The results indicate that when 'place' was predictable, the capuchins learned the spatial locations of food and nonfood sites rapidly (one-trial learning). In a second experiment, the positions of baited feeding sites were random. In the absence of other information, the capuchins used the presence of a local landmark cue (yellow block) placed at reward platforms to select feeding sites. In a final experiment, there was evidence that expectations regarding the amount of food available at a platform (2 bananas vs. 1/2 banana) had a significant influence on capuchin foraging decisions. Although the capuchins were extremely sensitive to changes in experimental conditions, when given conflicting cues, spatial information was predominant over other information in selecting feeding sites.
Key Words: cognition, foraging decisions, primates, feeding ecology, Costa Rica

Introduction
Ecological approaches to the study of learning provide important insight into the evolution of sensory capabilities and foraging behaviour [1-6]. In order to forage efficiently, an animal must be able to retain and integrate information regarding changes in the spatial and temporal availability of food resources. The information available to the forager, however, is likely to vary based on food type, the size and quality of the food patch, resource predictability, the presence/absence of potential competitors, and its own cognitive skills. Cues that may be used by one species to locate or select feeding sites, may provide little information to a second species [7]. For example, rufous hummingbirds (Selasphorus rufus) readily learn to associate spatial information (location of food source) with visual cues (a red light placed at some distance from the food source) to locate baited feeding sites [4]. In contrast, this type of visual cue is less salient to rats, who more easily associate olfactory cues with spatial information when foraging [8].
In the case of nonhuman primates, data from both captive and field settings offer evidence of species-level differences in foraging patterns, and the degree to which individuals and groups use vocal, visual, olfactory, and spatial information to coordinate travel and locate feeding sites [1,6, 9-15]. Tropical forests are characterized by highly complex, fluctuating, and seasonal patterns of food production [10,16]. The availability of feeding sites can vary on a time scale of hours, days, or months, and be distributed spatially across the landscape as within-patch, between-patch, and core area (preferential exploitation of certain zones of the home range) foraging choices. In some instances, the decision to feed in one patch may preclude the use of other patches during the same day [17]. Despite the existence of marked spatial-temporal heterogeneity in resource availability, several species of primates are reported to be highly efficient at finding rare and ephemeral foods [1,10,12,13, 18-28]. In most instances, however, there is little direct information on how feeding sites are located, and whether species differ in their hierarchical or ordered use of spatial, temporal, visual, and olfactory information. This results from the fact that traditional methods in primatological field research have rarely included the level of controlled experimental design required to test hypotheses regarding problem-solving skills and learning under natural conditions. Experimental field studies, however, offer an opportunity to control the level of environmental information available to the forager, and to examine directly species "differences in spatial learning, the development of foraging rules, and the hierarchy of perceptual cues used .... in making foraging decisions [6]." In this paper we report on a series of field experiments designed to examine the kinds of information white-faced capuchins (Cebus capucinus) use in selecting feeding sites (fig. 1). All species of capuchins exploit a diet of fruits, nuts, insects (including both adult and larval forms), and vertebrate prey [29-35]. Cebus has been described as a manipulative forager, and consumes foods that are hidden in tree holes, rotting wood, branches, termite nests, the base of palm fronds, bromeliads, and other epiphytes, and embedded under bark [22,30,32,36-38]. These resources are obtained through a process of extractive foraging that require fine sensory and motor control [39-44].

Evidence from both captivity and the wild indicates that capuchins have the highest degree of manual prehension and pollicial opposability among New World primates and exhibit more frequent and advanced tool using skills than any other species of monkey [43,45-48]. In their natural habitat, capuchins are reported to crack open palm nuts and hard fruits by striking them against an anvil [37,45]. At La Suerte Biological Field Station in Costa Rica, white-faced capuchins have been observed to use twigs to probe holes in tree branches searching for prey (Sara Garber, pers comm). In addition, these primates are noteworthy in their high degree of encephalization and complex patterning of cerebral fissures [49]. Enlargement of the neocortex in cebus is so extreme that when allometrically corrected for body size, they have a neocortex that is comparable to that of the chimpanzee [50]. The behavioural implications of neocortical expansion and visually directed object manipulation in wild capuchins has not been fully examined. However, Gibson [42: pg 112] has suggested that expansion of the visual, motor, and sensory cortices and "differential enlargement of neocortical association areas" in some primate lineages may have led to an enhanced "ability to construct relationships between multiple objects and multiple actions." In this study, we address questions concerning relationships between foraging, cognition, and learning in wild capuchins. This was accomplished by testing their ability to use visual, spatial, olfactory, and quantity information to predict the presence of food rewards at multiple feeding sites. Specifically, using a series of controlled field experiments we examined (1) the rate at which white-faced capuchins learn the spatial positions of feedings sites, (2) whether local landmark cues can be used to identify the presence of concealed food items, (3) the degree to which expectations concerning the amount of food can affect patch choice and foraging decisions, and (4) hierarchical patterns of decision-making (that is, when given conflicting information which cues are relied upon to a greater degree than others).

