OPTIMAL FORAGING on the ROOF of the WORLD: a FIELD STUDY of HIMALAYAN LANGURS a Dissertation Submitted to Kent State University

Home , Arunachal macaque, Brown greater galago, Capped langur, Colobinae, Cotoneaster frigidus, Douc

OPTIMAL FORAGING ON THE ROOF OF THE WORLD: A FIELD STUDY OF HIMALAYAN LANGURS

A dissertation submitted to Kent State University in partial fulfillment of the requirements for the degree of Doctor of Philosophy

Kenneth A. Sayers

May 2008

Dissertation written by Kenneth A. Sayers B.A., Anderson University, 1996 M.A., Kent State University, 1999 Ph.D., Kent State University, 2008

Approved by

______, Dr. Marilyn A. Norconk Chair, Doctoral Dissertation Committee

______, Dr. C. Owen Lovejoy Member, Doctoral Dissertation Committee

______, Dr. Richard S. Meindl Member, Doctoral Dissertation Committee

______, Dr. Charles R. Menzel Member, Doctoral Dissertation Committee

Accepted by

______, Dr. Robert V. Dorman Director, School of Biomedical Sciences

______, Dr. John R. D. Stalvey Dean, College of Arts and Sciences

TABLE OF CONTENTS

LIST OF FIGURES ...... vi LIST OF TABLES ...... viii ACKNOWLEDGEMENTS ...... x

Chapter I. PRIMATES AT THE EXTREMES ...... 1

Introduction: Primates in marginal habitats ...... 1 Prosimii ...... 2 New World monkeys ...... 3 Cercopithecinae...... 3 Colobinae and apes ...... 5 The gray langur: A brief history of a very adaptable colobine ...... 6 Himalayan gray langurs ...... 9 Goals of the project ...... 12

II. STUDY SITE AND SUBJECTS ...... 13

Langtang National Park, Nepal ...... 13 Vegetation types of main study area ...... 15 Climate of main study area ...... 17 Mammalian fauna of main study area ...... 17 Study subjects ...... 19 Taxonomic identification of study subjects ...... 21

III. DIET, ACTIVITY PATTERNS, AND RESOURCES ...... 24

INTRODUCTION ...... 24 METHODS ...... 25 Activity and feeding data ...... 25 Classification of dietary items ...... 27 Phenology ...... 28 Data analysis ...... 30 RESULTS ...... 31 Activity patterns ...... 31 Plant production ...... 32

iii Diet ...... 35 Seasonal diet ...... 38 Food selection ...... 40 Daily path lengths ...... 43 DISCUSSION ...... 43

IV. EXTRACTIVE FORAGING AND INTELLIGENCE ...... 51

INTRODUCTION ...... 51 METHODS ...... 56 Data collection ...... 56 Data analysis ...... 58 RESULTS ...... 59 Classes of extractive foraging in Himalayan langurs ...... 59 Correlations: food availability and extractive foraging ...... 66 Extractive foraging in other colobines ...... 68 DISCUSSION ...... 68

V. OPTIMAL FORAGING: CLASSICAL PREY MODEL ...... 77

INTRODUCTION ...... 77 METHODS ...... 83 Behavioral observations ...... 83 Nutritional analysis and currencies for the model ...... 84 The model ...... 87 Model predictions and statistics ...... 90 RESULTS ...... 93 Threshold for dropping items and the zero-one rule ...... 93 Preference for profitable food types ...... 112 Increased selectivity at higher encounter rates ...... 114 Selectivity independent from abundance of low-ranking food .....116 Comparison of currencies ...... 118 Conformance of model assumptions and predictions ...... 120 DISCUSSION ...... 120

VI. OPTIMAL FORAGING: SOCIAL PREY MODEL ...... 131

INTRODUCTION ...... 131 METHODS ...... 135 Behavioral observations ...... 135 Nutritional analysis and currency ...... 136 The model ...... 138 Model predictions and statistics ...... 138

iv RESULTS ...... 143 Patch use strategy, residence time and travel time ...... 143 Social foraging and the expanding specialist strategy ...... 145 Number of foragers within patch and gain rate at switch ...... 148 DISCUSSION ...... 150

VII. CONCLUDING REMARKS ...... 154

Introduction ...... 154 Diet, activity patterns, and resources ...... 155 Colobine cognition ...... 157 Optimal foraging theory ...... 160 Himalayan langurs, foraging theory, and human evolution ...... 162

REFERENCES ...... 166

LIST OF FIGURES

1. Topographical map of main study area ...... 14

2. Woody habitats of main study area ...... 16

3. Temperature and precipitation at Ghore Tabela, Nepal ...... 18

4. F troop near Langtang Monastery ...... 20

5. Variation in forearm coloration in F troop ...... 23

6. Phenology of broad-leaved vegetative structures at Langtang ...... 33

7. Phenology of reproductive plant parts at Langtang ...... 34

8. Seasonal daily paths for Himalayan langurs ...... 44

9. Extracted foods of Himalayan langurs ...... 61

10. Percent contribution of food types to extractive foraging ...... 62

11. Juvenile langur digging ...... 63

12. Adult female with underground storage organ ...... 64

13. MEO profitability threshold for juveniles ...... 106

14. MEH profitability threshold for juveniles ...... 106

15. CP profitability threshold for juveniles ...... 107

16. MEO profitability threshold for adult females ...... 107

17. MEH profitability threshold for adult females ...... 108

18. CP profitability threshold for adult females ...... 108

19. MEO profitability threshold for adult males ...... 109

vi 20. MEH profitability threshold for adult males ...... 109

21. CP profitability threshold for adult males ...... 110

22. MEO profitability threshold for a single adult male ...... 110

23. MEH profitability threshold for a single adult male ...... 111

24. CP profitability threshold for a single adult male ...... 111

25. Himalayan langur feeding on Cotoneaster frigidus ...... 141

26. Instantaneous rate of gain before switching to a less profitable food ...... 149

vii

LIST OF TABLES

1. Phenological sample ...... 29

2. Feeding records and food types ...... 36

3. Top ten food items in the Himalayan langur diet ...... 37

4. Seasonal diet ...... 39

5. Spearman correlations: plant part abundance and consumption ...... 41

6. Spearman correlations: non-seasonal food consumption and plant parts ...... 42

7. Stepwise multiple regressions: consumption, abundance, and daily path ...... 45

8. Semnopithecus entellus foraging behavior at long-term study sites ...... 46

9. Ultimate causation models of primate brain evolution ...... 52

10. Gibson’s (1986) classification of primate foragers ...... 55

11. Categories of extractive foraging for Langtang langurs ...... 60

12. Spearman correlations: extractive foraging and phenology scores ...... 67

13. Extractive foraging in colobine monkeys ...... 69

14. Assumptions of the classical prey model as modified for patch choice ...... 89

15. Variables: calculating profitability threshold for dropping items from diet ...94

16. Seasonal En/T and dietary contribution for a single adult male ...... 105

17. Spearman correlations: dietary contribution and profitability ...... 113

18. Spearman correlations: encounter rates and diet breadth ...... 115

19. Spearman correlations: encounter rates and dietary contribution ...... 117

20. Nutritional currencies compared to predictions of prey model ...... 119

viii 21. Seasonal comparison of assumptions met and model success ...... 121

22. Assumptions of the social prey model ...... 139

23. Patches where ≥ 2 food types were taken from a single patch ...... 144

24. Patch strategies and residence time for Cotoneaster frigidus leaf parts ...... 146

25. 2 × 2 table of two patch strategies and solitary versus social foragers ...... 147

ACKNOWLEDGMENTS

First off, I would like to thank L.S.B Leakey Foundation and Kent State

University School of Biomedical Sciences for their financial support. Ram Rimal and

Ranger Ming Mav Chhewang Tamang provided invaluable field assistance and

friendship, and the Langtang National Park staff provided welcome aid and

encouragement. Achyut Ahdikari waded through botanical minutiae, and Nina Jablonski

and Mukesh Chalise offered advice on Nepalese fieldwork. I also thank Daniel Taylor-

Ide and Robert Fleming, Jr. for their assistance in the selection of Langtang as a

Himalayan langur field site, Himalayan Glacier Trekking and Cemat Water Lab for

handling logistics, and Dindu Lama and family for everything else. This research was

conducted in conjunction with the Nepal Ministry of Forests and Soil Conservation and

Department of National Parks and Wildlife Conservation, and I thank them.

My family (Dad, Mom, Eric, Matt) provided support, and I will forever be

grateful, although I suspect they don’t really know exactly what I do. Since it is most

unlikely they will ever read this, that’s of little consequence. My fellow graduate

students have always been a joy to be around. These include Burt Rosenman, Phil Reno,

Maria Serrat, Maryann Raghanti, Tremie Gregory, Cynthia Thompson, and a host of

others. Nancy Lou Conklin-Brittain aided immensely in regards to the nutritional portion

of this project. The members of my committee, Owen Lovejoy, Rich Meindl, and Charlie

Menzel, have been eternally helpful and I have learned much from them. I would also

x like to thank Marilyn Norconk, my Dissertation Chair, for her unending help on this research and her exemplary example as a scientist.

xi CHAPTER I

PRIMATES AT THE EXTREMES

“I saw a troop of large monkeys gamboling in a wood of Abies brunoniana; this

surprised me, as I was not prepared to find so tropical an animal associated with a vegetation typical of a boreal climate.”

--- Sir Joseph Hooker, Himalayan Journals, 1855

Introduction: Primates in marginal habitats

When Sir Joseph Hooker, the well-known botanist and confidant to Charles

Darwin, came across monkeys living in coniferous forest at 9,000 feet in Sikkim, it is not

surprising that he was somewhat jarred (Hooker 1855). Nonhuman primates are

stereotypically confined to tropical or subtropical regions; this was known in the mid-

1800s and remains the conventional wisdom. Even today, maps depicting the

geographical distribution of extant primates often do not include the Himalayan area

traversed by Hooker over 150 years ago (Rowe 1996). Add to this that the Himalayan

langur monkey, the primate spotted by that famous scientist, is found not only in the

interior of Sikkim, but also at even higher altitudes and more northerly latitudes. It is not,

however, the only primate to inhabit such a harsh environment. There are others, and the

whole group remains little studied.

1 2

The scarcity of data on marginal-habitat species should likewise not be considered surprising, as relatively few nonhuman primates live in temperate or alpine environments.

Nonetheless, it is likely that the number of primates facing seasonal temperatures that are consistently near or below freezing, with the resultant seasonality in resource availability, is greater than is commonly assumed. These should be given more attention, as animals living in extreme environments can provide especially insightful examples of evolutionary adaptation in terms of behavioral and phenotypic plasticity (Cichy and Ford

1994).

Prosimii

South African Senegal bushbabies (Galago senegalensis) at Mosdene and brown

greater galagos (Otolemur crassicaudatus) at Louis Trichardt, both in the Northern

Transvaal, provide excellent examples of cold-weather primates (Harcourt 1986). Winter

minimum temperatures often approach freezing and can drop at Louis Trichardt to as low

as -10º C. During this time of year, insect availability is lower than in summer and gum

is harder and more crystalline in nature. In one study, the brown greater galago was

found to exploit mainly gum and reduce its activity levels in winter, while the smaller

Senegal bushbaby, at a slightly warmer site, exploited more insects and remained active.

Yet during a more severe winter at the same location, when insect availability was likely

further reduced, Senegal bushbabies were found to act like the brown greater galago, with

reduced activity and increased gum eating (Bearder and Martin 1980).

New World monkeys

Howler monkeys (Alouatta spp.) undoubtedly represent the most generalized of

platyrrhines, living in a wide variety of habitats with considerable variation in activity

patterns and especially diet. In southern Brazil, brown howlers (Alouatta guariba) at

Aracuri Ecological Station inhabit a temperate environment of broad-leaved and

coniferous forest with exceptionally cold winters. These howlers are more folivorous

than conspecifics at the subtropical Itapuá State Park, and these and other differences are

suspected to be linked with differing patterns of resource abundance (de Marques 2002).

Cercopithecinae

Numerous cercopithecines face seasonally cold temperatures, major resource

fluctuations, and dramatically reduced access to high-quality foods. While some of these

linger in academic obscurity, the same cannot be said for the famous snow monkeys of

Japan. Indeed, the superb, long-term work conducted on these animals has been

influential enough to convince some primatologists that Japanese macaques (Macaca

fuscata) are the only nonhuman primates to inhabit areas with snowy winters (McGrew

2004). While this is certainly not the case, it is true that some Japanese macaque

populations face up to 140 days of snow annually (Suzuki 1965), which may limit

movement and foraging on the ground (Watanuki and Nakayama 1993). Nakayama and

colleages (1999), studying macaques of the northern Shimokita Peninsula, found that

daily energy intake was up to five times lower in the winter than the fall, leading to

energy deficits (intake vs. expenditure) for all age-sex classes (Nakayama, Matsuoka, and

Watanuki 1999).

Barbary “apes” (Macaca sylvanus) in the Atlas Mountains of northern Africa forage on cedar cambium, cones, and needles during snowy periods (Richard 1985; Taub

1977). Other macaques (including Macaca assamensis, M. mulatta, M. arctoides, and M. thibetana in Asia) may face similar, if not as dramatic, challenges (Kumar, Karki, and

Ghimire 2001). The recently-described Arunachal macaque (Macaca munzala) of India, a close relative to, or synonym of, Macaca assamensis, lives up to 3500 m, the highest elevation yet reported for macaques on the Indian subcontinent (Sinha and others 2005).

The tremendous ecological plasticity of macaques has led some recent workers to propose them as putative models for some aspects of human evolution (Hart and Sussman

2005).

Geladas (Theropithecus gelada) inhabit desolate high-altitude stretches of

Afroalpine grassland (Fashing and Nguyen 2006). Living in areas almost devoid of trees and often facing nighttime temperatures near or below freezing, the gelada is unique among primates in having an almost completely graminivorous diet. Living in single- male, multi-female units that travel together in large bands, geladas get some 90% of their diet through grass blades, seeds, and rhizomes. They are also among the most terrestrial of nonhuman primates, with over 99% of feeding records occurring on the ground (Dunbar 1977).

The recently-discovered highland mangabey (Lophocebus kipunji) is another primate that lives in a relatively cold, marginal habitat (Jones and others 2005). The

long-haired monkey is found as high as 2600 m in the Southern Highlands and

Udzungwa Mountains of Tanzania, where temperatures drop to below -3º C. Little is yet known about its behavior.

Colobinae and apes

Some colobines are found in among the harshest habitats of all nonhuman

primates. The exemplars are the snub-nosed monkeys of China (Rhinopithecus spp.).

They live in deciduous and coniferous forests with snow cover up to half the year, and

two species (R. bieti and R. roxellana) have evolved the unusual adaptation of a lichen- based diet (Kirkpatrick 1998; see Vedder and Fashing 2002 for an interesting African colobine example). At Xiaochangdu, Tibet, Rhinopithecus bieti ranges between 3500 and

4250 m, where temperature averages below 0º C for four months of the year. Although lichens are indeed a staple, the monkeys at Xiaochangdu have a broad diet, and prefer leaves and other non-lichen food when it is available (Xiang and others 2007).

The ape representative is the mountain gorilla (Gorilla gorilla beringei) which is found in bamboo and montane forest to over 3500 m in the Virunga Mountains of Central

Africa. Thermoregulatory stress has been argued to be a significant problem for these primates, particularly during rainy periods when they are exposed to frequent moisture and nightly temperatures near freezing. There is a peak in mortality in these months for mountain gorillas, as well as for geladas, which face similar problems (Watts 1998).

Thermoregulatory stress is not the only impediment to living in such regions.

Although diverse in their habits, primates in marginal habitats share unique foraging

challenges which often result in seasonal shifts to low-quality foods (winter buds in

Macaca fuscata, Nakayama, Matsuoka, and Watanuki 1999) or year-round inclusion of largely non-seasonal resources (lichens in Rhinopithecus spp., Kirkpatrick 1998; but see

Liming 2006). This study focuses on foraging strategies of Himalayan langurs, a geographical variant of the common gray langur (Colobinae) of the Indian subcontinent.

The gray langur: A brief history of a very adaptable colobine

Gray langurs (Semnopithecus entellus) range from Sri Lanka to the Himalayas in

habitats ranging from semi-desert and subtropical forest to temperate forest and sub-

alpine scrub (Bishop 1977; Bishop 1979; Brandon-Jones 2004; Koenig and Borries

2001). Arguably, the only nonhuman primate species that rivals them in terms of habitats

occupied is the rhesus macaque (Macaca mulatta).

Early observations on gray langurs living in the Simla district of India were recorded by Dodsworth (1914), who noted a diet of fruits, buds and flowers for the

animal and commented that in captivity “it has not the vicious or depraved habits of S.

rhesus (the rhesus monkey)” (Dodsworth 1914:732). Perhaps the “depraved habits” of

confined rhesus monkeys as compared to gray langurs were related to what they were

being fed. As discussed by McCann (1928:192): “Doses of 5 to 10 grains of strychnine

have been administered to a Common Langur without effect, while the same dose kills a

Rhesus monkey in a very short time.” This is one of the first allusions to colobine

monkeys’ impressive ability to eat low-quality foods and even detoxify some otherwise

dangerous compounds. In a later and more detailed report, McCann (1933) noted some

features of gray langur social organization (from the Abu Hills, India) that have been borne out by subsequent observation, such as the existence of single-male troops and all- male bands at least at some sites.

The most detailed physical study of the monkey came soon after with Ayer’s publication of The Anatomy of Semnopithecus entellus in 1948 (Ayer 1948). In a review covering the entire body, he noted the large, sacculated stomach that is known to be an adaptation allowing for the digestion of high-fiber foods such as mature leaves. Also important for topics that we will consider later, Ayer commented: “This study of external morphology indicates that Semnopithecus entellus has probably the most advanced brain among the Cercopithecidae. A study of its internal structure is likely to be an extremely valuable addition to our knowledge of the primate brain” (Ayer 1948:149). While a preliminary and subjective statement, it is known today that this humble colobine has a larger neocortex relative to the rest of its brain than capuchins or spider monkeys, and most of its fellow Old World monkeys (Kudo and Dunbar 2001).

The modern era of gray langur studies, however, began with the influential work conducted by Phyllis Jay (later Phyllis Dolhinow) from 1958-60 (Jay 1965; Jay 1963) on langurs at Orcha (Madhya Pradesh) and Kaukori (Uttar Pradesh) in North India. At these sites, langurs were found to live in multi-male, multi-female groups (typically) that ranged in size from 5 to 120 or more individuals. Groups were cohesive, and remained constant with the exception of births, deaths, and the departure of adult males. In contrast to the reputation that the gray langur would later obtain, Jay described her subjects as relatively peaceful with a minimum of aggressive interaction. Jay also proposed a langur

age-sex classification system (dark infant, white infant, juvenile, subadult, adult) that is still in use today.

In the early 1960s, another research program was initiated by Yakimaru Sugiyama of the Japan-India Joint Project in Primates Investigation (Sugiyama 1964). This work focused on gray langurs and bonnet macaques (Macaca radiata) living near Dharwar of

Mysore State in southern India. Suggestive of the extreme social flexibility of gray langurs, Sugiyama found that close to ¾ of bisexual troops at Dharwar included only a single adult male, in contrast to Orcha and Kaukori. Males outside of these troops lived in all-male bands. In addition, this project provided the first detailed account of gray langur diet, which included leaves (over half of feeding records) as well as fruits, flowers, insects and bark (Yoshiba 1967).

Most spectacularly, Sugiyama (1965) made a series of observations that were to guide the study of gray langurs over the next four decades (Sugiyama 1965). These included aggressive troop takeovers by outside males and attacks on individuals in the group, most notably infants which were sometimes killed. Among other possibilities,

Sugiyama suggested that killing infants they did not sire would allow males to impregnate females sooner: “non-sucking-stimulus decreases the secretion of lactogenic hormone and correlatively increases gonadotropin, which urges a female to oestrus” (p.

413). In this statement Sugiyama presented the kernel of what was later to be developed into the sexual selection hypothesis for infanticide (Hrdy 1974). Bolstered by observations at Jodhpur by S.M. Mohnot (1971) and at Abu by Sarah Hrdy (1974), and later at a number of other sites and with other species, this hypothesis views infanticide as

an evolutionary (and thus genetically-influenced) reproductive strategy. It has also become one of the most controversial topics in primatology, a great “holy war” in the words of Volker Sommer (Sommer 2000:9). On the one hand are sociobiologists who view the sexual selection hypothesis as a powerful explanation of an important phenomenon (Hrdy 1974; Koenig and Borries 2001; van Schaik and Janson 2000) and on the other critics who view “infant killing” (the word “infanticide” is anathema to these workers) as a byproduct of general aggression sometimes brought about by social crowding and pathology (Bartlett, Sussman, and Cheverud 1993; Boggess 1979; Curtin and Dolhinow 1978).

Since the early studies of Jay and Sugiyama, gray langurs have been observed at a minimum of 30 sites (Koenig and Borries 2001). Most work has focused on social behavior in general and topics related to infanticide in particular (Newton 1992). Of the studies to date, only a small but important subset has investigated ecological questions.

In addition, most work, including all hitherto mentioned, has involved lowland langurs.

Himalayan gray langurs

Little is known about the Himalayan varieties of gray langur, even their

taxonomy. The first long-term investigation began with Naomi Bishop, a student of

Phyllis Dolhinow, and her husband John, who investigated langur social behavior in

predominately one troop at Melemchi, north-central Nepal (2442-3050 m) in 1971 and

1972 (Bishop and Bishop 1978; Bishop 1975; Bishop 1979). The habitat was mainly a

temperate, and relatively thick, Himalayan oak (Quercus semecarpifolia) forest. In

addition to collecting baseline data for high-altitude langurs, including troop composition and general behavior patterns, Naomi Bishop presented what remains the most detailed description of Himalayan langur auditory communication (Bishop 1975). In addition, during a census in August of 1972, Bishop noted langurs at 4050 m elevation at Routang, a high-altitude yak pasture, and received reports that on the previous day the same troop had crossed a ridge at 4250 m (Bishop 1977). These are among the highest altitudes yet reported for Himalayan langurs.

Next were the studies of Jane Boggess (1976, 1980) and Richard Curtin, again students of Phyllis Dolhinow, at Junbesi (2442-3505 m) in the Everest (Solu Khumbu) region of Nepal. Boggess studied social behavior and male membership changes in six troops from 1972 to 1974, and again in 1976 and 1978 (Boggess 1980; Boggess 1976;

Boggess 1982). She described what has become known as a Himalayan-typical pattern of social organization, with mainly multi-male, multi-female troops, extratroop males occurring as solitaries or pairs (and not all-male bands), and the lack of troop takeovers or infant killing. Instead, Boggess found “an alternating pattern of exclusions and introductions with gradual adult male replacement” (Boggess 1980:233). Perhaps because of the differences between these observations and those from the Indian sites of

Dharwar, Jodhpur or Abu, Boggess became one of the major early critics of the sexual selection hypothesis for infanticide. No matter one’s views on this contentious issue,

Boggess’ writings provide perhaps the most detailed and reasoned counterpoint to the hypothesis. Indeed, her critique in the volume Infanticide: Comparative and

Evolutionary Perspectives (Boggess 1984) was considered sufficiently grave that the editors saw fit to include two rebuttal chapters (Hrdy 1984; Sugiyama 1984).

Curtin collected data on ranging and preliminary information on langur foraging, mainly from one troop, at Junbesi from 1972 to 1974 (Curtin 1975; Curtin 1982). In addition to points that will be detailed later, he found that the langurs ranged over wider areas than those populations previously studied, particularly in the winter when little food was available. At the time, Curtin had provided one of the best examinations of the ecological problems faced by a generalist primate at the extreme of its range.

