Diet and Feeding Ecology of Woolly Monkeys in a Western Amazonian Rain Forest

Transcription

1 International Journal of Primatology, Vol. 25, No. 4, August 2004 ( C 2004) Diet and Feeding Ecology of Woolly Monkeys in a Western Amazonian Rain Forest Anthony Di Fiore 1,2 Received November 17, 2003; accepted December 23, 2003 I investigated the diet and feeding ecology of two social groups of woolly monkeys (Lagothrix lagotricha poeppigii) in Yasuní National Park, Ecuador between April 1995 and March Woolly monkeys in Yasuní were predominantly frugivorous, with fruits comprising ca. 77% of the yearly diet; the next most common food type in the diet was insect and other animal prey. The fruit diet of woolly monkeys in Yasuní is the most diverse yet recorded for any ateline primate, including spider monkeys (Ateles), which are often regarded as ripe fruit specialists: 208 distinct morphospecies of fruits were consumed by woolly monkeys either during the study or during several preceding months of pilot work. Nonetheless, close to one-third of the yearly diet came from just 3 plant genera Inga, Ficus, and Spondias and only 20 genera each contributed to 1% of the diet. For one study group, the proportion of ripe fruit in the diet each month was correlated with the habitat-wide availability of this resource, a pattern evidenced by several other ateline species. However, the relationship was not apparent in the second study group. The modal party size for feeding bouts on all food types was a single monkey, and, contrary to reports for other atelines, neither feeding party size nor the total number of feeding minutes that groups spent in food patches was well predicted by patch size. Both results highlight the independent nature of woolly monkey foraging. Given that woolly monkeys and closely-related spider monkeys focus so heavily on ripe fruits, their very different patterns of social organization are intriguing and raise the question of just how their ecological strategies differ. Two important differences appear to be in the use of animal prey and in the 1 Department of Anthropology, New York University and NYCEP (New York Consortium in Evolutionary Primatology), New York. 2 To whom correspondence should be addressed at Department of Anthropology, New York University, Rufus Smith Hall, Room 802/803, 25 Waverly Place, New York, New York 10003; anthony.difiore@nyu.edu /04/ /0 C 2004 Plenum Publishing Corporation

2 768 Di Fiore phytochemical composition of the ripe fruits that they consume: spider monkeys rarely forage for animal prey, and woolly monkeys seldom consume the lipid-rich fruits that are an important part of spider monkey diets. KEY WORDS: woolly monkeys; Lagothrix; diet; feeding ecology; phenology; atelines; spider monkeys; Ateles. INTRODUCTION A good deal of our understanding of the links between primate foraging ecology and resource distribution has emerged from long-term studies of ateline primates (Chapman, 1988, 1990; Leighton and Leighton, 1982; Milton, 1980; Peres, 1994, 1996; Strier, 1986, 1989; Symington, 1987, 1988). Atelines may be particularly amenable subjects for this type of research because 1. they typically live in seasonal environments where food abundance and distribution can vary greatly over the course of a year, 2. all atelines are characterized by some degree of flexibility in their grouping patterns, which provides an important dimension along which behavioral responses to changing ecological conditions can be monitored (Strier, 1990, 1992), and 3. despite their close phylogenetic relatedness, the various ateline genera differ markedly in diet and foraging strategy which facilitates comparisons among taxa that are less confounded by phylogeny. Among the atelines, howlers (Alouatta) are primarily folivorous with very short day ranges and small home ranges (Crockett and Eisenberg, 1987; Milton, 1980). In contrast, spider monkeys (Ateles) are ripe fruit specialists (Klein and Klein, 1975; Symington, 1987; van Roosmalen, 1985; van Roosmalen and Klein, 1988), and their brachiating pattern of locomotion has been interpreted as an efficient means to rapidly travel between widely spaced patches of ripe fruits (Cant, 1977) between which little feeding occurs. Like howlers, muriquis (Brachyteles) concentrate heavily on leaves (Milton, 1980; Strier, 1989), but are similar to spider monkeys in their locomotor behavior, which permits them to include more fruits in the diet than sympatric howlers do (Strier, 1992). In all 4 ateline genera, various aspects of foraging ecology, e.g., dietary composition, foraging party size, feeding party size, all correlate well with ecological variables describing the distribution and habitat-wide abundance of resources (Chapman, 1990; Chapman et al., 1995; Leighton and Leighton, 1982; Strier, 1989, 1991; Symington, 1987, 1988). Until recently, woolly monkeys (Lagothrix) have been the least studied of the atelines. Prior reports of their foraging behavior classified them as primarily frugivorous (Defler and Defler, 1996; Peres, 1994; Soini, 1986), but in some areas they consume large quantities of arthropod prey (Stevenson, 1992; Stevenson et al., 1994). Despite early reports of subgrouping

3 Feeding Ecology of Woolly Monkeys 769 behavior, suggesting they may live in fission-fusion societies similar to those of closely-related spider monkeys and of chimpanzees (Pan) (Kavanaugh and Dresdale, 1975), long-term studies confirm that woolly monkeys live in socially-cohesive, multimale-multifemale groups typically containing between 20 and 40 individuals (Defler and Defler, 1996; Di Fiore, 1997; Nishimura, 1990; Peres, 1994; Ramirez, 1980, 1988; Stevenson et al., 1994). Thus, it is unclear to what extent the correlations between resource availability and foraging behavior that characterize some of the other atelines apply to Lagothrix. I describe the diet and feeding ecology of a population of woolly monkeys (Lagothrix lagotricha poeppigii) in Yasuní National Park, Ecuador in the western Amazon Basin. I also summarize data on the floristic composition of the habitat, document the way in which resource availability changes within this environment over time, and examine how particular aspects of the feeding ecology of woolly monkeys such as dietary composition, feeding party size, and food patch utilization are adjusted to changing ecological conditions. Finally, I compare the feeding ecology of woolly monkeys in Yasuní to that of other populations of Lagothrix and of other ateline primates. I also discuss the implications of the study particularly for understanding the comparative socioecology of woolly monkeys and a closely related competitor, spider monkeys. METHODS Study Area and Subjects Yasuní National Park, Ecuador, is a 900,000-ha reserve of lowland tropical rain forest in the upper Amazon Basin. The study site is south of the Río Napo, 47 km along a road built in 1993 and 1994 as part of petroleum development in the region (75 28 W, 0 42 S) (Fig. 1). The roughly 650-ha primate study site comprised mostly primary, terra firme forest spread out over a series of ridges and drainages feeding several small but permanent streams. With the exception of occasional hunting in the past by seminomadic parties of indigenous hunters and gatherers and several isolated hunting incidents near the start of the study, populations of the 10 primate species in the study plot were undisturbed by humans. Rainfall at the site between September 1994 and April 1996 averaged 267 mm per month (SD = 94 mm, range = 111 to 461 mm), and during the primary study period (from April 1995 to March 1996) a total of 3274 mm of rain fell. Although rainfall was variable across months, monthly precipitation never totaled <149 mm, making the region less markedly seasonal than some

