Ecology and Genetics of Color Vision in Callicebus brunneus, a Neotropical Monkey

Transcription

1 Ecology and Genetics of Color Vision in Callicebus brunneus, a Neotropical Monkey By JOHN ANDREW BUNCE B.S. Bates College 2002 M.A. University of California, Davis 2004 DISSERTATION Submitted in partial satisfaction of the requirements for the degree of DOCTOR OF PHILOSOPHY in Anthropology in the OFFICE OF GRADUATE STUDIES of the UNIVERSITY OF CALIFORNIA DAVIS Approved: Lynne A. Isbell David Glenn Smith Gerald H. Jacobs Committee in Charge 2009 i

2 Copyright by John Andrew Bunce All rights reserved. ii

3 TABLE OF CONTENTS ACKNOWLEDGEMENTS v ABSTRACT ix CHAPTER 1: General Introduction CHAPTER 2: Characterization and distribution of opsin gene alleles affecting color vision in a wild population of Callicebus brunneus, a Neotropical monkey CHAPTER 3: The Importance of Spatial Memory and Lactation for the Evolution of Color Vision in Callicebus brunneus, a Neotropical monkey CHAPTER 4: Intergroup dietary differences in a wild population of the primate Callicebus brunneus and a new hypothesis linking natal dispersal with the selective maintenance of polymorphic color vision CHAPTER 5: General Conclusions TABLES FIGURES iii

4 APPENDIX A: Callicebus of the California National Primate Research Center (CNPRC) APPENDIX B: Habituation and Capture Procedures APPENDIX C: Plant species constituting YOR patches discovered by adult Callicebus brunneus from January to October APPENDIX D: Compiled list of foods in the diet of Callicebus brunneus at the Estación Biológica de Cocha Cashu, Manu National Park, Madre de Dios, Peru REFERENCES CITED iv

5 ACKNOWLEDGEMENTS I thank my advisor Lynne Isbell for constant encouragement, generous support, insightful critique, sound advice, swift and thorough revisions, and illuminating discussion through seven years of graduate study. The original inspiration for this dissertation was a bit of foam born of the swirling torrent of her unconventional ideas. David Smith and Jerry Jacobs, the other two members of my dissertation committee, provided much needed support throughout this project. Their expertise, patience, generosity, and encouragement were instrumental, not only in helping me to refine my ideas, but also in making this learning process enjoyable. In particular, David generously invited me into the UC Davis Molecular Anthropology Laboratory, allowing me the valuable opportunity to learn to genotype my own monkey samples and come to appreciate a well formed fecal pellet. Mark Grote invested considerable time and provided invaluable help in the early stages of this project with experimental design, and later with statistical analyses (although any errors are completely my own). This dissertation benefited immensely from Mark s expertise, breadth of interest, clarity of thought, patience, and extraordinary talent for teaching. This project owes much to Peter Rodman. Soon after entering the graduate program, Peter invited me to spend several academic quarters working with the data of Francis Bossuyt (below). This experience drove home the usefulness of electronic data collection methods, and eventually resulted in the data collection program for handheld computers that I wrote with advice from Michael Sokol and Peter. In 2003, Peter bought the pneumatic dart gun that I used throughout the field portion of this project. Unbeknownst to him, this allowed me the valuable opportunity, before going to the field, to practice my marksmanship in the long basement corridors of Young Hall in the early hours of the morning. Peter also loaned me the telemetry equipment and guitar amplifier used to locate the animals. His practical advice about fieldwork v

6 in Neotropical forests and his encouragement during this project were of great importance to its successful completion. Bill Mason, Sally Mendoza, Greg Vicino, and Carmel Stanko of the California National Primate Research Center (CNPRC) generously allowed me access to the Callicebus colony at the CNPRC in order to learn how to collect blood and hair samples for genetic analysis. This experience proved to be invaluable in the field. For generous help and support in genotyping, I thank the kind and patient denizens of the Smith Molecular Anthropology Lab of the University of California, Davis, especially Debbie George, John McDonough, Jessica Satkoski, Meradeth Snow, Wendy Garnica, Joy Erickson, and Cara Monroe. Additionally, Alison Surridge of the University of East Anglia (U.K.) and Maureen Neitz of the Medical College of Wisconsin generously imparted advice and duplicated the genotyping analysis as a check on my novice efforts. This dissertation was greatly improved through enlightening discussions with many people (in addition to those mentioned above), especially Amy Porter, Andy Marshall, Sandy Harcourt, the UC Davis Simian Seminar group, Bruce Winterhalder, Nate Dominy, Henry McHenry, Alison Surridge, Nick Mundy, Andrew Smith, and Caissa Revilla. I also thank the extraordinary staff of the UC Davis Department of Anthropology, especially Lucy Day, Peggy Slaven, Nancy McLaughlin, Royce McClellan, Barbara Raney, Edie Stasulat, Candy Clark, and Jeff Trask. Perhaps no other group of people was more instrumental in the success of this project than my coworkers in Peru. Caissa Revilla Minaya (2004, 2006), Inés Nole Bazán ( ), Amy Porter ( ), Cecilia Carrasco Ramírez (2004, 2005), Guillermo Añí Figueroa (2005), Edilberto Castro Gómez (2004), and Natalia Quinteros Casaverde (2003) all chased Callicebus vi

7 with me. Their dedication and enthusiasm were unfailing through rain, swamps, fermented lunch, giardia, malaria, bot flies, tick infestations, and unidentified worms in unmentionable places. Each brought his or her own particular area of expertise to the project. I consider it a great experience and an honor to have worked with them. I also thank the other longer term residents of Cocha Cashu, as well as those who passed through, for their support, advice, and for preventing us from taking ourselves too seriously. As part of his doctoral research on natal dispersal, Francis Bossuyt studied the Callicebus of Cocha Cashu intensively from 1996 until his tragic disappearance at the field station in All five monkey groups in this study were observed by Francis, and three groups still contained individuals bearing radio and identification collars fit by him. Francis detailed notes of Callicebus behavior, territory locations, and capture protocols were invaluable in the design and implementation of this project. His memory inspires even those of us who never had the chance to meet him. Several aspects of this project hinge on accurate plant identification. Fortunately, at Cocha Cashu I had access to experts who were both extremely generous with their time and patient with a neophyte. I am especially grateful to Patricia Alvarez Loayza, who not only helped identify specimens in the field, but also invested considerable time and effort matching my plant photographs to herbarium specimens at the Field Museum in Chicago. Tim Paine, Varun Swamy, John Terborgh, and Kyle Dexter also helped in the identification of plant specimens at the field station. Any errors in plant identification presented here are entirely the fault of the author. For botanical sample collection from trees taller than I could climb, I am grateful to Julian Huarancasi, who, I am convinced, can climb anything in the forest. For permission to work at the field station, as well as logistical support, I thank John Terborgh, Carmen Chávez, Rebekah Hren, and especially Verónica Chávez. I am grateful to Doña vii

8 Marta for turning limited provisions into good tasting food. For considerable help with Peruvian research permits and sample exportation I am grateful to Carmen Chávez, Verónica Chávez, Caissa Revilla, Miriam Minaya, Pachi Fuentealba, Natalia Quinteros, and Neri Fernandez. For generous hospitality in Peru, I thank Caissa and Natalia Revilla, Miriam Minaya, and Oscar Revilla, as well as Cecilia Carrasco, Berner Torres, Natalia Quinteros, Rowena Cerro, and Señor Klaus of Atalaya. Permission to conduct this investigation in Peru, as well as to export genetic samples to the U.S. for analysis, was granted by the Instituto Nacional de Recursos Naturales del Perú (INRENA), with the particular help and attention of Jessica Espinoza, Carmen Jaimes, and Karina Ramirez. This project benefitted from collaborative agreements with Daniel Gonzales Gamarra of the Universidad Nacional San Antonio Abad de Cusco, Cusco, Peru, and Martha Williams León de Castro of the Universidad Nacional Agraria La Molina, Lima, Peru. Various phases of this research were supported by UC Davis Department of Anthropology Summer Research Fellowships, Sigma Xi Society Grants in Aid of Research, the Organization for Tropical Studies Francis Bossuyt Memorial Fellowship, the University of California Marjorie and Charles Elliott Fellowship, the UCD and Humanities Graduate Research Award, a General Grant from the Leakey Foundation, and an Individual Research Grant from the Wenner Gren Foundation. The writing of this dissertation was greatly aided by a University of California Dissertation Year Fellowship. through this, Finally, I thank my family, especially my mother, father, and brother, for helping me and Caissa, for always being with me. viii

9 ABSTRACT The trichromatic color vision of many primates allows an individual to make distinctions among colors such as green, yellow, orange, and red. This capacity is unique among diurnal placental mammals. Most platyrrhine monkeys are characterized by X linked polymorphic trichromacy in which heterozygous females are trichromatic, while homozygous females and all males are dichromatic and thus unable to make trichromatic color discriminations. This dissertation examines the mechanisms through which natural selection maintains polymorphic trichromacy in a wild population of the platyrrhine monkey Callicebus brunneus in southeastern Peru. Genetic samples from five socially monogamous C. brunneus groups revealed the presence of three X linked photopigment gene alleles similar to those found previously in other platyrrhines. There was no strong evidence of disassortative mating or differential short term reproductive success among groups with dichromatic versus trichromatic adult females. This is contrary to the expectation if trichromats enjoy a fitness advantage over dichromats. The longdistance detection of yellow, orange, and red (YOR) food patches is hypothesized to be easier for trichromats than for dichromats. However, an analysis of C. brunneus foraging behavior revealed that most YOR food patches may have been located by means of spatial memory and non color cues, thereby minimizing any potential trichromatic advantage in encounter rate. For YOR patches unlikely to have been located by means of spatial memory, no difference in the encounter rates of di and trichromatic females was detected. However, there was a nonsignificant trend for lactational status to differentially affect the encounter rates of di and trichromatic females. These results are explained by a new hypothesis for the maintenance of polymorphic trichromacy in C. brunneus populations through temporally variable selection. This hypothesis implicates spatial memory of food resources and female dispersal behavior, as well ix

10 as temporal fluctuation in food abundance and predation, in the action of selection on color vision. The application of the hypothesis of temporally variable selection for trichromacy to other primates is discussed. Understanding how selection acts on color vision in extant primate populations is the first step in the development of hypotheses to explain the evolution of trichromatic color vision in our primate ancestors. x

11 1 CHAPTER 1 General Introduction The ability to perceive color differences in natural scenes is an important part of the lives of most humans, both for utilitarian reasons (e.g., detecting food edibility) and because color often plays a central role in many aspects of cultural expression (e.g., dress). With the exception of those with congenital or acquired color vision abnormalities, the vast majority of people have trichromatic color vision in which colors such as violet, blue, green, yellow, orange, and red can be distinguished in photopic light environments (e.g., daylight conditions) (Neitz and Neitz 2000; Sharpe et al. 1999). Trichromacy is a capacity that separates humans from most other diurnal placental mammals, the majority of whom have dichromatic vision and may have difficulty distinguishing among colors such as green, yellow, orange, and red in many contexts (Ahnelt and Kolb 2000; Jacobs 1993; Sharpe et al. 1999). As with evolutionary studies of other human morphological traits, such as opposable thumbs, bipedalism, and large brains, examination of the evolution of trichromacy can lead to a more complete understanding of our place in nature. Additionally, evolutionary studies of trichromacy contribute to an explanation of why humans perceive the world in a manner so different from that of most other animals. In photopic light environments, color vision is made possible by neural comparisons among retinal cone cells containing opsin proteins absorbing light over defined ranges of wavelengths (Jacobs 2008). The dimensionality of color vision (e.g., di or trichromacy) can often be inferred from the number of different kinds of opsins employed in such neural comparisons (although caution should be exercised in such inference). For instance, tetrachromacy, trichromacy, and dichromacy usually result from comparisons among classes of cones containing one of four, three, or two (respectively) functionally distinct opsin types (Jacobs 2008;

12 2 Neumeyer 1991). Ancestral vertebrates may have been characterized by cones containing four kinds of opsin (Collin and Trezise 2004), homologues of all of which have been retained in some fish, reptiles, and birds. In these animals, such a configuration yields tetrachromatic color vision (i.e., possibly allowing many more hue discriminations than humans can make) (Neumeyer 1991). Some marsupials may have retained three of the ancestral vertebrate opsins, and two species have been shown to be trichromatic (Arrese et al. 2006; Arrese et al. 2002). However, early in the evolution of placental (eutherian) mammals, two of these four ancestral opsin classes appear to have been lost. The resulting dichromacy perhaps reflects an early shift from a diurnal to a nocturnal niche, where cone based trichromacy may have been less useful (Jacobs 1993; Jacobs and Rowe 2004). As a result, the retinas of most extant diurnal eutherian mammals contain only two types of cone cell, containing either a short wavelength sensitive (S) opsin or a long wavelength sensitive (L) opsin with peak light absorbance in the range of approximately nm and nm, respectively (Jacobs 1993; Jacobs and Rowe 2004; Yokoyama and Radlwimmer 1999). Intriguingly, primates represent the only known eutherian lineage in which at least some members have re evolved trichromacy (Jacobs 1993). In some ancestral primates, this seems to have been accomplished by the development of a polymorphism at the location of the X linked gene coding for the L opsin (also known as the M/L opsin). The fact that this gene occurs on the X chromosome leads to a situation where heterozygous females, having a different L opsin allele on each of their two X chromosomes, are trichromatic, while homozygous females and all males have only a single L opsin allele and are dichromatic (reviewed in Jacobs 2007; Jacobs 2008; Surridge et al. 2003). There is currently much debate about whether this visual configuration, known as polymorphic trichromacy, evolved once in a basal primate and was subsequently lost in several extant lineages (Tan et al. 2005), or evolved independently several times (Heesy and

13 3 Ross 2001). However, in any case, polymorphic trichromacy appears to be the norm in extant Neotropical (platyrrhine) monkeys, and several Malagasy lemurs (strepsirrhines) (Jacobs 2007; Jacobs et al. 2002; Leonhardt et al. 2008; Tan and Li 1999; Veilleux and Bolnick 2009). Between approximately 43 and 31 million years ago, shortly after the platyrrhine and catarrhine (Old World monkeys, apes, and humans) lineages diverged (Steiper and Young 2006), a basal catarrhine is believed to have undergone a duplication of the L opsin gene onto an adjacent region of the X chromosome (Dulai et al. 1994; Hunt et al. 1998; Nathans et al. 1986b; Sharpe et al. 1999). It is unknown whether this basal catarrhine species was strictly dichromatic, in which case, after duplication, the sequence of one gene copy diverged such that it came to code for the functionally distinct M (middle wavelength sensitive) opsin, or whether this catarrhine was a polymorphic trichromat, in which case the duplication may have placed two different opsin alleles onto a single X chromosome such that functionally distinct M and L opsins could be expressed immediately (reviewed in Regan et al. 2001; Surridge et al. 2003). In either case, after the L opsin gene duplication, ancestral catarrhines produced three classes of opsin (S, M, and L) in their retinal cones, facilitating the routine (i.e., present in all individuals of both sexes) trichromatic color vision which appears to have been maintained to the present in all extant catarrhine primates (Jacobs and Deegan 1999; Jacobs and Williams 2001; Onishi et al. 1999). Despite the complex evolutionary history of color vision in primates, little is currently known about what may be (or was) the selective advantage of one visual configuration over another. Allen (1879) proposed that the color vision of many animals is (or was originally) intimately linked to their ecology, specifically, foraging for colorful plant products (e.g., fruit or flowers). In recent years, this hypothesis has been refined to posit that the evolution of trichromacy within primates has been driven by a fitness advantage of trichromats over

14 4 dichromats, realized through an enhanced ability to detect yellow, orange, or red objects (e.g., food, predators, conspecifics) against the dappled green background of tropical forest leaves (Mollon 1989). Although support for this hypothesis has come from studies modeling the color vision of primates and colorful ecologically important targets (de Araujo et al. 2006; Dominy and Lucas 2001; Lucas et al. 2003; Osorio et al. 2004; Osorio and Vorobyev 1996; Regan et al. 2001; Riba Hernandez et al. 2005; Riba Hernandez et al. 2004; Stoner et al. 2005; Sumner and Mollon 2000a), as well as from captive experiments isolating the effect of color vision on the detection of yellow and orange food objects (Caine and Mundy 2000; Leonhardt et al. 2008; Smith et al. 2003b), as yet no study of wild primates has demonstrated a convincing advantage of trichromats over dichromats for visually detectable targets in a natural environment (Dominy et al. 2003a; Hiramatsu et al. 2008; Melin et al. 2008; Smith et al. 2003a; Vogel et al. 2007). The present study investigates the hypothesized selective advantage of trichromacy over dichromacy in a wild population of the platyrrhine primate Callicebus brunneus. In Chapter 2 I present evidence that these monkeys, like most other platyrrhines, are characterized by polymorphic trichromacy, and that reproductively successful di and trichromatic females occur in the same population. In Chapter 3 I compare the foraging behavior of di and trichromatic females with regard to yellow, orange, and red (YOR) food patches. I show that very few YOR food patches exploited by C. brunneus are likely to be encountered without the aid of spatial memory, thus likely minimizing any advantage of trichromacy over dichromacy in terms of food patch encounter rate during routine foraging behavior. There is some suggestion, however, that a trichromatic advantage in the encounter rate for small, ephemeral YOR patches may be realized when females are energetically stressed due to heavy lactation. In Chapter 4 I develop a new hypothesis to explain how natural selection could maintain polymorphic trichromacy in C. brunneus populations, despite seemingly comparable short term reproductive success (Chapter

15 5 2) and food patch encounter rates (Chapter 3) among di and trichromatic study animals. This hypothesis of temporally variable selection implicates dispersal behavior in the maintenance of polymorphic trichromacy in C. brunneus, and is plausibly applicable to several other primates with this visual configuration. Understanding how selection acts on color vision in extant primates is the first step in the development of hypotheses addressing how selection has shaped the evolution of color vision in ancestral primates, and, ultimately, contributes to answering the question of why most humans, unlike the majority of placental mammals, are able to perceptually assign a variety of color attributes, such as green, yellow, orange, and red, to objects in the natural world.

16 6 CHAPTER 2 Characterization and distribution of opsin gene alleles affecting color vision in a wild population of Callicebus brunneus, a Neotropical monkey ABSTRACT The color vision of most platyrrhine primates is determined by alleles at the polymorphic X linked gene locus coding for the M/L retinal photopigment. Females who are heterozygous at the locus have trichromatic vision, potentially allowing them to distinguish among colors such as green, yellow, orange, and red that homozygous females and all males, all of whom carry a single allele type and are therefore dichromatic, might be expected to confuse. The number, frequency, and sequence of alleles at the M/L locus vary among platyrrhines, suggesting this may hold important implications for the way selection acts on color vision in these primates. This study represents the first reported characterization of M/L photopigment alleles in a wild population of the socially monogamous platyrrhine monkey Callicebus. Using blood, saliva, and fecal samples, I found evidence of three functionally distinct alleles, occurring at relatively high frequencies, and resulting in Callicebus pairs containing both di and trichromatic individuals. I found evidence of neither disassortative mating nor of a trichromatic advantage in short term reproductive success, as might be expected if there is strong selection for trichromacy over dichromacy in the study population. The results of this first analysis of the color vision of individual Callicebus in the wild suggest several avenues for future investigation of selection on color vision in this genus, including the role of extra pair copulations and habitat selection.

17 7 INTRODUCTION The diversity of color perception within the order Primates appears to be exceptional among mammals (Jacobs 1993). The ability of most humans, apes, and Old World monkeys (Catarrhini) to routinely distinguish among colors such as green, yellow, orange, and red is the result of a visual configuration known as trichromacy (Jacobs 2008; Sharpe et al. 1999). The color vision of other primates, however, is considerably more variable. Many Malagasy lemurs (Lemuroidea) and tarsiers (Tarsius), like the majority of diurnal placental mammals (Eutheria), appear to have dichromatic vision, in which colors such as green, yellow, orange, and red are apt to be confused, such that any of these colors can be matched to any other under appropriate ambient light conditions (Ahnelt and Kolb 2000; Hendrickson et al. 2000; Jacobs 1993; Jacobs 2008; Jacobs and Deegan 1993; Sharpe et al. 1999; Tan and Li 1999; Tan et al. 2005). Several nocturnal and cathemeral primates, including the strepsirrhine bush babies (Galago, Otolemur) of Africa and lorises (Nycticebus) of Asia, as well as the Neotropical (platyrrhine) night monkey (Aotus) are monochromats, able to make only luminance (rather than color) discriminations in many viewing situations (Jacobs et al. 1993a; Jacobs et al. 1996b; Kawamura and Kubotera 2004; Tan and Li 1999; Tan et al. 2005). One unusual lemur, the nocturnal aye aye (Daubentonia), appears to be a dichromat, although there is a suggestion that it may have the ability, unique among primates, to see in the ultra violet region of the spectrum (Hunt et al. 2007; Perry et al. 2007). In another variation, the platyrrhine howler monkey (Alouatta) is unique among noncatarrhine primates in having independently evolved full trichromacy, comparable to that of humans (Araujo Jr. et al. 2008; Hunt et al. 1998; Jacobs et al. 1996a; Kainz et al. 1998). However, one of the most interesting primate visual configurations, apparently shared by all platyrrhine monkeys (except Aotus and Alouatta, above) and several lemurs, is polymorphic trichromacy, a condition that results in high frequencies of dichromats and trichromats in the same population

18 8 (Jacobs 2007; Jacobs et al. 2002; Leonhardt et al. 2008; Mollon et al. 1984; Tan and Li 1999; Veilleux and Bolnick 2009). Polymorphic Trichromacy The color perception of polymorphic trichromats is the result of neural comparisons among the outputs of retinal S cone cells and those of one (dichromatic individuals) or two (trichromatic individuals) classes of retinal M/L cone cells (Jacobs 2007; Surridge et al. 2003). Cone cell light absorption is facilitated by photopigment proteins, known as S (short) and M/L (medium/long) opsins, the names referring to the relative wavelengths of peak absorbance. The primate S opsin is most sensitive to light at approximately 430 nm (except perhaps Daubentonia, above) and is controlled by an autosomal gene (Bowmaker et al. 1991; Mollon et al. 1984; Nathans et al. 1986a). M/L opsins are controlled by a single polymorphic gene on the X chromosome comprising (usually) two or three functionally distinct alleles coding for opsins with maximal absorbances between approximately 530 nm and 565 nm (Jacobs 2007; Jacobs et al. 1993b; Kawamura et al. 2001; Mollon et al. 1984; Neitz et al. 1991). As a result, males, who are hemizygous, and females who are homozygous, produce only one variant of M/L opsin. Comparisons of the outputs of S cones with a single class of M/L cone yields dichromacy. Heterozygous females, however, produce two variants of M/L opsin which segregate into different M/L cone cells in the retina. Together with the S cones, these two classes of M/L cone facilitate trichromacy (reviewed in Jacobs 2007; Jacobs 2008; Surridge et al. 2003). Thus, in wild primate populations of polymorphic trichromats, males can be one of several varieties of dichromat, while females can be one of several varieties of dichromat or trichromat (Jacobs 2007; 2008). In platyrrhines, the similarity and conservative maintenance of this polymorphism in nearly all extant genera suggest that it originated at least 20 million years ago (mya) in a basal

19 9 platyrrhine (Boissinot et al. 1998; Hunt et al. 1998) and is under stabilizing selection (Surridge and Mundy 2002). The selective advantage of trichromacy to wild primates is currently the subject of debate. A popular scenario is that trichromacy facilitates the detection of yellow, orange, and red food items (e.g., fruits and young leaves) against the green foliage background of tropical forests (Mollon 1989; Sumner and Mollon 2000a). Photopigment Alleles The association between the absorbance properties of platyrrhine M/L opsins and nucleotide changes in M/L gene alleles has been well studied. Non synonymous changes at only a few sites in three exons of the M/L gene cause the wavelength of opsin peak light absorbance to vary over a range of about 30 nm. For platyrrhines, the most important of these sites (where changes cause the greatest shift in peak absorbance) are 180 in exon 3 and 277 and 285 in exon 5, although several other sites may have minor effects (Asenjo et al. 1994; Hiramatsu et al. 2004; Neitz et al. 1991; Shyue et al. 1998; Yokoyama 2000). Table 2.1 shows opsin peak absorbances and corresponding allele haplotypes discovered in populations of representative platyrrhine monkeys, as well as the separate M and L opsin genes of Alouatta (howler monkeys) and humans. For platyrrhines, the correspondence between M/L haplotype, opsin expression in the retina, and behavioral evidence of color perception is so consistent that the color perception of these primates can often be inferred directly from their genotypes (Jacobs 2007; Jacobs et al. 1996a; Mollon et al. 1984; Saito et al. 2005a; Tovee et al. 1992; Williams et al. 1992), though this should be done cautiously (e.g., Jacobs 1993; Jacobs 2008; Leonhardt et al. 2008).