Two more studies of shorter duration should be mentioned. Christian Vogel compared and contrasted langur populations at the Indian sites of Kumaon-Hills

(Himalayan, 900-1500 m) and Sariska (Rajasthan, 400 m) based on 3 months of study in

1968 (Vogel 1971). Although not all of the differences found between the sites have been borne out by subsequent observation, he did note greater birth seasonality in the

Himalayan area compared to the lower site. Today it is known that birth seasonality is most pronounced not only in Himalayan areas, but also in lowland forest reserves where provisioning does not occur (Koenig and Borries 2001). Lastly, Sugiyama collected seasonal data over 5 months on Himalayan langur diet from the Himachal Pradesh, India

(1500-3200 m), and included a list of foods taken (Sugiyama 1976).

Given the above, Bishop characterized a Himalayan pattern of langur ecology and behavior that is different in many respects from lowland forms (Bishop 1979). To sum, highland langurs form predominately multi-male, multi-female troops, use expansive home ranges, employ different vocalizations than lowland langurs, and exhibit behavioral

and morphological buffers to cold weather (Bishop 1979). Aggressive troop takeovers and infanticide have not yet been observed, and outside males occur singly or in pairs.

Goals of the project

Detailed observations of Himalayan langur foraging behavior have yet to be

collected. For example, no long-term studies have included systematic quantification of

diet coupled with phenological sampling. There have also been no studies that have

attempted to identify Himalayan langur food preference or quantify the relationships

between resource abundance, activity patterns, diet, and ranging. In addition, there have

been no nutritional analyses of Himalayan langur foods or theoretical treatments of their

feeding behavior. The current study is an attempt to remedy the situation, investigating

Himalayan langur foraging decisions from the standpoint of current models of primate

socioecology and cognition, nutritional ecology, and optimal foraging theory.

CHAPTER II

STUDY SITE AND SUBJECTS

Langtang National Park, Nepal

Langtang National Park is located in north-central Nepal on the Tibetan border

(Figure 1). It was established in 1976 as Nepal’s first National Park, and at an estimated

1710 km2 is among its largest (Green 1981). With altitudes varying from approximately

800 m to over 7200 m, habitats range from subtropical forest to perpetual snow. A

glacier-fed river, the Langtang Khola, cuts through the northern section, forming a steep- walled valley. The region was first made known to the western world with explorer H.W.

Tilman’s Nepal Himalaya (Tilman 1952). This book details, among other sojourns, two trips to the Langtang, and contains an appendix discussing the natural history (mostly

botany) of the valley (Polunin 1952).

Although a comparable survey has yet to be completed for Langtang, the book

The Arun: A Natural History of the World’s Deepest Valley represents the best

description of the staggering ecological differences that can occur within small

Himalayan areas based on altitude (Cronin 1979). The Arun River flows through eastern

Nepal and where running between Mount Everest and Kanchenjunga is over 20,000 feet

deep. The fauna of the lower hills south of the Arun includes water buffalo and tigers,

while the upper altitudes are home to pika and snow leopard. Although the Arun survey

13 14

Figure 1. Location of Langtang National Park, Nepal, and a topographical map of the main study area.

15 investigated a relatively wide area, such abrupt ecological changes are typical of any large Himalayan valley, including Langtang.

Vegetation types of main study area

The Langtang Valley between Ghore Tabela (3033 m) and Langtang village

(3480 m) was the primary area of observation. Several vegetation types are present, with

different woody species characterizing each (Figure 2). On the north side of Langtang

Khola, mixed oak (Quercus semecarpifolia) forest predominates to approximately 3100 m. Himalayan oaks are evergreen and dispersed with smaller evergreens, such as

rhododendrons, and an assortment of deciduous woody plants (Malla 1976; Polunin and

Stainton 1997).

Above 3100 m oaks rarely occur, and smaller trees and shrubs make up much of

the woody plant cover. The relative density of plant species within this zone varies

markedly, and vegetation from 3100 m to 3200 m will be termed “scrub,” and above

3200 m as “high scrub.” Both scrub and high scrub habitats are characterized by small to

medium-sized deciduous trees and shrubs, with rhododendrons and other broad-leaves

representing the evergreen woody plants.

While coniferous trees are comparatively rare on the north side of the river, the

south side is largely coniferous forest, with Himalayan hemlock (Tsuga dumosa) and

silver fir (Abies spectabilis). Other habitat types include fields (cultivated areas and yak

pastures), rockslides and cliffs. The village of Langtang is located at the northeastern

extremity of the monkey range, but scattered human residences are found throughout the

Figure 2. Woody habitat types of the study area, with phenological plots indicated.

17 valley, mainly catering to foreign trekkers who travel through this area predominantly in spring and fall.

More detail on predominate vegetation types is included in the phenological analysis discussed later.

Climate of main study area

The climate at Ghore Tabela is highly seasonal (Figure 3). Mean annual

temperature for 2003, using monthly maximum-minimum midpoints, was 13.6° C.

Winters (December-March) were characterized by cold nights, usually near freezing, and

days that were generally clear and mild. During both spring (April-May) and fall

(October-November), nights were cool and days were warm and sunny. Although snow

occurred in varying amounts from November to May, 76% of the 1374 mm annual

precipitation occurred as rain during the monsoon months from June to September. Since

the monkeys moved over a considerable altitudinal range, temperature and precipitation

often depended on what part of the habitat they were exploiting. During spring, for

example, rain at 3000 m elevation often turned to snow at 3200 m.

Mammalian fauna in main study area

The mammalian fauna of Langtang National Park was surveyed by Michael J. B.

Green and the Durham University Himalayan Expedition from April 1976 to May 1977

(Green 1981). Above 3000 m are an assortment of insectivores (shrews), rodents,

carnivores, artiodactyls, and one lagomorph (the Himalayan mouse-hare or pika).

Figure 3. Temperature and precipitation at Ghore Tabela (3033 m), Langtang National Park, Nepal. Precipitation was snow/sleet or rain at the altitude the monkeys were traveling. February max-min temperatures were not recorded and are assumed to represent midpoints between January and March.

350 25 300 20

250 15 snow 200 10 rain 150 5 mean high 100 0 mean low Temperature (C) Precipitation (mm) Precipitation 50 -5 0 -10 JFMAMJJASOND Month

Notable species include red panda (Ailurus fulgens) and snow leopard (Panthera uncial), both carnivores, and the Himalayan tahr (Hemitragus jemlahicus), an artiodactyl (Green

1981).

The only nonhuman primate found above 3000 m is the Himalayan gray langur

(Semnopithecus entellus). The Durham expedition noted langurs at altitudes of 4120 m, which is comparable to observations that will be discussed here. The Assamese macaque

(Macaca assemensis) is found at lower elevations, up to approximately 2500 m where it is sympatric with the gray langur (Kumar, Karki, and Ghimire 2001; personal observation). The rhesus macaque (Macaca mulatta) is also found within the Park, but mainly at lower altitudes where all three monkeys are found (Green 1981; Kumar, Karki, and Ghimire 2001).

Study subjects

Age-sex classifications for the Langtang langurs follow Bishop (1975), as modified from Jay (1963; Dolhinow 1972). I chose the highest ranging troop in the main

Langtang Valley, in terms of altitude, as the main study group (F troop) (Figure 4). This troop was never sighted below 3000 m altitude, ranging from above Ghore Tabela (3033 m) to Langtang (3480 m), and moved on surrounding cliffs to altitudes estimated at 4000 m or higher. After 3 months of habituation, F troop could generally be approached to within 10 m, but observations in all months were limited more by weather and habitat characteristics than by habituation. Group size for F troop ranged from 27-33 individuals, with a modal number of 3 adult males and 10 adult females. Five infants

Figure 4. F troop near Langtang Monastery (approximately 3500 m).

21 were born during the main observation period in F troop, with births ranging from

December 2002 to early May 2003. Timing of reproduction at Langtang thus corresponds to the birth peak for Himalayan langurs as noted by Bishop (1979).

A secondary semihabituated troop, B troop, traveled between 2500 and 3200 m with overlap with F troop’s range on the northern side of Langtang Khola. Not as well known as F troop, B troop was followed opportunistically in the monsoon of 2003 when

F troop could not be located. The highest count for B troop included 55 monkeys and five adult males.

Taxonomic identification of study subjects

Although Bishop (1979) suggested that only one Himalayan langur type be

recognized, recent workers propose two high-altitude subspecies (Brandon-Jones 2004;

Brandon-Jones and others 2004; Napier 1985) or species (Groves 2001). These are the

pale-armed (Semnopithecus entellus schistaceus) and dark-armed (S. e. ajax) Himalayan langur. The pale-armed Himalayan langur is apparently very widespread, ranging from

Bhutan to possibly Afghanistan, and the dark-armed form is known only from specimens from Jammu and Kashmir, and Pakistan (Brandon-Jones 2004; Brandon-Jones personal communication). As the names suggest, the major feature delineating these forms is the darkness of the forearms. In the schistaceus variety, the forearms are similar in coloration to the upper arms and back or only slightly darker, while in the ajax variety the forearms are “dark brown or black” (Napier 1985:77).

The langurs of Himalayan north-central Nepal, including Langtang, are generally placed in Semnopithecus (or Presbytis) entellus schistaceus (Napier 1985) or S. schistaceus (Groves 2001). Brandon-Jones (2004), however, suggests that langurs in this region should be classified as the dark-armed Himalayan langur (Semnopithecus entellus ajax or S. ajax) based on photographic evidence –the presence of dark forelimbs in photographs of Melemchi langurs published in Bishop (1979; see also the cover and plates in Bishop and Bishop 1978 for variation). However, the only museum specimens from the Helambu Valley, where Melemchi is located, are referable to Semnopithecus entellus schistaceus (Brandon-Jones 2004).

To add to the confusion, Langtang langurs exhibit adult variation in forearm and back coloration, but in none did I interpret the differences in forearm and back shading as striking as that suggested by ajax descriptions (Figure 5). It is appropriate to tentatively retain them in schistaceus, although more data needs to be collected on intra- and inter- troop pelage variation (Oppenheimer 1977) from highland langurs across their range to adequately test the Brandon-Jones (2004) hypothesis. It should also be noted that

Himalayan langurs of both the schistaceus and ajax varieties are often described as brown in coloration (Brandon-Jones 2004; Groves 2001; Hill 1939; Napier 1985; Pocock

1928). The Langtang langurs are not brown or brownish, but gray, as are those at

Melemchi (Bishop and Bishop 1978) and Junbesi (Curtin 1975).

Figure 5. Variation in forearm coloration in F troop. (a) adult female, (b) adult male (facing).

(a)

(b)

CHAPTER III

DIET, ACTIVITY PATTERNS, AND RESOURCES

INTRODUCTION

In current models of primate socioecology, leaves are generally considered

ubiquitous or non-patchy resources that are unlikely targets of contest competition (Isbell

1991; Wrangham 1980). Folivorous primates are expected to exhibit shorter daily path lengths and smaller home ranges than frugivorous primates (Clutton-Brock and Harvey

1980), with rest-dominated, energy-minimizing, activity budgets (Oates 1987). Although

comparative studies sometimes support these generalizations (Sterck, Watts, and van

Schaik 1997), increasing evidence suggests that, under some circumstances, they are

unlikely to be correct.

When leaves occur on trees or shrubs that are separated from one another by areas

with little food they can be considered patchy (Astrom, Lundberg, and Danell 1990). The marginal value theorem (Charnov 1976a) predicts that as overall food abundance decreases, patches will be further spaced out, and both travel times between patches

(travel budget) and patch residence times (feeding budget) will increase (Stephens and

Krebs 1986). Thus, at low enough levels of resource abundance, a folivorous primate could be expected to have daily path lengths and activity budgets similar to those of frugivorous primates at less marginal sites. For example, black snub-nosed monkeys

(Rhinopithecus bieti) are conservatively estimated to travel an average of 1,310 meters 24 25

each day even though they feed heavily on lichens, which, like leaves, are often

considered ubiquitous and non-patchy (Kirkpatrick 2007).

I present ecological and behavioral results designed to: 1) quantify gray langur

diet, activity patterns and resource availability at an extreme of their geographic range, 2)

identify food preference by comparing plant part consumption and abundance based on

phenology scores, and 3) relate feeding budgets and daily path lengths to the abundance

and the consumption of various plant parts. The results are compared and contrasted with generalizations frequently made concerning the behavioral ecology of primate folivores.

METHODS

Activity and Feeding Data

The author and/or two Nepalese field assistants performed group follows on F

troop during 10 months between January 2003 and February 2004. Contact was

established with F troop during January 2003 and then monthly between March and May

2003, and from September 2003 to February 2004. Ideally, our group follows consisted

of locating a troop in the morning near their sleeping site, generally cliffs, and following

them until they entered another sleeping site that evening. Contact with the monkeys,

however, could never be guaranteed, and from June to August 2003 F troop was not

contacted despite hundreds of hours of search by both the research team and hired local

trackers. Thus, monsoon data from F troop is limited to observations from September

2003. In total, F-troop was followed for approximately 775 hours between January 2003

and February 2004. During that time, the average number of individuals observed per

scan, using monthly means, was 13.3 ± 4.2 (40.3 % of individuals, using modal group size).

B troop was followed for approximately 292 hours during the monsoon of 2003, monthly from June to September. B troop generally used large trees as sleeping sites. B troop group follows were similar to those performed on F troop, although more opportunistic due largely to monsoon weather. Most days were characterized by thick

fog that rolled into the valley during the morning hours and contact with the monkeys

was often lost if they ascended cliffs outside of the range of visibility or human climbing.

Due to comparatively lush monsoon vegetation, the proportion of B troop individuals that

could be seen per scan (25.8%, or 14.2 ± 1.1 individuals) was lower than that of F troop.

Observations were generally carried out by naked eye or through binoculars,

although a spotting scope facilitated observations when monkeys used cliff habitats. We

recorded general activity (feed, travel, rest, rest-huddle, rest-cling, groom, play, and

miscellaneous social behavior) by scan sampling at 20-minute intervals (Altmann 1974).

Activity for each visible individual was recorded at the moment it was first observed

during scans, which continued for 10 minutes. For each individual that was feeding

during scans, we recorded the food species and plant part. This method unfortunately

biases observations in favor of items that are more likely to be observable in the

Himalayan environment (Curtin 1982). At Langtang, observations may have been biased

towards arboreal feeding and terrestrial feeding in field habitats, and biased against

terrestrial feeding in oak forest, scrub, and high scrub habitats (e.g., herbaceous plant

exploitation), due to visibility.

Daily path lengths were estimated using GPS point-to-point sampling. For travel that could not be recorded with the unit, such as vertical movement on cliff faces, I estimated distance to the nearest 10 meters. GPS checks on flat or slightly sloping ground suggests the distance estimations were accurate to within 10 meters for distances of up to 100 meters.

Classification of dietary items

Distinctions among plant parts were often made based on information in field

guides on the flora of the region (Polunin and Stainton 1997; Stainton 1997). Broad-

leaves were categorized as mature or young (based on size, color and texture), bud

(generally dormant winter buds), or petiole, and were separated post-hoc into the

categories deciduous and evergreen. Fruit was categorized as ripe or unripe based on

color and size, or dehiscence state, and the portion(s) of the fruit eaten was noted.

Unopened flowers were considered flower buds.

Underground storage organs were classified as soft (mainly tubers) or hard

(mainly woody roots). The “miscellaneous” category was used if underground resources

could not be allotted to a more specific category. Herbaceous plant parts included herb

leaves, herb fruits, herb flowers, young fern furled tops, and epiphytic fern rhizomes.

Coniferous vegetation was labeled as needle or cone only. Other plant parts and food

items included: bark, young bamboo shoots, mosses and lichens, grass blades,

mushrooms, invertebrates, and earth (rock-licking).

Phenology

The author conducted phenological samples on nineteen occasions in nine plots

(total area = 0.75 ha) during the 2003 calendar year. Sampling took place on the first

and/or 15th of each month, for nineteen two-week “sampling periods” (see below). Three

16.5 × 50 m plots were established in coniferous forest, one in Quercus semecarpofolia forest, three in scrub (3100 m –3200 m), and two in high scrub (above 3200 m) (Figure 2,

Table I); field and cliff habitats were not sampled. All trees, shrubs, and climbers with diameter at breast height of 10 cm or greater were measured. Voucher specimens from each species were collected and later identified by plant scientists at the Central

Department of Botany, Tribhuvan University, Kathmandu, Nepal.

For broad-leaved species, I estimated abundance of mature leaves, young leaves, leaf buds, ripe fruit, unripe fruit, flower buds, and flowers, relative to maximum crown volume, for each species in each plot on a 0-5 scale in increments of 0.5 (modified from

Dasilva 1994). A score of 5 would indicate that the plant part in question was found in all parts of the crown, and covered all visible portions of the crown. In practice this makes 5 the highest possible combined score for vegetative structures (mature leaves, young leaves, leaf buds) and reproductive structures (fruits and flowers). Since reproductive structures generally are not found entirely throughout the crown, scores for these plant parts rarely approached the maximum value. This allows for estimation of absolute abundance of plant parts, and also allows for comparison between different plant part groups. I also divided broad-leaves post-hoc into two groups, evergreen and deciduous. Thus, the plant parts for these species included evergreen mature leaves,

Table 1. Phenological sample. Basal areas are given in m2 per hectare across all plots.

D = broad-leaved deciduous, E = broad-leaved evergreen, C = coniferous.

Genus Species Type Basal area Part(s) eaten % of diet Cotoneaster frigidus D 13.45 bark, fruit, leaf, leaf bud 22.0 Berberis aristata D 11.64 fruit, leaf, root 3.9 Elsholtzia fruticosa D 9.78 flower, root 0.2 Caragana gerardiana D 8.83 seed, root1 7.3 Tsuga dumosa C 6.49 bark, cone, needle 0.3 Sorbus cuspidata D 5.91 bark, fruit, leaf bud 6.9 Quercus semecarpifolia E 4.86 leaf2 0.1 Rhododendron arboreum E 3.56 Abies spectabilis C 3.13 Zanthoxylum nepalense D 2.93 bark, flower, leaf 13.8 Ribes sp. D 2.77 leaf, leaf bud 0.2

Hippophae rhamnoides D 2.36 fruit, leaf, leaf bud 6.3 Ilex dipyrena E 1.58 leaf, petiole 0.4 Viburnum cotinifolium D 0.96 leaf bud3 0.2 Acanthopanax cissifolius D 0.94 Salix tetrasperma D 0.53 flower, leaf, leaf bud 1.4 Cotoneaster acuminatus D 0.51 fruit, leaf, leaf bud 0.3 Rhododendron barbatum E 0.50 Jasminum humile D 0.33 bark, fruit, leaf, leaf bud 6.7 Rosa macrophylla D 0.21 fruit, leaf 1.3 Viburnum erubescens D 0.19 leaf 0.2 Unidentified D 0.18 Unidentified D 0.16 fruit, leaf, leaf bud 1.7 Rosa sericea D 0.14 fruit, leaf, leaf bud 1.6 Aster albescens D 0.12 Betula utilis D 0.10 Acer caudatum D 0.08 Viburnum nervosum D 0.08 Neillia thrysiflora D 0.05 Rabdosia sp. D 0.05 Pieris formosa E 0.04 Larix himalaica C 0.02 Rubus sp. D 0.01 leaf 0.1 Clematis acuminata D 0.01 Sum 74.9 Notes: 1. Flower taken by B troop in monsoon. 2. Seed taken by B troop in monsoon. 3. Fruit taken by B troop in monsoon

evergreen young leaves, deciduous mature leaves, deciduous young leaves, leaf buds,

ripe fruit, unripe fruit, flower buds, and flowers. Using the same sampling strategy I

measured the abundance of two plant parts on conifers, needles and cones.

I calculated species-specific contribution to forest production using:

Wi = ( ∑ Ai / ni ) * Bi

where Wi is the weighted abundance of a plant part, Ai is the phenological score, ni is the number of individuals, and Bi is the basal area per hectare in square meters, all for species

i (modified from Dasilva 1994). Summed totals for all species, and each plant part, were

utilized for estimates of overall vegetation abundance.

Data analysis

All analyses described here are limited to data from F troop, although the

monsoon diet and path lengths of B troop will be provided for reference. A sampling

period consisted of the day of phenological sampling and the two-week period following, and eleven sampling periods corresponded with feeding data for F troop. I calculated

Spearman rank correlation coefficients between the abundance score and the percent contribution of a plant part to diet during a given sampling period. Correlations between the abundance and consumption of specific plant parts have been used to assess large scale preferences in other primates (e.g., Dasilva 1994). Many of the correlations are not meaningful in the sense of langur food preference, because some resources that are available year round are only taken in some seasons at Langtang. For this reason, only the consumption of evergreen mature leaves and bark, two largely non-seasonal

31 resources, are compared with the abundance and consumption of other plant part groups.

Spearman correlations were used to identify the potential relationship between overall vegetation abundance and feeding budgets.

Kruskal-Wallis was used to test for seasonal differences in daily path lengths (n =

84) for winter, spring, monsoon, and fall. In order to examine the relationship between travel distance and diet when controlling for vegetation abundance, I performed stepwise multiple regression with daily path length and summed phenological scores for individual days (n = 76) as independent variables, and log transformed the percentages of food types ingested on those days as the dependent variable. Regressions were performed separately for each food type. To avoid taking the natural log of zero for those days when a certain food type was not eaten, a constant of 0.001 was added to each daily diet percentage before log transformation. Sensitivity analysis with constants 0.01 and 0.0001 did not alter inclusion or exclusion of variables. Unless stated otherwise, analyses are two-tailed and the level of significance is 0.05. Statistical analyses were performed with SPSS 13.0.

RESULTS

Activity patterns

The adult activity budget for F troop, based on 3379 records where age-sex identification was established, includes feeding (39.8%), travel (17.5%), resting (29.2%), huddling

(3.2%), grooming (9.5%), and miscellaneous social behavior (0.9%). A significant negative correlation was found between estimates of total vegetation abundance and frequency of feeding records (n = 11, Spearman, r = -0.96, p < 0.001).

Plant production

The 34 woody plant species in the phenological sample accounted for 74.9% of

langur diet (Table 1). Conifers bore needles and cones throughout the year with little

variation in abundance. Evergreen young leaves only appeared during a brief period of

leaf turnover during the early monsoon (June-August). Broad-leaved species, particularly

vegetative structures, showed a pattern of marked seasonality in plant production (Figure

6). Considering all plots, broad-leaved deciduous plants were most abundant by basal

area, and deciduous leaf portions were the most abundant plant parts for all seasons

except winter. The availability of young deciduous leaves peaked in June, but by July

most deciduous leaves were classified as mature. Leaf buds were available mainly in

winter and spring.

Reproductive plant parts, with abundance scores consistently lower than

vegetative parts, also showed seasonal variation in abundance (Figure 7). Flowers

showed two peaks in abundance, the first in monsoon and the second in fall. Although

spring flowering did occur in several species, monsoon flowering was characteristic of

most plants in the sample. Unripe fruit was available mostly in monsoon and fall (June-

November) with a peak in September. Ripe fruit, while available from August to April, was most abundant in late fall and winter (October-February).

Figure 6. Abundance of broad-leaved vegetative structures at Langtang as determined via phenological analysis.

250

200 leaf buds 150 deciduous young leaves deciduous mature leaves 100 evergreen young leaves evergreen mature leaves Abundance Units Abundance 50

0 JFMAMJJASOND Month

Figure 7. Abundance of reproductive structures at Langtang as determined via phenological analysis

40 35 30 flowers 25 flower buds 20 unripe fruit 15 ripe fruit 10 Abundance Units Abundance 5 0 JFMAMJJASOND Month

Diet

Members of F troop were observed feeding on plant foods from a minimum of 30

families, 39 genera, and 43 species. More than half (57.1%) of the nine-month sample

from F troop (March-May 2003, September 2003-February 2004: n = 9895 feeding

records) was made up of leaf parts (Table 2). This included deciduous mature leaves,

leaf buds, deciduous young leaves, evergreen mature leaves, and herb leaves, as well as

unidentified leaves, coniferous needles, and evergreen mature leaf petioles. Ripe, unripe,

and herbaceous fruit made up 22.4% of the total, with an average of 7.3% of monthly records representing seeds. Underground foods made up 7.7% of the diet, with 5.3% representing soft underground storage organs, and the rest miscellaneous underground resources and hard or woody underground storage organs. Flower parts, including flowers, herbaceous flowers, and flower buds, contributed 6.9% to the diet, and bark made up 5.4%. Other items included mosses and lichens, coniferous cones, epiphytic fern rhizomes, grass, young bamboo shoots, suspected invertebrates, and earth (rock- licking). The top ten items in the nine-month sample made up 58.5% of total feeding

records (Table 3).