4 770 Di Fiore Fig. 1. Geographic distribution of 4 subspecies of Lagothrix lagotricha and the locations of long-term studies of the species, based on Fooden (1963) and Defler and Defler (1996). Insets show Yasuní National Park with the location of the study site and a detailed overview of the set of trails and botanical transects within it. of other Amazonian sites where primates have been studied. Nonetheless, the site did undergo a marked dry and hot season from July to September of 1995 (Di Fiore and Rodman, 2001). At least 5 social groups of woolly monkeys, totaling ca. 110 individuals, used various portions of the study site; they were all of typical composition, containing 2 to 5 adult males, several subadult males, 9 to 11 reproductive age females, and 4 to 6 juveniles. The 2 groups that ranged nearest to the road were the primary focus of my study, though several other social groups were followed intermittently, and they contributed a handful of data to my characterization of the woolly monkey diet. Group 1 varied from 25 individuals at the start of the study to 24 individuals by the end, due to one death, one disappearance, and one birth. Group 2 had 23 individuals at both the start and end of the study as one individual was born and one individual was lost to a hunter (Di Fiore, 1997, 2001, Di Fiore and Rodman, 2001). Collection of Data on Diet and Feeding Ecology I collected systematic feeding data for 12 months beginning in April 1995 on one of my main study groups during two 5-day follows each month. The second main study group, habituated later, was sampled for 10 months, beginning in June Each 5-day follow consisted of 2 days of group scan

5 Feeding Ecology of Woolly Monkeys 771 sampling and 3 days of focal-animal sampling (Altmann, 1974). Two additional groups were also the subjects of sporadic scan sampling throughout the course of the study. During scan sampling, I collected scans at 10-min intervals between 0630 and 1730 h; scans lasted for 5 min and were followed by 5 min of inactivity. During a scan, I recorded the age and sex class (or individual identity, if possible) and behavior of each individual seen during the scan. I watched each monkey for 5 sec after it was first detected and noted its predominant behavior. I tried to collect behavioral records on as many individuals as possible during a scan by changing positions under the group frequently. However, groups were often spread out over a large area, so only a portion of the group was sampled during each scan (mean = 4.76, SD = 1.79, range = 1 to 15). Scan sampling yielded a total of 19,693 individual behavioral records during 4138 scans. Of them, 2165 scans were on group 1, 1566 scans on group 2, and the remainder either on other groups, on occasional multigroup associations, or on rarely-encountered solitary individuals. Of the total set of individual behavioral records, I classified 4074 as FEEDING RECORDS, during which a scanned monkeys processed or ingested food items. The number of feeding records collected in each month varied (mean = 339.5, SD = 76.9, range = 241 to 472), mainly as a result of increased hours of observation during the latter months of the study (Table I). For each feeding record, I noted the type of food consumed: fruit, flowers, leaves, seeds, insects. For feeding on plant items, I further noted the specific part consumed e.g., whole fruit, pericarp, mesocarp, aril, leaf petiole and the maturity of the item immature or mature whenever possible, and I collected the item for taxonomic identification if it was not immediately recognizable. Table I. Number of feeding records collected by month and group Month Group 1 Group 2 Other Total Apr May Jun Jul Aug Sep Oct Nov Dec Jan Feb Mar Total Average SD

6 772 Di Fiore The feeding records form the basis for the majority of my analyses; however, I supplement the data with additional information on FEEDING BOUTS collected whenever possible during both scan and focal animal sampling. A feeding bout is the more or less continuous occupation of a single food patch by 1 members of a woolly monkey group, where a patch is a discrete area within which food items occur, separated by areas of much lower food item density in which the monkeys did not feed (sensu: Leighton and Leighton, 1982). A food patch typically comprised a single tree crown bearing fruits, flowers, or new leaves, surrounded by other trees not showing comparable phenological activity. However, some food patches consisted either of a liana spanning several adjacent crowns or of 2 neighboring trees that were fed upon simultaneously by members of a group. For each feeding bout, I recorded the time at which the first animal entered the patch and then took censuses of the number of individuals feeding in the patch at specified intervals until the last feeder left. During scan sampling, I performed the tree crown censuses quickly at the start of each 5-min group scan, before searching for additional group members to sample, and again at the end of the group scan after reconnoitering the group, i.e., an interval duration of 5 min. During focal sampling, censuses were taken either at 1-, 2-, or 5-min intervals, depending on observation conditions and providing they could be made without interrupting my focal sampling schedule. For large-crowned feeding sources with many simultaneous feeders, I typically followed the 5-min interval schedule used during scan sampling; for smaller feeding sources or ones with fewer feeders, I would sample at more rapid 1- or 2-min intervals. I used the feeding patch censuses to calculate the total number of feeding-min a group spent in a patch by multiplying the mean number of feeders present during an interval by the interval duration and then summing across intervals. This method of calculating group feeding-min is similar to that used by Terborgh (1983) to determine the daily feeding times of various species of primates in Manu National Park, Perú, except that at Manu changes in the number of feeders in food patches were monitored continuously, while I monitored them only at set intervals. In some cases, all feeders would vacate a patch only to return a short time later; I therefore treated feeding bouts separated by <5 min of no feeding activity as the same bout and those separated by 5 min as separate bouts. In total, I collected data for 810 feeding bouts, 753 of which were on fruit resources. In some cases, a good portion of the crown in which monkeys were feeding was obscured from view, thus precluding an accurate census of the number of feeders. For these cases, I recorded the minimum number of feeders and used it to calculate group feeding-min. While this undoubtedly results in an underestimate of group feeding-min for these bouts, it is preferable to discarding the bout data entirely. The only bias that this should

