is not to say that these differences will be unimportant for understanding a given species’ visual ecology, but we have focused our review on the implications of the large-scale differences between a generic avian visual system and that of humans. The discovery that there are examples of many bird-relevant objects (e.g., plumage, fruit, and insects) which reflect UV, whereas many common backgrounds (e.g., leaves, earth, and bark) do not, suggested that UV vision may function in foraging and/or signaling. However, only recently has direct experimental evidence started to accumulate (see Section IV). In almost all cases in which the role of UV information in foraging or signaling has been experimentally investigated, a significant influence has been found. The obvious exceptions (Koivula et al., 1997) have concerned species such as owls, which probably do not rely on color vision when hunting at low light intensities (Martin, 1990) and may not possess UV-sensitive cones (Bowmaker and Martin, 1978). In many ways, it is perhaps not surprising that removal of UV wavelengths from the plumage reflectance of species such as blue tits and bluethroats has significant effects on the attractiveness of individuals so manipulated (see Section IV,C) because the maximum reflectance from both species’ plumage lies in the UV/violet (Andersson and Amundsen, 1997; Andersson et al., 1998; Hunt et al., 1998). However, it is notable that removal of UV wavelengths has also been found to affect mate choice in species such as the starling, in which UV differences are correlated with differences in the human-visible spectrum (Bennett et al., 1997; Cuthill et al., 1999). In this sense, UV information could have been considered redundant and removal of UV unlikely to have an effect on choice, but this was not the case (Fig. 12; Bennett et al., 1997). Even more striking, similar effects are found in the zebra finch (Fig. 10), which is colorful and sexually dichromatic to the human eye (Zann, 1996) with notable beak and plumage reflectance at long (orange-red) wavelengths (Bennett et al., 1996). Why does removal of UV affect the attractiveness of species which are not particularly UV reflective? It is important to remember that UV is not a color, although when talking of objects which reflect only in the UV it is useful shorthand to refer to them as ultraviolet. Because it is the relative

UV VISION IN BIRDS

199

amounts of different wavelengths (UV and human visible) that a visual system interprets as hue, the UV/violet receptor of birds is liable to contribute to opponent channels in which it is compared with the output of other cones (Palacios and Varela, 1992; Thompson et al., 1992; Vorobyev et al., 1998). Thus, for example, removal of all UV input may alter the appearance of the zebra finch’s white plumage (which reflects UV as well as the entire human-visible waveband; Bennett et al., 1996) and/or the contrast of the whole bird against UV-reflecting backgrounds. Thus, the fact that a species lacks distinctive ‘‘pure UV’’ reflectance is no guarantee that these wavelengths do not have a role in its visual signaling. Across different species, there are examples of red, yellow, green, blue, black, and white feathers that either do or do not reflect UV and thus will be perceived as quite different colors to birds, whereas they are classed as similar by humans (Burkhardt, 1996). Under what conditions will human perception be sufficient for studying bird coloration? If the goal is simply to score the presence/absence, or size, of plumage patches, then birds are likely to see any patches that humans can detect. We can be particularly confident if the color is produced by a pigment or structure of known optical properties. For example, melanin absorbs UV and human-visible wavelengths; therefore, the black bibs on sparrows (Rohwer, 1985; Møller, 1989; Veiga, 1993) are likely to appear dark, relative to the surrounding plumage, to sparrows and to humans. However, there could be additional patches, or within-patch variation, to which humans are blind. Nevertheless, to what extent are birds which are colorful to humans also colorful to birds? Vorobyev et al. (1998) addressed this question by comparing the separation in avian and human receptor space (see Section II,E) of the differently colored plumage regions of a range of bird species. Thus, for any one species, they calculated the distance between all possible pairs of spectra, first for human receptor space and then for both a passerine-type eye (true UV receptor) and a pigeon-type eye (violet receptor). Species of a fairly uniform color would score low and similar values, whereas species with many differently colored patches would score a wide range of values. Correlating these (estimated) perceptual separations between either type of bird visual system and that of humans gave a measure of the extent to which avian and human systems coincided in this measure of within-body colorfulness. The correlation was high for species such as the zebra finch, with most spectral variation being due to simple step-function rises in reflectance at long wavelengths. However, for species with considerable UV reflectance and/or complex multipeak reflection spectra, such as Princess Stephanie’s bird of paradise (Astarchia stephaniae), the correlation was weaker. Therefore, the results are mixed: For some plumage patterns, humans and birds may agree on the diversity

200

INNES C. CUTHILL et al.

of color variation present, but for others they will not. Of course, this analysis only considers colorfulness with respect to within-body contrast and thus does not relate directly to conspicuousness (Endler, 1990). The latter quality will depend on both the background and the light environment (Endler and The´ry, 1996). Often, however, behavioral ecologists are not simply interested in whether differently colored plumage patches are present or absent or whether or not there is sexual dichromatism. It is of interest to measure intensity of coloration and degree of difference and often to manipulate these qualities. Here, caution must be taken because matching paints or dyes by eye is no guarantee that they will match to the avian eye. For example, a yellow color band with no UV reflectance (Cuthill et al., 1999) is a very different (avian) color from the yellow of a blue tit’s chest (Fig. 9B). Preferably, the spectral reflectance of the colors should be measured directly or, if human color standards are being used to rank color variation, then these and the plumage regions being scored should be calibrated against objective reflectance measurements beforehand (Birkhead et al., 1998). Where the natural receiver is a bird, and color variation is being scored, as in a preference test or mate-choice trial, then the problem of human subjectivity may seem to have been avoided. However, the finding that the prevailing light environment, in terms of the presence or absence of UV, can have profound effects on observed mating preferences indicates caution (Bennett et al., 1997). That is, experiments conducted under UV-deficient conditions, such as under many types of artificial light or with UV-blocking glass or plastic, can produce repeatable results which are nonetheless unrelated to the patterns of choice in daylight. A clear example of this effect is shown by Hunt et al.’s (1997) investigation of the replicability of Burley et al.’s (1982) classic demonstration of band color preferences in zebra finches. Burley et al.’s results were reproducible under wide-spectrum lighting (UV ⫹ human visible), but the preference disappeared in the absence of UV cues. This potentially explains the inconsistent results in earlier investigations (reviewed by Collins and Ten Cate, 1996), in which different types of artificial light or one-way glass were used. More important, it sounds a warning for all who carry out experiments on birds under artificial lighting conditions. Not all such experiments are flawed. It is likely that birds will have a fair degree of color constancy with shifts in the illuminant and may well adapt to different lighting regimes over time. However, the extent and speed with which they might do so is currently unknown. The effect of sensory mechanisms and higher level processing on signal evolution is one of the most active areas in behavioral ecology (Endler and McLellan, 1988; Guilford and Dawkins, 1991, 1993; Endler, 1992a,b; Ryan

UV VISION IN BIRDS

201

and Keddy-Hector, 1992; Ryan and Rand, 1993; Ryan, 1997; Endler and Basolo, 1998). Now that we have the technology to study visual signals from beyond the narrow perspective of human color vision, we can attempt to understand the true significance of colors to birds. By linking the approaches of sensory physiology and behavioral ecology, not only will a better understanding of animal coloration be gleaned but also, perhaps, insight will be gained into why different color vision systems evolved in the first place.

VI. Summary Birds can see UV light because, unlike humans, their lenses and other ocular media transmit UV, and they possess a class of photoreceptor which is maximally sensitive to violet or UV light, depending on the species. Birds retain what appears to be the ancestral tetrapod, perhaps vertebrate, system of a single class of rod, subserving scotopic vision, and four spectrally distinct cone types, used for color vision under photopic conditions. Current evidence is consistent with the idea that birds have a tetrachromatic color space, as compared to the trichromacy of humans, and therefore will see a range of hues we cannot imagine. Birds, along with some reptiles and fish, also possess double cones in large numbers, a cone class the function of which is still unclear. We reviewed a range of behavioral experiments, from several species, which show that UV information is utilized in behavioral decisions, notably in foraging and signaling. Hidden sex differences in coloration have been found in species which are more or less monomorphic to humans; therefore, the extent of chromatic variation, both within and between species, may have been underestimated in the past. It is also significant that removal of UV wavelengths affects mate choice even in species which are colorful to humans. These studies emphasize that avian and human color perceptions are different and that use of human color standards, and even artificial lighting, may produce misleading results. However, genuinely objective measures of color are available, as are, importantly, models for mapping the measured spectra into an avian color space.

Acknowledgments Our research has been supported by grants from Biotechnology and Biological Sciences Research Council, the Natural Environment Research Council, Royal Society, and Nuffield Foundation. We thank Staffan Andersson, Dietrich Burkhardt, Ron Douglas, John Endler, Klaus Lunau, Erhard Maier, Justin Marshall, Graham Martin, Daniel Osorio, Tim Roper, and Misha Vorobyev for discussion and help with various aspects of our research.

202

INNES C. CUTHILL et al.

References Able, K. P., and Able, M. A. (1993). Daytime calibration of magnetic orientation in a migratory bird requires a view of skylight polarization. Nature 364, 523–525. Able, K. P., and Able, M. A. (1995). Manipulations of polarized skylight calibrate magnetic orientation in a migratory bird. J. Comp. Physiol. A 177, 351–356. Alatalo, R. V., and Mappes, J. (1996). Tracking the evolution of warning signals. Nature 382, 708–710. Allen, R. B., and Lee, W. G. (1992). Fruit selection by birds in relation to fruit abundance and appearance in the naturalized shrub Berberis darwinii. N. Z. J. Bot. 30, 121–124. Andersson, S. (1996). Bright ultraviolet coloration in the Asian whistling-thrushes (Myiophonus spp.). Proc. R. Soc. London B 263, 843–848. Andersson, S. (1999). Morphology of UV reflectance in a whistling-thrush: Implications for the study of structural colour signalling in birds. J. Avian Biol. 30, 193–204. Andersson, S., and Amundsen, T. (1997). Ultraviolet colour vision and ornamentation in bluethroats. Proc. R. Soc. London B 264, 1587–1591. ¨ rnborg, J., and Andersson, M. (1998). Ultraviolet sexual dimorphism and Andersson, S., O assortative mating in blue tits. Proc. R. Soc. London B 265, 445–450. Applebury, M. L., and Hargrave, P. A. (1986). Molecular biology of the visual pigments. Vision Res. 26, 1881–1895. Auber, L. (1957). The structures producing ‘‘non-iridescent’’ blue colour in bird feathers. Proc. Zool. Soc. London 129, 455–486. Avery, J. A., Bowmaker, J. K., Djamgoz, M. B. A., and Downing, J. E. G. (1983). UV sensitive receptors in a fresh-water fish. J. Physiol. 334, 23–24. Avery, M. L., Decker, D. G., Humphrey, J. S., Hayes, A. A., and Laukert, C. C. (1995). Color, size, and location of artificial fruits affect sucrose avoidance by cedar waxwings and European starlings. Auk 112, 436–444. Baker, R. R., and Parker, G. A. (1979). The evolution of bird coloration. Philos. Trans. R. Soc. London B 287, 63–130. Bennett, A. T. D., and Cuthill, I. C. (1994). Ultraviolet vision in birds: What is its function? Vision Res. 34, 1471–1478. Bennett, A. T. D., Judson, O. P., and Norris, K. J. (1992). The functions of ultraviolet vision in birds. Am. Zool. 32, 164. Bennett, A. T. D., Cuthill, I. C., and Norris, K. J. (1994). Sexual selection and the mismeasure of color. Am. Nat. 144, 848–860. Bennett, A. T. D., Cuthill, I. C., Partridge, J. C., and Maier, E. J. (1996). Ultraviolet vision and mate choice in zebra finches. Nature 380, 433–435. Bennett, A. T. D., Cuthill, I. C., Partridge, J. C., and Lunau, K. (1997). Ultraviolet plumage colors predict mate preferences in starlings. Proc. Natl. Acad. Sci. USA 94, 8618–8621. Birkhead, T. R., Fletcher, F., and Pellatt, E. J. (1998). Sexual selection in the zebra finch Taeniopygia guttata: Condition, sex traits and immune capacity. Behav. Ecol. Sociobiol. 44, 179–191. Blough, P. M., Riggs, L. A., and Schafer, K. L. (1972). Photopic spectral sensitivity determined electroretinographically for the pigeon eye. Vision Res. 12, 477–485. Bowmaker, J. K. (1977). The visual pigments, oil droplets and spectral sensitivity of the pigeon. Vision Res. 17, 1129–1138. Bowmaker, J. K. (1979). Visual pigments and oil droplets in the pigeon retina, as measured by microspectrophotometry, and their relation to spectral sensitivity. In ‘‘Neural Mechanisms of Behaviour in the Pigeon’’ (A. M. Granda and J. H. Maxwell, Eds.), pp. 287–305. Plenum, New York.

UV VISION IN BIRDS

203

Bowmaker, J. K. (1980). Colour vision in birds and the role of oil droplets. TINS 3, 196–199. Bowmaker, J. K. (1991a). The evolution of vertebrate visual pigments and photoreceptors. In ‘‘Vision and Visual Dysfunction’’ (J. R. Cronly-Dillon and R. L. Gregory, Eds.), Vol. 2, pp. 63–81. CRC Press, Boston. Bowmaker, J. K. (1991b). Visual pigments, oil droplets and photoreceptors. In ‘‘Vision and Visual Disfunction’’ (P. Gouras, Ed.), Vol. 6, pp. 108–127. CRC Press, Boston. Bowmaker, J. K., and Hunt, D. M. (1999). Molecular biology of photoreceptor spectral sensitivity. In ‘‘Adaptive Mechanisms in the Ecology of Vision’’ (S. N. Archer, M. B. A. Djamgoz, E. R. Loew, J. C. Partridge, and S. Valerga, Eds.). Chapman & Hall, London. Bowmaker, J. K., and Knowles, A. (1977). The visual pigments and oil droplets of the chicken retina. Vision Res. 17, 755–764. Bowmaker, J. K., and Martin, G. R. (1978). Visual pigments and colour vison in a nocturnal bird, Strix aluco (Tawny Owl). Vision Res. 18, 1125–1130. Bowmaker, J. K., and Martin, G. R. (1985). Visual pigments and oil droplets in the penguin, Spheniscus humboldti. J. Comp. Physiol. A 156, 71–77. Bowmaker, J. K., Kovach, J. K., Whitmore, A. V., and Loew, E. R. (1993). Visual pigments and oil droplets in genetically manipulated and carotenoid deprived quail: A microspectrophotometric study. Vision Res. 33, 571–578. Bowmaker, J. K., Heath, L. A., Wilkie, S. E., and Hunt, D. M. (1997). Visual pigments and oil droplets from six classes of photoreceptor in the retinas of birds. Vision Res. 37, 2183–2194. Bowmaker, J. K., Wilkie, S. E., and Hunt, D. M. (1999). Photoreceptors and molecular genetics of visual pigments. In ‘‘Proceedings of the 22nd Ornithological Congress, Durban, South Africa’’ (N. Adams, and R. Slotow, Eds.). University of Natal, Durban. Bownds, D. (1967). Site of attachment of retinal in rhodopsin. Nature 216, 1178–1181. Braekevelt, C. R. (1990). Retinal photoreceptor fine structure in the mallard duck (Anas platyrhynchos) Histol. Histopathol. 5, 123–131. Braekevelt, C. R. (1993). Fine structure of the retinal photoreceptors of the great horned owl (Bubo virginianus). Histol. Histopathol. 8, 25–34. Braekevelt, C. R. (1994). Retinal photoreceptor fine structure in the American crow (Corvus brachyrhynchos). Anat. Histol. Embryol. 23, 376–387. Braekevelt, C. R., Smith, S. A., and Smith, B. J. (1996). Fine structure of the retinal photoreceptors of the barred owl (Strix varia). Histol. Histopathol. 11, 79–88. Brines, M. L., and Gould, J. L. (1982). Skylight polarization patterns and animal orientation. J. Exp. Biol. 96, 69–91. Burkhardt, D. (1982). Birds, berries and UV. Naturwiss. 69, 153–157. Burkhardt, D. (1989). UV vision: A bird’s eye view of feathers. J. Comp. Physiol. A 164, 787–796. Burkhardt, D. (1996). Ultraviolet perception by bird eyes and some implications. Naturwiss. 83, 492–497. Burkhardt, D., and Finger, E. (1991). Black, white and UV: How birds see birds. Naturwiss. 78, 279–280. Burkhardt, D., and Maier, E. (1989). The spectral sensitivity of a passerine bird is highest in the UV. Naturwiss. 76, 82–83. Burley, N., Krantzberg, G., and Radman, P. (1982). Influence of colour-banding on the conspecific preferences of zebra finches. Anim. Behav. 30, 444–455. Cajal, S. R. (1893). La re´tine des verte´bre´s. Cellule 9, 17–257. Cameron, D. A., and Pugh, E. N. J. (1991). Double cones as a basis for a new type of polarization vision in vertebrates. Nature 353, 161–164.

204

INNES C. CUTHILL et al.

Campenhausen, M. V., and Kirschfeld, K. (1998). Spectral sensitivity of the accessory optic system of the pigeon. J. Comp. Physiol. A 183, 1–6. Chatfield, C., and Collins, A. J. (1995). ‘‘Introduction to Multivariate Analysis.’’ Chapman & Hall, London. Chen, D., and Goldsmith, T. H. (1986). Four spectral classes of cone in the retinas of birds. J. Comp. Physiol. A 159, 473–479. Chen, D. M., and Stark, W. S. (1994). Electroretinographic analysis of ultraviolet sensitivity in juvenile and adult goldfish retinas. Vision Res. 34, 2941–2944. Chen, D. M., Collins, J. S., and Goldsmith, T. H. (1984). The ultraviolet receptor of bird retinas. Science 225, 337–340. Chittka, L. (1992). The color hexagon: A chromaticity diagram based on photoreceptor excitations as a generalized representation of colour opponency. J. Comp. Physiol. A 170, 533–543. Chittka, L. (1996). Does bee color vision predate the evolution of flower color? Naturwiss. 83, 136–138. Chittka, L. (1997). Bee color vision is optimal for coding flower color, but flower colors are not optimal for being coded—Why? Israel J. Plant Sci. 45, 115–127. Chittka, L., and Menzel, R. (1992). The evolutionary adaptation of flower colours and the insect pollinators’ colour vision. J. Comp. Physiol. A 171, 171–181. Chittka, L., and Waser, N. M. (1997). Why red flowers are not invisible to bees. Israel J. Plant Sci. 45, 169–183. Chittka, L., Beier, W., Hertel, H., Steinmann, E., and Menzel, R. (1992). Opponent colour coding is a universal strategy to evaluate the photoreceptor inputs in Hymenoptera. J. Comp. Physiol. A 170, 545–563. Chittka, L., Shmida, A., Troje, N., and Menzel, R. (1994). Ultraviolet as a component of flower reflections, and the colour perception of hymenoptera. Vision Res. 34, 1489–1508. Church, S. C., Bennett, A. T. D., Cuthill, I. C., Hunt, S., Hart, N. S., and Partridge, J. C. (1998a). Does Lepidopteran larval crypsis extend into the ultraviolet? Naturwiss. 85, 189–192. Church, S. C., Bennett, A. T. D., Cuthill, I. C., and Partridge, J. C. (1998b). Ultraviolet cues affect the foraging behaviour of blue tits. Proc. R. Soc. London B 265, 1509–1514. Coemans, M. A. J. M., Vos Hzn, J. J., and Nuboer, J. F. W. (1994a). The orientation of the e-vector of linearly polarized light does not affect the behaviour of the pigeon Columba livia. J. Exp. Biol. 191, 107–123. Coemans, M. A. J. M., Vos Hzn, J. J., and Nuboer, J. F. W. (1994b). The relation between celestial colour gradients and the position of the sun, with regard to the sun compass. Vision Res. 34, 1461–1470. Collins, S. A., and Ten Cate, C. (1996). Does beak colour affect female preference in zebra finches? Anim. Behav. 52, 105–112. Crane, J. (1954). Spectral reflectance characteristics of butterflies (Lepidoptera) from Trinidad, B. W. I. Zool. N.Y. 39, 85–115. Cronin, T. W., Marshall, N. J., and Caldwell, R. L. (1994a). The intrarhabdomal filters in the retinas of mantis shrimps. Vision Res. 34, 279–291. Cronin, T. W., Marshall, N. J., Caldwell, R. L., and Shashar, N. (1994b). Specialization of retinal function in the compound eyes of mantis shrimps. Vision Res. 34, 2639–2656. Cronin, T. W., Marshall, N. J., and Caldwell, R. L. (1996). Visual pigment diversity in two genera of mantis shrimps implies rapid evolution (Crustacea, Stomatopoda). J. Comp. Physiol. A 179, 371–384. Cserha´ti, P., Sze´l, A., and Ro¨hlich, P. (1989). Four cone types characterized by anti-visual pigment antibodies in the pigeon retina. Invest. Ophthalmol. 30, 74–81.

UV VISION IN BIRDS

205

Cuthill, I. C., and Bennett, A. T. D. (1993). Mimicry and the eye of the beholder. Proc. R. Soc. London B 253, 203–204. Cuthill, I. C., Bennett, A. T. D., Partridge, J. C., and Maier, E. H. (1999). Plumage reflectance and the objective assessment of avian sexual dichromatism. Am. Nat. 160, 183–199. Dartnall, H. J. A. (1953). The interpretation of spectral sensitivity curves. Br. Med. Bull. 9, 24–30. Dartnall, H. J. A., Bowmaker, J. K., and Mollon, J. D. (1983). Microspectrophotometry of human photoreceptors. In ‘‘Colour Vision: Physiology and Psychophysics’’ (J. D. Mollon and L. T. Sharpe, Eds.). Academic Press, London. Darwin, C. (1859). ‘‘On the Origin of Species.’’ John Murray, London. Das, D. (1997). Visual pigments, oil droplets and opsin sequences from the canary, Serinus canaria. Unpublished PhD thesis, University of London, London. Deegan, J. F., and Jacobs, G. H. (1993). On the identity of the cone types of the rat retina. Exp. Eye Res. 56, 375–377. Dittrich, W., Gilbert, F., Green, P., McGregor, P., and Grewcock, D. (1993). Imperfect mimicry: A pigeon’s perspective. Proc. R. Soc. London B 251, 195. Douglas, R. H., and Marshall, N. J. (1999). A review of vertebrate and invertebrate ocular filters. In ‘‘Adaptive Mechanisms in the Ecology of Vision’’ (S. N. Archer, M. B. A. Djamgoz, E. R. Loew, J. C. Partridge, and S. Valerga, Eds.). Chapman & Hall, London. Dumbacher, J. P., Beehler, B. M., Spande, T. F., and Garraffo, H. M. (1992). Homobatrachotoxin in the genus pitohui: Chemical defense in birds. Science 258, 799–801. Durrer, H., and Villiger, W. (1970). Schillerfarben der Stare (Sturnidae). Elektronenmikroskopischer Beitrag zur Fa¨rbung und Systematik der Familie. J. Ornithol. 111, 133–153. Dyck, J. (1971a). Structure and colour production of the blue barbs of Agapornis roseicollis and Cotinga maynana. Z. Zellforschung 115, 17–29. Dyck, J. (1971b). Structure and spectral reflectance of green and blue feathers of the rosefaced lovebird (Agapornis roseicollis). Kongelige Danske Videnskabernes Selskab Biol. Skrifter 18, 1–67. Dyck, J. (1976). ‘‘Structural Colours.’’ Australian Academy of Sciences, Canberra. Eens, M., Pinxten, R., and Verheyen, R. F. (1990). On the function of singing and wingwaving in the European starling Sturnus vulgaris. Bird Study 37, 48–52. Eguchi, E., and Meyer-Rochow, V. B. (1983). Ultraviolet photography of forty-three species of Lepidoptera representing ten families. Ann. Zool. Jpn. 56, 10–18. Eisner, T., Silberglied, R. E., Aneshansley, D., Carrell, J. E., and Howland, H. C. (1969). Ultraviolet video viewing: The television camera as an insect eye. Science 166, 1172–1174. Emmerton, J., and Delius, J. D. (1980). Wavelength discrimination in the ‘‘visible’’ and UV spectrum by pigeons. J. Comp. Physiol. A 141, 47–52. Emmerton, J., Schwemer, J., Muth, J., and Schlecht, P. (1980). Spectral transmission of the ocular media of the pigeon (Columba livia). Invest. Ophthalmol. 19, 1382–1387. Endler, J. A. (1978). A predator’s view of animal color patterns. Evol. Biol. 11, 319–364. Endler, J. A. (1983). Natural and sexual selection on color patterns in poeciliid fishes. Environ. Biol. Fishes 9, 173–190. Endler, J. A. (1987). Predation, light-intensity and courtship behavior in Poecilia reticulata (Pisces: Poeciliidae). Anim. Behav. 35, 1376–1385. Endler, J. A. (1990). On the measurement and classification of colour in studies of animal colour patterns. Biol. J. Linnean Soc. 41, 315–352. Endler, J. A. (1992a). Sensory drive: Does sensory biology bias or constrain the direction of evolution? Introduction to the symposium. Am. Nat. 139, S1–S3. Endler, J. A. (1992b). Signals, signal conditions, and the direction of evolution. Am. Nat. 139, S125–S153.

206

INNES C. CUTHILL et al.

Endler, J. A. (1993). The color of light in forests and its implications. Ecol. Monogr. 63, 1–27. Endler, J. A., and Basolo, A. L. (1998). Sensory ecology, receiver biases and sexual selection. TREE 13, 415–420. Endler, J. A., and McLellan, T. (1988). The processes of evolution—Toward a newer synthesis. Annu. Rev. Ecol. Syst. 19, 395–421. Endler, J. A., and The´ry, M. (1996). Interacting effects of lek placement, display behavior, ambient light, and color patterns in three neotropical forest-dwelling birds. Am. Nat. 148, 421–452. Feare, C. (1984). ‘‘The Starling.’’ Oxford Univ. Press, Oxford. Finger, E. (1995). Visible and UV coloration in birds: Mie scattering as the basis of color in many bird feathers. Naturwiss. 82, 570–573. Finger, E., and Burkhardt, D. (1994). Biological aspects of bird coloration and avian colour vision including ultraviolet range. Vision Res. 34, 1509–1514. Finger, E., Burkhardt, D., and Dyck, J. (1992). Avian plumage colors. Origin of UV reflection in a black parrot. Naturwiss. 79, 187–188. Fischer, K. E., and Chapman, C. A. (1993). Frugivores and fruit syndromes—Differences in patterns at the genus and species level. Oikos 66, 472–482. Fox, D. L. (1976). ‘‘Animal Biochromes and Structural Colours.’’ Univ. of California Press, Berkeley. Fratzer, C., Dorr, S., and Neumeyer, C. (1994). Wavelength discrimination of the goldfish in the ultraviolet spectral range. Vision Res. 34, 1515–1520. Gendron, R. P. (1984). Searching for cryptic prey: The effect of a search image. Behav. Proc. 9, 303. Gendron, R. P. (1986). Searching for cryptic prey: Evidence for optimal search rates and the formation of search images in quail. Anim. Behav. 34, 898–912. Giurfa, M., Nunez, J., Chittka, L., and Menzel, R. (1995). Color preferences of flower-naive honeybees. J. Comp. Physiol. A 177, 247–259. Goldsmith, T. H. (1980). Hummingbirds see near ultraviolet light. Science 207, 786–788. Goldsmith, T. H. (1990). Optimization, constraint, and history in the evolution of eyes. Q. Rev. Biol. 65, 281–322. Goldsmith, T. H. (1991). The evolution of visual pigments and colour vision. In ‘‘Vision and Visual Disfunction’’ (P. Gouras, Ed.), Vol. 6, pp. 62–89. CRC Press, Boston. Goldsmith, T. H. (1994). Ultraviolet receptors and color vision: Evolutionary implications and a dissonance of paradigms. Vision Res. 34, 1479–1487. Goldsmith, T. H., Collins, J. S., and Perlman, D. L. (1981). A wavelength discrimination function for the hummingbird Archilochus alexandri. J. Comp. Physiol. A 143, 103–110. Goldsmith, T. H., Collins, J. S., and Licht, S. (1984). The cone oil droplets of avian retinas. Vision Res. 24, 1661–1671. Go¨tmark, F. (1992). Antipredator effect of conspicuous plumage in a male bird. Anim. Behav. 44, 51–55. Go¨tmark, F. (1994). Are bright birds distasteful? A reanalysis of H. B. Cott’s data on the edibility of birds. J. Avian Biol. 25, 184–197. Go¨tmark, F. (1996). Simulating a color mutation: Conspicuous red wings in the European blackbird reduce the risk of attacks by sparrowhawks. Funct. Ecol. 10, 355–359. Go¨tmark, F. (1997). Bright plumage in the magpie: Does it increase or reduce the risk of predation? Behav. Ecol. Sociobiol. 40, 41–49. Go¨tmark, F., and Hohlfalt, A. (1995). Bright male plumage and predation risk in passerine birds: Are males easier to detect than females? Oikos 74, 475–484. Go¨tmark, F., and Olsson, J. (1997). Artificial colour mutation: Do red-painted great tits experience increased or decreased predation? Anim. Behav. 53, 83–91.

UV VISION IN BIRDS

207

Go¨tmark, F., and Unger, U. (1994). Are conspicuous birds unprofitable prey? Field experiments with hawks and stuffed prey species. Auk 111, 251–262. Govardovskii, V. I. (1983). On the role of oil drops in colour vision. Vision Res. 23, 1739–1740. Govardovskii, V. I., and Zeuva, L. V. (1977). Visual pigments of chicken and pigeon. Vision Res. 17, 537–543. Greenewelt, C. H., Brandt, W., and Friel, D. D. (1960). Iridescent colors of hummingbird feathers. J. Opt. Soc. Am. A 50, 1005–1013. Griswold, M. S., and Stark, W. S. (1992). Scotopic spectral sensitivity of phakic and aphakic observers extending into the near ultraviolet. Vision Res. 32, 1739–1743. Guilford, T. (1986). How do warning colors work? Conspicuousness may reduce recognition errors in experienced predators. Anim. Behav. 34, 286–288. Guilford, T., and Dawkins, M. S. (1989). Search image versus search rate: Two different ways to enhance prey capture. Anim. Behav. 37, 163–165. Guilford, T., and Dawkins, M. S. (1991). Receiver psychology and the evolution of animal signals. Anim. Behav. 42, 1–14. Guilford, T., and Dawkins, M. S. (1993). Receiver psychology and the design of animal signals. TINS 16, 430–436. Hailman, J. P. (1977). ‘‘Optical Signals: Animal communication and Light.’’ Indiana Univ. Press, Bloomington. Ha´rosi, F. I., and Hashimoto, Y. (1983). Ultraviolet pigments in a vertebrate: A tetrachromatic cone system in the dace. Science 222, 1021–1023. Hart, N. S. (1998). Avian photoreceptors. Unpublished PhD thesis, University of Bristol, Bristol, UK. Hart, N. S., Partridge, J. C., and Cuthill, I. C. (1998). Visual pigments, oil droplets and cone photoreceptor distribution in the European starling (Sturnus vulgaris). J. Exp. Biol. 201, 1433–1446. Hart, N. S., Partridge, J. C., and Cuthill, I. C. (1999). Visual pigments, cone oil droplets, ocular media and predicted spectral sensitivity in the domestic turkey (Meleagris gallopavo). Vision Res. 39, 3321–3328. Hasson, O. (1991). Pursuit-deterrent signals: Communication between prey and predator. TREE 6, 325–329. Hawryshyn, C. W., and McFarland, W. N. (1987). Cone photoreceptor mechanisms and the detection of polarized light in fish. J. Comp. Physiol. A 160, 459–465. Heath, L. A., Wilkie, S. E., Bowmaker, J. K., and Hunt, D. M. (1997). The rod and green cone opsins of two avian species, the budgerigar, Melopsittacus undulatus, and the mallard duck, Anas platyrhynchus. Gene 204, 121–126. Hill, G. E. (1993a). Geographic variation in the carotenoid plumage pigmentation of male house finches (Carpodacus mexicanus). Biol. J. Linnean Soc. 49, 63–86. Hill, G. E. (1993b). Male mate choice and the evolution of female plumage coloration in the house finch. Evolution 47, 1515–1525. Hill, G. E. (1993c). The proximate basis of inter-population and intra-population variation in female plumage coloration in the house finch. Can. J. Zool. 71, 619–627. Hill, G. E., and Montgomerie, R. (1994). Plumage color signals nutritional condition in the house finch. Proc. R. Soc. London B 258, 47–52. Hill, G. E., Montgomerie, R., Inouye, C. Y., and Dale, J. (1994). Influence of dietary carotenoids on plasma and plumage color in the house finch—Intrasexual and intersexual variation. Funct. Ecol. 8, 343–350. Hinton, H. E. (1973). Natural deception. In ‘‘Illusion in Nature and Art’’ (R. L. Gregory and E. H. Gombrich, Eds.), pp. 161–191. Duckworth, London.