20 10 M/L Alleles and Selection Given the close genotype phenotype relationship in the platyrrhine color vision system, several authors have investigated the frequencies of alleles on a large scale in both wild and captive populations in order to draw conclusions about how selection operates on the polymorphic opsin locus (Cropp et al. 2002; Surridge et al. 2005b). For instance, if trichromatic color vision provides an absolute advantage over dichromatic vision, then selection should maximize the number of functionally distinct alleles at the M/L locus in platyrrhines with polymorphic trichromacy. This would increase the proportion of heterozygous females born into the population at Hardy Weinberg equilibrium. Similarly, a fitness advantage for trichromats would be expected to result in frequency dependent selection on alleles, i.e., an individual with a rare allele would have a relative fitness advantage because of an increased probability of siring a heterozygous trichromatic daughter. Thus, strong selection for trichromacy could, theoretically, result in a large number of M/L alleles with approximately equal frequency in the population. However, no more than three functionally distinct high frequency alleles at the M/L locus (resulting in three distinct classes of retinal M/L cones) is the norm for platyrrhine populations, though these three alleles may differ among species (reviewed in Jacobs 2007). Cropp et al. (2002) suggested that the absence of many more opsin alleles at the locus may be the result of a tradeoff between selection for trichromacy over dichromacy and selection for trichromats with wide separation of M/L opsin light absorbance maxima, coupled with constraints on shifting the peak light absorbance of the M/L photopigments beyond the range of about nm (e.g., a functional limit on peak absorbance in longer wavelengths and limits on the separation of peak absorbances due to quantum noise and chromatic aberration: Sumner and Mollon 2000a). The conclusion that trichromats with widely spaced (up to a point) M/L peak absorbances have an advantage over trichromats with more closely spaced pigments in

21 11 detecting colorful fruits and leaves against green leafy backgrounds, especially under dim illumination, is supported by the results of several target detection studies modeling the vision of wild primates (Osorio et al. 2004; Regan et al. 2001; Rowe and Jacobs 2004; Rowe and Jacobs 2007; Sumner and Mollon 2000a). Thus, a primate population with a greater variety of distinct M/L pigments having peak absorbances within the range of nm would have a higher probability of producing heterozygous females with closely spaced pigments and presumably inferior trichromatic color vision. In this sense, limiting the number of alleles in the population to three (and no more) may be close to the optimal equilibrium (Cropp et al. 2002). However, this argument fails if trichromacy, of any form, is always superior to dichromacy for ecologically relevant tasks (e.g., Sumner and Mollon 2000a). Thus, an alternative possibility is that there are some important tasks for which dichromacy is superior to trichromacy (Dominy et al. 2003b; Melin et al. 2007; Morgan et al. 1992; Saito et al. 2005b), for instance, discerning the form (or texture) of an edible insect or fruit whose yellow and green coloration matches that of the yellow and green background against which it is found, thereby potentially distracting a trichromat from detection on the basis of form. In this case, the spread of additional M/L alleles through the population could be impeded by frequency dependent selection at the level of the individual, favoring high proportions of both di and trichromats which excel at different tasks or in different micro environments (Mollon et al. 1984). Another important observation is that M/L alleles often occur in unequal frequencies when averaged across populations of wild and captive platyrrhines. This is contrary to the expectation of equal frequencies (when averaged across many independent populations) if selection acts to maximize the number of heterozygous trichromatic females born into populations (see above). For instance, pooling analyses of 326 X chromosomes from several genera of callitrichine platyrrhines (tamarins and marmosets), each with three M/L variants

22 12 having peak absorbances at approximately 543 nm, 556 nm, and 563 nm, Surridge et al. (2005b) found that the overall frequency of the 556 nm variant (0.17) was less than half that of the other two variants (543 nm: 0.39, 563 nm: 0.44). When compared to equal allele frequencies, these skewed frequencies have the effect (under Hardy Weinberg equilibrium and a balanced sex ratio) of decreasing the total percentage of heterozygous trichromats in the population from 33% to 31%. However, perhaps more importantly, such allele frequencies would also result in an increase in the percentage of trichromats with the widest possible separation of M/L pigments, in this case 543/563 nm trichromats, from 11% to 17% of the population (calculations following Rowe and Jacobs 2004). Surridge et al. (2005b) suggest that skewed callitrichine allele frequencies may be the result of a balance between selection for trichromats over dichromats, selection among trichromats for those with the widest peak light absorbance separation, and perhaps selection among dichromats for those with longer wavelength M/L pigments whose light absorbance is most different from that of the S pigment (e.g., Osorio et al. 2004). In a similar analysis for the single genus Saimiri (squirrel monkey), having three M/L variants with peak absorbances at approximately 535 nm, 550 nm, and 562 nm, Rowe and Jacobs (2004) pooled analyses (and inferences) of 362 X chromosomes and did not detect differences in the frequencies of the three photopigment variants. However, Cropp et al. (2002), examining wild and captive individuals of three Saimiri species, found a deviation from equal allele frequencies in one species, Saimiri oerstedii, after having examined 70 S. oerstedii X chromosomes. Interestingly, in contrast to the callitrichines, the M/L variant with the shortest peak light absorbance, 535 nm, was the least common in the S. oerstedii population. Cropp et al. (2002) suggest that unequal allele frequencies in this endangered Saimiri species are most likely the result of sampling error or recent genetic drift.

23 13 Aside from M/L allele frequencies at the species level, allele frequencies in local primate populations may be heavily influenced by chance events and mating behavior. For instance, Hiramatsu et al. (2005) found very different M/L allele frequencies in neighboring 20 to 30 member groups of Cebus (capuchins), which they attributed to intense inbreeding in one group. Thus, there may be few instances in wild primate populations where the number of heterozygous trichromats is maximized through equal allele frequencies. This could be due to non random mating, drift, and/or a balance between selection for trichromats over dichromats and selection within trichromats and dichromats (above). Callicebus Trichromacy In terms of polymorphic trichromacy, one of the most unusual platyrrhines may be Callicebus, the titi monkey or mono tocón. In contrast to the two or three M/L photopigment variants found in most other platyrrhine populations, five M/L variants have been reported in Callicebus. Jacobs and Deegan (2005) used electroretinogram flicker photometry to measure the absorbance of M/L cones in the retinas of 82 Callicebus individuals at the California National Primate Research Center (CNPRC), USA (see Appendix A for colony history and species identification). They found evidence of five distinct classes of M/L opsins having peak absorbances at approximately 530 nm, 536 nm, 542 nm, 551 nm, and 562 nm, occurring at population frequencies of 0.068, 0.203, 0.085, 0.288, and 0.356, respectively. As shown in the left hand side of Table 2.1, photopigments corresponding to each of these five M/L variants have been detected in other platyrrhine species. However, as noted by Jacobs and Deegan (2005), no other primate population is known to contain all five M/L variants simultaneously. Unusual M/L photopigments, created as the result of recombination events between more common photopigment variants, have been previously reported in wild primate populations. For

24 14 instance, in a sample of 93 X chromosomes from 54 individuals in a wild population of Saimiri boliviensis having three common M/L variants with peak absorbances at approximately 535 nm, 550 nm, and 562 nm (Mollon et al. 1984), Cropp et al. (2002) found a single example of a recombinant allele (i.e., splicing of a 535 nm allele with a 562 nm allele) presumably coding for an M/L pigment with a peak absorbance of 558 nm. Thus, by itself, the presence of more than three M/L variants in a platyrrhine population is not unexpected. However, what is striking about the CNPRC Callicebus population is that all five M/L variants occur at relatively high frequencies. This has the consequence of potentially increasing the number of heterozygous trichromatic females born into the population. Indeed, Jacobs and Deegan (2005) found over 80% of the Callicebus females to be trichromats, which is close to the 73% predicted under Hardy Weinberg equilibrium with the observed allele frequencies. There are currently no genetic or retinal absorbance data pertaining to M/L opsin variation in wild Callicebus populations, and it is unknown whether the presence and frequency of the five M/L opsins found in this captive Callicebus colony are artifacts of captivity (e.g., captive breeding and species/subspecies hybridization, see Appendix A). However, if such M/L variation is representative of wild populations, it would suggest the intriguing possibility (among others) that selection for trichromacy over dichromacy is considerably stronger in Callicebus than it is in other platyrrhine primates with polymorphic trichromacy. If true, this would point to Callicebus as an important subject to investigate the, as yet poorly understood, selective advantage of trichromacy in wild primate populations (see Chapter 3). Behavioral evidence of strong selection for trichromacy over dichromacy can manifest in several ways. For instance, individuals could increase the likelihood of producing heterozygous trichromatic daughters by mating with individuals who carry M/L alleles different from their own (i.e., disassortative mating). Surridge et al. (2005a) found evidence of disassortative mating

25 15 in a wild Saguinus population, in which social and reproductive groups of between three and five adults were formed such that M/L alleles present in one sex tended to be absent in the other sex at frequencies greater than chance. Surridge et al. (2005a) argued that this configuration was most likely the result of inbreeding avoidance, but that it had the effect of increasing the frequency of heterozygous females in the population. An additional observation of Surridge et al. (2005a) was that the single unstable and unsuccessfully reproducing Saguinus group in their study (of eight groups) did not contain heterozygous trichromatic adult females (although the random probability of such an observation was not determined). Taken together, these results suggest that trichromatic Saguinus may have a fitness advantage. Given the potential for strong selection for trichromacy in wild Callicebus populations, I hypothesize that such patterns of disassortative mating and differential reproductive success will be at least as pronounced in wild populations of Callicebus as they appear to be in this Saguinus population. The present study has two objectives. First, I characterize the allele haplotypes of the polymorphic M/L opsin gene in a wild population of Callicebus and attempt to match these haplotypes to the M/L photopigment absorbances recorded by Jacobs and Deegan (2005) in captive Callicebus. Second, I look for evidence of strong selection for trichromacy over dichromacy in the wild Callicebus population by noting the degree to which individuals mate disassortatively with regard to M/L alleles and by comparing short term reproductive success of di and trichromatic females. METHODS Study Site and Subjects The Estación Biológica de Cocha Cashu comprises a 10 km 2 study area located at approximately 350 m elevation in the interior of Manu National Park in the department of

26 16 Madre de Dios in southeastern Peru (11 52 S, W). The site lies within the 6 8 km wide meander belt of the Manu River and consists of a patchwork of seasonally flooded riparian and lacustrine successional stages amidst mature high ground tropical forest (Terborgh 1983). Mean annual rainfall is approximately 2000 mm, with most rain usually falling between October and April (the wet season). Mean annual temperature is approximately 23 C (range 13 C and 33 C) (Terborgh 1983; 1990). The immediate vicinity of the study site has been protected from hunting and logging since 1973 (Terborgh 1999) and contains a community of eleven primate species (Terborgh 1983). Brown titi monkeys, or monos tocones (Callicebus brunneus, Hershkovitz 1990), are small (approximately 1 kg), socially monogamous, arboreal omnivores with slight sexual dimorphism, and generally unspecialized morphology (Hershkovitz 1990; Kinzey 1981; Wright 1984). Callicebus brunneus devote most foraging time to fruit, but also eat leaves and insects (Kinzey 1981; Wright 1985, Chapter 3). The social unit consists of an adult female and male and their dependent offspring (Wright 1984), which together defend exclusive territories of 5 12 ha at this field site (Terborgh 1983 and Figure 2.1). During the majority of the study period group sizes ranged from two to three animals, not including dependent infants. Data and Sample Collection Five groups of C. brunneus in the vicinity of the research station were observed between August 2005 and October 2006 as part of a larger behavioral study (Chapter 3). All births, disappearances, and presumed subadult natal dispersal events were recorded during this time. Between August 2005 and March 2006, three adult females, one adult male, and two subadults from four groups were captured via darting and fit with collars to aid in individual identification (Appendix B for capture protocol). Several other individuals in three groups were identifiable by

27 17 means of collars fit as part of a previous investigation (Bossuyt 2002; Rodman and Bossuyt 2007, A. Porter, pers. comm.). Blood extracted from the femoral artery or saliva swabbed from the tongue and gums of captured animals was collected and dried (at ambient temperature) on FTA Classic Card nucleic acid preservation material (Whatman plc., Brentford, Middlesex, UK) (Smith and Burgoyne 2004). These samples were stored at ambient temperature with CaSO 4 CoCl 2 indicating desiccant (W. A. Hammond Drierite Co. Ltd., Xenia, OH, USA) until January 2007, and subsequently stored at 10 C until DNA extraction (May 2007 and June 2008). Additionally, between July and October 2006, well after all individuals in the five groups were habituated and individually recognizable (within groups), fecal samples were opportunistically collected from ten adults, one subadult, and one juvenile. All collected C. brunneus feces were pellets approximately 2 3 cm long and cm wide when fresh. Fecal pellets from each individual were placed directly in glass or plastic vials with CaSO 4 CoCl 2 desiccant, stored at ambient temperature until January 2007, and subsequently stored at 10 C until DNA extraction (February October 2007). In total, 22 independently collected genetic samples (feces and FTA ) were recovered from five adult females, six adult males, two subadults (male and female), and one juvenile female, distributed among the five socially monogamous groups. FTA Card DNA Extraction In the Molecular Anthropology Laboratory (MAL) of the University of California, Davis (USA), pieces of approximately 2 3 mm diameter were cut with a NaOCL sterilized razor blade from the FTA Classic Cards onto which drops of C. brunneus blood or saliva had been dried in the field. The standard manufacturer protocol for FTA sample preparation, consisting of a series of washes with FTA Purification Reagent (Whatman plc., Brentford, Middlesex, UK) and

28 18 TE 1 Buffer (10 mm Tris HCL, 0.1 mm EDTA, ph 8.0), followed by placement of the washed card piece containing the sample DNA directly into the PCR reaction tube, resulted in consistently unsuccessful PCR reactions for these samples. Consequently, based on recommendations from the manufacturer, I employed a Proteinase K digestion protocol with newly cut 3 mm FTA card sample pieces. The digestion protocol was as follows: 1) The sample piece was incubated for 30 min at 65 C in 200 μl FTA Purification Reagent containing 60 μg/ml Proteinase K. The solution was vortexed for 5 s, centrifuged for 5 s, mixed with a pipette, removed, and discarded. 2) 200 μl FTA Purification Reagent (not containing Proteinase K) was added to the sample piece, which was then incubated for 5 min at room temperature. The solution was vortexed for 5 s, centrifuged for 5 s, mixed with a pipette, removed, and discarded. This step was repeated for a total of three times. 3) 250 μl TE 1 Buffer (10 mm Tris HCL, 0.1 mm EDTA, ph 8.0) was added to the sample piece, which was then incubated for 5 min at room temperature. The solution was vortexed for 5 s, centrifuged for 5 s, mixed with a pipette, removed, and discarded. This step was repeated for a total of four times. 4) The sample piece was then dried at 65 C for 30 min on a heat block. Following this digestion protocol, DNA was eluted from the sample piece by adding 50 μl ddh 2 O and heating at 90 C for 30 min on a heat block. At the conclusion of this elution step, the sample card pieces containing C. brunneus blood were still reddish in color (according to the manufacturer, these pieces should ideally be colorless after the digestion/washing procedures), though this did not impede subsequent PCR and sequencing of the eluted DNA. Aliquots of the eluted DNA (still containing the card pieces) were stored at 10 C until PCR amplification.

29 19 Fecal DNA Extraction Between 40 and 90 mg of the outer layer of each desiccated fecal pellet was manually removed with a sterile razor blade in the MAL, following Wehausen et al. (2004). To reduce the likelihood of sample cross contamination, a new razor blade was employed for each pellet, and all razor blades and forceps were rinsed in a solution of NaOCl (household bleach) before processing each pellet. Once removed, the outer pellet material was placed in a 2 ml microcentrifuge tube containing 1.6 ml of ASL Buffer (QIAmp DNA Stool Mini Kit, Qiagen Inc., Valencia, CA, USA), incubated at 10 C for days, and manually mixed by tube inversion for 30 s each day. After incubation, 400 μl of fresh ASL Buffer was added to each tube, samples were vortexed for 1 min, and DNA was extracted using a QIAmp DNA Stool Mini Kit (Qiagen Inc., Valencia, CA, USA) following the manufacturer s recommended protocol beginning after the ASL Buffer digestion and homogenization step. Aliquots of extracted DNA were stored at 10 C until PCR amplification. Amplification and Sequencing Regions of exons 3, 4, and 5 of the C. brunneus M/L opsin gene, encompassing sites implicated in the absorbance of light by M/L opsins (see Table 2.1), were amplified and sequenced separately as follows. A 129 bp region of exon 3, from nucleotide position 921 to 1049 (numbering as in Nathans et al. 1986b) was amplified with the primer pair 5 CTCCAACCAAAGATGGGCGG 3 and 5 ATCACAGGTCTCTGGTCTCTG 3 (derived from Dulai et al. 1994). This region encompasses nucleotides corresponding to codons 143 to 185. Exon 4 was amplified in its entirety with the primer pair 5 GCCGGCCCTTCTCTCCAG 3 and 5 TGATTCAGGGGCAGAGAAGCTTA 3, annealing in the intron regions on either side of the exon (derived from Cropp et al. 2002) and encompassing nucleotides corresponding to codons 194 to

30 Exon 5 was amplified in its entirety with the primer pair 5' TCCACCCCCCGACTCACTATCC 3 and 5' ACGGTATTTTGAGTGGGATCTGCT 3 (M. Neitz, pers. comm.), annealing in the intron regions on either side of the exon and encompassing nucleotides corresponding to codons 249 to 328. Additionally, a 149 bp region of exon 5 from nucleotide position 1284 to 1432 was amplified separately with the primer pair 5 GAATCCACCCAGAAGGCAGAG 3 and 5 ACGGGGTTGTAGATAGTGGCA 3 (derived from Dulai et al. 1994). This region encompassed nucleotides corresponding to codons 264 to 312. PCR reactions contained 2.5 μl of approximately ng/μl DNA template (not quantified), 0.5 μl of 10 μm dntp mix (2.5 μm each dntp), 2.5 μl of 10x NH 4 Reaction Buffer without MgCL 2 (Bioline USA, Inc., Taunton, MA, USA), 1.7 μl (1.5 μl when amplifying the entire exon 5) of 25 mm MgSO 4, 0.65 μl of 0.1 μg/μl BSA, 0.5 μl of each (the forward and the reverse) primer at a concentration of 10 μm, 0.12 μl of 5 U/μL Platinum Taq DNA polymerase (Invitrogen Corporation, Carlsbad, CA, USA), and ddh 2 O to a final volume of 25 μl. The touchdown thermocycling profile consisted of an initial denaturing at 94 C for 3 min, followed by 60 cycles of denaturing at 94 C for 15 s, annealing at 64 C (exon 3), 62 C (exon 4), or 59 C (exon 5) for 30 s (dropping the annealing temperature 0.1 C per cycle), and extension at 72 C for 20 s. This was followed by a final extension at 72 C for 5 min. Each PCR was performed with a positive control consisting of human DNA and a negative control consisting of ddh 2 O instead of template DNA. Amplification of an appropriately sized fragment was confirmed by electrophoresis using a size ladder on a 1.3% agarose gel. PCR products were purified by adding 0.2 μl of 20 U/mL Exonuclease 1 enzyme, 0.2 μl of 10x NEBuffer for Exo 1 (New England Biolabs, Ipswich, MA, USA), and ddh 2 O to bring each PCR product to a final volume of 100 μl, and then heated at 37 C for 90 min followed by 20 min at 80 C. Samples were filtered through MultiScreen filter plates (Millipore Corporation, Billerica, MA, USA) by pulling a vacuum of approximately 4.7 kpa until

31 21 dry, and then re hydrated with 25 μl ddh 2 O. Sequencing (forward and reverse strands) was performed at the University of California DNA Sequencing Facility (Division of Biological Sciences, University of California, Davis, USA) with an ABI 3730 DNA Analyzer using BigDye Terminator v3.1 Cycle Sequencing chemistry (Applied Biosystems, Foster City, CA, USA). Sequences were aligned with Sequencher 4.7 DNA Sequence Assembly Software (Gene Codes Corporation, Ann Arbor, MI, USA). Genotype Determination Genotypes were assigned to individuals using site 180 in exon 3 and sites 277 and 285 in exon 5, as amino acid changes at these sites are known to cause large shifts in the M/L opsin peak light absorbance. Additionally, the majority of known platyrrhine M/L opsin classes can be discriminated by changes at these three sites (Table 2.1). It should be noted, however, that this genotyping procedure does not discriminate between potential alleles that differ only in exon 4, coding for pigments whose peak light absorbances are probably very close together (Table 2.1). Nor does this procedure discriminate between functional alleles that differ at sites as yet unknown to affect the absorbance of light by photopigments. The use of samples (e.g., feces) yielding low quality (e.g., due to degradation) and/or quantities of DNA to determine the M/L photopigment genotypes of wild platyrrhine monkeys increases the risk of allelic dropout, i.e., the failure of one of the two alleles carried by a heterozygous female to amplify during PCR (Surridge et al. 2002; Taberlet et al. 1996; Taberlet et al. 1999). Allelic dropout can result in a trichromatic female being classified as a dichromat. Alternatively, false alleles can be generated as a result of PCR and sequencing errors, such that homozygous (dichromatic) females may appear heterozygous (trichromatic). False alleles appear to be particularly common in genetic analyses employing DNA from non invasively

32 22 collected samples (Taberlet et al. 1999), such as feces (e.g., Lathuilliere et al. 2001; Taberlet et al. 1996). To reduce the risk of allelic dropout and false alleles, I followed Surridge et al. (2002) in performing at least two separate extractions for three of the five adult C. brunneus females, using either two independently collected fecal samples from the same individual, or corresponding fecal and FTA samples from captured females. This was not possible for two adult females (Groups 1 and S), assigned heterozygous and homozygous genotypes (respectively), from whom only single fecal samples were collected. Sequences from at least two independent PCR amplifications of site 180 in exon 3, and at least three independent amplifications of sites 277 and 285 in exon 5 were obtained for each adult female (somewhat less than the six independent sequences per individual of Surridge et al. 2002). Allele sequences from most varieties of heterozygous female would be expected to show heterozygous sites in exon 3, exon 5, or both (Table 2.1). Thus, this procedure potentially provides between two and five independent opportunities to verify that a female is heterozygous or homozygous. A female was typed as a heterozygote (and a trichromat) if there was evidence of heterozygosity at absorbance tuning sites in both forward and reverse sequences from at least one independent PCR amplification of exon 3 or exon 5 (although in practice there was always evidence from more than one amplification see Results below). An isolated heterozygous site in an otherwise homozygous exon sequence fragment was deemed a false allele (i.e., a PCR or sequencing error) if it was not present in the complementary (forward or reverse) sequence fragment nor in the sequences from other independent amplifications from that individual. Because males are known to be hemizygous at the M/L opsin locus, inaccurate genotyping due to allelic dropout and false alleles is less of a concern. Genotypes of four of the six adult males in the study were determined from a single amplification of exon 5 from each of

33 23 two independently collected fecal samples per individual, and, for three of these males, at least one amplification of exon 3. The genotypes of two adult males (Group W males 1 and 2 in Table 2.3, below) were inferred from a single amplification of exon 5 from one FTA sample, or two amplifications of exon 5 from a single fecal sample (respectively). Genotypes of the subadult female in Group 1, the subadult male in Group 3, and the juvenile female in Group 3 were inferred from one to two amplifications of exon 5, and thus should be interpreted with caution. Contamination and Mis assignment Checks The primers used in this study amplify the separate human M and L opsin genes in addition to the M/L opsin gene of C. brunneus. For this reason, all C. brunneus sequences from exons 3 and 5 were compared to exons 3 and 5 of human M and L genes in order to ensure that the putative monkey sequences were not the result of human contamination. In addition to several nucleotide differences between the human M and L opsin genes, which, when both genes are amplified together, appear as heterozygous sites not apparent in the monkey sequences, several consistent differences were found between all putative C. brunneus sequences and those of both human M and L genes. For instance, amino acid position 154 in exon 3 of human M and L opsin genes has the nucleotide triplet GTG (Nathans et al. 1986b). In contrast, all C. brunneus exon 3 sequences analyzed in this study had GTT at position 154. Similarly, amino acid position 275 of the C. brunneus exon 5 was found to vary between the nucleotide triplets GTG and ATG. In contrast, the human L opsin gene has TTT at position 275, while the human M opsin gene has CTG (Nathans et al. 1986b, Bunce, data not shown). There were two instances, one exon 3 amplicon and one exon 5 amplicon, where the resulting C. brunneus sequence for one individual matched the human M and L gene sequences and was dissimilar from all other exon 3 and exon 5 sequences obtained for that individual. These two

34 24 instances were attributed to human contamination after the extraction step and were excluded from all subsequent analyses. The consistent differences observed between the C. brunneus and human sequences give me confidence that the genotyping of individual monkeys was not compromised by human contamination. As a check on the correspondence between collected fecal samples and the individuals from whom they originated, Di Fiore (2005) s method of genetic sex determination through amplification of the amelogenin and SRY loci was performed on DNA extracted from several fecal samples. Despite the fact that fecal samples were collected toward the end of an intensive 15 month behavioral study when animals were well habituated and individually recognizable to the observers (within groups), two samples attributed to adult females were found to have originated from male animals. Because the only male in each group was the resident adult male at the time of both instances of sample deposition, it was concluded that these samples were produced by the adult males. In both instances, the M/L opsin genotypes of the disputed samples matched those determined from independently collected samples from the adult males and did not match the genotypes determined from independently collected samples from the adult females. More encouragingly, the results from five other fecal samples subjected to genetic sex determination were in accord with the sex of the individual to whom they were assigned. RESULTS Alleles and Genotypes Analysis of site 180 and sites 277 and 285 in exons 3 and 5, respectively, of the X linked M/L opsin gene from homozygous and hemizygous C. brunneus individuals revealed the presence of at least three functionally distinct alleles (i.e., each of these individuals had one of

