Recognition and categorization of biologicallysignificant objects by rhesus monkeys(Macaca mulatta): the domain of foodLaurie R. Santosa,*, Marc D. Hausera,b, Elizabeth S. SpelkeaaDepartment of Psychology, Harvard University, 33 Kirkland Street, Cambridge, MA 02138, USAbProgram in Neuroscience, Harvard University, Cambridge, MA 02138, USAReceived 4 May 2000; received in revised form 6 June 2000; accepted 15 June 2001AbstractTo survive, organisms must be able to identify edible objects. However, we know relativelylittle about how humans and other species distinguish food items from non-food items. Wetested the abilities of semi-free-ranging rhesus monkeys (Macaca mulatta) to learn rapidlythat a novel object was edible, and to generalize their learning to other objects, in a sponta-neous choice task. Adult monkeys watched as a human experimenter first pretended to eat oneof two novel objects and then placed replicas of the objects at widely separated locations.Monkeys selectively approached the object that the experimenter had previously eaten, exhi-biting a rapidly induced preference for the apparently edible object. In further experiments inwhich the same objects were used as tools or were manipulated at the face but not eaten, wefail to observe an approach bias, providing evidence that the monkeys’ pattern of approach inthe earlier experiments was specific to objects that were eaten. Subsequent experiments testedhow monkeys generalized their preference for an edible object by first allowing them to watcha human experimenter eat one of two objects and then presenting them with new objectscomposed of the same substance but differing from the original, edible object in shape orcolor. Monkeys ignored changes in the shape of the object and generalized from one edibleobject to another on the basis of color in conjunction with other substance properties. Finally,we extended this work to infant rhesus monkeys and found that, like adults, they too usedcolor to generalize to novel food objects. In contrast to adults, however, infants extended thispattern of generalization to objects that were acted on in other ways. These results suggest thatinfant monkeys form broader object categories than adults, and that food categories becomeL.R. Santos et al. / Cognition 82 (2001) 127–155 127Cognition 82 (2001) 127–155www.elsevier.com/locate/cognit0010-0277/01/$ - see front matter q 2001 Elsevier Science B.V. All rights reserved.PII: S0010-0277(01)00149-4COGNITION* Corresponding author. Fax: 11-617-496-7077.E-mail address: [email protected] (L.R. Santos).sharpened as a function of maturational or experiential factors. q 2001 Elsevier Science B.V.All rights reserved.Keywords: Biologically significant objects; Rhesus monkeys (Macaca mulatta); Domain of food1. IntroductionTo survive, human and non-human animals must classify the rich assortment ofphysical and social objects that surround them. Discovering appropriate categories isa challenge, however, because there are indefinitely many ways to sort objects in anyorganism’s environment. For example, when encountering a new fruit that turns outto be rotten, an animal might infer that the fruit’s unpalatability was signaled by anynumber of different properties or combinations of properties of the fruit, includingits smell, color, texture, shape, size, time of discovery, or proximity to other objects.Some of these properties would be beneficial for distinguishing ripe from rottenfruit, but most properties would not (see Hauser, Kralik, & Botto-Mahan, 1999;Keil, Smith, Simons, & Levin, 1998). Differentiating properties that are relevant forcategorization from the many detectable but irrelevant properties is made moredifficult by the fact that the properties relevant for categorizing objects in onedomain often are not relevant for categorizing objects in other domains. For exam-ple, although a chimpanzee could successfully use the feature of smell to judge theedibility of fruits, it could not apply this same feature to determine rank in males orrigidity in rocks. Given these difficulties, how do organisms discover effective waysto sort objects?Evolutionary biology and psychology suggest that animals may solve thisproblem by breaking it into parts. Instead of one general-purpose system for learningto categorize objects, animals may possess a set of specialized subsystems forcategorizing particular, ecologically significant kinds of objects (Gallistel, 1990;Hauser, 2000). Behavioral ecologists provide numerous examples of domain-speci-fic recognition systems in a wide array of animals, including recognition of compet-ing males by frogs (Ryan & Rand, 1999), recognition of different predator classes byvervet monkeys (Cheney & Seyfarth, 1990), and recognition of edible foods bychickens (Marler, Dufty, & Pickert, 1986). In some cases, domain-specific recogni-tion systems may be largely innately specified; in other cases they depend oninnately constrained learning systems (e.g. imprinting in chicks, Horn, 1998; songrecognition in sparrows, Marler, 1993; recognition of stellar configurations bymigratory birds, Emlen, 1996). If constrained, domain-specific recognition systemsallow human and non-human animals to categorize the significant objects in theirenvironment, then primates may form categories of some objects but not others, andthey may use different perceptual features to categorize objects of different kinds.A number of researchers in the field of cognitive development have embraced asimilar hypothesis about children’s categorization of objects, proposing that humansare born with biases that guide their perception and learning about different objectL.R. Santos et al. / Cognition 82 (2001) 127–155128kinds (Carey & Spelke, 1994; Gelman, 1990; Hirschfeld & Gelman, 1994; Keil,1989; Keil et al., 1998). When children encounter a new animal, for example, theymay attend to detailed features of its shape, composition, and manner of motion(Gelman, 1990; Mandler, 1992); when they encounter a tool, they may attend todifferent shape features and to the actions that others perform upon it (Brown, 1990;Landau, Smith, & Jones, 1988, 1998).Despite the interest in domain-specific categorization processes, the domain offood has received little attention in human or animal research. Although manyinvestigators of object recognition in humans and non-human primates proposethat distinct representational systems underlie recognition of animate objects