208

INNES C. CUTHILL et al.

Hunt, D. M., Dulai, K. S., Cowing, J. A., Julliot, C., Mollon, J. D., Bowmaker, J. K., Li, W. H., and Hewett-Emmett, D. (1998). Molecular evolution of trichromacy in primates. Vision Res. 38, 3299–3306. Hunt, S., Cuthill, I. C., Swaddle, J. P., and Bennett, A. T. B. (1997). Ultraviolet vision and band colour preferences in female zebra finches, Taeniopygia guttata. Anim. Behav. 54, 1383–1392. Hunt, S., Bennett, A. T. D., Cuthill, I. C., and Griffiths, R. (1998). Blue tits are ultraviolet tits. Proc. R. Soc. London B 265, 451–455. Hurlbert, A. (1986). Formal connections between lightness algorithms. J. Opt. Soc. Am. A 3, 1685–1693. Huth, H. H., and Burkhardt, D. (1972). Der Spektrale sehbereich eines Violetta Kolibris. Naturwiss. 59, 650. Jacobs, G. H. (1981). ‘‘Comparative Color Vision.’’ Academic Press, New York. Jacobs, G. H. (1983). Color-vision in animals. Endeavour 7, 137–140. Jacobs, G. H. (1992). Ultraviolet vision in vertebrates. Am. Zool. 32, 544–554. Jacobs, G. H., Neitz, J., and Deegan, J. F. (1991). Retinal receptors in rodents maximally sensitive to ultraviolet light. Nature 353, 655–656. Jane, S. D., and Bowmaker, J. K. (1988). Tetrachromatic colour vision in the duck (Anas platyrhynchos L.): Microspectrophotometry of visual pigments and oil droplets. J. Comp. Physiol. A 162, 225–235. Johnsen, A., Andersson, S., Ornborg, J., and Lifjeld, J. T. (1998). Ultraviolet plumage ornamentation affects social mate choice and sperm competition in bluethroats (Aves: Luscinia s. svecica): A field experiment. Proc. R. Soc. London B 265, 1313–1318. Johnson, S. G. (1991). Effects of predation, parasites, and phylogeny on the evolution of bright coloration in North American male passerines. Evol. Biol. 5, 52–62. Kawamura, S., and Yokoyama, S. (1996). Phylogenetic relationships among short wavelength sensitive opsins of American chameleon (Anolis carolinensis) and other vertebrates. Vision Res. 36, 2797–2804. Kawamuro, K., Irie, T., and Nakamura, T. (1997). Filtering effect of cone oil droplets detected in the P-III response spectra of Japanese quail. Vision Res. 37, 2829–2834. Kevan, P., Giurfa, M., and Chittka, L. (1996). Why are there so many and so few white flowers? TIPS 1, 280–284. Knowles, A., and Dartnall, H. J. A. (1977). ‘‘The Photobiology of Vision.’’ Academic Press, New York. Koivula, M., Korpimaki, E., and Viitala, J. (1997). Do Tengmalm’s owls see vole scent marks visible in ultraviolet light? Anim. Behav. 54, 873–877. Kreithen, M. L., and Keeton, W. T. (1974). Detection of polarized light by the homing pigeon, Columba livia. J. Comp. Physiol. 89, 83–92. Land, M. F. (1972). The physics and biology of animal reflectors. Prog. Biophys. Mol. Biol. 24, 77–123. Langley, C. M., Riley, D. A., Bond, A. B., and Goel, N. (1996). Visual search for natural grains in pigeons (Columba livia): Search images and selective attention. J. Exp. Psychol. Anim. Behav. Proc. 22, 139–151. Langmore, N., and Bennett, A. T. D. (1999). Strategic concealment of sexual identity in an estrildid finch. Proc. R. Soc. London B 266, 543–550. Lawrence, E. S. (1985a). Evidence for search image in blackbirds (Turdus merula L): Longterm learning. Anim. Behav. 33, 1301–1309. Lawrence, E. S. (1985b). Evidence for search image in blackbirds (Turdus merula L): Shortterm learning. Anim. Behav. 33, 929–937.

UV VISION IN BIRDS

209

Lee, W. G., Weatherall, I. L., and Wilson, J. B. (1994). Fruit conspicuousness in some New Zealand Coprosma (Rubiaceae) species. Oikos 69, 87–94. Liebman, P. A. (1972). Microspectrophotometry of photoreceptors. In ‘‘Photochemistry of Vision’’ (H. J. A. Dartnall, Ed.), Vol. 7, No. 1, pp. 481–528. Springer-Verlag, Berlin. Liebman, P. A., and Granda, A. M. (1971). Microspectrophotometric measurements of visual pigments in two species of turtle Pseudemys scripta and Chelonia mydas. Vision Res. 11, 105–114. Liebman, P. A., and Granda, A. M. (1975). Super dense carotenoid spectra resolved in single cone oil droplets. Nature 253, 370–372. Lipetz, L. E. (1984). A new method for determining peak absorbance of dense pigment samples and its application to the cone oil droplets of Emydoidea blandingii. Vision Res. 24, 597–604. Loew, E. R., and Lythgoe, J. N. (1985). The ecology of colour vision. Endeavour 14, 170–174. Lunau, K., and Maier, E. J. (1995). Innate color preferences of flower visitors. J. Comp. Physiol. A 177, 1–19. Lunau, K., Wacht, S., and Chittka, L. (1996). Colour choices of naive bumble bees and their implications for colour perception. J. Comp. Physiol. A 178, 477–489. Lythgoe, J. N. (1979). ‘‘The Ecology of Vision.’’ Oxford Univ. Press, Oxford. Maier, E. J. (1992). Spectral sensitivities including the ultraviolet of the passeriform bird Leiothrix lutea. J. Comp. Physiol. A 170, 709–714. Maier, E. J. (1993). To deal with the invisible: On the biological significance of ultraviolet sensitivity in birds. Naturwiss. 80, 476–478. Maier, E. J., and Bowmaker, J. K. (1993). Colour vision in the passeriform bird, Leiothrix lutea: Correlation of visual pigment absorbency and oil droplet transmission with spectral sensitivity. J. Comp. Physiol. A 172, 295–301. Majerus, M. E. N. (1998). ‘‘Melanism—Evolution in Action.’’ Oxford Univ. Press, Oxford. Mariani, A. P., and Leure-DuPree, A. E. (1978). Photoreceptors and oil droplet colours in the red area of the pigeon retina. J. Comp. Neurol. 182, 821–838. Marples, N. M., and Roper, T. J. (1996). Effects of novel color and smell on the response of naive chicks towards food and water. Anim. Behav. 51, 1417–1424. Martin, G. R. (1985). Eye. In ‘‘Form and Function in Birds’’ (A. S. King and J. McLelland, Eds.), Vol. 3. Academic Press, London. Martin, G. R. (1986). The eye of a passeriform bird, the European starling (Sturnus vulgaris): Eye movement amplitude, visual fields and schematic optics. J. Comp. Physiol. A 159, 545–557. Martin, G. R. (1990). ‘‘Birds by Night.’’ Poyser, London. Mason, J. R., Arzt, A. H., and Reidinger, R. F. (1984). Comparative assessment of food preferences and aversions acquired by blackbirds via observational learning. Auk 101, 796–803. Menzel, J. G., Wunderer, H., and Stavenga, D. G. (1991). Functional morphology of the divided compound eye of the honeybee drone (Apis mellifera). Tissue Cell 23, 525–535. Meyer, D. B. (1977). The avian eye and its adaptations. In ‘‘The Visual System in Vertebrates’’ (F. Crescitelli, Ed.), Vol. 7, No. 5, pp. 549–611. Springer-Verlag, Berlin. Meyer, D. B., Cooper, T. G., and Gernez, C. (1965). Retinal oil droplets. In ‘‘The Structure of the Eye: II, Symposium’’ (J. W. Rohen, Ed.). Schattauer-Verlag, Stuttgart. Meyer-Rochow, V. B. (1991). Differences in ultraviolet wing patterns in the New Zealand lycaenid butterflies Lycaena salustius, L. rauparaha, and L. feredayi as a likely isolating mechanism. J. R. Soc. N. Z. 21, 169–177. Meyer-Rochow, V. B., and Eguchi, E. (1983). ‘‘Flugelfarben, wie sie falter sehen’’: A study of UV and other colour patterns in Lepidoptera. Ann. Zool. Jpn. 56, 85–99.

210

INNES C. CUTHILL et al.

Meyer-Rochow, V. B., and Ja¨rvilheto, M. (1997). Ultraviolet colours in Pieris napi from northern and southern Finland: Arctic females are the brightest! Naturwiss. 84, 165–168. Møller, A. P. (1989). Natural and sexual selection on a plumage signal of status and on morphology in house sparrows, Passer domesticus. J. Evol. Biol. 2, 125–140. Morris, V. B. (1970). Symmetry in a receptor mosaic demonstrated in the chick from the frequencies, spacing and arrangement of the types of retinal receptor. J. Comp. Neurol. 140, 359–398. Morris, V. B., and Shorey, C. D. (1967). An electron microscope study of types of receptor in the chick retina. J. Comp. Neurol. 129, 313–340. Munro, U., and Wiltschko, R. (1995). The role of skylight polarization in the orientation of a day-migrating bird species. J. Comp. Physiol. A 177, 357–362. Munro, U., Munro, J. A., Phillips, J. B., and Wiltschko, W. (1997). Effect of wavelength of light and pulse magnetisation on different magnetoreception systems in a migratory bird. Aust. J. Zool. 45, 189–198. Muntz, W. R. A. (1972). Inert absorbing and reflecting pigments. In ‘‘Photochemistry of Vision’’ (H. J. A. Dartnall, Ed.), Vol. 7, No. 1, pp. 529–565. Springer-Verlag, Berlin. Murray, K. G., Winnettmurray, K., Cromie, E. A., Minor, M., and Meyers, E. (1993). The influence of seed packaging and fruit color on feeding preferences of American robins. Vegetatio 108, 217–226. Nakanishi, K. (1991). 11-cis-retinal, a molecule uniquely suited for vision. Pure Appl. Chem. 63, 161–170. Neumeyer, C. (1992). Tetrachromatic color vision in goldfish: Evidence from color mixture experiments. J. Comp. Physiol. A 171, 639–649. Neumeyer, C., and Arnold, K. (1989). Tetrachromatic colour vision in the goldfish becomes trichromatic under white adaptation light of moderate intensity. Vision Res. 29, 1719–1727. Ohtsuka, T. (1985). Relation of spectral types to oil droplets in cones of turtle retina. Science 229, 874–877. Okano, T., Kojima, D., Fukada, Y., Shichida, Y., and Yoshizawa, T. (1992). Primary structures of chicken cone visual pigments: Vertebrate rhodopsins have evolved out of cone visual pigments. Proc. Natl. Acad. Sci. USA 89, 5932–5936. Osche, G. (1983). Optical (visual) signals in the coevolution of plants and animals. Berichte Deutschen Bot. Gesellschaft 96, 1–27. Osorio, D., Marshall, N. J., and Cronin, T. W. (1997). Stomatopod photoreceptor tuning as an adaptation to colour constancy underwater. Vision Res. 37, 3299–3309. Osorio, D., Jones, C. D., and Vorobyev, M. (1999). Accurate memory for colour but not pattern contrast in chicks. Curr. Biol. 9, 199–202. Owens, E. D., and Prokopy, R. J. (1986). Relationship between reflectance spectra of host plant surfaces and visual detection of host fruit by Rhagoletis pomonella flies. Physiol. Entomol. 11, 297–307. Owens, I. P. F., and Hartley, I. R. (1998). Sexual dimorphism in birds: Why are there so many different forms of dimorphism? Proc. R. Soc. London B 265, 397–407. Palacios, A. G., and Varela, F. J. (1992). Color mixing in the pigeon (Columbia livia). 2. A psychophysical determination in the middle, short and near-UV wavelength range. Vision Res. 32, 1947–1953. Palacios, A. G., Goldsmith, T. H., and Bernard, G. D. (1996). Sensitivity of cones from a cyprinid fish (Danio aequipinnatus) to ultraviolet and visible light. Visual Neurosci. 13, 411–421.

UV VISION IN BIRDS

211

Parrish, J. W., Ptacek, J. A., and Will, K. L. (1984). The detection of near-ultraviolet light by nonmigratory and migratory birds. Auk 101, 53–58. Partridge, J. C. (1989). The visual ecology of avian cone oil droplets. J. Comp. Physiol. A 165, 415–426. ¨ lkugeln im Farbensehen der Sauropsiden. Ann. Peiponen, V. A. (1964). Zur Bedeutung der O Zool. Fenn. 1, 281–302. Peitsch, D., Fietz, A., Hertel, H., Desouza, J., Ventura, D. F., and Menzel, R. (1992). The spectral input systems of hymenopteran insects and their receptor-based color-vision. J. Comp. Physiol. A 170, 23–40. Phillips, J. B., and Borland, S. C. (1992). Behavioural evidence for use of a light-dependent magnetoreception mechanism by a vertebrate. Nature 359, 142–144. Phillips, J. B., and Borland, S. C. (1994). Use of a specialized magnetoreception system for homing by the Eastern red-spotted newt Notophthalmus viridescens. J. Exp. Biol. 188, 275–291. Phillips, J. B., and Moore, F. R. (1992). Calibration of the sun compass by sunset polarized light patterns in a migratory bird. Behav. Ecol. Sociobiol. 31, 189–193. Phillips, J. B., and Sayeed, O. (1993). Wavelength-dependent effects of light on magnetic compass orientation in Drosophila melanogaster. J. Comp. Physiol. A 172, 303–308. Plaisted, K. C., and Mackintosh, N. J. (1995). Visual search for cryptic stimuli in pigeons: Implications for the search image and search rate hypotheses. Anim. Behav. 50, 1219–1232. Pohl, H. (1992). Ultraviolet radiation—A zeitgeber for the circadian clock in birds. Naturwiss. 79, 227–229. Provencio, I., Loew, E. R., and Foster, R. G. (1992). Vitamin-A2-based visual pigments in terrestrial vertebrates. Vision Res. 32, 2201–2208. Prum, R. O., Torres, R. H., Williamson, S., and Dyck, J. (1998). Coherent light scattering by blue feather barbs. Nature 396, 28–29. Puckey, H. L., Lill, A., and O’Dowd, D. J. (1996). Fruit color choices of captive Silvereyes (Zosterops lateralis). Condor 98, 780–790. Radwan, J. (1993). Are dull birds still dull in UV? Acta Ornithol. 27, 125–130. Raven, P. H. (1972). Why are bird-visited flowers predominately red? Evolution 76, 674. Reid, P. J., and Shettleworth, S. J. (1992). Detection of cryptic prey: Search image or search rate? J. Exp. Psychol. Anim. Behav. Proc. 18, 273–286. Remington, C. L. (1973). Ultraviolet reflectance in mimicry and sexual signals in Lepidopetra. J. N. Y. Entomol. Soc. 81, 124–125. Ridley, H. N. (1930). ‘‘The Dispersal of Plants throughout the World.’’ Reeve, Ashford, London. Rodieck, R. W. (1998). ‘‘First Steps in Seeing.’’ Sinauer, Sunderland, MA. Rohwer, S. (1985). Dyed birds achieve higher social status than controls in Harris’ sparrows. Anim. Behav. 33, 1325–1331. Roper, T. J., and Marples, N. M. (1997a). Colour preferences of domestic chicks in relation to food and water presentation. Appl. Anim. Behav. Sci. 54, 207–213. Roper, T. J., and Marples, N. M. (1997b). Odour and colour as cues for taste-avoidance learning in domestic chicks. Anim. Behav. 53, 1241–1250. Rowe, C., and Guilford, T. (1996). Hidden color aversions in domestic chicks triggered by pyrazine odors of insect warning displays. Nature 383, 520–522. Ruderman, D. L., Cronin, T. W., and Chiao, C. C. (1998). Statistics of cone responses to natural images: Implications for visual coding. J. Opt. Soc. Am. A 15, 2036–2045. Ryan, M. J. (1997). Sexual selection and mate choice. In ‘‘Behavioral Ecology: An Evolutionary Approach’’ (J. R. Krebs and N. B. Davies, Eds.), pp. 179–202. Blackwell Sci., Oxford.

212

INNES C. CUTHILL et al.

Ryan, M. J., and Keddy-Hector, A. (1992). Directional patterns of female mate choice and the role of sensory biases. Am. Nat. 139, S4–S35. Ryan, M. J., and Rand, A. S. (1993). Sexual selection and signal evolution: The ghost of biases past. Philos. Trans. R. Soc. London B 340, 187–195. Sallabanks, R. (1993). Hierarchical mechanisms of fruit selection by an avian frugivore. Ecology 74, 1326–1336. Sanders, M. J., Owen-Smith, R. N., and Pillay, N. (1997). Fruit selection in the olive thrush: The importance of colour. S. Afr. J. Zool. 32, 21–23. Schultze, M. (1866). Zur Anatomie und Physiologie der Retina. Arch. Mikroskop. Anat. 2, 175–286. Schultze, M. (1867). Ueber Sta¨bchen und Zapfen der Retina. Arch. Mikroskop. Anat. 3, 215–247. Sibley, C. G., and Monroe, B. L. (1990). ‘‘Distribution and Taxonomy of Birds of the World.’’ Yale Univ. Press, New Haven, CT. Silberglied, R. E. (1979). Communication in the ultraviolet. Annu. Rev. Ecol. Syst. 10, 373–398. Silberglied, R. E. (1984). Visual communication and sexual selection among butterflies. In ‘‘The Biology of Butterflies. Symposium of the Royal Society of London, Number 11’’ (R. I. Vane-Wright and P. E. Ackery, Eds.), pp. 207–223. Academic Press, London. Sillman, A. J., Bolnick, D. A., Haynes, L. W., Walter, A. E., and Loew, E. R. (1981). Microspectrophotometry of the photoreceptors of palaeognathous birds: The emu and tinamou. J. Comp. Physiol. 144, 271–276. Snow, B., and Snow, D. (1988). ‘‘Birds and Berries: A Study of an Ecological Interaction.’’ Poyser, Calton, UK. Starace, D. M., and Knox, B. E. (1998). Cloning and expression of a Xenopus short wavelength cone pigment. Exp. Eye Res. 67, 209–220. Stark, W. S., Wagner, R. H., and Gillespie, C. M. (1994). Ultraviolet sensitivity of three cone types in the aphakic observer determined by chromatic adaptation. Vision Res. 34, 1457–1459. Stavenga, D. G., Smits, R. P., and Hoenders, B. J. (1993). Simple exponential functions describing the absorbance bands of visual pigment spectra. Vision Res. 33, 1011–1017. Stryer, L. (1987). The molecules of visual excitation. Sci. Am. 257, 32–40. Swaddle, J. P., and Cuthill, I. C. (1994). Female zebra finches prefer males with symmetrical chest plumage. Proc. R. Soc. London B 258, 267–271. Takao, M., Yasui, A., and Tokunaga, F. (1988). Isolation and determination of the chicken rhodopsin gene. Vision Res. 28, 471–480. Thompson, E., Palacios, A., and Varela, F. J. (1992). Ways of coloring: Comparative color vision as a case study for cognitive science. Behav. Brain Sci. 15, 1–74. Tove´e, M. J. (1995). Ultraviolet photoreceptors in the animal kingdom: Their distribution and function. TREE 10, 455–460. Veiga, J. P. (1993). Badge size, phenotypic quality, and reproductive success in the house sparrow: A study on honest advertisement. Evolution 47, 1161–1170. Viitala, J., Korpimaki, E., Palokangas, P., and Koivula, M. (1995). Attraction of kestrels to vole scent marks visible in ultraviolet light. Nature 373, 425–427. Vorobyev, M., and Osorio, D. (1998). Receptor noise as a determinant of colour thresholds. Proc. R. Soc. London B 265, 351–358. Vorobyev, M., Gumbert, A., Kunze, J., Giurfa, M., and Menzel, R. (1997). Flowers through insect eyes. Israel J. Plant Sci. 45, 93–101. Vorobyev, M., Osorio, D., Bennett, A. T. D., Marshall, N. J., and Cuthill, I. C. (1998). Tetrachromacy, oil droplets and bird plumage colours. J. Comp. Physiol. A 183, 621–633.

UV VISION IN BIRDS

213

Vos Hzn, J. J., Coemans, M. A. J. M., and Nuboer, J. F. W. (1994). The photopic sensitivity of the yellow field of the pigeon’s retina to ultraviolet light. Vision Res. 34, 1419–1425. Vos Hzn, J. J., Coemans, M. A. J. M., and Nuboer, J. F. W. (1995). No evidence for polarization sensitivity in the pigeon electroretinogram. J. Exp. Biol. 198, 325–335. Waelchli, G. (1883). Zur Topographie der gefa¨rbten Kugeln der Vogelnetzhaut. Arch. Ophthalmol. 29, 205–223. Wald, G., and Zussman, H. (1937). Carotenoids of the chicken retina. Nature 140, 197. Walls, G. L. (1963). ‘‘The Vertebrate Eye and Its Adaptive Radiation.’’ Hafner, New York. Wang, S.-Z., Adler, R., and Nathans, J. (1992). A visual pigment from chicken that resembles rhodopsin: Amino acid sequence, gene structure, and functional expression. Biochemistry 31, 3309–3315. Wang, Z., Asenjo, A. B., and Oprian, D. D. (1993). Identification of the Cl⫺-binding site in the human red and green colour vision pigments. Biochemistry 32, 2125–2130. Wehner, R. (1989). Neurobiology of polarization vision. TINS 12, 353–359. Wheelwright, N. T., and Janson, C. H. (1985). Colors of fruit displays of bird-dispersed plants in two tropical forests. Am. Nat. 126, 777–799. Whelan, C. J., and Willson, M. F. (1994). Fruit choice in migrating North American birds—Field and aviary experiments. Oikos 71, 137–151. Wilkie, S. E., Vissers, P. M. A. M., Das, D., DeGrip, W. J., Bowmaker, J. K., and Hunt, D. M. (1998). The molecular basis for UV vision in birds: Spectral characteristics, cDNA sequence and retinal localization of the UV-sensitive visual pigment of the budgerigar (Melopsittacus undulatus). Biochem. J. 330, 541–547. Willson, M. F., and Whelan, C. J. (1989). Ultraviolet reflectance of fruits of vertebratedispersed plants. Oikos 55, 341–348. Willson, M. F., and Whelan, C. J. (1990). The evolution of fruit color in fleshy-fruited plants. Am. Nat. 136, 790–809. Willson, M. F., Graff, D. A., and Whelan, C. J. (1990). Color preferences of frugivorous birds in relation to the colors of fleshy fruits. Condor 92, 545–555. Wiltschko, R., and Wiltschko, W. (1998). Pigeon homing: Effect of various wavelengths of light during displacement. Naturwiss. 85, 164–167. Wiltschko, W., and Wiltschko, R. (1993). Navigation in birds and other animals. J. Navigation 46, 174–191. Wiltschko, W., and Wiltschko, R. (1995). Migratory orientation of European robins is affected by the wavelength of light as well as by a magnetic pulse. J. Comp. Physiol. A 177, 363–369. Wiltschko, W., Munro, U., Ford, H., and Wiltschko, R. (1993). Red-light disrupts magnetic orientation of migratory birds. Nature 364, 525–527. Wright, A. A. (1972a). The influence of ultraviolet radiation on the pigeons’ color discrimination. J. Exp. Anal. Behav. 17, 325–337. Wright, A. A. (1972b). Psychometric and psychophysical hue discrimination functions for the pigeon. Vision Res. 12, 1447–1464. Wright, M. V., and Bowmaker, J. K. (1998). Visual pigments and oil droplets in the retinae of palaeognathus birds. Invest. Ophthalmol. 39, S1059. Yau, K.-W. (1994). Phototransduction mechanism in retinal rods and cones. Invest. Ophthalmol. 35, 9–32. Yokoyama, S., and Yokoyama, R. (1996). Adaptive evolution of photoreceptors and visual pigments in vertebrates. Annu. Rev. Ecol. Syst. 27, 543–567. Yokoyama, S., Radlwimmer, F. B., and Kawamura, S. (1998). Regeneration of ultraviolet pigments of vertebrates. FEBS Lett. 423, 155–158.

214

INNES C. CUTHILL et al.

Young, S. R., and Martin, G. R. (1984). Optics of retinal oil droplets: A model of light collection and polarisation detection in the avian retina. Vision Res. 24, 129–137. Zann, R. (1996). ‘‘The Zebra Finch: A Synthesis of Field and Laboratory Studies.’’ Oxford Univ. Press, Oxford. Zeigler, H. P., and Bischof, H.-J. (1993). ‘‘Vision, Brain and Behaviour in Birds.’’ MIT Press, Cambridge. Zuk, M., and Decruyenaere, J. G. (1994). Measuring individual variation in color: A comparison of two techniques. Biol. J. Linnean Soc. 53, 165–173.

ADVANCES IN THE STUDY OF BEHAVIOR, VOL. 29

What Is the Significance of Imitation in Animals? Cecilia M. Heyes and Elizabeth D. Ray department of psychology university college london london wc1e 6bt, united kingdom

I. Introduction For at least a century, and with increasing rigor and sophistication in recent years, psychologists and biologists have investigated whether nonhuman animals (henceforward ‘‘animals’’) are capable of a certain kind of social learning, namely, imitation (Thorndike, 1898; Galef, 1988; Tomasello, 1996; Tomasello et al., 1993b; Whiten and Ham, 1992). Imitation consists of response learning by observation, i.e., learning how to move the body by observing the behavior of others. Other varieties of social learning consist of stimulus learning by observation; these are means of acquiring information about the static or dynamic properties of objects; about their value, location, and motion (Heyes, 1994). Although this article includes a survey of some of the most interesting recent experiments on imitation in animals, our main purpose is not to address the question ‘‘Can animals imitate?’’ Instead, we offer an answer to a related and somewhat neglected question. What is the significance of imitation in animals; what would be the advantage of knowing whether animals are capable of response learning by observation? We will argue that the principal significance of this field of animal behavior research lies in what it can reveal about the cognitive mechanisms underlying imitation in humans and animals; specifically, that it has the potential to establish whether those mechanisms are ‘‘transformational’’ or ‘‘associative.’’ The greatest challenge for any theory of the cognitive mechanisms of imitation is to explain imitation of ‘‘perceptually opaque’’ actions, those actions which yield dissimilar sensory inputs when observed and executed. Section II explains why this is difficult, and Section III distinguishes two types of theory of imitation, transformational and associative, in terms of the way in which they attempt to meet this challenge. Although it is fairly clear whether each existing theory postulates transformational or associative processes, the models currently available are not specified in sufficient 215

Copyright 䉷 2000 by Academic Press All rights of reproduction in any form reserved. 0065-3454/00 $35.00

216

CECILIA M. HEYES AND ELIZABETH D. RAY

detail to allow empirical testing of conflicting predictions. Section IV sketches a new, associative sequence learning (ASL) theory which is more amenable to empirical testing, and Section V discusses the kind of evidence of imitation in animals that would favor this theory over transformational alternatives or vice versa. In Section VI we search for this kind of evidence among recent experiments on imitation in animals using two-action test procedures, and in Section VII we consider briefly the significance of imitation in animals with respect to culture rather than cognition.

II. Perceptual Opacity All behaviors or actions may be said to lie on a continuum of perceptual opacity; they vary in the degree to which they yield dissimilar sensory inputs when observed and executed (Fig. 1). Highly perceptually opaque actions, which usually generate highly dissimilar sensory inputs, include head movements and facial gestures. For example, under typical observation conditions, when a human observer (O), sees another person, a ‘‘demonstrator’’ (D ), raising an eyebrow, the sensory input to O is primarily visual and includes the movement of an arc (the eyebrow) in the upper portion of an elliptical frame (the face). In contrast, when O raises his own eyebrow, the sensory input to O is primarily kinesthetic, the movement is felt rather than seen, with any visual component consisting largely of an increase in the amount of light entering one eye. Perceptually transparent movements, those which are low on the dimension of perceptual opacity, yield relatively similar sensory inputs when observed and executed and typically include distal appendage movements and vocalizations. For example, although O receives kinesthetic input when he fans his fingers and not when he observes D performing the same finger movements, the pattern of visual input to

Fig. 1. Summary of the discussion of perceptual opacity, a new, imitation-relevant dimension of action.

WHAT IS THE SIGNIFICANCE OF IMITATION IN ANIMALS?

217

O is very similar in the two cases, especially when O stands behind or alongside D while observing. Perceptual opacity is not an invariant property of an action. It can vary with observation conditions, including the spatial orientation of O’s sense organs in relation to D and to the part of O’s body that moves during action execution, and via the use of instruments such as mirrors and video recorders. For example, mirrors and manual, tactile exploration of D’s and O’s body by O can render highly transparent actions, such as facial gestures, that are normally highly opaque. Imitation of perceptually opaque actions is more difficult to explain than imitation of perceptually transparent actions because in the former case it is not clear how O could derive the information necessary to produce a behavioral match. In principle, imitation of perceptually transparent actions could be achieved through a sensory matching process in which O generates variant actions, compares the sensory feedback from these ‘‘trials’’ with a concurrently present or memorial sensory representation of D’s action, and selects the variant(s) for which the discrepancy is minimal. However, a simple sensory matching process such as this cannot explain imitation of more perceptually opaque actions. In recognition of what we are calling the problem of perceptual opacity, Thorndike (1898) argued that ‘‘motor imitation’’ is a more impressive cognitive feat than ‘‘vocal imitation,’’ and Piaget (1951) suggested that imitation of ‘‘invisible’’ movements represents a later stage in cognitive development than imitation of visible actions such as finger movements. The dimension of perceptual opacity subsumes these earlier distinctions and may be more helpful in the development of an adequate theory of imitation for two reasons. First, unlike Piaget’s distinction, it draws attention to the fact that a simple sensory matching hypothesis is insufficient to explain imitation not only of actions without visual feedback but also of those which yield different visual inputs when observed and executed. Second, and in contrast with Thorndike’s distinction, it acknowledges that sensory matching can explain imitation of some nonvocal behaviors, and that this depends both on the body parts involved in the action and on the observation conditions such as viewing angle. Thus, the problem confronting theories of imitation, the problem of perceptual opacity, is broader than the visible/invisible distinction implies, narrower than the motor/vocal distinction implies, and may be more accurately described by a dimension than a dichotomy. III. Theories of Imitation Investigators of imitation in animals seldom refer to any explicit theory of imitation, and it is therefore surprising that at least a dozen such theories

218

CECILIA M. HEYES AND ELIZABETH D. RAY

have been elaborated in modern terms and some significant degree of detail. We divide these theories into two groups according to whether they attribute imitation to transformational or associative psychological processes. Generally, associative theories (e.g., Allport, 1924; Aronfreed, 1969; Gewirtz, 1971; Holt, 1931; Miller and Dollard, 1941; Mowrer, 1960; Skinner, 1953) claim that the information required to produce an imitative match between the behavior of an observer and that of its demonstrator (or model) is derived from experience. They suggest that the capacity to imitate a given action, X, now, derives from experience of simultaneously observing and executing X in the past. In contrast, transformational theories (e.g., Bandura, 1986; Meltzoff, 1990; Piaget, 1951) assert that a substantial portion of the information necessary to produce a behavioral match is internally generated by complex cognitive processes. These processes transform the sensory input from the demonstrator’s action into a ‘‘symbolic conception’’ (Bandura, 1986), ‘‘imaged representation’’ (Piaget, 1951), or ‘‘supramodal representation’’ (Meltzoff, 1985) which contains the information necessary to guide execution of matching behavior by the observer. Associative and transformational theories represent two important, plausible, alternative accounts of the processes underlying imitation. However, each existing theory has shortcomings which make it unsuitable as a framework for empirical investigation of whether, or to what extent, associative and transformational processes mediate imitation. A problem common to transformational theories is underspecification. They do not indicate how information in sensory input from the demonstrator is transformed into a representation capable of guiding production of matching behavior. In other words, although transformational theories claim that, even for perceptually opaque actions, the observer’s information processing system can translate sensory input from the demonstrator into a production code, they do not give any hint of the mapping functions involved. For example, Bandura’s (1986) ‘‘social-cognitive’’ theory suggests that information obtained during observation of a demonstrator’s action is first stored as a sensory representation and then transformed into a ‘‘symbolic conception’’ which ‘‘provides the internal model for response production and the standard for response correction.’’ In other words, the symbolic conception can both generate motor programs for approximately matching behavior and edit these programs to produce more precisely matching behavior using sensory feedback from action execution. Thus, Bandura’s theory suggests a three-part cognitive architecture for imitation consisting of a sensory representation, symbolic conception, and motor program; however, it is silent regarding the process or mechanism by which imitation of perceptually opaque actions is achieved. It does not say how a sensory

WHAT IS THE SIGNIFICANCE OF IMITATION IN ANIMALS?

