THE COGNITIVE ANIMAL
Empirical and Theoretical Perspectives on Animal Cognition
edited by
Marc Bekoff, Colin Allen, and Gordon M. Burghardt
A Bradford Book The MIT Press
Jacques Vauclair
Animals often behave adaptively in response to a between exemplars of a class. Thus, level 4 is novel stimulus because the stimulus resembles more complex than open-ended classification, others for which the appropriate response is al- the latter being related to the use of perceptual ready known. Such an adaptation expresses an dimensions of stimuli. [See Schrier et al. (1984) ability to categorize. In effect, in the absence of for an example in macaques, D'Amato and Van categorization, each object or event would be Sant (1988) for an example in cebus monkeys, perceived as unique, and generalizations would and Vauclair and Fagot (1996) for an example in be impossible. Therefore it is not surprising to Guinea baboons.] Level 5 of Herrnstein's cate-find categorizing abilities in various animal spe- gorization is attained when a subject is able to cies, although most of the empirical evidence use abstract relations not only between objects concerns birds (mainly pigeons) and primates. but also between concepts, such as in conceptual Since categorization is a fundamental aspect of matching or in conceptual identity (for example, information processing, its study is crucial for the mastery of a "sameness" relation). The evi-increasing our understanding of animals' cogni- dence for capacities to perform the first three
tive abilities. levels of categorization is abundant for several
This essay describes some of the studies that animal species (see Zayan and Vauclair 1998 and have been carried out in the past 5 years with two Thompson and Oden 2000 for reviews). It is, species of baboons, both in laboratory-controlled however, much less clear concerning levels 4 conditions (Guinea baboons, Papio papio) and in and 5.
outdoor settings (olive baboons, Papio anubis). These studies aimed to explore different levels
of categorizing behaviors (and their underlying Laboratory Studies with Guinea Baboons processes) in monkeys confronted with various
tasks. Furthermore, we were interested in com- Several experiments were conducted with ba-paring monkeys and humans tested with similar boons in order to assess the abilities of these stimuli and procedures. monkeys to discriminate objects on the basis of A useful general framework for the inves- their membership in a category and to study the tigation of these behaviors was provided by nature of the representations of categories the Herrnstein (1990), who described categorization baboons formed. In all the experiments reported abilities in animals in five levels of increasing in this section, we used a video task requiring the abstractness, including (1) discrimination, (2) baboons to manipulate a joystick that controlled categorization by rote, (3) open-ended catego- the movements of a cursor on a screen (Vauclair rization (namely, category formation resting on and Fagot 1994). Briefly, with this technique, the a perceptual similarity between individuals that subject was required to manipulate the joystick belong to a given class), (4) concepts, and (5) so as to "touch" with the cursor a response abstract relations. Herrnstein (1990) uses two stimulus that matched the sample stimulus on an criteria to define conceptual categorization (level arbitrary (experimenter-defined) basis.
Jacques Vauclair 240
this purpose, the baboons were first trained in Demonstration of prototype effects in animals a symbolic matching-to-sample task with 21 dif- is controversial. Some authors (von Fersen and ferent fonts of the characters "B" and "3" as Lea 1990) have described it in pigeons, while sample forms, and color squares as comparison others (e.g., Huber and Lenz 1993) did not find forms. After training, novel fonts were displayed. it. Only one study (Jitsumori 1994) has assessed The monkeys showed positive transfer of cate- prototype effects in nonhuman primates. In this gorizing performance to the novel stimuli of the research, artificial stimuli defined by three two-characters used in the original training. Such valued positive or negative dimensions (color, results demonstrate that the original learning shape, and background color) were presented to was not achieved by rote, because in that case rhesus monkeys. The stimuli used in training the animals would have demonstrated no trans- included only two of the three positive or nega-fer to the novel typefaces. Thus the baboons' tive dimension values. In transfer, both the pro-performance indicates that these monkeys were totypes and novel exemplars of the two categories able to exhibit level 3 behavior (open-ended cat- were shown to the animals. Three of the five egories in Herrnstein's sense). monkeys demonstrated a high level of transfer A proper assessment of categorization requires with the prototypes containing all three positive not only that subjects conceive that different or negative features. However, for those mon-objects have common class attributes but also keys, there was no statistical difference between that the subjects can discriminate among indi- the performance achieved with prototypes and vidual members within a category (Thompson that obtained with the nonprototypical stimuli.
