FORAGING BEHAVIOR CNY Alan C. Kamil, John R. Krebs, PUM RM OIE CM aT ETFORAGING BEHAVIOR Edited by Alan C. Kamil University of Massochuserts Amherst, Massachusetts John R, Krebs ‘Outer Uriversiy Ost England and H. Ronald Pulliam University of Georgia Athens, Georgia PLENUM PRESS ¢ NEW YORK AND LONDONLibrary of Congress Cataloging in Publication Data Foraging behavior, Based on a conference on foraging behavior which was held in June, 1984, in Providence, nl Bibliography: p. Includes index. |. Animals—Food—Congresses, |, Kamil, Alan C. I. Krebs, Jahn, R. Ul Pulliam, H, Rona. QL756.5.F66. 1987 591.53 87-6951 six 9 IS BNoL 3: 9781 -4613-1802 © 1987 Plenum Press, New York Saligover reprint ofthe hardener Est eon 1987 ‘A Division of Plenum Publishing Corporation 233 Spring Street, New York, N.Y. 10013 Alb rights reserved [No part of this book may be reproduced, sored in a retrieval system, or transmnitied im any form or by any means, electronic, mechanical, photocopying, microfilming, recording, or otherwise, without written permission from the PublisherCONTENTS OPTIMAL FORAGING THEORY . A Brief History of Optimal Foraging Ecology .....s61+ T.W. Shooner Faith and Foraging: A Critique of the aradigm Argument ‘THE PRODLEM OF SELECTIVITY 10... .sseeeeseeeenerersnseeereneeeetees A Test of the Energy Maximization Prenise of Optinal Foraging Theory D.B. Campbell Delivering Food to a Central Place: Three Studies of Bee-Eators Merops ApLAGEer ..+essesecscsesnanavereeeesensere J.R. Krebs, M.T. Avery and A.I. Houston Operant Conditioning Simulations of Foraging and the Delay-Reduction Hypothesis: ss csesereverssrsesesersesneersere E. Fantino Foraging Behavior of Sone Wintering Waders and Habitat Distribution . P.M. Meire Prey-Selection Foraging Time Constraints and Diet Choice ssessesererseeererseenene J.R. Lucas PATCH UTILIZATION .. Stochastic Models of Optimal Foraging ... RF. Green Starlings and Optimal Foraging Theor; Fractal World . Ay Kacelink and 1.¢, Cutht1L Modelling in a Tests of Optiaal Foraging Using an Operant Analogue ... J. Hangon Competition for Non-Depleting Resources: The Ideal Free Distribution in Sticklebacks .. M, Milingkt 141 143 173 193 aus 239 am 23 303 25 363Foraging Games in a Random Environment . T. Caraco THE REPRODUCTIVE CONSEQUENCES OF FORAGING ..........20ese00+ Reproductive Consequences of Dietary Specialization and Switching in an Ecological Generalist .sssseseseeeeees R. Pierotti and €. Annett The Consequences of Foraging for Reproductive Success « D.H. Marge and R,S, Frite Parents, Nestlings and Feedings Frequency: A Model of Optinal Parental Investment and Implications for Avian Reproductive Strategies WN. Mur LEARNING AND FORAGING . The Role of Learning in Honey Bee Foraging ToL, Gould Adaption to Reward ..sseeses J.E.R. Staddon and A.K. Redd Signal Detection Theory and Foraging for Cryptic or Mimetic Prey 7, Getty, A.C, Kamil and Panela G, Real Foraging Theory and Design . J.G. Ollason Should Foragers Renember Where They've Been? Explorations of a Simulation Model Based on the Behavior and Energetics of Territorial Hummingbirds ....... D.P, Armatrong, C,L, Gass and €.D. Sutherland Randon and Systenatic Fotaging, Experimental Studies of Depletion and Schedules of Retnforcenent WM, Baum CACHING BEHAVIOR so. e545, Cache-Frotecting Behavior of Food-Hoarding Animals s.seeeeseserseee S.B, Vander Wall and K,C. Smith Cache Site Menory in Birds ......0+eereee+s R,P, Balda, E.G, Bunch, A.C. Kamil, D,F. Sherry and DF, Tomback 389 a5 ar 443 457 a7 479 497 525 549 363 587 609 eu 645 667One of the most important factors contributing to the explosion of empirical interest in foraging behavior has been the development of optimal foraging theory (OFT), The vast majority of papers on foraging behavior in the last 10 years reference an optimal foraging model. In many cases, the reference may be peripheral. That is, the study being reported may have been carried out for reasons having little to do with OFT, but some refer- ence in the manuscript seemed appropriate (or politically advisable) to the authors. But in most cases, tt is clear that OFT provided an immediate stimulus for the research. As described by the two papers in this section, this influence has been widespread - laboratory and field studies, psycho- Logical, ecological, ethological and anthropological investigations have all been influenced by OFT. However, OFT remains controvereial, For some, it is anathema — tautological, misguided, and fruitless, For others, it 1s extremely exefting — programmatic, progressive, and fruitful. Others are less extreme, seeing OFT 5 « useful tool. How can a single approach produce not only an enormous range of opinion, but of emotion as well? This fs 4 complex issue. However, one often unrecognized factor that may contribute te this variability 4m reaction is the variety of different uses to which OFT has been put, and to the variation in the interests, and therefore paradiges and long range goals, of those participating in the debate, For evolutionary theorists, the main issue is the usefulness of OFT for understanding the design and evolutionary history of behavior; for ecologists, for understanding ecological systens; for behaviorigte, includ= ing ethologists and psychologists, for predicting and underatanding the complexities of behavior better. OFT may prove useful in one or many ways. But no oingle approach is Likely to serve so many diverse interests equelly well.There are two papers in this section. They take widely different views of OFT, In the first, Schooner provides a history of the development of OFT and an upbeat view of the prospects for future davelopnents within the OFT framework. Schoener particulatly emphasizes the relationships between OFT and several academic disciplines cuteide of biology, recent theoretieal advances and problems for future theory and research. In the second paper, Gray argues that OFT represents