Methods
Study Site
Field experiments were conducted on a troop of wild white-faced capuchin monkeys at La Suerte Biological Research Station in northeastern Costa Rica (10° 26'N, 83° 47'W). The site represents an area of wet tropical lowland forest located along the La Suerte River. Average yearly rainfall in this region is 3962 mm, with a mild dry season extending from February until April [51]. In addition to white-faced capuchins, the site also supports populations of mantled howler monkeys (Alouatta palliata) and black-handed spider monkeys (Ateles geoffroyi). The home range of the study troop was part of a 20 hectare advanced secondary forest that was selectively logged in the early 1970's. Since then, the forest has been left relatively undisturbed and protected. Common adult tree species at the site include Pentacelthra macroloba (Mimosaceae), Goethalsia meiantha (Tiliaceae), Calatola costaricensis (Icacinaceae), Pithecelobium longifolium (Mimosaceae), Ceiba pentandra (Bombacaceae), Apeiba aspera (Tiliaceae), Bactris gasipaes (Palmae), and Hampea appendiculata (Malvaceae).
Research Protocol
In June and July 1994, and continuing in early December 1994 we began following and habituating a troop of 10-12 capuchin monkeys at La Suerte. From these initial observations we were able to identify an area in the troop's home range suitable for conducting field experiments. The area selected offered high visibility and was located in a part of the forest commonly traversed by the capuchins.
The field experiments were designed to present the capuchins with a schedule of food availability that was analogous to patterns of resource availability they encounter naturally in the wild. The information available to the capuchins was controlled, however, and therefore we could test a series of hypotheses regarding the use of spatial information and perceptual cues in selecting feeding sites. Two programs of prebaiting were initiated. During the first prebaiting period, five feeding platforms were constructed in the study area. These platforms were baited with bananas twice daily for a period of 22 days. The prebaiting period functioned to habituate the capuchins to the feeding platforms and allowed us to collect base line data on capuchin social behaviour, foraging activities, and vocalizations.

A blind built of wood, palm thatch, and dark plastic was placed approximately 10 meters from the platforms. The blind offered an excellent vantage from which to collect data, and served to minimize any effects that observer presence might have on the behaviour of capuchins at the feeding platforms.

In the second prebaiting period, we replaced the 5 initial platforms with 13 new platforms. Two bananas were placed on each platform twice daily for a period of 3 days. This was done so that group members could explore all feeding sites prior to the first day of the field experiments. Platforms measured 1280 cm2 in area and were raised approximately 1.5 meters above the ground. The platforms were fixed in location and used in all subsequent feeding experiments. One group of 7 platforms (Station 1) was arranged in an oval configuration and located 15 meters south of the blind (fig.2). On average, platforms were positioned 3.2 meters apart. A second group of 6 platforms (Station 2) also arranged in an oval configuration, was located 10 meters north of the blind (fig. 2). Distances between platforms at Station 2 averaged 2.8 meters. At both Station 1 and 2, platforms were positioned in order to take advantage of natural access routes provided by the surrounding vegetation. In these experiments, a Feeding Station is analogous to a food patch, each with its own number of feeding platforms, pattern of food distribution, and rate of resource renewal.

No attempt was made to trap or mark the animals, but we were able to recognize certain individuals based on distinctive facial markings, body size (the largest animal in the group was an adult male), and/or presence of dependent offspring (by the end of the study, 3 females in the group were carrying and nursing infants). Information on idiosyncratic patterns of behaviour and the selection of platforms among members of the study group was collected ad libitum.

Feeding Experiments
A series of feeding experiments was designed to identify the ability of capuchins to use (a) spatial information, (b) visual cues, (c) olfactory cues, and (d) expectations regarding the amount of food at a feeding site in foraging decisions. Platforms were baited twice daily (5:15 AM - hereafter referred to, as morning session; and 9:00am - hereafter referred to, as afternoon session) according to a research protocol in which 3 of 7 platforms at Station 1 each contained 2 real bananas, and 2 of 6 platforms at Station 2 each contained 2 real bananas. The real bananas were not peeled. The remaining platforms each contained 2 plastic bananas. The plastic bananas were similar in size, shape, and colour to the real bananas. Although subtle visual differences between plastic and real bananas could be detected by human observers in full sunlight, in the shadowy and poorly illuminated conditions of the forest understory, visual discrimination between real and plastic bananas was considerable less reliable. During certain experiments, banana skins were included with plastic bananas. This was done to minimize or eliminate differences in olfactory information available to the forager.
Quantitative data were collected over a period of 40 consecutive days. Experiment 1 was divided into 3 test conditions, each lasting 5 days (total 15 days). The purpose of this experiment was to examine the ability of capuchins to use spatial information to predict the location of baited (real banana) and sham (plastic banana) feeding platforms. The spatial positions of real and sham feeding sites varied across each of the three 5 day test periods, but were constant within each 5 day test period. In Experiment 2 (total 13 days) platforms containing real and sham bananas were assigned randomly (place was no longer predictable), and a visual cue (a yellow block) was placed only at platforms containing real bananas. This experiment was designed to test the ability of capuchins to use a local landmark or nearby associative cue to predict the location of platforms containing concealed food rewards. In Experiment 3 (12 days) we tested the ability of capuchins to associate place with the presence, absence, and quantity of food at a feeding site (plastic banana vs. 1/2 banana vs. 2 bananas). The details of each feeding experiment are presented in table 1.

During the field experiments, all quantitative observations of capuchin behaviour were made from a blind. When concealed in the blind, the researcher could view both Feeding Stations and all feeding platforms simultaneously. Over 400 hours were spent in the blind. An additional 120 hours of quantitative data on diet, ranging, and positional behaviour were collected during periods of the day when the monkeys were exploiting other areas of their range.

The placement and configuration of Feeding Stations were designed to insure that once an animal located a Feeding Station, it had an equal probability of visiting any or all of the feeding platforms. A capuchin was recorded as visiting a platform if it was observed to (1) sit or stand directly on a platform and search for food, or (2) hang by tail and/or limbs on a substrate adjacent to a platform and search the platform for food. Visits to platforms containing real bananas were scored as correct choices, whereas those to platforms containing plastic bananas were scored as incorrect choices. Revisits to the same platform or visits by more than one animal to a platform during the same morning or afternoon session were noted, but scored as a single visit. Chi Square Tests were used to determine whether the number of correct choices differed significantly from random.