7 Feeding Ecology of Woolly Monkeys 773 produce is to underestimate feeding time on the largest patches, which, by virtue of their size, are more likely to have portions of the crowns obscured. Additionally, for some bouts, I was unable to record the precise time when either the first feeder entered or the last feeder left the patch, either because I arrived late, when feeders were already present, or had to leave to follow a focal monkey before the last feeder left. Rather than discarding these censored data, I still calculated minimum total group feeding-min for the bouts. Again, the only systematic bias should be to underestimate group feedingmin for bouts in the largest patches because, by virtue of their size, there was a greater chance of there still being other feeders in the patch when I left. Clearly the individual feeding record and feeding bout data are not independent, at least for scan sampling days. Moreover, individual feeding records collected during scans are also not necessarily independent of one another because 1. during a single scan, records could be collected simultaneously on multiple individuals in the same patch, and 2. multiple records could be collected on the same individual for the same patch if it fed in the patch for 10 min. However, since I am interested primarily in the relative proportions of different food types in the diet and in the relative consumption of different plant species, the only bias that should result from the nonindependence of records is if consumption of particular types of food or of particular plant species were systematically more likely to be seen than others, which was unlikely. For all plant food patches that were fed in for more than 5 group feedingmin, I tagged and mapped the location of the patch relative to previously mapped trails and trees, measured the diameter at breast height (DBH) of the tree constituting the patch (or the DBH of the supporting trunk if the feeding source was a liana, epiphyte, or hemiepiphyte) as a measure of patch size, and, whenever possible, collected botanical specimens for identification and vouchering. For the few cases in which patches spanned multiple crowns, e.g., large lianas, I measured and summed the DBH of each tree supporting the life form comprising the patch and used the sum as my measure of patch size. Plant food patches were identified to genus (or to species if possible) by botanists familiar with the Amazonian flora of lowland Ecuador. In cases where in specific identifications could not be made, I assigned distinctive types within a genus to different morphospecies based on gross morphological features of their leaves and fruit pending future classification. Floristic Survey and Phenological Sampling In order to obtain an independent estimate of the density, distribution, and abundance of plant resources available, I conducted a study of forest

8 774 Di Fiore composition and phenology to complement the behavioral observations. The floristic study was based on 2934 trees 10 cm in DBH located within 5 1-ha belt transects placed randomly throughout the study area at the onset of the investigation, with the restriction that no 2 transects lay within 100 m of each other (Fig. 1). Each transect measured 10 m 1000 m and was oriented along a SE-NW axis. The alignment assured that the botanical transects ran perpendicular or oblique to the major ridges and drainages within the study plot, and thereby allowed me to sample the range of altitudinal microhabitats in the area. I divided each of the 5 floristic transects into m 10-m blocks. I selected one-half of the blocks on each transect for phenological monitoring according to a stratified design in which 2 of every consecutive 4 blocks were chosen at random as a phenology block. From April 1995 to March 1996, the crowns of between 1416 and 1459 trees in the 2.5 ha were inspected with binoculars, and the presence and abundance of new leaf flush, flowers, and fruits in either the crown itself or in any associated epiphytic plants, e.g., vines and lianas, stranglers, bromeliads, and aroids was noted. The actual number of trees monitored varied slightly from month to month due to loss of some trees to inclement weather or natural death. Moreover, because the growth on intervening crowns sometimes obscured visibility, crowns that were visible one month occasionally were impossible to monitor in subsequent months, or vice versa. The abundance of new foliage was estimated as a percentage of the total crown volume made up of new leaves. Flower abundance was estimated similarly, according to the percentage of crown volume (or trunk and branch surface area for ramiflorous or cauliflorous plants) in which flowers occurred. Because Lagothrix at other sites are primarily frugivorous (Defler and Defler, 1996; Peres, 1994; Soini, 1986; Stevenson et al., 1994). The actual crop size (number of fruits) of each transect plant bearing fruits was also estimated by making repeated counts of subsections of the crown and extrapolating from them to full crown volume. All phenological data were collected by trained botanists and field assistants concurrently with my behavioral data collection regime. I used the following procedures to derive indices characterizing the monthly habitat-wide availability of different plant phenophases. For new leaves and flowers, I first used DBH to estimate the crown volume of each monitored tree via a regression equation determined previously (Di Fiore, 1997); I then multiplied the crown volume by the proportion of the crown comprising new leaves or flowers, summed the values across the set of monitored trees, and divided this value by the summed crown volumes of all trees monitored. For immature and mature fruits, I took the logarithm of the number of fruits estimated in each fruit-bearing crown, summed the log scores across the set of fruiting trees, and divided it by the total number