Insectivory was limited to one case of suspected arthropod foraging (sensu

Struhsaker 1978) during a winter scan and one case of a langur catching and consuming a

grasshopper during fall non-scan, ad libitum sampling.

Table 2. Monthly percentages and nine-month averages of feeding records for 12 food types for F troop. USO = underground storage organ, MUR = miscellaneous underground resource.

Deciduous Deciduous Evergreen Leaf Ripe Unripe Soft Herb Herb mature young Flowers Bark mature MURs Other buds fruit fruit USOs fruit leaves leaves leaves leaves March 0.0 45.3 3.5 0.0 0.0 0.0 14.0 0.5 19.9 0.0 0.0 14.5 2.3 April 0.0 18.5 0.0 36.7 0.0 27.5 8.4 0.0 0.0 0.0 4.4 3.1 1.2 May 0.0 0.6 0.0 55.2 0.0 31.8 7.5 0.0 0.0 0.0 4.9 0.0 0.0 September 64.4 0.0 9.3 0.0 4.7 0.0 1.1 0.0 0.0 4.2 3.9 0.0 12.9 October 48.4 0.0 5.5 0.0 13.8 1.0 0.1 14.3 0.0 12.3 3.9 0.0 0.7 November 40.7 0.0 20.6 0.0 18.7 0.8 0.6 12.5 0.5 2.3 2.7 < 0.1 0.4 December 33.3 5.1 19.8 0.0 28.8 0.0 0.9 9.0 0.0 1.2 0.0 1.2 0.6 January 0.0 41.7 34.5 0.0 0.0 0.0 7.5 8.6 3.9 0.9 0.0 0.0 2.9 February 0.0 62.6 9.8 0.0 0.0 0.0 7.9 2.4 15.1 0.2 0.0 0.0 2.1

Nine-month 20.8 19.3 11.4 10.2 7.3 6.8 5.4 5.3 4.4 2.3 2.1 2.1 2.6 average

Table 3. Top ten food items in the Langtang Himalayan langur diet, listed by average monthly percentage of feeding records from the nine-month sample. The plant part was included in a season if it contributed greater than 1.0% of records. W = winter, S = spring, M = monsoon, F = fall. USO = underground storage organ.

Family Genus Species Part eaten Season(s) % of diet Rosaceae Cotoneaster frigidus decid mature leaf W, M, F 12.6 Rutaceae Zanthoxylum nepalense decid young leaf/flower S 10.8 Leguminosae Caragana gerardiana seed M, F, W 7.3 Rosaceae Cotoneaster frigidus leaf bud W 6.6 Oleaceae Jasminum humile bark W, S 4.0 Rosaceae Sorbus cuspidata leaf bud W 3.9 Ericaceae Gaultheria sp. evergreen mature leaf W 3.8 Berberidaceae Berberis aristata ripe fruit W, F 3.4 Elaeagnaceae Hippophae rhamnoides decid mature leaf M, F 3.1 Elaeagnaceae Hippophae rhamnoides leaf bud W 3.0

Total 58.5

Seasonal diet

Winter (December-March) was characterized by the lowest scores for total

vegetation abundance. The majority of the diet was made up of leaf buds, particularly

from Cotoneaster frigidus and Sorbus cuspidata, and ripe fruit (e.g., Berberis aristata and Cotoneaster frigidus) (Table 4). Likely fallback foods include evergreen mature leaves (especially Gaultheria sp.). This resource is available all year, but regularly

exploited only in winter. Bark was taken from at least five woody plant species.

Spring (April-May) showed the first leaf flush and deciduous young leaves made

up much of the diet (Table 4). The young leaf and flower cluster, taken concurrently, of

Zanthoxylum nepalense was by far the most important spring food item, followed

distantly by the young leaves of Jasminum humile. Bark continued to be a relatively

important resource, making up approximately the same proportion of the diet in spring as

it did in winter. However, this was limited to one species; the green bark of Jasminum humile accounted for 136 of 140 bark feeding records.

Monsoon (June-September) marked the reduced availability of deciduous young

leaves, a brief period of evergreen young leaf availability, and the flowering and fruiting

of numerous plant species. F-troop data for this season are limited to September, when

deciduous mature leaves made up majority of the diet, followed by fruit. Observations from B-troop were collected in all months during the monsoon (Table 4).

Fall (October-November) showed a decrease in overall plant part availability as

deciduous leaf drop began. Deciduous mature leaves were the main dietary component

Table 4. Seasonal diets from March 2003 to February 2004: winter (December-March), spring (April-May), monsoon (June-September), and fall (October-November). All values reflect average monthly contribution to feeding records. All data are from F troop except as noted.

Late Monsoon Winter Spring Fall monsoon (B troop) Deciduous mature leaves 8.3 0.0 64.4 36.7 44.6 Deciduous young leaves 0.0 45.9 0.0 23.0 0.0 Evergreen mature leaves 9.8 0.0 0.0 0.1 0.3 Evergreen young leaves 0.0 0.0 0.0 5.1 0.0 Leaf buds 38.6 9.6 0.0 0.0 < 0.1 Ripe fruit 16.9 0.0 9.3 1.3 13.1 Unripe fruit 7.2 0.0 4.7 15.8 16.3 Flowers 0.0 29.7 0.0 9.5 0.9 Soft underground storage organs 5.1 0.0 0.0 0.0 13.4 Bark 7.6 8.0 1.1 0.0 0.4 Herbaceous leaves 0.0 4.7 3.4 4.5 3.3 Herbaceous fruit 0.6 0.0 4.2 0.9 7.3 Miscellaneous underground 3.9 1.6 0.0 0.0 0.0 resources Other 2.011 0.622 12.933 3.144 0.555 Notes: 1. Hard and/or woody underground storage organs (0.8%), mosses or lichens (0.8%), unclassified leaves (0.1%), coniferous needles (0.1%), evergreen mature leaf petioles (0.1%), grass blades (0.1%), and suspected invertebrates (< 0.1%). 2. Unclassified leaves (0.3%), epiphytic fern rhizomes (0.1%), flower buds (0.1%), and mosses or lichens (< 0.1%). 3. Unidentified fruit (11.4%), coniferous needles (0.7%), and coniferous cones (0.7%). 4. Young bamboo shoots (1.8%), unidentified fruit (0.7%), fern furled tops (0.3%), herb flowers (0.2%), and mushrooms (0.1%). 5. Herb flowers (0.4%), young bamboo shoots (0.1%), hard and/or woody underground storage organs (0.1%), and rock licking (< 0.1%).

for F troop (Table 4), with Cotoneaster frigidus and Zanthoxylum nepalense the primary

species exploited. Unripe fruit, particularly Caragana gerardiana legume seeds (husks

discarded), were frequently consumed, as were an assortment of fleshy ripe fruits.

Perhaps the most striking aspect of the fall diet, however, was the inclusion of soft

underground storage organs, herb fruits, and herb leaves. Potatoes (Solanum tuberosum)

from cultivated fields were used extensively, as were radishes (Raphanus sativus).

Food selection

Positive correlations between consumption and abundance were strongest (p <

0.001) for deciduous mature leaves (Table 5). Significant positive correlations (p < 0.05) were also found for leaf buds, flowers, deciduous young leaves, and ripe fruit. No significant positive correlations were detected for coniferous needles, unripe fruit, or evergreen mature leaves. Coniferous cones and flower buds were not exploited during any phenological sampling period, and evergreen young leaves were not available during observations of F troop.

We found that evergreen mature leaf consumption correlated negatively with flower consumption and abundance, herb leaf consumption, deciduous young leaf abundance, and total vegetation abundance (Table 6). Bark feeding correlated negatively with deciduous mature leaf consumption and abundance, unripe fruit consumption and abundance, ripe fruit abundance, soft underground storage organ consumption, herb fruit consumption, evergreen mature leaf abundance, flower abundance, and total vegetation abundance.

Table 5. Spearman rank correlation coefficients comparing consumption and abundance of plant part groups. * = significant at the 0.05 level, ** = significant at the 0.01 level. See text for details.

Consumption

deciduous deciduous evergreen leaf ripe unripe coniferous Abundance mature young mature flowers bark buds fruit fruit needles leaves leaves leaves deciduous mature leaves **0.98 *-0.67 -0.15 **-0.74 0.42 *0.73 -0.11 -0.2 **-0.86 deciduous young leaves -0.22 *0.69 *-0.63 -0.23 **-0.78 -0.2 **0.9 -0.05 0.12 evergreen mature leaves *0.64 -0.09 **-0.74 **-0.80 -0.12 **0.86 0.35 -0.26 *-0.71 leaf buds **-0.95 *0.66 0.01 *0.72 -0.49 -0.58 0.1 0.2 **0.82 ripe fruit **0.77 *-0.67 0.01 -0.51 *0.63 **0.78 -0.46 -0.25 **-0.77 unripe fruit *0.86 -0.35 -0.38 **-0.83 0.09 0.58 0.25 -0.23 **-0.8 flowers 0.44 0.33**-0.79 **-0.84 -0.32 0.42 *0.70 -0.36 *-0.62 flower buds **-0.99 *0.68 0.17 *0.72 -0.45 **-0.74 0.14 0.25 **0.88 coniferous needles 0.03 *0.69 **-0.87 -0.59 -0.53 0.2 **0.9 -0.36 -0.3 coniferous cones *-0.63 0.08 *0.67 **0.84 -0.09 **-0.91 -0.33 0.37 **0.74 overall vegetation 0.58 0.08 *-0.73 **-0.85 -0.27 0.58 0.59 -0.3 *-0.65 41

Table 6. Spearman rank correlation coefficients between evergreen mature leaf and bark consumption and the consumption of other plant part groups. * = significant at the 0.05 level, ** = significant at the 0.01 level.

Evergreen Bark Consumption mature leaf consumption consumption deciduous mature leaves -0.17 **-0.90 deciduous young leaves -0.47 0.36 evergreen mature leaves --- 0.40 leaf buds 0.52 **0.84 ripe fruit 0.54 -0.41 unripe fruit -0.42 **-0.74 flowers *-0.63 -0.02 coniferous needles 0.32 0.50 bark 0.40 --- soft underground storage organs -0.28 **-0.76 herbaceous fruit -0.49 **-0.87 herbaceous leaves *-0.66 -0.14 miscellaneous underground resources *0.63 **0.78

Daily path lengths

Using mean values for each month (n = 9), the average daily path length for

Himalayan langurs was 1.50 ± 1.00 km. Daily paths lengths differed significantly among

seasons (Kruskal-Wallis, p < 0.001), with the longest in winter and the shortest in

monsoon and spring (Figure 8). Fall values were intermediate between those seasons and

winter.

Daily path length was positively related to consumption of soft underground

storage organs, unripe fruit, ripe fruit, deciduous mature leaves, and herbaceous fruit in

the stepwise regression model (Table 7). Conversely, daily path length was negatively

related to the consumption of deciduous young leaves, flowers, and bark. Overall

vegetation abundance contributed significantly to all of the above models except ripe

fruit, flowers, and deciduous young leaves. Abundance was the only independent

variable included in the models for evergreen mature leaves, leaf buds, herb leaves, and

miscellaneous underground resources.

DISCUSSION

Semnopithecus entellus (sensu Brandon-Jones and others 2004) foraging behavior

has been the subject of at least ten long-term studies from a minimum of eight sites

(Table 8). Data from these studies substantiates their reputation as generalist feeders.

While these studies represent a wide range of habitats from Sri Lanka to the Himalayas,

the overall contribution of primary food types differs surprisingly little; leaf parts range

from 45 to 60% of the diet. Supplemental and fallback foods are more variable. Langurs

Figure 8. Seasonal daily paths for Himalayan langurs at Langtang. Median with the 10th and 90th percentiles and error bars. Outliers in solid circles.

2 Path length (km)

Winter Spring Monsoon Fall (F troop) (B troop)

Season

Table 7. Stepwise multiple regressions: % daily records on each plant part (log transformed dependent variable) on overall vegetation abundance and daily path length for those days (n = 76). Only significant predictors are given.

Unstandardized Standardized Model Dependent variable R square Predictors coefficient coefficient significance b beta constant -13.278 deciduous mature leaves 0.536 abundance 0.052 0.657 0.000 path length 2.607 0.259 constant -1.652 deciduous young leaves 0.120 0.002 path length -2.916 -0.346 constant 1.691 evergreen mature leaves 0.300 0.000 abundance -0.036 -0.548 constant 6.703 leaf buds 0.547 0.000 abundance -0.055 -0.739 constant -4.899 ripe fruit 0.118 0.002 path length 3.230 0.344 constant -14.292 unripe fruit 0.575 abundance 0.045 0.615 0.000 path length 3.503 0.380 constant -1.719 flowers 0.077 0.016 path length -2.345 -0.277 constant -10.723 soft underground storage organs 0.315 path length 4.097 0.468 0.000 abundance 0.018 0.262 constant 3.807 bark 0.259 abundance -0.028 -0.416 0.000 path length -2.104 -0.249 constant -8.028 herbaceous leaves 0.127 0.002 abundance 0.021 0.357 constant -13.645 herbaceous fruit 0.627 abundance 0.047 0.744 0.000 path length 1.612 0.200 constant 1.775 0.000 miscellaneous underground resources 0.337 abundance -0.035 -0.580 45

Table 8. Comparison of Semnopithecus entellus foraging behavior from long-term (≥ 6 months) field sites. “Feed” = percentage of diurnal activity devoted to feeding, L = all leaves, ML = mature leaves, YL = young leaves, LB = leaf buds, FR = fruit and seeds, FL = flowers.

Avg. path Site Habitat Feed (%) L ML YL LB FR FL Supplemental Sourcesa length (m) Polonnaruwa (Sri Lanka) Semi-deciduous tropical 48 21 27 45 7 earth, insects 1 Dharwar (India) Dry deciduous tropical 44b 60-1300c > 54 > 6 13 stalks, bark, insects 2 Kahna (India) Moist deciduous 26 1083 49 35 4 11 24 10 insects, gum 3 Singur (India) Village, agricultural 29 54 37 5 provisioned foods 4 Jodhpur (India) Village, semidesert 24 67 39 28 23 7 provisioned foods 5 Ramnagar (Nepal) Semi-evergreen sal 34 58 47 14 20 8 insects 6 Junbesi (Nepal) Himalayan 39 1179d >45 >31 >14 >1 crops, USOs, bark 7 Langtang (Nepal) Himalayan 40 1497 57 25e 12f 19 22g 7h crops, USOs, bark 8

a. 1. Hladik, 1977 (Figure 4 and text pp. 337-8). 2. Yoshiba, 1967 (Table 5 and text pp. 136 and 140). 3. Newton, 1992 (Tables I, II and III). 4. Oppenheimer, 1978 (text p. 337). 5. Srivastava, 1989, cited in Newton, 1992 (Table VI); feeding % from Winkler, 1988. 6. Koenig and Borries, 2001 (Table 1); here averaged from Podzuweit, 1994, Chalise, 1995, and Nikolei, unpublished data. 7. Curtin, 1975 (Table 21 and text p. 61), Curtin, 1982 (Table III). 8. This study. b. Based on 10 days of focal sampling. c. Range. Mode listed as 300-700 meters. d. Mean of four 3-month averages from January to December 1973. For reasons given in Curtin (1975, 1982), it is likely an underestimate of path length. e. Includes mature broadleaves, broadleaf petioles, coniferous needles, and unidentified leaves. f. Includes young broadleaves and herbaceous leaves. g. Includes fruit from woody species as well as herbaceous fruit. h. Includes flowers from woody species as well as herbaceous flowers. 46

include insects as a primary supplemental resource at lowland sites, but fallback foods in

the Himalayas are underground storage organs and bark.

In a recent review, Koenig and Borries (2001:125) note the positive correlation

between consumption and abundance for young leaves, flowers, and fruit in lowland gray

langur populations, and suggest they “feed on everything that is available except mature leaves.” The current study fits this pattern if evergreen mature leaves specifically, and several other resources, are substituted as the fallback foods. Himalayan langurs broaden the feeding repertoire of gray langurs by inhabiting an environment so marginal that

deciduous mature leaves are ingested whenever they are available. The ability to subsist

at least seasonally on non-preferred foods is likely one reason for the expansive

geographical and ecological range of gray langurs, including decidedly marginal habitats

such as the Himalayas. Ecological generalism and diversity in feeding techniques characterizes numerous wide-ranging primates, including howler monkeys and macaques

(Glander 1981; Nakayama, Matsuoka, and Watanuki 1999). Given these observations, recent statements that apes possess greater foraging flexibility than monkeys, or that this flexibility allows apes to inhabit environments where monkeys cannot live (Byrne 2001), should be viewed critically.

Most colobines, while having diverse diets, favor young leaves and/or seeds or whole fruits over mature leaves, and this is often related to generalizations concerning the chemical attributes of the plant parts in question (Kirkpatrick 1999; Waterman and Kool

1994). Himalayan langurs clearly prefer broad-leaved deciduous leaves (both mature and

young) to evergreen mature leaves, as the former are taken in close relation to their

48 abundance while the latter are not. In colobine dietary studies, the distinction between evergreen and deciduous broad-leaves is often not made, or is not singled out as a factor in diet selection. Oates (1977), however, noted high selection ratios for certain deciduous species eaten by Colobus guereza and suggested it is related to greater amounts of young leaves, retained over a longer period, than in evergreen species (Oates 1977). There may be a more general rule at play, here, however, as colonizer plant species (e.g., deciduous woody plants in this study), may devote less of their resources to the production of secondary compounds or other antifeedants than non-colonizers (Cates and Orians 1975;

Marsh 1981).

For several colobines, evidence has been provided to argue that seeds are selected more strongly than any other food type, including young leaves (Colobus satanus,

McKey and others 1981; C. polykomos, Dasilva 1994). Making generalizations about plant part quality in the absence of nutritional data, however, can prove problematic

(Schülke, Chalise, and Koenig 2006). In Himalayan langurs, fruits and seeds are clearly important seasonal foods. Given, however, that deciduous leaves, ripe fruits, flowers and leaf buds peaked in availability at different times of year, and all were taken in relation to their abundance, it is difficult to argue for gross preferences of one type over another.

Himalayan langur daily path lengths varied considerably over the course of the study, and we found this to be related to season and the proportion of certain foods that were consumed. When controlling for overall vegetation abundance, the langurs traveled longer distances on days when soft underground storage organs, fruits, and deciduous mature leaves were being consumed at higher rates.

Daily paths were shorter when deciduous young leaves, flowers, and bark were being exploited. Deciduous young leaves are abundant in spring, and we often observed the langurs during this season spending the entire day feeding from trees and shrubs within a single gully. The negative relationship between flower consumption and daily path is likely related to the spring exploitation of Zanthoxylum nepalense flowers, which were taken along with the young leaves of this species. These findings accord with

Curtin’s (1975) observation that Himalayan langurs at Junbesi, Nepal traveled further in the winter when meadow feeding on fruits, particularly Cotoneaster microphyllus, was especially important. Similar relationships between diet and ranging patterns have been noted for lowland gray langurs and several other Asian colobines (Kirkpatrick 2007).

Semnopithecus entellus at Kahna Tiger Reserve, India, had smaller ranges and traveled less when banqueting on mature leaves (Newton 1992). Food availability was also found to influence the day range of gray langurs in the Aravalli Hills (Chhangani and Mohnot

2006b). Travel distances in capped langurs (Trachypithecus pileatus) in Madhupar

National Park, Bangladesh are positively related to fruit consumption and negatively related to mature leaf feeding (Stanford 1991). Similar trends have been noted in banded

(Presbytis melalophos) and maroon leaf monkeys (P. rubicunda) (Bennett 1986; Davies

1984, cited in Kirkpatrick, 2007).

As Newton (1992) has noted, however, there is a danger in making broad generalizations about the consumption of leaves and its influence on primate socioecology. The common practice of labeling leaves as a ubiquitous or non-patchy resource is one example. Although deciduous mature leaves are relatively abundant

50 during monsoon and early fall at Langtang, they are increasingly less available in the months following. Indeed, it could be argued that in late fall and winter deciduous mature leaves are a resource that is more patchily distributed than ripe fruits at many subtropical or tropical primate field sites. This may account for why, when overall vegetation abundance is controlled for, deciduous mature leaf consumption is actually positively related to daily path length in Himalayan langurs.

Thus, while using broadly defined plant categories as a correlate to ranging behavior or competitive regime is a necessary step in first-generation models of primate socioecology (Wrangham 1980; Isbell 1991; Sterck et al. 1997), future work will need to incorporate the idea that under certain circumstances many foodstuffs, even leaf parts, can be a rare resource (e.g., young leaves for Procolobus badius, Snaith and Chapman

2005; mature leaves for Colobus satanas, McKey and Waterman 1982). This can lead to activity budgets not dissimilar to non-folivores. Indeed, in Himalayan and other gray langurs, the amount of time devoted to feeding and travel, and the distance traveled in a given day (this study and Table 8), overlaps those of highly frugivorous spider monkeys

(Suarez 2006a). The stereotype of the lazy leaf-eater must be applied with caution.

CHAPTER IV

EXTRACTIVE FORAGING AND PRIMATE INTELLIGENCE

INTRODUCTION

Debate has long surrounded the evolutionary origins of primate intelligence

(Tomasello and Call 1997). Compared with other mammals, primates are known for relatively large brains or neocortices (Barton 1996; Bush and Allman 2004; Jerison 1973;

Karlen and Krubitzer 2006), and much work has involved linking primate brain evolution to proposed measures of ecological or social complexity. Several models concerning the ultimate causation of primate intelligence (broadly construed) have been proposed, and all have received some support and suffer from various deficiencies (Table 9). While these hypotheses have been depicted as mutually exclusive, recent work has incorporated both social and ecological factors from multiple models (Walker and others 2006).

Parker and Gibson (1977) and Gibson (1986) noted that some primates can envision the presence of hidden resources and develop strategies to exploit them, and in turn relate this to primate brain evolution (Gibson 1986; Parker and Gibson 1977).

Called extractive foraging, it involves embedded foods which may require complex manipulation to harvest. Examples include foods removed from casings, such as seeds, eggs, or vertebrate flesh, underground storage organs dug from the ground, and pith or invertebrates that are removed from wood or soil. Primates utilizing more complicated extractive techniques are expected to score higher on various measures of brain size or 51

Table 9. Ultimate causation models of primate brain evolution.