a large advance over earlier attempts at functional explanation, He points out that OFT has mode the assumptions of adaptive argument explicit. But he feele that the main result of this advance has been to fully reveal the difficulties of many of these assumptions. He concludes that biologists must abandon the concept of optimal design, and sketches an alternative framework. Thus two very different views of the utility and validity of OFT are expressed in this ction. The editors believe that it is crucial for the reader to consider both of these chapters, and not just read the one that happens to agree with the 1. The debate about OFT can have positive reader's pre-existing bias: effecta if it forces us to consider some very basic issues about our enterprise, What constitutes support for a model? How do we evaluat aseuaptions? How does our theoretical bia affect what we do? And most important, how do we evaluste the usefulness of an approach? We must continually ask ourselves these questions. We hope that these two chap- tera, taken together, further that process, One point we would like to suggest the reader keep fn mind 18 that both Gray and Schoener indicate substantial areas that appear to deserve considerable future empirical and theoretical attention, Perhaps the ensible strategy now ts to attack these problems: enpirically and theoret- feally, In the long run, there may be less conflict between alternative approsehes than currently appears to be the case, For exanple, studies of the ontogeny of foraging strategies would be of great interest to many of us regardiess of where we stand on the question of OFT. Finally, one specific question must be addressed. How can different scientists review the some Literature and come to very different conclu- sions about the extent to which the daca support OFT? Part of the answer to this question involves a basic problem in tests of models, especially behavioral models in which the data often include mich "noise." Suppose thet 2 quantitative model 1s tested in an experiment. There are two typesof tests one could apply to the resulting data. One could test for a significant correlation between the predictions and the dat teat for significant deviations between the predictions and the date. If er one could both tests are significant, should the results be viewed as support for the model, or as nonsupport? Perhaps both views are justified, If there are significant differ- ences between behavior and the predictions, then clearly the model is incorrect. However, 4f there is alec qualitative agreement between model and data, the model may be partislly correct though incomplete. Thus qualitative agreement between a model and data can be an ambiguous situa tion. In some cases, further refinement and development of the model will prove worthwhile. In other cases, the model should be abandoned. Usually, We suapect, making the Judgment will require time. If succeeding veretons of the model seen t0 be leading to progress, then continued development is advisable. But if not, then at sone point the model (or general approach) needs to be dropped. Im the case of OFT, the argument {6 about « general approach, not specific model. OFT consists of a set of models which share certain features, WI the optimization approach prove useful in the long run? The editors believe that 1t probably will, especially in the realm of behavior. However, wo also believe that workers in the field must continue to question and re-evaluate the OFT approach. In particular, it ts nec sary to separate the different potential functions of OFT, and evaluate its usofulness in each aren separately. We hope that by presenting the differ- ent views of Schoener and Gray, ve will contribute to a continuing re-evaluae tion of assumptions, approaches and progress.A BRIEF HISTORY OF OPTIMAL FORAGING ECOLOGY Thomas W. Schoe: Department of Zoology University of California Davis, California 95616 Optinel foraging ecology began in 1966 with the back-to-back public tion in Anerican Naturalist of two papers, one by Mackrthur and Planka, the other by Enlen. Although different in detail, the papers were collectively unique im their proposal that prey selectivity cavld be understocd as driven by a tendency, selected through evolution, to maximize net onetay gained per unit tine feeding. The teaporal coincidence of the two papers suggests that the ecological world was as ready to receive this proposal as to invent ft. Indeed, in the 20 years aince, hundreds of papers mentioning opeinal foraging have been published--1ike other areas of science (Price 1963) and ecology (Schoener 1974s, 19838), optinel foraging enjoyed an exponential growth phase which ts only now beginning to inflect (Pyke 1984, and below). Two decades would ordinarily seem too short a tine to merit « history, but the intensity of research activity on optimal foraging perhaps suggests otherwise. Nonetheless, it is appropriate to ask, why a history, long or short, would be of import. T can think of four areas where the kind of history I au atteapting may have an effect. Firat, a history of concepts and ideas can highlight the logical connections between theories and gubtheories, As such, a philosophical function is served. There are perhaps surer ways to approach such philo~ sophical questions, but sometines 2 history can be heuristic. Second, a history can illustrate the motivations and creative proces: es of individual sedentists, As such, a kind of paychological function 4sserved. Because academics are forever seeking role models, they are often intensely interested in biographical details. The shortness of the tine span to be covered here is both advantageous and disadvantageous for this aspect: memories are still relatively fresh so accuracy is high, but as most participants are not yet dead, the historian has to tread cautiously— history with little lag time merges into