9 Feeding Ecology of Woolly Monkeys 775 of trees monitored that month. These procedures allowed me to take into account differences in crown size and in the proportion of a crown bearing a phenophase (in the case of new leaves and flowers) and differences in actual crop size (in the case of immature and mature fruits) among monitored trees. The set of phenology trees included 185 genera belonging to 56 families, but not all of these genera were food resources for woolly monkeys. Therefore, to characterize the habitat-wide availability of immature and ripe fruits, I derived 2 indices for these phenophases using different subsets of the full set of trees. For one index, I used only trees belonging to genera eaten at all by woolly monkeys over the course of the study (N = 916 to 934 trees), while for the second I used only genera that contributed to 1% of the diet or to 1% of the total number of marked feeding trees (N = 482 to 490 trees). Because I was unable to determine the taxonomic identity of most leaves and flowers eaten by woolly monkeys during the study, I used the full set of focal trees monitored each month to characterize monthly flower availability, and I did the same for new leaves but omitted all palms (family Arecaceae) from the calculation because I never saw woolly monkeys eat palm leaves. RESULTS Overall Diet I determined diets of each of the 2 main study groups by first calculating the proportion of records that each group fed on the various food types each month and then derived the yearly budget by averaging across months to account for the unequal intermonthly distribution of feeding records diet (Fig. 2(A)). For group 1, the number of feeding records collected on the various age-sex classes was nearly in proportion to their representation in group (Chi-square test: χ 2 = 4.88, ns, comparing the distribution of records among adult male, subadult male, adult female, and juvenile agesex classes); for group 2, however, juveniles were somewhat undersampled and adult females oversampled relative to their proportional representation in the group (χ 2 = 23.05, P < 0.01). Differences between the 2 groups in overall dietary composition were not significant (Wilcoxon signed-ranks test comparing monthly proportions of major food types between groups: Z = 0.866, N = 10, P = for fruits; Z = 0.560, N = 10, P = for flowers; Z = 1.682, N = 10, P = for leaves; Z = 0.459, N = 10, P = for prey), so I present the remainder of the overall dietary results based for the pooled set of 4074 feeding records for all groups combined. Table II shows the breakdown of feeding records according to item type. Over 90% of food items were of plant origin (N = 3689 records). Of

10 776 Di Fiore Fig. 2. (A) Yearly diets of 2 groups of woolly monkeys based on feeding records. (B) Temporal variation in the diet of group 1 and (C) temporal variation in the diet of group 2 over the study period.

11 Feeding Ecology of Woolly Monkeys 777 Fig. 2. Continued the remainder of the diet, 4.0% of records (N = 164) were positively indentified as animal prey mainly insects, with a modicum of feeding on vertebrate prey; another 5.3% of records (N = 215) were inferred to be insect prey based on the behavior of the monkeys either immediately preceding or following the feeding observations, and <1% (N = 6) records were of feeding on arboreal fungi. I observed scanned monkeys drinking water directly in only 3 cases (not included in the 4074 total feeding records), each time all from the center of a large arboreal bromeliad. Feeding on fruits constituted 76.7% of the total diet (N = 3123 records) and 84.7% of the non-animal portion of the diet, based on the pooled feeding records for all groups. Of the fruit feeding records, 93.1% (N = 2908) were on ripe or nearly-ripe fruits that possessed either a soft and easily penetrated pericarp or a succulent mesocarp inside a tougher husk. Another 1.5% of records (N = 46) involved feeding on the exudates of ripe seeds of Parkia. For many normally dehiscent fruits, particularly from Virola, Otoba, and Trichilia, the monkeys would consume the aril of the fruit before dehiscence, generally biting through the husk and tearing the fruit open to reach the aril inside. Nonetheless, they always consumed them after the fruit had matured to a stage when the aril was already well-formed, possessing its mature color and when the seed was hard and unlikely to be subject to predation. They ate the remaining 5.4% of fruits (N = 169 records) at an unknown stage of maturity. I observed no instance of Lagothrix taking obviously unripe fruits,

12 778 Di Fiore Table II. Overall composition of the Lagothrix diet Food Type # of records % of records Fruits Ripe fruit pericarp Ripe fruit mesocarp Ripe fruit aril Ripe whole fruit Exudates Undetermined fruit parts Fruits total Flowers Immature seeds Leaves New leaves & vine shoots Intermediate & mature leaves Leaf petioles Leaf buds Undetermined leaf parts Leaves total Other vegetative parts Pith of Aroid Epiphytes Undetermined plant parts Bark Other plant total Fungi Animal prey Frogs Other vertebrates Ants and termites Other insects and larvae Presumed insects Total animal prey Total though they occasionally consumed immature seeds of 2 palms (Iriartea deltoidea and Socratea exhorriza), which together accounted for <1% (N = 22) of plant feeding records. Leaves and other non-reproductive plant parts at various stages of maturity constituted 9.8% of the overall diet (N = 401 records), with >60% of them (6.0% of the total diet) being either new leaves or the tendrils and growing points of epiphytic vines and another 9% of them (<1% of the total diet) being the succulent growing shoots of aroid epiphytes. Behind fruits and new leaves, flowers and recently fertilized infloresences of palms and Cecropia were the next most frequently consumed plant items (3.5%, N = 143 records). In the analysis above, I determined the relative proportions of the various food items in the overall diet using each feeding record as an independent data point, regardless of the month from which it came. If the overall diet is instead determined by first calculating separate diets for each month of the

13 Feeding Ecology of Woolly Monkeys 779 study and then taking the grand mean of the monthly diets, the proportions of different classes of food in the diet are essentially unchanged. Calculated in this manner, fruits, including exudates, comprised 76.2% of the diet; immature seeds, leaves, and other nonreproductive plant parts comprised 10.6%, flowers 3.6%, and animal prey 9.6% of the diet. Dietary Diversity The 3123 fruit feeding records represent a minimum of 147 morphospecies of plants belonging to 80 different genera and 45 different families (Table III). Another 12 genera (14 morphospecies) from 4 additional families were eaten only during focal samples or during several months of pilot work before the study, and 47 other morphologically distinct fruits, not yet identified, were also consumed during either scan or focal sampling. Conservatively scoring each of these as a separate morphospecies without assigning generic identity brings the total number of different fruit species in the diet to 208 utilized during a ca. 15-mo period. Subsequent fieldwork undertaken since the end of this systematic study has further increased the number of genera in the fruit diet by ca. 15, and thus the number of species by at least that many as well. Despite its great breadth, the woolly monkey fruit diet in Yasuní is nonetheless strongly biased towards a small number of important fruits. Based on the feeding record data, only 10 genera, comprising a minimum of 46 morphospecies from 10 families, account for >50% of all fruit feeding records, and >75% of all fruit feeding records come from just 30 genera in 22 families. In total, only 22 genera from 19 families individually contributed to >1% of fruit feeding records, and only the top 3 genera Inga, Ficus, and Spondias each accounted for >5% of all feeding records and together for 28.1% of the fruit diet (Table III). Typically, just one to a few genera from each family were important food sources. Results from the feeding bout data closely match those derived from fruit feeding records scored in scan samples. Only 20 genera from 18 families individually accounted for >1% of the total fruit feeding time recorded in feeding bouts, and 18 of them 20 also appear in the list of the 22 top genera from feeding records (Table IV). Based on feeding bouts, just 7 genera together account for >50% of total fruit feeding time, and 20 genera for >75%; again, Inga, Ficus, and Spondias are the top 3 genera, together accounting for 33.1% of feeding time (Table IV). Despite the similarity between the 2 main study groups in their proportional use of various food types fruits, flowers, leaves, animal prey there were marked differences between them in the specific sets of fruit resources