35 25 three alleles). Table 2.2 shows the amino acid differences between these three presumptive alleles. In the right hand side of Table 2.1, the three allele haplotypes are matched with haplotypes of other known platyrrhine alleles based on sites 180, 277, and 285. The C. brunneus haplotypes appear to correspond most closely with platyrrhine alleles coding for M/L pigments with peak light absorbances at approximately 535 nm, 550 nm, and 562 nm (Table 2.1). This is in accord with the three most common M/L pigment classes reported by Jacobs and Deegan (2005) for the captive Callicebus colony at the CNPRC. However, without direct measures of the absorbance of M/L cone cells in the retinas of C. brunneus individuals of known genotype (e.g., Neitz et al for Saimiri and Saguinus) or in vitro reconstitution of photopigments from allele sequences (e.g., Hiramatsu et al for Cebus), I cannot exclude the possibility that the putative C. brunneus 535 nm allele actually codes for the 530 nm or 543 nm M/L opsin variants reported by Jacobs and Deegan (2005), or that the putative 550 nm C. brunneus allele actually codes for a 556 nm opsin variant analogous to that of Saguinus (Table 2.1). For this reason, the naming of the three opsin alleles found in this study, hereafter referred to as the 535 nm, 550 nm, and 562 nm alleles, should be viewed as a temporary convenience until accurate estimates of the peak light absorbances of the resulting photopigments can be made. Three of the five adult females (and the subadult female) in this study were found to be heterozygous at the M/L opsin locus. For two of these adult females, evidence of heterozygosity at sites 180 (exon 3) and 277 (exon 5) was found in both forward and reverse sequences derived from at least four PCR amplifications, each from an independently collected fecal and an FTA sample. For one adult female (Group 1), evidence of heterozygosity at site 285 (exon 5) was found in both forward and reverse sequences derived from five PCR amplifications of DNA from a single fecal sample. In contrast, homozygous females either displayed no heterozygosity at known absorbance tuning sites in exons 3 and 5, or displayed an isolated heterozygous site on a

36 26 single, otherwise homozygous, sequenced strand that could be attributed to a false allele according to the criteria above (see Methods: Genotype Determination). Table 2.3 shows the genotyping results for the 14 C. brunneus individuals in this study, as well as the sample types collected from each, validation of the sex of the sample donating animal, and the number of sequences and PCR amplifications used in the genotype assignments. As shown in Table 2.3, exon 5 was successfully amplified and sequenced for all samples. Amplification of exon 3 was generally successful, though it was not attempted for subadults, the juvenile, or adult male 1 from Group W. Sequenced strands from exon 3 were obtained from the adult male of Group 2 and adult male 2 of Group W, but these were not of sufficient length to contain site 180 and were therefore not used in genotyping. Despite repeated attempts, exon 4 could not be amplified for the adult female of Group 2. Additionally, presumed allelic dropout was observed in all exon 4 sequences from two of the three heterozygous adult females. For this reason, genotyping assignments were based solely on exons 3 and 5. No evidence of allelic dropout or false alleles was obtained from exon 3 sequences. However, generally only the forward strands sequenced from exon 3 extended far enough to contain site 180. For this reason, only a single exon 3 sequence fragment per amplification was used in genotyping (and is recorded in Table 2.3). Of the 23 sequenced exon 5 fragments from heterozygotes, representing 11 PCRs, two fragments showed evidence of allelic dropout. This constitutes an allelic dropout rate of 9% for sequenced fragments, or 18% for PCRs. Both of these cases of allelic dropout involved an adenine residue at site 277 on a reverse strand amplified using a primer pair annealing between nucleotide positions 1284 and 1432 (see Methods: Amplification and Sequencing). Of the 76 total fragments sequenced from exon 5, representing 47 PCRs, 12 fragments showed evidence of false alleles. This constitutes 16% of sequenced fragments and 26% of PCRs, although no strand with a false allele occurred in the

37 27 absence of a complementary strand with the correct allele from the same PCR. All cases of false alleles involved an adenine residue at site 277 on a forward strand amplified using a primer pair annealing between nucleotide positions 1284 and 1432 (above). Seven of the 12 false alleles occurred in males, who are hemizygous. The rates of allelic dropout and false alleles observed in this study for sequenced fragments (9% and 16%, respectively) are consistent with the findings of Lathuilliere et al. (2001), who reported allelic dropout and false allele rates of 3 6% and 13 20%, respectively, during an analysis of microsatellites using DNA derived from the feces of Barbary macaques (Macaca sylvanus). Several synonymous mutations were found in exons 3 and 5 of the 550 nm allele, at sites not known to affect light absorbance by photopigments. Because such mutations are unlikely to affect absorbance, these mutant alleles were not separated in the analysis. One nonsynonymous mutation was detected at amino acid site 269 in exon 5 of a nm heterozygous adult female (Group W). All 550 nm and 562 nm alleles from hemizygous and homozygous individuals in this study had the nucleotide triplet ATG (methionine) at position 269. However the heterozygous female displayed an ATT/ATG (isoleucine/methionine) polymorphism at this position which was reproduced in two sequences from separate PCR amplifications of DNA from fecal and FTA samples. Because position 269 is not known to affect light absorbance by photopigments (although its effect has yet to be investigated), this mutation was not considered further in this analysis. Allele Frequencies and Associations Among the 16 adult C. brunneus X chromosomes examined in this study, the 535 nm, 550 nm, and 562 nm alleles occurred at frequencies of 0.188, 0.625, and 0.188, respectively. For these three alleles, Jacobs and Deegan (2005) reported frequencies of 0.203, 0.288, and 0.356,

38 28 respectively, in a sample of 118 Callicebus X chromosomes. If the 530 nm and 542 nm alleles reported by Jacobs and Deegan (2005) at frequencies of and 0.085, respectively, are assumed to be absent from the wild C. brunneus sample (but present in the wild population from which it was drawn), then the allele frequencies observed in this study cannot be statistically distinguished from those reported by Jacobs and Deegan (2005) for the captive animals (Fisher s exact test, simulated p value of 0.13 based on 200,000 replicates, computed using the base package of R: R Development Core Team 2008). Disassortative mating by M/L genotype (e.g., through avoidance of inbreeding) might be expected if individuals who maximize the number of heterozygous trichromatic female offspring have a fitness advantage (Surridge et al. 2005a). Figure 2.2 shows how the M/L alleles were distributed among the five socially monogamous C. brunneus study groups. Considering adult pairs, in only one group (S) did the adult female and male have no M/L alleles in common. This was balanced by one group (2) in which both adults had one and the same allele. In the remaining three groups, females shared one of their two alleles with their male pair mates. Thus, in this small sample there appears to be no evidence for disassortative mating by M/L opsin genotype, such that, within groups, alleles found in one sex are less common in the other sex (contrast with Surridge et al. 2005a for Saguinus). The above results for allele frequencies and disassortative mating are presented as the first such data from a wild Callicebus population. However they must be interpreted with extreme caution due to the very small sample size. Reproduction in Groups As summarized in Table 2.4, among the five C. brunneus study groups observed between August 2005 and October 2006, nine births were recorded, three of which ended in

39 29 disappearance (presumed death) before the age of six months. Additionally, four subadults disappeared from their natal groups at ages (judging by size and behavior) that suggest natal dispersal. Two adult males disappeared over the course of the study and were replaced by new adult males. Noteworthy is the lack of a strong trend for trichromatic females to successfully produce more offspring than dichromatic females over the 15 month observation period. Both di and trichromatic females gave birth, lost infants, and successfully raised offspring to dispersal age. If short term reproductive success is counted as the number of existing offspring plus the number of births minus the number of infant/juvenile disappearances, then the most reproductively successful group (3) contained a trichromatic female and the least successful (S) contained a dichromatic female. However, due to the limited sample size, it would be premature to draw conclusions from this observation with regard to a population wide fitness difference between di and trichromatic females. The important point is that reproductively successful dichromatic females appear to exist at reasonably high frequencies in this C. brunneus population. DISCUSSION The results of this study suggest three important conclusions. First, at least three X linked M/L opsin alleles occur at relatively high frequency in this wild C. brunneus population, leading to the presence of di and trichromatic females. This is the first time, to my knowledge, that the color vision of wild Callicebus individuals has been investigated. Second, aside from the presence of the M/L polymorphism and high frequency alleles, I found no strong evidence for selection favoring trichromatic over dichromatic C. brunneus females, either in terms of large differences in short term reproductive success or evidence of adaptations facilitating disassortative mating in order to maximize the likelihood of heterozygous offspring. Third, the

40 30 methodology used in this study confirms the adequacy of non invasively collected fecal samples for the color vision genotyping of wild Callicebus. However, even for habituated and individually recognizable animals, care must be taken to ensure that collected samples actually originate from the desired individual. Callicebus Alleles and Selection Here I have provided molecular genetic evidence for the presence of at least three functionally distinct M/L opsin alleles in a wild C. brunneus population. Based on allele haplotypes consisting of sites known to affect the absorbance of light by photopigments, it appears that these C. brunneus alleles correspond to the three most common M/L photopigment variants reported by Jacobs and Deegan (2005) in a captive Callicebus population, with peak absorbances at approximately 535 nm, 550 nm, and 562 nm (Table 2.1). Three photopigment variants with similar absorbance properties have also been reported in Cebus (Jacobs and Deegan 2003), Saimiri (Mollon et al. 1984), Brachyteles (muriquis, Talebi et al. 2006), and Pithecia (sakis, Boissinot et al. 1998; Jacobs and Deegan 2003), while two of these three variants have been found so far in Ateles (Hiramatsu et al. 2005; Jacobs and Deegan 2001; but see Riba Hernandez et al. 2004), and Lagothrix (wooly monkeys, Jacobs and Deegan 2001). This contrasts with the three M/L photopigments common in callitrichines (tamarins and marmosets), having peak absorbances at approximately 544 nm, 556 nm, and 563 nm (reviewed in Jacobs 2007; Jacobs and Deegan 2003; Surridge et al. 2003). In the sample of 14 wild C. brunneus individuals, I found no evidence of the two least common M/L opsin variants reported by Jacobs and Deegan (2005) for captive Callicebus, having peak absorbances at approximately 530 nm and 542 nm. However, until a correspondence between allele sequences and in vivo light absorbances is determined for all five potential Callicebus M/L opsin variants, I cannot rule

41 31 out the possibility that this wild C. brunneus sample contains more than three functional alleles. Similarly, even if the five C. brunneus alleles would be separated by this haplotype analysis, the absence of two of the five alleles from this small C. brunneus sample would not exclude the possibility that the other two alleles occur in wild populations at frequencies comparable to those reported by Jacobs and Deegan (2005) in a captive colony. What this small C. brunneus sample does demonstrate, however, is that at least three functionally distinct M/L opsin alleles occur in wild populations at frequencies high enough that they were sampled multiple times in a collection of only 14 animals. The high frequency of three M/L alleles suggests that a large proportion of wild female C. brunneus are trichromatic, and this was supported by the fact that four out of the seven females (57%) in this study were found to be trichromats (i.e., close to the 54% predicted under Hardy Weinberg equilibrium with the observed allele frequencies). I found no evidence that adult C. brunneus increase their likelihood of producing trichromatic daughters by pairing disassortatively with respect to M/L opsin alleles, such that pair mates tend to posses alleles absent from their partners. This contrasts with the findings of Surridge et al. (2005a), who found that, within Saguinus groups containing multiple adult males and one or more adult females, M/L alleles absent from one sex tended to be present in the other sex. Additionally, mating tended to occur between Saguinus individuals with the most divergent alleles possible within the group. Surridge et al. (2005a) attributed the pattern to inbreeding avoidance in Saguinus, with the idea that more distantly related individuals have a higher probability of possessing different M/L alleles. Primates such as Saguinus and Callicebus may be able to use olfactory cues (as in lemurs: Charpentier et al. 2008) and/or auditory cues (as in macaques: Fugate et al. 2008) to differentiate between kin and non kin in contexts such as mate choice (e.g., Callicebus do engage in both scent marking and vocal communication: Mason 1968; Robinson 1981, pers. obs.). The lack of evidence for disassortative mating in this C.

42 32 brunneus sample may indicate that 1) such potential kin recognition mechanisms are not effective in this species; 2) there are constraints on outbreeding in wild Callicebus in general; 3) there are outbreeding constraints in this C. brunneus population in particular; and/or 4) disassortative mating is characteristic of C. brunneus, but this sample is not representative of the population. Data from more wild Callicebus groups are necessary to address this question. However, a relevant observation is that Callicebus, though socially monogamous, do not have a strictly monogamous mating system. Mason (1966), studying a population of Callicebus cupreus ornatus (revised classification by Hershkovitz 1990) reported several instances of extra pair copulations. Thus, pair membership may not accurately reflect the mating behavior in these animals. With this in mind, it is perhaps suggestive to note in Figure 2.2 that extra pair copulations between neighboring C. brunneus groups (overlapping circles in Figure 2.2) would result in a higher probability of trichromatic daughters for most individuals than would withinpair mating. While the genotypes of all three offspring sampled in this study were compatible with within group parentage, paternity analysis using other markers was not done. Therefore, I cannot rule out the possibility that some offspring were sired by extra group individuals, particularly from neighboring C. brunneus groups not included in this study. Information from a larger sample of wild Callicebus groups is needed to investigate the potential role of extra pair copulation in the maintenance of polymorphic color vision in this genus. Contrary to the hypothesis of a general selective advantage for trichromatic females over dichromatic females, I found no strong evidence that C. brunneus pairs containing trichromats had higher short term reproductive success than dichromat dichromat pairs. This contrasts with the results of Surridge et al. (2005a) who found that 1) older females in a wild Saguinus population tended to be trichromats; 2) seven out of eight study groups (presumably not chosen on the basis of M/L allele composition) contained trichromatic females; and 3) the

43 33 only group containing dichromatic females was unstable and failed to produce offspring over the 18 month study period. In contrast, two of the five stable and successfully reproducing C. brunneus pairs observed in this 15 month study contained dichromatic females. Over the study period, the group producing the most surviving offspring (Group 3, three offspring) contained a trichromatic female, while the group producing the fewest surviving offspring (Group 2, one offspring) contained a dichromatic female (Table 2.4). Observations of more Callicebus groups over longer time intervals are necessary to determine if trichromatic females, in general, have higher reproductive success than dichromatic females. Nonetheless, it seems likely that viable dichromat dichromat C. brunneus pairs occur at relatively high frequency in this population, such that an observer blind (with respect to subject color vision) sample of five groups would yield two such pairs. Evidence suggesting that completely dichromatic, reproductively successful groups occur in other platyrrhines has been reported by Smith et al. (2003a; 2005) for Saguinus and Hiramatsu et al. (2005) for Cebus. From Figure 2.2 it is interesting to note that the two C. brunneus groups with dichromatic females (2 and S) are neighbors, and are physically separated from the groups with trichromatic females. At the study site, the territories of Groups 2 and S bordered a river, while those of Groups 1, 3, and W were distributed around a lake further inland (Figure 2.1). A consequence of this arrangement was that the territories of Groups 2 and S comprised forests of similar types, e.g., early riparian successional stages and a Ficus trigona (Moraceae) swamp. In contrast, the territories of the other three groups comprised larger portions of mature high ground and late successional forest (e.g., Terborgh 1983:13). Thus di and trichromatic C. brunneus groups may be distributed non randomly on the landscape. Mollon et al. (1984) argued that polymorphic trichromacy in platyrrhine primates could be maintained by frequencydependent selection at the level of individuals if di and trichromatic monkeys have advantages

44 34 in different ecologically important tasks, such as foraging in different micro habitats or on different types of food. As a much simplified example, suppose feeding competition limits reproductive success in a fixed environment of two equally abundant forest types, one best exploited by dichromats, the other by trichromats. Then, all else being equal, the fitness of dichromats relative to trichromats will increase if the population frequency of dichromats falls below 0.5, and vice versa for trichromats. In this case, frequency dependent selection is contingent on advantages to both trichromats and dichromats in different circumstances. There is behavioral evidence that trichromatic primates may have an advantage in finding orange and yellow fruit against green leafy backgrounds (Caine and Mundy 2000; Smith et al. 2003b) while dichromats may have an advantage in finding color camouflaged food objects, such as insects (Melin et al. 2007; Morgan et al. 1992; Saito et al. 2005b). However, evidence of niche divergence in wild di and trichromatic platyrrhines has yet to be documented (e.g., Melin et al. 2008). The C. brunneus data suggest that more information about the spatial distribution and dietary composition of platyrrhine groups containing all dichromats compared to groups with a mix of di and trichromatic individuals, may prove useful in evaluating the hypothesis of frequency dependent selection on di and trichromacy in platyrrhines. Genotyping Wild Callicebus The use of fecal samples to determine the color vision genotypes of wild platyrrhine primates has become standard practice in recent years (e.g., Hiramatsu et al. 2005; Riba Hernandez et al. 2004; Smith et al. 2003a; Surridge et al. 2002; Vogel et al. 2007). The present study extends this methodology to the genus Callicebus. Through this analysis I was able to directly sequence at least one exon of the M/L opsin gene from all 22 collected fecal, blood, and saliva samples. When feces and blood or saliva were collected from the same individual,

45 35 comparable haplotypes were obtained from each sample type, indicating the reliability of fecal DNA in this analysis. However, in two instances a fecal sample was erroneously attributed to a study animal. This confusion occurred despite the fact that all individuals in the study were habituated and individually recognizable (within groups) after 15 months of behavioral observation. The risk of misattributing fecal samples is inherent in this methodology, and an increased likelihood of such misattribution may be a peculiarity of work with Callicebus, of work with the study population, or of the particular sample collector (the author, in this case). However, in situations where group sizes are small, such errors can often be easily identified and corrected using Di Fiore (2005) s method of genetic sex determination in primates. I would recommend this technique in future studies attempting to assign color vision genotypes obtained through feces to individual Callicebus in wild populations.

46 36 CHAPTER 3 The Importance of Spatial Memory and Lactation for the Evolution of Color Vision in Callicebus brunneus, a Neotropical monkey ABSTRACT The trichromatic color vision of many primates, allowing distinction among colors such as green, yellow, orange, and red, is unique among diurnal placental mammals, most of whom have dichromatic vision for which discrimination among such colors may be difficult. However, the ecological utility and selective advantage(s) afforded to primates by trichromacy remain poorly understood. I test the hypothesis that the advantage of trichromacy to wild forestdwelling primates lies in a superior ability to detect yellow, orange, or red (YOR) food patches at a distance. I employ the socially monogamous Neotropical titi monkey (Callicebus brunneus), which, like most platyrrhines, has X linked polymorphic trichromacy, to determine 1) whether long range detection of YOR patches is important for these animals in the wild; 2) whether trichromats have an advantage over dichromats in long range YOR patch detection; and 3) whether a trichromat advantage is most apparent in times of energy stress. I report that, although C. brunneus spent the majority of foraging time in YOR patches, the task of detecting such patches from a distance using only color vision is likely to have been relatively rare as a result of potentially extensive use of spatial memory. In general, females were likely to have encountered more patches by means of long range visual detection (in the absence of memory) than were males, and, on average, heavily lactating trichromatic females had higher encounter rates than their dichromatic male pair mates for non memorable visually detected YOR patches. However, no differences were found in the encounter rates of di and trichromatic females for non memorable YOR patches likely to have been visually detected, regardless of assumed

47 37 energetic stress due to season and lactation. These results illustrate the need to determine the prevalence of the task of YOR target detection for wild primates before developing hypotheses for the selective advantages of trichromacy based on the prevalence and importance of YOR targets in the environment. These results also suggest that, in the search for a trichromatic advantage in wild primate populations, it may be necessary to account for the lactational status of females. To more fully understand the evolution of trichromatic color vision in our primate ancestors, we would do well to understand its utility in extant primate populations. INTRODUCTION An unresolved puzzle in science involves the acquisition and selective advantage(s) of trichromatic color vision, allowing consistent distinction among colors such as green, yellow, orange, and red. Trichromacy in humans, apes and Old World monkeys (Catarrhini) is made possible by neural comparisons among the outputs of three classes of cone cell in the retina, known as S (short wave), M (middle wave), and L (long wave) cones, which absorb light over defined ranges of wavelengths (the names referring to relative wavelengths of peak absorbance). Cone cell light absorption is facilitated by photopigment proteins, known as S, M, and L opsins, which are most sensitive to light at approximately 430 nm, 535 nm, and 562 nm, respectively (reviewed in Jacobs 2008; Surridge et al. 2003). In contrast, the retinas of most other diurnal placental mammals (Eutheria) have only S and L cones with their respective opsins (Ahnelt and Kolb 2000; Jacobs 1993). The result is dichromatic vision, in which colors such as green, yellow, orange, and red may be difficult to discriminate under normal lighting conditions (Sharpe et al. 1999). The mammalian L opsin is controlled by an X linked gene, while the S opsin gene is autosomal (Nathans et al. 1986a for humans). The catarrhine M opsin was created by a duplication of the L opsin gene, presumably occurring between approximately 43 and 31 million

48 38 years ago (mya) in a common ancestor of all extant catarrhine primates (Dulai et al. 1994; Hunt et al. 1998; Nathans et al. 1986b; Steiper and Young 2006). The fact that the M opsin and trichromacy have been conservatively maintained in catarrhines is interpreted as evidence of stabilizing selection (Jacobs and Deegan 1999; Jacobs and Williams 2001; Onishi et al. 1999; Surridge et al. 2003). Most Neotropical monkeys (Platyrrhini) and some Malagasy lemurs (Strepsirrhini, Lemuroidea) have achieved trichromacy in the absence of a gene duplication. Instead, these primates have developed a polymorphism at the ancestral X linked L opsin locus, comprising (usually) two or three alleles coding for opsins with maximal absorbance between 530 nm and 562 nm. As a result, heterozygous females are trichromatic, while homozygous females and all males are dichromatic (reviewed in Jacobs 2007; Jacobs et al. 2002; Leonhardt et al. 2008; Tan and Li 1999; Veilleux and Bolnick 2009). In platyrrhines, the similarity and conservative maintenance of this polymorphism in nearly all extant genera suggest that it originated at least 20 mya in a basal platyrrhine (Boissinot et al. 1998; Hunt et al. 1998) and is under stabilizing selection (Surridge and Mundy 2002). As noted by Mollon et al. (1984), the maintenance of such polymorphic trichromacy need not entail a trichromat fitness advantage over dichromats in all circumstances, although the evolution of the polymorphism would be difficult to explain in the absence of a heterozygote advantage in at least some ecologically important context (Surridge et al. 2003). Selective Advantage of Trichromacy Allen (1879) suggested that color vision in animals coevolved with colorful fruits and flowers, thereby making such food items more easily detectable against green leaves. From the perspective of primates, this hypothesis has been expanded to posit that trichromacy provides

49 39 an advantage in the detection of any yellow, orange, or red target against the dappled green leaf background of forests (Mollon 1989). Suggestions of which colorful targets are (and have been) most important to primates include fruits (Dominy et al. 2003b; Sumner and Mollon 2003b), young leaves (Dominy and Lucas 2001; Lucas et al. 2003), conspecifics (Changizi et al. 2006; Sumner and Mollon 2003a), and predators (Caine 2002; Smith et al. 2005). Much recent effort is devoted to examining this hypothesis in the context of primate foraging by directly and indirectly comparing the food detection abilities of di and trichromatic individuals, especially in species with polymorphic trichromacy. Sumner and Mollon (2000a; 2000b) distinguish between two tasks for which vision may be employed by primates foraging for food in a forest. A food patch must first be detected from a distance, in which case food item coloration should contrast with the dappled background of leaves and trunks. Second, upon arriving in the food patch, the color of edible (e.g., ripe or riper) fruits must be distinguished from that of inedible (e.g., unripe or less ripe) fruits and the leaves of the specific patch. By modeling the detectability of food items for di and trichromatic primates, numerous studies have shown that the trichromatic vision of both catarrhines and heterozygous platyrrhines may be generally superior to that of dichromats for detecting both fruits (de Araujo et al. 2006; Osorio et al. 2004; Osorio and Vorobyev 1996; Riba Hernandez et al. 2005; Riba Hernandez et al. 2004; Stoner et al. 2005; Sumner and Mollon 2000a) (though see Regan et al. 2001) and young leaves (Dominy and Lucas 2001; Lucas et al. 2003; Sumner and Mollon 2000a) against a background of mature green leaves. Direct evidence for a trichromat advantage in distinguishing orange targets from a green/brown background comes from studies of captive marmosets (Callithrix) designed to mimic natural detection tasks at distances up to 6 m. Caine and Mundy (2000) showed that, under such circumstances, trichromats find more orange food items than dichromats. However, there is currently no evidence from wild primate

50 40 populations of a long distance trichromatic advantage in the detection of colorful food patches. For instance, Dominy et al. (2003a) found no evidence that presumably trichromatic female tamarins (Saguinus) were more likely than dichromatic males to detect yellow fruit patches. Similarly, Smith et al. (2003a) concluded that sex and individual characteristics other than color vision were better determinants of leadership than trichromacy in groups of foraging Saguinus, contrary to expectations if trichromacy provides an advantage in long distance food detection. For the separate task of making within patch distinctions between ripe and unripe (or less ripe) fruit, trichromacy is again predicted to be generally superior to dichromacy, though the advantage is likely to be less than that for detecting food patches against a leafy background (Sumner and Mollon 2000b). A modest advantage in within patch foraging for ripe colored fruit has been demonstrated for captive Saguinus in a naturalistic foraging task (Smith et al. 2003b). However, Vogel et al. (2007) found no evidence of a trichromat advantage in energy intake rate for wild capuchins (Cebus) foraging within food patches. Melin et al. (2008) found no evidence that wild di and trichromatic Cebus differ in the time devoted to foraging on conspicuously colored fruit. Similarly, Hiramatsu et al. (2008) did not detect a difference in the fruit feeding rates of wild di and trichromatic spider monkeys (Ateles), as might be expected if trichromats were more efficient at selecting edible fruit. These results might be explained by Dominy (2004) s suggestion that within patch foraging by primates in the wild may be guided more by smell and touch than by vision. In summary, trichromatic color vision must surely provide a selective advantage over dichromacy for primates in order to have been maintained for long periods of evolutionary time. However, the nature of this advantage remains elusive in studies of wild primate populations. It has long been recognized that trichromacy may serve different purposes in extant primate species with different diets and habitats (e.g., Osorio and Vorobyev 2008), and that any or all of