1995). Thus, in order to dismiss simple stimulus Given the contrasting evidence for the use of generalization, it must be demonstrated that prototypical representations in animals, we fur-stimuli to be classified in the same category are ther investigated prototype effects by testing two discriminably different from one another. This different species of primates. Identical polymor-control was used in our study (Vauclair and phous artificial stimuli were presented to both Fagot 1996), but it must be pointed that it humans and baboons in a symbolic matching-to-is rarely used in investigations of categorizing sample task. In line with Jitsumori (1994), the abilities of animals. We also demonstrated in rationale of our study with baboons (Dépy et al. further studies similar abilities of our monkeys 1997) was first, to train subjects to classify two to categorically process spatial relations such as out of three feature stimuli (color, shape, posi-"above" and "below" categories (Dépy et al. tion) and second, to assess transfer of perfor-1999) as well as "long" and "short" distances mance with the prototypes of each category.
(Dépy et al. 1998). Analyses of data searched for possible prototype
A hallmark of categorical processing in hu- effects in both species and focused on species mans is the ability to extract a prototype of a differences and similarities in the procedures given category. A possible way to investigate the creating the categories.
represented the central tendency of the class, but and tested on the natural category of food versus because they corresponded to the exemplars nonfood with real objects using an adapted ver-sharing the least common features with the al- sion of a Wisconsin General Test Apparatus. ternative category. In fact our monkeys used a The monkeys were first trained to categorize two mixed procedure that consisted in memorizing objects, one food and one nonfood; then 80 other salient cues between stimuli (e.g., the back- objects (40 food and 40 nonfood) were presented ground color) or specific associations between and the categorizing response to each object was exemplars and response associations. recorded. The baboons showed a high and rapid transfer of their categorizing abilities to the novel items. A similar performance for vervet monkeys Studies with Olive Baboons in a Natural Setting
was described by Zuberbühler et al. (1999). These abilities also persisted in subsequent experiments Tasks requiring the categorization of items on
in which we used cutout photos and various the basis of their functional similarity, such as
modes of picture presentations. This set of data the distinction between food and nonfood items,
further demonstrates the abilities of the baboons are good candidates for studying level 4
behav-to relate real objects behav-to their picbehav-torial repre-ior. This kind of classification indicates the
pres-sentations (Bovet and Vauclair 2000). ence of conceptual categorization given that the
The same procedure of successive simple dis-items to be classified do not necessarily resemble
criminations in a two-alternative forced-choice each other. The animal literature provides only
procedure was used in follow-up studies (Bovet few cases of conceptually based, functional cate- and Vauclair 2001). In a first experiment, the gorizations. In one study by Watanabe (1993),
monkeys had to judge two physical objects as pigeons were trained with real objects (4 edible
"same" or "different" (perceptual identity). For and 4 inedible stimuli) and then tested with
example, they were required to judge two apples printed color photographs (6 novel stimuli, each
as being the same, or an apple and a padlock as belonging either to the food or nonfood cate- being different. In a crucial test (second experi-gory). Generalization to the novel stimuli regard- ment) of conceptual identity (corresponding to less of the type of stimulus presented (picture or Herrnstein's level 5), the baboons had to com-real object) was taken as evidence that pigeons bine their previously acquired skills in order to displayed object-picture equivalence based on classify as "same" two (different) objects that functional classification. The evidence for similar belonged to the same functional category (food abilities in nonhuman primates comes from the or nonfood) and apply that learning to new work of Savage-Rumbaugh et al. (1980). Al- exemplars. For example, they had to classify as though the aim of these authors was to demon- "same" an apple and a banana, or a padlock and strate the mastery of reference in linguistically a cup, and as "different" an apple and a padlock. trained chimpanzees, their study offers clear-cut This ability corresponds to level 5 of Hermstein's evidence of categorizing abilities in apes. In this classification scheme. The monkeys attained a experiment, the chimpanzees were first trained to high level of performance at the end of the ex-classify real items in two categories (food and periment with totally novel objects (i.e., objects tools). Then the subjects easily transferred their novel in the task but left in the monkeys' enclo-categorization to novel objects and later to pic- sure before the experiment). Such results dem-tures and arbitrary symbols (lexigrams) of the onstrate the mastery of the "same-different" respective categories.