gossip. Third, 2 history can reveal sonerhing about the sociology of scien- tists, and in particular for this paper, how practitioners of one type of ocience are perceived by scientists in adjacent fields. Because optimal foraging ideas received a conspicuously negative reception from a number of ecologiste, yet an equally positive reception from many social scientista, sociological aspects promise to be intriguing. Fourth, by providing a chronology of how and why scientiste actually did some body of research, a history can provide 2 sciontifie service, Knowing the original motivations for constructing the theory can provide feedback that would be lacking were those motivations permanently forgot— ten. New directions may be suggested by a map of the route folleved so far, and the overview provides # perspective on our successes and failures, which at the time appear so overblown. I do not claim that what follows will serve any of these purposes well, but they may provide something for the reader to consider who is uninterested in the story tteelf, However well such purposes are served, T will definitely not attempt a review of the field, Reviews are best done non-chronologically, whereas histories by definition are chronological, Moreover, two excellent reviews have recently appeared (Krebs, Stephens & Sutherland 1983; Pyke 1984), and I am fortunate in being able to refer the reader to them for technical decails, The remainder of this article begins with the ecological world before optimal foraging ecology was invented. It then traces the development of optinal diet and pateh-use theory in ite classical form. This early theory forms the basis of many more recent developments, go it will be treated in unusual detail both with respect to mathematical formulation and early teste. T will then turn to somewhat more peripheral although st{11 impor= tant "early" theoretical concepts, such as nutrient constraints, long-term foraging behavior, and how such time to spend eating. ‘The subsequent explosion of tests of optimal foraging hypotheses and the concurrent invasion of anthropology and psychology will be described. How the clas=sical theory of diet and pateh use was godified or not in light of later theoretical analyses will then be analyzed, and I will eketch recent theoretical extensions, such ae central place foraging, optimal territory eize, and risk proneness and aversion. Repeated themes criticizing optimal foraging ecology will be examined for seientifie validity in Light of a systematic survey of the literature; sociological aspects will also be touched upon, Finally, an assessment will be made of the current status of the field and ite possible futures. ECOLOGY BEFORE OPTIMAL FORAGING Alehough it may be hard for recent practitioners to imagine, there was an ecology before optimal foraging was invented in 1966. How were the data that optimal foraging theory (OFT) purports to explain treated before its inception, and how were optimal foraging concepts once established incorpor- ated into the canons of ecology? ‘As Kubn (1970) and others have pointed out, what scientists thought they knew at a particular moment in a discipline's history can often be discovered by examining the then contemporary textbooks. Accordingly, I pulled a number of general ecology texts from my shelves and from those of ny (mainly older) colleagues. As was very frustrating to ecologists, in the years befora the seven= ties general ecology texts were rare. The most preeminent vas Odum's Fundamentals of Ecology, and I checked the first (1953) and second (1959) editions, which are similar, for references to feeding behavior. Thi volumes contain no sections at al1 on behavioral ecology; indeed, no portion was specifically devoted to any aspect of what we now consider individual ecology (evolutionary, behavioral and physiological), except a brief discussion of territoriality im the second edition, "Feeding" 4s not in the indices. In the third edition (1971), « chapter was added called "The Species and the Individual in the Ecosystem," and it contains nothing about feeding strategies, Behavior, rather than behavioral ecology, is treated; for example, the feeding of hydra illustrates a eropism." A second pre-1966 general ecology text 1s Clarke's (1954) Elements of Ecology. Unlike Odun's, Clarke's index does contain “adaptations for feeding," but the referenced discussion mainly concerns morphological specializations. However, other discussion focuses on the types of specialization, describ- ing which organiems are specialized and which not, but without providing a rationale. ‘The various kinds of nutrients are alco discussed.Other pre-1966 texts are restricted by title to animals. The earliest and perhaps best known d6 Elton's Aninal Ecology, firet published in 1927 and reprinted @ number of times, Again, no references to feeding behavior are in the index; food preferences are mentioned only in the methods section, and as a kind of datum to be gathered rather than to be explained. The next major text to appear on animal ecology was Allee, Emerson, Park, Fark and Schmidt's (1949) huge (837 pages!) Principles of Animal Ecology. Surely in a text this size something would be said about feeding behavior, and indeed feeding specialization is discussed, but in the section on gommunity organization, A brief account is given of what the authors call “a clearly drawn difference of opinion (p. 517),” namely, whether animale take food types in proportion to their availability, or whether selection occurs. McAtee (1932) interpreted data from his 80,000 bird stomachs as supporting the former view (but nothing Like @ rigorous test was performed), whereas Dunn (193) argued differently. (Were the former view correct, incidentally, OFT would have had nothing to explain.) Finally, the third major text on animal ecology before 1966, MacPayden's Animal Ecology, firet appeared in 1957, and