51 41 these purposes may be different from that/those which provided a selective advantage in the ancestral primate populations where trichromacy first evolved (Allen 1879; Fernandez and Morris 2007). However, without an understanding of how trichromacy is used by (any) extant non human primate species, it is difficult to generate and evaluate evolutionary hypotheses for the origin of primate color vision (e.g., Dominy et al. 2003b). The objective of this study is to determine the foraging context(s) for which trichromacy provides an advantage over dichromacy in a wild platyrrhine population. I focus on the task of long distance detection of food patches in a tropical forest. I hypothesize that: 1) the longdistance visual detection of food patches containing yellow, orange, or red (YOR) food items is an important part of the foraging behavior of these animals; 2) trichromats detect more YOR food patches than dichromats per unit time searching; and 3) any difference in YOR patch detection rates between di and trichromats will be most apparent during times of energy stress, either as a result of seasonal decreases in fruit abundance, or increases in metabolic requirements due to lactation. Determining how an extant primate species uses trichromatic color vision is essential to understanding the evolution of this visual configuration. METHODS Study Site and Subjects The Estación Biológica de Cocha Cashu comprises a 10 km 2 study area located at approximately 350 m elevation in the interior of Manu National Park in the department of Madre de Dios in Southeastern Peru (11 52 S, W). The site lies within the 6 8 km wide meander belt of the Manu River and consists of a patchwork of seasonally flooded riparian and lacustrine successional stages amidst mature high ground tropical moist forest (Terborgh 1983). Mean annual rainfall is approximately 2000mm, with most rain usually falling between October

52 42 and April (the wet season). Mean annual temperature is approximately 23 C (range 13 C and 33 C) (Terborgh 1983; 1990). There is a marked increase in overall forest fruit production during the wet season (Janson and Emmons 1990; Terborgh 1983; 1990). The immediate vicinity of the study site has been protected from hunting and logging since 1973 (Terborgh 1999) and contains a community of eleven primate species (Terborgh 1983). Brown titi monkeys, or monos tocones (Callicebus brunneus, Hershkovitz 1990), are small, socially monogamous, arboreal omnivores with slight sexual dimorphism, generally unspecialized morphology, and polymorphic color vision (Hershkovitz 1990; Jacobs and Deegan 2005; Kinzey 1981; Wright 1984). The mean (+/ s.d.) weight of three adult female animals captured during this study was 1040 (+/ 60) g, while one older adult male weighed 1090 g. Callicebus brunneus devote most foraging time to fruit, but also eat leaves and insects (Kinzey 1981; Wright 1985). The social unit consists of an adult female and male and their dependent offspring (Wright 1984), which together defend exclusive territories of 5 12 ha at this field site (Terborgh 1983 and Figure 2.1). After birth, an infant is carried almost exclusively by the adult male for about four to six months until it can travel independently (Wright 1984 and pers. obs.). During the majority of the study period group sizes ranged from two to three animals, not including dependent infants. Data Collection Behavioral Sampling Five habituated C. brunneus groups (1, 2, 3, S, and W) were followed for a total of 107 days from 23 January to 3 June and 29 June to 6 October This period spans the entire dry season (May September) and about half of the wet season. Each group was followed for a total of 18 to 23 weekdays on a randomly ordered schedule. Continuous focal animal sampling

53 43 (Altmann 1974) was conducted simultaneously for the adult female and male in each group by a two observer team. Each day observers were randomly assigned to a focal animal. Callicebus brunneus have a normal h daily activity cycle (Kinzey 1981). The mean (+/ s.d.) time per day that both adults were simultaneously observed was 7.7 (+/ 2.7) h, with a range of h. Location of the animals at the start of each day was facilitated by a TR 4 VHF receiver and RA 14 antenna (Telonics, Inc., Mesa, AZ, USA) coupled with 25 g MI 2 radio transmitter collars (Holohil Systems, Ltd., Carp, Ontario, Canada) custom modified with no external antennas on adult females from three groups (2, 3, and W), and recorded vocalization playbacks for two groups (1 and S) using a Legendary portable guitar amplifier (Pignose Gorilla, North Las Vegas, NV, USA). Details of the animal habituation and capture methods are presented in Appendix B. Behavior was assigned to one of the following exclusively defined categories: foraging, socializing (e.g., grooming), resting, traveling, and miscellaneous behavior. Time at all behavioral category transitions was recorded using Axim X30 handheld computers (Dell Inc., Round Rock, TX, USA) running custom data collection software written in Microsoft Visual Basic.NET (Microsoft Corporation, Redmond, WA, USA). For the present analysis I focus on foraging, which was defined as ingesting, manipulating (with the hands or mouth), carrying (in the mouth), searching (within a food patch) for, or usurping (from another group member) a food item. Due to the density of the forest, there were also instances where a foraging focal animal was either partly or completely hidden from view. The animal was scored as foraging out of view if falling food items could be heard or seen originating from the animal s location. A 35 to 90 min inter observer reliability trial was conducted once per month, where the behavior of a single focal animal was recorded simultaneously and independently by both observers from a similar vantage point.

54 44 Food Patch Classification I define a single food patch as a physical location (e.g., a tree) of a certain food type (e.g., a fruit species) visited by an individual or individuals for a continuous time period, even if temporally separated foraging events occurred during the visit. I attempted to identify and mark the locations of all food patches used by the focal animals, as well as the forest geometry (following Endler 1993) and weather conditions where and when the animals first began foraging in the patch. Where possible, food items dropped by the animals or collected from the same patch were matched to the same or closely related species whose detectability for di and trichromats has been modeled by Regan et al. (2001), Osorio et al. (2004 and D. Osorio, pers. comm.), or Hiramatsu et al. (2008). Patches containing food items predicted by these models to be more easily detected by trichromats were assigned to the YOR category (patches characterized by yellow, orange, or red). If a food item could not be matched to a previously modeled species, it was classified as YOR if a normal human trichromat (JAB) had described it as yellow, orange, red, or pink at the time of collection. Using human color classifications to represent animal vision is often inappropriate, especially when an animal s visual perception is known to be very different from that of humans (Bennett et al. 1994). However, in general, the set of colored food items distinguishable from background leaves by a normal human trichromat is predicted to match or encompass the set distinguishable to any polymorphic platyrrhine phenotype (de Araujo et al. 2006; Osorio et al. 2004; Riba Hernandez et al. 2004; Stoner et al. 2005) (though see Regan et al. 2001). Thus, this classification is conservative in that, at worst, the YOR category includes some patches that are equally (un)detectable to di and trichromatic study animals. Classifications also took into account features of the food patch that were not consumed by the animals, for instance brightly colored immature fruits (e.g., Stylogyne

55 45 cauliflora [Myrsinaceae], Figure 3.1) and peduncles (e.g., Huertea glandulosa [Staphyleaceae], Figure 3.2). Reproductive Status Births and nursing bouts were recorded, as well as infant carrying behavior by the adult male. The frequency and duration of nursing bouts were difficult to quantify as nursing often occurred in hidden areas of dense vegetation. Consequently, a female was judged to be lactating heavily from the day of birth to the last observed day of male infant carrying behavior, and this was termed the Infant Dependency Period (IDP). The IDP lasted for about four to six months (somewhat longer than the four months observed by Wright 1984) and corresponds to a period of rapid weight gain for infant Callicebus in captivity (Garber and Leigh 1997). Despite reports of captive Callicebus infants taking their first solid food at six weeks (Tardif 1994), I assume that the majority of wild infants nutrition during the IDP came from milk, as relatively little independent foraging was observed. Consequently, an adult female s daily energy requirements and energy intake rate are expected to be highest during this time. Higher energy intake rates during early lactation have been reported for wild Cebus (McCabe and Fedigan 2007) and captive Callithrix mothers, who also tended to lose weight during this period, suggesting that energy expenditures exceeded even the increased energy intake rates (Nievergelt and Martin 1999). Once traveling independently of the adult male, I observed young C. brunneus to engage in much exploratory foraging, although they returned to the female to nurse until about age eight to ten months (consistent with Wright 1984). This latter period of infant locomotor independence roughly corresponds to a gradual slowing in captive Callicebus infant weight gain (Garber and Leigh 1997), and I assume that female milk production and daily energy requirements are reduced, as suggested for wild Cebus (McCabe and Fedigan 2007).

56 46 Color Vision Determination All behavioral data collection and food patch color classification were completed before the color vision genotypes of the study animals were determined through analysis of blood, saliva, and fecal samples as described in Chapter 2. Evidence was found for three functionallydistinct opsin alleles, coding for photopigments having peak light absorbances of approximately 535 nm, 550 nm, and 562 nm (Chapter 2). Genotypes and presumed phenotypes of the study animals are presented in Table 3.1. As expected, all males were dichromats. The adult females of groups 1, 3, and W were trichromats, while the adult females from groups 2 and S were dichromats. Color vision genotypes have been shown to correspond well with color perception in platyrrhines (Saito et al. 2005a; Tovee et al. 1992) and at least one strepsirrhine (Leonhardt et al. 2008) with polymorphic trichromacy. Statistical Methods Inter observer reliability Intraclass correlation coefficients (ICCs) were used to assess inter observer agreement in the amount of time the focal animal spent in each behavioral category during reliability trials. Unobserved behavioral categories in each trial were excluded from analysis. Two way analysis of variance models with absolute agreement measures were employed (McGraw and Wong 1996; Shrout and Fleiss 1979) using the irr package (Gamer et al. 2007) of R (R Development Core Team 2008). ICCs for the eight inter observer reliability trials ranged between 0.96 and 1.00, indicating that the proportion of variance in the records contributed by the two observers was very small compared to the total variance in the data. ICCs of greater than 0.8 are considered indicative of good inter observer reliability (e.g., Slagle et al. 2002).

57 47 Regression Models The number of patch encounters by individual C. brunneus was modeled using Poisson regression with fixed and random effects. I was interested in the influence of several individuallevel predictors, or fixed effects (e.g., vision type), on the number of patch encounters by individuals. However, I hypothesized that unmeasured variables operating at the level of the group might also account for variation in individual patch encounters, e.g., variation in patch availability or predation risk between group territories. To account for these unmeasured variables, I included a group level random effect. This approach is also known as a varyingintercept multilevel model (Gelman and Hill 2007: 237). The number of patches encountered per day by an individual animal divided by the amount of time per day the individual spent searching for food patches is interpreted as the encounter rate. The amount of time spent searching for patches was estimated by the amount of time per day that an individual was observed to be traveling, and thus I included daily travel time as the exposure time in the Poisson model. The expected encounter rate was related to a set of binary predictor variables through the natural logarithm link function. These binary predictors included Sex, Vision (di or trichromacy), Season (wet or dry), and IDP (see Reproductive Status, above). It should be noted that the predictor Vision has the form of an interaction between sex and vision type. This is a consequence of the fact that females can be di or trichromatic, while males can only be dichromatic. The group level random effect was modeled by a Gaussian distribution with a variance estimated from the data. In these models, the random effect facilitates a paired comparisons design, i.e., patch encounter rate comparisons of di and trichromatic females relative to their dichromatic male pair mates. The effect of unmeasured group level variables is reflected in a group specific

58 48 intercept shared by the adult male and female in each group. Additionally, this shared intercept accounts for the correlated encounter rates of the two adults. Thus, predictors of interest for the females (e.g., Vision) can be examined in a group specific manner even though not all levels of the predictors (e.g., di and trichromatic females) are observed in each group. Multilevel Poisson regression models were fit to the patch encounter data using the lme4 package (Bates et al. 2008) in R (R Development Core Team 2008). First, a full model with all predictors, pair wise interactions, and the group level random effect was fit. This full model was reduced by sequentially eliminating predictors and interactions on the basis of hypothesized sign and magnitude of the coefficients and significance level (p < 0.05) in the model. The set of reduced nested models was compared using the Akaike Information Criterion (AIC) and likelihood ratios. Dispersion, a measure of Poisson model fit, was monitored via the sum of the squared standardized residuals (Gelman and Hill 2007: 114), also known as the Pearson statistic (e.g., Wang et al. 2002). Once a final model was chosen, residuals were plotted to check for systematic errors in fit. RESULTS Food Patch Classification A total of 1430 food patch encounters was recorded during the study period, with a mean (+/ s.d.) of 286 (+/ 150) patch encounters for each of the five C. brunneus study groups. Within this set of food patches, 151 plant and insect species could be distinguished. Sixteen fruit species were matched to the same or closely related species whose detectability to di and trichromatic platyrrhines has been modeled (Hiramatsu et al. 2008; Osorio et al. 2004; Regan et al. 2001). In all but one case, the YOR categorization used in this study coincided with model predictions of better detectability to trichromats than to dichromats. The exception was the ripe

59 49 fruit of the epiphyte Monstera obliqua (Araceae), whose color varied from pale yellow to pale orange at the study sight and was classified as YOR. The model of Osorio et al. (2004, and pers. comm.) predicted a slight, though not marked, trichromatic advantage for the detection of this species [0.44 Just Noticeable Difference (JND) in dim illumination]. Such an advantage is probably negligible (Osorio et al. 2004). However it is unclear if the range of color variation in Osorio et al. (2004) s P. obliqua samples from northeastern Peru is comparable to the variation observed at the study sight (Figure 3.3). In any case, the good correspondence between the YOR classification and the model results in most instances gives me confidence that the YOR classification is a reasonable approximation of those food species better detected by trichromats than by dichromats. Appendix C contains the list of species in discovered patches (see below) classified as YOR. Patches containing food items which could not be collected and for which no color classification could be recorded in the field were classified as non YOR patches. The vast majority of such patches probably consisted of cryptic solitary insects (see below), and therefore were unlikely to be characterized by YOR coloration. Reproductive Status An infant was born into four of the five C. brunneus groups during the study period between July and September One birth occurred just prior to the start of the study period (December 2005) in the fifth group. Between August 2005 and October 2006, a full Infant Independency Period (IDP), from birth to complete locomotor independence, was observed for two infants. The IDPs for these individuals lasted 19 and 26 weeks. One infant was observed from birth to weaning, which occurred at 38 weeks.

60 50 Foraging Overview General characteristics of C. brunneus foraging behavior are presented in Figure 3.4. For this analysis, the proportion of daily patch encounters and the proportion of daily foraging time involving food patches of a given category (e.g., fruit, insects, etc.) were calculated for all observation days. For a given patch category, the mean daily proportion of patch encounters and the mean daily proportion of foraging time were calculated across all observation days (i.e., pooling all groups across the nine month study period). Ninety five percent confidence intervals were calculated for these population wide mean daily proportions, and are plotted in Figure 3.4. Figure 3.4A shows that patches containing a range of different food classes were visited, though the animals spent 78% (95% CI: 74% to 82%) of daily foraging time in fruit patches, on average. Figure 3.4B shows that, on average, only 37% (95% CI: 33% to 42%) food patches visited over the course of a day were classified as YOR and were thus potentially more detectable for trichromatic females than for dichromatic individuals. However, on average, 66% (95% CI: 60% to 71%) of daily foraging time was devoted to these YOR patches, indicating their importance in the C. brunneus diet, their large quantity of food items per patch, and/or their long food item processing time. Figure 3.4C shows that, on average, 56% (95% CI: 52% to 61%) of food patches encountered over the course of a day were potentially located by the study animals by means of spatial memory in conjunction with non color visual landmarks (e.g., distinctive trees, rivers, swamps, etc.). These memorable patches fell into two classes: 1) patches visited on multiple occasions by any number of group individuals; and 2) patches visited simultaneously by more than one individual, but only once during the observation period. Most patches in this latter class were species that tended to produce fruit in at least moderate quantities over several days or weeks (pers. obs.). They probably represented important predictable resources to C.

61 51 brunneus and were likely to have been revisited on days when the group was not observed. Though random search and re discovery cannot be ruled out in most circumstances (see Janson 1998), productive patches exploited on multiple occasions are reasonable candidates for location by spatial memory in primates with small territories (e.g., Cunningham and Janson 2007), as such memory likely evolved to increase encounter efficiency for spatially and temporally predictable food patches (Milton 1981). On average, such potentially memorable patches accounted for 83% (95% CI: 79% to 87%) of the animals daily foraging time (see also Discussion: YOR Patch Detection and Memory). In contrast, Figure 3.4C shows that, on average, 35% (95% CI: 31% to 39%) and 9% (95% CI: 7% to 11%) of all patches encountered over the course of a day were observed to be exploited only once during the entire study period by the adult female or male, respectively, while foraging alone. The fact that such patches were exploited by only one individual suggests that they contained insufficient food items to make simultaneous exploitation energetically profitable. This conclusion is supported by the relatively minor amount of total daily foraging time (i.e., time per day that at least one adult was observed to be foraging) devoted to these patches [females: 13% (95% CI: 10% to 17%); males: 4% (95% CI: 2% to 5%)] (Figure 3.4C). I assume (as in Garber 1989) that spatial memory is less likely to be important for encounters of such small patches exploited opportunistically by lone individuals, though I cannot exclude the possibility that patches were revisited on non observation days. I assume that these discovered patches are likely to have been detected directly by means of vision each time they were encountered, rather than having constituted predetermined destinations located in relation to memorable landmarks. Included in the discovered category are patches exploited by lone individuals whose locations and coloration I could not record. This was usually because the focal animal ate a food item (presumably an insect) so quickly while traveling that I could not mark its location. I assume such foraging is also indicative of direct

62 52 visual detection rather than spatial memory. Of all food patches encountered by C. brunneus individuals over the course of a day, a significantly larger proportion, on average, was discovered by adult females than by adult males (Figure 3.4C, and above). Because I am interested in patch detection by means of color vision, I focus on discovered patches in order to reduce the effects of spatial memory which is likely to be comparable in di and trichromatic animals. However, on average, 85% (95% CI: 81% to 90%) of daily discovered patches did not contain YOR food items, and were therefore unlikely to be more detectable to trichromats than to dichromats. Thus, I limit all subsequent analyses to the approximately 15% of daily discovered patches classified as YOR. These 93 YOR patches represent at least 44 different plant species which are listed in Appendix C. On average, discovered YOR patches constituted 8% (95% CI: 5% to 10%) of C. brunneus daily patch encounters, and, across the study period, the animals devoted a mean of 6% (95% CI: 3% to 8%) of daily foraging time to these patches. On days when at least one YOR patch was discovered, 60% (95% CI: 48% to 72%), on average, of daily discovered YOR patches were encountered in forest configurations analogous to Endler (1993) s forest shade or woodland shade categories, where canopy leaves intervened, at least partially, between the food patch and the sky. Similarly, combining forest geometry with weather conditions, on average 52% (95% CI: 40% to 65%) of daily discovered YOR patches were encountered in Endler (1993) s open/cloudy light environment, where ambient light is essentially white. More detailed analyses of the light environments in which discovered YOR patches were encountered (e.g., Rowe and Jacobs 2007; Yamashita et al. 2005) were not pursued as a consequence of small sample sizes.

63 53 Comparison of Di and Trichromats Results of the multilevel Poisson regression for discovered YOR patch encounters are shown in Table 3.2. The unit of analysis for this regression is the animal day, defined as the observed behavior of one animal on one day. Simultaneous observations of both adults in each C. brunneus pair yield an N of 214 animal days. Because the number of patch encounters is expected to scale linearly with the amount of time per day that an individual spends searching, the exposure time (time spent traveling) is included in this model as an offset, i.e., its coefficient is fixed at one (Gelman and Hill 2007: 212). Sex, IDP, and the interaction of Vision and IDP were significant predictors in the final model. The predictor Season was eliminated because its coefficient was not significantly different from zero. However, the interpretation of the results is unchanged if Season is included in the model. The estimated random effect variance indicates considerable variation in encounter rate among the five C. brunneus groups, e.g., the square root of the variance is comparable in magnitude to the estimated predictor coefficients. This signals the importance of controlling for group level effects. Good model fit is indicated by a dispersion ratio of 0.99, i.e., very close to unity. This is calculated by dividing the Pearson statistic by (N k), where k is estimated at six degrees of freedom (Gelman and Hill 2007: 115, 525). A plot of the binned residuals revealed no systematic errors in model prediction. To address the hypothesis of a trichromatic female foraging advantage, population average discovered YOR patch encounter rates and 95% confidence intervals were calculated from the fitted regression model. These are shown in Figure 3.5. The average daily patch encounter rates (patch encounters per hour of travel) for di and trichromatic females could not be statistically distinguished either inside of the IDP [dichromatic females: 0.34 (95% CI: 0.13 to 0.86), trichromatic females: 0.77 (95% CI: 0.42 to 1.43)] or outside of the IDP [dichromatic females: 0.77 (95% CI: 0.37 to 1.59); trichromatic females: 0.45 (95% CI: 0.26 to 0.81)]. However,

64 54 there was a non significant trend for trichromatic females to have a higher average daily encounter rate during the IDP, while dichromatic females had a higher average rate when the infant was independently mobile (above and Figure 3.5). During the IDP, the average daily patch encounter rate of trichromatic females was significantly higher than that of their dichromatic male pair mates [trichromatic females: 0.77 (95% CI: 0.42 to 1.43); males: 0.10 (95% CI: 0.04 to 0.23)]. Interestingly, there was a strong, though non significant, trend for dichromatic females to have a higher average daily patch encounter rate than males outside of the IDP [dichromatic females: 0.77 (95% CI: 0.37 to 1.59); males: 0.22 (95% CI: 0.12 to 0.41)]. The wide confidence intervals in Figure 3.5 testify to the large individual differences in C. brunneus patch encounter rate independent of color vision, sex, infant dependence, and group membership. DISCUSSION The results of this study make three important points about the potential utility of trichromatic color vision to foraging C. brunneus. First, fruits, many of which are yellow, orange, and/or red and potentially better detected by trichromats, are an important part of these animals diet as measured by the time devoted to feeding on them. However, the task of having to detect (at a distance) patches containing YOR food items using only color vision without the aid of spatial memory is likely to be fairly rare, on average. Furthermore, patches found in this manner may represent only a small fraction of total foraging time. Second, adult female C. brunneus discover more food patches than do their male pair mates. This suggests that females may be more reliant on vision for the opportunistic detection of food patches than are males. Third, the opportunistic detection of small YOR food patches may be most important to females when they are lactating heavily, and this may be a promising context to look for a selective advantage of trichromatic color vision in primates. I consider these points and

65 55 conclude with the importance of these results to an understanding of the evolution of primate color vision. YOR Patch Detection and Memory Here I have shown that C. brunneus devotes most foraging time to food items, particularly fruits, which are likely to be more visually detectable to trichromats than to dichromats based on their coloration. This is in accord with Terborgh (1983) who found that the vast majority of fruit species eaten by Cebus, Saimiri (squirrel monkeys), and Saguinus at the study site were yellow, orange, or red in color. Interestingly, Snodderly (1979) found that nearly half of the most important fruit species (based on foraging time) of a Callicebus torquatus group in northern Peru were eaten when they were still immature and green/brown, thus likely offering no detection advantage to a trichromat. Though interpreted in a different manner, this result is not inconsistent with my finding that on average about 66% of C. brunneus foraging time was spent on YOR items, including (but not limited to) fruit (Figure 3.4B). I interpret these results to mean that, on average, much, if not most, of Callicebus foraging time is spent on YOR food items for which trichromacy may be useful. However, these results also show that percentage of foraging time is not a good indicator of the number of instances that food items are potentially detected with color vision in this species. Although, on average, most foraging time was spent in patches containing YOR items, most patches that were visited did not contain YOR items (Figure 3.4B). In addition, of the patches that did contain YOR items, most were observed (or assumed) to have been visited on multiple occasions. This opens the possibility that spatial memory in conjunction with non color landmarks was used to locate such patches. At the study site, C. brunneus have been reported to use traditional travel routes through their territories and often use the same feeding and

66 56 resting trees over consecutive years (Wright 1985), and even across generations (Chapter 4), attesting to the plausibility that spatial memory is extensively employed by these animals while foraging. Spatial memory has been hypothesized to be important for primates such as Callicebus and Saguinus living in relatively small home ranges and feeding on temporally reliable, evenly distributed, food patches (Terborgh 1983). Cunningham and Janson (2007) demonstrated this by showing that saki monkeys (Pithecia) living in a 13 ha home range (comparable to that of C. brunneus) used both spatial and episodic like memory when foraging on abundant food patches with differing productivities and states of renewal. Furthermore, Bicca Marques and Garber (2004) demonstrated that, for within patch foraging, Callicebus, Saguinus, and Aotus (night monkeys) all used spatial memory to locate food when its location was predictable. For this reason, it is likely that both di and trichromatic C. brunneus are equally adept at locating the majority of YOR patches via spatial memory because many such patches show at least some degree of temporal continuity in food production [see also Lucas et al. (2003) and Cropp et al. (2002) for variations of this argument]. In fact, I estimate that only about 8% of all patches encountered by C. brunneus, representing about 6% of total foraging time, would have been better detected by a trichromat and were simultaneously not likely to have been located on the basis of spatial memory (see Results: Foraging Overview). Thus, although YOR patches are certainly important to these animals, they may not need to be visually detected very often through YOR coloration. This result demonstrates that the importance and detectability of objects to wild primates can be deceiving if we are actually interested in the importance of the task of longrange detection of these objects. Simply because an YOR object is important and visually detectable, does not mean that primates must (or even usually) use color vision to locate it (see also Riba Hernandez et al. 2004). Thus, a selective advantage of trichromacy over dichromacy