Results
General Feeding Behaviour
The monkeys located the feeding platforms on the second day of the initial prebaiting period. Some individuals habituated rapidly to the platforms and fed daily. Other individuals appeared to be reluctant to descend onto the platforms and vocalized repeatedly from a height of 3-5 meters above the platforms. Over time, however, adult males, adult females with infants, and juveniles were all observed to feed from the platforms.
On the first few days of the experiments, all visits to feeding platforms involved the use of arboreal access routes. By Day 3, however, animals were observed to travel on the ground as well as use arboreal pathways to reach particular feeding platforms. In general, only one capuchin visited a given platform at a time, and there was little evidence of direct or aggressive competition for access to baited feeding sites. Nevertheless, the capuchins rarely fed while positioned on a platform. Rather, they carried 1 or 2 whole bananas into the canopy to feed. Even under conditions in which only 1/2 banana was present on a platform (Experiment 3), the monkeys always transported the food to a higher location in the canopy to feed.

Over the course of the three feeding experiments, the capuchins visited the Feeding Stations on 39 of 40 days. At Station 1, the total number of platforms visited was 322. At Station 2, the total number of platforms visited was 229. On average, the capuchins visited 4.3 (±1.6) platforms per session at Station 1 and 3.2 (±1.6) platforms per session at Station 2.

The behaviour of the capuchins at the Feeding Stations suggest that each was treated as a distinct food patch. This is supported not only by the ability of the capuchins to learn to use spatial and quantity information that was specific to each Station (see below), but in addition, on 9 occasions the capuchins visited platforms at Station 1 but did not visit any platform at Station 2. Moreover, it was not uncommon for all or most group members to concentrate their foraging activities on several platforms at one Feeding Station before visiting platforms at the other Station. Finally, on several occasions, the capuchins visited one Station, fed, rested for a period of time, and then travelled to the other Station to feed.

As indicated in table 2, there was evidence of significant temporal differences in the manner and efficiency in which the capuchins exploited the Feeding Stations. Over the first 10 days of the study, (Exp. 1A and B) latency from the time the animals first arrived at the Feeding Station until the time they first fed was approximately 15 minutes. During the remaining 29 test days, latency dropped to less than 7 minutes. Similarly, the length of time members of the study group remained in the immediate vicinity of the Feeding Stations after feeding was concluded decreased from 38.1 minutes in Experiment 1A to 5.6 minutes in Experiment 3 (table 2). Capuchin feeding rates (i.e. the number of reward platforms visited per time spent at the Feeding Stations) showed a general pattern of increase throughout the study (table 2).

Experiment 1
In Experiment 1A, 3 of 7 platforms at Station 1 and 2 of 6 platforms at Station 2 had real bananas. The probability of encountering a platform that contained a food reward by chance was .384 (5/13). The information available to the monkeys in selecting particular feeding platforms included spatial cues (place was constant), visual cues (subtle differences in the appearance of real and plastic bananas), and olfactory cues (potential differences in the smell of plastic and real bananas). On the first day of the experiment, the capuchins travelled to both Feeding Stations, and visited a total of 10 different platforms. Sixty per cent of these visits were to real sites and 40% to sham sites (fig. 3). During this 5 day experimental period, 79.4% of visits were to Station 1 and 20.6% were to Station 2. Overall, 85.1% (23/27) of the visits to Station 1 and 85.3% (6/7) of the visits to Station 2 were to platforms containing real bananas. Combined, these values are significantly above chance levels (X2=19.7, df=1, p<.01). Figure 3 illustrates the cumulative increase in the frequency at which troop members learned the location of banana baited platforms.
Experiment 1A, was not designed to identify directly the types of information capuchins used in selecting feeding sites. This was addressed in Experiments 1B and 1C. On the first day of Experiment 1B, the locations of real and sham food rewards were rotated (given the number of platforms and our protocol, 3 platforms that did not contain food in Experiment 1A also did not contain food in Experiment 1B). If visual and/or olfactory cues were the primary means by which capuchins identified feeding sites, then we expect that they would continue to locate real banana platforms at a frequency greater than chance. If however, spatial information was a primary cue in selecting feeding sites, then we would expect the capuchins to return to platforms that had contained bananas in Experiment 1A. The results (fig. 3) indicate that on the first day of Experiment 1B the capuchins visited 8 platforms in the morning, only 4 of which contained real bananas (50%). Initial platform visits were to sites that had contained real bananas in the previous experimental condition. Thus in terms of the hierarchical use of information, spatial cues appeared to be more salient than visual and olfactory cues in selecting feeding sites.

The new positions of baited sites were learned rapidly (fig. 3). When the capuchins revisited the Feeding Stations in the afternoon (2nd exposure to test conditions), 4 of 5 (80%) visits were to platforms containing real bananas. Over the remaining 4 days of this experiment, the animals selected reward platforms 90.3% of the time (28/31) (X2= 21.7, df=1, p<.01).

Experiment 1 C offered a second test of the importance of spatial information in locating feeding sites. In this condition, visual cues were eliminated by covering both real bananas and plastic bananas with a large leaf. Differences in olfactory cues were minimized by placing banana skins in with the plastic bananas. The positions of real and sham feeding sites were placed randomly on the morning of day 1. Therefore, on their first exposure to the test conditions, the capuchins lacked spatial, visual, or olfactory information that they could use to increase the likelihood of encountering real banana sites. Given that 'place' remained constant throughout the rest of this experiment, on all subsequent visits the capuchins could rely only on spatial information (landmark cues) as a guide to locate reward sites.

On the morning of the first day of Condition 1C, the capuchins encountered real bananas on 5 of 10 platforms visited (50%; fig. 3). Their ability to relocate these platforms increased to 77% by Day 2, and remained at that level throughout the Experiment (afternoon Day 1 through Day 5 = 39/52 or 75%). These results indicate that when presented with spatial information only, the capuchins were still able to locate real banana feeding sites at a rate significantly greater than chance (X2=18.0, df=1, p<.01). Moreover, there were no differences in performance on morning and afternoon trials (AM= 18/23 [78.2%] and PM= 16/21 [76.1%], Days 2-5; X2= .01, df=1, p>.05). Evidence of a significantly better performance in the afternoon would imply that the capuchins did not necessarily retain spatial information from one day to the next, but rather used spatial information obtained each morning to predict the location of real banana platforms each afternoon.