Model Description and Predictions Principal Support Notes Social Group living requires complex Within primates, there is a 1) The positive relationship between social brain mental coordination in individuals positive relationship variables and neocortex ratio may not apply to to achieve reproductive success, far between neocortex ratio, prosimians when examined alone (Barton 1996). Jolly beyond the capacities required for nonvisual neocortex ratio, 2) Group size estimates may be incorrect. In (1966) foraging or other ecological and other brain ratios and particular, prosimian group sizes are likely problems. In short, the more estimates of mean group underestimated (e.g., Nekaris and Bearder 2007). Humphrey individuals that are regularly size, grooming clique size, In addition, current methods of quantifying the (1976) interacted with, the greater the and frequencies of play and average social network of a given primate do not selective premium on other social behavior. include individuals known to others through Byrne and “intelligence.” Living in social (reviewed in Dunbar 2003; olfaction or other non-visual or non-tactile senses. Whiten groups thus influenced the Walker et al. 2006). 3) Other social variables may likewise be (1988) evolution of primate brain size or Estimated rates of “tactical incorrectly estimated. For example, there may be structure. Primates living in larger deception” are also a bias in the study or reporting of “tactical Cheney groups, or having larger values on positively related to deception” among differing primate taxa. The and other social measures (e.g., rates of neocortex size (Byrne and methodology underlying the identification of Seyfarth tactical deception) are expected to Corp 2004). deception and other facets of social cognition has (1990) score higher on measures of been questioned (Kummer et al. 1990). relative brain size or complexity. 4) Captive studies suggest that social complexity Dunbar Note that the formation of larger is not a linear function of group size (Kummer (1995) groups in itself may be related to 1975). ecological factors, such as predator 5) The sample used to test the model is biased avoidance. towards primates showing the “cercopithecoid pattern” of social grouping. In addition, apes in particular deviate from model predictions, the most exceptional example being Pongo (Rodman 1999; but see Dunbar 2003). Frugivory Common foods in primate diets, Within primates, relative 1) The relationship between assorted brain ratios and such as fruit, are rare temporally brain size positively and frugivory or home range size has been cognitive and spatially in the foraging associates with percent contested (Walker et al. 2006). mapping environment. Long-term memory frugivory and home range 2) The percent frugivory or home range size of a and cognitive mapping is at a size (Clutton-Brock and given primate may not be linearly related to Clutton- premium for the location of such Harvey 1980). A positive ecological complexity. There is little a priori Brock and foods, and as such drives brain relationship between evidence for the common assumption that fruit 52

Model Description and Predictions Principal Support Notes Harvey evolution. Primates with higher frugivory and relative exploitation is more cognitively demanding than (1980) percentages of fruit in the diet, or neocortex size has also been foraging on, for example, rare leaves or insects. having larger day or home ranges reported (anthropoids, 3) Does the complexity of primate food Milton (which in general are positively Sawaguchi 1992; acquisition quantitatively differ from that of non- (1981) correlated with fruit intake) are haplorrhines, Barton 1996). primates? expected to score higher in terms of See also Allman (1999). relative brain size or complexity. Extractive Some primate foods are hidden or Relationship between 1) Dunbar (1995) found that the relationship foraging embedded and require complex categorization of forager between neocortex ratio and foraging category sensorimotor skills to extract, and relative brain size and was dependent on two data points, Homo and Pan. Parker and driving primate brain evolution. “neocortical progression When these were removed from the sample, Gibson Primates using more complex index” (neocortex size significant differences disappeared (see text). (1977). sensorimotor coordinations are compared to that predicted 2) Is extractive foraging in primates qualitatively expected to score higher in terms of for an insectivore of similar different from extractive foraging in non-primates relative brain size or complexity body weight) (Gibson (King 1986)? (see text and Table II). 1986). All 1) Intelligence is not unidimensional and the ultimate separation of ultimate causation variables (i.e., causation social versus ecological) is problematic (Menzel models 1997; Tomasello and Call 1997). 2) The brain scaling method utilized influences which models are supported. All previous tests may be inconclusive (Deaner et al. 2000). 3) There is disagreement on the degree to which natural selection can act on individual parts of the brain (e.g., Finlay and Darlington 1995). 4) Neocortex or other brain ratios may be too coarse a measure. Differences in brain microstructure may be more relevant (Holloway 1966). 53

complexity. It is also argued that extractive primate omnivores, such as capuchins, are

able to inhabit marginal habitats that cannot be exploited by most other members of the

Order (Gibson 1986).

Gibson (1986) proposed a classification of primate foraging, with categories

differentiated by the complexity of stimuli utilized (Table 10). Skilled extractive foragers

(category names after Dunbar 1995) (Dunbar 1995) employ the most complex foraging

techniques, such as tool-aided extraction or other methods involving three or more

sensorimotor tasks (e.g., “tapping, probing, looking, and listening,” Gibson 1986:99).

Unskilled extractive foragers utilize less complicated extractive measures, such as removing seeds and turning over objects, and specialized extractive foragers focus on

one hidden food for which they have anatomical adaptations. Non-extractive foragers

feed on visually exposed foods. Gibson argued that relative brain and neocortex size is

largest in skilled extractive foragers and smallest in specialized and non-extractive

foragers (Table 9).

Dunbar (1995) expanded the number of genera considered non-extractive,

including all data-sufficient colobines (Table 10), and then tested for differences in

neocortex ratio between primates in the four categories. Significant differences were

detected, but there was evidence that this was related mainly to differences between

skilled extractive foragers, especially Homo and Pan, and the other three categories.

Among anthropoids, significant differences were no longer detected when Homo and Pan

were removed from the sample. Dunbar concluded that no relationship exists between

Table 10. Gibson’s (1986) classification of primate foragers; category names follow Dunbar (1995). The primates included in each category are the same in the Gibson and Dunbar papers unless noted otherwise. Gibson does not give specific in-text examples of non-extractive primate foragers; it is assumed here that all taxa appearing on her tables and not in her extractive categories can be considered non-extractive.

Skilled extractive Specialized extractive

Cebus Callithrix Daubentonia Gorilla Pan Homo

Unskilled extractive Non-extractive

Ateles1 Gibson (1986) Dunbar (1995) Miopithecus2 Colobus Colobine monkeys3 Lagothrix Prosimii except Daubentonia Prosimii except Daubentonia Macaca Alouatta Alouatta Papio Aotus Cebuella Pongo Cercopithecus Cercocebus Saimiri Cercopithecus Erythrocebus 1. Not listed in Dunbar (1995) Hylobates 2. Included in non-extractive by Dunbar (1995) Miopithecus Pithecia Saguinus

3. Nasalis, Presbytis, Procolobus, and Pygathrix 55

extractive foraging and neocortex ratio in primates generally, or within prosimians or

anthropoids when considered separately (see also Dunbar 1992).

The inclusion of colobine monkeys in the “non-extractive” category is not exceptional, as they are often depicted as obligate folivores with little use for such behavior (King 1986). Since there are few studies that have quantitatively measured extractive foraging in any primate, this conclusion may be premature (van Schaik,

Deaner, and Merrill 1999).

Here, I describe extractive foraging in Himalayan gray langurs. I also test the hypothesis that extractive foraging provides a seasonal fall-back for the langurs during

periods of resource scarcity, as such behavior has been predicted to be especially important for survival in such marginal habitats (Gibson 1986). Lastly, I review the colobine dietary literature to assess the frequency of extractive foraging in other populations of Semnopithecus entellus and other species within the subfamily.

METHODS

Data collection

All-day group follows (sleeping site to sleeping site) were conducted for a period of 13 months between January 2003 and February 2004, following a pilot study in

March-April 2000. Data were collected on F-troop for approximately 775 hours over 10

months including January 2003 and between March 2003 and February 2004. For three

months during the monsoon of 2003 (June to August), F-troop could not be contacted.

During that time, B-troop was followed for a total of 292 hours. Data are presented separately for the two groups.

Data on food species and plant part (or other foods) ingested were recorded for each individual that was in view during group scans taken at 20-minute intervals

(Altmann 1974). At the beginning of each 20-minute period, the activity state (feed, travel, rest, huddle, cling, groom, play, or social) was recorded for each visible individual at the moment of first observation. Feeding was defined as reaching for, holding of, or mastication of food. For each individual that was feeding, the species and plant part being consumed were also recorded.

The manner in which individual food types were collected by the monkeys (e.g., hidden versus exposed target) was noted ad libitum when they were first observed being eaten. Most items relevant to this paper were eaten more than once, allowing for confirmation concerning normal mode of acquisition. Based on this information, food types were sorted post-hoc into two resource classes: extracted and non-extracted (after

Gibson 1986; Tomasello and Call 1997). A given food type (e.g., species X mature leaf) is thus considered here as always extracted or always non-extracted. This is an acceptable assumption for the vast majority of Himalayan langur food items. All resources classified as “extracted” were further subdivided into more specific categories and actions pertaining to mode of acquisition. Categories (and actions, in parentheses) include: 1) Removing plant cover (removing fruit casing or peeling husk), 2) Excavation

(digging or surface scratching), 3) Prying or Picking (removing bark to reach objects

58 underneath), and 4) searching under obstacles (probing under rocks to remove objects underneath).

Temporal variation in food abundance was estimated by phenological sampling conducted in nine plots (0.75 ha total) during the 2003 calendar year (Chapter 3).

Data analysis

Since data were collected from F troop and B troop during two temporally discrete periods, they are presented separately unless indicated otherwise, and all statistical tests were performed on F troop data only. For F troop, averages for extractive foraging categories or actions are presented based on a nine-month period (March-May

2003, September 2003-February 2004). B troop data are presented as a monsoon, 4- month (June-September 2003) average. All data reflect pooled observations of all members of a particular troop.

Spearman correlations assessed the relationship between extractive foraging in

Himalayan langurs and the abundance or consumption of plant part classes which constituted greater than 1% of feeding records. These plant part classes include evergreen mature leaves, deciduous young and mature leaves, leaf buds, unripe and ripe fruits, and flowers. Extractive foraging was expressed as the mean percentage of individuals consuming extracted foods over all feeding records. Excavation and seed consumption were also expressed as mean percentage over all feeding records, as was plant part consumption. Plant part abundance was expressed as abundance units per plant part class, and overall vegetation abundance was expressed as the sum of all plant parts

59 from all plant part classes. All of the above were entered for given sampling periods, with a sampling period defined as the day of phenological sampling and the two-week period following it. All data from outside sampling periods were excluded from the correlation analysis.

Published data on African and Asian colobines were used to test for differences in seed eating among species from the two continents using a Mann-Whitney U test, with percentages of the diet comprising seeds for individual studies as the test variable.

All tests were two-tailed with significance set at 0.05. Statistical analyses were performed in SPSS 13.0.

RESULTS

Frequency and classes of extractive foraging in Himalayan langurs

Extracted foods, including a minimum of 10 plant genera, averaged 15.1% ±

14.5% of the F troop monthly diet (n = 9, range = 0.0 to 39.6%), and 16.0% ± 12.2% of B troop monsoon diet (n = 4, range = 8.2 to 34.0%) (Table 11, Figure 9). For F troop, the most frequent targets were seeds (7.3% of overall diet) and soft underground storage organs (6.1%), but extractive foraging was also utilized to harvest young bamboo shoots, hard underground storage organs, unidentified small objects and, on one occasion, presumed invertebrates (Table 11, Figures 10a, 11 and 12). Extractive foraging occurred regularly in all seasons except spring (April and May), when deciduous young leaves made up 46% of the diet. Only 1.6% of spring diet was extractive, and all such records were confined to April (Figure 9). Monsoon extractive foraging in B troop was limited to

Table 11. Categories of extractive foraging for Himalayan langurs at Langtang National Park, Nepal. USOs = underground storage organs, W = winter, S = spring, M = monsoon, F = fall. A specific action is included in a season if it averaged over 1% of feeding records for that season in either troop.

Nine-month % Monsoon % Category Action(s) Target Season(s) F troop B troop

removing fruit casing seeds W, M, F 7.31 14.22 Removing plant cover 3 peeling or stripping husk young bamboo M < 0.1 1.8 digging USOs W, F 6.14 ---- Excavation 5 surface scratching soft USOs, small objects W, S 1.7 ---- Prying and picking removing bark and wood suspected invertebrates ---- < 0.1 ---- Searching under obstacles6 probing under rocks suspected invertebrates ------sum 15.1 16.0

Notes: 1. Caragana gerardiana. The outer portion of Sorbus cuspidata fruit (2.3% of overall diet) was sometimes discarded; this was categorized as a “non- extracted” resource. 2. Quercus semecarpofolia (11.8%) and Caragana gerardiana (2.5%). 3. Arundinaria maling. 4. Soft USOs of Solanum tuberosum (2.6%), Saussurea sp. (1.7%), Raphanus sativus (0.6%, see below), Rumex sp. (< 0.1%), and unidentified forms (0.4%). Hard or woody USOs of Aconogonum molle (0.3%), Berberis aristata (0.1%), Caragana gerardiana (< 0.1%), and Elsholtzia fruticosa (< 0.1%). Unidentified USOs not categorized as soft or hard contributed an additional 0.3%. Raphanus sativus was acquired both by digging and by consuming exposed pieces placed by locals on rocks to dry; the percentage above includes all Raphanus sativus records. 5. Percentage includes only foods classified as “small objects.” 6. Observed in B troop during pilot project. 60

Figure 9. Extracted foods in Himalayan langurs from January 2003 to February 2004. B

= B troop; all other months represent F troop.

45 40 small objects 35 30 underground storage 25 organs 20 bamboo

% of diet % of 15 10 seeds 5 0 J 04 J 03 F 04 S 03 N 03 D 03 A 03 O 03 M 03 M 03 JB 03 JB 03 A 03 B S 03 B Month

Figure 10. Percentage contribution of food types to extractive foraging in: a) F troop

(nine-month average), and b) B troop (monsoon average). USOs = underground storage organs.

(a)

suspected small objects invertebrates 11% 0%

seeds USOs 49% 40%

bamboo 0%

(b)

bamboo 11%

seeds 89%

Figure 11. Juvenile langur digging.

Figure 12. Adult female with the underground storage organ of Saussurea sp., a naturally-occurring herbaceous plant.

65 seeds (14.2% of monsoon diet) and young bamboo (1.6%) (Table 11, Figure 10b).

The monkeys were never observed to use tools (Beck 1980) in the context of foraging.

Including pilot project data in addition to the main study period, Himalayan langurs utilized six specific extractive actions, divided here into four general categories (Table 11).

I. Removing plant cover. The seeds of Caragana gerardiana (Leguminosae) and Quercus semecarpofolia (Fagaceae) were extracted using hands and teeth.

Bamboo was extracted by stripping the sheath leaves of young Arundinaria maling

(Gramineae) with teeth and/or hands.

II. Excavation. Soft underground storage organs from cultivated and naturally occurring herbaceous plants were extracted via digging, pulled out, and eaten by hand. Hard underground storage organs, generally woody roots, were also exposed by digging and either twisted out entirely and eaten by hand, or bitten off and ingested in situ. Small food items lying near the surface were excavated by scratching the soil with the fingers. Many of the “small objects” were unidentifiable, but others were identified as tuber fragments or seeds lying just beneath the surface.

III. Prying and picking to locate hidden food was noted on only one occasion in March 2003. An adult male was observed manipulating bark and dead wood and eating objects within; invertebrates were the suspected target of this behavior (see

Struhsaker 1978).

IV. Searching under obstacles was noted in B-troop during the pilot project.

On three separate days, individuals gathered on the banks of Langtang Khola and

were observed searching under rocks and in crevices between rocks and pulling out objects, presumed invertebrates, and eating them. All non-infant age-sex classes

participated, and the three incidents lasted approximately 45, 40, and 50 minutes.

Correlations between plant part abundance/consumption and extractive foraging in

Himalayan langurs

Correlations between feeding and plant part abundance (assessed during phenology samples) and consumption suggest that extractive foraging in general, and specific distinctions at the category or action level, were somewhat seasonal (Table

12). Discussed here are the specific action of seed consumption and the general category of excavation, both of which are implicated in over 5.0% of F troop feeding records.

Extractive foraging in general was not related to overall vegetation abundance, but was negatively related to deciduous young leaf consumption, and positively related to the consumption and/or abundance of fruits and mature leaves

(Table 12). Seed eating was synonymous with consumption of plant parts classified as “unripe fruits,” negatively related to leaf bud consumption, and positively related to ripe fruit abundance and the abundance and/or consumption of mature leaves.

Excavation was negatively correlated with deciduous young leaf consumption.

Table 12. Correlations between extractive foraging and phenology scores. Spearman rank correlation coefficients between extractive foraging, seed consumption, and excavation with the abundance and consumption of plant part groups. * = significant at the 0.05 level, ** = significant at the 0.01 level.

Extractive Seeds Excavation Overall abundance 0.41 0.58 0.25

Evergreen mature leaf abundance *0.72 *0.86 0.35 consumption -0.22 -0.42 0.05 Deciduous mature leaf abundance **0.74 *0.73 0.52 consumption **0.77 *0.72 0.57 Deciduous young leaf abundance -0.41 -0.20 -0.28 consumption *-0.67 -0.40 *-0.67 Leaf bud abundance -0.60 -0.58 -0.46 consumption -0.48 *-0.71 -0.14 Ripe fruit abundance **0.78 **0.78 0.36 consumption 0.30 0.30 -0.02 Unripe fruit abundance 0.55 0.58 0.48 consumption **0.89 **1.00 0.36 Flower abundance 0.17 0.57 0.00 consumption -0.35 -0.01 -0.48

Extractive foraging in other colobines

Based on an extensive literature review, extractive foraging is found in a minimum of 25 colobine species (Table 13). Seeds are the most common embedded food

item, representing 15% of colobine diet over 42 studies where percentage was estimated.

Considerable variation was found, however, with seeds ranging from less than 1% of the

diet in Rhinopithecus bieti (n = 1) to 51%, on average, in Colobus satanas (n = 2). There

was no significant difference in seed exploitation for African versus Asian colobines

(Mann-Whitney U, p = 0.61).

Other extracted or potentially extracted foods were noted in studies of 14 colobine

species (Table 13). Of resources that by definition or description could be considered

hidden, pith is the most common (6-7 species, 11-12 sites), followed by presumed or

identified invertebrates (3 species, 6 sites) and underground storage organs (1 species, 2

sites minimum, see Table 13). Other extracted foods, each reported from only one site,

include soil located under the organic layer, peeled fruits, gum, vertebrate flesh, and eggs.

Non-seed extractive foods are almost invariably small components of the overall diet.

DISCUSSION

Colobine monkeys are clearly capable extractive foragers, and for a number of years there was a prominent revisionist view that categorized these monkeys as “seed- eaters” as opposed to “leaf-eaters.” Current thought simply stresses that colobines have

an eclectic diet, and thus considerable interspecific and intraspecific diversity

(Kirkpatrick 1999; Kirkpatrick 2007). Nonetheless, it is important to point out that while

Table 13. Extractive foraging in colobine monkeys. Extractive foods include those that by definition or author description meet Gibson’s (1986) criteria. Potentially extractive foods are those that are extractive in other primates, but in which author description in the study of interest is insufficient to determine whether the resource was visually hidden. For example, roots may be aerial and potatoes may be unearthed by humans. Borderline cases are also included in this category. NR = seeds not distinguished from whole fruits, Y = seeds taken but no percentage given, USOs = underground storage organs. Dashed lines (---) indicate studies for which only seed data were available, from a secondary source.

Other Potentially Species and Site Seeds (%) Notes References Extractive Extractive AFRICA

Colobus angolensis

(Angolan colobus) Ituri Forest, Zaire 22 ------1 Nyungwe, Rwanda 20 a 2 Nyungwe, Rwanda <1 3 Salonga, Zaire 50 4

Colobus guereza

(Guereza) Budongo, Uganda 12 ------b 5 Ituri Forest, Zaire 22 ------6 Kakamega, Kenya 1 b 7 Kibale, Uganda ≥1 aquatic plants 8 gum, aquatic Kibale, Uganda >1 v 9 plants

Colobus polykomos

(King colobus) Tiwai, Sierra Leone 32 pith 10

Colobus satanas

(Black colobus) Douala-Edea, Cameroon Y 11 Forêt des Abeilles, 41 pith gum 12 Gabon Lopé, Gabon 60 13

Colobus vellerosus

(White-thighed colobus) Boabeng-Fiema, Ghana NR pith sap c 14

Procolobus

(Piliocolobus) badius

Other Potentially Species and Site Seeds (%) Notes References Extractive Extractive (Red colobus) Abuko, Gambia 3 pith 15 Fathala Forest, Senegal 19 16 suspected Gbanraun, Nigeria 12 pith d 17 invertebrates suspected Gombe, Tanzania Y e 18 invertebrates Jozani, Zanzibar Y 19 Jozani, Zanzibar 1 pith f 20 suspected Kibale, Uganda 1 g 21 invertebrates Korup, Cameroon NR buds h 22 Salonga, Zaire 31 23 Tana River, Kenya 1 24 Tiwai, Sierra Leone 25 25

Procolobus (Procolobus) verus (Olive colobus) Tiwai, Sierra Leone 14 i 26

ASIA

Nasalis larvatus

(Proboscis monkey) Kalimantan, Indonesia ≥20 27 (Borneo) Sabah, Malaysia 7 ------28 (Borneo) Sarawak, Malaysia 15 29 (Borneo)

Presbytis comata

(Javan leaf monkey) Java, Indonesia 1 soil j 30

Presbytis hosei

(Hose's leaf monkey) Sabah, Malaysia 19 b 31 (Borneo)

Presbytis melalophos

(Banded leaf monkey) Kuala Lompat, 8 32 Peninsular Malaysia Kuala Lompat, 25 bamboo pith? k 33 Peninsular Malaysia Perawang, Indonesia 36 soil l 34 (Sumatra)

Other Potentially Species and Site Seeds (%) Notes References Extractive Extractive

Presbytis potenziana

(Mentawai leaf monkey) Muntei, Indonesia 4 35

Presbytis rubicunda

(Maroon leaf monkey) Kalimantan, Indonesia NR fruit m 36 (Borneo) Kuala Lompat, 30 bamboo pith 37 Peninsular Malaysia

Presbytis thomasi

(Thomas' leaf monkey) Bohorok, Indonesia Y n 38 (Sumatra) Ketambe, Indonesia Y o 39 (Sumatra)

Pygathrix nemaeus

(Douc) Nui Chua and Phuoc Y 40 Binh, Vietnam

Rhinopithecus avunculus (Tonkin snub-nosed monkey) Du Gia, Vietnam 5 cambium 41 Tuyen Qiang, Vietnam 15 p 42

Rhinopithecus bieti (Yunnan snub-nosed monkey) Baimaxueshan, China <1 43 invertebrates, Xiaochangdu, China Y roots, resin q 44 vertebrate flesh

Rhinopithecus brelichi (Guizhou snub-nosed monkey) Fanjing, China Y 45

Semnopithecus entellus

(Gray langur) Aravalli Hills, India NR gum 46 insect larvae, Dharwar, India Y r 47 bamboo Gir Sanctuary, India Y pith, eggs roots, sap 48 Jodhpur, India Y roots, sap 49 Junbesi, Nepal NR USOs, suspected s 50

Other Potentially Species and Site Seeds (%) Notes References Extractive Extractive (Himalayan) invertebrates Kanha, India NR gum, invertebrates t 51 USOs, bamboo, Langtang, Nepal 7 suspected 52 (Himalayan) invertebrates Melemchi, Nepal NR potatoes 53 (Himalayan) Rajaji, India 12 pith 54 Ramnagar, Nepal Y pith algae, gum 55 Simla, India (Himalayan) Y pith cambium 56 Singur, India NR potatoes 57

Trachypithecus auratus

(Silver langur) Java, Indonesia ≥7 u 58

Trachypithecus francoisi

(Francois' langur) Fusui, China <1 v 59 Nonggang, China 14 roots 60

Trachypithecus geei

(Golden langur) Kakoijana, India NR bamboo 61

Trachypithecus obscurus

(Dusky langur) Kuala Lompat, 3 62 Peninsular Malaysia

Trachypithecus phayrei

(Phayre's langur) Phu Khieo, Thailand ≥22 bamboo w 63

Trachypithecus pileatus

(Capped langur) Madhupur, Bangladesh Y 64 Madhupur, Bangladesh 9 sap x 65 water lily Pakhui, India Y 66 stalks, gum Notes: a. 300 member supergroup. b. Mean of two groups. c. Five groups studied. d. Galls consumed from Macaranga leaves; females observed “manipulating rotten wood” twice. e. Pulling dead bark and mouthing of undersurface/surface. f. Mean of two habitats. g. During possible arthropod foraging, hang upside-down and mouth the bottom of moss-covered branches, unroll curled mature leaves, search through moss and lichens, search undersurface of mature leaves, pull off bark and mouth undersurface, and probe wood with fingers. h. Manually “open up” buds, presumably to get at inner contents. i. Pentaclethra seeds, a small percentage, eaten only after pod dehiscence. j. Mean of three focal animals. Red soil eaten from small holes. k. Species not distinguished, of the two studied (Presbytis melalophos

and P. rubicunda). l. Mean of seven focal animals over two month-long sampling periods. Ingested gray soil under organic layer at depths of up to 20 cm. m. Fruit of Lansium and Garcinia were peeled and flesh consumed. n. Most observations on two heterosexual troops and one all-male group. Inner portions of mature rubber seeds eaten. o. Three groups studied. p. Based on 34 feeding observations. q. Search for invertebrates in dead wood and under bark and rocks, and search for nuts in leaf litter. Two observations of cannibalism of infants; one accompanied by observation of infanticide. r. Insect larvae in Terminalia tomentosa leaf galls eaten. s. Dig for underground storage organs, and search under moss and lichens on forest floor, possibly for insects. t. Induce increased gum flow by enlarging holes or removing hardened gum with canines, and turn over leaves before finding and consuming caterpillars. u. Data from Group 3. v. Data from 6 study groups. w. Young seeds comprise 22% of diet, based on focal sampling. x. Sap "oozing" from bark (non-extraction implied).