67 57 should not be inferred directly from the quantity or importance of objects in the environment that are better spotted by trichromats from a distance. One implication of this conclusion is that care should be taken when interpreting the many valuable studies modeling the long distance detectability of naturally occurring fruits, leaves, and conspecifics to di and trichromatic primates (e.g., de Araujo et al. 2006; Dominy and Lucas 2001; Hiramatsu et al. 2008; Lucas et al. 2003; Osorio et al. 2004; Regan et al. 2001; Riba Hernandez et al. 2005; Riba Hernandez et al. 2004; Stoner et al. 2005; Sumner and Mollon 2000a; Sumner and Mollon 2003a), most of which predict an overall advantage to trichromats in terms of the number of well detected food species. To evaluate the selective advantage of trichromacy, these data must be combined with population specific information about how often each particular class of target (e.g., fruit species) is likely to be directly located by the animals primarily on the basis of vision (rather than memory). In C. brunneus, trichromacy may provide little benefit over dichromacy in most foraging contexts, despite the fact that many important food items may be better detected by trichromatic eyes. General Sex Differences in Foraging On average, adult female C. brunneus discovered nearly four times as many food patches as adult males during the observation period. As mentioned above, such patches tended to be small and were not likely to have been located by spatial memory. Because there is little evidence that Callicebus uses olfactory cues when locating food sources from distances greater than a few meters (e.g., Bicca Marques and Garber 2004), it is likely that females were detecting these patches visually. It is thus noteworthy that the polymorphic color vision of these platyrrhines results in a proportion of females, but not males, with trichromatic color vision. However, across the study groups, an average of about 85% of discovered patches did not

68 58 contain YOR items (see Results: Foraging Overview), and thus any trichromatic advantage in their detection was probably minimal. Most of these discovered non YOR patches contained leaves, stems, insects, and items eaten so quickly that they could not be identified (although I suspect they were solitary cryptically colored insects). A female foraging advantage for insects has been documented previously in this C. brunneus population by Wright (1984, using the earlier C. moloch classification). She found that adult females caught three times more insects than adult males during periods of insect abundance when females were lactating and males were burdened with carrying an infant. Similarly, Tirado Herrera and Heymann (2004) found that a lactating Callicebus cupreus female devoted about twice as much of her total foraging time to insects as she did when gestating, and about twice as much as her male pair mate. Interestingly, foraging for cryptic surface dwelling insects is a task for which dichromats may have an advantage over trichromats, as demonstrated for Cebus by Melin et al. (2007). An analysis of a dichromat advantage for non YOR targets and frequency dependent selection for polymorphic trichromacy (Mollon et al. 1984) is beyond the scope of this chapter; however, I note its plausibility in this C. brunneus population. Lactation and Color Vision With regard to the small proportion of discovered patches containing YOR items, I found that trichromatic females had a significantly higher average encounter rate than did their male pair mates during the IDP, i.e., when females were lactating heavily and males were carrying an infant. The trends apparent in Figure 3.5 suggest that this was the result of a decrease in the male encounter rate coupled with an increase in the trichromatic female encounter rate, compared to outside of the IDP. Wright (1984) suggested that male C. brunneus foraging may be impeded when they are carrying an infant. In contrast, an increase in female

69 59 encounter rate during the IDP is most likely the result of attempts to increase food consumption in order to satisfy greater energy needs during lactation (e.g., Nievergelt and Martin 1999). In a wild population of Cebus, McCabe and Fedigan (2007) found that heavily lactating females increased their ingestion rate for all types of food (fruit, seeds, flowers, and insects), but did not appear to change the relative proportions of the food types they consumed. Notably, lactating females were not observed to increase the relative proportion of protein rich items in their diet (e.g., insects) relative to sugar rich food items (e.g., fruits). This suggests that fruits, many of which may have YOR coloration, continue to be a vital component of the diet during lactation. Additionally, McCabe and Fedigan (2007) did not detect a change in the proportion of time that lactating Cebus spent foraging compared to pregnant and cycling individuals. Thus, higher energy intake was achieved by increasing consumption rate, rather than foraging time (compare to Tirado Herrera and Heymann 2004 where a small effect of this nature might potentially have gone undetected). This suggests that there may be constraints on changes in the activity time budget. These results are in accord with my findings for C. brunneus, and suggest that females can increase their net energy intake during heavy lactation by increasing the rate at which they discover YOR food patches. This will be especially important in C. brunneus if lactating females cannot sufficiently increase their within patch ingestion rates while foraging simultaneously with their pair mate in larger patches, and if lactating females cannot devote substantially more time to foraging without disrupting group cohesion. In this context, any adaptation that increases a lactating female s encounter rate for small quickly exploited patches will likely provide a selective advantage. Trichromacy may serve this purpose. For C. brunneus, it appears that trichromacy would provide a potential advantage for only a small fraction of encountered food patches to which only a small fraction of total foraging time is devoted. However, small

70 60 foraging advantages taken at the population level, especially during periods of energy stress, may be sufficient for selection. The regression analysis of discovered YOR patches was not able to distinguish between the average encounter rates of di and trichromatic females, nor between the rates of either type of female inside or outside of the IDP. There was a non significant trend for trichromatic females to have higher average YOR patch encounter rates than dichromatic females during the infant dependency period, as would be predicted if trichromacy provided a foraging advantage during critical periods of increased energy requirements due to lactation. However, the uncertainty in my population level encounter rate estimates makes this inconclusive. I suggest that future studies examining the ecological utility of trichromacy in primates pay particular attention to periods of heavy lactation in females. This may be the context in which trichromacy is most useful in forest dwelling primates, and it may be the case (as suggested by Figure 3.5) that the foraging behavior of di and trichromatic females is differentially affected by heavy lactation. Evolution of Primate Trichromacy I hypothesize that trichromatic color vision in C. brunneus may be maintained because it enhances the ability to detect small, quickly exploitable YOR food patches from a distance. This enhanced detection may translate into a net increase in energy intake. More available energy, especially during periods of lactation induced energy stress, may provide a fitness advantage to trichromatic females. The idea that selection for trichromacy in primates acts most strongly during periods of energy stress is not new (e.g., W.G. Kinzey cited in Jacobs 1997) and has previously been suggested specifically in the context of finding colorful young leaves (Dominy and Lucas 2001; Dominy et al. 2003b) and orange and red fruits (Dominy et al. 2003a; Lucas et

71 61 al. 2003) which may serve as critical fallback foods during seasonal periods of food scarcity (see also Marshall and Wrangham 2007). If this were the case, then selection for trichromacy would be expected to be similar in mechanism for males and females (although the degree to which sexes are affected could differ, e.g., Boag and Grant 1981). Energy stress, however, can also result from lactation (e.g., Nievergelt and Martin 1999). This study suggests that lactation may be better than season at explaining variation in encounter rates for YOR patches across sexes in C. brunneus populations. This is consistent with Terborgh (1983) s prediction for relatively high temporal continuity in food resources throughout the year for small territorial platyrrhines. Thus, it may be the case that polymorphic trichromacy is maintained in most platyrrhine primates by selection on lactating females in addition to (or instead of) selection on both sexes during seasonal food shortages. For C. brunneus females, selection for trichromacy may be insignificant during most years, as evidenced by the observation that, despite the small sample size, reproductively successful dichromatic females were encountered in the study population (compare to Surridge et al. 2005a for Saguinus). However, more intense selection for trichromacy may occur during super annual cycles of acute energy stress, for instance if a staple fruit species exhibits inter annual variation in production such that occasional years of fruit dearth coincide with heavy lactation (see also Chapter 4). An extreme form of such periodic selection for foraging adaptations has been well documented in Galapagos finches (e.g., Boag and Grant 1981). To statistically detect any such selective advantage for trichromacy in C. brunneus (confirming the non significant trends found here), it may be necessary to observe a larger sample of di and trichromatic females during periods of energy stress caused by heavy lactation and/or food scarcity. Additionally, it is conceivable that trichromacy in male catarrhines and in males of the only fully trichromatic platyrrhine genus, Alouatta (howler monkeys, Araujo Jr. et al. 2008; Jacobs et al. 1996a), may have originally provided a selective

72 62 advantage that was different, in either magnitude or kind, from the hypothesized advantage(s) afforded by trichromacy to lactating females. This possibility warrants further investigation. Finally, it is well recognized that selective advantage(s) currently maintaining trichromacy in extant primate populations need not resemble the selective advantage(s) responsible for its spread and fixation in ancestral primate populations (e.g., Allen 1879; Fernandez and Morris 2007). Once evolved, adaptations can be co opted for other purposes and be thereby maintained and/or modified by selection in a context different from the original (Gould and Vrba 1982). This likely happened with trichromacy during the adaptive radiation of anthropoid primates. However, if there is a useful model for investigating the ecological utility of trichromacy in the ancestral population of small, arboreal, and omnivorous catarrhines living in the tropical paleoforests where full trichromacy is believed to have first evolved in primates (Bown et al. 1982; Kirk and Simons 2001; Nathans et al. 1986b; Teaford et al. 1996), the small, arboreal, omnivorous, and morphologically conservative C. brunneus living in a seasonal tropical evergreen forest may serve well.

73 63 CHAPTER 4 Intergroup dietary differences in a wild population of the primate Callicebus brunneus and a new hypothesis linking natal dispersal with the selective maintenance of polymorphic color vision ABSTRACT Most Neotropical monkeys are characterized by polymorphic trichromacy, a system in which heterozygous females are trichromatic and can make distinctions among colors such as green, yellow, orange, and red, whereas all males and homozygous females are dichromats and are apt to confuse such colors in many contexts. Polymorphic trichromacy is presumed to be maintained in wild primate populations by natural selection. Laboratory and modeling studies suggest that trichromacy provides an advantage over dichromacy in the context of foraging for colored food objects. However, no field study has yet detected such an advantage in the daily foraging behavior of wild primates. In an attempt to reconcile these conflicting results, I develop a new hypothesis of temporally variable selection for the maintenance of polymorphic trichromacy, using the ecology of the small platyrrhine monkey Callicebus brunneus as a template. This hypothesis requires: 1) an advantage of trichromacy over dichromacy when animals forage in unfamiliar areas; 2) a mechanism, such as spatial memory of food sources, to equilibrate di and trichromatic foraging success in familiar areas; and 3) a temporally variable probability that a dispersing female will attempt to establish a new territory in an unfamiliar area in the absence of knowledgeable or trichromatic conspecifics. To evaluate the first requirement, I compare dietary differences among five C. brunneus groups in order to estimate the likelihood that a forager in an unfamiliar area will encounter unfamiliar food species more readily recognized as edible by a trichromat than by a dichromat. I then discuss the plausibility that C. brunneus relies heavily on spatial memory when foraging, as well as the likelihood that

74 64 the establishment of new territories by solitarily dispersing females is contingent on temporally variable ecological factors such as forest fruit production and predation. I conclude that the polymorphic trichromacy of Callicebus, as well as several other platyrrhines and strepsirrhines, is plausibly maintained by temporally variable selection. INTRODUCTION There is much debate about how selection has shaped, and currently maintains, the unusual variety of color vision configurations within the order Primates. The trichromatic color vision of most humans, apes, and Old World monkeys (Catarrhini) is facilitated by three classes of cone cell in the retina, each containing one of three types of photopigment: S (short wave), M (middle wave), or L (long wave) opsins, most sensitive to light at approximately 430 nm, 535 nm, and 562 nm, respectively (reviewed in Jacobs 2008; Surridge et al. 2003). Neural comparisons among the outputs of these cone cells result in perceptual distinction among colors such as green, yellow, orange, and red (Jacobs 2008; Mollon 1989). The absence of one of the three classes of cone cell results in dichromatic vision, where discrimination among such colors may be difficult in natural light environments (Sharpe et al. 1999). Dichromatic vision seems to be the norm in diurnal placental mammals (Eutheria), including most diurnal members of the Strepsirrhini (lemurs), which appear to possess only S and L like opsins in their retinal cones (Ahnelt and Kolb 2000; Jacobs 1993; Tan and Li 1999). The presence of a single class of cone cell leads to monochromacy, where only luminance (lightness) distinctions can be made in many circumstances (Sharpe et al. 1999). This visual configuration characterizes several nocturnal strepsirrhines and Aotus, the Neotropical night monkey (Jacobs et al. 1993a; Jacobs et al. 1996b; Kawamura and Kubotera 2004).

75 65 For the mammals that possess them, the L opsin is controlled by an X linked gene, while the S opsin gene is autosomal (Kawamura et al. 2001; Nathans et al. 1986a). In an ancestral (stem) catarrhine population living approximately 43 to 31 million years ago (mya), the L opsin gene appears to have undergone a duplication which eventually resulted in the functionallydistinct M opsin gene (Dulai et al. 1994; Hunt et al. 1998; Nathans et al. 1986b; Steiper and Young 2006). The fact that the M opsin and trichromacy have been conservatively maintained for at least 30 million years in catarrhines is interpreted as evidence of stabilizing selection (Jacobs and Deegan 1999; Jacobs and Williams 2001; Onishi et al. 1999; Surridge et al. 2003). Interestingly, most Neotropical monkeys (Platyrrhini) and some lemurs have achieved trichromacy in the absence of a gene duplication. Instead, these primates have developed a polymorphism at the ancestral X linked L opsin locus, comprising (usually) two or three alleles coding for opsins with maximal light absorbance between 530 nm and 562 nm. Consequently, heterozygous females are trichromatic, while homozygous females and all males are dichromatic (reviewed in Jacobs 2007; Jacobs et al. 2002; Leonhardt et al. 2008; Tan and Li 1999; Veilleux and Bolnick 2009). The similarity and conservative maintenance of the polymorphism in nearly all extant Neotropical genera suggest that it originated at least 20 mya in a basal platyrrhine (Boissinot et al. 1998; Hunt et al. 1998) and is under stabilizing selection (Surridge and Mundy 2002). Two exceptions to the rule of platyrrhine polymorphic trichromacy have been reported. The first is the monochromat Aotus (above), which has lost a functioning S opsin gene and appears to have no M/L polymorphism (Jacobs et al. 1993a). The second exception is Alouatta (the howler monkey) which, independently of the catarrhines, has undergone a duplication of the M/L opsin gene and subsequently lost all polymorphism. In Alouatta the result appears to be invariable trichromacy (in both sexes), strikingly convergent with that of catarrhines (Araujo Jr. et al. 2008; Hunt et al. 1998; Jacobs et al. 1996a; Kainz et al. 1998).

76 66 Advantages of Trichromacy It is doubtful that primate trichromacy, whether full or polymorphic, could have evolved in the absence of a selective advantage for trichromats over dichromats. However, the precise ecological context(s) in which such an advantage is (and has been) realized remain(s) unclear. Mollon (1989) suggested that trichromacy would be particularly advantageous in detecting any yellow, orange, or red (YOR) object against a dappled background of green leaves. Suggestions of which colorful targets are (and have been) most important to primates include fruits (Allen 1879; Dominy et al. 2003b; Sumner and Mollon 2003b), young leaves (Dominy and Lucas 2001; Lucas et al. 2003), conspecifics (Changizi et al. 2006; Sumner and Mollon 2003a), and predators (Caine 2002; Smith et al. 2005). Much recent effort is devoted to examining this hypothesis in the context of primate foraging by directly and indirectly comparing the food detection abilities of di and trichromatic individuals, especially in species with polymorphic trichromacy. By modeling the detectability of food items for di and trichromatic primates, numerous studies have shown that the trichromatic vision of both catarrhines and heterozygous platyrrhines may be generally superior to that of dichromats for detecting YOR fruits (de Araujo et al. 2006; Osorio et al. 2004; Osorio and Vorobyev 1996; Riba Hernandez et al. 2005; Riba Hernandez et al. 2004; Stoner et al. 2005; Sumner and Mollon 2000a) (though see Regan et al. 2001) and young leaves (Dominy and Lucas 2001; Lucas et al. 2003; Sumner and Mollon 2000a) from a distance against a background of mature green leaves, as well for making within patch distinctions among ripe and unripe (or less ripe) fruit (Sumner and Mollon 2000b). A trichromat advantage in both food patch detection and within patch fruit selection is supported by direct evidence from experiments with captive platyrrhines and strepsirrhines (Caine and Mundy 2000; Leonhardt et al. 2008; Smith et al. 2003b). However, to date, no study of the day to day foraging behavior of

77 67 wild primates has demonstrated a convincing advantage of trichromats over dichromats in either long distance food patch detection (Dominy et al. 2003a; Smith et al. 2003a, Chapter 3) or within patch foraging (Hiramatsu et al. 2008; Melin et al. 2008; Vogel et al. 2007). There is a need to explain this discrepancy between the evidence from modeling and captive experiments supporting a selective advantage of trichromacy over dichromacy, and the observations of wild primate populations where no such advantage is readily apparent. In terms of within patch foraging for ripe fruits in the wild, it is possible that both di and trichromats rely more on olfactory and tactile cues than on color vision to select food items (Dominy 2004; Dominy et al. 2001). Such cues are often (purposefully) made uninformative in controlled laboratory experiments (e.g., Araujo Jr. et al. 2008; Caine and Mundy 2000; Caine et al. 2003; Leonhardt et al. 2008; Pessoa et al. 2003; Rowe and Jacobs 2004; Rowe and Jacobs 2007; Saito et al. 2005b). Similarly, the task of locating food resources from a distance is likely to be considerably aided by the use of non color cues and spatial memory in many wild primate populations (e.g., Cunningham and Janson 2007; Janson 1998, Chapter 3), and such techniques are available to individuals regardless of color vision. However, in laboratory investigations of color vision, experimental tasks are specifically designed to control for any non color cues and spatial memory (e.g., Caine and Mundy 2000). Thus, for most day to day foraging tasks in the wild, trichromatic primates may have little (if any) advantage over dichromats for the simple reason that other ways of locating and selecting food, such as memory, odor, texture, and social facilitation, are hardly ever thwarted to the extent that they are in controlled experiments specifically designed to isolate the effect of color vision (see also Jacobs 2007). However, if this is true, we are left with a puzzle: if trichromatic color vision does not provide a selective advantage in most natural foraging situations, how has it been conservatively maintained in primates with diverse diets, predators, and habitats over many millions of years?

78 68 Alternative Hypotheses One solution to this puzzle may be that a current selective advantage of trichromacy occurs in a context other than foraging. As has long been appreciated, once evolved, trichromatic color vision can be co opted for many other purposes specific to the ecology of a particular animal, and thus may be selectively maintained in extant populations for reasons very different from those for which it originally evolved (Allen 1879; Fernandez and Morris 2007; Gould and Vrba 1982). This is likely to be the case in many catarrhine primates, where the sophisticated use of YOR color signals is plausibly of considerable importance to individual fitness (e.g., the red perineal swelling of some female cercopithecines, colobines, and chimpanzees during estrus: Hrdy and Whitten 1987; Rowe 1996) and appears to have evolved in several lineages after the evolution of trichromacy (Fernandez and Morris 2007). As mentioned above, for platyrrhine primates with polymorphic trichromacy, several non mutually exclusive hypotheses have been proposed for the selective maintenance of trichromacy in contexts other than foraging. These include detection and recognition of conspecifics (Sumner and Mollon 2003a), enhanced predator detection (Caine 2002; Smith et al. 2005), and perception of health and emotional state (Changizi et al. 2006). Sumner and Mollon (2003a) have shown that the pelage of most Neotropical primate species may be better detected against background leaves by a trichromat than by a dichromat. Smith et al. (2005) showed that, when vigilant, trichromatic Saguinus (tamarins) maintained greater inter individual distances with other group members than did dichromats. This could be explained if conspecifics and/or predators were better detected by trichromats, who then had less need to remain close enough to other group members to monitor their positions using non color cues (Smith et al. 2005). Further studies are needed to assess the hypothesis that polymorphic trichromacy is maintained through enhanced

79 69 detection of conspecifics. In terms of predation, my observations in a population of small platyrrhine monkeys (Callicebus brunneus, below) are in accord with those Terborgh (1983) and Wright (1985), and suggest that the principal predators of many arboreal Neotropical monkeys are likely to be large raptors (see below), few of whom exhibit (or are likely to be detected by means of) YOR coloration. Thus, although more data are certainly needed, an advantage in predator detection seems unlikely to contribute substantially to the selective maintenance of polymorphic trichromacy in many platyrrhines. It should be noted, however, that the situation may differ for terrestrial catarrhines with full trichromacy, who may be preyed upon heavily by felids with yellowish pelage (Coss and Ramakrishnan 2000; Isbell 1990). Changizi et al. (2006) have argued that trichromacy may be selectively maintained in extant primates because it facilitates perception of blood oxygen concentration in exposed skin, which may constitute a selective advantage in social contexts where the health, reproductive status, or emotional state of conspecifics must be judged. Although plausible for some platyrrhines, such as Cacajao calvus (red uakari), characterized by large social groups, complex social interactions, and large regions of relatively hairless skin (Ayres 1989; Fontaine 1981; Rowe 1996), an advantage in such a context seems less compelling as an important selective advantage maintaining polymorphic trichromacy in platyrrhine species with small social groups and little bare skin (e.g., Callicebus: Figure 4.1). Thus, although alternative selective advantages are possible, foraging still appears to be one of the most promising contexts in which to look for a selective advantage of trichromacy in polymorphic primates. Despite the paucity of evidence from field studies of a trichromat advantage in day today foraging situations, there may be less common foraging contexts, as yet under investigated, in which trichromatic animals would be predicted to have important (and more easily observable) advantages. One such context may be the long range detection of small ephemeral

80 70 YOR food patches by individuals who are foraging alone. The locations of such non regenerating food patches may not be worth remembering, and senses other than vision are likely to be ineffective for their long range detection. In Chapter 3 I have shown that foraging in such contexts makes up a very small part of average daily foraging time in a wild population of the platyrrhine monkey Callicebus brunneus (brown titi monkey) with polymorphic trichromacy. However, I report suggestive evidence that a trichromat advantage of this nature might become important to the fitness of females during periods of energy stress induced by heavy lactation. Future studies, both in the field and in the laboratory, are needed to test this hypothesis. It is clear from the above discussion that the selective advantage(s) maintaining trichromacy in extant primate populations may differ depending on the ecology of particular species or populations. It is also clear that detecting such advantages in the wild is likely to be considerably more difficult than may be assumed from the results of visual modeling and captive experiments. The purpose of the present study is to assess the plausibility of a new hypothesis for the selective maintenance of polymorphic trichromacy in a wild population of Callicebus brunneus in southeastern Peru. The motivation for this hypothesis is the need to (better) explain how polymorphic trichromacy is maintained in a C. brunneus population despite the (statistical) similarity of food patch encounter success among di and trichromatic females (Chapter 3) and evidence of reproductively successful groups composed entirely of dichromatic adults (Chapter 2). Here I develop the hypothesis with specific reference to the ecology of the study population, and then discuss the applicability of the hypothesis to other primates with polymorphic trichromacy.