Experiment 2
This experiment tested the ability of capuchins to use a nearby associative cue (a yellow block) to predict the location of productive feeding sites. This was accomplished by presenting the monkeys with a set of conditions in which 'place' was no longer predictable. That is, although the number of platforms at each Feeding Station that contained plastic bananas and real bananas remained constant, the locations of food and sham food were assigned randomly each morning and each afternoon. In addition, platforms were covered with a large leaf to hide the bait, differences in olfactory cues minimized, and a yellow block was placed only on those platforms that contained real bananas. In this experiment, the presence or absence of a yellow block on each platform provided the only reliable information that could be used to successfully identify feeding sites. In the wild, capuchins may rely on nearby visual cues to increase the likelihood of encountering prey concealed or embedded in microhabitats such as palm fronds, dead wood, and the base of bromeliads.
Initial analysis of the data failed to indicate that the capuchins had learned to associate the yellow block with the presence of a food reward. Over the first 6 days only 21 of 47 platforms (44.6%) visited at Station 1 contained real bananas (table 3). This was not significantly different from random (42.8%; X2=.04, df=1, p>.05). At Station 2, 45.4% of visits (15/33) were to real banana platforms (table 3). Once again this did not deviate from chance expectations (33%; X2=1.55, df=1, p>.05).

During the final 6 days of Experiment 2 (days 7-12), the capuchins visited the feeding platforms 130 times. However, as in the case of the first 6 days, they failed to select banana sites above chance levels (table 3; 45.3% Station 1 and 40.0% Station 2). In Experiment 1C, when 'place' was predictable the capuchins visited an average of 6.4 platforms per day. When an identical amount of food was available, but 'place' no longer predictable (Experiment 2), the number of platforms visited per day increased to 9.3. Unlike in previous experiments, some group members continued to search unexplored platforms even after all sites with yellow blocks had already been visited.

Closer inspection of the data showed that at least some individuals did associate these nearby landmarks with a food reward. Given that 3 of 7 platforms contained real bananas at Station 1 and 2 of 6 platforms contained real bananas at Station 2, foraging behaviour was rescored to include only the first 3 platform visits and first 2 platform visits respectively. This was done because it became apparent that the first animals to visit feeding platforms often went directly to yellow block/banana sites, whereas animals arriving later went from platform to platform in search of food. Analyzing the data in this way showed that during both morning and afternoon sessions, the initial set of platforms visited (61/111 = 54.9%) were associated with the presence of yellow blocks (X2=7.8, df=1, p<.01). Moreover, a breakdown of which platforms were selected on a day by day basis offers insight into the rate at which this information was learned (fig. 4). For example, during the first 3 days of Experiment 2, sites containing yellow blocks were among the initial set of platforms visited 46.6% of the time (expected was 42.8%) at Station 1 and 28.5% of the time (expected was 28.5%) at Station 2 (fig. 4). From day 4 through day 12, however, 57.4% of the first 3 platforms visited at Station 1 and 60% of the first 2 platforms visited at Station 2 were associated with yellow blocks. These values were significantly above chance levels (X2=14.9, df=2, p< .05).

Platform choices were also examined in terms of combined or joint probability in foraging success. That is, we calculated the expected probability that each of the first 3 platforms, 2 of the first 3 platforms, 1 of the first 3 platforms, and none of the first 3 platforms visited at Station 1 in a morning or afternoon trial would have real bananas. This was also done for Station 2, and the results of both analyses are presented in table 4. The evidence indicates that during the final 9 days of Experiment 2, the capuchins did rely on nearby landmark cues to discriminate between food and nonfood sites.

Experiment 3
In the previous experimental condition, the spatial positions of real and plastic bananas were random during morning and afternoon trials. In Experiment 3, 'place' was once again constant, differences in olfactory cues minimized, and the bait was concealed. The purpose of this experiment was to determine whether the capuchins would associate the location of a food source with information concerning the amount of food expected at a feeding site. In this case we instituted a new baiting protocol, in which the amount of food at particular platforms was unequal. Hence, two platforms at Station 1 contained 2 bananas each, one platform contained 1/2 banana, and 4 contained plastic bananas. At Station 2, one platform contained 2 bananas, two platforms each contained 1/2 banana, and 3 platforms contained plastic bananas.
When place was constant in previous experiments, the capuchins learned to discriminate between food and nonfood platforms rapidly. During the first 5 days of this experiment, however, there was no evidence that platforms containing real bananas were visited more frequently than expected by chance. Six of 13 platforms (46.1%) offered a banana reward and 52.5% (51/97) of capuchin visits were to these platforms (X2=0.86, df=1, p>.05). By day 6, however, the capuchins showed a change in behaviour that was consistent with knowledge of the locations of real and sham feeding sites. The data indicate that from day 6-12 there was a marked reduction in the mean number of platforms visited daily (Day 1-5 = 9.1; Days 6-12 = 6.0), and the percentage of visits to real banana platforms increased to 75% (63/84) (X2=15.1, df=1, p<.05). Moreover, if we examine foraging patterns based on the first 3 choices during Days 6-12, then 80.5% (29/36) of the platforms selected at Station 1 and 84.3% of platforms selected (27/32) at Station 2 contained real bananas.