219

representation is converted into a symbolic conception capable of minting motor programs for matching behavior. As a consequence of their underspecification, transformational theories do not make predictions about the conditions of imitation—about which actions can and cannot be imitated, under what observation conditions, and by which species or individuals—and this, in turn, makes them resistant to empirical evaluation. Associative theories have a variety of weaknesses. For example, Mowrer’s (1960) two-factor theory and Aronfreed’s (1969) template theory postulate sensory matching processes, and the authors apply them only to perceptually transparent actions. This may be interpreted as an implicit prediction that perceptually opaque actions cannot be imitated. If so, it seems that these theories are falsified by the common experience that adult humans can, for example, imitate facial expressions. If not, two-factor theory and template theory are radically incomplete because they simply do not address the question of how perceptually opaque actions are imitated. Holt’s (1931) associative theory is more clearly inconsistent with what is known about human imitative competence. Holt suggested that the capacity to imitate is acquired through social interactions in which an adult faces a child and mirror imitates the child’s actions. For example, when the child moves his right arm to the right of his body, the adult moves her left arm to the right of the child’s body. The child’s gaze follows the adult’s imitative movement and thus, according to Holt, the child associates the stimuli that initiated his own movement (nature unspecified) with sensory feedback from visual tracking to the right of his body. As a consequence of this association, the child will subsequently respond to sight of a demonstrator’s arm moving toward the right of the child’s visual field with an arm movement to the same location in space. Holt’s theory is ingenious, but it makes the false prediction that humans will be incapable of transposition imitation, e.g., imitating right arm movements with their right arm, not their left, when they are facing the demonstrator. To achieve this capacity under the conditions and via the mechanisms specified by Holt’s theory it would be necessary for adults regularly to imitate the actions of infants while the adults have their backs turned to the infants. In fact, humans are capable of transposition from age 7 or 8, and this mode of imitation is preferred to mirroring from approximately age 14 onwards (Gordon 1922/1923; Heyes et al., 1999; Wapner and Cirillo, 1968). Other associative theories, like transformational theories, are untestable as a result of their underspecification. Miller and Dollard’s (1941) copying theory (not to be confused with their matched-dependent theory) has this problem, and since it is probably the most fully elaborated account of

220

CECILIA M. HEYES AND ELIZABETH D. RAY

imitation to date it is worth considering in some detail both the postulates and the key weaknesses of copying theory. Using the example of a person being taught to sing, Miller and Dollard’s (1941) copying theory offered an account of how this O is able to match any note, or sequence of notes, sung by another person. In stage 1 of training to match a single note (e.g., C), the demonstrator/teacher detects sameness and difference cues and O responds to them randomly. Thus, when O voices a note higher or lower than C, the teacher says ‘‘No’’ or ‘‘That’s wrong’’; this makes O feel anxious and initiates random variation in the note he is producing. When O finally hits on C, the teacher says ‘‘Yes’’ or ‘‘Good,’’ and the learner feels relieved. In stage 2, the teacher says ‘‘Too high’’ or ‘‘Too low’’ when O is producing the wrong note, and O responds directionally by producing a higher or lower note. In stages 3 and 4, O becomes able to detect the sameness and difference cues. When the trainer says ‘‘No,’’ ‘‘Yes,’’ ‘‘Too high,’’ or ‘‘Too low,’’ O repeats these words to himself and then experiences the twinge of anxiety or feeling of relief originally provoked by the trainer’s utterance. The learner’s implicit repetition of the words occurs in close temporal proximity to reinforcement (an increase or decrease in anxiety) and therefore, according to Hullian learning theory, these responses become ‘‘anticipatory’’; they begin to occur in direct response to the sameness and difference cues. After some practice guided by direct detection of the sameness and difference cues, the learner will be able to match a C reliably at first attempt, and he can then move on to other notes. Miller and Dollard (1941) asserted that it would be easier for the learner to match each successive note attempted in training because the sameness and difference cues would have something in common with those of previous notes in the sequence, and therefore ‘‘generalization’’ would occur. Having learned to copy every single musical note in stage 5, in stage 6 O learns, through the processes described for stages 1–5, to copy sequences of notes. Stage 6 learning is facilitated, via generalization, by prior training to match single notes. After learning to match an unspecified range of sequences, Miller and Dollard (1941) claimed that the one-time novice would be an expert capable of copying without practice novel sequences of notes. Copying theory has several virtues relative to earlier theories of imitation, but it also has two significant weaknesses—the first distinctive and the second in common with some other associative theories that assign an important role to reinforcement (Gewirtz, 1971; Gewirtz and Stingle, 1968; Miller and Dollard, 1941; matched-dependent theory). The distinctive weakness relates to Miller and Dollard’s claim that, at stage 3 in learning to copy, O’s behavior comes under the direct control of

WHAT IS THE SIGNIFICANCE OF IMITATION IN ANIMALS?

221

sameness and difference cues via repetition of D’s instructions by O. This implies that before an O can learn to copy any other actions, he or she must be able to copy, albeit implicitly, D’s instructions—the behavior through which Ds communicate whether a putative imitation is right or wrong. If this is the case, however, how do Os achieve this initial feat of imitation? How do they come to be able to copy the instructions themselves? The second problem with copying theory is that it assigns a critical role to generalization without specifying dimensions of generalization. This problem is almost immediately apparent in other reinforcement theories, but it is concealed in Miller and Dollard’s (1941) copying theory because their exposition deals almost exclusively with imitative singing of musical notes. Actions in this domain (e.g., ‘‘singing C’’ and ‘‘singing D’’) have two unusual properties: (i) They are defined, differentiated one from another, in terms of their sensory consequences rather than the effectors involved, and (ii) the sensory consequences that define the actions can be ordered on a known scale. These features of singing are important because, in combination, they make it plausible that generalization would, for example, make it easier to learn to copy a D than an E note after learning to copy a C. It is possible to specify the psychophysical dimensions on which the sameness cue DD is more like CC than is EE and on which the difference cue DB is more like CB than is EB. However, most actions cannot be ordered on a scale (or at least we do not know the scale on which they can be ordered) and therefore, for most actions, the claim that there is generalization of learning to copy could be tested only on the basis of ancilliary hypotheses about generalization gradients; currently, these would be very difficult even to formulate. Consider, for example, a person who has been trained to copy a curling movement of his or her left index finger and is now learning to copy a curling motion of the left ring finger. Will this person’s prior training help him or her, via generalization, more or less than if it had involved imitation of a curling movement of the right ring finger, a curling movement of the third toe on either foot, or a rigid, up and down movement of the left ring finger? Because the action ‘‘curling a finger’’ is not part of a known scale, copying theory does not make any obvious predictions. Skinner (1953) provided a sketch of an associative, reinforcementbased theory of imitation which does not use the concept of generalization but, unfortunately, suffers from underspecification. Skinner suggested that in order to imitate an action now, such as a dance step, O must have performed the very same dance step in the past and been rewarded while observing the dance step performed by a D. This means that imitation, response learning by observation, is impossible, and that an

222

CECILIA M. HEYES AND ELIZABETH D. RAY

individual cannot expand his or her behavioral repertoire through action observation. However, Skinner implied through one of his examples that novel sequences of actions can be acquired through observation. Thus, he suggested that a skilled dancer who has previously learned to imitate each step in a dance sequence can copy the whole sequence when it is demonstrated by an instructor. This would be latent, or ‘‘behaviorally silent’’ learning (Dickinson, 1980) (learning without action), and therefore it would be inconsistent with Skinner’s behaviorist analysis of learning. Therefore, it is perhaps not surprising that he did not suggest a mechanism through which it could be achieved.

IV. Associative Sequence Learning Theory In the preceding section, we argued that existing theories of imitation successfully delineate two kinds of processes (transformational and associative) that could mediate imitation, but that none of these theories are couched in a way that makes it possible to test empirically whether, or to what extent, transformational and/or associative processes are responsible for imitation in the real world. In this section, we outline a new ASL theory of imitation which incorporates components from several preceding associative models. The principal purpose of ASL theory is to stimulate the development of other theories (associative and transformational) that are (i) consistent with what is already known about the conditions of imitation and (ii) sufficiently well specified to generate testable predictions. We hope that ASL theory meets these conditions, but we would be very surprised if it turned out to provide a fully accurate account of imitation. A. Action Units ASL theory is schematically represented in Fig. 2. It assumes that, rather than being unitary, the vast majority of actions comprise sequences of component actions or ‘‘action units.’’ Thus, although it is conventional to think and speak of ‘‘an action’’ being imitated, ASL assumes that it is always a sequence of action units that is imitated. The hand icons at the top of Fig. 2 represent a sequence of hand movements: pointing, followed by splaying the fingers, followed by a victory sign. We use this action sequence for illustrative purposes, but two considerations should be borne in mind. First, the action units involved in any given case of imitation may be smaller (e.g., closing one finger toward the palm) or larger (e.g., incorporating pointing and splaying). Second, ASL theory applies to rela-

WHAT IS THE SIGNIFICANCE OF IMITATION IN ANIMALS?

223

Fig. 2. Schematic representation of the associative sequence learning theory of imitation learning.

tively perceptually opaque actions such as facial expressions as well as to relatively perceptually transparent actions such as hand movements. B. Horizontal Processes Suppose that an observer sees the set of hand movements in Fig. 2 (point, splay, and victory) for the first time, i.e., the sequence is novel. ASL theory suggests that two sets of associative processes, resulting in ‘‘horizontal’’ and ‘‘vertical’’ links, determine whether and to what extent the observer will be able to imitate the sequence. Through the horizontal processes, associations are formed which link sensory, in this case visual, representations of the action units in the sequence (SENS1...n ). These visual representations may be associated with one another in a chain such that, for example, activation of SENS1 activates SENS2 directly, but studies of list learning in humans have shown that such chaining models seldom apply (Henson et al., 1996). It is more likely that the horizontal processes conform to a context-based model (Brown, 1997) in which sensory representations of successive action units become associated with successive states of a time-varying context or control signal, such as the output of an internal clock. This distinction between chaining and context-based horizontal association is not crucial for the present purposes, but the lines connecting sensory representations in Fig. 2 are dashed and curved to signify that context-based association is more likely to occur. Through the horizontal processes, the observer could be said to learn what the action sequence ‘‘looks like’’; the O learns a stimulus sequence.

224

CECILIA M. HEYES AND ELIZABETH D. RAY

For example, if the observer were given a jumbled set of cards, each showing an action component (Carroll and Bandura, 1982), he or she would be able to pick out the components that appeared in the sequence and put those cards in the appropriate order. However, the horizontal processes are not unique to imitation (formation of visual representations, and their association with one another, proceeds as in other cases in which the individual learns a sequence of visual stimuli), and they are not sufficient to enable the observer to reproduce the novel action sequence. For imitation of the observed action sequence, the vertical processes are also necessary. C. Vertical Processes The vertical processes operate before the novel sequence is observed and result in a sensory representation of each action component (e.g., SENS1 ) becoming associated with a motor representation of the same component (e.g., MOTOR1 ). ASL theory makes minimal assumptions about the content of motor representations. They may encode kinesthetic feedback from performance of the action unit and/or a motor program. What is important is that ASL assumes that a motor representation of an action unit can be formed only through performance of that unit, and therefore that the functioning of the vertical processes is such that the accuracy of imitative performance is directly related to the proportion of the components in an action sequence executed prior to sequence observation. Vertical processes can result in sensory and motor representations becoming associated directly or indirectly. Direct associations are formed when an action unit is contiguously observed and executed (seen and done). There are three major sources of such experience: self-observation, mirrors, and synchronous action. Self-observation provides the kind of contiguous experience that will support imitation only for relatively perceptually transparent actions. For example, when an observer looks at her hand while moving her fingers, she receives contiguous experience of seeing and doing the finger movements. Mirrors and synchronous action (i.e., performing the same action at the same time as another individual) provide contiguous experience of observing and executing perceptually opaque and perceptually transparent actions. Behavioral synchrony may result from imitation of the observer by the model or simultaneous responding to a common environmental stimulus. Indirect links between sensory and motor representations of the same action unit are formed when a second stimulus, distinct from sensory input arising from observation of the action unit, is consistently paired on some occasions with sensory input from the action unit and on other occasions

WHAT IS THE SIGNIFICANCE OF IMITATION IN ANIMALS?

225

with performance of that unit. This is the ‘‘acquired equivalence’’ route (Hall, 1996) to imitation, and in humans the second stimulus is usually a word or phrase (VERB1...n ). D. Summary of ASL Theory ASL theory suggests that imitation consists of the reproduction of a succession of action units, the sequence of which is novel—i.e., that prior to imitative performance, O has not executed the action units in the modeled order. It postulates that two sets of associative, contiguity-based processes are necessary for imitation. The horizontal processes operate during observation of the novel action sequence and do not require any overt action on the part of the observer. Hence, this is an associative but not a behaviorist theory. The horizontal processes (which also operate when information is acquired about the serial order of nonaction stimuli) mediate observational (i.e., behaviorally silent; Dickinson, 1980) learning of what the sequence looks like, but are not sufficient to support imitation. Reproduction of the action sequence will be possible to the extent that sensory representations of the sequence components have become associated, via vertical processes, with motor representations of the same components. Associations of this kind are formed when, in the course of self-observation, mirror exposure, and/or synchronous action, the observer contiguously observes and executes an action unit, or they are formed through acquired equivalence training (i.e., experience in which observation and execution of an action unit have each been paired with a common stimulus such as a word or phrase). To the extent that such vertical links have been formed, exposure to the novel action sequence, or recollection of that sequence mediated by the horizontal processes, will activate motor representations in the order appropriate for sequence reproduction (i.e., imitation). This activation gives the learner the potential to imitate the observed action sequence (represented by the bottom row of icons in Fig. 2)—the information necessary to reproduce the action sequence. E. Learning and Performance It is important to note that ASL theory is a theory of learning and not of performance. It specifies inputs and processes which result in an observer being able to imitate a novel sequence of action units, but just because the observer can imitate does not necessarily mean that he or she will imitate. Performance will be governed by additional motivational processes. However, just as there is no reason to suppose that distinctive perceptual and attentional mechanisms operate on sensory input from body movement

226

CECILIA M. HEYES AND ELIZABETH D. RAY

stimuli, there is also no reason to assume that imitative performance is regulated by special motivational processes. Some of the nonimitative behavior of people and animals is reflexive or habitual, whereas other nonimitative behavior is goal directed (Dickinson, 1994), and we would expect imitative behavior to represent a similar combination. Thus, in some species and circumstances, if the processes specified by ASL have occurred, exposure to a novel action sequence may be sufficient to elicit imitation of that sequence. In other taxa and conditions, imitative performance (but not imitative learning) may depend on a representation of the consequences of the observed action or even on a ‘‘metarepresentation’’ (Whiten and Byrne, 1991), i.e., a representation of the demonstrator’s representation of the consequences of the action. Thus, ASL theory is not necessarily inconsistent with research emphasizing the goal directedness or intentionality of imitation (Tomasello and Call, 1997; Tomasello, 1999; Whiten and Byrne, 1991). This research is concerned primarily with the motivation to imitate and not, as is ASL theory, with the origins of the information necessary to achieve behavior matching. F. Redefining Imitation Science organizes phenomena, things to be explained, into categories not merely on the basis of their manifest properties but also in accordance with hypotheses about underlying causal mechanisms. If definitions are to play a useful role in research, they must be to some degree theory driven. Therefore, it is not surprising that ASL theory redefines imitation, suggesting that the proper reference of the term is slightly different from that previously assumed. ASL theory implies that ‘‘imitation’’ refers to instances of social learning in which observation of D performing a sequence of action units enables O to execute the same action units in the same order, where the sequence is novel, i.e., the units have not been executed in the same order by O in the past. Four features of this characterization of imitation are noteworthy. First, it is a refinement of the definition of imitation given at the beginning of this article as response learning by observation. The ASL definition specifies in addition what is learned by observation about the response, i.e., the order in which action units are to be executed. It also makes clear that imitation is unlikely to be the only kind of response learning by observation. The same horizontal and vertical processes that yield matching behavior when O has had correlated experience of seeing and doing the same action units could yield systematically nonmatching behavior when O has had correlated experience of seeing and doing different action units. For example, when O regularly steps forward with the right foot, while a social

WHAT IS THE SIGNIFICANCE OF IMITATION IN ANIMALS?

227

interactant steps backward with the left. There is no convenient term for systematically nonmatching behavior of this kind, but that does not necessarily mean that it is infrequent or of little functional significance relative to imitation. Its relative obscurity may be due instead to its being more difficult to detect than imitation. Second, the ASL definition is compatible with two time-honored definitions of imitation: Thorndike’s (1898) characterization of imitation as learning ‘‘to do an act from seeing it done’’ and Thorpe’s (1956) definition of ‘‘true imitation’’ as ‘‘the copying of a novel or otherwise improbable act or utterance.’’ Rather than contradicting these definitions, the ASL view merely adds specification of the dimension of novelty—of the content of what is learned about the act from seeing it done (i.e., the sequence of the action units). However, ASL theory is inconsistent with a common interpretation of Thorpe’s definition, which states that imitation refers to cases in which a completely novel action (whatever that might mean) is acquired through observation. ASL theory implies that this interpretation circumscribes an empty set of behavioral phenomena; that actions the components of which have not previously been executed, and executed during contiguous observation of the same act or in the context of acquired equivalence training, cannot be reproduced on the basis of information acquired through observation. Third, like Thorpe (1956), ASL theory distinguishes imitation from ‘‘social facilitation.’’ Imitation occurs when an observer reproduces a novel sequence of action units as a result of learning vertical associations between sensory and motor representations of the units and horizontal associations between sensory representations of successive units. By contrast, in cases of social facilitation, the sequence is not novel (e.g., the observer has previously observed and executed the target sequence as a whole), or experience does not play a significant role in formation of the horizontal and vertical associations mediating response reproduction. In the latter case, the associations are ‘‘hardwired’’ or, as Meltzoff (1990) stated, response reproduction is mediated by an ‘‘innate releasing mechanism.’’ Finally, ASL theory does not support a distinction between imitation and ‘‘reflexive’’ reproduction of a novel action sequence. The terms ‘‘mimicry’’ (Aronfreed, 1969) and ‘‘response facilitation’’ (Byrne and Tomasello, 1995) are sometimes used to refer to hypothetical cases of reflexive reproduction of a novel action sequence and to distinguish these from imitation, which is then defined as a goal-directed phenomenon. Because ASL theory focuses on learning rather than performance (see Section IV,E) it elides this distinction, emphasizing instead that, whether or not it is goal directed, imitation occurs through complex cognitive processes. They are complex by virtue of involving horizontal and vertical associations, in addition to the myriad

228

CECILIA M. HEYES AND ELIZABETH D. RAY

perceptual and motor functions involved in parsing and producing action sequences, and cognitive in the sense of being ‘‘behaviorally silent’’ (Dickinson, 1980)—not directly observable in behavior.

V. Predictions and Theory Testing Even in its current rudimentary form, ASL theory makes testable predictions. The two principal predictions are that imitation, reproduction of a novel sequence of actions, will be possible to the extent that the observer (i) is capable of stimulus sequence learning (i.e., learning what an action sequence looks like) and (ii) has contiguous experience of observation and execution of components of the action sequence or experience in which observation and execution have each been paired with a common (e.g., verbal) stimulus. If these predictions are inconsistent with the evidence, it suggests that ASL theory is wrong, and that different associative processes, or transformational processes, mediate imitation. If they are fulfilled, it would not necessarily indicate that ASL theory is correct, but it would strengthen its position relative to other untested and untestable theories. The second, and more distinctive, of ASL theory’s two principal predictions can be tested more readily through research on animals than in human experiments. This is because empirical evaluation of ASL theory requires experimental control of participants’ previous experience of correlated observation and execution of units in the to-be-imitated action sequence, and this is very difficult to achieve for human subjects. For example, Ishikura and Inomata (1995) provided one of many demonstrations that humans are capable of response learning by observation in an experiment in which adult subjects were instructed to reproduce a sequence of seven balletic poses, demonstrated by a trained dancer. The Os were successful in carrying out these instructions and, since there was a close topographic match between the Os’ and the Ds’ behavior and the actions were not directed toward an environmental object, unlike most putative imitation in animals, this success could not have been due to the Os having learned by observation something about the static or dynamic properties of an environmental object. Thus, Ishikura and Inomata’s participants provided evidence of response learning by observation but, since it would be practically impossible to assess the extent of their past experience of seeing-and-doing components of the sequence, their experiment does not provide a basis for testing ASL theory. Training experiments, in which Os are given varying degrees of experience of seeing-and-doing action components before an imitation test, are likely to provide the most effective means of testing ASL theory, but these

WHAT IS THE SIGNIFICANCE OF IMITATION IN ANIMALS?

229

would be laborious with human participants. The training phase would need to be long and/or intensive enough to prevent masking of its effects by prior, spontaneous experience of a similar kind. This problem can be minimized in research with many animal species by witholding until the experiment begins the opportunity to see, do, and see-and-do components of the to-be-imitated action. Training studies of two types could be used to test ASL theory. In the first type, the target, to-be-imitated action would be ‘‘sequentially novel’’ for all subjects (i.e., it would consist of action units in a sequence that had not previously been observed or executed by any subject), whereas the extent to which the target action is ‘‘combinatorial novel’’ would vary across groups. The combinatorial novelty of an action for an O is inversely related to the proportion of the action’s components which the O has contiguously observed and executed (or which have acquired equivalence through pairing with a third stimulus). If the proportion is low, the action has high combinatorial novelty for the O, and if the proportion is high it has low combinatorial novelty for O. For example, some observers would have prior experience of contiguously seeing and doing each of the components in the test sequence (low combinatorial novelty), others would have this experience for some but not all units (medium), and others would have no such experience (high). In this type of experiment, ASL theory would predict better imitative performance in the low group than in the medium group and the worst performance in the high group. The second type of training experiment would vary not the degree but rather the kind of pretest experience of correlated observation and execution of action units. Thus, before observing a set of units in a novel sequence, AABBAA, some subjects would have correlated experience of seeing A and doing A, seeing B and doing B (group AA/BB), whereas others would see A while doing B and vice versa (group AB/BA). A control group would see and do A and B as often as the other two groups, but observation and execution would be uncorrelated. ASL theory would predict that, after observation of the target sequence, group AA/BB would execute a sequence more like AABBAA than that of controls, whereas group AB/BA would perform a sequence more like BBAABB than that of controls. To our knowledge, neither of these two types of experiment have been conducted with any species, and therefore the predictions of ASL theory have not been tested directly. However, within the existing literature on imitation in animals there may be evidence that is inconsistent with ASL and/or experimental paradigms (i.e., species and method combinations) that could be used in the future to test ASL theory directly. Evidence inconsistent with ASL would suggest that animals are capable of learning to perform a relatively perceptually opaque action by observation of that

230

CECILIA M. HEYES AND ELIZABETH D. RAY

action (i.e., response learning by observation) without prior experience of contiguously observing and executing the whole action or a significant proportion of its components (i.e., when the action is combinatorially novel). For future use in testing ASL theory against alternative models, an experimental paradigm needs to be reliable and accessible; it should reliably yield evidence of response learning by observation and be accessible in the light of ethical and cost considerations. In the next section we seek evidence of this kind: studies of imitation in animals which demonstrate response learning by observation when the test behavior is combinatorially novel or the procedure is reliable and accessible.

VI. Survey of Two-Action Tests In approximately the past 10 years, the use of two-action tests has substantially increased the rigor of research investigating imitation in animals. These tests are designed to isolate response learning by observation (imitation) from stimulus learning by observation, and they typically begin with observers being exposed to a demonstrator operating on a single object in one of two different ways. After this observation experience, each subject is given access to the object, and a record is made of the number of times he or she responds to the object using the same action as the demonstrator and using the alternative action, the one that he or she did not observe. A bias in favor of the former, of demonstrator-consistent responding, is prima facie evidence of imitation. The current survey is confined to experiments using the two-action method because compelling evidence of the kind we are seeking (see Section V) is most likely to come from these. For convenience, studies of nonhuman primates (henceforward ‘‘primates’’), rodents, and birds are considered separately. This categorization does not imply that there is sufficient evidence even to speculate about the phylogenetic distribution of the capacity to imitate. A. Primates Whiten and colleagues conducted a series of two-action tests with chimpanzees (Pan troglodytes) (Whiten et al., 1996; Whiten, 1998) and capuchin monkeys (Cebus apella) (Custance et al., 1999) using a puzzle box or ‘‘artificial fruit.’’ In the most recent of these studies (Whiten, 1998), four chimpanzees observed a human demonstrator removing two pairs of objects from the exterior of a transparent box containing food. The objects in each pair were situated close together and at some distance from the other pair. The

WHAT IS THE SIGNIFICANCE OF IMITATION IN ANIMALS?

231

objects in one pair were bolts and in the other pair were T-bars. Two of the chimpanzees observed bolt removal followed by T-bar removal, whereas for the other two animals the T-bars were removed first. One chimpanzee in each of these conditions saw the bolts twisted and pulled out of position and the T-bars spun or turned, whereas the other animal saw the bolts poked out of their lodgings and the T-bars spun or turned. When given access to the objects, the chimpanzees showed a tendency to approach the objects in the observed sequence (e.g., those that observed bolt removal followed by T-bar removal tended to approach the bolts before the Tbars), but there was no reliable evidence that the chimpanzees preferentially used the observed actions to remove the objects. The first of these findings, that the chimpanzees tended to approach the objects in the observed order, is interesting because it suggests that these apes have the first of the two necessary conditions for imitation specified by ASL theory (i.e., that they can learn a stimulus sequence by observation). On the other hand, the second finding is disappointing because it means that the artificial fruit studies have not (yet) provided compelling evidence of response learning by observation. If it had been possible to test a larger sample of chimpanzees, a significant tendency to use the same action as the demonstrator may have emerged. However, as Whiten and colleagues acknowledge, such an effect, whether it was found in capuchins (Custance et al., 1999) or chimpanzees, could be due to emulation learning (Tomasello et al., 1993b) rather than response learning by observation. In other words, the observers may be learning by observation the affordances or dynamic properties of the objects manipulated rather than the action applied. This is a possibility in the case of the artificial fruit procedure because, in the studies to date, object movement has been confounded with demonstrator movement. For example, when a bolt was twisted out, it rotated in one direction, and when it was poked out, it translated in the opposite direction (relative to compass points, the box, and the observer). Three other ape studies have reported failure to find evidence of imitation using two-action tests. Tomasello, Call, and colleagues found that, in contrast with 2-year-old children, juvenile and adult chimpanzees (Nagell et al., 1993) and orangutans (Pongo pygmaeus) (Call and Tomasello, 1994) showed no tendency to copy the action used by a human demonstrator to rake in food. Similarly, there were no signs of imitation when juvenile and adult orangutans observed a stick, which was protruding from a box, being manipulated in one of four ways by a human demonstrator (Call and Tomasello, 1995; Tomasello, 1996). (Procedures in which different groups of subjects are exposed to two or more distinct actions and assessed in terms of their subsequent performance of these actions are conventionally

232

CECILIA M. HEYES AND ELIZABETH D. RAY

known as two-action tests, regardless of the number of levels of the independent variable.) Finally, two primate studies using procedures similar to those of the twoaction method are worthy of consideration. In the first of these, Bugnyar and Huber (1997) gave marmosets (Callithrix jacchus) a test in which they could obtain food by pulling a door (hinged at the top) toward their bodies or by pushing it away. Prior to this test, one group of subjects observed a demonstrator pulling the door, whereas the other group had no previous exposure to the apparatus. Careful analysis of the test data revealed suggestive, but not conclusive, evidence that some marmosets in the observer group were influenced by the action they observed. However, in comparison with the nonexposed animals, the observer group did not show a reliable bias in favor of pulling the door. In the second study, Myowa (1996) tested a single, infant chimpanzee (5–15 weeks of age) for imitation of human facial gestures in a procedure modeled on that of Meltzoff and Moore (1977). In weekly testing sessions, the chimpanzee was exposed to 15-s periods in which an adult human demonstrated tongue protrusion, lip protrusion, and mouth opening. These observation phases alternated with 20-s test periods in which the model adopted a passive face, and the facial gestures of the chimpanzee were recorded. Myowa reported that, for each of the three facial gestures, the chimpanzee was more likely to exhibit the recently observed gesture than the other two during test periods in Weeks 5–10. These results are difficult to interpret because Myowa’s (1996) study involved a single subject and the reliability of similar effects in human neonates has been questioned. In an extensive review and reanalysis of data on human neonatal imitation, Anisfeld (1991) found reliable evidence of demonstrator-consistent responding only in the case of tongue protrusion and pointed out that this effect could be due not to imitation but to a rebound effect. This hypothesis suggests that the baseline rate of tongue protrusion is depressed during observation periods by attention to the demonstrator and, in a compensatory fashion, increases above baseline in the subsequent test periods. However, even if observation of tongue protrusion causes tongue protrusion in neonates, in this isolated case the link between observation and execution could be innate. Taken at face value, Myowa’s (1996) findings could not be due to rebound effects because during the test periods there was a selective increase in the frequency of the previously demonstrated response. Furthermore, since this selective increase in frequency occurred for each of three actions, it is implausible, although not impossible, that the results were due to the operation innate stimulus-response links. Thus, Myowa’s study is very interesting,

WHAT IS THE SIGNIFICANCE OF IMITATION IN ANIMALS?

233

but it needs to be convincingly replicated before it can support any firm conclusions about imitation in primates. The absence of clear evidence that primates are capable of response learning by observation does not, of course, imply that they are unable to do so. Indeed, it is more likely to reflect their inaccessibility relative to many rodents and birds. The costs of primate research often prohibit careful experimental investigation with large samples, and it is seldom possible to repeat a procedure several times with different groups of subjects of the same taxa and thereby to establish effective parameters and incorporate additional controls. For the same reasons, and because it is almost as difficult to monitor and control the preexperimental experience of nonhuman primates as it is for humans, a primate paradigm is unlikely to provide a suitable basis for analytic experiments (i.e., for research on the psychological mechanisms of imitation in general and investigation of the role of combinatorial novelty in particular). B. Rodents In the first two-action test involving rodents, Collins (1988) allowed male mice (Mus musculus) to observe a female conspecific demonstrator from behind as the latter moved a pendulum door to the left or to the right for food reward. When subsequently given access to the door and rewarded for pushing it in either direction, the observer mice showed a reliable bias in favor of pushing the door in the same direction as did their demonstrator. This demonstrator-consistent response bias could have been due to response learning by observation, but it could also be that the mice learned about the action of the door rather than of the animal operating on the door. Substantiating this stimulus learning interpretation, Denny et al. (1983, 1988) reported that rats (Rattus norvegicus) exposed to a pendulum bar that moved, in the absence of a demonstrator, to the right for food and the left for no food or vice versa subsequently tended to push the bar in the direction that had been followed by reward. Bidirectional control experiments, inspired by Grindley (1932) and conducted by our own group, seemed until recently to control for stimulus learning by observation and thereby to provide stronger evidence of imitation in rats (Heyes and Dawson, 1990; Heyes et al., 1992, 1993, 1994; see Heyes, 1996, for a review). In the basic bidirectional control procedure, the observer confronted the demonstrator as the latter pushed a rigid pendulum or joystick to the ‘‘left’’ or to the ‘‘right’’ for food reward and was subsequently given access to the joystick from the position previously occupied by the demonstrator (Fig. 3). On test, the observers tended to push the joystick in the same direction relative to the actor’s body as had

234

CECILIA M. HEYES AND ELIZABETH D. RAY

Fig. 3. Diagram of the apparatus used in bidirectional control tests for imitation in rats. Reprinted with the permission of the Experimental Psychology Society from Heyes and Dawson, 1990.

their demonstrators, despite the fact that this action resulted in the joystick moving in the opposite direction within the observer’s visual field to that in which it had moved during observation (Heyes and Dawson, 1990). Furthermore, the effect persisted when the joystick was moved between observation and testing such that, on test, it moved in a plane perpendicular to that in which it had moved when acted on by the demonstrator (Heyes et al., 1992). In these circumstances, when the observers moved the joystick in the same direction as the demonstrator, relative to the actor’s body, it moved to a different location in space to that in which it had moved during conspecific observation. However, the results of recent experiments (Mitchell et al., 1998; Gardner, 1997) indicate that scent cues may influence rats’ performance in the bidirectional control procedure and thereby cast doubt on this evidence of imitation. Mitchell et al. (1998) found that when the joystick is rotated 180⬚ within its mounting between observation and test, observers show reliable demonstrator-inconsistent responding (i.e., a systematic tendency to respond in the opposite direction to their demonstrators). A plausible explanation for this is that the demonstrators deposit attractive odor cues on the side of the joystick contralateral to its direction of motion, and that exploration of these cues by the observers promotes a demonstratorconsistent response bias when the joystick remains in the same position on test and a demonstrator-inconsistent bias when it is rotated 180⬚.