81 71 Overview: Temporally Variable Selection for Trichromacy For forest living primates that rely heavily on predictable food sources whose locations are memorable, trichromacy may, for most of the lifespan, provide little advantage over dichromacy in terms of the rate at which these food sources are encountered. However, if, in the process of dispersal, individuals forage for unfamiliar food in unfamiliar areas without recourse to knowledgeable conspecifics, trichromacy may provide a decisive fitness advantage by facilitating the long range detection and recognition of YOR food items. I hypothesize that the importance of this selective advantage, and the likelihood that it can maintain trichromacy in a primate population over time, is dependent on the frequency with which individuals find themselves alone and searching for food in unfamiliar areas. In C. brunneus, a socially monogamous, territorial monkey in which subadults of both sexes disperse from their natal territory (Bossuyt 2002; Kinzey 1981; Mason 1966), obligate foraging in unfamiliar areas without recourse to resident conspecifics will occur whenever a disperser cannot immediately immigrate into an existing territory and must instead attempt to establish a new territory. If immigration into an existing territory is always the preferred dispersal option, then the probability that a disperser must attempt to establish a new territory will depend on the number of vacancies in existing territories at the time of dispersal. If the number of vacancies in existing territories varies over time (e.g., from year to year) contingent on local population dynamics, then the selective advantage of trichromacy will also vary over time. This hypothesis of temporally variable selection for trichromacy requires that: 1) trichromats have an advantage over dichromats when foraging in unfamiliar areas; 2) di and trichromats have equivalent foraging success, on average, in familiar areas; and 3) the probability of long term displacement to an unfamiliar area, without recourse to resident conspecifics, varies over time. The current investigation assesses the plausibility of the first requirement (above) for the C. brunneus study

82 72 population; namely, I investigate whether a trichromatic foraging advantage is likely to exist in unfamiliar areas. Following this, prospects for the second and third requirements will be discussed in relation to the study population. Foraging in an Unfamiliar Area When foraging alone in an unfamiliar region of forest, a frugivorous or folivorous primate must distinguish among types of trees from a distance and recognize which of those trees is likely to contain edible food items. Both color and non color cues can, theoretically, be used to complete both of these tasks for YOR food items. However, in the situation where an animal has little or no experience eating many of the YOR food species available in the unfamiliar region, I hypothesize that the task of recognizing trees as potentially containing edible items will be better solved by a trichromat. For instance, consider the fig Ficus trigona (Moraceae) (Figure 4.2). This large tree produces fruit which turns from green to red to dark purple as it matures (Figure 4.3) and constitutes an important food resource for many primates and birds in Western Amazonia (e.g., Terborgh 1983). In lowland forests, F. trigona is most commonly found in seasonally flooded swamps where it is the dominant tree species, and is easily recognized by its unique form of sprawling trunks and numerous supporting roots (for a vivid description of a F. trigona swamp see Terborgh 1983:11). A primate individual born into a group whose home range includes a F. trigona swamp will learn to associate food with the distinctive form of the tree, which can be spotted at a considerable distance. This will be true regardless of the color vision of the individual. If this monkey is forced to forage alone in an unfamiliar area that includes a F. trigona swamp, it is likely to detect and exploit F. trigona trees regardless of whether, from a distance, it can distinguish the reddish mid ripe fruits against a green leaf background. Now consider an

83 73 individual with no previous experience eating F. trigona fruit who is forced to forage alone in an unfamiliar area containing a F. trigona swamp. If the individual is a dichromat, it may distinguish F. trigona trees from a distance based on trunk shape, but may have no reason to suspect that the tree is worthy of investigation as a source of food. In contrast, a trichromat may detect the F. trigona trees based on trunk shape, and may also notice the color contrast of reddish fruits against green leaves. If, based on previous foraging experience with other species of red fruits, the trichromat has a cognitive association between red objects and food (i.e., signal significance of color in Jacobs 1981), then it is likely to investigate the F. trigona trees as potential food sources (and its curiosity will be rewarded). In such a situation, the trichromat may gain a fitness advantage over the dichromat because the trichromat s encounter rate (i.e., the number of foraging visits per unit time) of food trees of unfamiliar species will be higher, leading to a higher rate of long term net energy gain which can be translated into increased reproductive effort. This example illustrates that, even when foraging in an unfamiliar area where spatial memory of food resources is nonexistent, di and trichromats may be equally successful at encountering (i.e., exploiting) familiar food species that can be detected from a distance using non color cues. It has yet to be determined which (and to what extent) non color cues of potential plant food resources may be utilized by wild primates foraging in complex forest environments. However, under the assumption that YOR coloration provides a recognizable signal of potential edibility to wild trichromatic primates (Jacobs 1981; Jacobs 1993), it seems reasonable (from a conservative point of view) to conclude that a trichromat advantage in the encounter rate of food resources is likely when foraging for unfamiliar YOR food species in unfamiliar areas.

84 74 The present objective is to show that a C. brunneus individual is likely to have little or no previous experience eating the majority of available YOR food species when foraging in an unfamiliar area. I do this by examining differences in the species composition of the diet among five C. brunneus groups in a wild population. Under the assumption that a C. brunneus group exploits all, or nearly all, of the available food species in its territory, intergroup differences in dietary species composition provide a reasonable estimate of the proportion of available food resources that are likely to be unfamiliar to an individual foraging in an unfamiliar area. Under the above hypothesis, a trichromat foraging advantage in an unfamiliar area will be proportional to the fraction of available YOR food resources that consist of unfamiliar species. METHODS Study Site and Subjects The Estación Biológica de Cocha Cashu comprises a 10 km 2 study area located at approximately 350 m elevation in the interior of Manu National Park in the department of Madre de Dios in southeastern Peru (11 52 S, W). The site lies within the 6 8 km wide meander belt of the Manu River and consists of a patchwork of seasonally flooded riparian and lacustrine successional stages amidst mature high ground tropical forest (Terborgh 1983). Mean annual rainfall is approximately 2000 mm, with most rain usually falling between October and April (the wet season). Mean annual temperature is approximately 23 C (range 13 C and 33 C) (Terborgh 1983; 1990). There is a marked increase in overall forest fruit production during the wet season (Janson and Emmons 1990; Terborgh 1983; 1990). The immediate vicinity of the study site has been protected from hunting and logging since 1973 (Terborgh 1999) and contains a community of eleven primate species (Terborgh 1983).

85 75 Callicebus brunneus (Hershkovitz 1990), or brown titi monkeys (monos tocones), are small, socially monogamous, arboreal omnivores with slight sexual dimorphism, generally unspecialized morphology, and polymorphic color vision (Hershkovitz 1990; Jacobs and Deegan 2005; Kinzey 1981; Wright 1984). The mean (+/ s.d.) weight of three adult female animals captured during this study was 1040 (+/ 60) g, while one older adult male weighed 1090 g. Callicebus brunneus devote most foraging time to fruit, but also eat leaves and insects (Kinzey 1981; Wright 1985, Chapter 3). The social unit consists of an adult female and male and their dependent offspring (Wright 1984, Chapter 2), which together defend exclusive territories of 5 12 ha at this field site (Terborgh 1983 and Figure 2.1). After birth, an infant is carried almost entirely by the adult male for about four to six months until it can travel independently (Wright 1984 and pers. obs.). The particular C. brunneus population at the study site has been the subject of several ecological studies (e.g., Kinzey 1981; Terborgh 1983; Wright 1985), most recently by F. J. Bossuyt between 1996 and 2000 (Bossuyt 2002; Rodman and Bossuyt 2007). Data on the demography of the study population were obtained through records of individuals marked by Bossuyt (A. Porter, pers. comm.), as well as census and behavioral observations by JAB between 2003 and Foraging Data Collection Five habituated C. brunneus groups (1, 2, 3, S, and W), each with two to three adult or subadult individuals, along with dependent infants, were followed for a total 1062 h over 126 days from 23 January 3 June and 28 June 6 October This period spans the entire dry season (May September) and about half of the wet season. Each group was followed for h on weekdays on a randomly ordered schedule as part of a larger behavioral study

86 76 (Chapter 3). Continuous focal animal sampling (Altmann 1974) was conducted simultaneously for the adult female and male in each group by a two observer team. Each day observers were randomly assigned to a focal animal. Callicebus brunneus have a normal h daily activity cycle (Kinzey 1981). Across the five groups, the mean (+/ s.d.) observation time per day was 8.6 (+/ 1.5) h. Location of the animals at the start of each day was facilitated by a TR 4 VHF receiver and RA 14 antenna (Telonics, Inc., Mesa, AZ, USA) coupled with 25 g MI 2 radio transmitter collars (Holohil Systems, Ltd., Carp, Ontario, Canada) custom modified with no external antennas on adult females from three groups (2, 3, and W), and recorded vocalization playbacks for two groups (1 and S) using a Legendary portable guitar amplifier (Pignose Gorilla, North Las Vegas, NV, USA). For the present analysis I focus on foraging, which was defined as ingesting, manipulating (with the hands or mouth), carrying (in the mouth), searching (within a food patch) for, or usurping (from another group member) a food item. Due to the density of the forest, there were also instances where a foraging focal animal was either partly or completely hidden from view. The animal was scored as foraging out of view if falling food items could be heard or seen originating from the animal s location. Diet Sampling I define a single food patch as a physical location (e.g., a tree) of a certain food type (e.g., a fruit species) visited by an individual or individuals for a continuous time period, even if temporally separated foraging events occurred during the visit. I attempted to identify and mark the locations of all food patches used by the study animals on observation days. Where possible, food items dropped by the animals or collected from the same patch were matched to the same or closely related species whose detectability for di and trichromats has been modeled by

87 77 Regan et al. (2001), Osorio et al. (2004 and D. Osorio, pers. comm.), or Hiramatsu et al. (2008). Patches containing food items predicted by these models to be more easily detected by trichromats were assigned to the YOR category (patches characterized by yellow, orange, or red). If a food item could not be matched to a previously modeled species, it was classified as YOR if a normal human trichromat (JAB) had described it as yellow, orange, red, or pink at the time of collection. Using human color classifications to represent animal vision is often inappropriate, especially when an animal s visual perception is known to be very different from that of humans (Bennett et al. 1994). However, in general, the set of colored food items distinguishable from background leaves by a normal human trichromat is predicted to match or encompass the set distinguishable to any polymorphic platyrrhine phenotype (de Araujo et al. 2006; Osorio et al. 2004; Riba Hernandez et al. 2004; Stoner et al. 2005) (though see Regan et al. 2001). Thus, this classification is conservative in that, at worst, the YOR category includes some patches that are equally (un)detectable to di and trichromatic study animals. Classifications also took into account features of the food patch that were not consumed by the animals, for instance YOR colored immature fruits (e.g., Stylogyne cauliflora [Myrsinaceae], Figure 3.1) and peduncles (e.g., Huertea glandulosa [Staphyleaceae], Figure 3.2). Analysis of Dietary Similarity Calculation of the Jaccard Estimator Similarity in the food species composition of the diets of the five C. brunneus study groups was assessed using all observed food patches, and separately using only food patches in the YOR category. For each pair of groups, the abundance based Jaccard index estimator of Chao et al. (2005; 2006) was calculated. This estimator, hereafter referred to as the Jaccard estimator, is based on the frequency with which patches of a given food species were observed

88 78 to be encountered by the animals across the study period, and includes an adjustment for those food species which were likely to be present in the actual diets of both groups being compared, but which were not observed in the diet of at least one of the two groups during the study (i.e., unseen shared species). Following Chao et al. (2005; 2006), I define the Jaccard estimator,, as: 1 where and where a total of food patch encounters was observed for one C. brunneus group (e.g., Group 1) and food patch encounters was observed for a second group (e.g., Group 2). The frequencies of the various food species in the collection of observed patch encounters for Group 1 are represented by the set,,, and for Group 2 by the set,,,. The pairs of frequencies for food species observed to have been eaten at least once by both groups (i.e., the shared food species) comprise the set,,,, where is the

89 79 number of pairs in this set of shared species. The variable represents the number of shared food species that were observed to be eaten only once (singletons) by Group 1 and at least once by Group 2. The variable represents the number of food species observed to be eaten exactly twice (doubletons) by Group 1, and at least once by Group 2. Similarly, and represent the numbers of shared singleton and doubleton species, respectively, in the observed diet of Group 2. expression is an indicator function such that 1 if expression is true and 0 if expression is false. Standard error estimates for the Jaccard estimator, following the bootstrapping procedure of Chao et al. (2006: supplementary online material), were calculated with 3000 resamples using the base package of R (R Development Core Team 2008). Interpretation of the Jaccard Estimator The value of the Jaccard estimator approaches zero if the diets of the two C. brunneus groups have no shared food species, and reaches one if all food species are shared (although food species may have different frequencies in the two groups). The adjustment for unseen shared species (i.e., the second terms in equations 2 and 3) will increase the value of the Jaccard estimator if there is a large number of observed singletons in the diets of the C. brunneus groups, indicating the presence of many rare food species. The adjustment will be especially strong if the observed singleton food species for one group are commonly encountered by the other group, indicating that larger samples (e.g., more observation days) would likely reveal that many common food species observed in only one group are actually shared by both groups. The Jaccard estimator can be interpreted as follows: Take one randomly chosen foraging event from each of two C. brunneus groups. The Jaccard estimator represents the probability that, given that at least one of the two foraging events involves a food species shared by both groups, both foraging events involve food species shared by both groups. For instance, a Jaccard estimator

90 80 value of 0.6 for the observed diets of two C. brunneus groups would suggest that, on average, for at least 40% (1 0.6) of all foraging events by the two groups, at least one of the two groups was eating a food species never eaten by the other. Application of the Jaccard Estimator Chao et al. (2005; 2006) used the Jaccard estimator to measure the similarity in species composition among plant assemblages in several forest types using 1 ha sample plots. In this study, I use the Jaccard estimator to measure the similarity in food species composition among the diets of C. brunneus groups using randomly scheduled observation days over the course of the study period as random samples of the actual diets of the C. brunneus groups during this time. The Jaccard estimator assumes that sampling is done with replacement. This assumption is violated in the present study (e.g., an observed food patch encounter cannot be replaced in the pool of all food patch encounters and thus be potentially re sampled). However, the assumption of sampling with replacement can be relaxed if the population (e.g., the set of all food patch encounters by a C. brunneus group during the study period) is large relative to the size of the sample (e.g., the set of all observed food patch encounters by a particular group) (Plotkin and Muller Landau 2002). A further complication of the present data is the fact that the set of food species available to a C. brunneus group on any given observation day does not represent a random sample of all observed food species for the group over the course of the study period. This temporal heterogeneity in available food species is a consequence of seasonal variation in food (e.g., fruit) production in the monkey home ranges, and is analogous to the spatial clustering of plant species in tropical forests (e.g., Condit et al. 2000) that is often encountered by studies estimating the similarity in plant species composition between two forest regions (Plotkin and

91 81 Muller Landau 2002). Plotkin and Muller Landau (2002) showed that most traditional similarity indices tend to overestimate similarity between two assemblages when species are clustered in space. It is unclear how the marked temporal variation in the present data affects the behavior of the Jaccard estimator. However, bias in estimation is likely to be reduced by the fact that at least 23 dietary samples (i.e., observation days) were conducted for each group across the study period, thereby sampling much of the temporal heterogeneity in diet, and by the adjustment for unseen shared species, which would partially account for any un sampled food species with small temporal windows of availability or use. For stability of the estimate, Chao et al. (2006) recommend that the Jaccard estimator be used when there are at least ten shared species in the assemblages being compared, and at least one assemblage contains a doubleton. These conditions were satisfied in all cases except three comparisons using YOR food items: Groups 1 and 2, 1 and S, and 3 and S had nine shared YOR species. Thus, these three comparisons should be viewed with caution. Unidentified Food Items In addition to all identified food species, the Jaccard estimator calculations included four categories of unidentified food items: Unknown, Unknown Fruit, Unknown Flower, and Unknown Leaf/Stem. The Unknown category consisted almost entirely of single small items eaten so quickly (usually while an animal was traveling) that they could not be identified, but many may have been small, solitary, surface dwelling invertebrates. Assuming this is the case, the Unknown category would comprise food items different from any other identified food, as nearly all other identified invertebrate prey were large (e.g., katydids [Tettigoniidae]), social insects (e.g., army ants [Eciton sp.]), or required distinctive and observable processing (e.g., insect/spider egg sacks). Because they were not seen directly, no items in the Unknown

92 82 category were classified as YOR. The other three unknown categories, for fruit, flowers, and leaves/stems, contained items observed to be eaten by the animals, but which were not collected for definitive species identification. However, observations in the field suggested that these items were different from any identified species (and morphospecies). In this analysis, all items included within each unknown category were treated as if they were the same species. Because food items in each unknown category, except Unknown Flower, were recorded for all C. brunneus groups, the present analysis will likely lead to over estimation of dietary similarity if not all items in the unknown categories were actually of the same species (probable), or underestimation if many items in the unknown categories actually belonged to previously identified species (less probable). Thus, in this analysis, dietary similarity based on the Jaccard estimator should be viewed as an approximation, and probable over estimate, of actual dietary similarity. RESULTS Demography The territories of all five C. brunneus groups observed in this study were occupied during a previous study by F. J. Bossuyt between 1996 and 2000 (Bossuyt 2002; Rodman and Bossuyt 2007). Many feeding and resting locations marked by Bossuyt in these territories were still in use, even in cases where the territory residents appeared to have changed in the six year interval between the past and present studies. The adult pair from one group (1) and the adult males from two other groups (2 and S) still bore radio and identification collars originally fit by Bossuyt. This indicates that adult tenure in this population commonly lasts at least six years, and perhaps considerably longer. Definitive change in pair membership was evident in two territories where Bossuyt had collared individuals. The original (i.e., at the time of Bossuyt) adult female of Group 3 was observed in December 2004, but had been replaced by a new female by

93 83 July The original adult male in Group 3 had been replaced by a new male by July The adult female in Group W had been replaced by a new female by October Additionally, over the course of this study (January October 2006), replacements of resident adult males were confirmed in Groups 3 and W (Chapter 2). The history of this C. brunneus population suggests that replacement of adults, either through death (e.g., predation) or eviction, usually occurs to one adult in a pair at a time. As explained below, this has important consequences for the transmission of spatial memory of food resources in a territory as well as for the establishment of new territories by dispersing individuals. Diet Composition A total of 1754 C. brunneus food patch encounters was recorded during the study period, with a mean (+/ s.d.) of 352 (+/ 183) observed patch encounters for each of the five groups. Appendix D lists the 158 distinct food species (or morphospecies) observed to be eaten during the study period, hereafter referred to as distinct observed species, and which were included in the analysis of dietary similarity (Appendix D, no asterisks). Of these, 136 food species were identified at least to the level of family (plants) or order (invertebrates). Opportunistic feeding observations recorded while habituating the C. brunneus groups prior to the start of the study (July December 2005) and during non observation days (e.g., when capturing the animals) increases the number of distinct observed food species to 180. These additional 22 plant species (Appendix D, single asterisk) were not included in the analysis of dietary similarity (below) however, as diet sampling was not systematic during habituation and capture. Across all five groups, many patch encounters fell into the unidentified food categories (not listed in Appendix D): Unknown (672 total encounters, range by group: ), Unknown Fruit (34 encounters, range by group: 5 9), Unknown Flower (5 encounters, range by group: 0 3),

94 84 and Unknown Leaf/Stem (251 encounters, range by group: 36 77). Under the assumptions (above) that all food items in each unknown category were the same species, and each unknown category contained a species not counted among the distinct observed species in Appendix D, then C. brunneus in this study were observed to eat 184 food species. The dietary species richness observed in this study (over 180 species) is consistent with Wright (1985) s report of over 100 fruit and flower species (of which 88 were identified at least to genera) eaten by two C. brunneus groups across a total of 15 months at the study site. Richness appears to be somewhat higher than the approximately 70 food plant species observed by Lawrence (2007), following seven groups of C. brunneus for over a year in terra firme forest on a river system adjacent to that of the study site. Of the food plants identified by Lawrence (2007), ten genera and three species (the palms [Arecaceae] Iriartea deltoidea and Euterpe precatoria, and the myristicac tree Virola calophylla) were observed to be eaten during the present study. In addition to the 180 distinct observed food species I recorded over nine months, Appendix D also includes 26 plant genera and one insect order that were reported to be eaten by a single C. brunneus group at the study site during 57 h by Kinzey (1981, four asterisks), eight days by Terborgh (1983, three asterisks), and nine months by Wright (1985, two asterisks), in addition to a second group at the study site observed by Wright (1985) for 11 months (two asterisks). Genera, rather than species, are listed from previous studies (even if species were originally reported) in order to reduce duplicate recording in Appendix D as a consequence of changes in accepted plant taxonomy over 30 years. Taken together, these results suggest that the C. brunneus population at the study site has been observed to consume over 198 different plant species and at least five orders of invertebrates.

95 85 Included in Appendix D are the months in which particular food species were observed to be eaten in this study, the C. brunneus groups (1, 2, 3, S, and W) observed to eat them, and total number of patch encounters observed for each food species. It is evident that few species were eaten for more than a few months of the year, likely reflecting seasonal production of preferred fruit species. It is also notable that no plant species (or invertebrate order) was observed to be eaten by all five groups during the study period. For several species, such as the small tree Lunania parviflora (Flacourtiaceae) and the palm Ireartea deltoidea (Arecaceae) which are fairly common in C. brunneus inhabited forest (Gentry and Terborgh 1990), the lack of observed patch encounters for several groups may be the result of sampling error. For several other species, such as the mid size trees Guatteria discolor (Annonaceae) and Pseudolmedia laevis (Moraceae), peak fruit production of most individuals probably occurred outside of the nine month study period (Wright 1985 and pers. obs.). The absence of observed patch encounters for these two species by Group 1 is likely due to the fact that individual trees in the home range of Group 1 were not yet producing a sufficient crop of fruit to attract attention. Pseudolmedia laevis was observed to be eaten by Group 1 during habituation in October and November of For plant species with more restricted or patchy distributions, the lack of foraging observations from one or more C. brunneus groups may be the result of the absence of the species from a group s home range. This is especially likely for species such as the bamboo Guadua weberbaueri (Poaceae, growing in large stands), the large tree Sloanea obtusifolia (Elaeocarpaceae), and the swamp specialist Ficus trigona (Moraceae). The presence of such species, even when not producing fruit or edible shoots, would probably have been noticed by observers following a C. brunneus group around a 12 ha territory for several days each month. Thus, the lack of observed foraging events for such species by several groups is probably the result of a lack of availability rather than sampling error or a lack of preference on the part of

96 86 the animals. This interpretation must be viewed with considerable caution, however, until extensive (and exhaustive) botanical surveys of each C. brunneus territory can verify the absence of particular food species. The important observations from Appendix D are that the C. brunneus diet has high richness and appears to vary considerably among groups. Of the 158 distinct observed food plant species recorded during the nine months of this study, 81 species were classified as occurring in YOR patches. Eighteen C. brunneus food plant species were matched to the same or closely related species whose reflectance spectra has been previously measured and whose detectability to di and trichromatic primates has been modeled (Hiramatsu et al. 2008; Osorio et al. 2004; Regan et al. 2001). In all but one case, the YOR categorization used in this study coincided with model predictions of better detectability to trichromats than to dichromats. The exception was the ripe fruit of the epiphyte Monstera obliqua (Araceae), whose color varied from pale yellow to pale orange at the study site (Figure 3.3) and was classified as YOR. The model of Osorio et al. (2004, and pers. comm.) predicted a slight, though not marked, trichromatic advantage for the detection of this species (0.44 just noticeable difference [JND] in dim illumination). Such an advantage is probably negligible (Osorio et al. 2004). However it is unclear if the range of color variation in Osorio et al. (2004) s P. obliqua samples from northeastern Peru is comparable to the variation observed at the study site. In any case, the good correspondence between the YOR classification and the model results in most instances gives me confidence that the YOR classification is a reasonable approximation of those food species better detected by trichromats than by dichromats. Intergroup Dietary Similarity Each of the five C. brunneus groups was observed to eat an average of 54.8 plant and invertebrate species (including unknown categories) over the nine month study period. This is

97 87 approximately 1/3 of the 162 food species (158 distinct observed species plus four unknown categories) observed to be eaten across all groups over the study period. Group 3 had the most species rich diet (83 species), while Group S had the most species poor diet (29 species). Comparing the observed diets of each pair of groups, an average of 17 food species were shared, with a range of 11 shared species (Groups 1 and S) to 24 shared species (Groups 1 and 3). Table 4.1 presents the values of the Jaccard estimator for each pair of C. brunneus groups, calculated separately using only YOR food items and all observed food items. For the YOR component of the diet, values ranged between 0.26 (Groups 1 and 2) and 0.50 (Groups S and W), with a mean (+/ bootstrap SE) of 0.36 (0.19) across all ten pairs of groups. This suggests that, on average, for at least 64% (1 0.36) of YOR food patch encounters by a pair of C. brunneus groups, at least one group was eating a species never eaten by the other. When all observed food species were examined, similarity estimates tended to be higher, ranging between 0.55 (Groups 1 and S, 2 and 3, 3 and S) and 0.68 (Groups S and W) with a mean (+/ bootstrap SE) of 0.62 (0.32) across the ten pairs of groups. This suggests that, on average, for at least 38% of all food patch encounters by a pair of C. brunneus groups, at least one group was eating a species never eaten by the other. The large standard errors obtained through bootstrapping procedures (Chao et al. 2006: supplementary online material), indicate that the exact values of the Jaccard estimator reported here must be interpreted with caution. A comparison of the Jaccard estimator calculations to the number of shared (identified) food species among pairs of groups (above), suggests that, although relatively few food species were likely to be shared in the diets of any two C. brunneus groups, species that were shared were likely to be exploited often by both groups. This is consistent with previous observations of C. brunneus foraging behavior, in which, on a given day, most group foraging time was directed

98 88 toward a few important food species (Kinzey 1981; Lawrence 2007; Wright 1985). Taken together, however, the richness and Jaccard similarity comparisons support the interpretation of Appendix D, i.e., that dietary species composition varied among C. brunneus groups, especially with regard to YOR foods likely to be more detectable to trichromats than to dichromats. DISCUSSION Trichromatic Advantage in Unfamiliar Areas Callicebus brunneus in the study population have diets rich in plant and invertebrate species, including many YOR species for which trichromatic color vision is predicted to be advantageous. Furthermore, many species present in the diet of one group are likely to be absent from the diet of another randomly chosen group. Thus a C. brunneus individual foraging in an unfamiliar territory would be expected to have little or no experience eating many food species, especially YOR species, that would be eaten by a group familiar with the area. Consequently, using only non color cues, such an individual may be unlikely to recognize as worthy of investigation many patches containing edible food items. However, for a trichromatic individual who associates YOR coloration with potential edibility, many patches containing unfamiliar YOR food species would be detected at a distance and recognized as potentially containing food. Callicebus brunneus in this study population spend an average of 66% of daily foraging time in YOR food patches (Chapter 3). Thus, a trichromat foraging alone in an unfamiliar area may have a considerable advantage over a dichromat in a similar situation, not only because more food patches containing unfamiliar YOR species are likely to be exploited, but also because such food patches are likely to be those of greatest importance (measured in terms of time spent foraging) in a diet that would be optimal for the area. It appears reasonable to