In order to determine whether the capuchins used information about the quantity of food available in selecting feeding sites, we compared evidence of preferential visits to platforms expected to contain greater food rewards. This was done by ranking reward platforms at each Feeding Station according to the order they were visited, and whether they contained 2 real bananas or 1/2 of a real banana. During days 1 through 5, the first platform visited at Station 1 contained 2 real bananas in 10% (1/10) of morning and afternoon trials (table 5). Similarly, in only 2 of the first 10 trials (20%) at Station 2 did the capuchins select the platform containing 2 real bananas. In contrast, during the final 7 days the first feeding platform visited contained 2 bananas on 8 of 12 trials at Station 1 (66.6%) and 7 of 12 trials (58.3%) at Station 2 (table 5). Given that the likelihood of selecting a platform containing 2 bananas at random from Station 1 was 28.5% (2/7) and from Station 2 was 16.6% (1/6), it appeared that at least some capuchins integrated spatial and quantity information in their foraging decisions.

Discussion
The primate infraorder Platyrrhini represents an extremely successful and diversified radiation of small to medium sized arboreal monkeys, and includes several species that exhibit social patterns and foraging skills reminiscent of the great apes (e.g. the fission-fusion social system of atelines; extractive foraging, tool use, and high degree of vocal coordination among dispersed foragers in Cebus [15,29,30,39,41,42,45,52,53]. Currently, there exists only a limited amount of information to assess the degree to which high levels of communication and cognition in some platyrrhines and great apes are a prerequisite for these complex behavioural patterns [41]. However, given significant differences in diet, feeding ecology, encephalization, rates of infant development, and prehension among closely related platyrrhine taxa (e.g. howler monkeys and spider monkeys; or capuchins and squirrel monkeys), the study of New World primates offers an opportunity to explore adaptive and functional relationships between sensory skills, cognitive abilities, and foraging behaviour [6,10,11,30]. As expressed by Meador et al. [54: pg 21], "The adoption of an ecological perspective appears particularly useful in the comparative study of the learning capacities of New World monkeys."
The aim of the present set of experiments was to test the ability of wild capuchins (Cebus capucinus) to use spatial information and perceptual cues to solve foraging problems. Capuchins were presented with a controlled set of feeding and foraging conditions that were analogous to conditions they may naturally encounter in the wild. For example, many foods eaten by these primates are distributed in small, ephemeral, or scattered food patches [26,37,55]. These feeding sites are often associated with hidden or embedded prey, and may exhibit high variance in daily food rewards. Skills required to exploit these resources include a high degree of manual dexterity, as well as the ability to distinguish between similar feeding events that occur at different points in time and in different localities within the home range (e.g., integrate spatial and temporal information with information on food type, renewal rates, and quantity of food rewards). As the number of feeding sites visited increases, greater integration and cognitive skills may be needed to track food intake in several patches simultaneously in order to predict future foraging success [5,42]. In each of our field experiments, the ability of capuchins to locate real banana feeding sites was directly related to their ability to store, categorize, and associate disparate types of environmental information.

The results showed that individual capuchins learned spatial information rapidly and associated platform location with the presence or absence of a food reward (i.e. win-stay and lose-shift foraging pattern). For example, after a single exposure to a new set of resource distributions (Exp 1C) the capuchins returned to real banana feeding sites on 77% of platform visits. This was accomplished in the absence of other cues (food was concealed and differences in olfactory cues minimized) and required that the monkeys not only differentiate between the spatial positions of 13 individual feeding platforms, but also distinguish between past feeding success and present feeding success at each platform. That is, certain platforms that had offered a reward over the previous 10 days no longer offered a food reward. Some platforms that were previously unproductive, however, now offered a food reward. In exploiting these platforms, the capuchins exhibited a pattern of flexible learning and took advantage of changes in resource information as they became available. Evidence of flexible learning and extreme sensitivity to changes in food availability and distribution have also been reported in nonexperimental field studies of Cebus capucinus [26,33,34].

The results of Experiment 1 clearly indicated the predominance of spatial information over other cues in selecting feeding sites (fig. 3). On the first day of Experiment 1B (rotation), for example, when spatial information conflicted with visual and olfactory information, the capuchins continued to return to platforms that had previously contained food rewards. This occurred despite the fact that these platforms presently contained plastic bananas, and offered visual and olfactory cues that were different than the visual and olfactory cues offered by platforms containing real bananas. Using spatial information to return to feeding sites that were productive on the previous day is likely to be an important tactical component of foraging in many animal species [13].

In the second experiment 'place' was no longer predictable. Platforms containing real and sham bananas were changed randomly during each morning and each afternoon trial. Under these conditions, the same win-stay and lose-shift foraging strategy that was effective in Experiment 1 was no longer the most efficient way to exploit these resources. This is especially the case for an individual foraging within the context of a social group (i.e. within group feeding competition see [34,35]). In our experiments, the costs in time and energy of visiting an incorrect or sham feeding platform were probably negligible. However, due to the limited number of feeding sites at each food patch (only 3 of 7 platforms at Station 1 and 2 of 6 platforms at Station 2 had real bananas), each visit to a sham platform might significantly decrease an individuals' chances of encountering a reward platform that had not already been exploited by another group member. In Cebus capucinus, unlike Cebus apella, a single dominant individual rarely controls access to a feeding site [33,55-57]. White-faced capuchins have been observed to feed alone at small food patches [55]. However, under conditions in which only a limited set of productive feeding sites are available, the presence of several group members simultaneously exploiting the same patch is likely to increase the cost to each individual of selecting an unproductive feeding site [31,33].

In Experiment 2, a single yellow block placed on reward platforms was the only information available to identify real versus sham feeding sites. The evidence indicates that within a period of 3 days (6 trials), at least some capuchins used these local landmark cues to find the real bananas. Problem-solving under these conditions involved learning the association between the location of a hidden food reward and the presence/absence of a nearby object in the environment. Given the rate at which capuchins learned this association, we argue that these primates applied a set of pre-existing foraging 'rules' to our experimental setting. Such 'rules' or 'presolutions' "represent behavioural responses to food acquisition problems that are frequently encountered in their environment..." [6]. It is likely that spatial association learning [4], or the use of nearby cues to identify concealed feeding sites, plays an important role in the ability of capuchins to exploit insect and vertebrate prey successfully.