WHAT IS THE SIGNIFICANCE OF IMITATION IN ANIMALS?

235

An odor hypothesis of this kind, even one which assumes that demonstrators can deposit aversive and attractive scent cues on the joystick, does not explain all the published results of the bidirectional control procedure (Mitchell et al., 1998). However, unless or until the basic effect is demonstrated with appropriate controls for scent cues (which may involve joystick cleaning or ‘‘box swapping’’; Heyes et al., 1998), the bidirectional control experiments cannot be regarded as providing sound evidence of response learning by observation in rats. As a result, and given the ambiguity of the findings of Collins (1988) and Denny et al. (1983), the rodent literature does not currently include a paradigm with high potential for analytic investigation of the mechanisms of imitation. C. Birds The most promising evidence of imitation in animals presently comes from studies of birds, in particular, budgerigars, grackles, starlings, pigeons, and quail. We outline each of these studies and then consider them as a group in terms of whether they demonstrate response learning by observation of a combinatorially novel action or involve the use of a potentially accessible and reliable paradigm. Dawson and Foss (1965) initiated, and Galef et al. (1986) developed, the use of the two-action method in their studies of budgerigars (Melopsittacus undulatus). In the experiment by Galef et al., budgerigars that had observed a conspecific demonstrator removing the cover from a food dish using its beak or its feet showed a significant bias in favor of using the same appendage (beak or feet) as had their demonstrators. This effect was detected when performance on the first two test trials was combined but not on subsequent test trials. It is unlikely to have resulted from the D attracting the O’s attention to a particular part of the apparatus because the published report suggests that the beak-using and feet-using demonstrators made contact with the same part of the cover. Furthermore, since birds are relatively insensitive to olfactory cues, and the observers and demonstrators worked on physically distinct pieces of apparatus, it is very unlikely that the observers’ behavior was influenced by scent cues on the manipulandum. In principle, the subjects may have learned by observation about the trajectory of the cover (emulation learning) rather than about the actor’s body movement (imitation). This possibility arises because the apparatus used for demonstrations was such that foot operations tended to tip the cover off the food cup, whereas beak operations tended to result in the cover sliding out of position. It is difficult to evaluate because there is no record of whether, on test trials, the observers used their beaks to slide the cover

236

CECILIA M. HEYES AND ELIZABETH D. RAY

and their feet to tip it off the food cup. However, it is unlikely that learning about the dynamic properties of the cover was solely responsible for demonstrator-consistent appendage use because the apparatus used on test trials allowed both trajectories to be achieved using both appendages. Like budgerigars, Carib grackles (Quiscalus lugubris) and European starlings (Sturnus vulgaris) have provided evidence of imitation which is strong relative to that obtained from nonavian species but which could, in principle, have been influenced by emulation learning. Lefebvre et al. (1997) allowed grackles to observe one of two techniques being used to remove a stopper from an inverted tube containing food. One technique, which was demonstrated by conspecifics, involved open-beak probing and pulling movements, whereas the other demonstrated by Zenaida doves (Zenaida aurita) consisted of closed-beak pecking. On first contact with the tube apparatus, more of the observers of the closed-beak demonstration removed the stopper with a closed beak than with an open beak, and observers in this group made more closed-beak pecks than observers of the open-beak demonstration. The starling experiment (Campbell et al., 1998) employed a ‘‘two-object/ two-action’’ test (Ray 1997). Observer birds were exposed to a conspecific demonstrator removing one of two stoppers (red or black) from a box containing food either by pulling the stopper up with a closed beak or by pushing it down into the box with an open beak. Thus, there were four groups of observers: red-up, red-down, black-up, and black-down. To remove the stopper for the first time, and in the course of three subsequent test trials, observers showed reliable tendencies both to remove the same stopper as their demonstrator removed and to do so using the same up/ closed-beak or down/open-beak action. A final group of bird studies, involving pigeons (Columba livia) and Japanese quail (Coturnix japonica), excluded the possibility of emulation learning (Zentall et al., 1996; Kaiser et al., 1997; Akins and Zentall, 1996, 1999). In these experiments, observers were exposed to a conspecific demonstrator pressing a treadle with its beak (group Pecking) or with its feet (group Stepping) 50 times for food reward. The treadle, which was mounted on an operant panel, moved through the same trajectory regardless of whether the action effecting treadle depression was pecking or stepping. Immediately after the demonstration session, observers were given access to the treadle, allowed to press it at least 50 times, and rewarded for each response regardless of whether it consisted of stepping or pecking. In these circumstances, both pigeons (Zentall et al., 1996) and quail (Akins and Zentall, 1996) tended to use the same action as their demonstrators. Among the pigeons, 5 of 10 birds in group Pecking made pecking responses and 5 made stepping responses, whereas none of the 10 birds in group Stepping pecked the lever. In the case of the quail, the frequency of pecking was

WHAT IS THE SIGNIFICANCE OF IMITATION IN ANIMALS?

237

significantly greater than the frequency of stepping in group Pecking, whereas the frequencies of stepping and pecking did not differ in group Stepping. In comparison with previous studies, all these bird experiments provide strong prima facie evidence of imitation. That is, their results suggest that exposure to a demonstrator’s behavior (and not to the object of that behavior) increased the probability that the observer would produce the same behavior (and not that it would simply be more active or direct its behavior to particular objects). The budgerigar, grackle, and starling studies did not completely exclude the possibility of emulation learning, and it would certainly be desirable for future experiments to do so. However, we consider the avian data to be strong for two reasons. First, there is no independent evidence that birds (or members of any other nonhuman species) are capable of the necessary kind of emulation learning (i.e., of learning to discriminate complex, dynamic properties of objects under conditions comparable to those of the budgerigar, grackle, and starling experiments). Second, the evidence that pigeons and quail show demonstrator-consistent responding in the absence of the opportunity for emulation learning supports the hypothesis that this kind of learning is not solely responsible for the effects reported in the other avian species. The avian data are certainly strong enough to make it worthwhile to consider whether the target actions were combinatorially novel. Any post hoc assessment of combinatorial novelty is bound to be speculative, and our speculation is that it was low in these studies—that many of the components of the imitated actions had been contiguously observed and executed by the subjects prior to the experiments. In all cases, the observers had lived in the laboratory, in groups, or with visual access to other birds prior to the experiments. In these circumstances it is likely that they consumed the same foodstuffs at the same time as other birds and therefore that, through synchronous action, they had the opportunity simultaneously to observe and execute various foraging behaviors. It might be objected that at least one of the avian imitation effects, treadle stepping in pigeons and quail, is unlikely to have arisen from previous experience of seeing and doing the target response because stepping is a relatively arbitrary foraging behavior. However, this is not a compelling argument for imitation of combinatorially novel actions for two reasons: (i) The evidence that observation of stepping promotes stepping in pigeons and quail is not conclusive and, even if it were, (ii) if birds have a tendency to peck when they see other birds peck, whether based on an innate link or correlated experience of seeing and doing, this would increase the likelihood that they have correlated experience of observation and execution of other lower frequency foraging behaviors, including stepping.

238

CECILIA M. HEYES AND ELIZABETH D. RAY

Regarding the first point, the results reported by Zentall et al. (1996) and Akins and Zentall (1996) suggest that observation of pecking may have increased the observer birds’ tendency to peck the treadle on test, whereas observation of stepping had no effect on test performance, and two subsequent experiments have not ruled out this possibility. Kaiser et al. (1997) reported that control pigeons that had no pretest exposure to a conspecific pressing the treadle exhibited low rates of stepping relative to that observed in group Stepping by Zentall et al. (1996). This cross-experimental comparison may indicate that observation of stepping promotes stepping on test, but it is also possible that the difference between group Stepping and the control birds was due to the demonstrators in group Stepping drawing their observers’ attention to the treadle. Since the control birds did not have their attention drawn to the treadle by a conspecific, they directed little behavior of any kind to the manipulandum on test. The second follow-up experiment (Akins and Zentall, 1999) measured the proportions of pecking and stepping responses made by quail on test and showed that the proportion of stepping responses by observers of stepping was numerically, but not statistically, higher than that of control birds with no pretest exposure to a conspecific pressing the treadle. Thus, further research is necessary to establish whether pigeons and quail show effects of observing stepping and pecking, but, regarding the second reason discussed previously, even conclusive evidence of this kind would not indicate that birds can imitate combinatorially novel actions. It is likely that the sight of pecking elicits pecking in birds by virtue of an innate link between observation and execution of this response or, as argued previously, that pecking is imitated on the basis of past experience of seeingand-doing components of the action or the action in its entirety. In either case, innate link or correlated experience, a tendency to peck when others peck would increase the probability that when two or more birds feed in close proximity on common foodstuffs they will simultaneous observe and execute other foraging behaviors because this tendency reduces the range of behavioral variants from which each bird is independently selecting its responses at any given time. It is important to determine whether the members of a particular species exhibit matching behavior in response to more than one action, and the work of Zentall and colleagues is admirable in that it has come closer than any other two-action test experiments to demonstrating effects of both target actions. However, the significance of matching two or more behaviors lies in the fact that it reduces the plausibility that each example of matching is due to an innate stimulus–response link. The fact that more than one behavior can be matched does not reduce the probability that all of the matched behaviors have low combinatorial novelty.

WHAT IS THE SIGNIFICANCE OF IMITATION IN ANIMALS?

239

Perhaps the clearest point to emerge from our speculation about combinatorial novelty is that the only truly effective way of assessing its role in imitation is through experiments explicitly designed to do so, and in particular by varying pretest experience of seeing-and-doing components of the target action (see Section V). In our view, avian two-action tests have significant potential to provide a basis for analytic experiments of this kind. Budgerigars, starlings, pigeons, and quail (although not grackles) are all relatively easy to acquire and maintain as laboratory animals, and therefore it is possible to test a good-sized sample in each experiment and to repeat experiments (in the same and different laboratories) to adjust parameters and check the reliability of findings. Research to date suggests that, of the studies involving accessible species, the two-object/two-action test with starlings and the treadle test with quail are likely to yield the most reliable effects. In their budgerigar study, Galef et al. (1986) reported a transitory effect, confined to the first couple of test responses, and noted that they had been unable to find a stronger or more durable effect in the course of 2 years of parameter variation. Similarly, Akins and Zentall (1996) switched from pigeons to quail because they found that a substantial proportion of observer pigeons either failed to respond on test or exhibited demonstrator-inconsistent behavior. In contrast, the effects found in starlings and quail have persisted throughout the test session, and each test has been replicated (A. Goldsmith et al., unpublished results; Akins and Zentall, 1999). The strength of the treadle test is that it controls for emulation learning, whereas the advantage of the two-object/two-action test is that it allows simultaneous investigation of stimulus learning and response learning by observation and thereby provides a method of comparing the conditions favoring nonimitative and imitative social learning (Ray, 1997). The treadle test could be adapted to the same purpose simply by adding a second treadle to the apparatus and allowing quail to observe pecking or stepping of the left or the right treadle. In a complementary way, a ‘‘ghost control’’ condition or transfer test could be added to the two-object/two-action test to examine the role of emulation learning (Heyes et al., 1992, 1994). Ghost control starlings might be exposed, in the presence of a feeding conspecific, to the stopper rising up out of the food box or moving down into the box automatically (e.g., through the operation of fine wires). In a transfer test, starlings that had observed up/open-beak or down/closed-beak demonstrations could be tested using a manipulandum that does not move in the vertical plane. In conclusion, this brief review suggests that (i) recent use of two-action tests has brought a new vigor and incisiveness to research on imitation in animals, (ii) the current evidence is consistent with the hypothesis that

240

CECILIA M. HEYES AND ELIZABETH D. RAY

imitation depends on previous experience of contiguously seeing-and-doing components of the to-be-imitated action sequence, and (iii) several avian paradigms are likely to be sufficiently reliable and accessible to support analytic experiments investigating the mechanisms of imitation learning.

VII. Postscript: Imitation and Culture It has often been argued that imitation in animals is important with respect to our understanding of both cognition and culture; that in addition to involving complex psychological mechanisms, imitation plays a key functional role in the transmission and accumulation of information across individuals and generations (Tomasello et al., 1993a; Tomasello and Call, 1997; Tomasello, 1999; Boyd and Richerson, 1985; Richerson and Boyd, 1999). In this article, we focus on the cognitive significance of imitation, but this is not to deny that there is any kind of special relationship between imitation and culture. The most thorough contemporary analyses of the link between imitation and culture state or imply that (i) the psychological mechanisms of imitation are distinct from those of other forms of social and individual learning, in terms of both their ‘‘rules of operation’’ (Sherry and Schacter, 1987) and their evolutionary origins, and (ii) these mechanisms mediate information transmission with sufficient fidelity to support cultural evolution, a process analogous to natural selection which promotes behavioral adaptation. In contrast with the first of these hypotheses, ASL theory suggests that associative mechanisms of common phylogenetic origin underlie individual learning, nonimitative social learning, and imitation [see Lefebvre (1999) for data consistent with this hypothesis]. However, ASL theory also suggests that the capacity to imitate represents ontogenetic specialization of these general processes; that when, in the course of ontogeny, inputs to the general processes include contiguous observation and execution of action units (and/or acquired equivalence training), they yield an ‘‘imitation repertoire’’—a set of action units that can be imitated when observed in novel sequences. ASL theory therefore implies that, to the extent that experience of seeing-and-doing action units derives from exposure to mirrors and to imitation of the observer’s behavior by others, culture supports imitation (Heyes, 2000). Of course, in addition to being supported by culture, imitation may play a special role in promoting cultural evolution. However, as argued by Heyes (1993), and in contrast with the second implication discussed previously, it would seem that fidelity of information transmission requires processes supporting faithful acquisition of information by one individual, O, from

WHAT IS THE SIGNIFICANCE OF IMITATION IN ANIMALS?

241

another, D, and faithful retention processes ensuring that O does not lose or change the information before it is retransmitted to a third party. For the information contained in certain kinds of behavior, imitation may be an unusually effective acquisition process (compared with nonimitative social learning or language), but additional psychological processes promoting conformity to group norms (Wilson, 2000) are likely to be necessary to achieve the transmission fidelity required for cultural evolution.

VIII. Summary Actions vary on the dimension of perceptual opacity. Highly perceptually opaque actions, such as facial expressions, give rise to dissimilar sensory inputs when observed and executed, whereas highly perceptually transparent actions, such as vocalizations and distal appendage movements, yield similar sensory inputs when observed and executed. The most significant challenge for any theory of the psychological mechanisms of imitation learning is to explain imitation of perceptually opaque actions. The theories that have addressed this problem in the past century are of two kinds: Transformational theories suggest that most of the information necessary to achieve a behavioral match is generated internally by complex cognitive processes, whereas associative theories claim that this information is derived principally from experience. These theories delineate plausible alternative accounts of the psychological mechanisms of imitation, but they do not provide a satisfactory framework for empirical inquiry because each theory either does not make testable predictions or is inconsistent with what is already known about the conditions of imitation. The ASL theory suggests that imitation is mediated by associative processes which form links between sensory representations of successive components in an observed action sequence (horizontal processes) and between sensory and motor representations of individual action components (vertical processes). It predicts that reproduction of a novel sequence of action units is possible to the extent that the subject (i) can learn a stimulus sequence by observation and (ii) has prior experience of contiguously observing and executing components of the novel sequence and/or acquired equivalence training for those components (i.e., when the sequence has low combinatorial novelty). The latter prediction can be tested more readily with animal than with human subjects, and a survey of research using twoaction tests of imitation suggests that several paradigms involving avian subjects are sufficiently reliable and accessible to support analytic experiments of the relevant kind.

242

CECILIA M. HEYES AND ELIZABETH D. RAY

Thus, the principal significance of imitation in animals lies in its potential to provide insight into the psychological mechanisms of imitation learning, and recent methodological innovations have brought it to the brink of realizing this potential. The most pressing current requirement is to formulate both transformational and associative theories of imitation which make empirically testable predictions. Acknowledgments This research was supported by the Biotechnology and Biological Sciences Research Council and by the Economic and Social Research Council funded Research Centre for Economic Learning and Social Evolution. We are grateful to Martin Eimer, Bennett Galef, and the editors for their comments on an earlier draft of this manuscript, to Anthony Dickinson, Dorothy Einon, David Shanks, and Phil Reed for their contributions in discussion, and to Veli Avdji for his assistance with Fig. 2.

References Akins, C. K., and Zentall, T. R. (1996). Imitative learning in male Japanese quail using the two-action method. J. Comp. Psychol. 110, 316–320. Akins, C. K., and Zentall, T. R. (1999). Imitation in Japanese quail: The role of demonstrator reinforcement. In press. Allport, F. H. (1924). ‘‘Social Psychology.’’ Riverside, Cambridge, MA. Anisfeld, M. (1991). Neonatal imitation. Dev. Rev. 11, 60–97. Aronfreed, J. (1969). The problem of imitation. Adv. Child Dev. Behav. 4, 209–319. Bandura, A. (1986). ‘‘Social Foundations of Thought and Action: A Social Cognitive Theory.’’ Prentice Hall, Englewood Cliffs, NJ. Boyd, R., and Richerson, P. J. (1985). ‘‘Culture and the Evolutionary Process.’’ Chicago Univ. Press, Chicago. Brown, G. D. A. (1997). Formal models of memory for serial order. In ‘‘Cognitive Models of Memory’’ (M. A. Conway, Ed.). Psychology Press. Bugnyar, T., and Huber, L. (1997). Push or pull: An experimental study of imitation in marmosets. Anim. Behav. 54, 817–831. Byrne, R. W., and Tomasello, M. (1995). Do rats ape? Anim. Behav. 50, 1417–1420. Call, J., and Tomasello, M. (1994). The social learning of tool use by orangutans. Hum. Evol. 9, 297–313. Call, J., and Tomasello, M. (1995). The use of social information in the problem-solving of orangutans and human children. J. Comp. Psychol. 109, 301–320. Campbell, F. M., Heyes, C. M., and Goldsmith, A. (1998). A demonstration of simultaneous stimulus learning and response learning by observation in the European starling using a two-object/two-action method. Anim. Behav. 58, 151–158. Carroll, W. R., and Bandura, A. (1982). The role of visual monitoring in observational learning of action patterns: Making the unobservable observable. J. Motor Behav. 14, 153–167. Collins, R. L. (1988). Observational learning of a left–right behavioral asymmetry in mice. J. Comp. Psychol. 102, 222–224. Custance, D., Whiten, A., and Fredman, T. (1999). Social learning of artificial fruit processing in enculturated capuchin monkeys. J. Comp. Psychol. 113, 13–23.

WHAT IS THE SIGNIFICANCE OF IMITATION IN ANIMALS?

243

Dawson, B. V., and Foss, B. M. (1965). Observational learning in budgerigars. Anim. Behav. 13, 470–474. Denny, M. R., Bell, R. C., and Clos, C. F. (1983). Two-choice, observational learning and reversal in the rat: S–S versus S–R effects. Anim. Learning Behav. 11, 223–228. Denny, M. R., Clos, C. F., and Bell, R. C. (1988). Learning in the rat of a choice response by observation of S–S contingencies. In ‘‘Social Learning: Psychological and Biological Perspectives’’ (T. R. Zentall and B. G. Galef, Eds.). Erlbaum, Hillsdale, NJ. Dickinson, A. (1980). ‘‘Contemporary Animal Learning Theory.’’ Cambridge Univ. Press, Cambridge, UK. Dickinson, A. (1994). Instrumental conditioning. In ‘‘Handbook of Perception and Cognition: Animal Learning and Cognition (N. J. Mackintosh, Ed.). Academic Press, San Diego. Galef, B. G. (1988). Imitation in animals: History, definition and interpretation of data from the psychological laboratory. In ‘‘Social Learning: Psychological and Biological Perspectives (T. R. Zentall and B. G. Galef, Eds.). Erlbaum, Hillsdale, NJ. Galef, B. G., Manzig, L. A., and Field, R. M. (1986). Imitation learning in budgerigars: Dawson and Foss (1965) revisited. Behav. Processes 13, 191–202. Gardner, M. R. (1997). Imitation: The methodological adequacy of directional control tests. Unpublished doctoral thesis, University of London, London. Gewirtz, J. L. (1971). Conditional responding as a paradigm for observational, imitative learning and vicarious-reinforcement. Adv. Child Dev. Behav. 6, 273–304. Gewirtz, J. L., and Stingle, K. G. (1968). Learning of generalized imitation as the basis for identification. Psychol. Rev. 75, 374–397. Gordon, H. (1922/1923). Hand and ear tests. Br. J. Psychol. 13, 283–300. Grindley, G. C. (1932). The formation of a simple habit in guinea pigs. Br. J. Psychol. 23, 127–147. Hall, G. (1996). Learning about associatively activated stimulus representations: Implications for acquired equivalence and perceptual learning. Anim. Learning Behav. 24, 233–255. Henson, R. N. A., Norris, D. G., Page, M. P. A., and Baddeley, A. D. (1996). Unchained memory: Error patterns rule out chaining models of immediate serial recall. Q. J. Exp. Psychol. 19, 80–115. Heyes, C. M. (1993). Imitation, culture and cognition. Anim. Behav. 46, 999–1010. Heyes, C. M. (1994). Social learning in animals: Categories and mechanisms. Biol. Rev. 69, 207–231. Heyes, C. M. (1996). Genuine imitation? In ‘‘Social Learning in Animals: The Roots of Culture’’ (C. M. Heyes and B. G. Galef, Eds.). Academic Press, San Diego. Heyes, C. M. (1999). Questions about the evolution of cognition. In ‘‘Evolution of Cognition’’ (C. M. Heyes and L. Huber, Eds.). MIT Press, Cambridge. Heyes, C. M., and Dawson, G. R. (1990). A demonstration of observational learning using a bidirectional control. Q. J. Exp. Psychol. 42B, 59–71. Heyes, C. M., Dawson, G. R., and Nokes, T. (1992). Imitation in rats: Initial responding and transfer evidence. Q. J. Exp. Psychol. 45B, 81–92. Heyes, C. M., Jaldow, E., and Dawson, G. R. (1993). Observational extinction: Observation of nonreinforced responding reduces resistance to extinction in rats. Anim. Learning Behav. 21, 221–225. Heyes, C. M., Jaldow, E., Nokes, T., and Dawson, G. R. (1994). Imitation in rats: The role of demonstrator action. Behav. Processes 32, 173–182. Heyes, C. M., Ray, E. D., Mitchell, C., and Nokes, T. (1998). Stimulus enhancement: Controls for social facilitation and local enhancement. Submitted for publication. Heyes, C. M., Ray, E. D., and McCarthy, J. (1999). Effects of viewing angle on mirror and transposition imitation of body movement sequences. Manuscript in preparation.

244

CECILIA M. HEYES AND ELIZABETH D. RAY

Holt, E. B. (1931). ‘‘Animal Drive and the Learning Process. Holt, New York. Ishikura, T., and Inomata, K. (1995). Effects of angle of model-demonstration on learning of motor skill. Perceptual Motor Skills 80, 651–658. Kaiser, D. H., Zentall, T. R., and Galef, B. G. (1997). Can imitation in pigeons be explained by local enhancement together with trial-and-error learning? Psychol. Sci. 8, 459–460. Lefebvre, L. (1999). Feeding innovations and their cultural transmission in bird populations. In ‘‘Evolution of Cognition’’ (C. M. Heyes and L. Huber, Eds.). MIT Press, Cambridge. Lefebvre, L., Templeton, J., Brown, K., and Koelle, M. (1997). Carib grackles imitate conspecific and zenaida dove tutors. Behaviour 134, 1003–1017. Meltzoff, A. N. (1985). The roots of social and cognitive development: Models of man’s original nature. In ‘‘Social Perception in Infants’’ (T. M. Field and N. Fox, Eds.), pp. 1–30. Ablex, Norwood, NJ. Meltzoff, A. N. (1990). Towards a developmental cognitive science: The implications of crossmodal matching and imitation for the development of representation and memory in infancy. Ann. N. Y. Acad. Sci. 608, 1–37. Meltzoff, A. N., and Moore, M. K. (1977). Imitation of facial and manual gestures by human neonates. Science 198, 75–78. Miller, N. E., and Dollard, J. (1941). ‘‘Social Learning and Imitation.’’ Yale Univ. Press, New Haven, CT. Mitchell, C. J., Heyes, C. M., Dawson, G. R., and Gardner, M. R. (1998). Limitations of a bidirectional control procedure for the investigation of imitation in rats: Odour cues on the manipulandum. Q. J. Exp. Psychol. 52B, 193–202. Mowrer, O. H. (1960). ‘‘Learning Theory and the Symbolic Processes.’’ Wiley, New York. Myowa, M. (1996). Imitation of facial gestures by an infant chimpanzee. Primates 37, 207–213. Nagell, K., Olguin, R., and Tomasello, M. (1993). Processes of social learning in the imitative learning of chimpanzees and human children. J. Comp. Psychol. 107, 174–186. Piaget, J. (1951). ‘‘Play, Dreams and Imitation in Childhood.’’ Norton, New York. Ray, E. D. (1997). Social and associative learning. Unpublished doctoral thesis, University of London, London. Richerson, P. J., and Boyd, R. (1999). Climate, culture and the evolution of cognition. In ‘‘Evolution of Cognition’’ (C. M. Heyes and L. Huber, Eds.). MIT Press, Cambridge. Sherry, D. F., and Schacter, D. L. (1987). The evolution of multiple memory systems. Psychol. Rev. 94, 439–454. Skinner, B. F. (1953). ‘‘Science and Human Behavior.’’ Free Press, New York. Thorndike, E. L. (1898). Animal intelligence. Psychol. Rev. Monogr. 2(8). Thorpe, W. H. (1956). ‘‘Learning and Instinct in Animals.’’ Methuen, London. Tomasello, M. (1996). Do apes ape? In ‘‘Social Learning in Animals: The Roots of Culture’’ (C. M. Heyes and B. G. Galef, Eds.). Academic Press, San Diego. Tomasello, M. (1999). Two hypotheses about primate cognition. In ‘‘Evolution of Cognition’’ (C. M. Heyes and L. Huber, Eds.). MIT Press, Cambridge. Tomasello, M., and Call, J. (1997). ‘‘Primate Cognition.’’ Oxford Univ. Press, Oxford. Tomasello, M., Kruger, A. C., and Ratner, H. H. (1993a). Cultural learning. Behav. Brain Sci. 16, 495–592. Tomasello, M., Savage-Rumbaugh, S., and Kruger, A. C. (1993b). Imitative learning of actions on objects by children, chimpanzees, and enculturated chimpanzees. Child Dev. 64, 1688– 1705. Wapner, S., and Cirillo, L. (1968). Imitation of a model’s hand movements: Age changes in transposition of left–right relations. Child Dev. 39, 887–894. Whiten, A. (1998). Imitation of the sequential structure of actions by chimpanzees. J. Comp. Psychol. 112, 270–281.

WHAT IS THE SIGNIFICANCE OF IMITATION IN ANIMALS?

245

Whiten, A., and Byrne, R. W. (1991). The emergence of metarepresentation in human ontogeny and primate phylogeny. In ‘‘Natural Theories of Mind’’ (A. Whiten and R. W. Byrne, Eds.). Basil Blackwell, Oxford. Whiten, A., and Ham, R. (1992). On the nature and evolution of imitation in the animal kingdom: Reappraisal of a century of research. Adv. Study Behav. 21, 239–283. Whiten, A., Custance, D. M., Gomez, J.-C., Teixidor, P., and Bard, K. A. (1996). Imitative learning of artificial fruit processing in children and chimpanzees. J. Comp. Psychol. 110, 3–14. Wilson, D. S. (1999). Gossip and other forms of talk as multilevel adaptations. In ‘‘Evolution of Cognition’’ (C. M. Heyes and L. Huber, Eds.). MIT Press, Cambridge. Zentall, T. R., Sutton, J. E., and Sherburne, L. M. (1996). True imitative learning in pigeons. Psychol. Sci. 7, 343–346.