References: 1. Bocian 1997, in Kirkpatrick 1999. 2. Vedder and Fashing 2002. 3. Fimbel et al. 2001. 4. Maisels et al. 1994. 5. Plumptre, in Fashing 2007. 6. Bocian 1997, in Kirkpatrick 1999. 7. Fashing 2001. 8. Oates 1977. 9. Harris 2005, 2006. 10. Dasilva 1994. 11. McKey et al. 1981. 12. Gautier-Hion et al. 1997; Fleury and Gautier-Hion 1999. 13. Harrison 1986; Oates 1994. 14. Wong et al. 2006. 15. Starin 1991. 16. Gatinot 1978; Oates 1994. 17. Werre 2000. 18. Clutton-Brock 1975. 19. Mturi 1993. 20. Siex 2003, 2005; Fashing 2007. 21. Struhsaker 1975, 1978. 22. Usongo and Amubode 2001. 23. Maisels et al. 1994. 24. Marsh 1981. 25. Davies et al. 1999. 26. Oates 1988. 27. Yeager 1989. 28. Boonratana 1993, in Kirkpatrick 1999. 29. Bennett and Sebastian 1988. 30. Ruhiyat 1983. 31. Mitchell 1994. 32. Curtin 1980. 33. Davies et al. 1988. 34. Megantara 1989. 35. Sangchantr 2004. 36. Supriatna et al. 1986. 37. Davies et al. 1988; Davies 1991. 38. Gurmaya 1986. 39. Ungar 1995. 40. Hoang and Baxter 2006. 41. Wright et al. 2006. 42. Boonratana and Le 1998. 43. Kirkpatrick 1996. 44. Xiang and Grueter 2007; Xiang et al. 2007. 45. Bleisch and Xie 1998. 46. Chhangani and Mohnot 2006. 47. Sugiyama 1964; Yoshiba 1967. 48. Rahaman 1973; Starin 1973, in Oppenheimer 1977. 49. Mohnot, in Oppenheimer 1977. 50. Curtin 1975. 51. Newton 1992. 52. This study. 53. Bishop, in Oppenheimer 1977. 54. Kar-Gupta and Kumar 1994. 55. Chalise 1995; Schülke et al. 2006. 56. Sugiyama 1976. 57. Oppenheimer 1977, 1978. 58. Kool 1993. 59. Li and Rogers 2006. 60. Zhou et al. 2006. 61. Biswas and Bhattacharjee 2004. 62. Curtin 1980. 63. Suarez 2005, 2006a, 2006b. 64. Islam and Husain 1982. 65. Stanford 1991. 66. Kumar and Solanki 2004.

many primatologists are aware of colobine digestive specializations that aid digestion of

high-fiber foods like leaves (Bauchop and Martucci 1968), much less attention has been

directed towards dental specializations that aid in the destruction of certain types of seeds

(Lucas and Teaford 1994). Some colobine monkeys would be most aptly described as

specialized extractive foragers (Gibson 1986; Dunbar 1995). However, the data set

reviewed here suggests that at least several species possess the capacity for more diverse

modes.

The other labels, unskilled and skilled extractive foragers (Gibson 1986; Dunbar

1995), are differentiated by the complexity of the stimuli confronted while foraging for

multiple types of hidden food. Consider the classic study of red colobus (Procolobus

[Piliocolobus] badius) at Kibale, Uganda (Struhsaker 1975). In addition to a small

dietary contribution of seeds, red colobus were observed to engage in probable

“arthropod foraging,” a small (6%) but perhaps nutritionally important component of the

diet (Struhsaker 1978). The foraging modes utilized during this behavior were varied,

and included hanging upside-down and mouthing the bottoms of moss-covered branches,

unrolling curled mature leaves, searching through moss and lichens and the bottoms of mature leaves, pulling off bark and mouthing the undersurface (seen also at Gombe,

Tanzania, Clutton-Brock 1975), and probing wood with fingers (Struhsaker 1975). Red

colobus appear capable of unskilled or even skilled extractive foraging.

Gray langurs (Semnopithecus entellus) are also adept at multiple modes of

extractive foraging, as illustrated by the Himalayan langurs at Langtang. Seeds were

extracted from acorns and legume pods, bamboo leaf sheaths were removed from young

shoots, underground storage organs and other foods were located by digging and surface

scratching, and presumed invertebrates were located by picking through wood and

probing under rocks. Although extractive foraging in general occurs throughout the year,

excepting spring, different categories or actions predominate in certain seasons as diet

breadth changes. Seed-eating occurs whenever favored species are available, while

excavation is negatively related to young leaf consumption. Although comparative data

are scarce, this is in agreement with Gibson’s (1986) suggestion that extractive foraging

can be expected to be important in marginal or seasonal habitats, as is the recent finding

of multiple modes of extractive foraging in Rhinopithecus bieti (Xiang and others 2007).

It should be clear that some of the categorizations used in testing the extractive

foraging hypothesis not only ignore examples from the colobines, but other primates as

well, e.g., pitheciins and cercopithecines. Pithecia was categorized by Dunbar (1995) as

a non-extractive forager despite possessing dental specializations for exploiting seeds,

which make up from 26% to 61% of the diet (Kinzey and Norconk 1990; Norconk 2007).

Although considered an “anatomical extractor” by Singleton (2004:312), Pithecia (as well as closely related Chiropotes) engage in other forms of extraction as well, such as the exploitation of pith (Norconk 1996; Peetz 2001). Cercocebus, also categorized as a non-extractive forager (Dunbar 1995), was described by Jolly (2007) as a “forest-floor gleaner” (Jolly 2007). This feeding niche includes extractive actions such as searching through dead wood and leaf litter in order to locate insects and fallen fruit or seeds

(Waser 1984).

Although primates could be re-categorized and the putative relationship between

extractive foraging and various brain ratios freshly tested, some authors have called for a more quantitative accounting of extractive foraging in primate field studies (King 1986;

van Schaik, Deaner, and Merrill 1999). The simplest procedure would be that employed

in this study; namely, categorizing feeding behavior and giving the percentages of

various extractive actions in relation to the overall diet. Although data are sparse, it is

likely that most primates engage in extractive foraging at least occasionally, and that the

complexity of such actions in many species has been underestimated.

CHAPTER V

OPTIMAL FORAGING THEORY: THE CLASSICAL PREY MODEL

INTRODUCTION

Optimal foraging theory (OFT) utilizes mathematical models to predict behavior

and operates on the assumption that feeding behavior has been molded by natural selection (Stephens and Krebs 1986). Although OFT, first developed in the 1960s and

1970s (Charnov and Orians 1973; Emlen 1966; MacArthur and Pianka 1966; Schoener

1971) has been embroiled in controversy (Gray 1987; Perry and Pianka 1997; Pierce and

Ollason 1987), few can deny its impact on behavioral ecology. Indeed, OFT has transformed largely descriptive work on feeding behavior into studies that convert measurable variables such as time and energy intake into quantitative, testable predictions

concerning animal behavior. As a result, the breadth and applicability of its models

continues to grow (Stephens, Brown, and Ydenberg 2007). Secondly, the major

criticisms leveled at OFT likely reflect misunderstanding rather than genuine

disagreement. OFT, for example, does not argue that animals are optimal; rather, it uses

a mathematical tool (optimization) to denote how an animal should behave under

specified conditions (Ydenberg, Brown, and Stephens 2007). Deviations in animal

behavior from that predicted by OFT models directs attention towards novel lines of

research and a more complete understanding of the mechanisms of foraging (Stephens

and Krebs 1986). 77 78

A seminal model in foraging theory is the classical prey model, variously called

the attack, optimal diet, or contingency model, which predicts which foods in a set should

be accepted by a forager under given conditions (Charnov 1976b; MacArthur and Pianka

1966; Schoener 1971). Food types (prey) are rank-ordered by a given currency (often,

but not necessarily, some characterization of energy) divided by the handling time it takes

to consume each item. The higher the ratio, the more “profitable” the food is considered

to be. Food types are then entered into the “prey algorithm,” which includes the variables

of currency, handling time, and encounter rate, in the order of their profitability. As each new food type is entered, the algorithm gives the overall rate of intake if only these types were taken. This set of food types prioritizes prey with the highest overall rate of intake and defines the optimal diet. The model therefore provides a quantitative, sliding

“threshold” of profitability below which foods should not be taken. In addition, this model predicts preference for more profitable food types and increased selectivity as

encounter rates with high-ranking foods increase. Finally, the model predicts that encounter rates with low-ranking prey is unrelated to their inclusion in the diet (Stephens and Krebs 1986). An analogous version of this model exists for patch choice (Schoener

1987; Schoener 1974).

Human behavioral ecologists have applied variants of the classical prey model to modern human hunter-gatherers and, to a lesser extent, the archaeological record

(Kennett and Winterhalder 2006; Winterhalder and Smith 1981). Hawkes and colleagues

(1982), for example, used this approach to predict the caloric profitability threshold for food items to be included in the diet of Aché hunter gatherers (Hawkes, Hill, and

O'Connell 1982; Kaplan and Hill 1992). Kurland and Beckerman (1982), in a similar

vein, utilized the model to argue that selection favored reciprocity and information

exchange in early hominid evolution due to its effects on reducing search costs.

Interestingly, researchers of nonhuman primates have rarely applied OFT to their

subjects, but this is not necessarily due to lack of interest. The data required to test even the most basic foraging models, such as intake rate, may be difficult to gather from primates that are nocturnal, difficult to habituate, or living in high canopy. Even under the best of conditions, the quantitative testing of an optimal foraging model is arduous, and most primatologists who reference foraging theory use it as an a posteriori tool to explain observed behavior. Hamilton III and colleagues (1978), for example, noted that chacma baboons (Papio ursinus) specialized on highly profitable foods (insects) when they were abundant, but that the abundance of presumably low-ranking foods (leaves) did not influence their inclusion or exclusion, as predicted by the classical prey model

(Hamilton III, Buskirk, and Buskirk 1978). Actual quantitative testing of predictions from this model coupled with food type ranking has yet to be undertaken with any nonhuman primate. Grether and colleagues (1992), however, have quantitatively tested several assumptions and predictions of another touchstone model of classical foraging theory, the marginal value theorem or patch exploitation model, which predicts when animals should leave one patch to travel to another. This work involved lar gibbons

(Hylobates lar) and siamang (Hylobates [Symphalangus] syndactylus) (Grether,

Palombit, and Rodman 1992).

OFT has been questioned on multiple fronts. In one popular edited volume

devoted largely to foraging theory (Kamil and Sargent 1981), the single contribution by a

primatologist suggested that mantled howler (Alouatta palliata) diets were too complex

from a nutritional standpoint to be accounted for by maximizing one variable such as energy (Glander 1981; see also Milton 1979). Post (1984) offered a useful evaluation on

the limitations of OFT that is required reading for all interested in the subject, but this

contribution focused mainly on situations where primate foraging may violate common

assumptions of classical OFT (see also Janson and Vogel 2006; Richard 1985). For

example, while the classical prey model assumes a “fine-grained” environment, where

resources are evenly distributed and are encountered in proportion to their abundance in

the environment, many animals (including many primates) actually inhabit “coarse-

grained” environments where the encounter rate with a given resource may change

throughout the day as the animal enters different parts of its habitat (Post 1984). This can

fundamentally alter the predictions of the model.

While dietary and habitat complexity are important considerations, they should be

considered challenges, not impediments, to investigating OFT with nonhuman primates.

It is true that classical models often maximize a single currency, such as energy, while nonhuman primates and many other animals face the problem of balancing critical nutrients, toxins and digestion inhibitors. Linear programming models can be utilized to handle such multiple requirements (Altmann 1998; Belovsky 1978), but there are likely to be many situations where a single currency may be sufficient to describe the general feeding patterns of a given animal. For example, the “alternative” (linear programming)

research program that Post (1984) recommended at the conclusion of his critical

evaluation of OFT is itself an optimality model, where the optimal diet in n-dimensional

space is calculated from information on the nutritional requirements of a given animal.

The closer an individual is to this point in space the higher its reproductive success is

predicted to be (Altmann and Wagner 1978). In a non-primate example of this approach,

Columbian ground squirrels (Spermophilus columbianus) that approached a calculated

“optimal diet” had six times greater reproductive success than deviators (Ritchie 1990).

Similarly, S. Altmann (1998), in a landmark study, found that energy shortfall as yearlings accounted for 96% of variability in fecundity and 81% in reproductive success for yellow baboon (Papio cynocephalus) females (Altmann 1998). Although many nutritional and anti-feedant parameters were considered, it could be argued that energy alone would make a perfectly reasonable currency for maximization in yellow baboons.

A single currency would also be acceptable when multiple nutrients are highly correlated within food types, in other words, when maximizing one nutrient maximizes many

(Glander 1981; Stephens and Krebs 1986).

The “fine-grained” versus “coarse-grained” environment problem may also be less severe than Post (1984) implied. For example, the classical prey model can be applied separately to different parts of an environment that have variable resource abundances (Stephens and Krebs 1986). In an extensive review of tests, it has been noted that predictions from the model are most often upheld in foragers that feed on immobile prey (e.g., fruit or young leaves), a category which would accommodate the

diets of many primates. In addition, the model appears to be fairly robust, and often

withstands violations of some of its assumptions (Sih and Christensen 2001).

Here I compare predictions of the classical prey model, modified for patch choice

(Schoener 1987; Schoener 1974), with the behavior of Himalayan gray langurs

(Semnopithecus entellus) living at a high altitude (3000-4000 m) site at Langtang

National Park, Nepal. The gray langur is a colobine monkey possessing a large, multi-

chambered stomach with symbiotic gut microorganisms which aid in the digestion of

high-fiber foods (Bauchop and Martucci 1968; Kay and Davies 1994). Although

colobines are popularly described as “leaf-eating monkeys,” gray langurs have an

eclectic, generalist diet that varies seasonally (Koenig and Borries 2001), and this is

particularly true of Himalayan populations (Curtin 1982). This provides an ample

opportunity to investigate predictions of the classical prey model as they pertain to

behavioral shifts in response to changes in the abundance of foods. Field observations

included continuous recording of feeding bouts and between-patch travel times, and

laboratory work included standard nutritional analysis of langur foods. I apply a simple

modified version of the classical prey model with corrections for search costs, and use

three currencies (kcal, kcal with a flat correction for neutral detergent fiber fermentation,

and crude protein; for rationale, see Methods). Since the classical prey model has not

been examined in detail in any nonhuman primate, it is appropriate to begin with this

simple, but potentially robust, model before moving to a more complex one with added

constraints (Grether, Palombit, and Rodman 1992).

METHODS

Behavioral observations

All behavioral observations were dictated into a cassette recorder between

December 2002 and December 2003 and subsequently transcribed. Only data from F troop is considered here. A different focal individual was chosen for each sample day (n

= 54) and data were collected on each food patch that was observed to be entered by this

individual. Focal individuals were rotated among non-adults (listed as “juveniles”), adult

females, and adult males. A patch is defined as an area of food concentration separated

from other patches by areas with little or no food. In general, each tree, shrub or herb

clump can be considered a separate patch (Astrom, Lundberg, and Danell 1990; Stephens

and Krebs 1986). There are, however, some situations where multiple plants grow

contiguously and an animal can feed simultaneously in more than one food source.

These were considered as single patches and when the foods types differed they were

treated as simultaneous encounters with multiple patch types (see below).

Because the length of time in which individuals could be followed varied

extensively based on topography, feeding data were collected from other individuals chosen at random whenever the focal animal was not visible. Whenever possible, individual identification was recorded.

When a focal individual was observed to enter, or was already feeding in, a food patch, the following data were dictated into the recorder: food species, plant part ingested, the time and size of each bite, within-patch travel, and time of patch departure.

Bite size refers to the number of food items (leaves, fruit, etc.) put into the mouth, and

when number could not be deciphered, the average number of items per bite for that

patch was later substituted. Periods when ingestion could not be observed were

considered missing time and discarded (Grether, Palombit, and Rodman 1992). When

the focal individual left a patch, it was followed, whenever possible, until it entered

another food patch, and recording ceased only when the individual stopped feeding or

moving (e.g., began resting, grooming, etc.). When necessary, observations were aided by binoculars, or, rarely, a spotting scope. These data allow calculation of intake over time in a second-by-second fashion for each patch or patch type, as well as average travel

time between food patches. In total, 403 langur feeding bouts were recorded that included age-sex data and foods in which all nutritional analyses have been performed

(see below). Ninety-seven (97) between-patch travel times were estimated.

Food types were collected and weighed wet, field dried, and after laboratory

drying. Laboratory drying was completed at Peabody Museum, Harvard University.

Plant identifications were conducted by plant scientists at the Central Department of

Botany, Tribhuvan University, Kathmandu, Nepal.

Nutritional analysis and currencies for the model

Nutrient (crude protein, water soluble carbohydrate, lipids, hemicellulose) and non-

nutrient (cellulose, cutin, lignin, total tannins) analyses were conducted by the author on 55

Himalayan langur food types at the Nutritional Ecology Laboratory in the Department of

Anthropology, Peabody Museum, Harvard University (Conklin-Brittain, Wrangham, and

Hunt 1998; Wrangham, Conklin-Brittain, and Hunt 1998). Crude protein (CP) was

determined using the Kjeldahl procedure for total nitrogen and multiplying by 6.25 (Pierce

and Haenisch 1958) instead of using the 4.3 conversion factor (Conklin-Brittain and others

1999; Norconk and Conklin-Brittain 2004).

The detergent system of fiber analysis (Goering and van Soest, 1970) as modified

by Robertson and van Soest (1980) was used to determine the neutral-detergent, or total cell

wall fraction (NDF) that includes hemicellulose (HC), cellulose (Cs), sulfuric acid lignin

(Ls) and cutin (Goering and Van Soest 1970; Robertson and Van Soest 1981). Total ash, an

estimate of overall mineral content, was measured in accordance with Williams (1984).

Lipid content was measured using petroleum ether extraction for four days at room

temperature, a modification of the method of the Association of Official Analytical

Chemists (Williams 1984). Free simple sugars (FSS) (formerly referred to as water soluble

carbohydrates, Conklin-Brittain et al. 1998) were estimated using a phenol/sulfuric acid

calorimetric assay of Dubois et al. (1956) as modified by Strickland and Parsons (1972)

(DuBois and others 1956; Strickland and Parsons 1973), with sucrose as the standard. Total

nonstructural carbohydrates (TNC) were calculated as follows: TNC = 100 - %NDF -

%lipids - %CP - %ash (Conklin-Brittain et al. 1998). The results of the analyses are utilized

as a percentage of organic matter (OM), which excludes inorganic materials (ash).

Currencies for use in the foraging models include: 1) zero-fermentation

metabolizable energy (MEO, kcal/100g organic matter) = (4 × % total nonstructural

carbohydrate) + (4 × % crude protein) + (9 × % lipids), 2) high-fermentation metabolizable energy (MEH, kcal/100g organic matter) = MEO + (2.0 × %NDF)

(Conklin-Brittain, Knott, Wrangham 2006; National Research Council 2003), and 3)

crude protein (CP). Energy is a convenient currency that is applicable in many

situations, and has the added advantage that search costs can also be reported in kilocalories. Although nutritional analyses of colobine foods that include estimates of

energetic value are rare, this variable has been suggested to be a major component of

food selection for some colobines e.g., Colobus polykomos (Dasilva 1994), and

Himalayan langurs live in a marginal environment where energetic considerations are

likely to be important. However, due to the foregut fermentation of colobine monkeys,

they are likely able to derive more energy from fibrous foods than is suggested by the

standard MEO equation (Kay and Davies 1994). Conklin-Brittain (2006) calculated MEH

in chimpanzees, which can digest approximately half of the NDF in their diet through

hindgut fermentation, as MEH = MEO + (1.6 × NDF) (Conklin-Brittain, Knott, and

Wrangham 2006). Foregut fermenters, however, show greater apparent digestibility of

fiber than do hindgut fermenters (Edwards and Ullrey 1999), with values of at least

68.9% of NDF (National Research Council 2003). Therefore, here we use MEH = MEO +

(2.0 × %NDF) as a conservative correction to account for colobine fermentation. There

are likely to be problems with this “flat” correction applied equally to all food types, as

foods with differing nutritional characteristics may be assimilated in differing fashions.

However, at this point very little is known about the differences in assimilation of

different colobine foods, other than a general preference for lower-fiber leaves over

higher fiber leaves (Waterman and Kool 1994). We argue the “flat” correction is a

reasonable starting point for the investigation of such questions.

Crude protein has long been considered to play a role in diet choice for colobine

monkeys (Milton 1979) and herbivores in general (Newman 2007). Note from the above that CP is in itself a component of ME calculations and the two measurements are often correlated. The primary evidence for the importance of CP in colobine food selection lies

in the finding that a protein-to-fiber ratio is useful in predicting colobine leaf choice

(Milton 1979) or even biomass (Chapman and others 2002), although alternative ratios

(such as MEO-to-fiber) are rarely considered. The primary limitation to the use of a ratio is that it is often unclear whether it is the numerator or denominator, or both, that is driving food selection, and thus we limit ourselves to CP in the foraging model.

The model

Although there are a number of derivations of the classical prey model, we choose a modified version that treats patches as analogous to prey, and includes search costs

(Charnov 1976b; Paulissen 1987; Schoener 1987; Schoener 1974). The formula for the

model is as follows:

∑λ 1 1 ∑ λ

Where En/T is the net energy (or other currency) acquired over time foraging, λ is the encounter rate with patches of type i, ei is the energy (or other currency) acquired from

patch type i, hi is the time spent handling items of patch type i, and Cs is the cost of

searching for food (kilocalories per second).

Encounter rates (λ were determined by dividing the number of patches of type i

entered by total search time (average travel time between patches for that season × total number of patches for that season). Only patches where at least one bite of food was

taken were considered “encountered.” This provides an estimate of encounter based on

actual animal observations and is the approach taken in some of the more detailed tests of

the classical prey model (Paulissen 1987). Encounter rates are expressed as patches per

second of search time. The currency (ei) is expressed as kilocalories or grams organic matter (for CP) acquired while exploiting patch type i, and handling time is expressed as

seconds spent exploiting patch type i.

Search costs (Cs) were determined using a general equation of mass-specific cost

of terrestrial locomotion from Taylor et al. (1982) (Taylor, Heglund, and Maloiy 1982):

. . 10.7 + 6.03 (2)

where vg is velocity in meters per second and has units of watts/kg, which were

then converted to kilocalories. The average velocity was estimated at 1.25 m/s; a

“comfortable walking speed” for most primates (Steudel-Numbers 2003:257). Adult

males were estimated at 19.5 kg, adult females 16.1 kg, and juveniles as ¾ the weight of

adult females (12.1 kg) (Bishop 1975). Although basic metabolic costs are retained here

in the calculation, it is still likely an underestimate of the relative cost of search, particularly in the Himalayan environment. Search costs are not included in the model

applications using CP as currency.