14 Table III. Plant genera used for fruit feeding by woolly monkeys in Yasuní National Park Months fruits consumed d % of total Min # of Life # feeding fruit feeding Apr-95 May-95 Jun-95 Jul-95 Aug-95 Sep-95 Oct-95 Nov-95 Dec-95 Jan-96 Feb-96 Mar-96 Family Genus a,b morphospecies form c Part eaten Pilot work records records Acanthaceae Mendoncia 1 L Mesocarp X Anacardiaceae Spondias 1 CT Pericarp X X X X X X Anacardiaceae Tapirira 1 CT Mesocarp Mar-95 X Annonaceae Dichlinanona cf. 1 CT Mesocarp X Annonaceae Duguetia 1 ST Mesocarp X X X Annonaceae Guatteria 5 CT Pericarp Feb-95 X X X X X Annonaceae Oxandra 1 CT Pericarp X X X Annonaceae Porcelia 1 CT Mesocarp X X Annonaceae Rollinia 1 CT Mesocarp X X Annonaceae Ruizodendron 1 CT Pericarp X Apocynaceae Lacmellea 1 CT Mesocarp X Araliaceae Schefflera 1 CT Whole fruit X X Areceae Attalea 1 ST Mesocarp X Areceae Iriartea 1 CT Mesocarp X X S X S XX S X S X S XX S X Areceae Socratea 1 CT Mesocarp X S X X Bombacaceae Matisia 2 ST Mesocarp Mar-95 X X X X X X Bombacaceae Quararibea 1 CT Mesocarp Jan-95 to Mar-95 X F F X X F Boraginaceae Cordia 1 CT Mesocarp Feb-95 X X X X X Burseraceae [unidentified] 1 Mesocarp X X Burseraceae Crepidospermum cf. 1 ST Mesocarp X Burseraceae Protium 3 CT Aril Jan-95 X X X Burseraceae Tetragastris 1 ST Aril X X Cecropiaceae Pourouma 4 CT Mesocarp X X X X X Celastraceae Maytenus 1 CT Aril X X Clusiaceae Garcinia 2 CT Mesocarp X X X Clusiaceae Quapoya cf. 1 CT Pericarp X Clusiaceae Tovomita 1 CT Aril Mar-95 Combretaceae Buchenavia 1 CT Mesocarp X X X Convolvulaceae Dicranostyles 1 L Mesocarp, Cotyledons X X X Cucurbitaceae [unidentified] 1 L Mesocarp X X X Cucurbitaceae Cayaponia 1 L Whole fruit X F Dichapetalaceae Tapura or Stephanopodium cf. 1 CT Mesocarp X Ebenaceae Diospyros 1 ST, CT Mesocarp X X X Euphorbiaceae Glycidendron 1 CT Pericarp X X Euphorbiaceae Hyeronima 1 CT Pericarp X X

15 Table III. Continued Months fruits consumed d % of total Min # of Life # feeding fruit feeding Apr-95 May-95 Jun-95 Jul-95 Aug-95 Sep-95 Oct-95 Nov-95 Dec-95 Jan-96 Feb-96 Mar-96 Family Genus a,b morphospecies form c Part eaten Pilot work records records Euphorbiaceae Paradrypetes 1 CT Pericarp Jan-95 X X Euphorbiaceae Richeria 1 CT Mesocarp X Fabaceae [unidentified] 1 Mesocarp X Fabaceae Cassia 1 CT Mesocarp X Fabaceae Dialium 1 CT Mesocarp X Fabaceae Enterolobium 1 CT Mesocarp X X X Fabaceae Inga 16 CT Mesocarp Jan-95 to Mar-95 X X X X X X X X Fabaceae Parkia 2 CT Exudate X X Flacourtiaceae Hasseltia 1 ST Pericarp X X X X X Flacourtiaceae Pleuranthodendron 1 ST Pericarp X X X Loganiaceae Strychnos 2 L Mesocarp X X X X X X Malpighiaceae Byrsonima 1 CT Pericarp X X X X X X Marcgraviaceae Souroubea 1 E Whole fruit X X Meliaceae Trichilia 4 CT Aril Jan-95, Mar-95 X X X X X X X X Menispermaceae Disciphania 1 L Mesocarp X Monimiaceae Siparuna 1 CT Undetermined fruit part X Moraceae Brosimum 1 CT Undetermined fruit part Nov-94 Moraceae Castilla 1 CT Mesocarp X Moraceae Clarisia 1 CT Pericarp X X Moraceae Ficus 15 H Whole fruit X X X X X X X X X Moraceae Maquira 1 CT Mesocarp X Moraceae Naucleopsis 2 CT Mesocarp X Moraceae Perebea 2 ST Pericarp X X X Moraceae Pseudolmedia 3 CT Pericarp Feb-95, Mar-95 X X Myristicaceae Iryanthera 1 CT Aril Mar-95 X Myristicaceae Otoba 1 CT Aril Mar-95 X X X X X X X Myristicaceae Virola 4 CT Aril X X X X X Myrsinaceae cf. Myrsine cf. 1 ST Pericarp X X Myrtaceae Eugenia 1 ST Pericarp X Myrtaceae Plinia 2 CT Mesocarp X X X X X Nyctaginaceae Guapira 1 CT Pericarp X X Opiliaceae Agonandra 1 ST Mesocarp X X X Passifloraceae Passiflora 2 L Mesocarp X Polygalaceae Moutabea 1 L Mesocarp X Polygonaceae Coccoloba 2 CT Pericarp Feb-95 X X X X Rhamnaceae Rhamnidium 1 CT Mesocarp X Rhamnaceae Zizyphus 1 CT Mesocarp X X Rosaceae Prunus 1 CT Pericarp Nov-94 X X Rubiaceae Borojoa 1 ST Mesocarp X X X X X Rubiaceae Coussarea 2 ST Mesocarp X X X X Rubiaceae Pentagonia 1 ST Mesocarp X X Rubiaceae Posoqueria 1 ST Mesocarp X X