This Page Intentionally Left Blank

ADVANCES IN THE STUDY OF BEHAVIOR, VOL. 29

Vocal Interactions in Birds: The Use of Song as a Model in Communication Dietmar Todt and Marc Naguib institut fu¨r verhaltensbiologie freie universita¨t berlin 12163 berlin, germany

I. Introduction Since the publications of Darwin (1859) and Wallace (1889) there has been continuing interest in the communication of animals. Any understanding of the behavior of organisms requires information on when, how, what, and why individuals communicate. Communication is central to a wide range of questions about social behavior, including mate attraction, mating behavior, spacing between competitors, and defense of resources. Most interest in communication has been centered around those signals that have evolved as conspicuous and elaborate displays, although soft and subtle vocalizations deserve to be studied as well. Among conspicuous displays, birdsong is an outstanding system that has proven to be an excellent model with which to address the classical questions raised in animal behavior. Within this framework, we focus on the use of vocal signals during interactions with competitors or mates in passerine birds. We will describe the different modes of interactions and evaluate how our understanding attempts to answer proximate and functional questions on animal signaling. Interactions among individuals as a specific case in communication are among the most important social behaviors and provide a particularly fascinating aspect of communication in particular. In fact, interactions provide the clearest evidence that individuals are communicating. Participating animals often respond to each others’ signals in very specific ways and each of the interacting individuals usually takes the role of both a signaler and a listener. In addition, we will show that interactions are an excellent system with which to study decision processes that result from intrinsic and external influences and to study proximate and functional questions. 247

Copyright 䉷 2000 by Academic Press All rights of reproduction in any form reserved. 0065-3454/00 $35.00

248

DIETMAR TODT AND MARC NAGUIB

Among the many ways in which animals interact with each other, studies on long-range signaling have been particularly fruitful. Long-range signaling often plays a central role in obtaining a mate and in regulating the spacing between competitors. Thus, it has important evolutionary implications since it is linked directly to an individual’s reproductive potential and thus its fitness. In addition, long-range signaling provides unique opportunities to study communication. Long-range signals are conspicuous and information flow is restricted to a specific signal modality, such as the visual or, most commonly, acoustic channel. Thus, the full signal can be recorded and the context of signaling often can be defined clearly because it does not change as rapidly as it does during close-range interactions. This provides a unique opportunity to study causes and effects of specific signal patterns or signaling strategies. The framework of this chapter is derived from an approach which considers that ‘‘genuine’’ interaction processes typically encompass both timespecific and pattern-specific relationships among the exchanged signals (Fig. 1; Todt and Hultsch, 1994, 1996, 1998, 1999). For instance, in a dyadic interaction a bird can respond to the other by adjusting the onset time of its song output to that of the other singer (time-specific responses; Todt, 1970a; Hultsch and Todt, 1982). Concurrently, a bird can use song patterns that are in some way linked to the patterns used by its counterpart (patternspecific responses; Todt, 1970a; Wolffgramm and Todt, 1982). This approach resulted from numerous studies on song structures and communication by song in species such as the blackbird (Turdus merula) and the nightingale (Luscinia megarhynchos) that have songs of about 3 s duration separated by silent intervals of similar duration (Todt, 1970a,b, 1971a, 1981; Hultsch, 1980). Here, we will apply this approach to singing interactions among territorial rivals (i.e., to vocal duels). Their temporally patterned signaling renders a performance process ideally adapted to vocal interactions with other birds. Since the framework applies equally well to singing interactions among cooperating singers (Todt et al., 1981), we will also apply it to the duets often found within pairs of mated individuals in tropical passerines. Finally, the approach is effective for analyzing comparative issues on vocal interactions when signals are related to each other in a dialogue-like manner. A. Vocal Long-Range Interactions Vocal signaling over a long distance is central to many biological issues. Studies on diverse taxa, including insects (Ro¨mer, 1993; Greenfield, 1993, 1994a,b), anurans (Klump and Gerhardt, 1992; Gerhardt, 1994; Grafe, 1996; Ryan, 1997), mammals (Todt et al., 1988; McComb, 1992; Tyack, 1996; Janik and Slater, 1997), and birds (Catchpole and Slater, 1995), have contributed

VOCAL INTERACTIONS IN BIRDS

249

Fig. 1. Interaction scheme. Bars symbolize different patterns (songs). Different shadings symbolize different types of patterns or songs. Interactions involve time- and pattern-specific responses. Time-specific response refers to a singer’s ability to start a song after different latencies after the onset or offset of the song of the counterpart [indicated by (a) alternating and (b) overlapping of songs and by the stacked symbols for bird 2]. Pattern-specific responses refer to a male’s ability to reply with a song that is specifically linked to the song of its counterpart (e.g., a song of the same type or a song of the shared repertoire; see also Figs. 4 and 5). Solid arrows indicate the influence of a previous song on the timing and pattern of the subsequent song. Dashed arrows indicate the influence of a song on choice and timing of the subsequent song of the counterpart. Songs on the right (song type C) symbolize a matching response.

significantly to resolving many research issues from sensory physiology to sexual selection and evolutionary ecology. In most species, the males use vocal long-distance signals to attract mates and/or to regulate the spacing to competitors. In addition, in some species, such as many tropical passerines, mates duet with each other in elaborate ways. In both cases, individuals settle within signaling distance and use their vocal signals not just as general displays but also to interact with each other. Studying the properties of such interactions provides important insights into the mechanisms underlying signaling behavior and into the functions of different signaling strategies. Good understanding of these kinds of systems can then provide a framework that is of general applicability to other forms of interactions, including those with concurrent signaling in different modalities. One of the best studied interaction processes performed by conspecific competitors is the singing behavior of passerine birds. During their evolution, songbirds have elaborated their species-specific vocal patterns into a

250

DIETMAR TODT AND MARC NAGUIB

signal system which, in many aspects, is outstanding in the animal kingdom. Their unit of interaction normally is the ‘‘song,’’ which is a temporally coherent signal pattern typically 2–4 s in duration (Fig. 2). Songs can be identified and distinguished by their phonological and temporal structure such as by the combination of their acoustic constituents, usually called elements or syllables. Songs that differ in these parameter configurations are categorized as different song types (Fig. 2), and there is evidence that such categorization can be biologically relevant (Todt, 1974; Todt and Hultsch, 1978; Kroodsma, 1982). The variety in song structure and size of vocal repertoires across species is striking. In some species each individual or even each population has only one song in its vocal repertoire. In contrast, many other species develop and perform signal repertoires ranging from a few to many different types of songs (Kroodsma, 1982; Catchpole and Slater, 1995). Numerous studies have shown that the singing behavior of birds is a multifunctional display. It has often been questioned whether or not the use of a particular vocal pattern (e.g., a song) may convey a particular message and, accordingly, whether or not different types of songs convey different messages. Currently, this question cannot be answered satisfactorily. In most species, the acoustic features of a single song provide an indication of the species, or the population, to which a singer belongs (Krebs and Kroodsma, 1980; Becker, 1982; Lynch, 1996; Payne, 1996) as well as an indication of the identity of the singer (Stoddard, 1996; Naguib and Todt, 1998). In addition, there is evidence for a differential use of at least categories of song types with reference to social context (Sossinka and Bo¨hner, 1980; Schroeder and Wiley, 1983a,b; Spector, 1992; Wiley et al., 1994), time of day (Schroeder and Wiley, 1983b; Highsmith, 1989; Morse, 1989; Hultsch, 1993), and season (Lampe and Espmark, 1987; Weary et al., 1994) or the singer’s location within its territory (Lein, 1978; Wiley et al., 1994). Furthermore, the vocal repertoire as a whole may reveal additional information such as a male’s age and quality (Catchpole, 1980; Hasselquist et al., 1996; Searcy and Yasukawa, 1996). Nowicki et al. (1998) suggested that the size of song repertoire could reflect conditions during posthatching development and thus reveal the genetic and/or phenotypic quality of a male. Most commonly, however, the current state or social status of an individual are signaled by features of several successive songs and by their use of songs during an interaction (Smith and Norman, 1979; Hultsch, 1980; Hultsch and Todt, 1982; d’Agincourt and Falls, 1983; Schroeder and Wiley, 1983a,b; Popp, 1989; Brindley, 1999; McGregor et al., 1992; Shackleton and Ratcliffe, 1994; Wiley et al., 1994; Nielsen and Vehrencamp, 1995; Dabelsteen et al., 1996, 1997; Naguib and Todt, 1997; Naguib et al., 1999; Naguib, 1999).

Fig. 2. Spectrograms of song from passerines singing with immediate (a, b) and eventual variety (c, d). (a) Sonagrams of a song sequence from a nocturnally singing nightingale. Note that each song type is different. Each male has a repertoire of approximately 200–300 different song types. Songs at approximately 45 s and at approximately 113 s are whistle songs (see text). The terminal section of a song can be prolonged to more than 10 s (see song at approximately 125 s). We assume that frequent lengthening of songs indicates a bird’s arousal. (b) Sonagrams of a song sequence from a European blackbird. Note that each song type is different. The tonal part is the motif section. The variable part is the twitter section. (c) Sonagrams of three consecutive songs of a chaffinch. Chaffinches sing in a repetitive mode (with eventual variety). All three consecutive songs are of the same type. The singer will eventually switch to another type. Chaffinches have repertoires of approximately 2–5 song types. (d) Sonagrams of three consecutive songs of a Carolina wren. Carolina wrens sing in a repetitive mode (with eventual variety). They have a repertoire of approximately 25–40 different song types. A song type may be repeated more than 100 times before a singer switches to a song of another type. Sonagrams show a switch to a song of a different type between the second and the third song.

252

DIETMAR TODT AND MARC NAGUIB

Fig. 2 (Continued )

VOCAL INTERACTIONS IN BIRDS

253

B. Outline In this chapter, we will discuss how birds respond to each other and the proximate and ultimate consequences of different singing strategies during an interaction. Interactions are characterized by specific relations between the signals of two or more signalers and they are commonly composed of a series of acoustic signals emitted by at least two individuals. As we will discuss in more detail in the final section, vocal interactions can be analyzed best on the basis of a dialogue model in which two components are crucial to characterize and interpret the process: (i) the timing of signals in relation to conspecifics’ signals and (ii) the use of specific song patterns when replying to another singer. Both the timing of songs and pattern-specific responses can result in highly complex and dynamic singing interactions. We will deal with these interactions by describing and evaluating the rules of song control and retrieval as well as the relationships of the signals of interacting individuals to each other. Responses are often obvious but they can also be subtle, delayed, or may only be evident in long vocal interactions. In addition, interactions are a sequential process. Thus, the probability of a response and its meaning might depend on what information has been accumulated since the beginning of the interaction. In other words, much information could be extracted sequentially or could be cumulative. By focusing on single events one would thus miss information that could be important for the sender and the listener. On the other hand, it might be the rare events with their high information value that are functionally most relevant. Thus, assessing the salience of specific responses is not a trivial task because the salience of rare events cannot be detected by using standard statistical procedures. For assessing the function of signals during an interaction, we also need to consider the signaling context (Fig. 3). The context is often crucial for the information conveyed and for the functional and evolutionary significance of specific singing strategies. For instance, we will argue that song matching (replying with the same song pattern) may have different functions, depending on the context and the social relationship between the interacting individuals. Similarly, the timing of songs may carry very important information in some contexts, such as during close-range disputes, but it may be irrelevant in other situations. Here, we will focus on long-range vocal interactions in which singers do not approach each other but respond to each other vocally. We will present examples from various studies conducted on many species in order to discuss the striking diversity found in birds’ singing behavior. Since many of our ideas and concepts have been derived from vocal interactions in nightingales, we will exemplify certain principles in communication by open-

254

DIETMAR TODT AND MARC NAGUIB

Fig. 3. Context in which vocal interactions between temperate-zone territorial passerines occur: (a) interactions during establishment of a territory, (b) boundary dispute between neighbors, (c) long-range countersinging, and (d) interactions between intruder and territory holder.

ing a wider window into this species. Nightingales provide an excellent model with which to study vocal interactions since they interact with each other at night over many hours without changing song posts. Thus, the relationship between the songs of the interacting singers can be studied descriptively and experimentally in a unique manner without changes in the external context confounding the analysis. Birds often interact vocally in unambiguous ways. In a community of territorial individuals, however, there are also many other cases in which it is difficult to decide whether or not two or more singers are truly interacting. The dawn chorus in which many individuals sing more or less simultaneously is an example. Species in which birds alternate long singing bouts with each individual holding a ‘‘monolog’’ for a considerable time, such as Acrocephalus warblers (Catchpole, 1976), are another example. In such cases, it will be difficult to decide to what extent, if at all, singers respond to each other. Consequently, we will focus first on cases that are more or less obvious (dyadic interactions). We will focus on male–male interactions

VOCAL INTERACTIONS IN BIRDS

255

among territorial passerines and then deal with intersexual interactions between mates of duetting species. We then expand the information drawn from these dyadic interactions to more complex forms of interactions. In addition, we will discuss if and how information exchanged during vocal interactions affects other listening and potentially eavesdropping individuals.

II. Dyads: Interactions between Two Individuals The most apparent and widespread forms of interactions are between two individuals. Thus, it is not surprising that in birdsong the vast majority of studies have focused on such dyads. In fact, most of what we know about mechanisms and functions of interactions is drawn from either descriptive studies on dyadic interactions or from playback experiments in which one individual is confronted with playback of conspecific song. In a particularly clear case of a dyad, both individuals sing during the same time period so that from a formal perspective there are two domains to consider (Fig. 1). One bird may respond to the other by adjusting the timing of his song output to that of the other singer (time-specific responses; Todt, 1970a; Hultsch and Todt, 1982), and he may use a song pattern that in some way linked to the pattern used by his counterpart (pattern-specific responses; Todt, 1970a; Wolffgramm and Todt, 1982). These kinds of responses are not mutually exclusive, but for practical reasons we will first treat them separately. Then, we will discuss the combination of patternand time-specific responses, a feature that has been considered in only a few studies (reviewed in Todt and Hultsch, 1994, 1996, 1999).

A. Time-Specific Responses Consider two males singing in the center of their own territories. Each male may sing without paying attention to the other singer. Thus, they may more or less accidentally overlap some songs and accidentally alternate some songs. A lack of a relation of the output of the two singers to each other raises questions of whether or not the two singers are truly interacting with their songs. There are many cases, however, in which the evidence for true interactions is remarkably clear (Todt and Hultsch, 1994, 1999). Many species, for instance, can perform many time-specific responses that allow one to distinguish several different strategies of interaction by song. In some contexts, singers may also switch between different singing strategies. For instance, birds may react specifically to a counterpart only during

256

DIETMAR TODT AND MARC NAGUIB

short sections of an often long vocal interaction. Findings such as this show that analyzing the time domain of song exchange can be a challenging task. 1. Alternating Singing Alternating singing occurs not only between individuals of the same species but also among individuals that share the same acoustic space. It is widely accepted that one advantage of alternating singing is to avoid acoustic interference and consequently to increase broadcasting range and quality (Cody and Brown, 1969; Ficken et al., 1974, 1985; Wasserman, 1977; Gochfeld, 1978; Hultsch, 1980; Popp et al., 1985; Popp, 1989; Wiley, 1994). Such avoidance of interference also sets the context in which more subtle information can be exchanged because each participant can listen to any response. Thus, the advantage of alternating songs is not only that others can better perceive one’s own signal but also allows the individual to listen to others’ signals and to potentially decode subtle information. It is not well known how a bird’s hearing physiology deals with its own songs since they are broadcast at 90–100 dB. However, without doubt any such loud signal, in combination with the apparent dampening of the middle ear (Nottebohm, 1991), will impair perception of signals emitted by distant conspecifics. a. The Nightingale. Nightingales usually alternate songs during their long nocturnal vocal interactions. These interactions are intriguing. Usually, nocturnal song begins between about 10 pm and 1 am and continues until dawn. The long period during which they interact vocally provides a unique opportunity to study vocal interactions without the context of the vocal responses changing because singers switch locations (Todt, 1971a; Hultsch, 1980; Naguib, 1999). Here, as with other vocal interactions, the timing of songs is an important parameter to assess if two individuals are truly interacting. Analyses of natural interactions (Hultsch and Todt, 1982) and playback experiments (Hultsch, 1980; Hultsch and Todt, 1982; Naguib, 1999) have revealed important insights into how songs are timed during a vocal interaction. In addition, experiments have identified how the timing of songs during alternating singing reveals information about the singer. In an early interactive experiment, for instance, a territorial nightingale was exposed to alternating playback of conspecific songs with varying duration (short and long songs). The results showed that the singer adjusted the timing of its song onsets to the offsets of the playback songs (Hultsch, 1980; Todt and Hultsch, 1999). This adjustment then affected the length of the subject’s silent intervals between songs. It resulted in a shortening or extension of the subject’s silent intervals between songs, respectively, dependent on the sum of onset latency and the duration of playback songs (Fig. 2 in Todt and Hultsch, 1999). At the end of stimulation trials there

VOCAL INTERACTIONS IN BIRDS

257

was a striking aftereffect: Regardless of whether a subject’s silent intervals between songs had been triggered to give shorter or longer interval duration (compared to unaffected singing), the bird did not switch back immediately to its normal timing of song delivery. Instead, in both cases it tended to prolong these intervals before gradually approaching its ‘‘normal’’ interval duration. Together with the fact that the subjects ‘‘tolerated’’ considerable variation in the song onset latencies of the playback songs, the latter effect demonstrates that subjects modified their temporal pattern of singing as a result of being in an interactive state rather than behaving according to a simple stimulus/response mechanism. The state of the individual, such as its readiness to interact vocally, is an important variable. In some cases, birds may not interact with playback in such an obvious manner but instead increase song rate without a clear relation to the playback or another singer. This is shown when nightingales, during their daily dawn chorus, modify their nocturnal singing style by drastically reducing the duration of their silent intersong intervals (Hultsch and Todt, 1982). Detailed analyses of singing before dawn and playback experiments conducted in the field allowed us to uncover three different types of interaction strategies. These were first distinguished by formal criteria and labeled by terms such as ‘‘inserting,’’ ‘‘overlapping,’’ and ‘‘autonomous singing’’ (Hultsch and Todt, 1982). Inserters generally begin their songs during the silent intersong intervals of a neighbor. They do so by starting their songs more frequently than expected by chance about 1 s after a neighbor has terminated its preceding song. Overlappers, on the other hand, begin a song before a neighbor has terminated its preceding song. These singers also have preferred song onset latencies, but in this case about 1 s after a neighbor has initiated a song. In contrast to these birds, autonomous singers do not adjust their song output in relation to the timing of their neighbors’ songs. There is evidence that these different behaviors serve different functions and that, in particular, the latency at which singers start their songs has a specific meaning (Table I; Todt and Hultsch, 1994, 1999; Naguib and Todt, 1997; Naguib, 1999; Naguib et al., 1999; see also next section). In a study to further examine such interaction strategies, Naguib (1999) found that the timing of songs during alternating singing depended on the kind of experience a bird had had with its counterpart. In an interactive playback experiment subjects sang at higher rates during an alternating treatment when they had previously been overlapped by the playback than did subjects that were not previously overlapped by playback. These experiments suggest that the timing of songs during alternating singing reflects the arousal or general state of the singer and functionally its readiness to escalate an interaction, as discussed later.

258

DIETMAR TODT AND MARC NAGUIB

TABLE I Overview of Different Characteristics of Vocal Interactionsa Characteristic Song alternating

Context Widespread

Proximate aspects

Message/meaning

No signal masking, permits mutual listening

Interactive state Low or medium agonistic motivation Meaning depends on exact timing Song Close-range disputes Masking effects high Directing response overlapping Mutual listening is Agonistic signal, threat impaired Long-distance Masking effects low Agonistic attitude singing Mutual listening is Lack of interest in listening? impaired Autonomous Dawn/dusk chorus, Unclear, challenge? Noninteractive state, high singing established males status? established territory holder? Song matching Close-range dispute Attracts attention Addressed response Long-distance Agonistic signal, threat or singing possibly friendly signal depending on exact timing Song type Close-range dispute Attracts attention Directing response switching Low to medium agonistic signal a The meaning of a specific strategy will depend on several contextual factors and on the relation between the singers. See text for details.

b. The Timing of Songs during Alternating Singing. As the experiments on nightingales suggest, the timing of songs during alternating singing might have more meaning than just avoidance of acoustic interference or the enabling of an individual to listen to a response. The timing of song onset is a variable feature that may carry information. Birds might start their song immediately after the end of their counterparts’ songs or they may take more time before they reply. Smith and Norman (1979) interpreted such asymmetry in the timing of songs as being associated with differences between the two individuals. Their observations on red-winged blackbirds (Agelaius phoeniceus) suggested that the more threatened bird started its song with a shorter latency after the counterpart’s song had terminated than vice versa. Popp (1989) came to a similar conclusion in his study on ovenbirds (Seiurus aurocapillus). Recent experiments with nightingales also suggest that asymmetries in the timing of songs during alternating singing can carry important information. Nightingales that already had experience

VOCAL INTERACTIONS IN BIRDS

259

with a rival as being aggressive timed their songs with shorter intervals than did subjects that did not have such experience (Naguib, 1999). In addition, in a different experiment, territorial males responded more strongly to the leader of two simulated rivals that were alternating songs asymmetrically (Naguib et al., 1999), as discussed in Section IV. These experiments indicate that the timing of songs during alternating singing is an important variable in vocal interactions. Their results are also in line with the general finding that asymmetries in displays are an important cue for distinctions between individuals with different qualities (Maynard Smith and Parker, 1976; Hammerstein, 1981; Naguib et al., 1999b). In summary, the benefits in terms of gaining signal space and improving transmission quality by alternating songs are straightforward. In addition, the timing of songs during alternating singing can be highly specific. The timing of songs could reflect reaction times in certain contexts and this might reflect the motivation or even the quality of a singer. It is a feature of delivering songs that can potentially be varied along a continuous scale so that much more information might be encoded in the timing of song during alternating singing than has been assessed to date. So far, the studies that have addressed functional consequences of asymmetries in timing of song during alternating singing have found it to be of functional importance. This should encourage continuing studies on this topic. 2. Song Overlapping Although song alternation is the most common singing form, birds frequently overlap songs of conspecifics (Fig. 4). In trying to understand the different singing strategies, song overlapping appears to be puzzling at first. Song overlap occurs when a singer starts a song before its counterpart has terminated its song (Fig. 1b). Such overlap can result in jamming or masking the counterpart’s song, dependent on song structures (e.g., the difference in frequencies of the overlapping and overlapped syllables) and on the distance between singers. This strategy is curious because not only the addressee but also the overlapper can suffer from signal masking. Thus, there are several hypotheses concerning this problem, all of which are based on the idea that the benefits a bird gains by overlapping outweigh the disadvantages or possible costs of doing so (reviewed by Todt and Hultsch, 1994, 1996, 1999). Song overlapping has the following proximate consequences on the behavior of the singer whose song is overlapped: 1. Birds whose songs are overlapped tend to interrupt that song or even stop singing altogether. Thus, being overlapped may result in a loss of signal time. This in turn will result proximately in relatively more signal

260

DIETMAR TODT AND MARC NAGUIB

Fig. 4. Sonagrams of a vocal interaction between two free-ranging territorial nightingales. The sequence exemplifies several different forms of replies that are common in vocal interactions in nightingales. Singer 1’s songs (top) overlap most of the songs of singer 2 (bottom). Overlapping events are indicated by arrows. The first song of singer 2 is a nonoverlapping match. The last song of singer 1 is an overlapping match followed by a nonoverlapping match countermatched by singer 2. The long song of singer 1 in the center is an overlapping match that is also overlapped by the subsequent song of singer 2.

time for the overlapper. Song interruption might occur either as a result of an avoidance response or because a bird has decided to listen to a neighbor’s song (Hultsch and Todt, 1982). 2. It is possible that overlappers perceive the masking of their own song as less disturbing. Since masking of its own song is the result of the singer’s own behavior, the bird is presumably prepared to be masked and, importantly, is also in a state that tolerates such masking. Negative consequences of an animal’s own behavior, or negative reafferences, seem to be much easier to deal with than ‘‘negative’’ effects that result from external stimuli. In addition, overlappers do not overlap every song. Thus, a bird will always be uncertain of whether or not its song will be overlapped. This uncertainty might be an important reason for birds becoming ‘‘uncomfortable’’ when being overlapped. This might contribute to an avoidance reaction as described previously. 3. Songs that are overlapped may suffer more from being jammed than does the overlapping song. The cost to the overlapper is that the first part of its own song may be masked. In many species, song types differ from each other most in their second part, thus reflecting an informational hierar-

VOCAL INTERACTIONS IN BIRDS

261

chy. In many thrushes, for instance, the different song types in an individual’s repertoire are often initiated by the same syllables but differ in their middle and terminal sections (Todt, 1969, 1970a,b; Todt and Hultsch, 1978, 1999; Hultsch, 1980; Naguib et al., 1991; Naguib and Kolb, 1992). Consequently, middle and terminal sections bear greater potential information than the initial sections of a song. Thus, in some species an overlapper often overlaps the more informative section of its counterpart’s song but risks masking of only the first part of his own song, saving the more complex part from being jammed (Hultsch and Todt, 1982). This is true in particular for species such as the Eurasian blackbird or the bluethroat (Luscinia svecica) in which the terminal section is more complex and less loud than the initial section of a song (Todt, 1970d; Naguib and Kolb, 1992). A similar argument might also hold for species with a syntactically relatively simple song structure such as the great tit (Parus major) and the coal tit (Parus ater). Both species vary the length of their song and it has been argued that song length may be important for communication (Lambrechts and Dhondt, 1988; Weary et al., 1990; Adhikerana and Slater, 1993). Overlapping may impair perception (by others) of the length of the song that is overlapped. Although this argument also holds for the overlapper (the beginning of whose song will be difficult for others to perceive), he might still be in a better position. Overlappers might benefit by attracting the attention of third individuals, such as females that attend to an interaction (Todt, 1981; Naguib and Todt, 1997; Naguib et al., 1999; Todt and Hultsch, 1999; Otter et al., 1999). 4. Although overlapping is most common during relatively close-range disputes, it can also regularly be observed in long-range interactions. What is the objective of overlapping a song during such long-range interactions? The overlapping song will be attenuated to relatively low amplitude so that masking effects will be small. It is possible that overlapping might still be perceived as aggressive or ‘‘unfriendly.’’ It might be a signal to indicate the state of the singer, which might reveal information on its way of interacting in other contexts. At the very least it indicates that the overlapper is not ‘‘interested’’ in listening to the counterpart’s full signal. Song overlapping is usually interpreted as a directed agonistic signal. For instance, it can often be observed in agonistic duels at close range during the establishment of a territory or during boundary disputes. In addition, birds commonly react in intense and aversive ways when overlapped by playback (Todt, 1981; Hultsch and Todt, 1982; Brindley, 1991; McGregor et al., 1992; Dabelsteen et al., 1996, 1997; Naguib, 1999). In playback experiments on laboratory-kept blackbirds, Todt (1981) showed that these birds avoided song posts on which their songs were overlapped. The degree of

262

DIETMAR TODT AND MARC NAGUIB

avoidance of these song posts also increased with the volume of playback. This suggests that avoidance was indeed primarily to prevent the bird’s own song from being masked. In addition, birds might avoid being overlapped because it impairs their perception of details of the overlapper’s songs. a. The Nightingale. Playback experiments on nocturnally singing nightingales have shown that birds deviate more from their singing rhythm when overlapped by playback than when playback stimuli are alternated with their songs. Hultsch and Todt (1982) showed that nightingales interrupted their singing when songs were played to them at high rates. In subsequent interactive playback experiments, nightingales responded correspondingly, but in addition sang at higher rates in overlapping treatments than in playback treatments in which songs were alternated with them (Naguib, 1999). During the overlapping treatment, they also used more extreme lengths of silent intervals between songs: Subjects used significantly more short silent intervals (⬍3 s) and significantly more long silent intervals (⬎5 s) between songs. In addition, they interrupted significantly more songs during overlapping treatments than they did during alternating treatments. Such results suggest that birds can use mixed strategies to avoid being overlapped. A similar strategy was found in great tits (P. major), which increased the variation of song length when they were overlapped by playback (Dabelsteen et al., 1996). As mentioned previously, nightingales use different singing strategies during nocturnal singing (Hultsch and Todt, 1982). Inserters started most songs in the silent intervals between their counterpart’s songs, whereas overlappers began most of their songs while the counterpart was still singing its song (Figs. 1 and 4). There is evidence that these different strategies might reflect relative differences among males in terms of state or status, and that, for instance, autonomous singers who, by definition, do not modify their singing in such ways might be well-established individuals of high status (Hultsch and Todt, 1982). Individuals that interact in specific ways (e.g., alternate or overlap) might do so to probe the counterpart and to test and establish the relationship. From this perspective, there might be less need to interact in such specific ways for well-established individuals. Interestingly, some males that were overlappers toward one neighbor changed to the inserter role when interacting with another neighbor (Hultsch and Todt, 1982). Such asymmetries in singing interactions, which depend on the individual with whom the singer is interacting, suggest that neighbors have developed specific relations to each other. In fact, differences in relations among territory holders might be common in territorial birds. Males differ in state and quality, such as in their resource-holding potential and in their potential to compete for mates. These differences might well be reflected in the way the individuals use their songs during

VOCAL INTERACTIONS IN BIRDS

263

an interaction. Such asymmetric relations might be similar to dominance relations, with the constraint that here each territorial male retains strong site-dependent dominance. b. Timing of Songs during Overlapping. So far, we have dealt with song overlapping as one response category. However, birds can vary how much of the counterpart’s songs they overlap. Do they overlap just the end or do they overlap as much as possible? A series of playback studies on European blackbirds (Todt, 1970a,b; Wolffgramm and Todt, 1982) and nightingales (Todt, 1971a; Hultsch and Todt, 1982) found that birds overlap a potential rival with preferred latencies indicating that it is a voluntary signal. The degree of overlap can reflect reaction times that presumably mirror a male’s state or status. In addition, the degree of overlapping might be a crucial component for the functional relevance of masking effects (Fig. 1b). If the overlapper starts its song too early, the overlapper will risk having its own full song masked by the song that he is overlapping. On the other hand, overlapping just the end of a song jams only minor parts of the counterpart’s song. In addition, if singers interact over long distances, the overlapper would need to take into account the time delay caused by the slow speed of sound. Thus, minor overlapping at the overlapper’s song post might not be perceived as such by the singer that is overlapped. In the studies on blackbirds, the latency with which subjects overlapped broadcast songs depended on whether or not they matched the song. Blackbirds and nightingales overlap and match a song concurrently, a response that Todt (1981) and Wolffgramm and Todt (1982) suggested had particular signal value, as discussed in Section II,C. In summary, there is convincing evidence that high rates of song overlapping are used as a threat signal and perceived as such. However, overlapping might even carry important information when it occurs only rarely. It will be interesting to see if future research can address such more subtle aspects of timing of songs during an interaction. B. Pattern-Specific Responses (Song Matching and Others) Most passerine species have repertoires of different song types that are delivered in one of two species-specific ways. In some species, individuals sing in a versatile manner (with immediate variety) by usually not immediately repeating the same song type (Figs. 2a and 2b). After each song, individuals choose a different song from their repertoire, which may comprise more than 100 different song types. In other species, males sing in a repetitive manner (with eventual variety) by usually repeating the same song type a few to several hundred times (Figs. 2c and 2d) before switching to a song of a different type. There is a tendency for species with large

264

DIETMAR TODT AND MARC NAGUIB

song repertoires to deliver their song with immediate variety. Species with small repertoires tend to sing with eventual variety (Kroodsma, 1982). These differences in singing style have profound implications for the use of song during interactions, and it is thus best to deal to some extent separately with species using these different singing styles. How do we detect pattern-specific responses and how can we decide if a song is such a kind of response to the perceived song? A formal approach to interactions has proved to be valuable in this case. Such a formal analysis of interaction processes begins with an evaluation of pattern-specific relationships between process constituents. In vocal interactions these constituents are usually song types or more basic vocal units sung by two signalers. This maneuver requires one to estimate the size of the respective song type repertoires. This is necessary to assess the probability with which two individuals would match each other by chance when not responding specifically to each other. Second, the ‘‘normal’’ singing programs or styles of the singers have to be assessed. Several studies have shown that the probability that a singer sings a specific song depends on what it has just sung and on the similarity between the perceived song and the singer’s own version of that song type (Todt and Hultsch, 1996, 1998, 1999). 1. Song Matching Most researchers who have studied the role of pattern-specific responses have focused on song matching—a situation in which interacting individuals reply to their counterpart with the same song type or song pattern. Song matching is one of the most conspicuous phenomena during vocal interactions and it can be experienced even by ear most clearly every spring in temperate zone passerines. Here, we use the term song matching exclusively to describe singing interactions in which birds reply with the pattern or song type that another bird has just sung. Although song matching requires that singers share parts of their song repertoire, birds with similar song repertoires may never match each other when singing. Thus, we do not use the term to describe song sharing among neighbors (Hultsch and Todt, 1981) or similarities among repertoires per se (Payne, 1982). In order to avoid terminological confusion, Todt (1970b, 1971a) suggested the term ‘‘equivalent vocal response for song matching’’ which was contrasted to the term ‘‘convalent vocal response.’’ For the same reason other authors have used the term ‘‘matched countersinging’’ when referring to the actual performance of singing birds (McGregor, 1991). With a few exceptions (McGregor, 1986), in all species studied interacting birds match each other with varying frequencies or reply with other specific patterns, depending on the context (Bre´mond, 1968; Lemon, 1968, 1974; Todt, 1968a,b, 1969, 1970a,b, 1971a, 1974, 1981; Dixon, 1969; Bertram, 1970;

VOCAL INTERACTIONS IN BIRDS

265

Falls, 1985; Kroodsma, 1979; Hultsch, 1980; Krebs et al., 1981; Slater, 1981; Falls et al., 1982, 1988; Wolffgramm and Todt, 1982; Kramer and Lemon, 1983; Schroeder and Wiley, 1983a,b; Whitney and Miller, 1983; Simpson, 1985; Horn and Falls, 1986, 1988a,b; Hultsch and Todt, 1986; McGregor, 1986, 1991; Weary et al., 1990; McGregor et al., 1992; Stoddard et al., 1992; Shackleton and Ratcliffe, 1994; Todt and Hultsch, 1994, 1996, 1999; Nielsen and Vehrencamp, 1995; Beecher et al., 1996). Some studies that have found birds to avoid matching even demonstrate a response equivalent to song matching (Whitney and Miller, 1983; Whitney, 1990; Capp, 1992; Dufty and Pugh, 1994). In singers that repeatedly sing the same song type, birds that avoid matching by switching to another song when matched show a response that is also pattern specific. 2. Other Kinds of Pattern-Specific Responses Birds might respond to each other with distinctly different patterns that are linked to each other in specific ways. A striking case was recently described for song sparrows (Melospiza melodia). Males responded to songs of their neighbors not necessarily with an exact match but with a different song type from that particular neighbor’s repertoire (Beecher et al., 1996). This kind of response is much more subtle and hard to recognize than vocal matching. It requires good knowledge of the repertoires and singing styles of subjects. Such prerequisites are also crucial for an identification of another type of pattern-specific reaction termed ‘‘convalent response,’’ or synonymously ‘‘vocal supplementing’’ (Todt, 1971a; Todt et al., 1981). During vocal supplementing a bird responds to a song with a different type of song but that has a high probability of following the perceived pattern in the bird’s own (or maybe its counterpart’s) normal singing program (i.e., song type succession). Vocal supplementing, which is common in both duels and duets by song (see Section II,F), has also been described for calling in brown-headed cowbirds (Molothrus ater; Dufty and Pugh, 1994). In general, over time neighbors may develop specific relations to each other that can result in a mode of interaction that is very specific to the particular individuals. This can be true for interactions by song in blackbirds and nightingales as well (Todt 1970b, 1971a). 3. Intrinsic Rules of Song Type Delivery Given the different and elaborate forms of pattern-specific responses, we must determine why and when birds interact with each other and in what ways. The fact that birds do not always match each other indicates that matching is more than simply an automatic feedback system in which a bird sings the same song that it has just heard. We argue that, depending on the species, matching and other forms of pattern-specific responses result