It is instructive to compare these results on white-faced capuchins to a related experimental field study of foraging decisions in moustached tamarins (Saguinus mystax) [6]. Using a similar (but not identical) research design, Garber and Dolins [6] found that tamarins also learned the spatial positions of reward and sham platforms rapidly, and that olfactory cues did not play a significant role in locating feeding sites. As in the case of the capuchins, when given conflicting visual, olfactory, and spatial information, callitrichines responded to changes in the positions of real and plastic bananas by returning to platforms that had previously contained food rewards [6]. Moreover, there was strong evidence that the tamarins adopted a range of complex foraging rules (i.e. win-stay and lose shift; win-shift and lose-return) in selecting feeding sites. Similar patterns of rule-based learning and complex use of spatial information have been reported in captive groups of saddle-back tamarins [58,59] and cotton-top tamarins [60].

Despite many similarities in response to the conditions imposed by the field experiments, capuchins appeared to differ from tamarins in their ability to associate nearby landmark cues with the presence of hidden food rewards at feeding sites. Capuchins learned this association rapidly (3 days). Although the tamarins were not tested under identical conditions (red flags were used instead of yellow blocks and 16 platforms were used in the experiment rather than 13), during the course of a 6 day experiment with local landmarks cues present, the tamarins did not select reward platforms at a rate greater than expected by chance [6]. In the wild, moustached tamarins rarely exploit concealed or embedded prey [61]. It is possible that these primates rely primarily on direct visual sighting of foods and distant landmarks rather than local landmarks as a guide in locating feeding sites. Support for this conclusion will require additional field experiments. Capuchins, however, appear to use distant spatial information [22-30], as well as local microenvironmental information (possibly evidence of insect damage, condition of arboreal nests, subtle differences in colour or texture of holes or tunnels on tree branches and trunks) in foraging decisions. The ability to integrate both types of information has been reported in common chimpanzees when extracting embedded prey [48].

In the final experiment, there was evidence that the capuchins learned and used information about the quantity of food in selecting feeding sites (1/2 banana vs. 2 bananas). Acquisition of this information occurred over a 5 day period. This was longer than the time it took to associate spatial information with the presence/absence of a food reward (one-trial learning) or to associate a yellow block with the presence of food at randomly changing feeding sites (3 days). Delay in learning might reflect several factors. It is possible that (a) quantity information is less salient to the capuchins than other information and is more difficult to learn, (b) the difference between 1/2 banana and 2 bananas was not large enough to stimulate goal-directed behaviour, (c) switching from a random baiting protocol in Experiment 2 to a predictable baiting protocol in Experiment 3 required an extended period of relearning patterns of spatial predictability, or (d) a change in social foraging tactics resulted in subordinate individuals selecting sites that offered smaller food rewards in order to avoid competing with dominant individuals (e.g. risk sensitive forgaing). At present, our data base is insufficient to test these alternatives. We do, however, plan to explore these important issues through additional field experiments.

In conclusion, the results of this study indicate that wild white-faced capuchins used associative cues (local landmarks), spatial cues, and quantity information to discriminate between real banana and plastic banana feeding sites. In the presence of conflicting information, however, spatial cues appeared to predominate over other forms of information. We feel, that in responding to the conditions of these feeding experiments, the capuchins applied a set of problem-solving behaviours that they otherwise use in exploiting resources in their natural environment. These problem-solving behaviours may include certain spatial, temporal, and quantity based representational 'rules' of foraging such as return to a feeding site that had previously offered a food reward (win-stay), avoid feeding sites that were unproductive (lose-shift), return to feeding sites that previously contained near-to-ripened foods (lose-return), and avoid feeding sites that previously contained food but were unlikely to offer additional food rewards (win-shift). In this regard, controlled field experiments offer a powerful tool for studying individual and species differences in learning, and the use of spatial-temporal information and sensory cues in foraging.

Acknowledgements
This study was conducted with the permission of La Suerte Biological Research Station, Limon, Pococi, Costa Rica. We thank Rene Molina, Lilian Molina, Alvaro Molina, and Federico Molina for their support and encouragement. Assistance in the field was provided by Joanna Lambert, Jennifer Rehg, Gerardo Rivera, and Ricardo V‡squez. Comments on earlier drafts of this manuscript were provided by Steven Leigh, Joanna Lambert, Hilary Box, and Hannah Buchanan-Smith. We thank wildlife photographer Michael P. Turco for permission to use the photo in Figure 1. P.A. Garber wishes to thank Sara and Jenni for sharing their insights on primate foraging decisions and unpredictable dietary patterns.
FIGURE CAPTIONS
Fig. 1 - A white-faced capuchin monkey (Cebus capucinus) from Costa Rica. Photo by Michael P. Turco.
Fig. 2 - Schematic representation of the spatial relationship between Feeding Stations, feeding platforms, and the blind used in this experimental field study.

Fig. 3 - Cumulative frequency of visits to platforms containing real bananas (correct platforms). Data are combined for Feeding Stations 1 and 2 and are based on the number of visits to reward platforms divided by total number of visits to all platforms (banana plus sham).

Fig. 4 - Percent correct visits to platforms containing a nearby landmark cue (yellow block) and associated food reward. Data for Feeding Stations 1 and 2 are presented separately. Values represent percent correct visits (total number of visits to banana platforms divided by the total number of visits to banana plus sham platforms) for each 3 day period.

References
1. Menzel EW: Chimpanzee spatial memory. Science 1973, 182: 943-945.
2. Kamil AC, Yoerg SI: Learning and foraging behaviour; in Bateson PPG, Klopfer PH (eds): Perspectives in Ethology, Volume 5: Ontogeny. New York, Plenum Press, 1982, pp 325-364 .