16 Table III. Continued Months fruits consumed d % of total Min # of Life # feeding fruit feeding Apr-95 May-95 Jun-95 Jul-95 Aug-95 Sep-95 Oct-95 Nov-95 Dec-95 Jan-96 Feb-96 Mar-96 Family Genus a,b morphospecies form c Part eaten Pilot work records records Santalaceae Acanthosyris 1 CT Mesocarp X Sapindaceae Allophylus 3 ST, CT Mesocarp, Pericarp Nov-94 X X X X X Sapindaceae Matayba 1 CT Whole fruit X X Sapindaceae Paullinia 6 L Aril X X X X X X X Sapindaceae Talisia cf. 1 CT Mesocarp Mar-95 X X Sapotaceae Micropholis 1 CT Mesocarp X X X X Sapotaceae Pouteria 2 CT Mesocarp Jan-95 X X X Sapotaceae Pouteria or Chrysophyllum cf. 1 CT Mesocarp X X Simaroubaceae Simaba 1 CT Mesocarp X Simaroubaceae Simarouba 1 CT Pericarp X X Solanaceae [unidentified] 1 L Mesocarp X X Solanaceae Cestrum 1 CT Whole fruit X Sterculiaceae Gauzuma 1 CT Mesocarp Oct-94 X Sterculiaceae Theobroma 1 ST Mesocarp X Tiliaceae Apeiba 1 CT Mesocarp Oct-94 X X X Ulmaceae Celtis 1 L Mesocarp Feb-95 X X Verbenaceae Vitex 1 CT Mesocarp X Violaceae Leonia 2 ST Mesocarp Feb-95, Mar-95 X Vitaceae Cissus 1 L Mesocarp X X [unidentified] [unidentified] Total a Genera in boldface individually contributed to at least 1% of feeding records b Genera that are underlined were recorded either during focal samples only or during pilot work in the months prior to the start of systematic data collection c CT = canopy tree, ST = subcanopy tree, L = liana, H = hemiepiphyte, E = epiphyte d Months in which fruits or fruit parts were consumed are indicated by X; superscripts F and S indicate months in which flowers and immature seeds, respectively, of these genera were consumed

17 Feeding Ecology of Woolly Monkeys 783 Table IV. Plant genera individually contributing to at least 1% of fruit feeding bout time and their ranks based on feeding records % of total fruit # of # of Rank based on Genus Family feeding bout time minutes bouts feeding records Inga Fabaceae Ficus Moraceae Spondias Anacardiaceae Otoba Myristicaceae Trichilia Meliaceae Coccoloba Polyganaceae Byrsonima Malpighiaceae Pourouma Cecropiaceae Quararibea Bombacaceae Plinia Myrtaceae Paullinia Sapindaceae Guatteria Annonaceae /13 Parkia Fabaceae Pouteria Sapotaceae Talisia cf. Sapindaceae Cordia Boraginaceae Hyeronima Euphorbiaceae Cissus Vitaceae Buchenavia Combretaceae Strychnos Loganiaceae /22 Total used (Table V), particularly in fruits that were rarely consumed. Nonetheless, 11 common genera each contributed to 1% of the feeding records scored for each group. Probably the reversal of the top 2 genera Inga and Ficus between group 1 and group 2 is a result of the fact that group 2 was not sampled systematically (and no feeding individuals were recorded) during the first 2 months of the study, a period of heavy concentration on Inga by group 1. An alternative explanation is that the 2 genera were present at different densities within the ranges of the 2 groups, a condition that is certainly true for many of the plant genera that feature less prominently than Inga in the fruit diet. Unfortunately, the large overlap in home range between adjacent woolly monkey groups (>45%: Di Fiore, 2003), the way in which the floristic composition of the site was evaluated (using belt transects that cut across several groups ranges), and the overall low density of Ficus in the site (Di Fiore, 1997) makes it difficult to address the alternative possibility even qualitatively. Characteristics of Feeding Bouts, Feeding Parties, and Feeding Sources Feeding bouts on fruits lasted an average of 26.4 group feeding-min (range = 1 to 522, SD = 43.3 min, N = 753 feeding bouts, including both