266

DIETMAR TODT AND MARC NAGUIB

from a series of factors, of which triggering by the received song that is matched is only one. Distances between singers (or even other potential listeners) and other contextual variables are also important for matching to occur. In general, song matching has been studied from two perspectives: (i) identification of matching during territorial interactions with neighbors, strangers, or their simulations by playback (the functional perspective) and (ii) identification of decision processes that underlie choice of song types such as those that match the received pattern (the cybernetic perspective). In this section we will discuss the latter approach. Decision processes involved in song type choice during singing are as fascinating as they are complex in songbirds in which males have large song repertoires and deliver them with immediate variety. Several studies have shown that birds follow specific rules in their succession of song types, and that in many cases these are very similar across species. Examples are the American redstart (Setophaga ruticilla; Lemon et al., 1993), the bluethroat (Naguib and Kolb, 1992), the canary (Serinus canaria; Wolffgramm, 1975), the chaffinch (Fringilla coelebs; Hinde, 1985; Slater, 1981), the common nightingale (Todt, 1970b, 1971a; Hultsch 1980), the common redstart (Phoenicurus phoenicurus; Thimm, 1973), the Eurasian blackbird (Todt, 1968a, 1969, 1975a), the marsh wren (Cistothorus palustris; Verner, 1976), the cardinal (Richmondena cardinalis; Lemon and Chatfield, 1971), the mistle thrush (Turdus viscivorus; Isaac and Marler, 1963), the stonechat (Saxicola torquata; Schwager and Go¨ttinger, 1984), the thrush nightingale (Luscinia luscinia; Naguib and Kolb, 1992), and the whinchat (Saxicola rubetra; Schwager and Go¨ttinger, 1984). Thus, other than in bout singers, after each song a singer must select a new song type from the available pool. The choice of song patterns during an interaction raises intriguing questions on how intrinsic rules of song type delivery are managed when birds reply to conspecifics since they often will have to deviate from their normal singing program or singing style. These rules, in fact, might impose constraints on a singer’s ability to match a song. Consider the vocal virtuosity of the nightingale. Theoretically, a singer needs to decide in each pause of about 3 s between songs which of its 200–300 different song types to sing next. However, there are mechanisms that are likely to help a bird to solve this problem, e.g., by subdividing the large repertoire into subsets of sequentially associated song types (Hultsch, 1993). The subdivision reflects a hierarchical memory organization which is thought to facilitate the retrieval of a particular set of songs and then a particular type of song (Todt and Hultsch, 1999). There is evidence that this organization also influences the occurrence of convalent responses (vocal supplementing) and vocal matching. These mechanisms have at least

VOCAL INTERACTIONS IN BIRDS

267

three implications. The mechanisms can lead to specific vocal interactions, but at the same time also impose constraints on the flexibility with which a singer can respond in pattern-specific ways to external stimuli. By promoting an independent song retrieval, however, the mechanisms can help an individual to sing autonomously and thereby to challenge a conspecific neighbor. Interactions between singers that deliver songs with immediate variety are thus more complex than those between singers that sing with eventual variety. These differences in singing style are important considerations when principles underlying matching are generalized or when the functional significance of song matching is discussed. In our view, the problems faced by a singer that sings with immediate variety when trying to match a counterpart have not been considered sufficiently in discussions on song matching. The reason is simple: Most field studies on the function of song matching have been conducted on species that sing in a repetitive manner (with eventual variety; see Section II,B,6). However, evaluating both singing styles is essential to comprehend song matching as a general phenomenon. In the Cybernetic parts that follow we will focus on variable singing by concentrating on studies that have investigated interactions among blackbirds and nightingales. Beforehand, however, we will briefly highlight a methodological feature that is usually not considered by most investigators but has turned out to be a valuable tool in playback experiments with either species. In many studies our subjects were exposed to a given set of auditory stimuli for a relatively long time. Although birds over time reduced the intensity of their territorial response, such as approach to the loudspeaker, neither blackbirds nor nightingales gave up their vocal responses. These vocal responses even became more elaborate and eventually reached a form that in the field can be found between two familiar neighbors. In other words, our results suggest that repeated presentation of the same stimuli is one of the few useful methods for attempts to uncover the mechanism of vocal interaction by song. 4. A Candidate Species: The Eurasian Blackbird The way blackbirds interact has allowed important insights into the mechanisms of song control. A series of playback experiments showed that both the subject’s rules of song succession (self-program) and the nature of the heard song are crucial components in their vocal responses (Todt, 1970a,b, 1975a; Todt and Wolffgramm, 1975). Figure 5 exemplifies these phenomena. Birds that share part of their repertoire interact by matching each other. After awhile, however, birds stop matching each other and continue to sing other songs from their repertoire that are not shared by the opponent. Which factors lead both singers to repeatedly sing shared songs and which

268

DIETMAR TODT AND MARC NAGUIB

Fig. 5. Sequence (symbolized) of two interacting blackbirds. Songs are indicated by capital letters in the left column with shared ones in the shaded area. Both birds sing the shared patterns at the same time but then both start to sing unshared songs. In the first bout of shared songs, bird 2 (⫻) sang a shared song first (arrow 1). In the second bout of shared songs, bird 1 (䊊) sang first a shared song (arrow 2).

factors lead them both to switch to unshared songs? When interacting in pattern-specific ways with a rival, as shown in Fig. 5, a singer will have to restrict its singing to those shared songs. An intriguing question is the following: How long can a singer do this and when are there internal components of song control that ‘‘force’’ a bird to also sing other song types, including those that are not shared? The primary goal of the studies on song matching was to investigate mechanisms that underlie song control during solitary singing and during vocal interactions (Todt, 1969, 1970a, 1974, 1975a; Wolffgramm and Todt, 1982). These experiments were based on playbacks with modified and unmodified sequences of a subject’s own songs. Although subjects will not hear their own song from an external source in natural encounters, they readily interact with playback of their own song. This provides the opportunity to investigate factors that affect choice of specific song patterns. Playback of modified sequences of the subjects’ own songs then allowed assessment of the properties of the rules of song type delivery (Todt and Hultsch, 1978, 1996). To what extent can a blackbird change its normal song succession when interacting with its own song broadcast from an external source? In a systematic investigation on factors affecting choice of song types, several factors were identified that influenced which song type will be sung

VOCAL INTERACTIONS IN BIRDS

269

next (e.g., specific intervals between renditions of the same song type and high transition probabilities between specific song types; Todt, 1968a, 1969, 1970a,b, 1974, 1980; Todt and Wolffgramm, 1975). These studies led to a cybernetic model that provided explanations of how transition probabilities and periodicities of song type renditions are processed when a bird interacts with playback. The playback experiments showed that the probability of whether or not a bird matched a song did not only depend on what the bird had just sung but also on the perceived song and its similarity to the bird’s own song and on how many songs had elapsed since it had sung a song of that particular type. These studies are among the few in which several rules of song type delivery have been identified and for which a model has been developed that takes into account the interaction of these rules (Todt, 1975; Todt and Wolffgramm, 1975). In playback experiments, Todt (1970a) and Wolffgramm and Todt (1982) included both pattern-specific responses (song matching) and time-specific responses and were able to show that subjects’ choice of song patterns was not independent from the time delay in their response, as we will discuss in Section II,C. The concept of adjusting the playback stimulus to a subject’s own vocal output has yielded important functional insights into singing interactions through the use of computers or digital recorders that provide an easier and more flexible choice of broadcast stimuli (Dabelsteen and Pedersen, 1990; Dabelsteen, 1992; McGregor et al., 1992; Nielsen and Vehrencamp, 1995; Naguib, 1999). 5. A Candidate Species: The Common Nightingale During their extended nocturnal singing episodes, nightingales may respond to each other by adjusting the timing of their songs, as described previously, and by pattern-specific responses such as matching or vocal supplementing (Fig. 4). Their large vocal repertoire and a medium level of repertoire sharing (앑30%; Hultsch and Todt, 1981) results in a chance probability near zero for song type matching. Thus, each matching event and the frequency of matching might have specific signal value. In addition, due to vocal learning, neighbors may also share parts of their singing programs (i.e., preferred sequential successions of song), and this increases the probability for both matching and vocal supplementing to occur (Todt and Hultsch, 1996, 1998, 1999; Naguib and Todt, 1998). Territorial nightingales often switch roles, with one individual being the matcher for a few songs and the other individual being the matcher on other occasions (Fig. 4). Thus, dyadic singing is characterized by matching events being either clustered or scattered over a singing performance. The following is a question of interest in song matching: Are there song types that are more likely to elicit a specific response and are possibly more

270

DIETMAR TODT AND MARC NAGUIB

likely to be matched? Each male nightingale has one category of songs that comprises approximately 20% of the songs sung during a nocturnal bout (Hultsch, 1980). These so-called whistle songs not only stand out in their structure (Fig. 1a; Hultsch and Todt, 1996) but also can be matched and overlapped concurrently by a song containing whistles of the same pitch (Hultsch, 1982). Whistle songs also increase the probability that rivals start to sing. We recently conducted playback experiments on 28 territorial male nightingales that were not singing at the beginning of the playback. We found that subjects started to sing with significantly shorter latencies when whistle songs were broadcast than when nonwhistle songs were broadcast (D. Todt, H. Hultsch, and M. Naguib, unpublished data). In fact, when trying to localize silent birds while mapping territories early in the season, a human that whistles may often stimulate a nightingale to sing. Whistle songs can also probably play a significant role in attracting conspecific females (Hultsch, 1981; M. Metz and H. Hultsch, unpublished data). Sometimes, songs are matched successively by several singing neighbors. Thus, if one bird sings song type A, its neighbor might match it and the next neighbor will match it again. This results in a song type being passed along a series of territories. Functionally such ‘‘multiple matching’’ might result in birds being identified by outsiders as familiar or established individuals. Here, matching might strengthen relationships between individuals that interact while concurrently being aversive to outsiders. In addition to matching each other, nightingales may respond in the way we refer to as vocal supplementing (synonymous with convalent responses; Todt, 1971a). Consider two birds, bird 1 and bird 2. When bird 1 hears song type A it may reply by either matching it or by singing song type B, which it would normally sing after having vocalized song type A (Fig. 1). If bird 2 also usually sings song type B in close sequential association to song type A, the vocal supplementing of bird 1 might increase the probability that bird 2 will match bird 1. In this case, two factors known to be relevant to song type choice will favor the use of song type B: The normal singing style or program of the bird (its ‘‘preferred’’ song type succession) and the auditory input would both favor production of song type B. Birds may even stimulate matching by singing songs for which the probability of being matched by a specific singer is higher than that for other songs or for other singers. The functional significance of such very subtle ways of interacting is unclear. In this case, being matched could be perceived as a positive signal rather than an aggressive signal, as is often implied (see Section C). Here, matching could also result from auditory facilitation. Specifically, it would be interesting to know if birds have different expectations of being matched depending of which song type they sing and with which singer they are interacting.

VOCAL INTERACTIONS IN BIRDS

271

Vocal supplementing can be striking when birds are confronted with playback of their own song (Todt, 1971a), but it can also be observed during natural interactions. Although nightingales only rarely sing song types in a strictly consistent order, some song types always have a high probability of being followed within a short sequential interval by one to several specific song types (Todt and Hultsch, 1996; Fichtel and Todt, 1998). This clustering of song types is typical for many thrushes and is described in detail by Todt and Hultsch (1996, 1999). When such individauls interact they can also obtain information on the way in which the shared song types are clustered. Our studies on vocal interactions in nightingales gained a lot from investigations that concentrated on the mechanisms controlling song performance (Todt, 1971a, Hultsch, 1980; Hultsch and Todt, 1996; Riebel and Todt, 1997) and on song acquisition and song development (Todt et al., 1979; Hultsch and Todt, 1989a,b, 1992, 1996; Hultsch, 1985, 1993; Todt and Bo¨hner, 1994: Todt and Hultsch, 1996, 1998; Hultsch et al., 1999a,b). These studies showed, for instance, that song type clusters can be a consequence of early processes of song acquisition. Here, small clusters are a result of a so-called package formation, whereas large clusters consist of song types that are acquired from the same learning context. Thus, not only sharing of vocal repertoires but also sharing of sequential associations among song types may be mediated by learning. This conclusion has interesting implications for an interpretation of the various vocal interactions described previously. 6. Singers That Deliver Songs with Eventual Variety Many song birds repeat the same song type or pattern several times before they switch to another pattern or type. This repetitive singing is commonly referred to as bout singing or singing with eventual variety (Figs. 2c and 2d). Song matching in these species has a different kind of complexity compared to that in singers that deliver their repertoire with immediate variety. Here, singers that match each other can become locked into bouts in which both individuals repeatedly alternate with the same song type (Nielsen and Vehrencamp, 1995). These bouts end as soon as one individual switches to a different song type. Such a switch might be followed by the other bird so that a new bout of matched countersinging results. Numerous studies have investigated matching in species with such a singing style, such as Abyssinian ground thrushes (Geokichla piagiae; Todt, 1971c), blackcapped chickadees (Parus atricapillus; Shackelton and Ratcliffe, 1994), cardinals (R. cardinalis; Lemon, 1974), Carolina wrens (Thryothorus ludovicianus; Simpson, 1985), chaffinches (F. coelebs; Slater, 1981), great tits (Krebs et al., 1981; Falls et al., 1982; Weary et al., 1992; McGregor et al., 1992), Heuglin’s robin chats (Cossypha heuglini; Todt et al., 1981), plain

272

DIETMAR TODT AND MARC NAGUIB

titmice (Parus inornatus; Dixon, 1969), and calls of red-winged blackbirds (Beletsky et al., 1986), song sparrows (Kramer and Lemon, 1983; Kramer et al., 1985; Stoddard et al., 1992; Nielsen and Vehrencamp, 1995), tufted titmice (Parus bicolor; Lemon, 1968; Schroeder and Wiley, 1983a,b), and western meadowlarks (Sturnella neglecta; Falls, 1985; Falls and Krebs, 1975; Falls and d’Agincourt, 1982; Horn and Falls, 1986, 1988a,b; Falls et al., 1988). Once a bird has switched to a matching song, it can be difficult to determine if it responded only by switching but then ceased to respond and just continued ‘‘as usual’’ with the bout it had just started. However, experiments suggest that most birds continue to respond after switching in a similar way to that which we described for blackbirds and nightingales. The probability that western meadowlarks would switch to a song of the same type as a playback song depended on how many songs they had already sung in their current bout (Falls and d’Agincourt, 1982). If they had just started a new bout before the playback began, the probability of an immediate switch was rather low. In contrast, when they were further into a bout, the probability of switching to a song of the same type as the playback song was much higher. Thus, despite the striking difference in the delivery of the repertoire between bout singers and variable singers, it is clear that in both cases song type choice depends on intrinsic rules (here, the number of songs sung within a bout) and on external stimuli. Most bout singers have small to medium-sized song repertoires. Thus, the probability that two singers concurrently sing the same song type is high, depending on their degree of repertoire sharing. Playbacks with western meadowlarks (Falls, 1985) and great tits (Falls et al., 1982) showed that matching depends on the similarity of the received song to a song in the bird’s own vocal repertoire and on the familiarity with the interacting partner. Subjects were more likely to match playback songs when these songs were more similar to their own songs. The highest matching rates were obtained in playbacks with their own songs. Interestingly, both species matched neighbors’ songs less frequently than strangers’ songs. This difference in matching was interpreted as strangers being a stronger threat and, thus, one that required more attention. Stoddard et al. (1992) found similar results in song sparrows and argued that birds might match a rival when they are uncertain about the rival’s identity. This would explain low matching rates among neighbors and medium matching rates to self songs that are familiar in structure but less familiar if heard from a distant location. Strangers are more likely to sing unfamiliar song patterns and to sing from unfamiliar locations. However, this argument may not be generalized. Singers often continue to match each other on specific occasions, even when they are established neighbors.

VOCAL INTERACTIONS IN BIRDS

273

This difference exemplifies the diversity of contexts in which matching occurs and the diversity of interpretations that may result. Most studies on song matching in bout singers have argued that matching is a threat signal. Krebs et al. (1981) suggested on the basis of experiments with great tits that the probability with which a singer matches its counterpart indicates its readiness to attack. In contrast, Nielsen and Vehrencamp (1995) concluded from experiments with song sparrows that prolonged song matching reflects a low level of threat but serves to ‘‘probe and reveal the identities and relative status of the two birds.’’ In summary, there is ample evidence that matching rate increases with the intensity of the interaction. However, the frequent occurrence of matching outside agonistic contexts indicates that it is also used to signal other kinds of information. Clearly, these different interpretations are not mutually exclusive. The variety in findings and interpretations provides good evidence for our argument made at the outset that matching is not a simple response but rather leads to a highly complex form of signaling. Its occurrence and meaning also seem to strongly depend on the species and the context, as discussed in more detail in Section II,E. In addition to song matching and vocal supplementing, singing interactions in bout singing species provide another pattern-specific response feature. Because birds sing the same song type repeatedly before switching to another type (Fig. 1c), they can respond to each other by synchronizing their switches. Thus, switching between song bouts can be synchronized without the birds singing the same song type. A large body of research has shown that song switching, independent of song matching, is a common feature during vocal interactions (Todt, 1971c; Falls and Krebs, 1975; Kroodsma and Verner, 1979; Smith and Reid, 1979; Yasukawa, 1981; Falls and d’Agincourt, 1982; Kramer and Lemon, 1983; Kramer et al., 1985; Horn and Falls, 1986, 1988). Synchronization of switches is usually interpreted in an equivalent way to song matching. By doing so, birds can address other singers. The occurrence of synchronized switching has been found to increase during close-range agonistic interactions in a similar way to that described for song matching. High switching rates often correlate with the intensity of disputes, and it has been suggested that they indicate the readiness to escalate a contest (Kramer and Lemon, 1983; Nielsen and Vehrencamp, 1995). In addition, which of the two singers switches first seems to play a role. Birds may take leader and follower roles during interactions in which one singer tends to switch first and the other tends to switch thereafter (Horn and Falls, 1988). These leader–follower relations could signal a difference in ‘‘dominance’’ in a way similar to that suggested for song matching in marsh wrens by Kroodsma (1979).

274

DIETMAR TODT AND MARC NAGUIB

C. Pattern- and Time-Specific Interactions Because interactions involve both time- and pattern-specific responses, these cannot be fully separated when the dynamics of interactions are studied (Figs. 1 and 4). However, few studies have considered both timeand pattern-specific responses. The first evidence of a relationship between matching and a particular timing of this response was documented for blackbirds (Todt, 1969b, 1970a), and this finding stimulated a series of experiments in other species, such as nightingales (Hultsch, 1980) and western meadowlarks (Falls, 1985). The rationale of these studies was to use song matching as a tool to obtain insights into the song control system of birds. Detailed investigations in blackbirds and nightingales showed that these birds matched a perceived vocal pattern preferentially either after a short latency (range: 0.4–1.2 s) or after a longer latency (Todt, 1970a, 1971a, 1974, 1975a, 1981; Hultsch, 1980; Wolffgramm and Todt, 1982). When choosing a short latency, birds matched the first part of a heard song and thus overlapped the following parts of the song (rapid matching). Due to the song structure of these species (see Section II,A,2), the stimulus song and the matching reply could still differ in their second part. In the case of a longer latency, however, the response was given after the stimulus song had ended, and thus it resulted in alternating singing (delayed matching). In the latter case, the response was usually a complete match of the stimulus pattern. On the basis of these results, Todt (1981) hypothesized that matching serves to address a particular neighbor, but only the timing encodes a more specific message. Rapid matching of a nearby neighbor, for example, was interpreted as an addressed keep-out signal or a ‘‘vocal threat,’’ respectively, whereas delayed matching was considered to be more friendly and to serve as a kind of ‘‘vocal greeting.’’ This explanation was founded on both the proximate consequences of the responses among close neighbors and features related to the execution of a matching response (Hultsch and Todt, 1986). Rapid matching seemed to be associated with high arousal. Delayed matching, in contrast, occurred when birds seemed to be more relaxed (Todt and Hultsch, 1994, 1996). Since rapid matching clearly requires a remarkable vocal competence, such as the ability to instantly identify a heard song and to immediately retrieve and produce an adequate response, this type of matching was suggested to have an impact on individuals other than the addressed neighbor. This assumption is currently being tested. D. Song Matching and Ranging Several authors have suggested a causal relation between song matching and judging the singer’s distance (ranging); (Krebs et al., 1981; Falls et al.,

VOCAL INTERACTIONS IN BIRDS

275

1982; Morton, 1982; Falls, 1985; Horn and Falls, 1988a). The argument was derived from the early assumption by Morton (1982) that ranging requires a singer to have the perceived song type in its own vocal repertoire. According to this idea, a bird that matches the song of a rival does so in order to provide the rival with accurate information about its distance from the rival. Concurrently, by matching, a singer would indicate to its counterpart that it can also range its counterpart’s location. Thus, it has been suggested that birds that match a song might do so specifically to exchange information on their distance to each other (Krebs et al., 1981; Falls et al., 1982; McGregor and Falls, 1984; McGregor, 1991, 1994). Although this argument is interesting, it has not received experimental support (Naguib, 1998; Wiley, 1998). Several studies have shown that birds can range songs that they do not themselves sing (McGregor et al., 1983; McGregor and Falls, 1984; McGregor and Krebs, 1984). It has also been shown that birds require little or no experience with a song type they hear in order to approximate the singer’s distance (Wiley and Godard, 1996; Naguib, 1996b, 1997a; Philmore et al., 1998). Furthermore, the mechanisms underlying assessment of a signal’s degradation do not explicitly require prior knowledge of the specific song type (Naguib, 1995, 1996a, 1997a,b; Wiley and Godard, 1996; Philmore et al., 1998). Thus, experimental evidence does not suggest that ranging plays a significant role in song matching, unless matching enhances accuracy of ranging on a level not yet assessed. So far, few studies have attempted to measure directly the accuracy of ranging (e.g., Nelson and Stoddard, 1998; Naguib et al., in press). In addition, as we have shown for species that sing with immediate variety, matching occurs at specific instances during singing bouts. These bouts are often delivered without changing song post. All features of the interactions suggest that information on the distance of a counterpart is not restricted to matching events. In addition, in some populations, singers might share only a low proportion of their repertoires (Hughes et al., 1998), whereas in other populations of the same species a high degree of repertoire sharing is common (Beecher, 1996). It seems unlikely that selection favors a mechanism for ranging that would result in some individuals, populations, or even species not being able to range each other. This situation, however, is the logical consequence of arguments central to Morton’s (1982) ranging hypothesis. Thus, current evidence does not support the idea that song matching is causally linked to sending distance information. In light of the diverging results obtained in ranging experiments in different species, it is certainly possible that knowledge of the details of undegraded songs might be of varying importance for the accuracy of ranging among species (Naguib, 1996b, 1997a). However, song matching is a ubiquitous feature in interactions in a wide range of species and, as a general phenomenon, it oc-

276

DIETMAR TODT AND MARC NAGUIB

curs in visual displays and is also a conspicuous feature of human vocal and nonvocal behavior. Thus, we must expect that selective forces other than ranging have been more important for the evolution of matching behavior. E. Summary of Functional Aspects In the previous sections, we discussed examples showing where, when, and how singers interact with each other in time- and pattern-specific ways. Here, we will summarize the functional aspects of interactions and complement them by discussing evolutionary issues. Among the different ways of timing signals during an interaction, alternating singing is the most common and permits mutual listening and responding. It avoids signal masking and thus increases the detectability of the signal. Asymmetries in the timing of songs between the singers are common and reflect differences in state, such as motivation or arousal, or social status. Overlapping appears to serve to challenge and repel a competitor, often by jamming the other bird’s signals. Overlapping can increase the singers’ relative signal time and space when singers that are overlapped interrupt their songs. However, over long distances masking effects will be low. Overlapping might also indicate a sort of ignorance or lack of interest in the counterpart’s song or a level of agonistic tendencies that is not based on effects of signal masking. Autonomous singing, in which songs are not timed specifically in relation to those of the counterpart, could be a strategy of territorially well-established males that have less need to interact with their neighbors or lack interest in doing so (Hultsch and Todt, 1982). However, it is possible that autonomous singing conveys specific messages that have not yet been assessed. Of the several functions proposed for song matching, only one is commonly acknowledged to be relevant for all species studied: Matching the song of a counterpart addresses the response to that individual. There is widespread evidence that birds match songs of their counterparts during agonistic and more or less relaxed vocal interactions. Some other authors, however, interpreted song matching as an agonistic signal. Their argument has been derived from increased rates of song matching found during intense disputes. High rates of matching presumably signal high arousal and a readiness to escalate a contest. Low rates of matching, in contrast, might have the opposite effect. However, although during an agonistic interaction song matching might be used as a signal to direct a response, it is not necessarily an agonistic signal per se. Kroodsma (1979) suggested that song matching may reflect dominance– subordinance relationships. A more specific interpretation was given by

VOCAL INTERACTIONS IN BIRDS

277

Hultsch and Todt (1986), who argued that the particular meaning of matching is encoded in the timing of this response. As mentioned in Section II,B,5, they postulated a keep-out message for rapid matching but a form of ‘‘vocal greeting’’ for delayed matching. On the other hand, the signal value of song matching might also be different when birds interact at close range during boundary disputes from when they interact over long distance. Our observation that matching may also result in songs being passed along a series of territories, support this argument. Thus, in some contexts, matching might be an important mechanism to establish a specific relation to specific neighbors. Finally, matching might be used to probe the neighbor’s state or status or to obtain information on its kind of vocal responses. Matching might also function as a signal to other individuals that are not involved directly in the interaction. Established neighbors are expected in certain contexts to have a mutual interest in preventing strangers from settling too close (Axelrod and Hamilton, 1981; Hultsch and Todt, 1981; Godard, 1993). One way to do this is to signal a strong dyad, in which the singers take turns in matching each other. This argument is not trivial, at least not for species such as nightingales in which the song repertoire is organized in such a way that we assume that not every song can be accessed spontaneously. Individuals that match each other frequently and take turns doing so can possibly do this only after having had sufficient experience with each other. In this case, high matching rates would signal that the singers have had a long, well-established relationship. Matching could have an aversive effect on outsiders. That this argument has not been applied to vocal interactions in birds is surprising insofar as it is thought to be an important factor in relations between members within and outside a group of animals or humans. Effects of vocal learning (Todt and Hultsch, 1996, 1999) and convergence of acoustic signals among group members in flocks of passerines are examples for this argument (Nowicki, 1989). Matching is even a ubiquitous mechanism in a wide range of other nonvocal social behaviors. Human society is full of such examples. In summary, these very different hypotheses on song matching reflect the high degree of functional complexity that can result from one simple mechanism—reply with the same signal one hears. However, a crucial question remains unanswered: What are the consequences of song matching and asymmetries in the timing of songs, such as overlapping or asymmetric alternating, with respect to the outcome of an interaction? In fact, no study has shown that song overlapping or high rates of song matching during a natural interaction increase the probability of winning a dispute over a resource. Are overlappers and individuals that match in certain ways more attractive to females? Do they have better or larger territories? Are their songs learned preferentially by fledglings? Does it contribute to an individu-

278

DIETMAR TODT AND MARC NAGUIB

al’s fitness? Studies on anurans and insects have shown that different calling strategies are highly relevant in female choice (Klump and Gerhardt, 1982; Grafe, 1996). We expect that the situation is more complex in songbirds, in which the use of song during an interaction at the most is just one aspect of female choice; it will surely be exciting what future research will reveal. F. Duets Duets are striking examples of time- and pattern-specific coordination of vocally interacting individuals. The precision, as a rule, is much higher than that in the interactions among territorial males that we have described so far. The vocal interactions occur mostly between paired individuals. These distinctly elaborated performances are commonly termed antiphonal duets (Thorpe and North, 1965; Todt, 1970c, 1975b; Thorpe, 1972; Wiley and Wiley, 1977; Todt and Fiebelkorn, 1979; von Helversen, 1980; Todt et al., 1987; Farabaugh, 1982). Duetting includes a variety of vocal styles of singing or calling (Price, 1998, 1999) that are developed especially between the members of a pair or group. In this section we will concentrate on duetting by song, which provides an appropriate basis for a formal comparison to the vocal duels described previously. Vocal duets between members of a pair may vary in the different types of motifs that mates mutually perform or in the temporal relationships between duet contributions. The flexibility depends on the context and presumably on the state of the signalers. In this section, we will not review the vast literature on this topic but rather exemplify principles of duetting behavior from studies on Heuglin’s robin chats (C. heuglini; Todt, 1971b; Wickler, 1974; Hultsch, 1983; Fig. 6) and Hunter’s warblers (Cisticola hunteri; Todt 1970c; Fig. 7). We selected these species because they provide a good comparison. The robin chats especially have been studied in terms of more different aspects than any other duetting songbird. Studies included analyses of song development (Todt et al., 1981) and experiments on both pair bond persistence (Todt, 1975b) and the role of visual stimuli in the field (Hultsch and Todt, 1984). Additional experiments with physiological treatments, such as deafening and hypoglossus sectioning, showed that constraints in hearing the mate or in producing the normal duet contributions did not impair the pair bond (Todt and Hultsch, 1982). Finally, there were neurobiological inquiries also into the song control system of this species (Brenowitz et al., 1985; Brenowitz and Arnold, 1989). Duetting species are found in various species of different families (Kunkel, 1974; Farabaugh, 1982). Interestingly, these species share several biological characteristics. With some exceptions, they are restricted to tropical and subtropical habitats. Most of them have prolonged pair bonds and are territorial

VOCAL INTERACTIONS IN BIRDS

279

Fig. 6. Duet of Heuglin’s robin chat. Solid element symbols refer to vocal patterns of the male; open element symbols refer to vocalizations of the female. Female vocalizations can be escorted by wing beats that seem to stimulate the male. Position and amplitude of wing beats are symbolized by triangles (top).

throughout the year, and many of them are cooperative breeders. Their duetting behavior can be subdivided into the following basic categories: 1. Duet contributions are not exchangeable among members of the pair: Each mate contributes a specific part to a duet (for illustrations see Figs. 6 and 7). Often, these different parts are distinctly sex specific and birds do not perform the duet part performed by their partners. Thus, this signaling has been interpreted as an acoustic sexual dimorphism. Occasionally, one mate may utter its particular vocal pattern when singing alone.

Fig. 7. Duet and counterduet of the Hunter’s warbler. Open symbols are male and female motifs of one pair. Solid symbols show motifs of the other pair. Patterns within and between pairs are highly coordinated. Females use these to initiate duets.