99 89 conclude from these data that a trichromatic foraging advantage is likely to exist in this population for C. brunneus forced to forage in unfamiliar areas. This satisfies the first requirement for temporally variable selection for trichromacy.. Spatial Memory Within Territories The second requirement for temporally variable selection is that di and trichromats have equivalent success, on average, when foraging in familiar areas. This condition may be satisfied for C. brunneus if spatial memory, in conjunction and non color cues, can be used to locate all important food patches within the territory of a resident monkey group. In Chapter 3 I showed that, on average, approximately 8% of food patches encountered over the course of a day by C. brunneus were characterized by YOR coloration and, simultaneously, were not likely to have been located by means of spatial memory (i.e., discovered YOR patches). Such patches were potentially better detected from a distance by trichromats than by dichromats. However, on average, discovered YOR patches accounted for only approximately 6% of daily foraging time (compare to the approximately 66% of daily foraging time, on average, in all YOR patches, above) and there was no statistical difference in the average patch encounter rates of di and trichromatic females. In contrast, the vast majority of C. brunneus food patch encounters (approximately 92%) involved patches that: 1) did not (or were presumed not to) have YOR coloration; 2) were observed to have been exploited on multiple occasions; or 3) were exploited simultaneously by two or more individuals. Trichromacy is expected to provide little advantage in the encounter rate of such patches. For instance, patches that are visited on multiple occasions suggest that food items (e.g., fruit) are produced over a series of days or weeks, indicating temporal predictability. Similarly, patches visited simultaneously by multiple individuals suggest the production of large quantities of food items, many of which are probably

100 90 also available over a series of days. The locations of both of these types of patches are likely to be worth remembering by animals, regardless of color vision, as they are renewable and predictable over the short term (Chapter 3). An animal s use of spatial memory to locate food resources should not be assumed without an independent measure of species and habitat specific patch detection distance, in order to rule out sequential rediscovery of patches observed to be exploited by animals on multiple occasions (Janson 1998; Janson and Di Bitetti 1997). Spatial memory (as well as temporal memory) of fruit resources has been objectively demonstrated for primates foraging over areas comparable in size to those of C. brunneus territories (Cunningham and Janson 2007), as well as for primates with much larger ranges (Janmaat et al. 2006; Janson 1998). Bicca Marques and Garber (2004) showed that, at least for within patch foraging, Callicebus appear to be adept at learning the spatial locations of food items, in some contexts requiring only a single exposure to a food stimulus in a fixed location. Thus, the use of spatial memory by C. brunneus in the location of food patches is certainly plausible. In fact, Terborgh (1983) has proposed that small territorial platyrrhines, such as Callicebus and Saguinus, which rely on predictable food resources evenly distributed both temporally and spatially, are especially likely to rely on spatial memory for efficient foraging. At the study site, C. brunneus foraging and ranging behavior is consistent with the idea that spatial memory is commonly employed while foraging. Wright (1985) described C. brunneus as engaging in slow, methodical journeys from one small fruit resource to another, and reported that certain travel paths and fruit trees observed to be used by the same group of monkeys five years prior to her study were still actively used by the animals. Additionally, many traditional pathways through the territory were used by Wright s study groups for three consecutive years (Wright 1985). These observations accord well with my own (see below). Thus, it appears plausible (if not likely) that nearly all food resources exploited

101 91 by C. brunneus foraging within familiar territories can be encountered with nearly equal success, on average, by di and trichromatic individuals using a combination of spatial memory and noncolor cues. Even if spatial memory and non color cues are used extensively by resident C. brunneus, trichromatic individuals may have a foraging advantage during the initial phase of memory formation, however. For instance, trichromacy may be advantageous in the initial detection of a YOR food patch, the location of which can then be subsequently memorized. Such a situation is likely to occur infrequently in the study population, as, regardless of color vision, the spatial memory of a young individual born into a population can be developed by foraging alongside adults, particularly the adult male (Kinzey 1981; Kinzey et al. 1977; Wright 1984, and pers. obs.). Additionally, there is evidence (below) that a new immigrant into an established territory (i.e., a successfully dispersing subadult) can almost always rely on the knowledge of its new pair mate (i.e., resident adult) to develop spatial memories of the locations of food resources in the initially unfamiliar territory. Thus, few individuals in established territories may have to independently discover important YOR food patches. This is suggested by the demography of the study population: Individual replacement in established territories appears to occur one adult at a time over a period of years (see Results: Demography). Coupled with the fact that C. brunneus pair mates tend to forage within sight of each other (Mason 1966 and pers. obs.), piecemeal replacement of residents would allow knowledge of the locations of important food sources to be transmitted from a resident adult to any entering disperser who replaces his or her established pair mate (see Smith et al. 2003a for a related suggestion). For instance, spatial knowledge originally acquired by a trichromatic female can be learned by her dichromatic male pair mate, who can then transmit the spatial knowledge (probably passively) to a dichromatic

102 92 female who replaces the original trichromatic female. In this way, spatial knowledge of food resources in a stable territory could potentially span the tenures of many adults, regardless of their color vision. Thus, it appears likely that, in this C. brunneus population, di and trichromatic individuals have comparable success, on average, when foraging in familiar territories. This satisfies the second requirement for temporally variable selection for trichromacy. Dispersal and Immigration The final requirement for temporally variable selection is variation in the probability of long term displacement to an unfamiliar area without recourse to resident conspecifics. As mentioned above, natal dispersal appears to be the primary cause of long term displacement to unfamiliar areas in C. brunneus. Natal dispersal by both sexes is the norm in C. brunneus, as in other members of the genus Callicebus (Kinzey 1981; Mason 1966). In order to reproduce, a dispersing individual of either sex must immigrate into an established territory of an existing adult pair and fill a vacancy left by a dead adult, or permanently evict the same sex adult through direct confrontation. Alternatively, a disperser could unite with a dispersing (or evicted) individual of the opposite sex and establish a new territory in a currently unoccupied area. A third possibility is for a subadult to remain in the natal group and replace the same sex adult. However, this is probably unusual for reasons of incest avoidance. Because Callicebus territories generally have minimal overlap (Kinzey et al. 1977; Wright 1985) (though see Price and Piedade 2001 for a possible exception in C. personatus), a disperser will, in general, have no spatial knowledge of food resources once it leaves its natal territory (though see Easley and Kinzey 1986 for a possible exception in C. torquatus). Similarly, aside from (generally hostile) intergroup encounters with groups whose territories border the natal territory (e.g., Mason 1966; Wright 1985), a disperser is unlikely to be familiar with other conspecifics in the area. Thus, dispersal in

103 93 Callicebus is usually both locational and social (Isbell and Van Vuren 1996), in that a disperser permanently leaves both a familiar area and a familiar social group. As discussed in the previous section, di and trichromatic dispersers who have successfully integrated into an established territory with a knowledgeable previously resident pair mate, are likely to have equivalent foraging success, on average. Thus, from the perspective of newly dispersing female, a dichromatic female resident in an established territory should, on average, be no more susceptible to eviction (e.g., due to under nourishment) than a trichromatic female territory resident. Consequently, the opportunities for a dispersing female to gain entrance into an established territory, and, once resident, to resist eviction by subsequent dispersers, is expected to be the same regardless of whether the female is dichromatic or trichromatic. Immigration into an established group provides a dispersing C. brunneus female with immediate reproductive opportunities as well as access to the pair mate s spatial knowledge of food resources, which likely results in much more efficient foraging. Thus, immigration into an established territory is expected to be the favored option of any C. brunneus disperser. However, in the event that no vacancy or evictable resident female can be found among the established territories of a C. brunneus community, the only (reproductively compatible) options for a dispersing female are to attempt to establish a new territory in an unoccupied region of forest and attract a male, or join a male who is attempting to establish a new territory. In either case, the establishment of the new territory is likely to take place in a region of forest unfamiliar to both individuals. This scenario is consistent with a report (J. Terborgh cited in Wright 1985:121) that new C. brunneus groups at the study site occasionally appear in areas where they had apparently not existed previously. New group formation in other Callicebus species, however, may occur differently. Easley and Kinzey (1986), showed that a C. torquatus group gradually shifted its territory to a completely new area over seven years. The authors suggest

104 94 that such range shifts may allow offspring to establish new territories in areas that partly overlap familiar regions of their previous natal range (Easley and Kinzey 1986). In an unusual situation of supernumerary males, Bicca Marques et al. (2002) reported the formation of a new C. cupreus group by a dispersing male who established a territory adjacent to the natal territory. However, the apparent stability of territories in the C. brunneus study population over several years suggests that most dispersers who are forced to establish new territories probably do so in unfamiliar areas at a distance from the natal territory. As the results of this study suggest, trichromatic female dispersers attempting to establish new territories in unfamiliar regions are likely to enjoy greater initial foraging success than their dichromatic counterparts. The costs associated with establishing a new territory are likely to be high. For instance, in a 13 year study of the small territorial platyrrhine Saguinus (tamarin) with bisexual natal dispersal, Goldizen et al. (1996) estimated that perhaps as many as 12 dispersing subadult females (13 subadult disappearances 1 successful immigrant from outside the population), representing approximately 40% of all dispersal aged females, were unable to immigrate into existing territories, and only two new territories were likely to have been established in the study area during the observation period (Goldizen et al. 1996). This suggests that perhaps 80% (10/12) of female dispersers who must attempt to establish new territories are unsuccessful and probably die. High mortality during locational dispersal in mammals is often attributed to predation and lack of knowledge of food resources (see Isbell and Van Vuren 1996). An adaptation such as trichromacy can improve foraging efficiency by enabling a small monkey to reduce the time it spends searching for food patches in unfamiliar areas, thereby also reducing exposure to predators (see below). The strength of natural selection for trichromacy in a population will be contingent on how often dispersers must attempt to establish new territories.

105 95 Likelihood of New Territory Establishment Under the assumption that the first choice of a dispersing C. brunneus individual is always immigration into an established territory, then, at any given time, the number of founder individuals, F, that will attempt (successfully or unsuccessfully) to establish new territories in novel areas can be expressed as: F = D (E + W) 4 where D is the total number of dispersing individuals in a population at a given time. E is the number of resident adults in established territories that are evictable, and, once evicted, will die without successfully entering or establishing any additional territories. W is the number of widowed individuals, i.e., resident adults in established territories whose pair mate has died. W represents the number of current vacancies in established territories that dispersers can fill. In the interpretation of Equation 4 I make several assumptions: 1) an equal proportion of the values of all variables represents females; 2) any disperser in D is equally capable of evicting any same sex resident adult in E, and would be equally accepted as a pair mate by any widowed opposite sex resident adult in W (i.e., widowed resident adults cannot or do not discriminate among potential immigrants on the basis of photopigment allele complement, see Chapter 2); and 3) resident adults who have been evicted and do not die (i.e., who are capable of secondary dispersal and/or new territory establishment) are included in D. In most populations, it may be reasonable to assume that E is a constant, as, in the absence of abnormal population wide disease, E is likely to be strongly correlated with species typical rates of senescence (e.g., weak older adults who will die once evicted). In the C. brunneus population at the study site, W is most likely determined by the rate of predation on adults, as most predators of these monkeys

106 96 (see below) appear largely incapable of capturing (and carrying) two adult individuals simultaneously in a single predation attempt on a group. Therefore, successful predation on adult C. brunneus usually results in widowed individuals. D is dependent on the birth rate and juvenile survival. D may be increased by an increase in the productivity of the environment (e.g., a particularly productive fruit season resulting in better nourished infants) and/or a decrease in predation pressure on infants and juveniles. Thus, in this C. brunneus population, at any given time, the probability that a disperser must attempt to establish a new territory (F/D), and consequently, the selective advantage of trichromatic color vision, may be influenced to a large extent by predation regimes and fruit and leaf production cycles. To satisfy the final requirement of temporally variable selection for trichromacy, such ecological conditions must be shown to change over time. Temporal Variation in Food Production and Predation Adult female C. brunneus in the study population usually give birth to a single infant once per year (Chapter 2). The probability that a female will have a successful pregnancy and that her offspring will survive from birth to dispersal age (approximately three years, e.g. Kinzey 1981) may be influenced by the amount of available food in the territory at critical periods, such as peak lactation and weaning (see Chapter 3). There are indications that both total forest fruit production and the fruit production of individual tree species can vary substantially from year to year in forests populated by C. brunneus. For instance, Terborgh (1983; 1990) provides evidence for a general trend of higher total fruit availability from September through April and lower availability from May through August at the study site. However, the amplitude of the fruit availability cycle, as measured by fruit collected in traps positioned 0.4 m above the forest floor (Terborgh 1983), appears to vary from year to year (e.g., Terborgh 1990: Figure 3.1). Such

107 97 changes in amplitude, as measured by fruit fall, could be due to changes in the total production of fruit by trees in different years and/or by yearly changes in the total consumption of fruit by frugivores (e.g., primates and birds) foraging in the canopy (see Terborgh 1983). Super annual cycles of fruit production by Neotropical trees have been documented through long term phenological studies (e.g., Bendix et al. 2006), and this is often interpreted as an evolved response of trees to heavy seed predation (e.g., Janzen 1970). However, regardless of whether inter annual variation in fruit availability is due to variation in fruit production or competition with other frugivores, the effect for a small monkey such as C. brunneus, which may spend nearly 80% of daily foraging time on fruit (Chapter 3), is likely to be the same, namely, interannual variation in net energy intake. This may translate into inter annual variation in birth rate and infant/juvenile survival, which in turn may (after a lag of approximately three years) lead to variation in the number of potential dispersers in any given year. It is probable that predation pressure on C. brunneus also varies from year to year. At the study site, the principal predators of C. brunneus are likely to be raptors (see Terborgh 1983; Wright 1985) and possibly Cebus sp. (capuchin monkeys, e.g., Lawrence 2007; Wright 1985, and pers. obs.), with potentially minor predators being felids, e.g., ocelots (Felis pardalis), the mustelid tayra (Eira barbara), and snakes, e.g., boas (Boa sp.) (Lawrence 2007; Terborgh 1983; Wright 1985, and pers. obs.). Terborgh (1983) lists six sympatric eagle and hawk species that have been observed to attack primates at the study site, and Wright (1985) observed attacks on C. brunneus by at least three of these species. From August 2005 to October 2006 I (and my coworkers) observed six unsuccessful avian predation attempts on four of the five C. brunneus study groups (all but Group S). At least five attacks were by ornate hawk eagles (Spizaetus ornatus), a medium sized eagle with a diverse diet of mammals, birds, and reptiles (Klein et al. 1988; Terborgh 1983). Robinson (1994) suggests that eagles such as S. ornatus are likely to have

108 98 ranges greater than 100 ha at the study site. Thus, C. brunneus territories (5 12 ha) are small relative to the hunting ranges of its primary avian predators. The likelihood that a particular C. brunneus group will be preyed upon by eagles such as S. ornatus may be related to the proximity of the monkey territory to an active eagle nest site (A. Porter, pers. comm.), as one or both members of a nesting S. ornatus pair must return to the nest site with food for the offspring on most days (Klein et al. 1988; Lyon and Kuhnigk 1985). Klein et al. (1988) report that S. ornatus offspring are dependent on food provisioning and probably do not stray more than 200 m from the nest for nearly a year after hatching. Thus, C. brunneus territories close to an active eagle nest may experience elevated predation pressure for many months. At present, there is circumstantial evidence that S. ornatus may periodically change nest sites within its hunting range. Lyon and Kuhnigk (1985) report that, over the course of two years, an active S. ornatus nest in Guatemala was abandoned, and a new nest was established approximately 5 km away. Although it is unknown whether the same pair of eagles constructed both nests, it is plausible under the assumption that S. ornatus ranges are comparable in size to the minimum estimated by Robinson (1994) for larger eagles at the study site. If true, then eagle predation on local populations of C. brunneus may vary considerably over several years, depending on where in the large hunting range an S. ornatus pair constructs a nest, and how often the nest site is moved. Thus, it seems likely that ecological factors such as food availability and predation pressure vary considerably from year to year for C. brunneus at the study site. This will affect the probability that a disperser will have to attempt to establish a new territory in an unfamiliar area, which will in turn determine the selective advantage of trichromatic color vision. It appears that temporally variable selection for trichromacy is plausibly operant in C. brunneus.

109 99 Temporally Variable Selection for Trichromacy In summary, I have argued that, in the C. brunneus study population, 1) trichromats are likely to have an advantage over dichromats in the long distance detection and recognition of patches containing unfamiliar YOR food items, and this can lead to an increased food patch encounter rate when foraging in unfamiliar areas; 2) as a result of a heavy reliance on spatial memory and non color cues, di and trichromatic residents of established territories are likely to have comparable food patch encounter rates, and, by extension, comparable fitness; and 3) dispersal can result in a female being forced to attempt to establish a new territory in an unfamiliar area, a situation where trichromacy is likely to be beneficial, and the frequency of which is likely to vary from year to year depending on fluctuations in food availability and predation pressure. This hypothesis of temporally variable selection for trichromacy has several important implications. First, under the hypothesis, trichromacy provides an average selective advantage over the long term, and the polymorphism at the M/L opsin gene locus responsible for polymorphic trichromacy in a C. brunneus population can be maintained. This is because, across many C. brunneus generations, as long as more dispersing individuals are produced than can fill available vacancies or evict established residents, more trichromatic than dichromatic dispersers will, on average, successfully become territory residents (either in new or existing territories). This will result in frequency dependent selection on opsin alleles. For instance, if there is random mating, an adult carrying a rare allele will have a higher probability of producing a trichromatic daughter than will an adult carrying a common allele. On average (over generations), dispersing trichromatic daughters will have higher fitness than dispersing dichromatic daughters. Thus, any rare allele will tend to increase in frequency until (ideally) reaching equilibrium in the population. In this way, polymorphic trichromacy in the C. brunneus population can be maintained.

110 100 A second implication of the hypothesis of temporally variable selection is that, at any given point in time, dichromatic female territory residents may be found at relatively high frequency in a C. brunneus population (e.g., Chapter 2). These dichromatic females will have fitness equal, on average, to that of trichromatic female territory residents. This is a consequence of the assumptions that: 1) dichromatic females are just as likely as trichromatic females to fill vacancies or replace evictable adults in existing territories; 2) once having successfully immigrated into an existing territory, dichromatic females can rely on the spatial knowledge of their previously resident pair mates to achieve YOR food patch encounter rates comparable to those of trichromatic females; 3) depending on ecological factors such as food availability and predation pressure, in some years (e.g., high adult predation and/or low food availability) few if any dispersers will attempt to establish new territories, and dichromatic dispersers will be just as successful as trichromatic dispersers in becoming reproductivelysuccessful territory residents; and 4) under conditions of random mating, di and trichromatic females are equally likely to produce trichromatic daughters (an interesting consequence of the X linked polymorphism). Finally, this hypothesis suggests a change in the focus of field researchers looking for a selective advantage of trichromacy in C. brunneus populations. Any advantage of trichromacy over dichromacy in long distance food patch detection and recognition may only be discernable in dispersing females who are attempting to establish new territories in unfamiliar areas. The hypothesis predicts that a comparison of the day to day foraging activities of di and trichromatic female territory residents would yield little difference, on average (e.g., Chapter 3). Instead, a trichromat fitness advantage might only become evident through long term population studies, tracking the fates of dispersing di and trichromatic females (see below).

111 101 The hypothesis presented above has been developed by incorporating specific features of the social system and ecology of a single C. brunneus population in southeastern Peru. However, polymorphic trichromacy characterizes many primates with very different social systems and ecology. Consequently, it is useful to examine the applicability of the hypothesis to other primates with this visual configuration. Applicability to Other Primates Overview Temporally variable selection can plausibly maintain polymorphic trichromacy only in primate species: 1) characterized by extensive use of spatial memory and non color cues when foraging for YOR food items; 2) that exhibit both locational and social dispersal (Isbell and Van Vuren 1996) by females; and 3) that live in groups with low limits on the allowable number of co resident females. Without these three characteristics, it seems difficult to imagine how diand trichromatic females could have comparable foraging success in familiar areas, divergent foraging success in unfamiliar areas, and routinely find themselves in situations where they are forced to forage for YOR items in unfamiliar areas without recourse to conspecifics who may be trichromatic and/or familiar with the area. For instance, if spatial memory is relied upon extensively when foraging for YOR items in familiar areas, then a trichromatic foraging advantage may be realized only in circumstances where spatial memory cannot be used (e.g., unfamiliar areas). However, if YOR food items, such as ephemeral fruit patches or mobile insects, must be routinely discovered (i.e., their locations are not remembered), even when foraging in familiar areas, then a trichromatic advantage may be realized in day to day foraging and temporally variable selection may not be applicable. Similarly, for primate species relying extensively on spatial memory when foraging in familiar areas, a trichromatic advantage may

112 102 only be realized if females must occasionally forage in unfamiliar areas without recourse to knowledgeable or potentially trichromatic conspecifics. This may not occur if, for instance, females are philopatric and spend their whole lives in a familiar area or in the company of other females (some of whom may be trichromats). Thus it seems likely that temporally variable selection can potentially maintain polymorphic trichromacy only in species in which at least some females engage in locational and social dispersal. Finally, females who are dispersing both locationally and socially must attempt to establish new territories only when they cannot gain entry into an established group with breeding opportunities. This is especially likely to occur if there is a low, ecologically imposed limit on the number of adult females any given group may contain (e.g., due to intra group feeding competition). For primate species in which the numerical limits on female group membership are relatively high (e.g., Cebus and Saimiri, below), dispersing females may be able to immigrate directly into any neighboring group at any time without having to replace or displace a resident female. Thus, such females may never have to forage in the absence of female conspecifics (some of whom may be trichromatic), apart from the (probably short) time needed to transfer from one group to another. Temporally variable selection is unlikely to be the mechanism maintaining polymorphic trichromacy in such species, as the context in which a trichromatic advantage is most important (under the hypothesis), namely, foraging in unfamiliar areas in the absence of knowledgeable or trichromatic conspecifics, will be very rare. Interestingly, Isbell (2004) has shown that the characteristics of extensive use of spatial memory, female natal dispersal, and the number of female group residents may be causally linked. Primates relying heavily on spatial memory to locate most food sources are expected to defend territories from intruding conspecifics in order to ensure that food is reliably encountered in familiar feeding areas (see also Terborgh 1983). Such species are expected to be

113 103 energy limited and characterized by energetically efficient goal directed travel between familiar and temporally reliable food patches, rather than energetically less efficient wandering in which unfamiliar (potentially unreliable) food patches are routinely discovered. Goal directed travel based on spatial memory, as opposed to wandering and discovery, may not be conducive to foraging in unfamiliar areas. Consequently, females employing a goal directed foraging strategy may not be able to expand the boundaries of the current home range (or territory) in order to accommodate the foraging needs of adult offspring. As a result, such species are predicted to exhibit female (or bisexual) natal dispersal and relatively small social groups (Isbell 2004). Isbell (2004) categorized primates with a (presumed) heavy reliance on spatial memory when foraging and obligate female (or bisexual) natal dispersal as stingy species, which usually live in groups with one (or sometimes two) reproducing female(s). Species whose foraging behavior is primarily based on spatial memory but also includes a substantial wandering component are labeled incomplete suppressors, and are characterized by small groups of reproducing females, some of whom may be kin. Species relying extensively on wandering behavior while foraging tend not be territorial and are classified as generous, facilitator, or indifferent, depending (respectively) on whether females forage alone but with ranges overlapping those of reproducing female kin, together with (usually many) reproducing female kin, or together with (usually many) reproducing female non kin. For platyrrhine and strepsirrhine primates with polymorphic trichromacy, temporally variable selection for trichromacy appears most applicable to those species conforming to Isbell (2004) s stingy classification and least applicable to species characterized by habitual wandering behavior while foraging.