18 784 Di Fiore Table V. Plant genera contributing to at least 1% of the fruit feeding records for each study group individually Group 1 Group 2 % of Rank of genus % of Rank of genus Rank Genus a records for Group 2 Rank Genus a records for Group 1 1 Inga Ficus Ficus Inga Spondias Otoba /5 Trichilia 5.5 7/8/9 4 Spondias /5 Quararibea Coccoloba Pourouma Guatteria Coccoloba /8/9 Trichilia Matisia 3.2 7/8/9 Pouteria Otoba /8/9 Cissus /11/12 Iriartea /15 10 Byrsonima /11/12 Paullinia Hyeronima /11/12 Plinia Garcinia Parkia Pourouma Tapirira /15 Iriartea /11/12 15/16 Guatteria /15 Virola /16 Strychnos /17 Cordia /18/19 17/18/19 Cordia /17 16/17 Schefflera /18/19 Pentagonia Plinia /11/12 17/18/19 Tetragastris /20 Buchenavia /20 Allophylus /22/23/24 Talisia cf /22/23/24 Pseudolmedia /22/23/24 Guapira /22/23/24 Zizyphus 1.1 a Genera in boldface individually contributed to at least 1% of records for both groups. censored and noncensored bouts). Approximately 30% of the bouts lasted 5 group feeding-min, while only 12% lasted >60 min. For the smaller set of fruit feeding bouts where neither the beginning nor the end points were censored, the mean bout duration was only 23.7 group feeding-min (range = 1 to 229 min, SD = 33.3 min, N = 445 bouts), while the mean for censored bouts was higher (31.0 min, SD = 54.3 min, N = 308 bouts), which is consistent with the suggestion that bouts in larger patches, which tended to be longer, were also more likely to be censored. Woolly monkeys tended to feed longer on flowers (mean = 48.0 min, SD = 49.6 min, N = 18 bouts) and shorter on leaves (mean = 9.7 min, SD = 5.7 min, N = 18 bouts) than they did on fruits. Among fruit feeding bouts, those on figs (Ficus sp.) averaged much longer than bouts on other fruit types (mean = 67.1 min, SD = 61.3 min, N = 31 bouts). Feeding party size, which I defined as the maximum number of simultaneous feeders observed in a single food patch, varied from 1 to 12 individuals. The mean feeding party size, regardless of food type, was 2.66 individuals

19 Feeding Ecology of Woolly Monkeys 785 (SD = 1.88, N = 810 bouts), while the modal party (34% of all bouts) was a single individual. Mean and modal party sizes were similar for the smaller set of only fruit feeding bouts (mean = 2.66, SD = 1.87; mode = 1, seen in 35% of bouts), though during bouts of feeding on figs, party sizes were considerably larger (mean = 4.58, SD = 2.45). Feeding party sizes generally peaked several minutes after the start of a feeding bout, thus recorded feeding party sizes are unlikely to have been influenced by brief censoring of either the beginning or end of a bout. Feeding in marked feeding sources those fed in for >5 group feedingmin accounted for 88% of the total set of fruit feeding records (N = 2747 out of 3123 records) and 91.6% of fruit feeding bouts (N = 690 out of 753 bouts). Marked feeding trees were, on average, much larger in diameter and possessed much larger crowns than those in the forest as a whole: the mean DBH for 799 marked fruit feeding trees and trunks supporting epiphytic fruit sources was 43.5 cm (range = 7.4 to cm, SD = 23.8 cm), versus that of 20.6 cm (range = 10.0 to cm, SD = 14.2 cm) for 2896 measured trees on the phenology transects. Trees (excluding figs) in which fruit feeding occurred were larger than those supporting bouts of leaf feeding (leaf feeding tree mean DBH = 26.7 cm, SD = 17.4, N = 45), and the average size of the supporting trunk for hemiepiphytic figs was ca. 75% larger than the average for other fruit sources. Similarly, feeding on the exudate of pods of Parkia, a major item in the diet of Lagothrix at another Amazonian site (Peres, 1994), also generally took place in large-diameter, overstory trees, as did bouts of flower feeding. The maximum number of feeders per fruit patch, the size of fruit patches, and the total group feeding-min spent in each fruit patch all varied over the course of the year. Monthly mean party size and monthly mean group feeding-min per fruit patch were related (Spearman rank correlation: r s = 0.888, N = 12, P < 0.005), but neither was significantly correlated with mean patch size (r s = 0.336, N = 12, P = for party size; r s = 0.203, N = 12, P = for group feeding-min). Temporal Variation in Resource Availability and Its Influence on Diet The habitat-wide abundance of young leaves, flowers, immature fruit, and ripe fruit all varied over the course of the study (Fig. 3). The two indices of fruit abundance (based on all genera in the diet versus 1% genera) were highly correlated (Spearman rank correlation: r s = 0.902, N = 12, P < for ripe fruits; r s = 0.993, N = 12, P < for immature fruits) and indicated a similar pattern of seasonal variation in fruit abundance. Not

20 786 Di Fiore Fig. 3. (A) Temporal variation in 2 indices describing the habitat-wide availability of ripe and immature fruits. (B) Temporal variation in indices describing the habitat-wide availability of flowers and new leaves.

21 Feeding Ecology of Woolly Monkeys 787 surprisingly, ripe fruit availability peaked shortly after the peak availability of immature fruit. The availability of new leaves peaked in August and September of 1995 during the period of lowest availability of ripe fruits (Fig. 3); the second, higher peak is anomalous, entirely due to a single enormous tree (Dussia tessmannii), the largest in the botanical transects, that dropped all its leaves in the preceding months and then flushed new foliage throughout the crown in November As with new leaves, the abundance of flowers also peaked during the season of lower ripe fruit availability. Although there was no clear relationship between rainfall and the abundance of either leaves, flowers, or fruits, the period of lowest ripe fruit abundance coincided directly with the 2 driest and hottest months of the year. In the combined data set of all feeding records, ripe fruits constituted the bulk of woolly monkey diet across the year, varying between 64% and 89% of the monthly diet. For these combined data, the proportion of ripe fruit in the diet followed the same general pattern as ripe fruit availability in the study area: high initially, then dropping during the central portion of the study and increasing again towards the end. However, the most precipitous drop in the proportion of fruits in the diet came between July 1995 and August 1995, 2 months later than the sharpest decline in fruit availability. Indeed, the proportion of fruit in the diet was actually highest during July 1995, one of the months during which ripe fruit was most scarce. Thus the monthly proportion of fruit in the diet was not significantly correlated with the index of ripe fruit abundance based on all genera in the diet (Spearman rank correlation: r s = 0.385, N = 12, P = 0.202), though it was correlated with the index derived from the smaller set of genera contributing to 1% of the diet or 1% of marked feeding trees (r s = 0.601, N = 12, P < 0.05). The feeding records for the 2 main study groups indicate similar overall patterns of dietary change across the year. (Fig. 2(B) and (C)), except for the intensive flower use by group 2 during December 1995 due to their exploitation of flowers of Cayaponia, a resource that group 1 did not consume at that time. Moreover, the dietary proportions of all major food classes showed positive though nonsignificant correlations between groups, providing support for the idea that across the habitat, seasonal changes in the diets of the 2 groups were related (Spearman rank correlation: r s = 0.527, N = 10, P = for fruits; r s = 0.567, N = 10, P = for flowers; r s = 0.632, N = 10, P < for leaves; r s = 0.539, N = 10, P = for prey). However, the relationship between habitat-wide fruit availability and the percentage of fruit in the diet differed between groups: group 1 exhibited a clear, positive association between the variables (r s = 0.587, N = 12, P = for the index based on all consumed genera; r s = 0.657, N = 12, P < 0.05 for the 1% genera), while no association was evident for group 2 (r s = 0.030, N = 12,