relation and the ability to conceptually judge as In one of our studies (Bovet and Vauclair
same or different objects in the previously learned 1998), olive baboons living in small social groups
Jacques Vauclair 242
Discussion and Prospects for Future Work Third, category formation and some of its un-derlying cognitive processes have been discussed A number of issues can be raised about catego- so far mainly with respect to physical objects or rization processes from the studies reported in features. There is, however, another context in this essay. They concern in turn (1) the presence which animals elaborate natural categories, that of categorization procedures in animals outside of their relations with conspecifics, a domain of visual perception, (2) the adaptive value of considered to allow the most complex expression categorization treatments, (3) the interest and of cognitive processes (e.g., Humphrey 1976). importance of studying categorization in the Animals may represent the classes of their con-social domain, and (4) the limits of categoriza- specifics as well as the classes of relationships tion procedures in animals, and possible differ- established between conspecifics or between the entes with human categorization. latter and themselves. As claimed elsewhere First, it is important to stress that the exam- (Vauclair 1996; Zayan and Vauclair 1998), two ples provided here dealt with the processing of types of social categories could be fruitfully visual objects. However, acoustical stimuli are studied: intraspecies recognition, including rec-also categorically organized, not only in non- ognition of individual conspecifics, and repre-human primates (e.g., May et al. 1989; Ramus sentation of dominance hierarchies and their et al. 2000), but also in rodents (e.g., Ehret 1992) transitivity in linear orders. Moreover, it is par-and in birds (Weary 1989). Note also that the ticularly important to study categorizing pro-involvement of categorization procedures for cesses in monkeys and apes because of their organizing information in the environment goes phylogenetic proximity to humans and because well beyond perception and concerns the coding of their complex cognitive abilities.
of objects and space. [An example of the cate- Rare experimental studies are available on gorization of geometric versus nongeometric social categorization in nonhuman primates; information in a reorientation task by rhesus they concern the mother-infant bond and macaques can be found in Gouteux et al. (2001).] dominance-subordination relation in vervets Second, our experiments with two baboon (Cheney and Seyfarth 1980) and macaques species show that they are able to categorize (Dasser 1988). However, because of the contro-stimuli by using varied kinds of procedures, the versies surrounding the results and interpreta-extent of which may depend on the species and Lions of these experiments (e.g., Thompson 1995) individual experiences with the objects and pro- and the limited number of subjects tested, these cedures. In all cases, categorizing strategies ap- experiments need to be both replicated and pear to be adaptive given that they allow the extended. In particular, it would be useful to test animals to respond efficiently to sets of different and compare monkeys and apes confronted with similar with limited cognitive effort and memory. milar problems for which they would have to This latter feature can be exemplified in the infer a kin or a hierarchy relation in their con-experiment with dimensional stimuli (Dépy et al. specifics as well as their ability to use transitive
1997). Baboons adopted a strategy for process- inference within these relations. For example,
ing these stimuli based on the use of background movies of conspecifics could be shown to these color as a dominant cue for discrimination. It animals in which animal A is dominant over B could be stated that the multiplicity of these andBis dominant over C. Several indexes could strategies, as well as their opportunistic use, be used to assess dominance relationships, such ensures the production of responses that are as avoidance, access to food, and facial expres-adapted to changing and varied environmental sions or submissive gestures (de Waal 1982). The
pro-cedure to choose the dominant individual in a tween strategies expressed by humans, who used pair. Then they could be tested for their abilities propositional reasoning for categorization of to infer (transitive inference), for example, that A polymorphic stimuli (see also Fagot et al. 1998)
is dominant to C. and those expressed by baboons, who relied on
Fourth, the experiments that we carried out on perceptual discriminations. Such discrepancies, monkeys tested in two different setups (labora- which were also observed when we compared tory and more natural conditions) permit one to baboons and 3-year-old children (Bovet et al. evaluate in some way the respective role of these 2001), indicate the limits of a comparative ap-different environments in the expressions of cat- proach to categorization processes.