appeared as a second edition in 1963. The latter, but not the former, contains dominated by phystological energetics, "Choice experiments" are discussed section of feeding, a section, however, in a methodological appendix as a way to describe diets but not as a vay to test hypotheses, During 1972-74, well past the entree of optimal foraging into the ecological literature, « relative explosion of general ecology texts hit the market. In the vanguard was Kreb's (1972) Ecology. Very little behavioral ecology 1s treated in its ow right: a brief chapter on habitat preferences and another (under community ecology) on evolutionary ecology are the only exanples. Nothing on feeding strategies is mentioned anywhere, a situation continuing through the second (1978) and even ehird (1985) editions. Although not aa conceptual a text as Krebs's, Saith's Ecolo; and Field Biology (1966, 1974) also ignores feeding-strategy ecology (as dos the topical "text" by Colinvaux (1973]). Smith's treatment has veferences to food preferences in a few places and treats behavior as classical ethology, with discussions of fixed-action patterns and the like. In short, at least two ecologists seem recently to have considered feeding- strategy ecology @ flash in the pan, although by 1980 it was more of a conflagration,‘The other ecology texts emerging in the early seventies all cover optinal foraging ecology extensively. In the firet edition of Ricklefs! Ecology (1973), the Mackrchur-Pdanka (1966) model de presented in detatt, although somewhat surprisingly in s chapter on comunity ecology. In the second edition (1979), the same ie true, except that foraging strategies are now discussed in a nev chapter on the miche, This text has sections on other aspects of feeding behavior, including one in which Ivlev's (1561) experiments on fishes are discussed extensively. These experiments predate optimal foraging theory, but they deal with selectivity and are quite relevant to it (e.g., Schoener 1969a), In contrast to RickLefe's text, McNaughton and Wolf's (1973) firet edition of General Ecology discusses optimal foraging ecology in the chapter on predator~prey relationships. It 1s found in a section entitled "Energy Efficiency." In the second edition (1979), the general locus of the optimal foraging discussion is the same, But now a section entitled "Optimality Models" has been added. Clearly in this series of texte optimal foraging ecology is coming into its own. The other two texts published during 1972-74 are by co-inventors of the field. In Enllen's (1973) Ecology: An Evolutionary Approach, feeding is given a separate chapter under the general subdivision "Ecology of Individuals," and various aspects of optimal foraging are treated therein. eine In the f4rot edition of Pianka's (1974) Evolutionary Eeology, foraging ideas appear in three separate sections—-one on population ecology, one on the niche, and one on biogeography. ‘This situation renained essen- tially unchanged through the 3rd edition (1983). The preceding synopsis seems to show the following. First, the data optimal foraging theory tries to explain are often absent from pre-1966 texts, Ironically, during that time, paychologiats, especially behaviorists, were generating data closer to the type demanded by the theory than were many ecologiste—-the irony, as detailed below, is that the ecological theory was eventually applied a posteriori to many of the psychological data, Second, of the post-1966 texts, at least one conceptual book ignores optimal foraging theory altogether, as well as omitting the phenonens OFT tries to explain, The other conceptual texts include optimal foraging, but disagree on where it should be discussed. Only Ealen's text, which pro~ ceeds from individuals through populations to communities, has an entire chapter on feeding, and it is only under his organization that concepts of behavioral ecology seem easily to fit. (Emlen, incidentally, organized his text fh what the phtlosopher Beckner [1974] calls a "perfect hierarchy," = highly logical way to treat ecological phenomena [see Sehoener 1986].) Inshort, while optimal foraging ecology has now been canonized, there is wide Afeageeement regarding inte which canen it should be placed. ORIGINS OF OPTIMAL DIET THEORY: "REDISCOVERING THE WHEEL’ Optimal diet theory fs that branch of optimal foraging theory con cerned with which kinds, of a set of available kinds of food, should be aten by a forager. The generative papers on optimal foraging deal entire- ly (Emlen 1966) or largely (MacArthur & Pianka 1966) with optimal diet. The formulation in MacArthur and Pianka’s paper ts = graphical version of what has been called "the fundamental theorem of OFT (Charnov, personal communication)" and "the contingency model (Schoener 1974b)," among other designations, Mackrthur was fond of expressing ideas graphically (Schooner 1972), particularly because qualitative generality is often possible QMackrthur 1972). Th this case, the graphical model is somevhat awkward and rather mote difficult to intuit than the subsequent algebraic formu lation (although basically identical to the lacter except for omfesion of the energy cost of searching and the assumption that all items have equal net energies). As 1s ao often the case with initial formulations, Machrthur and Pianka doubtless did not appreciate the importance and extent of appli cability of the theory they derived, Rather than behavioral ecology, application of the model focuses on population ecology and evolutionary ecology. How will an invading epecies affect diet and patch use of a resident species? What fa the limiting similarity of conpetitors? Even the title, "On the Optimal Use of a Patchy Environment," portends little of the branch of behavioral ecology the psper was co-founding, Of all Kac- Acthur's papers (as opposed to books), this one has the most citations (G79) in the citation indices of 1970-83. In