Model assumptions and likely causes of deviation are given in Table 14.

Table 14. Assumptions of the classical prey model as applied to patch choice (from Stephens and Krebs 1986) with notes concerning likely deviations.

Assumption Description Notes Search and handling While exploiting a Herbaceous foods, which are sometimes mutually exclusive patch, animal does not interspersed with other herbs, may violate. search for or gather Also patch types encountered information on other simultaneously. patches.

Patches encountered Patches are encountered Definite cases of simultaneous encounter, sequentially one after another and are in violation of this assumption, were noted. not alternatives. These include different food types (e.g., leaves and fruit) taken from one tree/shrub alternatively and two or more patch types growing contiguously (see Results).

Patches encountered Forager comes upon Along with the “fine-grained environment” randomly patches without prior assumption,” has been suggested to be the knowledge or a most unlikely for nonhuman primates (Post predetermined travel 1986). Larger trees and cultivated fields, path. which were sometimes revisited, are likely deviations for Himalayan langurs.

Complete information Forager knows model Assumption may be less reasonable as variables such as greater numbers of food types are taken encounter rates and within a sampling period. patch identity/quality.

Homogeneous, fine- Patches of similar type Rendered less feasible when more than one grained environment are not “clumped,” but habitat type is entered during a given are evenly distributed in sampling period. the environment.

Model predictions and statistics

The classical prey model makes four predictions (Stephens and Krebs 1986).

Prediction 1: Quantitative estimation of profitability threshold for dropping items from diet. For each of eight seasonal time periods, patch types were entered into

Equation 1 in the order of their profitability (MEO or MEH in kcal/second, CP in grams organic matter/second). The eight time periods chosen were predicated on the basis of sample sizes as well as environmental considerations, and thus represent varying lengths of time. These include late winter (late December-March), spring (April-May), monsoon

(September), fall 1-4 (October and November, divided into four two-week samples), and early winter (early December). The set of patch types that results in the highest /T defines the optimal diet. This procedure was carried out with patches from 1) all juveniles, 2) all adult females, 3) all adult males, and 4) a single adult male. Because the classical prey model is designed to predict the behavior of a solitary forager, and that the

“optimal diet” may differ between individuals, it is expected that the last condition will most closely fit the model (Krebs and McCleery 1984). This single male was also the alpha male, which should reduce dominance effects on diet selection. A corollary of this prediction is the zero-one rule; food types, under given environmental conditions, should either never be taken or always be taken upon encounter.

Predictions 2-4 will be tested using Spearman rank order correlations.

Prediction 2: More profitable patch types will be preferred. Correlations were used to assess the relationship between patch type profitability by MEO, MEH and CP and percent contribution to annual diet by both organic matter and time spent feeding. To

account for temporal effects in food availability, correlations were also performed for all

seasonal time periods where ≥ 5 food items were eaten.

Prediction 3: Higher encounter rates with profitable foods will result in increased

selectivity. The patch types exploited by members of each age-sex class, and by one

individual male, were divided into two categories, “high-ranking” (top half) or “low-

ranking,” (bottom half) based on their profitability across all patches and seasons by

MEO, MEH or CP. Encounter rates with high-ranking foods were then correlated with the

number of patch types (species and plant part) and food parts (plant part only) taken during seasonal time periods. For the latter condition, plant part categories included: 1) deciduous and herbaceous leaf parts, 2) evergreen leaf parts, 3) dormant leaf buds, 4) fruit and seeds, 5) soft underground storage organs, 6) woody underground storage organs, 7) bark, and 8) flowers. The classical prey model predicts a negative correlation.

Data were also visually inspected to see if low-ranking foods that were available over much of the year were taken only when encounter rates with high-ranking foods were low.

Prediction 4: Selectivity is not dependent on encounter rates with low-ranking patch types. The encounter rates with low-ranking foods were correlated with the percent of the diet made up of low-ranking foods by organic matter and time. The model predicts no correlation. Seasonal encounter rates with high-ranking foods were then correlated with the percent of the diet made up of low-ranking foods by both organic matter and time. The model predicts a negative correlation.

Comparison of currencies. For each age-sex class combination, conformation of langur behavior to predictions of the model was examined under MEO, MEH and CP. For each prediction, currencies were given a rank of 1 to 3, with 1 = closest to model predictions and 3 = furthest from model predictions. For the quantitative threshold for dropping items from the diet, the percent contribution of foods considered in the predicted optimal diet was compared for each season and currency. This was performed both with and without the inclusion of search costs for the energetic currencies. The currency which included the highest percentage of diet in the predicted optimal set by organic matter and time was given a rank of 1. For the remaining predictions, strength of correlation in the direction predicted was examined for each currency. Preference was ascertained by annual correlations between dietary contribution and profitability based on the three currencies. For all of the above, situations where organic matter and time measures resulted in different conclusions were classified as ties. For profitability, when such discrepancies occurred for annual contribution, all seasons with ≥ 5 feeding sessions were examined.

Deviations from model assumptions and success of the model in predicting diet.

As the model is being applied seasonally, it is important to note whether the model performs better when more of its assumptions are met. To examine this question, I compared seasonal estimates of likely deviations from model assumptions with the success of the model in predicting diet for a single adult male. Herbaceous vegetation was considered the most likely food type to deviate from the “exclusivity of search and handling” assumption and food types other than from shrubs, herbs, and climbers the

most likely to deviate from the random encounter assumption (but see Menzel 1991;

Stephens and Krebs 1986). The seasonal percentages of definite simultaneous encounters were noted, and greater numbers of patch types exploited were viewed as rendering the

“complete information” assumption more unlikely. In addition, the number of woody habitats exploited per season was noted, with the assumption that feeding within one habitat is more likely to approximate the assumption of a fine-grained environment than multiple habitats. Spearman rank-order correlations were then used to assess the relationship between seasonal rankings of “assumptions met” and the percentage of foods predicted in the optimal diet by organic matter and time.

All statistical tests are two tailed with p < 0.05, and were performed in SPSS 13.0.

RESULTS

Threshold for dropping items and the zero-one rule

Data used for seasonal rankings is given on Table 15. An example of patch type ranking and profitability threshold calculation (from Equation 1) is given in Table 16 for a single adult male and kilocalories (MEO) as currency. Langurs, including this adult male, consistently exploited items below the profitability threshold. The monkeys included foods not predicted in the “optimal set” in twenty-three of 24 (95.8%) age-sex- specific, seasonal applications of the model where n ≥ 5 feeding sessions, and this rate of failure was the same whether the currency utilized was MEO, MEH, or CP (Figures 13-

24). The predicted optimal diet differed based on currency utilized in 15/24 (62.5%) of the applications of the prey algorithm. Although MEO and MEH differed from one

Table 15. Variables entered into Equation (1) to estimate the profitability threshold for dropping items from the diet, arranged by season, age-sex classification, and plant part. Encounter rates (λ) are given in patches (n) per second of search time (search time = total estimated travel time between patches for the sample). Handling times (hi) are given in seconds. For the three alternative currencies, zero-fermentation metabolizable energy (MEO) and high-fermentation metabolizable energy (MEH) are given in kilocalories; crude protein (CP) in grams organic matter. Profitability is presented as currency per minute over all patches of that food type [(currency/hi) 60]. Food types for each season and age-sex class are listed by MEO profitability. USO = underground storage organ, e.m. = evergreen mature. All other broad-leaves are deciduous.

species part n

Late winter adult females search time = 635 s Hippophae rhamnoides m. leaf 1 0.00158 19.94 25.72 2.18 276 4.34 5.60 0.4753 Cotoneaster frigidus bark 1 0.00158 3.82 5.71 0.11 165 1.39 2.07 0.0407 Cotoneaster frigidus ripe fruit 3 0.00473 18.45 29.34 0.76 965 1.15 1.82 0.0471 Gaultheria sp. petiole 1 0.00158 0.69 1.03 0.02 53 0.78 1.16 0.0259 Elsholtzia fruticosa USO hard 2 0.00315 6.68 17.17 0.41 948 0.42 1.09 0.0259 Hippophae rhamnoides leaf bud 1 0.00158 3.99 7.52 0.56 1314 0.18 0.34 0.0256 Gaultheria sp. e.m. leaf 4 0.00630 5.18 12.75 0.81 1777 0.17 0.43 0.0274

Late winter adult males search time = 115 s

Cotoneaster frigidus ripe fruit 2 0.01733 9.19 14.62 0.38 317 1.74 2.77 0.0713 0.3062 Spring juveniles search time = 247 s

Zanthoxylum nepalense y. leaf/flower 7 0.02832 106.03 131.13 13.93 2730 2.33 2.88 94

species part n

Jasminum humile leaf bud 1 0.00405 2.32 3.03 0.22 120 1.16 1.52 0.1097 Clematis montana y. leaf 1 0.00405 0.52 0.71 0.07 50 0.63 0.85 0.0804 Cotoneaster frigidus y. leaf 2 0.00809 0.22 0.50 0.05 67 0.20 0.45 0.0451

Spring adult females search time = 725 s

Zanthoxylum nepalense y. leaf/flower 19 0.02622 152.49 188.59 20.04 3296 2.78 3.43 0.3647 Jasminum humile y. leaf 2 0.00276 15.65 17.91 1.04 425 2.21 2.53 0.1472 Clematis montana y. leaf 1 0.00138 2.97 4.01 0.38 97 1.84 2.49 0.2361 Sorbus cuspidata leaf bud 1 0.00138 1.14 1.69 0.04 46 1.48 2.19 0.0515 Rosa macrophylla y. leaf 7 0.00966 19.83 25.67 1.91 921 1.29 1.67 0.1242 Salix tetrasperma flower 1 0.00138 3.87 8.47 0.86 188 1.23 2.70 0.2757 Jasminum humile leaf bud 5 0.00690 15.28 20.02 1.45 808 1.14 1.49 0.1075 Compositae herb leaf 2 0.00276 6.28 8.21 0.86 428 0.88 1.15 0.1213 Berberis aristata y. leaf 1 0.00138 1.26 1.49 0.15 105 0.72 0.85 0.0881 Cotoneaster frigidus y. leaf 2 0.00276 3.17 7.21 0.73 401 0.47 1.08 0.1085 Rosa sericea y. leaf 1 0.00138 0.18 0.25 0.02 24 0.45 0.63 0.0373

Spring adult males search time = 630 s

Zanthoxylum nepalense y. leaf/flower 13 0.0206 99.89 123.54 13.13 2519 2.38 2.94 0.3126 Berberis aristata y. leaf 1 0.0016 11.17 13.27 1.37 511 1.31 1.56 0.1610 Rosa macrophylla y. leaf 1 0.0016 2.56 3.32 0.25 120 1.28 1.66 0.1232 Rosa sericea y. leaf 1 0.0016 1.01 1.39 0.08 98 0.62 0.85 0.0505 Jasminum humile leaf bud 1 0.0016 0.62 0.81 0.06 61 0.61 0.80 0.0575 Cotoneaster frigidus y. leaf 1 0.0016 0.22 0.50 0.05 57 0.23 0.52 0.0523

Spring one adult male 95 search time = 151 s

species part n

Zanthoxylum nepalense y. leaf/flower 5 0.0330 21.07 26.06 2.77 494 2.56 3.17 0.3365 Rosa macrophylla y. leaf 1 0.0066 2.56 3.32 0.25 120 1.28 1.66 0.1232 Cotoneaster frigidus y. leaf 1 0.0066 0.22 0.50 0.05 57 0.23 0.52 0.0523

Monsoon juveniles search time = 74 s

Sorbus cuspidata unripe fruit 1 0.0135 4.00 7.17 0.13 34 7.03 12.60 0.2291 Tsuga dumosa bark 1 0.0135 0.72 1.56 0.02 54 0.81 1.73 0.0247 54 fruit 2 0.0270 3.69 8.38 0.26 596 0.37 0.84 0.0260

Monsoon adult females search time = 93 s

Zanthoxylum nepalense m. leaf 1 0.01079 1.85 2.24 0.11 37 3.00 3.63 0.1729 54 fruit 4 0.04315 8.09 18.39 0.57 574 0.85 1.92 0.0593

Fall 1 juveniles search time = 196 s

Zanthoxylum nepalense m. leaf 5 0.02552 60.22 72.71 3.46 1077 3.36 4.05 0.1930 Cotoneaster frigidus m. leaf 1 0.00510 21.36 27.46 1.49 945 1.36 1.74 0.0947 Rosa macrophylla m. leaf 1 0.00510 0.67 0.76 0.02 80 0.50 0.57 0.0154

Fall 1 adult females search time = 109 s

Solanum tuberosum USO soft 2 0.01832 170.32 174.66 2.81 591 17.30 17.74 0.2855

Malva sp. herb leaf 1 0.00916 13.97 15.79 2.04 80 10.42 11.77 1.5201 96 Zanthoxylum nepalense m. leaf 2 0.01832 18.54 22.38 1.07 280 3.97 4.79 0.2281

species part n

Hippophae rhamnoides m. leaf 2 0.01832 5.40 6.97 0.59 161 2.01 2.59 0.2196 Cotoneaster frigidus m. leaf 4 0.03664 10.77 13.85 0.75 409 1.58 2.03 0.1103 Cotoneaster frigidus ripe fruit 1 0.00916 1.77 2.82 0.07 149 0.71 1.13 0.0292

Fall 1 adult males search time = 289 s

Solanum tuberosum USO soft 3 0.01037 373.92 383.45 6.17 850 26.39 27.06 0.4354 Caragana gerardiana unripe fruit 1 0.00346 2.71 3.12 0.50 14 11.49 13.21 2.1069 Rumex nepalensis? herb leaf 2 0.00691 4.12 4.70 0.44 32 7.80 8.90 0.8410 Zanthoxylum nepalense m. leaf 1 0.00346 17.05 20.59 0.98 290 3.53 4.26 0.2029 Raphanus sativus herb leaf 2 0.00691 2.14 2.65 0.33 42 3.06 3.79 0.4670 Cotoneaster frigidus m. leaf 6 0.02074 18.96 24.38 1.32 978 1.16 1.50 0.0812 Berberis aristata ripe fruit 1 0.00346 1.11 1.48 0.07 159 0.42 0.56 0.0271

Fall 1 one adult male search time = 134 s

Solanum tuberosum USO soft 1 0.00745 98.02 100.51 1.62 236 24.93 25.56 0.4113 Caragana gerardiana unripe fruit 1 0.00745 2.71 3.12 0.50 14 11.49 13.21 2.1069 Rumex nepalensis? herb leaf 2 0.01490 4.12 4.70 0.44 32 7.80 8.90 0.8410 Cotoneaster frigidus m. leaf 4 0.02979 15.13 19.45 1.06 802 1.13 1.45 0.0790

Fall 2 juvenile search time = 159 s

Hippophae rhamnoides m. leaf 1 0.00628 25.27 32.60 2.77 235 6.45 8.31 0.7055 Euphorbia sp. ripe fruit 1 0.00628 10.78 12.37 0.27 175 3.69 4.24 0.0918 Zanthoxylum nepalense m. leaf 2 0.01255 14.83 17.90 0.85 276 3.23 3.89 0.1855 Sorbus cuspidata ripe fruit 1 0.00628 2.61 4.21 0.08 66 2.37 3.83 0.0746 Cotoneaster acuminatus ripe fruit 1 0.00628 3.34 5.39 0.13 106 1.90 3.06 0.0761 97

species part n

Hippophae rhamnoides ripe fruit 1 0.00628 9.30 12.69 0.86 480 1.16 1.59 0.1075 Aconogonum molle herb leaf 1 0.00628 2.77 3.96 0.28 165 1.01 1.44 0.1009 Theropogon pallidus herb fruit 1 0.00628 1.77 2.51 0.07 129 0.82 1.16 0.0320 Aconogonum molle herb flower 1 0.00628 0.13 0.19 0.01 12 0.64 0.95 0.0371

Fall 2 adult females search time = 161 s

Solanum tuberosum USO soft 2 0.01245 283.86 291.09 4.68 1228 13.86 14.22 0.2288 Saussurea sp. USO soft 1 0.00622 31.42 54.42 5.12 172 10.97 19.00 1.7857 Caragana gerardiana unripe fruit 1 0.00622 18.98 21.81 3.48 121 9.40 10.81 1.7247 Clematis montana m. leaf 2 0.01245 20.65 22.92 1.16 154 8.04 8.93 0.4508 Elsholtzia fruticosa flower 1 0.00622 9.09 12.36 0.69 91 6.02 8.19 0.4545 Zanthoxylum nepalense m. leaf 1 0.00622 26.32 31.78 1.51 292 5.41 6.53 0.3110 Euphorbia sp. ripe fruit 3 0.01867 60.13 69.03 1.50 1128 3.20 3.67 0.0796 Cotoneaster frigidus m. leaf 1 0.00622 5.11 6.57 0.36 151 2.03 2.61 0.1415 Fagopyrum esculentum herb leaf 1 0.00622 2.80 3.33 0.31 101 1.66 1.98 0.1864 Fagopyrum esculentum herb flower 1 0.00622 2.64 4.53 0.47 102 1.55 2.67 0.2798 Cotoneaster acuminatus ripe fruit 1 0.00622 0.35 0.57 0.01 15 1.43 2.30 0.0574 Theropogon pallidus herb fruit 1 0.00622 0.41 0.58 0.02 24 1.04 1.47 0.0403

Fall 2 adult males search time = 494 s

Solanum tuberosum USO soft 4 0.00810 519.14 532.36 8.57 1477 21.09 21.63 0.3481 Clematis montana m. leaf 2 0.00405 61.93 68.74 3.47 270 13.78 15.30 0.7725 Rumex nepalensis? herb leaf 1 0.00203 5.67 6.46 0.61 41 8.30 9.46 0.8947 Hippophae rhamnoides m. leaf 1 0.00203 34.50 44.49 3.78 321 6.44 8.31 0.7049 Elsholtzia fruticosa flower 1 0.00203 17.36 23.60 1.31 172 6.04 8.21 0.4558

Caragana gerardiana unripe fruit 4 0.00810 54.76 62.95 10.04 708 4.64 5.34 0.8516 98 Zanthoxylum nepalense m. leaf 2 0.00405 47.45 57.30 2.73 626 4.55 5.49 0.2617

species part n

Theropogon pallidus herb fruit 1 0.00203 1.91 2.70 0.07 26 4.40 6.24 0.1714 Raphanus sativus herb leaf 1 0.00203 1.22 1.51 0.19 21 3.54 4.38 0.5404 Malva sp. herb leaf 1 0.00203 40.15 45.38 5.86 732 3.29 3.72 0.4805 Compositae herb leaf 1 0.00203 1.06 1.39 0.15 21 3.07 4.02 0.4227 Fagopyrum esculentum herb leaf 1 0.00203 0.18 0.21 0.02 4 2.59 3.08 0.2904 Cotoneaster frigidus m. leaf 7 0.01418 49.95 64.21 3.49 1561 1.92 2.47 0.1340 Hippophae rhamnoides unripe fruit 1 0.00203 1.26 1.72 0.12 71 1.07 1.46 0.0990

Fall 2 one adult male search time = 96 s

Solanum tuberosum USO soft 3 0.03124 475.57 487.68 7.85 1366 20.89 21.42 0.3446 Clematis montana m. leaf 1 0.01041 45.15 50.12 2.53 198 13.71 15.22 0.7687 Rumex nepalensis? herb leaf 1 0.01041 5.67 6.46 0.61 41 8.30 9.46 0.8947 Hippophae rhamnoides m. leaf 1 0.01041 34.50 44.49 3.78 321 6.44 8.31 0.7049 Elsholtzia fruticosa flower 1 0.01041 17.36 23.60 1.31 172 6.04 8.21 0.4558 Caragana gerardiana unripe fruit 1 0.01041 3.25 3.74 0.60 43 4.58 5.26 0.8396 Theropogon pallidus herb fruit 1 0.01041 1.91 2.70 0.07 26 4.40 6.24 0.1714 Raphanus sativus herb leaf 1 0.01041 1.22 1.51 0.19 21 3.54 4.38 0.5404 Malva sp. herb leaf 1 0.01041 40.15 45.38 5.86 732 3.29 3.72 0.4805 Compositae herb leaf 1 0.01041 1.06 1.39 0.15 21 3.07 4.02 0.4227 Fagopyrum esculentum herb leaf 1 0.01041 0.18 0.21 0.02 4 2.59 3.08 0.2904 Hippophae rhamnoides unripe fruit 1 0.01041 1.26 1.72 0.12 71 1.07 1.46 0.0990 Cotoneaster frigidus m. leaf 1 0.01041 0.37 0.47 0.03 24 0.92 1.18 0.0639

Fall 3 juveniles search time = 203 s

Sorbus cuspidata ripe fruit 3 0.01481 27.78 44.83 0.87 328 5.08 8.21 0.1600 Caragana gerardiana unripe fruit 2 0.00987 26.57 30.54 4.87 479 3.32 3.82 0.6098 99 Cotoneaster frigidus m. leaf 4 0.01974 31.22 40.13 2.18 1152 1.63 2.09 0.1135

species part n

Cotoneaster frigidus ripe fruit 1 0.00494 0.68 1.08 0.03 64 0.64 1.02 0.0264

Fall 3 adult females search time = 618 s

Sorbus cuspidata ripe fruit 5 0.00809 242.86 391.90 7.64 812 17.94 28.95 0.5645 Euphorbia sp. ripe fruit 1 0.00162 4.96 5.69 0.12 27 10.95 12.57 0.2724 Myrsine semiserrata e.m. leaf 1 0.00162 40.84 47.25 2.40 233 10.52 12.17 0.6185 Rosa sericea ripe fruit 1 0.00162 13.63 26.31 0.50 95 8.63 16.66 0.3165 Caragana gerardiana unripe fruit 10 0.01617 412.04 473.69 75.58 4125 5.99 6.89 1.0995 Saussurea sp. USO soft 1 0.00162 6.28 10.88 1.02 75 5.00 8.66 0.8145 Vibernum cotinifolium ripe fruit 1 0.00162 7.59 10.16 0.19 95 4.77 6.39 0.1182 Allium wallichii herb fruit 2 0.00323 14.65 20.71 0.73 187 4.70 6.65 0.2358 Solanum tuberosum USO soft 1 0.00162 11.83 12.13 0.20 269 2.64 2.71 0.0436 Cotoneaster frigidus m. leaf 13 0.02102 98.34 126.43 6.86 2453 2.40 3.09 0.1678 Gaultheria sp. fruit 1 0.00162 2.96 4.64 0.14 82 2.17 3.40 0.1008 Berberis aristata ripe fruit 1 0.00162 2.95 3.95 0.19 233 0.76 1.02 0.0491 Rubia manjith herb fruit 2 0.00323 1.65 2.30 0.11 239 0.41 0.58 0.0285 Gaultheria sp. e.m. leaf 1 0.00162 0.40 0.98 0.06 137 0.17 0.43 0.0274

Fall 3 adult male Search time = 481 s

Solanum tuberosum USO soft 2 0.00416 370.29 379.72 6.11 720 30.87 31.66 0.5095 Caragana gerardiana unripe fruit 6 0.01248 72.65 83.52 13.33 660 6.60 7.59 1.2113 Elsholtzia fruticosa flower 2 0.00416 7.44 10.12 0.56 101 4.43 6.02 0.3341 Cotoneaster frigidus m. leaf 6 0.01248 15.50 19.93 1.08 379 2.45 3.15 0.1711 Allium wallichii herb fruit 2 0.00416 1.26 1.78 0.06 41 1.83 2.58 0.0916 100 Berberis aristata ripe fruit 2 0.00416 2.57 3.44 0.17 136 1.14 1.52 0.0735

species part n

Aconogonum molle? herb flower 1 0.00208 0.13 0.20 0.01 9 0.90 1.34 0.0524 Theropogon pallidus herb fruit 1 0.00208 3.81 5.40 0.15 306 0.75 1.06 0.0291