egorical competence. Thus the studies on func- Categorizing procedures used by animals tional categorization with untrained animals are simultaneously dependent on environmental (Bovet and Vauclair 1998, see earlier discussion) constraints on their cognitive competence (for suggest that when they were tested with bio- example, the nature and the salience of available logically relevant stimuli (grouping objects in stimuli) and their past experience. Attempts at food and nonfood categories has obvious eco- homogenizing stimuli and methods seem to be logical significance for these animals), monkeys insufficient to elicit the use of homologous pro-were able not only to judge the sameness between cedures in the species under comparison. Such physical objects but also the sameness of func- effort is, however, needed, along with a consid-tional concepts. Such a high degree of abstraction eration of the cerebral structures involved in and conceptualization by monkeys has, to my solving the task (Roberts 1996) in order to de-knowledge, not yet been reported in the litera- termine if the processing is homologous or if it is ture. Note, however, that chimpanzees were suc- an effect that directly derives from interspecific cessful in a conceptual matching task (Thompson cognitive differences.
et al. 1997), whereas sea lions approached this In any case, it can be expected, through such level of abstraction when tested in an equivalence investigations, to obtain novel data and inter-class membership task (Schusterman and Kastak pretations about categorization and to increase
1998). our knowledge concerning general cognitive
Moreover, and contrary to Premack's (1983) abilities and the levels of representation of the contention, we demonstrated that cognitive com- social relationships of monkeys and apes. More petence comparable to relational matching does generally, this enterprise of comparative cogni-not necessarily require previous training with tion should lead to a better comprehension of explicit tokens and symbols. In effect, our mon- human cognition, given that the latter is the keys' previous training only involved categorizing product of both our ontogenetic history and our objects that belonged to one of two categories, phylogenetic past.
and using the same-different relation between objects (within and between the two categories).
A final issue that needs to be briefly considered References
is related to the existence of likely differences be- Bovet, D. and Vauclair, J. (1998). Functional catego-tween monkeys' and humans' strategies in solving rization of objects and of their pictures in baboons categorization problems. Our experiments most (Papio anubis). Learning and Motivation 29: 309-322. of the time involved both monkeys and human Bovet, D. and Vauclair, J. (2000). Picture recognition participants who were tested with similar pro- in animals and in humans: A review. Behavioral Brain cedures. This practice, which is seldom followed Research 109: 143-165.
Jacques Vauclair 244
Bovet, D., Vauclair, J., and Blaye, A. (2001). Catego- Jitsumori, M. (1994). Category discrimination of rization and abstraction abilities in three-year-old chil- artificial polymorphous stimuli by rhesus monkeys dren and baboons. Submitted. ( Macaca mulatta). Quarterly Journal of Experimental
Cheney, D. L. and Seyfarth, R. M. (1980). Vocal rec- Psychololgy 47B: 371-386.
ognition in free-ranging vervet monkeys. Animal Mackintosh, N. J. (1995). Categorization by people Behaviour 28: 362-367. and pigeons: The twenty-second Bartlett memorial
lec-D'Amato, M. R. and van Sant, P. (1988). The person ture. Quarterly Journal of Experimental Psychololgy
concept in monkeys (Cebus apella). JournalofExperi- 48B: 193-214.
mental Psychology: Animal Behavior Processes 14: 43- May, B., Moody, D. B., and Stebbins, W. C. (1989).
55. Categorical perception of conspecific
communica-Dasser, V. (1988). A social concept in Java monkeys. tion sounds by Japanese macaques, Macaca fuscata. Animal Behaviour 36: 225-230. Journalofthe Acoustical SocietyofAmerica 85: 837-Dépy, D., Fagot, J., and Vauclair, J. (1997). Catego- 847.
risation of three-dimensional stimuli by humans and Premack, D. (1983). The codes of man and beasts. baboons: Search for prototype effects. Behavioural Behavioral and Brain Sciences 6: 125-167.
Processes, 39: 299-306. Ramus, F., Hauser, M. D., Miller, C., Morris, D., and
Dépy, D., Fagot, J., and Vauclair, J. (1998). Compara- Mehler, J. (2000). Language discrimination by human tive assessment of distance processing and hemispheric newborns and by cotton-top tamarin monkeys. Science specialization in humans (Homo sapiens) and baboons 288:349-351.