114 104 Likely Candidates for the Hypothesis Several primates in Isbell (2004) s stingy and incomplete suppressor categories appear to be good candidates for temporally variable selection. For instance, like Callicebus, most callitrichines (tamarins and marmosets) have polymorphic trichromacy (Jacobs and Deegan 2003; Surridge et al. 2005b; Tovee et al. 1992), probably rely heavily on spatial memory when foraging within their vigorously defended minimally overlapping territories (Bicca Marques and Garber 2004; Garber 1989; Smith et al. 2003a; Terborgh 1983), and are characterized by dispersing females, some of whom appear to establish new territories in unfamiliar regions when no breeding opportunities can be found in existing territories (e.g., Goldizen et al. 1996). Similarly, Milne Edwards sifakas, Propithecus edwardsi (whose congener P. verreauxi has been verified as a probable polymorphic trichromat, Jacobs et al. 2002; Tan and Li 1999), may be a good strepsirrhine candidate for temporally variable selection. As with Callicebus and most callitrichines, P. edwardsi live in non overlapping territories containing one to two reproducing females, and female immigration into groups is usually associated with the eviction of a resident adult female (Morelli et al. 2009). Morelli et al. (2009) suggest that, in years when the number of female dispersers exceeds the number reproductive vacancies or evictable females in existing territories, P. edwardsi females may delay dispersal (and reproduction) and remain in their natal territory. However, it also seems plausible that, in such a situation, some females may attempt to establish new territories in regions unfamiliar to them, a scenario in which trichromacy might provide a foraging advantage for YOR food patches. To build a stronger case for the role of temporally variable selection in the maintenance of polymorphic trichromacy in callitrichines and P. edwardsi, more information is needed to establish that, when foraging for YOR items in an unfamiliar area, non color cues are less useful than YOR cues (or are not at all useful) in detecting food patches from a distance and/or recognizing such patches as potentially

115 105 containing food. In any case, at present, it seems likely that the hypothesis of temporally variable selection for trichromacy may have wider applicability than the C. brunneus population for which it was originally developed here. Unlikely Candidates for the Hypothesis For many platyrrhines and strepsirrhines with polymorphic trichromacy, there is still insufficient information about the use of spatial memory and the circumstances surrounding dispersal and new home range establishment to judge the applicability of the hypothesis of temporally variable selection for trichromacy (e.g., Ateles and Lagothrix: Asensio et al. 2008; Di Fiore et al. 2009; Symington 1988; Varecia: Vasey 2007). However, two well studied platyrrhine genera with polymorphic trichromacy, Saimiri (Cropp et al. 2002; Jacobs et al. 1993b; Mollon et al. 1984) and Cebus (Jacobs and Deegan 2003; Saito et al. 2005a), classified in Isbell (2004) s facilitator and indifferent categories, seem unlikely candidates for the hypothesis. In both of these genera, nearly all dispersing females (which are rare in Cebus and some Saimiri species) seem able to successfully immigrate directly into neighboring groups (Boinski et al. 2005; Jack and Fedigan 2009). This suggests that females will almost never have to forage in an unknown area in the absence of other females who are familiar with the area and/or are potentially trichromatic. Similarly, for Verreaux s sifaka (Propithecus verreauxi), a strepsirrhine with polymorphic trichromacy (Jacobs et al. 2002; Tan and Li 1999) characterized by Isbell (2004) as an incomplete suppressor, a dispersing female may occasionally establish a new home range with a dispersing male (Richard et al. 1993). However, Richard et al. (1993) found that all such new home ranges completely overlapped the home ranges of neighboring groups, and, since males tended to disperse in the vicinity of their natal range, at least one of the founding pair members was probably familiar with the area. Thus, it appears unlikely that temporally variable

116 106 selection is currently maintaining polymorphic trichromacy in Cebus, Saimiri, and P. verreauxi populations. Conclusions In summary, the hypothesis of temporally variable selection for the maintenance of polymorphic trichromacy appears plausible for several platyrrhine and strepsirrhine primates, in addition to Callicebus for which it was developed. Temporally variable selection may be especially important in primates whose extensive use of spatial memory may minimize a trichromatic advantage in food patch encounter rate during day to day foraging behavior, and for whom female dispersal may constitute one of the most important contexts for the action of selection on color vision. Because the strength of selection on dispersing females may vary across generations depending on the dynamics of the local population, long term studies of such primates may be necessary in order to document a selective advantage of dispersing trichromats over dichromats. Temporally variable selection may explain why field studies of the routine foraging behavior of Callicebus and callitrichine individuals in established territories (Dominy et al. 2003a; Smith et al. 2003a, Chapter 3) have failed to detect trichromatic foraging advantages analogous to those demonstrated for these animals in captivity (Caine and Mundy 2000; Smith et al. 2003b). The temporally variable nature of selection may also explain why reproductively successful social groups composed entirely of dichromats have been reported in some natural populations of Callicebus and Saguinus (Smith et al. 2003a, Chapter 2). For primates in Isbell (2004) s stingy and incomplete suppressor categories, the strongest evidence in support of the hypothesis that temporally variable selection is currently maintaining polymorphic trichromacy can only come through behavioral observations of dispersing females. Following dispersing primates is notoriously difficult in the wild, even when

117 107 technology such as radio telemetry is employed (e.g., Di Fiore et al. 2009). However, Bossuyt has demonstrated that it is feasible for Callicebus (e.g., Bossuyt 2002), albeit with a considerable investment of time and effort. The hypothesis of temporally variable selection predicts that, on average, dispersing trichromatic females will have higher food patch encounter rates than will dichromatic females when initially establishing new territories in unfamiliar regions, and that both di and trichromatic dispersers will have equivalent foraging success, on average, once resident in established groups. Indirect evidence in support of the hypothesis might come from annual population wide visual and genetic censuses of appropriate primate species documenting the establishment of new groups in previously unoccupied areas. If temporally variable selection for trichromacy is maintaining polymorphic trichromacy in the population, one would expect that, on average, more trichromats than dichromats successfully establish new territories, while di and trichromatic dispersers have approximately equal (on average) success in immigrating into established territories. Such censuses of primate populations over a series of years might also address the prediction that the number of dispersing (or dispersal aged) individuals in a population relative to the number of potential vacancies in established territories varies from year to year. Additional indirect evidence in support of the hypothesis of temporally variable selection might come from field or laboratory experiments demonstrating that di and trichromatic females have equivalent encounter rates, on average, for YOR patches whose locations are known and/or which can be detected and recognized at a distance using non color cues, while trichromats have higher encounter rates for patches of unfamiliar YOR species and/or for YOR patches whose locations are not known and which cannot be easily detected at a distance using non color cues. Thus, data from both long term population wide studies and laboratory experiments can surely be used to address the role of temporally variable

118 108 selection in the maintenance of polymorphic trichromacy in extant primates such as Callicebus and the callitrichines. In contrast, for primates such as Cebus and Saimiri, where the hypothesis of temporally variable selection appears less applicable, a trichromatic advantage may be discernible in dayto day foraging behavior within established groups. Important field studies of Cebus have thus far focused on within patch foraging and have not detected an advantage of trichromacy over dichromacy within YOR patches (Melin et al. 2008; Vogel et al. 2007). Here and elsewhere (Chapter 3) I suggest that trichromacy may be most important in the long distance detection and recognition of YOR food patches, and that such an advantage may be especially important during times of energy stress such as heavy lactation (Chapter 3) and periods of food scarcity (Dominy et al. 2003a; Dominy and Lucas 2001; Dominy et al. 2003b; Jacobs 1997; Lucas et al. 2003). Future studies are needed to clarify our understanding of the utility of trichromacy in such primate species. It is difficult to know if the selective advantages presumed to be currently maintaining polymorphic and full trichromacy in extant primate populations are similar to those which were important in its initial evolution in ancestral primate populations (e.g., Allen 1879; Dominy et al. 2003b; Fernandez and Morris 2007). Here I have proposed that such selective advantages are likely to be intimately linked to the details of the ecology and social system of particular primate species. Because an animal s morphology can only be used to define its ecological potential (Janson and Boinski 1992), fossil remains of ancestral primates are likely to reveal only the broad outlines of how these animals interacted behaviorally with their environments and with each other. This may be insufficient to determine the relevance of the hypothesis of temporally variable selection (or any other hypotheses) for the evolution of trichromacy in various primate lineages. However, regardless of the difficulty of reconstructing the evolutionary history of

119 109 trichromatic color vision in primates, it certainly would benefit from an understanding of the current utility of trichromacy in extant populations.

120 110 CHAPTER 5 General Conclusions In Chapter 2 I showed that at least three distinct opsin gene alleles occur in a wild population of the platyrrhine monkey Callicebus brunneus, potentially resulting in a proportion of females with trichromatic color vision, while other females and all males have dichromatic vision. Genetic characterization of these alleles revealed that they are similar in sequence to the three most common alleles of other platyrrhines such as Cebus and Saimiri, where they code for photopigments with peak light absorbance at approximately 530 nm, 550 nm, and 562 nm. I found no evidence of the two additional opsin gene alleles reported to occur at low frequencies by Jacobs and Deegan (2005) in a captive colony of Callicebus, although I cannot exclude the possibility that such alleles occur in wild populations at frequencies comparable to those recorded in captivity. In a sample of five socially monogamous C. brunneus groups, I was unable to detect evidence of strong selection for trichromacy over dichromacy, either through disassortative mating behavior (presumed through group composition) to increase the probability of producing heterozygous trichromatic daughters, or substantially greater shortterm reproductive success for C. brunneus pairs containing trichromatic adult females compared to pairs with dichromatic females. Behavioral studies of color vision are necessary to definitively confirm the presence of trichromacy in wild populations of C. brunneus (or any other animal: Neumeyer 1991), although such experiments are usually not feasible in the wild. However, the results of this study, together with the evidence of Callicebus trichromacy provided by Jacobs and Deegan (2005) s physiological study of retinal light absorbance, strongly suggest that heterozygous female Callicebus brunneus in the study population are trichromatic. Larger samples of wild C. brunneus populations are necessary to draw conclusions about the strength

121 111 of selection for trichromacy. I suggest that it may be useful to consider the role of extra pair copulation and habitat selection when looking for fitness advantages of free ranging trichromatic C. brunneus. In Chapter 3 I showed that C. brunneus spend a large proportion of daily foraging time in food patches containing yellow, orange, or red (YOR) items. Such patches are potentially better detected from a distance by trichromats than by dichromats against a dappled forest leaf background (Mollon 1989; Osorio et al. 2004; Regan et al. 2001). However, I suggest that dichromatic C. brunneus can potentially make use of spatial memory and non color cues to locate the vast majority of YOR patches at rates comparable to those of trichromats. For YOR food patches whose locations were less likely to be remembered by the study animals, I was unable to detect a difference in the patch encounter rates of dichromatic and trichromatic females after controlling for territory level effects, season, and reproductive status. On average, C. brunneus females had significantly higher encounter rates for such patches than did their dichromatic male pair mates, and there was an interesting (but non significant) trend for lactational status to differentially affect the patch encounter rates of di and trichromatic females. These results do not provide strong support for the hypothesized selective advantage of trichromacy over dichromacy in the context of long distance patch encounters while foraging. However, they do suggest that it may prove profitable in future field studies to look for just such an advantage, namely, one enjoyed by heavily lactating females foraging alone for small, ephemeral YOR food patches. Additionally, this study demonstrates the limitations of shortterm behavioral field studies employing small sample sizes in the detection of a selective advantage which may become pronounced in populations only at large temporal scales. In Chapter 4 I provided evidence that there is considerable dietary diversity among C. brunneus groups in the study population, even among those with adjacent territories. Using this

122 112 observation I developed a new hypothesis to explain the results of this study, namely, that polymorphic trichromacy may be maintained in C. brunneus populations by temporally variable selection, such that a trichromatic foraging advantage is only realized when dispersing females attempt to establish new territories in unfamiliar areas, a situation whose probability may vary over the course of several years according to ecological conditions such as fruit abundance and predation pressure. The hypothesis of temporally variable selection accounts for the observation that the three opsin gene alleles found in this study appear to occur at relatively high frequency in the C. brunneus population (i.e., suggesting frequency dependent selection on alleles and a trichromatic fitness advantage), as well as the observations (counter to the original hypothesis of a trichromatic fitness advantage) that reproductively successful dichromatic females appear to be fairly common in this population and that dichromatic and trichromatic females have statistically indistinguishable YOR patch encounter rates. In addition to Callicebus, temporally variable selection may also be applicable to other primates with polymorphic trichromacy, including callitrichines and the lemur Propithecus edwardsi. Evidence in support of this hypothesis could come from future field studies examining long term population dynamics, as well as the foraging behavior of dispersing females in primate species with polymorphic trichromacy, small social groups, and small stable territories. In conclusion, this investigation has generated several new hypotheses for how polymorphic trichromacy may currently be maintained in populations of C. brunneus and other primates with similar socio ecology. I suggest that future studies examining the selective advantage of trichromacy in wild populations pay particular attention to mating patterns, habitat selection, spatial memory of food resources, reproductive status, and dispersal behavior. Some or all of these characteristics may play important roles in selection for trichromacy and, most importantly, their consideration may guide the development of more sophisticated

123 113 hypotheses and innovative ways to test them. Only when the nature of selection for color vision in extant primate populations becomes clearer will we have a reasonable starting point for the development of hypotheses explaining how trichromatic color vision evolved in our primate ancestors.

124 114 TABLES Table 2.1. Correspondence between M/L opsin peak light absorbances and amino acid changes at sites thought to influence the absorbance of opsins in humans and representative platyrrhine primates. Translated Callicebus haplotypes described in this study are shown in bold. Approximate Wavelength of Peak Light Absorbance a Exon 3 Absorbance Tuning Site and Peak Absorbance Shift b Exon 4 Exon 5 Genus 530 nm 535 nm 543 nm 550 nm 556 nm 562 nm A S 6 nm Y H 28 nm Homo X S H Y I I A Y T N A Y Alouatta X S H c Y F L G Y T N A Y Cebus X S H Y F L G Y T N A Y Ateles X S H D F L G Y T N A Y Saguinus X S H Y F L G Y T N A Y Callicebus X S H Y F L G Y T N A Y? F I 2 nm I T 2 nm S G 1 nm F Y 9 nm A T 15 nm? S A 27 nm Y F 2 nm Saguinus X A H Y F L S Y T N A Y Cebus X A H Y I L S F T N A Y Ateles X S d H D I L S F T K A Y Callicebus X A H Y I F S F T N A Y (continued)

125 115 Table 2.1. (continued) Approximate Wavelength of Peak Light Absorbance a Exon 3 Absorbance Tuning Site and Peak Absorbance Shift b Exon 4 Exon 5 Genus 530 nm 535 nm 543 nm 550 nm 556 nm 562 nm A S 6 nm Y H 28 nm Saguinus X A H Y I L S Y A N A Y Callicebus X? F I 2 nm I T 2 nm S G 1 nm F Y 9 nm A T 15 nm? S A 27 nm Y F 2 nm Cebus X A H Y I L S F A N A Y Callicebus X A H Y I F S F A N A Y Homo X A H Y I T S F A N A F Alouatta e X A H c Y I L S F A N A Y Callicebus X a In vivo absorbance measures for Homo: Stockman and Sharpe (2000); Alouatta (howler monkeys): Jacobs et al. (1996a); Cebus (capuchins): Jacobs and Deegan (2003); Ateles (spider monkeys): Jacobs and Deegan (2001); Saguinus (tamarins): Jacobs et al. (1987); Callicebus (titi monkeys): Jacobs and Deegan (2005). Note that in vivo measurements tend to be slightly longer than in vitro measurements of reconstituted photopigment (Jacobs 2007).

126 116 b Translated haplotypes for Homo: Nathans et al. (1986b) and Bunce, unpublished; Alouatta: Hunt et al. (1998); Cebus and Ateles: Hiramatsu et al. (2005) (GenBank AB AB193784, AB193790, AB193796); Saguinus fuscicollis: Surridge and Mundy (2002) (GenBank AY AY142410); Callicebus brunneus: this study, haplotypes tentatively matched with absorbances. Dashes indicate photopigments found by Jacobs and Deegan (2005) whose sequences have yet to be determined. The Ateles 535 nm opsin reported by Riba Hernandez et al. (2004) is excluded because its precise absorbance and haplotype have yet to be published. Amino acid sites in exons 3, 4, and 5 determined to be important to photopigment tuning in primates (and mammals): bold and underlined, Neitz et al. (1991); bold, Yokoyama and Radlwimmer (2001); underlined, Shyue et al. (1998); italics: Kawamura et al. (in press); plain, Asenjo et al. (1994). Approximate shifts in photopigment peak absorbance corresponding to amino acid changes are taken from the above references and are shown directly below the site number. Absorbance shifts shown here are approximately additive and reversible, except for sites 230 and 309, where reversing the direction of the amino acid change has a negligible effect on pigment absorbance (Asenjo et al. 1994).

127 117 c Not reported by Hunt et al. (1998), but presumed, as site 197 appears to be monomorphic in platyrrhines d It remains to be determined if this Ateles haplotype found by Hiramatsu et al. (2005) corresponds to the in vivo absorbance measurements of Jacobs and Deegan (2001). Here it is provisionally placed in the 550 nm peak absorbance class. e It is currently unclear which absorbance tuning sites serve to differentiate the presumed Alouatta 530 nm allele from the Cebus 535 nm allele.

128 118 Table 2.2. Variable amino acid sites found in three Callicebus brunneus M/L opsin alleles a Exon 3 Exon 4 Exon 5 Allele L V S M F L G V V Y V T 550 V I A V I F S M M F L T 535 V V A M I F S M V F V A Absorbance tuning sites used to discriminate among the three alleles are shown in bold. a Allele names refer to presumed approximate peak light absorbance (in nanometers) of the resulting photopigment (see Table 2.1).

129 119 Table 2.3. Color vision genotyping results for 14 wild Callicebus brunneus individuals from five socially monogamous groups Group Individual Sample Type a Verified Sex b Exons Sequenced c Alleles d Inferred Vision 1 Adult Female feces f 3 (2,2), 4 (2,1), 5 (9,5) trichromacy 1 Adult Male feces 1, feces 2 m 3 (3,3), 4 (2,1), 5 (8,5) 535 dichromacy 1 Subadult Female feces, saliva F 5 (4,2) trichromacy 2 Adult Female feces, blood F 3 (4,4), 5 (9,7) 550 dichromacy 2 Adult Male feces 1, feces 2 m 4 (1,1), 5 (4,2) 550 dichromacy 3 Adult Female feces, blood F 3 (2,2), 4 (2,1), 5 (5,3) trichromacy 3 Adult Male feces 1, feces 2 3 (1,1), 4 (1,1), 5 (4,3) 562 dichromacy 3 Subadult Male blood M 5 (2,1) 562 dichromacy 3 Juvenile Female feces f 5 (3,2) 562 dichromacy S Adult Female feces f 3 (2,2), 4 (2,1), 5 (8,5) 550 dichromacy S Adult Male feces 1, feces 2 m 3 (1,1), 4 (3,2), 5 (9,5) 535 dichromacy W Adult Female feces, saliva F 3 (2,2), 4 (2,1), 5 (5,4) trichromacy W Adult Male 1 saliva M 5 (2,1) 550 dichromacy W Adult Male 2 e feces 4 (1,1), 5 (4,2) 550 dichromacy

130 120 a Blood and saliva samples were collected from captured animals on FTA DNA preservation cards. When two fecal samples were collected independently from the same animal, they are labeled with numbers. b Upper case letters: blood and saliva samples directly collected from captured animals (M = male, F = female); lower case letters: fecal samples subjected to the genetic sex determination procedure of Di Fiore (2005) (m = male, f = female); no letter: fecal samples collected from individuals whose sex was inferred from behavior. c Names of exons sequenced (3, 4, or 5) are in bold. In parentheses after each sequenced exon are: 1) the number of sequenced fragments (forward and reverse strands counted separately) spanning sites 180 (exon 3), sites 277 and 285 (exon 5), or any part of exon 4 ; and 2) the number of independent PCR amplifications yielding the sequenced fragments. d Allele names correspond to approximate peak opsin light absorbance in nanometers, see Table 2.1. e The adult male in Group W disappeared during the study and was replaced by another, see Table 2.4.

131 121 Table 2.4. Important events in five Callicebus brunneus groups observed between August 2005 and October 2006 Group Adult Female Vision Offspring Existing August 2005 Births Infant/Juvenile Disappearance s Presumed Subadult Dispersals Adult Male Replacement s 1 trichromacy trichromacy W trichromacy a 1 a 2 dichromacy S dichromacy a It was unclear whether a subadult male replaced the resident adult male, or whether the subadult dispersed after the disappearance of the resident adult male and the adult female subsequently paired with a new adult male.

132 122 Table 3.1. Genotypes and presumed phenotypes of adult Callicebus brunneus individuals Group Sex Allele a Vision 1 Female 535/550 Trichromacy 1 Male 535 Dichromacy 2 Female 550 Dichromacy 2 Male 550 Dichromacy 3 Female 550/562 Trichromacy 3 Male b 562 Dichromacy S Female 550 Dichromacy S Male 535 Dichromacy W Female 550/562 Trichromacy W Male c 550 Dichromacy a Refers to the presumed approximate wavelength of peak light absorbance, in nanometers, of the resulting photopigment b The adult male was replaced over the course of the study. The first adult male's genotype is unknown. c The adult male was replaced over the course of the study. Both males had the same genotype.

133 123 Table 3.2. Mixed effects Poisson regression for YOR food patches discovered by Callicebus brunneus adults Fixed Effect Predictor Coefficient (SE) Intercept 1.4 (0.48)** Sex = 1.23 (0.4)** Vision = Trichromacy 0.83 (0.55) IDP = Independent Infant 0.83 (0.4)* Trichromacy and Independent Infant 1.36 (0.49)** Random Effect Variance 0.18 Log Likelihood 93.3 Deviance N (animal days) 214 Pearson statistic *p value < 0.05 **p value < 0.01

134 124 Table 4.1. Abundance based Jaccard similarity estimator of Chao et al. (2005; 2006) calculated for pairs of Callicebus brunneus groups using observed yellow, orange, and red dietary items (above diagonal) and all observed dietary items (below diagonal). Bootstrap standard errors are shown in parentheses. The adult females of Groups 2 and S were dichromats (italicized). Those of 1, 3, and W were trichromats (Chapter 2). Group S W (0.14) 0.31 (0.14) 0.28 (0.16) 0.41 (0.19) (0.30) 0.29 (0.16) 0.48 (0.18) 0.38 (0.15) (0.28) 0.55 (0.32) 0.28 (0.17) 0.43 (0.17) S 0.55 (0.30) 0.67 (0.31) 0.55 (0.33) 0.50 (0.20) W 0.64 (0.32) 0.63 (0.33) 0.61 (0.33) 0.68 (0.33)

135 125 FIGURES Cocha Cashu 1 3 S W 2 N 200 m Río Manu (left bank) Figure 2.1. Territories of five Callicebus brunneus groups (1: circles, 2: triangles, 3: crosses, S: hatches, and W: diamonds) in the vicinity of the Estación Biológica de Cocha Cashu in Manu National Park, Madre de Dios, southeastern Perú. Points represent trees in which the animals were observed to feed or rest (often on multiple occasions) during opportunistic follows while habituating the them (July December 2005 for most groups) and during systematic behavioral follows (January October 2006). All data in this figure are based on GPS coordinates which, in

136 126 most cases, were accurate to a 20 m radius, and should thus be interpreted with caution. The scale in the figure is approximate. Coordinates of the lake and the river were taken in the dry season when water levels were lowest. Territory perimeters were drawn by hand connecting peripheral feeding and resting trees and should be viewed as approximations of the areas over which groups ranged. Approximate territory areas (in hectares) are: 11.8 (Group 1), 6.0 (Group 2), 9.5 (Group 3), 9.0 (Group S), and 5.4 (Group W), with an overall mean of 8.3 ha.

137 127 1 W?? 3??? 2? S???? Figure 2.2. Distribution of opsin haplotypes among five Callicebus brunneus groups between August 2005 and October 2006 (graphic following Surridge et al. 2005a). Black: 535 nm allele; stripes: 550 nm allele; white: 562 nm allele. Large symbols: adult pair; small symbols: juveniles and subadults; question marks: infants, juveniles, and subadults who were not genotyped. Groups whose territories shared a border are shown with overlapping circles. An un genotyped adult male in Group 3 disappeared in March 2006 after presumably fathering the two genotyped offspring and was replaced by the adult male represented in this figure. The adult male in Group W was replaced in March 2006 by a male with the same genotype. During the majority of the study period group sizes ranged between two and three individuals (not counting dependent infants), so several infants, juveniles, and subadults shown in this figure were not present at the same time.

138 128 Figure 3.1. Stylogyne cauliflora (Myrsinaceae) branch with ripe (black) and immature (red) fruit. Scale is in intervals of 1 cm. 6 September 2006.

139 129 Figure 3.2. Huertea glandulosa (Staphyleaceae) fruit (black). Note the yellow peduncles. Scale is in intervals of 1 cm. 25 April 2006.

140 130 Figure 3.3. Monstera obliqua (Araceae) fruit in situ. Note the variation in color from pale yellow (left) to pale orange (right). August 2006.

141 131 A B C Figure 3.4. A: Population wide mean daily proportions of Callicebus brunneus food patch encounters (black circles) and daily foraging time (grey circles) by food class. Data for this figure comes from 1430 patch encounters over 105 observation days. Two of the 107 total observation days were excluded because no foraging was observed. Bars in this figure represent 95% confidence intervals about the population wide mean daily proportions. Unknown patches were most likely small solitary cryptic insects snatched from leaf and branch surfaces, as they were eaten too quickly to be identified. B: Population wide mean daily proportions of patch encounters (black circles) and foraging time (grey circles) involving food items with no yellow, orange, or red (YOR) external characteristics compared to patches containing YOR food items. C: Population wide mean daily proportions of patch encounters (black circles) and foraging time (grey circles) involving memorable food patches that were potentially located by means of spatial memory, having been exploited simultaneously by both adults in the pair and/or visited

142 132 by one or more adults on multiple occasions, compared to discovered food patches for which spatial memory was probably not employed, having, in my estimation, been exploited only once during the study period by either the adult female or male alone.

143 133 Patch Encounters Per Hour Travel M Di F Tri F M Di F Tri F Dependent Independent Figure 3.5. Population average encounter rates for YOR patches discovered by adult male (M), dichromatic female (Di F), and trichromatic female (Tri F) Callicebus brunneus per hour of observed travel during periods of infant dependency (Dependent) and independency (Independent). Infant dependency is defined as the period from birth to the last observed day on which the adult male carried the infant. Population averages and 95% confidence intervals are calculated from a multilevel Poisson regression model (see text for details).

144 134 Figure 4.1. Callicebus brunneus from the Estación Biológica de Cocha Cashu, Manu National Park, Madre de Dios, southeastern Perú. Left: captured adult female of Group W (2 February 2006). Right: 12 month old juvenile female of Group 3 (22 August 2006). Male C. brunneus have similar coloration.

145 135 Figure 4.2. Ficus trigona (Moraceae) individual in a swamp at the Estación Biológica de Cocha Cashu, Manu National Park, Madre de Dios, southeastern Perú. 21 July 2006.

146 136 Figure 4.3. Ficus trigona (Moraceae) fruit at various stages of ripeness. Scale is in intervals of 1 cm. 19 May 2006.