3. Krebs JR, Inman AJ: Learning and foraging: Individuals, groups, and populations. Am Nat 1992, 240: S63-S84.

4. Brown GS, Gass CL:Spatial association learning by hummingbirds. Anim Behav 1993, 46:487-497.

5. Real LA: Information processing and the evolutionary ecology of cognitive architecture; in Real LA (ed.): Behavioural Mechanisms in Evolutionary Ecology. Chicago, University of Chicago Press, 1994, pp 99-132.

6. Garber PA, Dolins FL: Testing learning paradigms in the field: evidence for use of spatial and perceptual information and rule-based foraging in wild moustached tamarins; in Norconk M, Rosenberger AL, Garber PA (eds): Adaptive Radiation of Neotropical Primates. New York, Plenum Press, in press.

7. Kamil AC: A synthetic approach to the study of animal intelligence; in Real LA (ed.): Behavioural Mechanisms in Evolutionary Ecology. Chicago, University of Chicago Press, 1994, pp 11-45 .

8. Galef BG Jr: Utilization by Norway rats (Rattus norvegicus) of multiple messages concerning distant foods. J Comp Psychol 1983, 97: 364-371.

9. Fragaszy DM: Comparative studies of squirrel monkeys (Saimiri) and titi monkeys (Callicebus) in travel tasks. Zeit Tierpsych 1980, 54:1-36.

10. Milton K: Distribution patterns of tropical plant foods as an evolutionary stimulus to primate mental development. Am Anthropol 1981, 83: 535-543.

11. Milton K: Foraging behaviour and the evolution of primate intelligence; in Byrne RW, Whiten A (eds): Machiavellian Intelligence: Social Expertise and the Evolution of Intellect in Monkeys, Apes, and Humans. Oxford, Clarandon Press, 1988, pp 285-305 .

12. Boesch C, Boesch H: Mental map in wild chimpanzees: an analysis of hammer transports for nut cracking. Primates 1984, 25: 160-170.

13. Garber PA: Role of spatial memory in primate foraging patterns: Saguinus mystax and Saguinus fuscicollis. Am J Primatol 1989, 19:203-216.

14. Westergaard GC: The subsistence technology of capuchins. Int J Primatol 1994, 15: 899-906.

15. Boinski S, Campbell A: Vocal coordination of troop movement in white-faced capuchins: a second field test. Behaviour, in press.

16. Terborgh J: Community aspects of frugivory in tropical forests; in Estrada A, Fleming TH (eds):Frugivores and Seed Dispersal. Dordrecht, Dr W Junk Publishers, 1986, pp 371-384.

17. Post DG: Feeding behaviour of yellow baboons (Papio cynocephalus) in the Amboseli National Park, Kenya. Int J Primatol 1982; 3:403-430.

18. Menzel EW: Cognitive mapping in chimpanzees; in Hulse SH, Fowler H, Honig WK (eds): Cognitive Processes in Animal Behaviour. New York, Academic Press, 1978, pp 375-422.

19. Altmann SA: Baboons, space, time and energy. Am Zool 1974, 14: 221-248.

20. Sigg H, Stolba A: Home range and daily march in a hamadryas baboon troop. Folia Primatol 1981, 36:40-75.

21. Janson CH, Terborgh J, Emmons LE: Nonflying mammals as pollinating agents in the Amazonian forest. Biotropica 1981, 13: 1-6.

22. Robinson JG: Seasonal variations in use of time and space by wedge-capped capuchin monkeys, Cebus olivaceus: implications for foraging theory. Smith Contrib Zool 1986, 431:1-60.

23. Oates JF: Food distribution and foraging behaviour; in Smuts BB, Cheney DL, Seyfarth RM, Wrangham, RW, Struhsaker TT (eds): Primate Societies. Chicago, University of Chicago Press, 1987, pp 197-209.

24. Garber PA: Foraging decisions during nectar feeding in tamarins monkeys (Saguinus mystax and Saguinus fuscicollis, Callitrichidae, Primates) in Amazonian Peru. Biotropica 1988, 20:100-106.

25. Garber PA: Seasonal patterns of diet and ranging in two species of tamarin monkeys: stability versus variability. Int J Primatol 1993, 14: 145-166.

26. Chapman CA: Patterns of foraging and range use by three species of Neotropical primates. Primates 1988; 29:177-194.

27. Menzel CR: Cognitive aspects of foraging in Japanese monkeys. Anim Behav 1991, 41: 397-402.

28. Leighton M: Modeling dietary selectivity by Bornean orangutans: evidence for integration of multiple criteria in fruit selection. Int J Primatol 1993, 14: 257-314.

29. Robinson JG, Janson CH: Capuchins, squirrel monkeys, and atelines: socioecological convergence with Old World primates; in Smuts BB, Cheney DL, Seyfarth RM, Wrangham, RW, Struhsaker TT (eds): Primate Societies. Chicago, University of Chicago Press, 1987, pp 69-82 .

30. Janson CH, Boinski S: Morphological versus behavioural adaptations for foraging in generalist primates: the case of the cebines. Am J Phys Anthropol 1992, 88:483-498.

31. Chapman CA, Fedigan LM: Dietary differences between neighboring Cebus capucinus groups: local traditions, food availability, or responses to food profitability. Folia Primatol 1990; 54:177-186.

32. Fedigan LM: Vertebrate predation in Cebus capucinus: meat-eating in a Neotropical monkey. Folia Primatol 1990; 54:196-205.

33. Fedigan LM: Sex differences and intersexual relations in adult white-faced capuchins (Cebus capucinus). Int J Primatol 1993; 14:853-877.

34. Rose LM: Sex differences in diet and foraging behaviour in white-faced capuchins (Cebus capucinus). Int J Primatol 1994; 15:95-114.