Fall 3 one adult male search time = 72 s

Cotoneaster frigidus m. leaf 3 0.041585 7.119791 9.152953 0.496908 136.96 3.119067 4.009763 0.217688

Fall 4 juvenile search time = 208 s

Sorbus cuspidata ripe fruit 1 0.00482 36.46 58.84 1.15 246 8.89 14.35 0.2799 Rosa sericea ripe fruit 1 0.00482 47.10 90.89 1.73 389 7.26 14.02 0.2663 Euphorbia sp. ripe fruit 3 0.01445 62.07 71.25 1.54 884 4.21 4.83 0.1048 Caragana gerardiana unripe fruit 1 0.00482 43.91 50.49 8.06 636 4.14 4.76 0.7598 Hippophae rhamnoides m. leaf 1 0.00482 14.96 19.29 1.64 253 3.54 4.57 0.3875 Cotoneaster frigidus m. leaf 2 0.00963 5.66 7.28 0.39 243 1.40 1.80 0.0975 Berberis aristata ripe fruit 6 0.02889 2.80 3.75 0.18 826 0.20 0.27 0.0131

Fall 4 adult females search time = 548 s

Sorbus cuspidata ripe fruit 2 0.00365 189.80 306.29 5.97 900 12.66 20.43 0.3984 Solanum tuberosum USO soft 1 0.00182 26.02 26.68 0.43 129 12.12 12.43 0.2000 Caragana gerardiana USO hard 2 0.00365 28.01 63.01 4.33 167 10.05 22.60 1.5526 Rosa macrophylla m. leaf 1 0.00182 2.00 2.29 0.06 13 9.57 10.94 0.2963 Rosa sericea ripe fruit 1 0.00182 4.96 9.57 0.18 34 8.87 17.12 0.3254 Hippophae rhamnoides m. leaf 2 0.00365 65.39 84.33 7.16 446 8.80 11.35 0.9629 Clematis montana m. leaf 1 0.00182 5.16 5.73 0.29 36 8.70 9.66 0.4878 Caragana gerardiana unripe fruit 4 0.00730 43.37 49.86 7.96 361 7.20 8.28 1.3210 Cotoneaster acuminatus ripe fruit 1 0.00182 7.92 10.05 0.32 74 6.40 10.32 0.2569 101

species part n

Euphorbia sp. ripe fruit 1 0.00182 4.96 5.69 0.12 56 5.33 6.12 0.1326 Saussurea sp. USO soft 2 0.00365 26.93 46.64 4.38 328 4.92 8.53 0.8017 Jasminum humile m. leaf 2 0.00365 4.70 5.35 0.14 61 4.61 5.26 0.1409 Cotoneaster frigidus ripe fruit 2 0.00365 6.32 10.05 0.26 133 2.86 4.55 0.1173 Cotoneaster frigidus m. leaf 8 0.01459 40.31 51.82 2.81 965 2.51 3.22 0.1749 Berberis aristata ripe fruit 1 0.00182 3.19 4.27 0.21 293 0.65 0.87 0.0422 Cotoneaster frigidus leaf bud 1 0.00182 0.90 2.18 0.14 95 0.56 1.38 0.0878 Rubia manjith herb fruit 1 0.00182 0.59 0.81 0.04 66 0.53 0.74 0.0365

Fall 4 adult males search time = 1528 s

Hippophae rhamnoides m. leaf 4 0.00262 133.35 171.99 14.60 552 14.50 18.71 1.5876 Sorbus cuspidata ripe fruit 3 0.00196 203.14 327.80 6.39 1150 10.60 17.10 0.3336 Solanum tuberosum USO soft 7 0.00458 384.81 394.61 6.35 2684 8.60 8.82 0.1419 Caragana gerardiana unripe fruit 6 0.00393 116.02 133.38 21.28 1016 6.85 7.87 1.2564 Allium wallichii herb fruit 1 0.00065 1.26 1.78 0.06 23 3.26 4.61 0.1635 Cotoneaster frigidus ripe fruit 5 0.00327 43.75 69.59 1.80 857 3.06 4.87 0.1257 Cotoneaster frigidus m. leaf 20 0.01309 171.07 219.93 11.94 3988 2.57 3.31 0.1796 Euphorbia sp. ripe fruit 3 0.00196 17.67 20.29 0.44 434 2.44 2.80 0.0607 Cotoneaster acuminatus ripe fruit 3 0.00196 10.21 16.45 0.41 326 1.88 3.03 0.0755 Rubia manjith herb fruit 7 0.00458 17.40 24.18 1.20 1062 0.98 1.37 0.0676 Berberis aristata ripe fruit 6 0.00393 10.72 14.36 0.69 1194 0.54 0.72 0.0348

Fall 4 one adult male search time = 591 s

Hippophae rhamnoides m. leaf 2 0.00338 54.00 69.65 5.91 158 20.51 26.45 2.2446 Caragana gerardiana unripe fruit 1 0.00169 3.80 4.36 0.70 13 17.79 20.45 3.2631 Sorbus cuspidata ripe fruit 1 0.00169 96.36 155.50 3.03 341 16.94 27.34 0.5333 102 Solanum tuberosum USO soft 6 0.01015 377.55 387.17 6.23 2645 8.56 8.78 0.1413

species part n

Cotoneaster frigidus ripe fruit 2 0.00338 14.29 22.73 0.59 247 3.47 5.52 0.1424 Allium wallichii herb fruit 1 0.00169 1.26 1.78 0.06 23 3.26 4.61 0.1635 Cotoneaster frigidus m. leaf 9 0.01523 104.87 134.82 7.32 2055 3.06 3.94 0.2136 Euphorbia sp. ripe fruit 1 0.00169 8.41 9.65 0.21 189 2.66 3.06 0.0662 Cotoneaster acuminatus ripe fruit 2 0.00338 1.41 2.27 0.06 61 1.38 2.22 0.0552 Berberis aristata ripe fruit 1 0.00169 0.55 0.74 0.04 86 0.38 0.51 0.0249

Early winter juveniles search time = 44 s

Caragana gerardiana USO hard 1 0.02275 5.80 13.05 1.06 117 2.99 6.72 0.5479

Early winter adult females search time = 131 s

Hippophae rhamnoides m. leaf 1 0.00764 47.08 60.71 5.15 372 7.59 9.79 0.8304 Caragana gerardiana unripe fruit 3 0.02292 20.06 23.06 3.68 281 4.28 4.92 0.7857 Hippophae rhamnoides ripe fruit 1 0.00764 0.30 0.40 0.03 6 2.80 3.82 0.2587 Cotoneaster frigidus m. leaf 2 0.01528 10.55 13.56 0.74 232 2.73 3.51 0.1907 Cotoneaster frigidus ripe fruit 1 0.00764 3.81 6.06 0.16 93 2.47 3.93 0.1013 Aconogonum molle herb leaf 1 0.00764 11.86 16.98 1.19 300 2.37 3.39 0.2376 Cotoneaster frigidus bark 1 0.00764 0.24 0.36 0.01 13 1.14 1.71 0.0335 Cotoneaster frigidus leaf bud 2 0.01528 3.04 7.41 0.47 184 0.99 2.42 0.1546 Viburnum erubescens ripe fruit 2 0.01528 1.13 2.00 0.12 264 0.26 0.46 0.0281

Early winter adult males search time = 665 s

Hippophae rhamnoides m. leaf 1 0.00150 51.17 66.00 5.60 227 13.53 17.45 1.4805

Caragana gerardiana unripe fruit 3 0.00451 54.22 62.33 9.94 474 6.87 7.90 1.2601 103 Cotoneaster frigidus m. leaf 6 0.00902 72.34 92.99 5.05 1355 3.20 4.12 0.2236

species part n

Aconogonum molle USO hard 1 0.00150 4.95 6.89 0.12 120 2.48 3.46 0.0584 Cotoneaster frigidus ripe fruit 3 0.00451 11.16 17.75 0.46 286 2.34 3.72 0.0960 Cotoneaster frigidus leaf bud 3 0.00451 0.43 1.04 0.07 57 0.45 1.10 0.0701

Early winter one adult male search time = 422 s

Hippophae rhamnoides m. leaf 1 0.00237 51.17 66.00 5.60 227 13.53 17.45 1.4805 Caragana gerardiana unripe fruit 2 0.00474 25.48 29.29 4.67 299 5.11 5.88 0.9374 Aconogonum molle USO hard 1 0.00237 4.95 6.89 0.12 120 2.48 3.46 0.0584 Cotoneaster frigidus ripe fruit 2 0.00474 9.48 15.08 0.39 231 2.46 3.92 0.1010 Cotoneaster frigidus m. leaf 1 0.00237 3.10 3.99 0.22 93 2.00 2.57 0.1397 Cotoneaster frigidus leaf bud 2 0.00474 0.35 0.85 0.05 41 0.51 1.25 0.0799

104

105

Table 16. Seasonal patch types, overall rate of gain (En/T, calculated from Equation 1), and dietary contribution for a single adult male. Patch types listed in order of their profitability with kilocalories over time (MEO) utilized as currency. En/T shows the rate of gain if only that patch type and those of greater profitability were taken; patch types included in the predicted optimal diet for each season are given in bold face. Only seasons with ≥ 5 feeding sessions are shown. OM = organic matter.

E /T % of diet season patch type n % of diet (OM) (kcal/min) (time) Zanthoxylum nepalense YL/FL 2.289 86.3 73.6 spring Rosa macrophylla YL 2.245 11.5 17.9 Cotoneaster frigidus YL 2.204 2.2 8.6 Solanum tuberosum USO 15.107 85.5 21.8 Caragana gerardiana seed 14.974 1.4 1.3 fall 1 Rumex nepalensis HL 13.958 2.3 2.9 Cotoneaster frigidus ML 2.702 10.8 74.0 Solanum tuberosum USO 20.358 73.8 45.0 Clematis montana ML 20.060 6.6 6.5 Rumex nepalensis HL 19.951 0.9 1.3 Hippophae rhamnoides ML 19.038 6.5 10.6 Elsholtzia fruticosa FL 18.583 3.1 5.7 Caragana gerardiana seed 18.463 0.5 1.4 Theropogon pallidus HF 18.389 0.4 0.9 fall 2 Raphanus sativus HL 18.328 0.2 0.7 Malva sp. HL 16.414 7.5 24.1 Compositae HL 16.366 0.2 0.7 Fagopyrum esculentum HL 16.357 0.0 0.1 Hippophae rhamnoides UF 16.172 0.2 2.3 Cotoneaster frigidus ML 16.109 0.1 0.8

Hippophae rhamnoides ML 5.747 9.1 2.7 Caragana gerardiana seed 5.914 0.5 0.2 Sorbus cuspidata RF 8.899 23.3 5.9 Solanum tuberosum USO 8.589 44.1 45.4 Cotoneaster frigidus RF 8.445 3.3 4.2 fall 4 Allium wallichii HF 8.438 0.2 0.4 Cotoneaster frigidus ML 5.687 17.9 35.3 Euphorbia sp. RF 5.671 1.2 3.3 Cotoneaster acuminatus RF 5.656 0.3 1.1 Berberis aristata RF 5.644 0.1 1.5 Hippophae rhamnoides ML 3.338 55.0 22.5 Caragana gerardiana seed 4.188 20.4 29.6 Aconogonum molle USO 4.039 6.1 11.8 early winter Cotoneaster frigidus RF 3.641 14.1 22.9 Cotoneaster frigidus ML 3.561 3.4 9.2 Cotoneaster frigidus LB 3.437 0.9 4.1

106

Figure 13. The seasonal profitability of foods taken (data points) and calculated threshold for inclusion in diet (line) using MEo for juveniles. For all seasons with n ≥ 5 feeding sessions the percentage of feeding time spent on foods in the predicted set is given, with percentage organic matter in brackets.

MEO juvenile

92.0 51.2 25.0 16.2 18.3 10 [96.6] [71.2] [48.2] [43.2] [56.0]

kcal/min 4

0 spring monsoon fall 1 fall2 fall3 fall4 early winter Season = threshold

Figure 14. Seasonal food profitability and threshold, using MEH, for all juveniles.

MEH juvenile

92.0 51.2 25.0 16.2 18.3 [96.6] [71.2] [48.2] [43.2] [56.0] 16

8 kcal/min 6

0 spring monsoon fall 1 fall2 fall3 fall4 early winter

= threshold

107

Figure 15. Seasonal food profitability and threshold, using CP, for all juveniles.

CP juvenile 92.0 51.2 14.3 23.7 18.3 [96.6] [71.2] [35.7] [21.5] [13.5] 0.8

0.6

0.4 CP(gOM)/min

0.2

0.0 spring monsoon fall 1 fall2 fall3 fall4 early winter

= threshold

Figure 16. Food profitability and threshold, using MEO, for all females.

MEO adult female 25.6 48.9 100.0 35.3 34.3 9.0 29.1 21.3 20 [56.0] [65.1] [100.0] [71.4] [54.1] [40.2] [60.2] [46.0]

kcal/min 8

0 late spring monsoon fall 1 fall2 fall3 fall4 early winter winter

= threshold

108

Figure 17. Seasonal food profitability and threshold, using MEH, for all adult females.

MEH adult female 25.6 48.9 100.0 35.3 39.1 9.0 26.5 21.3 35 [56.0] [65.1] [100.0] [71.4] [65.2] [40.2] [58.2] [46.0]

kcal/min 15

0 late spring monsoon fall 1 fall2 fall3 fall4 early winter winter

= threshold

Figure 18. Seasonal food profitability and threshold, using CP, for all adult females.

CP adult female

5.0 48.9 100.0 4.8 8.2 45.5 12.7 37.5 [22.7] [65.1] [100.0] [7.0] [15.0] [35.4] [13.2] [60.6] 2.0

1.8

1.6

1.4

1.2

1.0

0.8 CP (gOM)/min 0.6

0.4

0.2

0.0 late spring monsoon fall 1 fall2 fall3 fall4 early winter winter

= threshold

109

Figure 19. Seasonal food profitability and threshold using MEO, for all adult males.

MEO adult male 74.8 36.0 24.4 30.6 12.8 27.8 35 [86.5] [90.8] [59.8] [75.0] [40.0] [49.0]

kcal/min 15

0 late spring monsoon fall 1 fall2 fall3 fall4 early winter winter

= threshold

Figure 20. Seasonal food profitability and threshold using MEH, for all adult males.

MEH adult males 74.8 36.0 24.4 30.6 12.8 27.8 [86.5] [90.8] [59.8] [75.0] [40.0] [49.0] 35

kcal/min 15

0 late spring monsoon fall 1 fall2 fall3 fall4 early winter winter

= threshold

110

Figure 21. Seasonal profitability and threshold using CP, for all adult males.

CP adult males

74.8 40.0 16.8 28.1 12.8 27.8 [86.5] [92.4] [13.3] [15.8] [39.7] [49.0] 2.5

2.0

1.5

1.0 CP (gOM)/min CP

0.5

0.0 late spring monsoon fall 1 fall2 fall3 fall4 early winter winter = threshold

Figure 22. Seasonal profitability and threshold using MEO, for a single adult male.

MEO single adult male

73.6 21.8 45.0 8.8 52.1 30 [86.3] [85.5] [73.8] [32.9] [75.4]

15 kcal/min

0 late spring monsoon fall1 fall2 fall3 fall4 early winter winter

= threshold

111

Figure 23. Seasonal profitability and threshold using MEH, for a single adult male.

MEH single adult male

73.6 21.8 45.0 8.8 52.1 [86.3] [85.5] [73.8] [32.9] [75.4] 30

15 kcal/min

0 late spring monsoon fall1 fall2 fall3 fall4 early winter winter

= threshold

Figure 24. Seasonal profitability and threshold using CP, for a single adult male.

CP single adult male

73.6 26.0 19.8 2.9 52.1 [86.3] [89.2] [14.4] [9.6] [75.4] 3.5

3.0

2.5

2.0

1.5 CP(g OM)/min CP(g

1.0

0.5

0.0 late spring monsoon fall 1 fall2 fall3 fall4 early winter winter

= threshold

112

another in only 3/24 (12.5%) of cases, the CP optimal diet differed from MEO and MEH in

15/24 (62.5%) and 14/24 (58.3%) of applications, respectively.

Many studies investigating the classical prey model have shown that animals exhibit “partial preferences,” where small amounts of foods not predicted in the optimal diet are nevertheless consumed (Krebs and McCleery 1984; Stephens and Krebs 1986).

Himalayan langurs of all age-sex classes exhibited partial preferences throughout the study. This is illustrated by the diet of a single adult male, which resemble the pooled age-sex results, and show seasonal differences in the extent to which the model could account for observed feeding behavior (Table 16, Figures 22-24). The model performed best in spring, where, under all three currencies, only one food type was predicted in the

optimal diet. This item, Zanthoxylum nepalense young leaf and flower clusters, made up

86.3% of dietary organic matter and represented 73.6% of foraging time. For other

seasons, however, the model failed to varying degrees based on the currency entered into

Equation (1) and/or the method used to quantify diet. Most strikingly, under both

energetic and crude protein currencies this male spent considerable amounts of time

exploiting foods not predicted in the optimal diet, even when organic matter contribution

of these foods were low.

Preference for profitable food types

For all age-sex classes (juveniles, females, males and single male), contribution of

food types to annual diet by percentage organic matter was positively related to MEO,

MEH and CP profitability (Table 17). Conversely, annual percent feeding time was not

113

Table 17. Spearman rank order correlation coefficients between contribution to diet of food types (by % organic matter [OM] or % feeding time) and food item profitability by age-sex class and season. Only seasons where ≥ 5 food types were taken are shown. JJ = all juveniles, ♀♀ = all females, ♂♂ = all adult males, ♂ = one adult male, MEO = zero- fermentation metabolizable energy, MEH = high-fermentation metabolizable energy, CP = crude protein. * = significant at the 0.05 level, ** = significant at the 0.01 level.

% OM % time n MEO MEH CP MEO MEH CP JJ Annual 23 0.77** 0.72** 0.67** 0.35 0.27 0.37 Fall 2 9 0.85** 0.85** 0.88** 0.48 0.48 0.82** Fall 4 7 0.96** 0.96** 0.43 -0.07 -0.07 -0.14

♀♀ Annual 47 0.51** 0.54** 0.49** -0.06 -0.06 0.00 Late Winter 7 0.21 0.25 0.50 -0.68 -0.64 -0.21 Spring 11 0.56 0.62* 0.67* 0.36 0.36 0.54 Fall 1 6 0.89* 0.89* 0.77 0.26 0.26 -0.03 Fall 2 12 0.83** 0.83** 0.49 0.69* 0.64* 0.29 Fall 3 14 0.69** 0.66** 0.75** -0.10 -0.22 -0.01 Fall 4 17 0.51* 0.57* 0.67** -0.03 -0.01 0.25 Early Winter 9 0.50 0.57 0.63 0.27 0.30 0.45

♂♂ Annual 30 0.64** 0.63** 0.47** 0.36* 0.35 0.20 Spring 6 1.00** 0.94** 0.83* 1.00** 0.94** 0.83* Fall 1 7 0.50 0.50 -0.04 -0.25 -0.25 -0.75 Fall 2 14 0.49 0.42 0.08 0.24 0.14 -0.02 Fall 3 8 0.81* 0.81* 0.79* 0.62 0.62 0.60 Fall 4 11 0.66* 0.70* 0.61* -0.01 0.02 0.04 Early Winter 6 0.77 0.83* 0.77 0.49 0.60 0.54

♂ Annual 24 0.66** 0.70** 0.53** 0.33 0.36 0.27 Fall 2 13 0.78** 0.73** 0.46 0.59* 0.53 0.24 Fall 4 10 0.55 0.64* 0.39 0.08 0.20 -0.07 Early winter 6 0.94** 1.00** 0.71 0.71 0.83* 0.49

114

significantly correlated with profitability with the exception of MEO in the adult male

category. Seasonal contribution to diet by percentage organic matter was, in general,

positively correlated with profitability under all three currencies. Again in contrast,

significant positive seasonal relationships between feeding time and profitability were the

exception rather than the rule.

For the single adult male, annual organic matter contribution was positively

related to profitability under all three currencies (Table 17). Correlation coefficients

between annual feeding time and profitability were also positive, but not statistically

significant. Within seasons, significant positive relationships were detected between

organic matter contribution and MEO and/or MEH profitability. A significant positive relationship between feeding time and MEO or MEH was apparent in 2 of 3 seasons.

Crude protein profitability was not significantly correlated with seasonal percentages by

either organic matter or feeding time.

Increased selectivity at higher encounter rates with profitable foods

Under the model, diet breadth is expected to decrease as food abundance

increases. Contrary to expectations, neither the number of patch types nor plant parts exploited by grouped age-sex classes or a single adult male were significantly related to

seasonal encounter rates of high ranking foods under any currency (Table 18). This is

likely related to the fact that many profitable foods were available simultaneously in the

fall seasons, while in the winter and spring seasons there were fewer food types of any

kind available.

Table 18. Spearman rank order correlation coefficients between seasonal encounter rates with high-ranking foods λ under three currencies and the number of food types or plant parts included in the diet. Shown are all seasons with 5 feeding sessions for that age sex-class; sample size reflects number of seasons that meet this criteria.

JJ MEO MEH CP ♀♀ MEO MEH CP ♂♂ MEO MEH CP ♂ MEO MEH CP Food types n=5 0.31 0.31 -0.15 n =8 0.17 0.17 0.17 n =6 0.75 0.75 0.75 n=5 0.70 0.70 0.10 Plant parts -0.26 -0.26 0.37 -0.16 -0.17 -0.16 0.29 0.29 0.10 0.53 0.53 -0.16 115

116

Nevertheless, visual inspection of data (Table 15) from all age-sex classes

suggests that this prediction was partially supported when food availability is considered.

Certain food items that were considered of low profitability under all currencies, such as

Gaultheria evergreen mature leaves and petioles and Elsholtzia fruticosa woody roots,

were available throughout the year but taken almost exclusively when encounter rates

with higher ranking foods were lowest (late winter).

Selectivity is independent of the abundance of low-ranking food

For grouped data, juvenile foraging behavior most clearly ran counter to model

predictions (Table 19). Encounter rates with low-ranking foods were positively related to

feeding time on low-ranking foods irrespective of currency, and encounter rates with high ranking foods did not show a significant negative correlation with the contribution of

low-ranking foods by either organic matter or time. In fact, for juveniles, encounter rates

with high-ranking foods were less-closely related to dietary contribution than low- ranking foods.

The behavior of adult females was consistent with the model under MEO (Table

19). Using this currency, encounter rates with low-ranking foods were not significantly

related to the dietary contribution of low-ranking foods, while the encounter rates with

high-ranking foods showed a strong positive relation to the contribution of high-ranking

foods by both organic matter and time. For adult males, similar agreement with the

model was detected under both MEO and MEH. For both adult sexes in the pooled data

set, the utilization of CP as currency provided slightly weaker conformation to model

117

Table 19. Spearman rank order correlation coefficients between seasonal encounter rates with high or low-ranking foods under three currencies and proportion of organic matter (OM) and feeding time devoted to them. * = significant at the 0.05 level, ** = significant at the 0.01 level.