contrast to MacArthur and Pianka's paper, the intended application of Emlen's (1966) paper was much closer to behavioral ecology--the diet of ‘2 snail (Thais emarginata), anong other phenomena, explanation by his model. Even the title of Emlen's paper, "The Role of Tine and Eneray dn Pood Preference," is squarely in the bailiwick of modern OFT. While conceptually ancestral for optimal diet theory, the Emlen model was a mathematieal dead end. The major reason 1ieo in its rather unintu= ultimately given an itive probsbilistte formulation and the consequently somewhat hidden 10accumption (Schoener 1971) that che forager ie "prescient," in the sense that it knows precisely which type of prey it will next encounter (not Just the encounter probabilities as the contingency model assumes). Moreover, there appears to be an error in one of the steps leading to the final result (Schoener 1971). Enlen (1968) later published a somewhat more general but related model of foraging. This one too has problems: the itemspecific energies pet time (e/t's; see below) are integrated, rather than calculating the integral of the energies over the integral of the times. (Thus unlike all papers fot which Templeton and Lewlor [1981] claim that "the fallacy of the averages” 1s committed, Enlen’s [1968] paper, not cited by them, seems in fact to comit it.) Even though the sathenstics were not generative, both of Enlen's papers and the discussion in his text (Emlen 1973) were biologically stimulating to the general field and more of a harbinger than the paper by MacArthur and Planka. The first strictly algebraic formation of the contingency model is apparently in Schoener (1969a). The formulation is all-but-buried in a lengthy paper whose ultimate purpose was to model optinal body size for predators. The paper distinguished Type I (eit-and-wait) and Type IT (actively searching) predators; because the latter (but not the former) was assumed to be feeding contimuously and not simultaneously performing other activities while searching, @ contingeney-model approach vas suggested. The contingency equation was never written doun as such, but instead the time to satisfy a given metabolic requirement was minintzed. This quantity was directly proportional to the expected time per item eaten and inversely proportional to the expected net energy per dten eaten, essentially giving the inverse of T, of the contingency model (Equation 1 below). ULeinately, my motivation for writing this paper was to understand the evolution of size differences between and within species, with special reference to Anolis lizards, for which T had recently obtained a large set of size data by measuring miseun specimens. Thus the optimal foraging aspect was a “eubmodel," and the properties of the submedel pero: mostly not considered. Here even more than in MacArthur and Pianka, the wore range of applicabflity of the optimal foraging model was scarcely imagined. Asa side note, about a year before publication of the optimal size paper I apent an afternoon at Princeton trying to float the model by Mackrthur and Henry Horn, At the end of @ lengthy and rather elaborate Presentation, Hackrthur said something to the effect that T now had come up with @ theory of optimal size, but that the final paper should be short, no Wnore than 12 printed pages, Disregarding thie completely, 1 published a 36-page paper, more in the style of « Holling than a MacArthur paper. (Holling was in fact one of the reviewers, and he called for mare informa- tion, implying a still longer paper! Richard Lewontin, vho waa then the editor of American Naturaliet, let me do what I wanted, #0 I did not expand. Thus @ compromise was reached between the two prevailing theoreti~ cal styles in ecology at the time, although rather more tovard one extrome than the other.) To elaborate on this interchange a bit mote, after the oral presen tion Macarthur felt that I probably had taken expected values incorrectly, and that what I really wanted was expectations of the e/t ratios for the individual food kinds, 1 thought about this for two weeks, did not see how Mackrthur could possibly be correct, and finally in desperation telephoned him to argue. After a while he agreed that his notion was a "red herring: in fact, here again » "fallacy of the averages" wae committed, although I am quite sure that were Nachrthur to have worked up an algebrate version of his graphical theoty, the etrot would not have been made. The mext event in the development of the contingency model for diet was the explicit writing out of the basic equation. This was done fn Schoener (1971) as the following (notation as in Schoener 1974p): ay where ty 1s the net energy gained per unit tine while feeding, e, Is the net energy (potential minus pursuit and handling-avallowing costs) for @ single iten of Type 1, t, fe the time to pursue, handle and suallow a single tten of Type i, py is the frequency of Type 1 in the environment (og + Ny/TN,, where Ny 48 absolute abundance for sone area, A), T, {6 the mean search time per item of available (encountered) food, and C, is the search cost per unit search time. T, = K/D, where 1/K is the rate of search and D ie overall food density (D = EN,/A), This summation over "diet" means only those items in the diet are included. As in previous presentations, the major mathematical properties of this equation were mostly not appreciated, although here at least the equation was put in contemporary context. The title of the 1971 paper was "Theory of Feeding Strategies," end except for the typical substitution of "foraging" for "feeding," was the first printed usage of the term that ts 412now so familiar. The paper, however, would not have existed were it not for the foresight of one or more persons on the editorial staff of Annual Review of Ecology and Systematics (Richard Johnston was editor). The paper specifically addressed OFT and its