35. Rose LM: Benefits and costs of resident males to females in white-faced capuchins, Cebus capuninus. Am J Primatol 1994; 32:235-248

36. Parker ST: Opportunism and the rise of intelligence. J Hum Evol 1978, 7: 596-608.

37. Freese C, Oppenheimer JR: The capuchin monkeys, genus Cebus; in Coimbra-Filho AF, Mittermeier RM (eds): Ecology and Behaviour of Neotropical Primates Vol 1. Rio de Janeiro, Academia Brasileira de Ciences, 1981, pp 331-390.

38. Terborgh J: Five New World Primates: A study in comparative ecology. Princeton, NJ. Princeton University Press, 1983.

39. Westergaard GC, Fragaszy, DM: The manufacture and use of tools by capuchin monkeys (Cebus apella). J Comp Psych 1987, 101:159-168.

40. Costello MB, Fragaszy DM: Prehension in Cebus and Saimiri: I. Grip type and hand preference. Am J Primatol 1988, 15:235-245.

41. Visalberghi E, Fragaszy DM: Do monkeys ape? in Parker ST, Gibson KR (eds): Language and Intelligence in Monkeys and Apes. Cambridge, Cambridge University Press, 1990, pp 247-273.

42. Gibson KR: New perspectives on instincts and intelligence: brain size and the emergence of hierarchial mental constructional skills; in Parker ST, Gibson KR (eds): Language and Intelligence in Monkeys and Apes. Cambridge, Cambridge University Press, 1990a, pp 97-128.

43. Antinucci F: The comparative study of cognitive ontogeny in four primate species; in Parker ST, Gibson KR (eds): Language and Intelligence in Monkeys and Apes. Cambridge, Cambridge University Pres, 1990, pp 157-171.

44. Fragaszy DM, Adam-Curtis LE: Environmental challenges in groups of capuchins; in Box HO (ed): Primate Responses to Environmental Change. London, Chapman and Hall, 1991, pp 239-264.

45. Izawa K, Mizuno A: Palm fruit cracking behaviour of wild black-capped capuchins (Cebus apella). Primates 1977, 18: 773-792.

46. Struhsaker TT, Leland L: Palm nut smashing by Cebus a. apella in Colombia. Biotropica 1977, 9: 124-126.

47. Antinucci R, Visalberghi E: Tool use in Cebus apella: a case study. Int J Primatol 1986, 7: 349-361.

48. Gibson KR: Tool use, imitation, and deception in a captive cebus monkey; in Parker

ST, Gibson KR (eds): Language and Intelligence in Monkeys and Apes. Cambridge, Cambridge University Press, 1990b, pp 205-218

49. Hershkovitz P: Living New World Monkeys (Platyrrhini), Volume 1, Chicago, University of Chicago Press, 1977.

50. Stephan H: Evolution of primate brains: a comparative anatomical investigation; in Tuttle RH (ed.): The Functional and Evolutionary Biology of Primates. Chicago, Aldine¥Atherton, 1972, pp 155-174 .

51. Sanford R, Paaby P, Luvall J, Phillips E: Climate, geomorphology, and aquatic systems; in McDade L, Bawa K, Hespenheide H, and Hartshorn G (eds): La Selva: Ecology and Natural History of a Neotropical Rain Forest. Chicago, University of Chicago Press, 1994, pp 19-33.

52. Strier KB: Atelinae adaptation: behavioural strategies and ecological constraints. Am J Phys Anthropol 1992, 88: 515-524.

53. Chapman CA, Wrangham RW, Chapman LJ: Ecological constraints on group size: an analysis of spider monkey and chimpanzee subgroups. Beh Ecol Sociobio 1995, 36: 59-70.

54. Meador DM, Rumbaugh DM, Pate JL, Bard KA: Learning, problem solving, cognition, and intelligence; in Mitchell G, Erwin J (eds): Comparative Primate Biology - Volume 2, Part B: Behaviour, Cognition, and Motivation. New York, Alan R. Liss, 1987, pp 17-84 .

55. Phillips KA: Resource patch size and flexible foraging in white-face capuchins (Cebus capucinus). Int J Primatol 1995; 16:509-521.

56. Janson CH: Aggressive competition and individual food consumption in the borwon capuchin monkey (Cebus apella). Behav Ecol Sociobiol 1985; 18:125-138.

57. Janson CH: Ecological consequences of individual spatial choice in foraging groups of brown capuchin monkeys Cebus apella. Anim Behav 1990b, 40: 922-934.

58. Menzel EW, Juno C: Marmosets (Saguinus fuscicollis): are learning sets learned? Science 1982, 217:750-752.

59. Menzel EW, Juno C: Social foraging in marmoset monkeys and the question of intelligence. Phil Trans Roy Soc Lond 1985, B308:145-157.

60. Dolins FL: Spatial Relational Learning and Foraging In Cotton-top Tamarins. Ph.D. Thesis, University of Stirling, Scotland, 1993.

61. Garber PA: Feeding ecology and behaviour of the genus Saguinus; in Rylands AB (ed): Marmosets and Tamarins: Systematics, Ecology and Behaviour. Oxford, Oxford University Press, 1993, pp 273-295.


5 Tables and 4 Figures [ 1 - Department of Anthropology, University of Illinois, Urbana, IL 61801.
2 - DPAS, State University of New York, Stony Brook, NY 11794.

Running Head: Memory and Learning in Wild Capuchins

Send all correspondence to:

Dr. Paul A. Garber
Department of Anthropology
109 Davenport Hall, 607 S. Mathews Ave.
University of Illinois, Urbana, IL 61801.
Tele: (217) 333-3616
FAX: (217) 244-3490
email: pgarber@ux1.cso.uiuc.edu

Revised March 15, 1996

Submitted for review and publication in Folia Primatologica Abstract