280

DIETMAR TODT AND MARC NAGUIB

2. Duet contributions are exchangeable between the members of a pair: The basic sound pattern of a duet is segmented into successive sequential units (e.g., syllables and motifs) that the mates mostly vocalize in an alternating manner. Normally, one mate sings the initiating duet part, whereas birds may change roles later in the duet. In exceptional cases, a bird may perform the whole sequence alone. Thus, this duet category shows striking similarities to the vocal supplementing described previously for interactions between neighboring nightingales. The majority of studies have documented vocal duets that belong to the first category (Farabaugh, 1982). Duets of the second category seem to be less common in songbird species but can occur, for example, in parrots (Todt, 1975c). However, although duet contributions are often gender specific, males of at least some species are able to mimic the ‘‘female’’ part in certain contexts. Hultsch (1983) showed in Heuglin’s robin chats that adult males are able to perform both parts of an antiphonal duet. One male may sing the male part and the other male (normally a familiar, subdominant individual) will take the female part. Heuglin’s robin chats can also perform a kind of pseudoduet, in which only one male sings the whole sequence. Findings such as this support Thorpe’s (1972) argument that antiphonal duets, despite their normal structural perfection and precise coordination, are flexible vocal displays. An interesting kind of duet initiation has been documented for Heuglin’s robin chats (Todt et al., 1981; Hultsch, 1983). Here, the male may be singing alone and then gradually change the tempo, the pitch, and the sound pressure level of its successively repeated motifs. Such changes presumably depend on his motivational state. High territorial and agonistic motivation were positively associated with a rapid and strong increase in the pitch and peak amplitude of motifs. The female then appeared to join in at specific parameter combinations (pitch, amplitude, and duration of intervals between patterns) during the male’s performance (Todt et al., 1981). Once these duets are initiated, they may continue for several seconds or less. In the latter case, their duration matches what is typical for the songs of territorial signalers. It has also been shown that parameters of the vocal duet change with other behavioral and contextual features (Todt et al., 1981). In Heuglin’s robin chats, duets that were both ‘‘well-coordinated and short’’ occurred after reuniting of previously separated mates, for instance, in the early morning. Normally, duets are performed within dense vegetation and may function as greeting ceremonies. Duets that are ‘‘wellcoordinated and long’’ are always displayed on exposed perches and, as a rule, are accompanied by rhythmical body and wing movements (Todt et al., 1981). Obviously, these duets serve territorial advertisement. A third

VOCAL INTERACTIONS IN BIRDS

281

group of duets, which are ‘‘badly coordinated and short,’’ occur only in contests with territorial intruders. During territorial contests between pairs of robin chats, a series of short duets usually follow each other in rapid succession. Interspersed are display flights, approaches toward intruders, or silent intervals. Possibly, these intervals are used to watch neighbors or to listen to their vocalizations. Evidence for the latter stems from the specific vocal responses mutually exchanged between pairs. Such counterduets have been described, e.g., for Hunter’s warblers (Todt, 1970c) but also occur in other species (Wiley and Wiley, 1977). In these counterduets, one pair acts essentially as one system and responds to the other one in a way similar to that described for vocal interactions by song among territorial passerines. Counterduets seem to demonstrate that a territory is occupied by an alliance of two individuals. Such an alliance, however, has to be backed up by optical signaling, and in this context spatial proximity between mates plays an important role (Hultsch and Todt, 1984). As for the male’s song in nonduetting species, experiments have revealed that vocal duetting is territorially effective only when the message is confirmed from time to time by the pair. Since Thorpe (1972) stimulated the study of duetting almost 30 years ago, different hypotheses have been proposed to explain the functions of this behavior. Tests conducted to examine the phenomenon have revealed that duetting is basically a territorial display, cooperatively performed by mates that are normally in close contact. Nevertheless, within-pair functions may be significant, such as in terms of synchronization and stimulation of reproductive behavior (Todt and Hultsch, 1982; Levin, 1996a,b). III. Interactions with More Than Two Individuals A. Dawn Chorus The dawn chorus provides an outstanding situation for vocal interactions. The concurrent high singing rates of most conspecific and many heterospecific males results in an enormously complex acoustic scenario. Staicer et al. (1996) provided an extensive overview of the various hypotheses put forward to explain the dawn chorus. In general, it remains a mystery whether or not males are truly interacting during such noise and complex acoustic conditions. In fact, except for anecdotal observations, it appears to be difficult with the current techniques to study interactions when many individuals sing in a relatively uncoordinated manner. Birds might just be singing at high rates without paying attention to what others sing. They may alternate rapidly between different singing strategies by responding in turns to different singers. Due to this complexity, most researchers that

282

DIETMAR TODT AND MARC NAGUIB

have experimentally investigated singing interactions have avoided the dawn chorus. B. Triads and More So far, we have focused on dyadic interactions between territorial passerines. Often, other individuals within signaling range are also vocalizing. Thus, dyads are not always a closed system as discussed so far. Territories and song posts are sometimes arranged such that three or more singers are at similar distance to each other. Consider three territorial males (A–C), all of which sing at the same time. B might alternate songs with A but concurrently overlap the songs of C, or the different singers may switch back and forth in responding to one or the other of two singers. Such sequential dyads when viewed over sufficient time could have the properties of triadic interactions. Such events are rare and thus difficult to analyze empirically. The same applies to true triadic interactions, in which three birds alternate songs. However, they are striking when they are observed. Despite this rarity, or even because of it, interactions among three or more individuals might be of particular importance for the relations between the individuals involved. Unfortunately, most of what we know of such interactions is anecdotal. Triads are presumably more common among tropical duetting birds than among territorial passerines in the temperate zones. Thorpe and North (1965) mentioned trios for several African duetting species, and Todt (1970c) documented trios and interactions with four participants in Hunter’s warblers, which included cases of vocal matching (Fig. 7). Interactions with four individuals in general seem to be common in duetting species when two pairs interact vocally during boundary disputes (Wiley and Wiley, 1977; Hultsch and Todt, 1984).

IV. Relevance for Other Listeners Obviously, the loud vocal interactions among territorial birds can often be heard by other conspecifics (Fig. 8). This ability to listen to others’ interactions potentially allows the listeners to eavesdrop on the information exchanged. Although this possibility has been addressed for a long time (Otte, 1974; Todt, 1981), it has not received more specific attention until recently (Endler, 1992; McGregor and Dabelsteen, 1996; Naguib and Todt, 1997; Naguib et al., 1999; Otter et al., 1999). Thus, instead of viewing communication as occurring only within dyads, we have to expect information to flow beyond the dyad, possibly in a web-

VOCAL INTERACTIONS IN BIRDS

283

Fig. 8. Signal space of two territorial birds that are interacting (birds 1 and 2). Circles indicate the active space of each of the singers. Some conspecifics (䊉) will hear both. This allows them to listen to the interaction and to potentially extract information from it. Other conspecifics (䊊) can only hear one of the singers and thus cannot obtain as much information from the interaction. Some may also be interacting with other males (dashed arrows).

or network-like manner. In their study on different singing strategies during vocal interactions in nightingales, Hultsch and Todt (1982) concluded ‘‘we guess that they [performance roles in terms of timing of songs] may reflect a social network in which individuals differ in terms of state or status.’’ Stalcer et al., (1996), in their review on the dawn chorus, derived a similar conclusion: ‘‘The social relations among territorial birds can be visualized as a complex and dynamic web.’’ In their chapter devoted specifically to eavesdropping, McGregor and Dabelsteen (1996) were more iconic when arguing ‘‘for this reason, communication using long-range signals is best considered as occurring within a network of signalers and receivers.’’ The theoretical advantages of attending to others’ interactions are straightforward. A listener can obtain information at low cost without investing time and energy by participating. As we have shown, these asymmetries in vocal interactions might reflect differences in motivation or even differences in quality or strength of the interacting individuals. Such information on differences among individuals is particularly valuable for

284

DIETMAR TODT AND MARC NAGUIB

an individual’s decisions on future actions. Females may choose males on relative criteria and not on absolute criteria. Males that expect to interact with any one of the conspecifics to which they are listening can obtain information on their relative state or competitive abilities. Relative differences among individuals presumably can be perceived more reliably (depending on the kind of display) when these individuals interact with each other than when they signal separately. In fact, there is evidence that songbirds extract and use information obtained by listening to other’s interactions. In a series of playback experiments with nightingales we investigated whether or not subjects would use information from asymmetries in the vocal interactions of conspecifics. We used two loudspeakers to simulate an interaction between two males appearing close to a subject’s territorial boundary (Fig. 9). We expected that if territorial males would perceive and use information provided by asymmetries in the rival’s timing of song, they would react differently with respect to the two loudspeakers. Like other territorial passerines, male nightingales approach intruders and their simulations by playback during daytime. In the first experiment, we simulated two rivals: One singer was overlapping the songs of the other singer (Naguib and Todt, 1997). Thus, the loudspeaker that broadcast the overlapping songs was intended to simulate a more aggressive or threatening intruder. Subjects as a group

Fig. 9. Symbolized succession of songs on the two channels in three playback experiments: (a) symmetric alteration of songs, (b) asymmetric alternation of songs (Naguib et al., 1999), and (c) song overlapping (Naguib and Todt, 1998). Open bars represent songs on channel 1 and hatched bars represent songs on channel 2, respectively. Song length was variable.

VOCAL INTERACTIONS IN BIRDS

285

responded more intensely at the loudspeaker that broadcast the overlapping songs. This experiment indicated that nightingales perceived the simulated asymmetries and responded more strongly to the presumably more salient rival. In a subsequent experiment, we investigated if such asymmetries also result in discrimination when birds do not overlap each other (Naguib et al., 1999a). We used the same setup, but this time songs of the two singers alternated, either asymmetrically or symmetrically. In the asymmetric design, songs of one singer followed those of the other singer after shorter latencies than vice versa. All except 1 of the 17 territorial subjects started to sing their first bout at the loudspeaker that played the leading songs, and subjects responded significantly more at the loudspeaker that broadcast the leading songs. In contrast, in simulated symmetric interactions subjects did not discriminate between the two loudspeakers (Naguib et al., 1999a). These experiments indicate that asymmetries in dyadic interactions can affect responses by third individuals. Interestingly, responses in the two experiments were in opposite directions, depending on the kind of asymmetry simulated—either to the overlapper (that always had the ‘‘last word’’) or to the leader (that always had the ‘‘first word’’). These context-dependent differences in response that in reaction to preceding or nonpreceding songs (and also following overlapping songs) suggest that the responses are not an incidental consequence of attention to the first- or last-heard stimulus but more likely reflect adaptive strategies. This argument is important to consider because attention to a last-heard stimulus presumably is under direct selection in other contexts, such as during detection of prey and avoidance of predators. Thus, we assume that responses either to the last or to the first songs are functionally relevant for the context studied. Importantly, such listening to relative differences among potential competitors or mates can provide important information that cannot be obtained as easily when the interacting individuals are sampled separately.

V. Final Remarks By reviewing and discussing a wide range of studies on vocal interaction by song in passerines, we had two goals: (i) compiling information on the complexity of vocal interactions within and between species by summarizing and reviewing a range of different studies and (ii) emphasizing the importance of addressing the mechanisms underlying the choice of signal patterns and the functional consequences of different signaling strategies. Future research, in particular on functional issues concerning long-distance vocal interactions, will be challenging. These interactions are often not followed immediately by other kinds of interactions so that there will be no immedi-

286

DIETMAR TODT AND MARC NAGUIB

ate ‘‘loser’’ or ‘‘winner,’’ and thus it is more challenging to link the vocal behavior to the outcome of a contest. From this perspective, analyzing functional consequences of singing interactions in songbirds is more complex than in similar studies conducted on insects or anurans. In these taxa, signaling strategies can be linked much more directly to female choice and thus reproductive success (Klump and Gerhardt, 1992; Ryan, 1997). In addition, interactions are particularly intriguing when, over time, individuals develop relationships the properties of which may involve a wide range of different kinds of responses, including very subtle ones. The kind of interaction among neighbors might change as the relationship between them develops. Thus, the way they interact can reflect their relationship on a more general level. Understanding the specific ontogenetic trajectories for each dyad may open an important window for understanding how a bird uses its song during an interaction with a specific individual (Wiley, 1981; Pepperberg, 1991; Todt and Hultsch, 1996). Interactions by song invite a comparative approach that includes contrasting them formally with other kinds of nonverbal dialogues. Dialogues are interaction processes par excellence (Todt and Hultsch, 1994, 1999) and are distinguished by the following features: (i) Each contributor has more than one choice of response and (ii) any signal used results from a decision influenced by two kinds of variables: external components, such as perceived acoustic patterns, and internal components, such as those underlying a signaler’s singing style or singing program (Fig. 1). In other words, neither stereotyped responding nor fixed signaling meet the dialogue criteria. Finally, there should also be a temporal change in roles that can be described as ‘‘turn-taking.’’ This chapter documented many time- and pattern-specific features occurring in avian vocal interactions. Referring to these features and to the criteria mentioned previously, we conclude that some of the vocal duels (i.e., the vocal interactions among male territorial passerines) fulfill the formal criteria of a dialogue, whereas others, such as stereotyped antiphonal duets, do not. On this basis, it seems essential to ask whether or not and how other properties of a dialogue, such as the messages and meanings involved, can be addressed during a series of exchanged songs. To date, questions on the message and meaning of signals have been addressed most commonly by associating the signals with the context of the communication (Smith, 1965, 1991). A dialogue perspective complements this approach by also deriving the message and meaning of signals from the relation of the exchanged signals to each other. Currently, investigators of mammal vocal communication are also concentrating on this issue, for instance, for cetaceans (Tyack, 1996) and various species of nonhuman primates (Maurus et al., 1988; Snowdon, 1988). It will be exciting to see how such comparative

VOCAL INTERACTIONS IN BIRDS

287

ideas will influence the development of our understanding of vocal communication in animals.

VI. Summary The aim of this chapter was to review results on forms and functions of vocal interactions in territorial songbirds. The framework adopted was from an approach that considers vocal interactions in passerines as a dialogue. Here, participants have more than one option of responding, and the meaning of signals depends on the relation of signals to each other. Within this framework, we discussed current knowledge about rules of song composition and rules of interaction by song. We focused on time-specific and pattern-specific relationships between mutually exchanged signal patterns. The proximate causes and consequences of such relationships were discussed and we dealt with functional aspects of selected response categories, such as vocal matching, song type switching, and song overlapping. In addition, this chapter supplemented the paradigm of dyadic interactions by a model that addresses polyadic interactions and the role of vocal interactions in a community of territorial birds. We hope that this review contributes to an advanced understanding of how and why birds develop and use a sophisticated system of rules when interacting by song.

Acknowledgments We thank Henrike Hultsch, Roger Mundry, and Joe Waas for many stimulating discussions and Henrike Hultsch; Peter Slater, Charles T. Snowdon, and R. Haven Wiley for helpful comments on the manuscript.

References Adhikerana, A. S., and Slater, P. J. B. (1993). Singing interactions in coal tits, Parus ater: An experimental approach. Anim. Behav. 46, 1205–1211. Axelrod, R., and Hamilton, W. D. (1981). The evolution of cooperation. Science 211, 1390– 1396. Becker, H. P. (1982). The coding of species-specific characteristics in bird sounds. In ‘‘Acoustic Communication in Birds’’ (D. E. Kroodsma and E. H. Miller, Eds.), pp. 213–252. Academic Press, New York. Beecher, M. D. (1996). Birdsong learning in the laboratory and field. In ‘‘Ecology and Evolution in Acoustic Communication in Birds’’ (D. E. Kroodsma and E. H. Miller, Eds.), pp. 61–78. Cornell Univ. Press, Ithaca, NY. Beecher, M. D., Stoddard, P. K., Campbell, S. E., and Horning, C. L. (1996). Repertoire matching between neighboring song sparrows. Anim. Behav. 51, 917–923.

288

DIETMAR TODT AND MARC NAGUIB

Beletsky, L. D., Higgins, B. J., and Orians, G. H. (1986). Communication by changing signals: Call switching in red-winged blackbirds. Behav. Ecol. Sociobiol. 18, 221–229. Bertram, B. (1970). The vocal behaviour of the Indian hill mynah. Anim. Behav. Monogr. 3, 81–192. Bre´mond, J. C. (1968). Recherche sur la se´mantique et les e´le´mentes vecteurs d’information dans les signaux acoustiques du rouge-gorge (Erithacus rubecula). Terre Vie 2, 109–220. Bremowitz, E. A., A. P. Arnold, and R. Levin. 1985. Neural correlates of female song in tropical duetting birds. Brain Res. 342:104–112. [114, 291, 296] Bremowitz, E. A., A. P. Arnold, 1989. Accumulation of estrogen in a vocal control brain region of a duetting song bird. Brain Res. 480:119–125. [294, 296] Brindley, E. L. (1991). Response of European robins to playback of song: Neighbour recognition and overlapping. Anim. Behav. 41, 503–512. Capp, M. S. (1992). Tests of the function of the song repertoire in bobolinks. Condor 94, 468–479. Catchpole, C. K. (1976). Temporal and sequential organisation of song in the sedge warbler (Acrocephalus schoenobaenus). Behaviour 59, 226–246. Catchpole, C. K. (1980). Sexual selection and the evolution of complex song among European warblers of the genus Acrocephalus. Behaviour 74, 149–166. Catchpole, C. K., and Slater, P. J. B. (1995). ‘‘Birds Song: Biological Themes and Variations.’’ Cambridge Univ. Press, New York. Chaiken, M. (1990). The ontogeny of antiphonal calling in European starlings. Developmental Psychology 23, 233–246. Cody, M. L., and Brown, J. H. (1969). Song asynchrony in neighbouring bird species. Nature 222, 778–780. Dabelsteen, T. (1992). Interactive playback: A finely tuned response. In ‘‘Playback Studies of Animal Communication’’ (P. K. McGregor, Ed.), pp. 97–109. Plenum, New York. Dabelsteen, T., and Pedersen, S. B. (1990). Song and information about aggressive responses of blackbirds, Turdus merula: Evidence from interactive playback experiments with territory owners. Anim. Behav. 40, 1158–1168. Dabelsteen, T., McGregor, P. K., Shephard, M., Whittaker, X., and Pedersen, S. B. (1996). Is the signal value of overlapping different from that of alternating during matched singing in great tits? J. Avian Biol. 27, 189–194. Dabelsteen, T., McGregor, P. K., Holland, J., Tobias, J. A., and Pedersen, S. B. (1997). The signal function of overlapping singing in male robins. Anim. Behav. 53, 249–256. d’Agincourt, L. G., and Falls, J. B. (1983). Variation of repertoire use in the eastern meadowlark, Sturnella magna. Can. J. Zool. 61, 1086–1093. Darwin, C. (1859). ‘‘On the Origin of Species by Means of Natural Selection.’’ Murray, London. Dixon, K. L. (1969). Patterns of singing in a population of the plain titmouse. Condor 71, 94–101. Dufty, A. M., and Pugh, J. K. (1994). Response of male brown-headed cowbirds to broadcast of complete or partial flight whistles. Auk 111, 734–739. Endler, J. A. (1992). Some general comments on the evolution and design of animal communication systems. Philos. Trans. R. Soc. London B 340, 215–225. Falls, J. B. (1985). Song matching in western meadowlarks. Can. J. Zool. 63, 2520–2524. Falls, J. B., and d’Agincourt, L. G. (1982). Why do meadowlarks switch song types? Can. J. Zool. 60, 3400–3408. Falls, J. B., and Krebs, J. R. (1975). Sequence of songs in repertoires of western meadowlarks (Sturnella neglecta). J. Can. Zool. 53, 1165–1178. Falls, J. B., Krebs, J. R., and McGregor, P. K. (1982). Song matching in the great tit (Parus major): The effect of similarity and familiarity. Anim. Behav. 30, 997–1009.

VOCAL INTERACTIONS IN BIRDS

289

Falls, J. B., Horn, A. G., and Dickinson, T. E. (1988). How western meadowlarks classify their songs: Evidence from song matching. Anim. Behav. 36, 579–585. Farabaugh, S. M. (1982). The ecological and social significance of duetting. In ‘‘Acoustic Communication in Birds’’ (D. E. Kroodsma and E. H. Miller, Eds.), pp. 85–124. Academic Press, New York. Fichtel, C., and Todt, D. (1998). The singing of wild birds can reflect properties of song learning and memory. Ostrich 69, 249–250. Ficken, R. W., Ficken, M. S., and Hailman, J. P. (1974). Temporal pattern shifts to avoid acoustic interference in singing birds. Science 183, 762–763. Ficken, R. W., Popp, J. W., and Matthiae, P. E. (1985). Avoidance of acoustic interference by ovenbirds. Wilson Bull. 97, 569–571. Gerhardt, H. C. (1994). Selective responsiveness to long-range acoustic signals in insects and anurans. Am. Zool. 34, 706–714. Gochfeld, M. (1978). Intraspecific social stimulation and temporal displacement of songs of the lesser skylark, Alauda ququla. Z. Tierpsychol. 48, 337–344. Godard, R. (1993). Tit for tat among neighboring hooded warblers. Behav. Ecol. Sociobiol. 33, 45–50. Grafe, T. U. (1996). The function of call alternation in the African reed frog (Hyperolius marmoratus): Precise call timing prevents auditory masking. Behav. Ecol. Sociobiol. 38, 149–158. Greenfield, M. D. (1993). Acoustic dueling in tarbush grasshoppers: Settlement of territorial contests via alternation of reliable signals. Ethology 95, 309–326. Greenfield, M. D. (1994a). Synchronous and alternating choruses in insects and anurans: Common mechanisms and diverse functions. Am. Zool. 34, 605–615. Greenfield, M. D. (1994b). Cooperation and conflict in the evolution of signal interactions. Annu. Rev. Ecol. Syst. 25, 97–126. Hammerstein, P. (1981). The role of asymmetries in animal contests. Anim. Behav. 29, 193–205. Hasselquist, D., Bensch, S., and von Schantz, T. (1996). Correlation between male song repertoire, extra-pair paternity and offspring in the great reed warbler. Nature 381, 229–232. Highsmith, R. T. (1989). The singing behavior of golden-winged warblers. Wilson Bull. 101, 36–50. Hinde, R. A. (1985). Alternative motor patterns in chaffinch song. Anim. Behav. 6, 211–218. Horn, A. G., and Falls, J. B. (1986). Western meadowlarks switch song types when matched by playback. Anim. Behav. 34, 927–929. Horn, A. G., and Falls, J. B. (1988a). Repertoires and countersinging in western meadowlarks (Sturnella neglecta). Ethology 77, 227–343. Horn, A. G., and Falls, J. B. (1988b). Responses of western meadowlarks, Sturnella neglecta, to song repetition and contrast. Anim. Behav. 36, 291–293. Hughes, M., Nowicki, S., Searcy, W. A., and Peters, S. (1998). Song-type sharing in song sparrows: Implications for repertoire function and song learning. Behav. Ecol. Sociobiol. 42, 437–446. Hultsch, H. (1980). Beziehungen zwischen Struktur, zeitlicher Variabilita¨t und sozialem Einsatz des Gesanges der Nachtigall (Luscinia megarhynchos B.) Dissertation, Fachbereich Biologie, Freie Universita¨t Berlin. Hultsch, H. (1981). Zur tageszeitlichen Variabilita¨t des Vogelgesanges am Beispiel der Nachtigall (Luscinia megarhynchos). Verh. Dtsch. Zool. Ges. 74, 240. Hultsch, H. (1982). Pitch and rhythm matching: Auditory fascilitation and triggering effects of acoustic stimuli. In ‘‘Proceedings of the 17th Congress of International Ornithology,’’ Moscow, p. 1113.

290

DIETMAR TODT AND MARC NAGUIB

Hultsch, H. (1983). Behavioural significance of duet interactions: Cues from antiphonal duetting between males (Cossypha heuglini H.). Behaviour 86, 89–99. Hultsch, H. (1985). Formation of sub-repertoires: A strategy to support development and performance of large vocal repertoires. Verh. Dtsch. Zool. Ges. 78, 229. Hultsch, H. (1993). Psychobiological and ecological aspects of song learning and memorization. Etologia 3, 309–323. Hultsch, H., and Todt, D. (1981). Repertoire sharing and song-post distance in nightingales (Luscinia megarhynchos B.). Behav. Ecol. Sociobiol. 8, 183–188. Hultsch, H., and Todt, D. (1982). Temporal performance roles during vocal interactions in nightingales (Luscinia megarhynchos). Behav. Ecol. Sociobiol. 11, 253–260. Hultsch, H., and Todt, D. (1984). Spatial proximity between allies: A territorial signal tested in the monogamous duet singer Cossypha heuglini. Behaviour 91, 286–293. Hultsch, H., and Todt, D. (1986). Signal matching: Zeichenbildung durch mustergleiches Antworten. Z. Semiotik, 8, 223–244. Hultsch, H., and Todt, D. (1989a). Memorization and reproduction of songs in nightingales (Luscinia megarhynchos): Evidence for package formation. J. Comp. Physiol. A 165, 197–203. Hultsch, H., and Todt, D. (1989b). Context memorization in the song learning of birds. Naturwissenschaften 76, 584–586. Hultsch, H., and Todt, D. (1992). Time window and unit capacity: Dual constraints on the acquisition of serial information in song birds. J. Comp. Physiol. A 170, 275–280. Hultsch, H., and Todt, D. (1996). Rules of parameter variation in homotype series of birdsong can indicate a ‘‘sollwert’’ significance. Behav. Processes 38, 175–182. Hultsch, H., Schleuss, F., and Todt, D. (1999a). Auditory-visual stimulus pairing enhances perceptual learning in a songbird. Anim. Behav. 58, 143–149. Hultsch, H., Mundry, R., and Todt, D. (1999b). Learning, representation, and retrieval of rule-related knowledge in the song system. In learning; A. Friedeorci and R. Mensel (eds.) de Gruyter, Berlin. pp. 89–115. Isaac, D., and Marler, P. (1963). Ordering sequences of singing behavior of mistle thrushes in relationship to timing. Anim. Behav. 11, 179–188. Janik, V. M., and Slater, P. J. B. (1997). Vocal learning in mammals. Adv. Study Behav. 26, 59–99. Klump, G., and Gerhardt, H. C. (1992). Mechanisms and function of call-timing in male–male interactions in frogs. In ‘‘Playback Studies of Animal Communication’’ (P. K. McGregor, Ed.), pp. 153–174. Plenum, New York. Kramer, H. G., and Lemon, R. E. (1983). Dynamics of territorial singing between neighboring song sparrows (Melospiza melodia). Behaviour 85, 198–223. Kramer, H. G., Lemon, R. E., and Morris, M. J. (1985). Song switching and agonistic stimulation in the song sparrow (Melospiza melodia). Anim. Behav. 33, 135–149. Krebs, J. R., and Kroodsma, D. E. (1980). Repertoires and geographical variation in bird song. Adv. Study Behav. 11, 134–177. Krebs, J. R., Ashcroft, R., and van Orsdol, K. (1981). Song matching in the great tit Parus major L. Anim. Behav. 29, 918–921. Kroodsma, D. E. (1979). Vocal duelling among male marsh wrens: Evidence for ritualized expression of dominance/subordinance. Auk 96, 506–515. Kroodsma, D. E. (1982). Song repertoires: Problems in their definition and use. In ‘‘Acoustic Communication in Birds’’ (D. E. Kroodsma and E. H. Miller, Eds.), Vol. 2, pp. 125–146. Academic Press, New York. Kroodsma, D. E., and Verner, J. (1978). Complex singing behaviours among Cistothorus wrens. Auk 95, 703–716.

VOCAL INTERACTIONS IN BIRDS

291

Kunkel, P. (1974). Mating systems of tropical birds: The effects of weakness or absence of external reproduction-timing factors, with special reference to prolonged pairbonds. Z. Tierpsychol. 34, 265–307. Lambrechts, M., and Dhondt, A. A. (1988). The anti-exhaustion hypothesis: A new hypothesis to explain song performance and song switching in the great tit. Anim. Behav. 36, 327–334. Lampe, H. M., and Espmark, Y. O. (1987). Singing activity and song pattern of the redwing Turdus iliacus during the breeding season. Ornis Scand. 18, 179–185. Lein, M. R. (1978). Song variation in a population of chestnut-sided warblers (Dendroica pensylvanica): Its nature and suggested significance. J. Can. Zool. 56, 1266–1283. Lemon, R. E. (1968). Coordinated singing by black-crested titmice. Can. J. Zool. 46, 1163–1167. Lemon, R. E. (1974). Song dialects, song matching, and species recognition by cardinals, Richmondena cardinalis. Ibis 116, 545–548. Lemon, R. E., and Chatfield, C. (1971). Organization of song in cardinals. Anim. Behav. 19, 1–17. Lemon, R. E., Dobson, C. W., and Clifton, P. G. (1993). Songs of American redstarts (Setophaga ruticilla): Sequencing rules and their relationships to repertoire size. Ethology 93, 198–210. Levin, R. N. (1996a). Song behaviour and reproductive strategies in a duetting wren Thryothorus nigicapillus: I. Removal experiments. Anim. Behav. 52, 1093–1106. Levin, R. N. (1996b). Song behaviour and reproductive strategies in a duetting wren Thryothorus nigicapillus: II. Playback experiments. Anim. Behav. 52, 1107–1117. Lynch, A. (1996). The population memetics of birdsong. In ‘‘Ecology and Evolution of Acoustic Communication in Birds’’ (D. E. Kroodsma and E. H. Miller, Eds.), pp. 181–197. Cornell Univ. Press, Ithaca, NY. Maurus, M., Streit, K. M., Barclay, E., Wiesner, E., and Ku¨hlmorgen, B. (1988). A new approach to finding components essential for intraspecific communication. In ‘‘Primate Vocal Communication’’ (D. Todt, P. Goedeking, and D. Symmes, Eds.), pp. 69–87. Springer-Verlag, Berlin. Maynard Smith, J., and Parker, G. A. (1976). The logic of asymmetric contests. Anim. Behav. 24, 159–175. McComb, K. (1992). Playback as a tool for studying contests between social groups. In ‘‘Playback and Studies of Animal Communication’’ (P. McGregor, Ed.), pp. 111–119. Plenum, New York. McGregor, P. K. (1986). Song types in the corn bunting Emberiza calandra: Matching and discrimination. J. Ornithol. 127, 37–42. McGregor, P. K. (1991). The singer and the song: On the receiving end of bird song. Biol. Rev. 66, 57–81. McGregor, P. K. (1994). Sound cues to distance: The perception of range. In ‘‘Perception and Motor Control in Birds’’ (M. N. O. Davies and P. R. Green, Eds.), pp. 74–94. Springer-Verlag, Berlin. McGregor, P. K., and Dabelsteen, T. (1996). Communication networks. In ‘‘Ecology and Evolution of Acoustic Communication in Birds’’ (D. E. Kroodsma and E. H. Miller, Eds.), pp. 409–425, Cornell Univ. Press, Ithaca, NY. McGregor, P. K., and Falls, J. B. (1984). The response of western meadowlarks (Sturnella neglecta) to the playback of degraded and undegraded songs. Can. J. Zool. 62, 2125. McGregor, P. K., and Krebs, J. R. (1984). Sound degradation as a distance cue in great tit (Parus major) song. Behav. Ecol. Sociobiol. 16 49–56. McGregor, P. K., Krebs, J. R., and Ratcliffe, L. M. (1983). The reaction of great tits (Parus major) to playback of degraded and undegraded songs: The effect of familiarity with the stimulus song type. Auk 100, 898–906.

292

DIETMAR TODT AND MARC NAGUIB

McGregor, P. K., Dabelsteen, T., Shephard, M., and Pedersen, S. B. (1992). The signal value of matched singing in great tits: Evidence from interactive playback experiments. Anim. Behav. 43, 987–998. Morse, D. H. (1989). Song patterns of warblers at dawn and dusk. Wilson Bull. 101, 26–35. Morton, E. S. (1982). Grading, discreteness, redundancy, and motivation–structural rules. In ‘‘Acoustic Communication in Birds’’ (D. E. Kroodsma and E. H. Miller, Eds.), pp. 183–212. Academic Press, New York. Naguib, M. (1995). Auditory distance assessment of singing conspecifics in Carolina wrens: The role of reverberation and frequency-dependent attenuation. Anim. Behav. 50, 1297–1307. Naguib, M. (1996a.) Ranging by song in Carolina wrens Thryothorus ludovicianus: Effects of environmental acoustics and strength of song degradation. Behaviour 133, 541–559. Naguib, M. (1996b). Auditory distance estimation in song birds: Implications, methodologies, and perspectives. Behav. Processes 38, 163–168. Naguib, M. (1997a). Ranging of songs in Carolina wrens: Effects of familiarity with the song type on use of different cues. Behav. Ecol. Sociobiol. 40, 385–393. Naguib, M. (1997b). Use of song amplitude for ranging in Carolina wrens (Thryothorus ludovicianus). Ethology 103, 723–731. Naguib, M. (1998). Perception of degradation in acoustic signals and its implications for ranging. Behav. Ecol. Sociobiol. 42, 143–146. Naguib, M. (1999). Effects of song overlapping and alternating on nocturnally singing nightingales. Anim. Behav. 58, in press. Naguib, M., and Kolb, H. (1992). Vergleich des Strophenaufbaus und der Strophenabfolge and Gesa¨ngen von Sprosser (Luscinia luscinia) und Blaukehlchen (L. svecica). J. Ornithol. 133, 133–145. Naguib, M., and Todt, D. (1997). Effects of dyadic vocal interactions on other conspecific receivers in nightingales. Anim. Behav. 54, 1535–1543. Naguib, M., and Todt, D. (1998). Recognition of neighbors’ song in a species with large and complex song repertoires: The thrush nightingale. J. Avian. Biol. 29, 155–160. Naguib, M., Hultsch, H., and Kolb, H. (1991). Hierarchische Verzweigungsstruktur in den Gesangsstrophen der Vo¨gel. Verh. Deutsch. Zool. Ges. 84, 477. Naguib, M., Fichtel, C., and Todt, D. (1999a). Nightingales respond more strongly to vocal leaders of simulated dyadic interactions. Proc. R. Soc. London B 266, 537–542. Naguib, M., Mundry, R., Ostreiher, R., Hultsch, H., Schrader, L., and Todt, D. (1999b). Cooperatively breeding Arabian babblers call differently when mobbing in different predator-induced situations. Behav. Ecol. 10, in press. Naguib, M., Klump, G. M., Hillmann, E., Grießmann, B., and Teige, T. Assessment of auditory distance in a territorial song bird: accurate feat or rule of thumb? Anim. Behav., in press. Nelson, B. S., and Stoddard, P. K. (1998). Accuracy of auditory distance and azimuth perception by a passerine bird in natural habitat. Anim. Behav. 56, 467–477. Nielsen, B. M. B., and Vehrencamp, S. L. (1995). Responses of song sparrows to song-type matching via interactive playback. Behav. Ecol. Sociobiol. 37, 109–117. Nottebohm, F. (1991). Reassessing the mechanisms and origins of vocal learning in birds. Trends Neurosci. 14, 206–211. Nowicki, S. (1989). Vocal plasticity in captive black-capped chickadees: The acoustic basis and rate of call convergence. Anim. Behav. 37, 64–73. Nowicki, S., Peters, S., and Podos, J. (1998). Song learning, early nutrition and sexual selection in songbirds. Am. Zool. 38, 179–190. Otte, D. (1974). Effects and functions in the evolution of signaling systems. Annu. Rev. Ecol. Syst. 5, 385–417.