applications mostly within the realm of behavioral ecology. The lengthy reference and data base for the paper was pulled together in three months with a certain amount of agony under the usual publisher's gun, Of all ay papere, this one is listed the most (583 times) in the citation indices covering 1970 to 1983. During 1973 and 1974, @ curious trend developed with respect to the contingency model. It appeared, and reappeared, over and over again, at least six times. Each time the notation was different, and each time its Presentation seemed in partial or total ignorance of contemporary or atlice algebraic formulations (Charney 1973, Chernov 6 Orfans 1973, Pearson 1974, Timin 1973, Maynard Smith 1974, Pullism 1974, Werner & Hall 1974; a more complete treatment of Schoener's [1971] equation also appeared in Schoener 1974b). Some of the activity was clearly correlated: three unpublished documents emanated fron the University of Washington (Charnov 1973, Charnov & Ortans 1973, Pearson 1974), Other formulations were apparently done in total isolation (¢.g., Maynard Smith 1974), as if from another planet. Some of the derivations (Charny 1973, Gharnoy 6 Orians 1973, Pulliam 1974) were attempts to derive OFT from a sounder, more explicitly probabilistic framevork than before, such as renewal theory, Charnov's several papers (also see Charnov 1976a) were notable in fret clearly stating the mathematical properties of the contingency model ag we know them today, His equation (see aleo Pulliam 1974) 4s writeen in « renewal theory type of notation a Fa PAP: @) TTF On vhere E,/T is the ratio of net energy intake, 1, is the number of Prey Type 4 encountered in one unit of search time, B,# 1s the expected net energy from one iten of Type 4, hy* 1s the expected handling time for an item of Type {, P, is the probability that the predator goes after an iten of Type 4 when ome 19 encountered, and E, is the cost per-unit-tine of searching. This formulation combines presentations in Charnov snd Orians (1973) and Chernov (19762) and is identical to Equation 1 under the following equive- + and E, = 6. tence statements: Ay = py/Tyr ByAPy = eye hyMPy = ty 13,The properties of Equation 1 and 2 in somevhat more contemporary form are as follows: 1, To compute the optimal dict, items are ranked in decreasing order of e/t and the diet is computed for successively larger sete of deem kinds, beginning with chac diet having only the highest ranked item and proceeding to add item kinds in order of decreas- ing rank, The procesa continues until e,/t, for an item kind 18 smaller than the Y, for the diét with only higher ranked item Kinds, The diet then consists of all items with rank higher than Teen Kind x. 2, The greater the overall abundance of food, the more specialized the diet, provided proportions (p,'s) stay constant, 3, Whether or not a food type should be eaten is Independent of its own abundance and depends only on the absolute abundances of food ‘types of higher e/t, 4, Partial preferences do not exist, f.e., an item kind fe either always taken when encountered or never taken, ‘Charnov (1976a) modestly asserts that "the...model has been derived independently by several people, all of whom proved the theorem (Pulliam 1974, Werner & Hall 1974, Charnov 1973, Timin 1973, Pearson 1974, Schoener 19698, 1971, Naynard Smith 1974)." Tn fact, the theorem, as expressed in the four statenents above, was not proven in its entirety by at least sone ‘of those Listed (Schoener 1971 [also Schoener 1974b]; Werner & Hall 1974, Maynard Smith 1974, Timin 1973). As @ side note, Charnov's theoretical work on optinal dict was quite delayed in getting into press. A long document, "Optimal Foraging: some ‘Theoretical Explorations," by Charnov and Orians appeared in 1973 as an intended book. But in the face of very mixed reviews (only I really liked it), the book was withdrawn after submission to Princeton University Press. sol was 26 years old and all Charney was to have a In Charnov's (personal communication) words " these famous ecologists were jumping on my hea similar experience for hie “marginal value theorem," as detailed below, ‘The 1973-74 buret of papers fs interesting from several viewpoints. ‘The coincidence of the treatments and their virtual identity (except for 14search energy) might seen to imply that the time was right for the develop- ment of an optimal diet model for tts own sake. Indeed, all of the 1973-74 treatments except Timin's had as their motivatton the explanation of diets and food preferences per se, and many of them were partly or elaborately tested with specific behavioral data (see below). If the derivations were truly independent, their simultaneity ts Antereeting from a second, sociological viewpoint: how could persons dnterested in the same set of concepts, apparently then coming into some prominence, communicate se little that the fundamental model was derived so often from actateh? I do not know the answer, although I have some specu- lations. First, derivations were not truly out of the blu MacArthur and Pianka's graphical model as a starting point. Second, most cited Charnov and Pulliem, at least, vere trying to derive a foraging model from stochastic considerations, and they happened to end up with the same less rigorously derived earlier formulations. Because of differences in nota— tton, this was not alvays obvious. For example, in 1972 Pullian sent me a draft of his 1974 paper which claimed that his model was fundamentally different froa my 1969-71 model. Doubtless the reason for the confusion the ehange in notation (from Equation 1 to Equation 2), and in a later communication Pulliam wrote, "I agree that ny diet model...essentially we restates your earlier model, though it took me sone tine to see this. The value of the restatement 46 only in that the stochastde form predicts the variances (and higher moments) which are important for testing the model and lead to some non-obvious theoretical conclusions..." A eimtlar exchange of letters occurred with Charnov, during which he referred to the A third explanation for the rederivactons (e,g-, the reformulation of Werner & Hall) te chat the entire phenomenon as "rediscovering the wheel.’ initial derivations (espectally Schoener 1969a) were not fn particularly useful form—but then mefther vere sone of the rederivations (Maynard Smith 1974). A fourth explanation atens from the observation that ecological theore iefane often do not read one another's paper, perhaps from arro- gence, perhaps because it would spoil the problem-solving fun, Whatever the e7 se, progress in this field 1s not as monotonic as it could be. optimal diet theory thus became firmly established only by 1974, Any stut es done much before that time might not be considered legitinate tests, and indeed Krebs et al.'s (1983) review excludes then. Honetheless, the theory did seem conspicuously in agreement with # number of pre-1974 studies, at least qualitatively, To give two examples: 1) Ivlev's (1961) extensive experinents with fishes (e.g-. carp) showed that diets became 15more specialized, the more overall food density increased (in conformance with Property 2 above) and 2) Orians (1966) showed that during midday, parent yellov-hesded blackbirds (Xanthocephalue xanthocephalus) fed their young alnost exclusively on the helplees and large teneral (newly hatched) odonates, vhereas at other tines, when this especially profitable food item was unavailable, they depended on snaller and otherwise less desirable prey (in conformance with Property 3). The first @ priori test of predictions from an optimal diet model was that of Werner and Hall (1974), They presented bluegill sunfish (Lepomis werochirus) with three izes of Daphnia as potential prey. Their intent was to vary overall abundance (inversely related to T, in Equation 1) while keeping relative abundance constant. They first translated the actual abundance into that perceived by the fish: because of vioual Limitations, larger items can be seen farther avay, so such items have a greater encoun- ter rate than expected on the basis of actual availebilities. They then tested Fish under three abundance regimes and found that they behaved both qualitatively and quantitatively in accord with their version of Equation 1 (without search energy). The quantitative agreement obtained was perhaps especially surprising for 2 science whose practitioners used to deny that they sought precise models. As documented below, quantitative agreenent was not to be restricted to the Werner and Hall study, and indeed, not to be very restricted at all, Because they only varied overall abundances (1 each effective Ky changed by the same factor), Herner and Hall did not test perhaps the most interesting prediction of OFT, namely, that a food kind's own abun= dance does not determine whether it 18 ineluded or not fn the optimal diet: rather this is determined only by absolute abundances of more highly ranked itens. ws In 1975, Charnov (personal communication) gave a seminar at Oxford on the diet model, in response to which the model was labeled "trivial," bue also in response te which the above deficiency was pointed out. A skeptical Krebs, in combination vith @ hopeful Charnov, shortly thereafter, along with Erichson and Webber, designed what was to be the acid test (Krebs, Erichson, Webber & Charnov 1977}. Subjects were great tits (Parus major), stationed along conveyor belts bearing two sizes of “prey; in fact, these were pieces of mealvorus tuck together 4m two quantities in such a way as te increase handling tine significantly. The procedure was a hard-core attempt at falsification--prey were presented in several abundance sets: 161) evo sete such that both kinds should be selected by the average bird, according to OFT; 2) one set Just above the threshold such that the better (Gavger) prey only should be selected; and 3) two sets with the sane absolute abundance of large prey as in (2), but with greater or much greater quantities of the smaller prey. After the first set of tests, an ecstatic Charnov seribbled the results in a letter to Orians (dated October &, 1975), exclaiming that "the theory really works!" More subjects were run, and a clumsy individual with an exceptionally long handling time (the only one that was hend-reared) was the exception that proved the rule--OFT predicted it showld never specialize and dt never did. In all cases (5 birds in 5 tests), qualitative support or consistency was found for the model, although partial preferences existed (as in Werner & Hall), contra~ dicting Property 4 above (but see Krebs et al. 1983 for some rationaliz~ ing). Quantitative agreement for some individuals was very good, although Krebs et al. did not present randomly arranged prey, and there is sone uncertainty as to whether the non-random version of the diet theory they used was close enough to optimal to be valid for all tests. (This uncer tainty still exiats, as repeated failed attempts to dissipate it have shown (review in Pyke 1984]). OPTIMAL PATCH CHOICE: A NOT-SO-OPERATIONAL ANALOGY TO DIET Macarthur and Pianka's (1966) paper also dealt with how a forager should select patches vithin which to feed. Their approach, and subsequent algebraic formulations (Schoener 1974b), assumes that patches can be treated analogously to prey types, so far as encounter and value are concerned. For example, Equation | for diet is an equation for patches With the symbols redefined: T, 18 now the mean travel time between two. available patehes (whether entered or not), C, i the energy cost per unit travel tine, py ia the relative abundance of Patch Type 1, e, is the energy gained from feeding in a patch of Type i, and t, is the time it takes to gain that energy. Then an algebra mathematically identical to that for food types exists for patch types. While = formal analogy is possible, nature is euch that the patch nodel is less plausible than the diet model. For one thing, patch less likely to be encountered randomly in even the moderately long run, and hen random encounter does not occur, the contingency model fails to a greater or lesser extent (e.g., Pulliam 1974, MeWair 1979). For another thing, patches are not gobbled up 1fke food items; rather, they are de 7pleted gradually and typically abandoned when energetic return rate becomes too low (see below). The “marginal” rate of energy gain at which a patch ia abandoned {s a property of the environment in some models and teat that e and t are net conatant between environments. Moreover, by its om 20 action, the forager is changing the distribution of patches with various values of eft, This makes it difficult, even experimentally, to meet the assumptions of the simple model, Doubtless because of these problems (see also Pyke 1984), quantitative teste of the MacArthur-Pianka patch model and its algebrate counterparts were net performed in the mid-70's, and indeed, they are apparently still Jacking (Krebs et al, 1983, Pyke 1984), However, the qualitative idea that more profitable patches will be chosen over less profitable ones, consistent with the MacArthut-Pianka model as well as other models of the sane general nature, was tested in Light of the new theory. For exemple, Smith and Sweatman (1974) showed in saveral experiments that great tits generally preferred richer to poorer patches, measured either in terns of prey density or prey size. As usual, however, partial preferences existed, which can be argued to be inconsistent with the patch model (but again see Krebs et al. 1983), ‘The new way of looking at feeding in patches also caste a different Light on old de complete ignorance of MacArthur and Planks (1966)--it 1s not cited—Reyana (1970) reinterpreted Tinbergen’s (1960) data on birds feoding upon various species of caterpillars. Tinbergen found anong other things that the less abundant types of caterpillars in a pine woods were fed upon disproportton— ately less often by titaice (Farus). His explanation was that a search image existed, whereby the more abundant prey types vere more frequently For example, in what was possibly an analysis in reinforced and hence more easily located by the foraging bird. Tinbergen's data would seen to contradict the third prediction of optimal diet theory, that a food kind's own abundance does not affect its incluston in the diet. Reyama, on the other hand, convincingly argued that the caterpillar types cecurred in more-or-less separate cluups rather than being distributed randomly throughout the entire wood. The clumps can be thought of as patches, and were birds to spend more tine in denser patches, Tinbergen"s data could be explained as optimal patch use. This early attenpt to replace an ethological or psychological hypothesis with one in line with OFT hag been repeated a musbor of times since, with varying results, as we shall see below.The fact that foragers deplete patches rather than consume them as they do items is frustrating to experinenters, but it does provide a general rationale for the see of predictions now enshrined ae “the compres— " Derivation of the compression hypothesis vas an objec~ sion hypothests, tive of the Macarthur-Pianks (1966) paper, although it vas not so christened until MacArthur and Wilson published their Theory of Island Biogeography in 1967, The basic idea 1s simple and can be seen using Equation 1 (see also Schoener 1974b). Individuals of competing species, when invading the geographic range of a resident species, by assumption lower the food avatlabiliey for individuals of the latter, This heppens in evo ways, First, certain kinds of itens within a homogeneous area (a patch) become rarer. Second, certain kinds of patches become poorer in quality. If Equation 1 is interpreted as referting to food types, then this reduction can be seen to cause the diet to expand (4f the foods the competitor pre- fers are ranked highly enough by the resident) or stay the same, in accord= ance with Property 3 above. In contrast, if Equation I is interpreted as referring to pateh types, then the effect of a conpeting species will be to change the e,/t, values, posstbly altering the ranking of the pateh types. Since tt is a patch type's e/t chat ds cructal in determining its inclusion in the itinerary, a lowered e/t could easfly cause a patch type to be dropped by the resident. If other patch types are not added, an overall narrowing of breadth will resule, snd in any event the two species’ patch types will be " between the species in food types will increase or stay the same, whereas "compressed" away from one another. In short, overlap overlap in patch types will decrease as a result of the interaction. This scheme vag used to interpret certain sorts of biogeographical data. However, as pointed dut a nusber of tines subsequently (Schoener 19740, Schoener, Huey and Pianka 1979), it {8 typically valid ever behavicral but not evolutionary time, The reason is that over evolutionary time, selection may modify the e's and t's for food kinds (e.g-, by changing morphology) such that rank will also change for them and a net divergence engue. The hypothesis will be valid over evolutionary time for = parapatric species in the thin overlap region when characteristics of the organisms there are controlled mainly by gene flow from the outeide (Schoener et al. 1979). The biogeographic context into which this portion of OFT was initially placed may be responsible for why textbook authors frequently put the whole (or such) of OFT into chapters on communities or biogeography rather chan on behavioral ecology. Moreover, given its largely behavioral domain, it has always struck me as curfous that the coapression hypothesis vas experimentally tested so infrequently. Possibly the difficulty of