(Papio papio). Brain and Cognition 38: 165-182. Roberts, A. (1996). Comparison of cognitive functions
Dépy, D., Fagot, J., and Vauclair, J. (1999). Processing in human and non-human primates. Cognitive Brain of above/below categorical spatial relations by baboons Research 3: 319-327.
(Papio papio). Behavioural Processes 48: 1-9. Rosch, E. and Mervis, C. B. (1975). Family
resem-de Waal, F. B. M. (1982). Chimpanzee Politics. Lon- blances: Studies in the internal structure of categories.
don: Jonathan Cape. Cognitive Psychology 7: 573-605.
Ehret, G. (1992). Categorical perception of mouse-pup Savage-Rumbaugh, E. S., Rumbaugh, D. M., Smith, ultrasounds in the temporal domain. Animal Behaviour S. T., and Lawson, J. (1980). Reference: The linguistic
43:409-416. essential. Science 210: 92-925.
Fagot, J., Dépy, D., Vauclair, J., and Kruschke, J. K. Schrier, A. M., Angarella, R., and Povar, M. (1984). (1998). Associative learning in baboons and humans: Studies of concept formation by stumptail monkeys: Species differences in learned attention to visual fea- Concepts monkeys, humans and letter A. Journal of tures. Animal Cognition 1: 123-133. Experimental Psychology: Animal Behavior Processes
Gouteux, S., Thinus-Blanc, C., and Vauclair, J. (2001). 10: 564-584.
Rhesus monkeys use geometric and nongeometric Schusterman, R. J. and Kastak, D. (1998). Functional information during a reorientation task. Journal of equivalence in a California sea lion: Relevance to ani-Experimental Psychology: General 130: 505-519. mal social and communicative interactions. Animal
Herrnstein, R. J. (1990). Levels of stimulus control: A Behaviour 55: 1087-1095.
functional approach. Cognition 37: 133-166. Thompson, R. K. R. (1995). Natural and relational
Huber, L. and Lenz, R. (1993). A test of the linear concepts in animals. In Comparative Approaches to feature model of polymorphous concept discrimination Cognitive Science, H. Roitblat and J. A. Meyer, eds., with pigeons. Quarterly JournalofExperimental Psy- pp. 175-224. Cambridge, Mass.: MIT Press.
chology 46B: 1-18. Thompson, R. K. R., Oden, D. L., and Boysen, S. T.
Humphrey, N. (1976). The social function of intellect. (1997). Language-naive chimpanzees (Pan troglodytes) In Growing Points in Ethology, P. P. G. Bateson and R. Judge relations between relations in a conceptual A. Hinde, eds., pp. 303-317. New York: Cambridge matching-to-sample task. Journal of Experimental
Thompson R. K. R. and Oden, D. L. (2000). Categor-ical perception and conceptual judgments by non-human primates: The paleological monkey and the analogical ape. Cognitive Science 24: 363-396. Vauclair, J. (1996). Animal Cognition: Recent Devel-opments in Modern Comparative Psychology. Cam-bridge, Mass.: Harvard University Press.
Vauclair, J. and Fagot, J. (1994). A joystick system for the study of hemispheric asymmetries in nonhuman primates. In Current Primatology: Behavioral Neuro-science, Physiology and Reproduction. Vol. 3, J. R. Anderson, J.-J. Roeder, B. Thierry, and N. Herrensch-midt, eds., pp. 69-75. Strasbourg: University Louis Pasteur.
Vauclair, J. and Fagot, J. (1996). Categorization of alphanumeric characters by baboons (Papio papio): Within and between class stimulus discrimination. Current Psychology of Cognition 15: 449-462. von Fersen, L. and Lea, S. E. G. (1990). Category dis-crimination by pigeons using five polymorphous fea-tures. Journal of the Experimental Analysis of Behavior 54: 69-89.
Watanabe, S. (1993). Object-picture equivalence in the pigeon: An analysis with natural concept and pseudo concept discriminations. Behavioural Processes 30: 225-232.
Weary, D. M. (1989). Categorical perception of bird song: How do great tits (Parus major) perceive tempo-ral variation in their song? Journal of Comparative Psychology 103: 320-325.
Zayan, R. and Vauclair, J. (1998). Categories as para-digms for comparative cognition. Behavioural
Pro-cesses 42: 87-99.