MEO MEH CP MEO MEH CP JJ (n = 5) % low OM -0.20 -0.20 -0.20 0.60 0.60 0.80 % low time -0.20 -0.20 -0.70 *0.90 *0.90 **1.00 % high OM 0.20 0.20 0.20 -0.60 -0.60 -0.80 % high time 0.20 0.20 0.70 *-0.90 *-0.90 **-1.00

♀♀ ( n = 8) % low OM *-0.79 -0.69 *-0.79 0.43 *0.76 *0.76 % low time *-0.83 -0.52 *-0.79 0.10 0.62 0.62 % high OM *0.79 0.69 *0.79 -0.43 *-0.76 *-0.76 % high time *0.83 0.52 *0.79 -0.10 -0.62 -0.62

♂♂ (n = 6) % low OM *-0.83 *-0.83 *-0.89 0.14 0.20 0.71 % low time **-0.94 **-0.94 -0.54 0.26 0.43 **0.94 % high OM *0.83 *0.83 *0.89 -0.14 -0.20 -0.71 % high time *0.94 *0.94 0.54 -0.26 -0.43 **-0.94

♂ (n = 5) % low OM **-1.00 -0.70 **-0.90 0.00 0.30 -0.70 % low time -0.70 -0.30 -0.70 0.30 0.70 -0.10 % high OM **1.00 0.70 *0.90 0.00 -0.30 0.70 % high time 0.70 0.30 0.70 -0.30 -0.70 0.10

118

predictions, as low-ranking foods as described by this currency were taken in closer

proportion to their encounter rates.

Results from a single adult male were generally consistent with the model (Table

19). Seasonal encounter rate with low ranking foods, under all three currencies, was not

significantly related to percent contribution of low-ranking foods. Also in compliance, a

significant negative relationship between encounter rate with high-ranking foods (under

both MEO and CP) and organic matter contribution of low-ranking foods was detected.

Correlation coefficients concerning encounter rates of high ranking foods and percentage

of time feeding on low-ranking foods were negative but not statistically significant.

Comparison of currencies

Although there was variation, using MEO as currency most closely approximated classical prey model predictions for grouped age-sex classes and a single adult male

(Table 20). However, for the “threshold” prediction, search costs were only included in

the models for MEO and MEH, and increasing search costs can result in a broader

predicted diet (Lifjeld and Slagsvold 1988). When search costs were removed, all

currencies performed equally well for juveniles and adult females, while MEO alone

conformed most closely for pooled adult males and a single adult male.

Table 20. Three currencies (MEO, MEH and CP) compared to predictions of the classical prey model. 1 = closest to model predictions and 3 = furthest from model predictions. Blank cells represent ties and checkmarks (99) indicate the currency to which the model best conforms over all predictions. Asterisks (**) indicate currencies that conform equally well to model predictions when search costs are removed from the MEO and MEH calculations for the seasonal thresholds for dropping items from the diet.

JJ ♀♀ ♂♂ ♂

MEO MEH CP MEO MEH CP MEO MEH CP MEO MEH CP threshold 1 2 2 1 1 3 1 1 3 1 1 3 profitability - - - 3 2 1 1 2 3 2 1 3 increased selectivity ------independence - - - 1 3 2 1 2 3 1 3 2 overall 9 * * 9 * * 9 9 119

120

Conformance of model assumptions and predictions

All age-sex classes engaged in foraging behavior that likely caused deviations from model assumptions. Over all individuals, for example, 78 of 403 patches (19.4%) were definite simultaneous encounters. For a single adult male, seasonal rankings based on assumptions met were positively related to rankings based on the percentage of the diet (MEO) predicted under the model (n = 5, rs = 0.90, p < 0.05) (Table 21).

DISCUSSION

Himalayan langurs generally included patch types not predicted by the classical prey model. In many cases these were rare foods that were well beneath the calculated

“acceptable profitability” and were taken only sporadically even within a season. For example, Cotoneaster frigidus leaf buds were sometimes consumed in the fall, but not on every occasion when foragers entered a tree of this species. In this respect such foods represent partial preferences. These are deviations from the zero-one rule, which states that foods should always be taken or never be taken when they are encountered (Stephens and Krebs 1986). Partial preferences have been observed in almost all tests of the model, both in lab and field (Sih and Christensen 2001).

Krebs and McCleery (1984) provide an abridged list of reasons for partial preferences, some of which may be relevant to this study. Their list includes: 1) discrimination errors, where different food types are confused, 2) long-term learning, where accurate estimates of model variables are only possible after many days of exposure to similar conditions, 3) inherent variation in the animal, 4) runs of bad luck,

Table 21. Seasonal comparisons of likely deviation and compliance with model assumptions compared with success of the model in predicting diet for a single adult male. The rankings of percentage diet predicted are based on the average of the organic matter and feeding time rankings. See Stephens and Krebs (1986).

early Assumption Measure of deviation spring fall 1 fall 2 fall 4 winter search and handling mutually exclusive % herb 0 25.0 40.0 3.8 0 sequential encounter % simultaneous 0 25.0 80.0 15.4 22.2 random encounter % non-shrubs, herbs, climbers 14.3 62.5 40.0 80.8 66.7 complete information # food types 3 4 13 10 6 homogeneous, fine-grained environment # woody habitat types 1 2 1 2 2

rank - assumptions met 1 3 5 4 2

rank - % of diet predicted (MEO) 1 3 4 5 2 121

122

where the animal uses a short-term rule to determine encounter rates and thus

underestimates habitat quality after repeated exposure to unprofitable foods, 5) simultaneous encounters, and 6) averaging across individuals. Little is known about how primates discriminate between food types and to what degree their taxonomies correspond to that used by researchers, i.e., the biological species and both objective

(e.g., leaf) and subjective (e.g., young leaf) descriptions of parts (Menzel 1997). There are certainly situations where discrimination errors could pose a problem for Himalayan langurs. During winter, for example, deciduous plants are largely or completely devoid of foliage, and discriminating between different types of woody shrub roots, which have differing nutritional characteristics, may be difficult. It is unlikely that discrimination errors alone, however, account for the general feeding patterns of the langurs. Primate knowledge of their natural habitats can be substantial (Menzel 1991).

For most primates, including Himalayan langurs, it is possible that conditions change so rapidly that estimates of model variables are one-step behind the environment, in a kind of cognitive slant on Van Valen’s “Red Queen hypothesis” (Van Valen 1973).

A single adult male, for example, continued to exploit potatoes in late fall even after they had been severely depleted and their handling time (digging time) increased to the point

where they were no longer included in the optimal diet. With regard to partial

preferences, little is known on how inherent variations in an animal’s internal clock may influence foraging behavior, or whether primates use short-term rules to estimate model parameters.

123

A minimum of 19.4% of Himalayan langur patches involved “simultaneous encounters” where more than one food type was exploited in the same patch. The actual frequency may exceed this as any situation in which a forager detects two food types at once (e.g., visually spotting a near shrub and a far shrub of differing species) can be considered simultaneous encounters (Krebs and McCleery 1984). These situations may account for some of the partial preferences exhibited by the monkeys. The model performed best when the fewest percentage of undoubted simultaneous encounters were recorded (spring, see below). When simultaneous encounters are the rule rather than the exception, a modified version of the classical prey model suggests that foods cannot be ranked based on their inherent (e.g., nutritional) characteristics (Engen and Stenseth

1984). Although averaging across individuals undoubtedly account for many of the discrepancies observed in the pooled age-sex data, partial preferences were also exhibited by a single adult male.

There are additional causes of partial preferences beyond those suggested by

Krebs and McCleery (1984), some of which may be of importance both in this study and future applications of OFT to primates. For example, nutritional requirements or anti- feedant avoidance can lead to partial preferences (Westoby 1978) and this is clearly the most cited argument cautioning the application of simple foraging models to primate behavior (Glander 1981; Milton 1979; Richard 1985). Guerezas (Colobus guereza), for example, travel long distances outside of their normal range to exploit rare foods rich in sodium (Fashing, Dierenfeld, and Mowry 2007). Although mineral contents are not yet available for Himalayan langur foods, this may be an explanation for the rock-licking

124

behavior seen at Langtang (Chapter 3). No simple nutritional rationale, however, can be

pointed to in terms of most of the deviations observed in this study, at least those

variables for which there are data. An analysis of langur food preference in terms of the

chemical composition of foods and feeding records (Sayers, unpublished) shows a

positive relationship only between energy or protein content and food choice, and these were the bases of the currencies utilized. There was no relationship between langur food

preference and percentages of lipids, free simple sugars, or tannins. There was, however,

a negative correlation between dietary contribution and all fiber variables except

hemicellulose. The highest ranking foods in the seasonal diets in this study were generally low in fiber, however, making this an unlikely reason for departures from model predictions.

The most-likely deviation that may be linked to nutritional factors concerns the exploitation of potatoes from cultivated fields during the fall months. For much of this season, potatoes were the only food in the predicted adult optimal diets by energetic currencies, yet the langurs included many other patch types as well. It has been noted that high-starch diets fed to captive primates, in particular foregut fermenters like colobines, can lead to excessive fermentation, stomach problems, and in extreme cases even death (National Research Council 2003). In fact, it is surprising that Himalayan langurs exploited potatoes as much as they did, by organic matter by far the most copious food in the fall diet. Another possible reason for the expansion of the diet beyond potatoes in this season (and hence another reason for partial preferences) involves variable levels of predation danger or fear associated with different patch types. While

125

feeding in potato fields, Himalayan langurs were exposed to potential aerial predators and also to local farmers wielding sling-shots and stones (see Brown and Kotler 2007).

Partial preferences can also be expected when foragers sample food types to gain information about them, which is a deviation from the “complete information” assumption. The langurs, for example, were observed taste-testing Sorbus cuspidata fruit before acceptance or rejection. Such sampling, whether by taste, touch, or smell, has been observed in other primates (e.g., Alouatta palliata, Glander 1981; Pithecia pithecia,

Norconk, personal observation; see also Dominy and others 2001) and may be especially important in situations where nutritional quality of a “food type” varies spatially or temporally between plants of the same species or even within a single plant (Houle,

Chapman, and Vickery 2007). Patch sampling is also a possible explanation for rare foods that appear to have little nutritional value, although it may be difficult to completely rule out other possibilities.

In this study it was assumed that langurs are globally omniscient, i.e., they know the expected encounter rates with foods over the entire habitat they range in during a given season. Berec and Křivan (2000) have modeled a situation where a forager is only omniscient within the range of its perception, which results in partial preferences.

Nonhuman primates are almost certainly intermediate between these two extremes.

Some foods taken beneath the threshold, however, were not merely “partial preferences,” they were routinely and consistently exploited. The most striking example involves the mature leaves of Cotoneaster frigidus. This was ranked first or second by annual feeding time for all age-sex classes, and in scan samples taken concurrently

126

(Chapter 3) represented the highest percentage of feeding records over an annual cycle.

Yet this resource was not predicted even once to be a part of the optimal diet for any age-

sex class, season, or currency. It is likely that Himalayan langurs perceive their

environment as poorer than suggested by the variables used in this study, i.e., the

“thresholds” from the monkeys’ perspective are lower than those shown on Figures 13-24.

For example, increasing search costs results in a broader predicted diet, and it is likely

that the general equation used here underestimates them (Taylor, Heglund, and Maloiy

1982). The Himalayan environment is obviously characterized by extreme changes in topography which the langurs must negotiate during travel between patches, making movement more costly than would be expected in flatter terrain (Sprague 2000). In a similar vein, underestimates of search time or overestimates of the encounter rates with high-ranking foods would also result in a narrower predicted diet than actually would be observed. It is in this regard that non-random encounter could cause violations from model predictions. Revisiting patches will result in an exaggerated estimated encounter rate and, if it is a high-ranking food, could result in an overly-narrow predicted diet breadth. Most important for this study, potato fields were revisited in the fall months.

Other predictions of the model were generally qualitatively or quantitatively upheld. Strong positive correlations were detected between profitability and organic matter contribution to diet, while correlations between profitability and feeding time were generally positive but weaker. This suggests that “profitability” as defined in optimal foraging models (but not necessarily as perceived by the animals) is driven largely by intake rate (Schülke, Chalise, and Koenig 2006). Foods of low profitability that were

127 available over the entire year were generally taken only when encounter rates with profitable patch types were lowest. With the exception of juveniles, high-ranking foods were taken in close relation to their abundance, while low-ranking foods were not.

The three currencies (MEO, MEH, and CP) conformed equally well to model predictions for pooled juveniles and females, while an energetic currency (MEO) performed best for pooled and a single adult male(s). These findings run counter to some of the colobine literature which argues for the primacy of crude protein in colobine food selection. Wasserman and Chapman (2003), for example, found no correlation between the energy content of food and foraging effort, and a positive relationship between protein-to-fiber content and foraging effort, in red colobus (Procolobus badius) and guerezas (Colobus guereza) at Kibale, Uganda. In addition, they found that estimates of energy consumption were higher than estimates of expenditure for these monkeys and suggested that energy was of minor importance. Although this may be the case, I do not accept their conclusion that these results demonstrate “the importance of protein over other nutritional characters” (p. 657). My reasons include: 1) Wasserman and Chapman looked only at the protein-to-fiber ratio, not crude protein alone, 2) in their calculations of energy consumption, intake rates for plant parts were not estimated directly, but taken from studies of howler monkeys, 3) they assumed that surplus energy is unnecessary, an unlikely scenario in a stochastic environment (Stephens and Krebs 1986), and 4) in any event, they provided no evidence to suggest that crude protein is a limiting variable (see

Oftedal and others 1991). Although protein is generally positively related, and fiber negatively related, to food selection in colobines, it is still an open question as to the

128

relative importance of each of these variables. Fewer studies still have examined calories

or intake rate. It appears that energetic factors, as well as protein, fiber and intake rate, play an important role in Himalayan langur food selection.

Contrary to expectations, however, metabolizable energy with a correction for fermentation (MEH) did not unilaterally outperform the energetic currency without this

correction (MEO). Undoubtedly, being able to ferment higher amounts of fiber than other

primates influences colobine food choice, and leaves (stereotypically a high-fiber food)

make up a significant proportion of the diet at most study sites (Kirkpatrick 1999).

Nonetheless, colobines have consistently shown a preference for lower-fiber over higher

fiber leaves (Davies, Bennett, and Waterman 1988; Fashing, Dierenfeld, and Mowry

2007). My interpretation of this is that fiber exerts a “sliding scale” on colobine food

preference. At low levels, fiber may be nearly completely digested, while at high levels

fiber will subtract from food value either through incomplete digestion, an increase in gut

retention time, or the overproduction of volatile fatty acids which could alter fore-

stomach pH (Lambert 1998). Unfortunately, few data currently exist to test this

hypothesis or to develop a more specific energetic currency for colobine monkeys that

includes variables such as item-specific assimilation (National Research Council 2003).

Elucidating such factors should be one long-term goal for applications of OFT to

primates.

Schoener (1987) noted several potential problems for applying the classical prey

model to patch choice, as performed here. One is that patches (such as trees in this study)

are less likely to be encountered randomly than individual prey items (such as a solitary

129

grasshopper) and can result in departure from model predictions. Another is that patches

are depletable and a forager can alter the profitability of a patch while exploiting it. In

this study, all patch types were assigned a mean value with no account taken of decreases

in intake rate over time (patch depression). In other words, the “patches as prey”

approach is more likely to result to deviations from the assumptions of the model that the standard usage.

In a wide-ranging review, Sih and Christensen (2001), noted that the classical prey model has proven to be quite robust (even in “patches as prey” applications) and often withstands deviations from the assumptions of the model. In this study, however,

the model performed better when the animals were more closely “playing the same game

as the model” (Stephens and Krebs 1986: 204). Future applications of OFT to nonhuman

primates should also attempt to determine which assumptions are being violated and its

effects on model performance.

Even simple models may have relevance to nonhuman primate foraging behavior

(Barton and Whiten 1994). In this study, the classical prey model was applied to

seasonal time periods from weeks to months in length, and it is probable that at study

sites where a single individual could be followed for entire days, the model could be

doubly informative. Given the temporal and spatial heterogeneity of primate habitats, I

would suggest applying the model freshly on a weekly or even daily basis, to individual

primates and habitat types if sample sizes allow. Although not applied here, recent game-

theory applications attempting to predict the behavior of social foragers appear to be

especially relevant to group-living primates, and hopefully will generate enthusiasm

130 among primate researchers. Unfortunately, there are not (as of yet) social models as general as those from classical foraging theory, at least pertaining to diet choice

(Giraldeau and Caraco 2000). To date, the classical prey model has proven to be informative in animals as diverse as invertebrates and human hunter-gatherers (Sih and

Christensen 2001; Stephens and Krebs 1986), and it is hoped this study will encourage further applications of patch choice and exploitation models to other members of the

Order Primates.

CHAPTER VI

OPTIMAL FORAGING THEORY: THE SOCIAL PREY MODEL

INTRODUCTION

The classical prey model has proven to be a useful tool for investigating dietary choices in a wide range of foragers. This generality is clearly one of its main appeals; it is refreshing to have one model that can be utilized to investigate the feeding decisions of sea stars (Campbell 1987), Himalayans langurs, and human hunter-gatherers (Kaplan and

Hill 1992). Yet it is clear that the true optimum for an animal may differ from what the simple variables in the classical prey model predict. One such situation involves group feeding, where the actions of others may have an influence on an individual forager’s decisions (Giraldeau and Caraco 2000).

Giraldeau and Caraco (2000) provide a synthesis of social foraging theory, which utilizes mathematical models, often derived from game theory, to predict the behavior of animals feeding in groups. In relation to the classical prey model, they note that the presence of others may influence the encounter rates λ with, or profitability (ei/hi) of, given foods to a forager. In other words, these variables may not be constant, but instead

“exhibit a frequency dependence generated by the different group members’ dietary choices” (p. 222). Feeding competitors, for example, can reduce the abundance of food in a given area and may interfere with an individual forager’s ability to estimate encounter rates. In addition, for social animals which possess distinctive dominance 131 132 hierarchies, highly-profitable foods may be inconsistently available to individuals of lower rank. In lowland gray langurs at Ramnagar, Nepal, most patches are sized as to allow only a subset of the group to exploit them, and low-ranking females are less likely to have access to the most valued food resources (Koenig and others 1998; Koenig and

Borries 2001). In terms of patch choice, other foragers may depress and reduce the profitability of a given type--by increasing the handling time required to ingest a given amount of food-- making it less likely that all other resources could be “ranked” relative to it (Charnov, Orians, and Hyatt 1976). This would be especially important when the forager exploits many foods, some of which vary only subtly in terms of nutrients or handling time. This situation likely characterizes many primate foods.

Although not all these factors have been included in a single model as of yet, several investigators have tackled the question of diet selection in groups—the results of this combined work has been called “the social prey model” (Giraldeau and Caraco

2000:225). Heller (1980) utilized computer simulations to track the success of four differing tactics for a solitary forager in a depleting patch containing two food types of differing profitability (Heller 1980). Intake rates were presumed to decline as foods in the patch were consumed. The four tactics considered included: 1) specialist 1 – feed only on the more profitable type, 2) specialist 2 – feed only on the less profitable type, 3) expanding specialist – include only the more profitable food early in the patch bout but take both types (generalize) at a later point, and 4) generalist – take both food types as encountered from the beginning of the patch session.

133

After 500 runs for each tactic, Heller concluded that for short patch residence times, solitary animals fared better when specializing only on the more profitable food

(specialist 1). For intermediate patch residence times foragers did better when including the poorer item late in the patch bout (expanding specialist), and for long patch residence times improved their success by using a generalist strategy from the outset. Using rationale from the marginal value theorem (Charnov 1976a; Parker and Stuart 1976) and allowing for multiple patches to be visited, Heller expanded this argument to travel time and resource quality, the latter defined as the initial densities of the profitable and less profitable food within patch types. For short travel times or high resource quality, animals should utilize the specialist 1 strategy, for intermediate conditions the expanding specialist strategy, and for long travel times or poor resource quality the generalist strategy (Heller 1980).

Heller (1980) then modified the simulations to include two individuals feeding simultaneously within the patch. He found that in competitive situations the expanding specialist strategy was either indistinguishable from or superior to all other strategies regardless of travel times, patch residence times or initial food densities. Note that the expanding specialist strategy is a violation of the zero-one rule from the classical prey model (Stephens and Krebs 1986) or its patch choice analogue (Schoener 1987), which do not take into account patch depression.

Mitchell (1990) investigated situations analogous to those discussed by Heller

(1980). First, using a different model, he investigated the case of a solitary forager depleting two food types. Mitchell found that as patch residence time increases the

134

optimal behavior changes from specialization to partial preferences (small amounts of the

less profitable food) to complete generalization (taking the less profitable food whenever

it is encountered). He also noted that when patch residence time is held constant, the

optimal policy will change based on the initial abundance of the more profitable food. As the density of this resource increases, the predicted behavior goes from generalization to partial preferences to specialization. These results differ from Heller’s in that an

expanding specialist strategy is not predicted under any patch residence time or resource

quality for a solitary forager.

Using optimal control theory, Mitchell (1990) then analyzed the effects of adding

varying numbers of competitors in the patch, all of which behave as generalists. In these

competitive situations, the optimal forager qualitatively changes its behavior and acts as

an expanding specialist as argued by Heller (1980). Adding increasing numbers of

competitors to the patch lowers the critical density of the profitable food at which the

switch from specialist to generalist is predicted to occur (Mitchell 1990). Visser (1991)

analyzed this problem using a deterministic simulation model and, when expressed in a

common currency, gives essentially the same quantitative predictions as Mitchell (Visser

and Sjerps 1991).

Himalayan langurs were observed to feed on multiple foods within individual

patches, both singly and in groups, which allows the quantitative testing of predictions

outlined above. This represents the first direct field test of the social prey model with any

animal group.

135

METHODS

Behavioral observations

All behavioral observations were dictated into a cassette recorder between

December 2002 and December 2003 and subsequently transcribed. A different focal individual was chosen for each sample day (n = 54) and data were collected on each food

patch that was observed to be entered by this individual. Focal individuals were rotated

among non-adults, adult females, and adult males. A patch is defined as an area of food

concentration separated from other patches by areas with little or no food. In general,

each tree, shrub or herb clump can be considered a separate patch (Astrom, Lundberg,

and Danell 1990; Stephens and Krebs 1986). There are, however, some situations where multiple plants grow contiguously and an animal can feed simultaneously in more than one food source. These were considered single patches. Unless stated otherwise, all

patches included in this chapter involve a subset of the data where two or more food

types were taken from a single patch.

Because the length of time in which individuals could be followed varied extensively based on topography, feeding data were collected from other individuals chosen at random whenever the focal animal was not visible. Whenever possible, individual identification was recorded.

In addition, the number of individuals feeding within a patch at the beginning of a

136

feeding session was recorded. This number was updated whenever individuals entered or left the patch in question. Bite size refers to the number of food items (leaves, fruit, etc.) put into the mouth, and when number could not be deciphered, the average number of items per bite for that patch was later substituted. Periods when ingestion could not be observed were considered missing time and discarded (Grether, Palombit, and Rodman

1992). When the focal individual left a patch, it was followed, whenever possible, until it entered another food patch, and recording ceased only when the individual stopped feeding or moving (e.g., began resting, grooming, etc.). When necessary, observations were aided by binoculars, or, rarely, a spotting scope. These data allow calculation of intake over time in a second-by-second fashion for each patch and food type, as well as average travel time between food patches.

Food types were collected and weighed wet, field dried, and after laboratory drying, which was completed at Peabody Museum, Harvard University. Plant identifications were conducted by plant scientists at the Central Department of Botany,

Tribhuvan University, Kathmandu, Nepal.

Nutritional analysis and currency

Nutrient (crude protein, water soluble carbohydrate, lipids, hemicellulose) and non-

nutrient (cellulose, cutin, lignin, total tannins) analyses were conducted by the author on 55

Himalayan langur food types at the Nutritional Ecology Laboratory in the Department of

Anthropology, Peabody Museum, Harvard University (Conklin-Brittain, Wrangham, and

Hunt 1998; Wrangham, Conklin-Brittain, and Hunt 1998). Crude protein (CP) was

137

determined using the Kjeldahl procedure for total nitrogen and multiplying by 6.25 (Pierce

and Haenisch 1958) instead of using the 4.3 conversion factor (Conklin-Brittain and others

1999; Norconk and Conklin-Brittain 2004).

The detergent system of fiber analysis (Goering and van Soest 1970) as modified by

Robertson and van Soest (1980) was used to determine the neutral-detergent, or total cell