VOCAL INTERACTIONS IN BIRDS

293

Otter, K. A., McGregor, P. K., Terry, A. M. R., Burford, F. R. L., Peake, T. M., and Dabelsteen, T. (1999). Do female great tits Parus major assess males by eavesdropping? A field study using interactive song playback. Submitted for publication. Payne, R. B. (1982). Ecological consequences of song matching: Breeding success and intraspecific song mimicry in indigo buntings. Ecology 63, 401–411. Payne, R. B. (1996). Song traditions in indigo buntings: Origin, improvisation, dispersal, and extinction in cultural evolution. In ‘‘Ecology and Evolution of Acoustic Communication in Birds’’ (D. E. Kroodsma and E. H. Miller, Eds.), pp. 454–473. Cornell Univ. Press, Ithaca, NY. Pepperberg, I. M. (1991). Learning to communicate: The effects of social interaction. Perspect. Ethol. 9, 119–164. Philmore, L. S., Sturdy, C. B., Ramsay, S. M., and Weisman, R. G. (1998). Discrimination of auditory distance cues by black-capped chickadees (Poecile atricapillus) and zebra finches (Taeniopygia guttata). J. Comp. Psychol. 112, 1–10. Popp, J., Ficken, R. W., and Reinartz, J. A. (1985). Short-term temporal avoidance of interspecific acoustic interference among forest birds. Auk 102, 744–748. Popp, J. W. (1989). Temporal aspects of singing interactions among territorial ovenbirds (Seiurus aurocapillus). Ethology 82, 127–133. Price, J. J. (1998). Family- and sex-specific vocal traditions in a cooperatively breeding songbird. Proc. R. Soc. London B 265, 497–502. Price, J. J. (1999). Recognition of family-specific calls in stripe-backed wrens. Anim. Behav. 57, 483–492. Riebel, K., and Todt, D. (1997). Light flash stimulation alters the nightingale’s singing style: Implications for song control mechanisms. Behaviour 134, 789–808. Ro¨mer, H. (1993). Environmental and biological constraints for the evolution of long-range signalling and hearing in acoustic insects. Philos. Trans. R. Soc. London B 340, 179–185. Ryan, M. J. (1997). Sexual selection and mate choice. In ‘‘Behavioural Ecology: An Evolutionary Approach’’ ( J. R. Krebs and N. B. Davies, Eds.), pp. 179–202. Blackwell Sci., Oxford. Schroeder, D. J., and Wiley, R. H. (1983a). Communication with shared song themes in tufted titmice. Auk 100, 414–424. Schroeder, D. J., and Wiley, R. H. (1983b). Communication with repertoires of song themes in tufted titmice. Anim. Behav. 31, 1128–1138. Schwager, G., and Gu¨ttinger, H. R. (1984). Der Gesangsaufbau von Braunkehlchen (Saxicola rubetra) und Schwarzkehlchen (S. torquata) im Vergleich. J. Ornithol. 125, 261–278. Searcy, W. A., and Yasukawa, K. (1996). Song and female choice. In ‘‘Ecology and Evolution of Acoustic Communication in Birds’’ (D. E. Kroodsma and E. H. Miller, Eds.), pp. 454–473. Cornell Univ. Press, Ithaca, NY. Shackleton, S. A., and Ratcliffe, L. (1994). Matched counter-singing signals escalation of aggression in black-capped chickadees (Parus major). Ethology 97, 310–316. Simpson, B. S. (1985). Effects of the location in territory and distance from neighbors on the use of song repertoires by Carolina wrens. Anim. Behav. 33, 793–804. Slater, P. J. B. (1981). Chaffinch song repertoires: Observations, experiments and a discussion of their significance. Z. Tierpsychol. 56, 1–24. Smith, D. G., and Norman, D. O. (1979). ‘‘Leader–follower’’ singing in red-winged blackbirds. Condor 81, 83–84. Smith, D. G., and Reid, F. A. (1979). Roles of the song repertoires in red-winged blackbirds. Behav. Ecol. Sociobiol. 5, 279–290. Smith, W. J. (1965). Message, meaning, and context in ethology. Am. Nat. 99, 703–716. Smith, W. J. (1991). Singing is based on two markedly different kinds of signalling. J. Theor. Biol. 152, 241–253.

294

DIETMAR TODT AND MARC NAGUIB

Snowdon, C. T. (1988). Communications as social interaction: Its importance in ontogeny and adult behavior. In ‘‘Primate Vocal Communication’’ (D. Todt, P. Goedeking, and D. Symmes, Eds.), pp. 108–122. Springer-Verlag, Berlin. Sossinka, R., and Bo¨hner, J. (1980). Song types in the zebra finch Poephila guttata castanotis. Z. Tierpsychol. 53, 123–132. Spector, D. A. (1992). Wood-warbler song systems. A review of paruline singing behaviors. In ‘‘Current Ornithology’’ (D. M. Power, Ed.), Vol. 11, pp. 199–238. Plenum, New York. Staicer, C. A., Spector, D. A., and Horn, A. G. (1996). The dawn chorus and other diel patterns of acoustic signalling. In ‘‘Ecology and Evolution of Acoustic Communication in Birds’’ (D. E. Kroodsma and E. H. Miller, Eds.), pp. 426–453. Cornell Univ. Press, Ithaca, NY. Stoddard, P. K. (1996). Vocal recognition of neighbors by territorial passerines. In ‘‘Ecology and Evolution of Acoustic Communication in Birds’’ (D. E. Kroodsma and E. H. Miller, Eds.), pp. 356–376. Cornell Univ. Press, Ithaca, NY. Stoddard, P. K., Beecher, M. D., Campbell, S. E., and Horning, C. L. (1992). Song-type matching in the song sparrow. Can. J. Zool. 70, 1440–1444. Thimm, F. (1973). Sequentielle und zeitliche Beziehungen im Reviergesang des Gartenrotschwanzes (Phoenicurus phoenicurus L.). J. Comp. Physiol. 84, 311–344. Thorpe, W. H. (1972). Duetting and antiphonal song in birds: Its extent and significance. Behav. Monogr. Suppl. 18, 1–305. Thorpe, W. H., and North, M. E. W. (1965). Origin and significance of the power of vocal imitation: With special reference to the antiphonal singing in birds. Nature 208, 219–222. Todt, D. (1968a). Zur Steuerung unregelma¨ßiger Verhaltensabla¨ufe. In ‘‘Kybernetik’’ (H. Mittelstaed, Ed.), pp. 465–485. Oldenbourg, Munich. Todt, D. (1968b). Korrespondenz zwischen wechselsingenden Amseln. Z. Naturforsch. 23, 1619. Todt, D. (1969). Interactions between periodicity and positive feedback controlling the vocalization of alternative motor-patterns in blackbird song. In ‘‘Biokybernetik’’ (H. Drischel and N. Tiedt, Eds.), pp. 168–175. Fischer-Verlag, Jena. Todt, D. (1970a). Gesangliche Reaktionen der Amsel (Turdus merula L.) auf ihren experimentell reproduzierten Eigengesang. Z. Vergl. Physiol. 66, 294–317. Todt, D. (1970b). Gesang und gesangliche Korrespondenz der Amsel. Naturwissenschaften 57, 61–66. Todt, D. (1970c). Die antiphonen Paargesa¨nge des ostafrikanischen Grassa¨ngers Cisticola hunteri prinioides Neumann. J. Ornithol. 111, 332–356. Todt, D. (1970d). Zur Ordnung im Gesang der Nachtigall (Luscinia megarhynchos L.). Verh. Deutsch. Zool. Ges., 64, 249–252. ¨ quivalente und konvalente gesangliche Reaktionen einer extrem regelTodt, D. (1971a). A ma¨ßig singenden Nachtigall (Luscinia megarhynchos). Z. Vergl. Tierpsychol. 71, 262–285. Todt, D. (1971b). Zur gesanglichen Korrespondenz der Schma¨tzerdrossel Cossypha heuglini Hartlaub. Behaviour 38, 146–153. Todt, D. (1971c). ‘‘Abstraktionsleistung’’ einer gesanglich korrespondierenden abessinischen Erddrossel (Geokichla piagiae Bouvier). Z. Tierpsychol. 28, 59–61. Todt, D. (1974). Zur Bedeutung der ‘‘richtigen’’ Syntax auditiver Muster fu¨r deren vokale Beantwortung durch Amseln. Z. Naturforsch. 29c, 157–160. Todt, D. (1975a). Short term inhibition of outputs occurring in the vocal behaviour of blackbirds. J. Comp. Physiol. 98, 289–306. Todt, D. (1975b). Effect of territorial conditions on the maintenance of pair contact in duetting birds. Experentia 31, 648–649.

VOCAL INTERACTIONS IN BIRDS

295

Todt, D. (1975c). Social learning of vocal patterns and modes of their application in grey parrots (Psittacus erithacus). Z. Tierpsychol. 39, 178–188. Todt, D. (1981). On functions of vocal matching: Effects of counter-replies on song post choice and singing. Z. Tierpsychol. 57, 73–93. Todt, D., and Bo¨hner, J. (1994). Former experience can modify social selectivity during song learning in the nightingale (Luscinia megarhynchos). Ethology 97, 169–176. Todt D., and Fiebelkohn, A. (1979): Display, Timing and Function of Wing Movements Accompanying Antiphonal Duets of Cichladusa guttata. Behaviour 72, 82–106. Todt, D., and Hultsch, H. (1978). Functional aspects of sequence and hierarchy in song structure. Acta XVII Congr. Int. Orn. Berlin, 663–670. Todt, D., and Hultsch, H. (1982). Impairment of vocal signal exchange in the monogamous duet-singer Cossypha heuglini (Turdidae): Effects on pairbond maintenance. Z. Tierpsychol. 60, 265–274. Todt, D., and Hultsch, H. (1994). Biologische Grundlagen des Dialogs. In ‘‘Kommunikation und Humanontogenese’’ (K. F. Wessel and F. Naumann, Eds.), pp. 53–76. Kleine Verlag, Bielefeld. Todt, D., and Hultsch, H. (1996). Acquisition and performance of song repertoires: Ways of coping with diversity and versality. In ‘‘Ecology and Evolution of Acoustic Communication in Birds’’ (D. E. Kroodsma and E. H. Miller, Eds.), pp. 79–96. Cornell Univ. Press, Ithaca, NY. Todt, D., and Hultsch, H. (1998). How songbirds deal with large amounts of serial information: Retrieval rules suggest a hierarchical song memory. Biol. Cybernetics 79, 487–500. Todt, D., and Hultsch, H. (1999). How nightingales develop their interactional competence. In ‘‘Proc. 22nd Intern. Ornithol. Congr.’’ (N. J. Adams and R. H. Slotow, Eds.), pp. 122–137. Birdlife, Johannesburg, South Africa. ¨ berpru¨fungen von Steuerungssystemen zur StrophenTodt, D., and Wolffgramm, J. (1975). U wahl der Amsel durch digitale Simulierung. Biol. Cybernetics 17, 109–127. Todt, D., Hultsch, H, and Heike, D. 1979. Conditions affecting song acquisition in nightingales (Luscinia megarhynchos). Z. Tierpsychol. 51:23–35. [85, 86, 369] Todt, D., Hultsch, H., and Duvall, F. P. (1981). Behavioural significance and social function of vocal and non-vocal displays in the monogamous duet-singer. Cossypha heuglini H. Zool. Beitr. 27, 421–448. Todt, D., Goedeking, P., and Symmes, D. (1988). ‘‘Primate Vocal Communication.’’ SpringerVerlag, Berlin. Tyack, P. L. (1996). Studying how cetaceans use sound to explore their environment. In ‘‘Perspectives in Ethology, Vol. 12’’ (D. H. Owings, M. D. Beecher, and N. S. Thomson, Eds.), pp. 251–297. Plenum, New York. Verner, J. (1976). Complex song repertoires of male long-billed marsh wrens in eastern Washington. Living bird 14, 263–300. von Helversen, D. (1980). Structure and function of antiphonal duets. Acta XVIII. Int. Orn. Congr. Berlin, 682–688. Wallace, A. R. (1889). ‘‘Darwinism.’’ Macmillian, London. Wasserman, F. E. (1977). Intraspecific acoustical interference in the white-throated sparrow Zonotrichia albicolis. Anim. Behav. 25, 949–952. Weary, D. M., Falls, J. B., and McGregor, P. K. (1990). Song matching and the perception of song types in great tits, Parus major. Behav. Ecol. 1, 43–47. Weary, D. M., Lemon, R. E., and Perreault, M. L. (1994). Different responses to different song types in American redstarts. Auk 111, 730–734. Whitney, C. L. (1990). Avoidance of song matching in the wood thrush: A field experiment. Wilson Bull. 103, 96–101.

296

DIETMAR TODT AND MARC NAGUIB

Whitney, C. L., and Miller, J. (1983). Song matching in the wood thrush (Hylocichla mustelina): A function of song dissimilarity. Anim. Behav. 31, 457–461. ¨ ber die Beeinflussung des Partners im Duettgesang der Schma¨tzerdrossel Wickler, W. (1974). U Cossypha heuglini. Z. Tierpsychol. 36, 128–136. Wiley, R. H. (1981). Social structure and individual ontogenies: Problems of description, mechanism, and evolution. In ‘‘Perspectives in Ethology’’ (P. P. G. Bateson and P. H. Klopfer, Eds.), pp. 105–133. Plenum, New York. Wiley, R. H. (1994). Errors, exaggeration and deception in animal communication. In ‘‘Behavioral Mechanisms in Ecology’’ (L. Real, Ed.), pp. 157–189. Univ. Chicago Press, Chicago. Wiley, R. H. (1998). Ranging reconsidered. Behav. Ecol. Sociobiol. 42, 143–146. Wiley, R. H., and Godard, R. (1996). Ranging of conspecifics by Kentucky warblers and its implications for interactions of territorial males. Behaviour 133, 81–102. Wiley, R. H., and Wiley, M. S. (1977). Recognition of neighbors’ duets by stripe-backed wrens Campylorhynchus nuchalis. Behaviour 62, 10–34. Wiley, R. H., Godard, R., and Thomson, A. D. (1994). Use of two singing modes by hooded warblers as adaptations for signalling. Behaviour 129, 243–278. Wolffgramm, J. (1975). Lautmustersequenzen und periodische Beziehungen im RollerkanarieGesang. J. Comp. Physiol. 35, 65–88. Wolffgramm, J., and Todt, D. (1982). Pattern and time specificity in vocal responses of blackbirds Turdus merula. Behaviour 81, 264–286. Yasukawa, K. (1981). Song repertoires in the red-winged blackbird (Agelaius phoeniceus): A test of the Beau Geste hypothesis. Anim. Behav. 29, 114–125.

Index

A

imitation studies, 230–233 song context analyzing, 49–51 description, 47–49 female choice assumptions, 60–61 direct benefits, 86–88 experimental approach, 83–85 field studies, 61, 63 indicators, 89–90 indirect benefits, 88–89 integrated approach, 85–86 laboratory studies, 64–68 male quality age, 81–83 brain structure, 75–78 fertility, 80–81 parasites, 73–75 parental effort, 79–80 territory, 79 viability, 81–83 mating system, 58–60 reproductive success, 68–73 species differences, 51–58 song interactions dawn chorus, 254–255, 281–282 delivery type, rules, 265–267 duets, 278–281 function, 276–278 long-range, 248–253 matching patterns, 264–265 overview, 253–255 ranging, 274–276 relevance, others, 282–285 responses, 253 time-specific responses alternation, 255–259 overlapping, 259–263 patterns, 263–274 triads, 282 variety, 271–273

Associative sequence learning theory action units, 222–223 cultural aspects, 240–241 description, 218–219 horizontal processes, 223–224 learning, 225–226 performance, 225–226 redefining imitation, 226–228 summary, 225 training tests, 228–230 two-action tests birds, 230–233 primates, 230–233 rodents, 233–235 vertical processes, 224–225 Attachment, 100–101 Attention index, 126–127

B Bees, 176 Behavior communication role, 247 diposotism, 140–141 perceptual opacity, 216–217 sex differentiation, 141–145 social diversity, 121–123 uniqueness, 101–102 Birds, see also specific birds color vision characteristics, 159–161 mechanism color space, 174–179 double cones, 173–174 interspecific variation, 171–173 oil droplets, 165–171 pigments, 162–165 study, 176–177 297

298

INDEX

Birds (continued) UV sensitivity characteristics, 161–162 functions, 177–179 Blackbirds, songs alternating, 258 overlapping, 263 pattern responses, 267–269

C Carbohydrates deficiency, 18–20 intake, 30–31 regulation, 17–18 Carroll, Lewis, 134 CFA, see Correspondence factorial analysis Choice, female song and assumptions, 60–61 direct benefits, 86–88 experimental approach, 83–85 field studies, 63 indicators, 89–90 integrated approach, 85–86 laboratory studies, 64–68 Cognition attachment, 100–101 imitation, 215 socialization, 147 Color vision mechanism color space, 174–179 double cones, 173–174 interspecific variation, 171–173 oil droplets, 165–171 pigments, 162–165 nonanthropocentrical analysis, 176–177 perception, 197 requirements, 159–160 Communication behavior role, 247 primate, 101 Correspondence factorial analysis application, 130–134 description, 128–129 Cues, visual, 18–21 Culture, imitation, 240–241

D Darwin, Charles, 45–47, 247 Dawn chorus, 254–255, 281–282 Defecation, 7 Demographics, social ontogeny, 135 The Descent of Man and Selection in Relation to Sex, 45 Development, see Social development Diposotism, behavioral, 140–141 Diversity indexes, 121–127

E Endogenous rhythms, 6–7 Environment, feeding stimuli, 7 Estradiol implants, 65–68 Extra-pair copulations, 69–70 Extra-pair young, 69–70

F Feedbacks, nutrient, 31–34 Feeding patterns analysis, 2–3 ending factors excitation levels, 9, 11 ingestion rate, 8–9 negative feedback, 11 time, 8–9 initiation factors causal, 6–8 nutritional, 5 physiological, 6 size, 5 time, 5 intake, 13–14 integrative simulation, 11–13 nonnutrient dimensions, 35 nutrition intake targets compromises, 24–27 dynamic, 27–31 intake, 24 rails, 23–24 space, 23–24 interactions, 21–23 learning cues, 14, 18–21

299

INDEX

regulatory system, 34–35 spatial complexity, 35–37 taste model, 31–34 UV vision, 180–187 Fertility, warbler, 80–81 Fisher effect, 88 Fisherian models, 46 Food rails, 23–24 Foraging, UV vision, 180–187

H Hailman framework, 106 Hamilton-Zuk hypothesis, 74 Humans color vision, 159–162 cone types, 197 perceptual opacity, 216–217 UV sensitivity, 161–162

I Imitation cognitive mechanism, 215 culture, 240–241 perceptual opacity, 217 redefining, 226–228 theories associative action units, 222–223 bird studies, 230–233 cultural aspects, 240–241 description, 218–219 horizontal processes, 223–224 learning, 225–226 performance, 225–226 primate studies, 230–233 redefining imitation, 226–228 rodent studies, 233–235 summary, 225 training tests, 228–230 vertical processes, 224–225 classification, 218 copying, 220–221 development, 217–218 reinforcement-based, 221–222 social cognitive, 218–219 transformational, 218–219

Infants, primate attachment, 100–101 interactions age factors, 140 dynamics, 120–121 locomotion factors, 114–116 physical environment, 134 roles, 116–117 sex factors, 140–145 uterine environment, 106–107 socialization, 103–104 Intrasexual selection theory, 45–47

L Learning theory, see Associative sequence learning theory Locomotion, 114–115 Locusta migratoria, see Locusts Locusts defecation, 7 description, 1–2 endogenous rhythms, 6–7 feeding patterns analysis, 2–3 ending factors excitation levels, 9, 11 ingestion rate, 8–9 negative feedback, 11 time, 8–9 initiation factors causal, 6–8 nutritional, 5 physiological, 6 size, 5 time, 5 intake regulation, 13–14 integrative simulation, 11–13 learning cues, 14, 18–21 no-choice assays, 14, 16–17 nonnutrient dimensions, 35 regulatory system, 34–35 spatial complexity, 35–37 taste model, 31–34 molting, 29–30 nutrition intake targets compromises, 24–27 dynamic, 27–31

300

INDEX

Locusts (continued) intake, 24 rails, 23–24 space, 23–24 interactions, 21–23

evolutionary factors, 30–31 model advantages, 24 physiological factors, 27–29 rails concept, 23–24 space factors, 23–24 interactions, 21–23

M Mating systems female choice direct benefits, 86–88 experimental approach, 83–85 field studies, 61, 63 indicators, 89–90 indirect benefits, 88–89 integrated approach, 85–86 laboratory studies, 64–68 male quality age, 81–83 brain structure, 75–78 fertility, 80–81 parasites, 73–75 parental effort, 79–80 territory, 79 viability, 81–83 reproductive success, 68–73 species differences, 58–60 Molting, 29–30 Mothers, primate –infant, interactions dynamics, 120–121 locomotion factors, 114–116 physical environment, 134 roles, 116–117 uterine environment, 106–107 N Nightingales, song alternating, 256–257 overlapping, 262–263 pattern responses, 269–271 virtuosity, 266–267 Nutrients choice patterns, 14, 16–21 feedbacks, 31–34 intake targets compromises, 24–27 developmental factors, 29–30

O Oil droplets, 165–171 Ontogeny, see Psychological ontogeny; Social ontogeny Opacity, perceptual, 216–217 Opsin, 162–165 Orientation, 179

P Parasites, 73–75 PCAIV, see Principal component analysis with instrumental variables Perceptual opacity, 216–217 Performance, 225–226 Phenotypes control function, 105–106 group social, 109 social ontogeny, 147 Primates communication, 101 imitation studies, 230–233 infants attachment, 100–101 interactions age factors, 140 dynamics, 120–121 locomotion factors, 114–116 physical environment, 134 roles, 116–117 sex factors, 140–145 uterine environment, 106–107 socialization, 103–104 socialization, 102–103 social ontogeny description, 105–108 index approach, 116–121 longitudinal studies, 110–111 multivariate approach advantages, 145–149

INDEX

description, 127–129 techniques, 129–134 processes, 108–110 univariate approach, 113–116 Principal component analysis with instrumental variables application, 135 description, 128–129 Proteins blood levels, 14, 16–17 deficiency, 18–20 intake, 30–31 regulation, 17–18 Psychological ontogeny, 105 R Reproduction, song role, 68–73 Retinal, 163 Rhodopsin, 164 Rodents, imitation studies, 233–235 S Schistocerca gregaria, see Locusts Sensory exploitation hypothesis, 46–47 Sensory physiology, 31–34 Sex differentiation development, 141–145 infant socialization, 140–141 Signaling, UV vision, 187–196 Social development functions, 113–114 networks, 125–126 sex differentiation, 141–145 –socialization differences, 102–103 unifying, 104–105 Socialization cognitive aspects, 147 definition, 100–102 distinctions, 102–103 –social development differences, 102–103 unifying, 104–105 Social ontogeny demographic factors, 135 index approach, 116–121

longitudinal studies, 110–111 multivariate approach advantages, 145–149 description, 127–129 techniques, 129–134 processes, 108–110 –social development, 105–108 univariate, 113–116 Social organization, 128 Society, definition, 100 Songs analyzing, 49–51 dawn chorus, 254–255 description, 47–49 mating system female choice assumptions, 60–61 direct benefits, 86–88 experimental approach, 83–85 field studies, 61, 63 indicators, 89–90 indirect benefits, 88–89 integrated approach, 85–86 laboratory studies, 64–68 male quality age, 81–83 brain structure, 75–78 fertility, 80–81 parasites, 73–75 parental effort, 79–80 territory, 79 viability, 81–83 reproductive success, 68–73 species differences, 58–60 social interaction dawn chorus, 281–282 delivery type, rules, 265–267 duets, 278–281 function, 276–278 long-range, 248–253 matching patterns, 264–265 overview, 253–255 ranging, 274–276 relevance, others, 282–285 time-specific responses alternation, 255–259 overlapping, 259–263 patterns, 263–274

301

302

INDEX

Songs (continued) triads, 282 variety, 271–273 species differences, 51–58 Sperm, warbler, 80–81

T Taste model, 31–34 Territory, warbler, 79

U Ultraviolet vision blocking filters, 165 color perception, 197 functions analysis, 177–178 foraging, 180–187 orientation, 179 signaling, 187–196 sensitivity, 161–162

V Vision color mechanism color space, 174–179 double cones, 173–174 interspecific variation, 171–173 oil droplets, 165–171 pigments, 162–165 requirements, 159–160

UV blocking filters, 165 functions foraging, 180–187 orientation, 179 signaling, 187–196 sensitivity, 161–162 Visual cues, 18–21

W Wallace, Alfred, 247 Warblers, song analyzing, 49–51 description, 47–49 mating system female choice assumptions, 60–61 direct benefits, 86–88 experimental approach, 83–85 field studies, 61, 63 indicators, 89–90 indirect benefits, 88–89 integrated approach, 85–86 laboratory studies, 64–68 male quality age, 81–83 brain structure, 75–78 fertility, 80–81 parasites, 73–75 parental effort, 79–80 territory, 79 viability, 81–83 relation, 58–60 reproductive success, 68–73 species differences, 51–58

Contents of Previous Volumes

Volume 18 Song Learning in Zebra Finches (Taeniopygia guttata): Progress and Prospects PETER J. B. SLATER, LUCY A. EALES, AND N. S. CLAYTON Behavioral Aspects of Sperm Competition in Birds T. R. BIRKHEAD Neural Mechanisms of Perception and Motor Control in a Weakly Electric Fish WALTER HEILIGENBERG Behavioral Adaptations of Aquatic Life in Insects: An Example ANN CLOAREC The Cicadian Organization of Behavior: Timekeeping in the Tsetse Fly, A Model System JOHN BRADY Volume 19 Polyterritorial Polygyny in the Pied Flycatcher P. V. ALATALO AND A. LUNDBERG Kin Recognition: Problems, Prospects, and the Evolution of Discrimination Systems C. J. BARNARD Maternal Responsiveness in Humans: Emotional, Cognitive, and Biological Factors CARL M. CORTER AND ALISON S. FLEMING The Evolution of Courtship Behavior in Newts and Salamanders T. R. HALLIDAY Ethopharmacology: A Biological Approach to the Study of Drug-Induced Changes in Behavior A. K. DIXON, H. U. FISCH, AND K. H. MCALLISTER

Additive and Interactive Effects of Genotype and Maternal Environment PIERRE L. ROUBERTOUX, MARIKA NOSTEN-BERTRAND, AND MICHELE CARLIER Mode Selection and Mode Switching in Foraging Animals GENE S. HELFMAN Cricket Neuroethology: Neuronal Basis of Intraspecific Acoustic Communication FRANZ HUBER Some Cognitive Capacities of an African Grey Parrot (Psittacus erithacus) IRENE MAXINE PEPPERBERG Volume 20 Social Behavior and Organization in the Macropodoidea PETER J. JARMAN The t Complex: A Story of Genes, Behavior, and Population SARAH LENINGTON The Ergonomics of Worker Behavior in Social Hymenoptera PAUL SCHMID-HEMPEL ‘‘Microsmatic Humans’’ Revisited: The Generation and Perception of Chemical Signals BENOIST SCHAAL AND RICHARD H. PORTER Lekking in Birds and Mammals: Behavioral and Evolutionary Issues R. HAVEN WILEY Volume 21 Primate Social Relationships: Their Determinants and Consequences ERIC B. KEVERNE 303

304

CONTENTS OF PREVIOUS VOLUMES

The Role of Parasites in Sexual Selection: Current Evidence and Future Directions MARLENE ZUK Conceptual Issues in Cognitive Ethology COLIN BEER Responses 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 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 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 Parasites and the Evolution of Host Social Behavior ANDERS PAPE MØLLER, REIJA DUFVA, AND KLAS ALLANDER The Evolution of Behavioral Phenotypes: Lessons Learned from Divergent Spider Populations SUSAN E. RIECHERT Proximate and Developmental Aspects of Antipredator Behavior E. CURIO 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 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 Cultural Transmission in the Black Rat: Pine Cone Feeding JOSEPH TERKEL 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 Sociality, Group Size, and Reproductive Suppression among Carnivores SCOTT CREEL AND DAVID MACDONALD Development and Relationships: A Dynamic Model of Communication ALAN FOGEL

CONTENTS OF PREVIOUS VOLUMES

305

Why Do Females Mate with Multiple Males? The Sexually Selected Sperm Hypothesis LAURENT KELLER AND HUDSON K. REEVE

Physiological, Sensory, and Experiential Factors of Parental Care in Sheep F. LE´VY, K. M. KENDRICK, 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 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 Somatosensation and Maternal Care in Norway Rats JUDITH M. STERN Experiential Factors in Postpartum Regulation of Maternal Care ALISON S. FLEMING, HYWEL D. MORGAN, AND CAROLYN WALSH 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

Parental Investment in Pinnipeds FRITZ TRILLMICH Individual Differences in Maternal Style: Causes and Consequences of Mothers and Offspring LYNN A. FAIRBANKS 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 ANDRE´ S. GILBURN Vocal Learning in Mammals VINCENT M. JANIK AND PETER J. B. SLATER Behavioral Ecology and Conservation Biology of Primates and Other Animals KAREN B. STRIER How to Avoid Seven Deadly Sins in the Study of Behavior MANFRED MILINSKI Sexually Dimorphic Dispersal in Mammals: Patterns, Causes, and Consequences LAURA SMALE, SCOTT NUNES, AND KAY E. HOLEKAMP

306

CONTENTS OF PREVIOUS VOLUMES

Infantile Amnesia: Using Animal Models to Understand Forgetting H. MOORE 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 Understanding the Complex Song of the European Starling: An Integrated Ethiological Approach MARCEL EENS Representation of Quantities by Apes SARAH T. BOYSEN 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 Stress and Decision-Making under the Risk of Predation: Recent Developments from Behavioral, Reproductive, and Ecological Perspectives STEVEN L. LIMA

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

Volume 28 Sexual Imprinting and Evolutionary Processes in Birds: A Reassessment CAREL TEN CATE AND DAVE R. VOS 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 MoleRats, 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

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

Olfaction in Birds TIMOTHY J. ROPER

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

Intraspecific Variation in Ungulate Mating Strategies: The Case of the Flexible Fallow Deer SIMON THIRGOOD, JOCHEN LANGBEIN, AND RORY J. PUTMAN