22 788 Di Fiore P = for all consumed genera; r s = 0.333, N = 12, P = for the 1% genera). For both groups, the proportion of flowers in the diet was significantly correlated with the habitat-wide abundance of flowers (r s = 0.776, N = 12, P < 0.01 for group 1; r s = 0.656, N = 10, P < 0.05 for group 2). The proportion of leaves in the diet of each group each month was also positively associated with the overall availability of new leaves, though for group 1 this relationship only approached significance (r s = 0.515, N = 12, P = for group 1; r s = 0.661, N = 10, P < 0.05 for group 2). Patch Size, Party Size, and Feeding Bout Duration I used simple linear regression to examine the relationship between DBH as an index of patch size and feeding party size defined as the maximum number of simultaneous feeders recorded in the feeding source during the bout; both variables are log transformed. While DBH was a significant predictor of feeding party size, there was notable scatter in the relationship, with differences in patch size explaining only 6.7% of the variance in feeding party size for all plant types combined (F 1,719 = 51.9, P < 0.001) and 6.8% of the variance in feeding party size for bouts on ripe fruits (F 1,675 = 49.6, P < 0.001). Using both uncensored and censored bouts, similar regressions of log (group feeding-min), a measure of feeding bout duration, on log (DBH) revealed that patch size explained 7.4% of the variance in feeding bout duration for all plant items (F 1,719 = 57.4, P < 0.001) and 7.0% of the variance in the length of fruit-feeding bouts (F 1,675 = 51.1, P < 0.001), also with notable scatter in the relationship. Repeating the analysis using the smaller set of only uncensored bouts yielded the same result: patch size was a significant predictor of bout duration, but explained very little of the variance (F 1,379 = 15.3, P < 0.001, R = for all bouts; F 1,356 = 13.0, P < 0.001, R 2 = for fruit-feeding bouts). Feeding party size and group feedingmin were not significantly related to patch size for bouts of either flower or leaf eating (R 2 = 0.187, F 1,14 = 3.224, ns for log (party size) on log (DBH) for flowers; R 2 = 0.117, F 1,14 = 1.856, ns for log (GFM) on log (DBH) for flowers; R 2 = 0.016, F 1,16 = 0.257, ns for log (party size) on log (DBH) for leaves; R 2 = 0.050, F 1,16 = 0.844, ns for log (GFM) on log (DBH) for leaves). DISCUSSION Woolly Monkey Foraging Ecology Compared to Other Atelines For woolly monkeys in Yasuní National Park, Ecuador, ripe fruit comprised roughly three-fourths of the yearly diet, and they ate fruits from 200 species of woody plants during a period ca. 15 mo; this represents the most

23 Feeding Ecology of Woolly Monkeys 789 diverse fruit diet yet recorded for any ateline species. For one of 2 main study groups, the proportion of the diet comprising ripe fruit was positively associated with the habitat-wide availability of ripe fruits, but the pattern did not characterize the second study group. Although juveniles were represented less often and females more often than expected in the feeding records collected on the latter group, this factor is unlikely to explain the difference between groups. If anything, we might predict that the relationship between ripe fruit availability and the fruit proportion of the diet would be stronger in group 2 in which adult females the age-sex class expected to be most limited by access to high-quality resources (Trivers, 1972) contributed more to the estimation of monthly diets. For both study groups, the proportional representation of alternative plant foods e.g., flowers, new leaves in the diet was positively associated with their availability, which tended to be greatest during hotter and drier months when ripe fruits were less abundant. For several other populations of ateline primates living in seasonal environments, the proportion of fruit in the diet correlates well with fruit availability, with fruit consumption dropping as fruits become more rare. For example, grey woolly monkeys (Lagothrix lagotricha cana) in a terra firme forest of the central Amazon showed a clear positive correlation between the proportion of ripe fruit in the diet and both the density of trees bearing ripe fruit and a phenological index of availability that weights each fruiting tree by a subjective abundance score (Peres, 1994). Similarly, Symington (1987) found that the proportion of ripe fruits in the diet of black spider monkeys (Ateles chamek) in Manu National Park, Perú was significantly positively correlated with fruit availability, and Strier (1991) found the same is true for muriquis (Brachyteles hypoxanthus). In contrast, the study of Lagothrix lagotricha lugens by Stevenson et al. (1994) revealed no clear relationship between the proportion of ripe fruit in the diet each month and the number of trees bearing ripe fruit that month, an alternative measure of fruit availability. The disparate results concerning the relationship between diet and fruit availability for the 2 Yasuní groups, and the variation in the nature of this relationship within and between ateline genera, serve to highlight the fact that there can be substantial variation in primate ecological strategies both within and between populations. The lack of a tight relationship between patch size and party size in Yasuní Lagothrix lagotricha poeppigii contrasts with results of similar analyses with spider monkeys, muriquis, and another population of woolly monkeys (Peres, 1996; Strier, 1989; Symington, 1988). For example, patch size explained 36% of the variance in feeding party size for Ateles chamek (Symington, 1988) and 43% of variance in party size for Lagothrix lagotricha cana feeding on fruits and flowers (Peres, 1996), versus <7% of the variance in Yasuní feeding party size. The different result that I obtained may be due