Judgment of conceptual identity in monkeys
6
Copyright 2001 Psychonomic Society, Inc. 470 The purpose of our experiments was to investigate ab- stract concept formation in baboons. One kind of con- cept formation is the judgment of conceptual identity, whereby the subject has to judge the relationship be- tween concepts’ sameness/difference at a conceptual level (Thompson & Oden, 2000). For example, the pair of letters AA is conceptually similar to the pair of letters BB rather than CD. In this case, the subjects are match- ing on the basis of the relations between the relations. In the few available experimental investigations of concep- tual matching, human infants (Tyrrell, Stauffer, & Snow- man, 1991; Tyrrell, Zingaro, & Minard, 1993) and chim- panzees (Premack, 1983; Thompson, Oden, & Boysen, 1997) have succeeded on this task, but monkeys have consistently failed on this kind of conceptual matching task (Grant-Webster, Gunderson, & Burbacher, 1990; Thompson & Oden, 1996, 2000; Tomasello & Call, 1997). According to Herrnstein (1990), categorization abilities in animals can be described at five levels, with increas- ing abstractness, including (1) discrimination, (2) catego- rization by rote, (3) open-ended categorization, (4) con- cepts, and (5) abstract relations. Level 5 of Herrnstein’s categorization is attained when a subject is able to use abstract relations not only between objects, but also be- tween concepts, as in conceptual matching or in concep- tual identity. In previous studies, using either a response general- ization paradigm (e.g., D’Amato & Van Sant, 1988) or a habituation/dishabituation paradigm (e.g., Zuberbühler, Cheney, & Seyfarth, 1999), monkeys have been shown to perceive items as equivalent or belonging to the same category. These experiments showed monkeys’ ability to form open-ended categories (D’Amato & Van Sant, 1988) or to use concepts (Zuberbühler et al., 1999), which correspond to Levels 3 and 4, respectively, of Herrn- stein’s classification scheme; however, monkeys have not demonstrated the ability to explicitly judge relational identity. To our knowledge, the only nonhuman animal for which evidence for conceptual identity has been found has been the chimpanzee. The main goal of our study was to test whether mon- keys are able to master conceptual identity of a some- what different type—namely, a same/different relation- ship between two functional categories. In a previous experiment, 4 Olive baboons were trained to categorize objects into two different functional classes, foods and nonfoods. Subsequent generalization tasks demonstrated positive transfer to novel exemplars of the classes (Bovet & Vauclair, 1998), a Level 4 ability according to Herrn- stein (1990). The same procedure, successive simple dis- criminations in a two-alternative forced choice proce- dure, is used in the present study. In Experiment 1, the monkeys had to judge two physical objects as same or different (perceptual identity). For example, they judged two apples as being the same or an apple and a padlock as being different. In a crucial test of conceptual identity (Experiment 2), the baboons had to combine their previ- ously acquired skills in order to classify as same two (dif- ferent) objects that belong to the same functional cate- gory (food or nonfood) and apply that learning to new exemplars. For example, they had to classify as same an apple and a banana or a padlock and a cup and as differ- Part of the experiments were carried out with a grant from the Lilly Foundation to the first author. We are grateful to William Hopkins and David Washburn for their suggestions on improving the manuscript and to Pierre Lucciani for giving us access to the monkeys at the Station de Pri- matologie of the CNRS at Rousset. Correspondence concerning this ar- ticle should be addressed to D. Bovet, Station Biologique, Univer- sité de Rennes 1, 35380 Paimpont, France (e-mail: [email protected]. univ-rennes1.fr).. BRIEF REPORTS Judgment of conceptual identity in monkeys DALILA BOVET University of Provence, Aix-en-Provence, France and Georgia State University, Atlanta, Georgia and JACQUES VAUCLAIR University of Provence, Aix-en-Provence, France Baboons (Papio anubis) were tested on categorization tasks at two different conceptual levels. The monkeys showed their ability (1) to judge as identical or different the objects belonging to two cate- gories, on a perceptual basis, and (2) to perform a judgment of conceptual identity—that is, to use the same/different relation between two previously learned categories. This latter experiment represents the first demonstration of judgment of conceptual identity in a monkey species. Psychonomic Bulletin & Review 2001, 8 (3), 470-475
-
Upload
independent -
Category
Documents
-
view
1 -
download
0
Transcript of Judgment of conceptual identity in monkeys
Copyright 2001 Psychonomic Society Inc 470
The purpose of our experiments was to investigate ab-stract concept formation in baboons One kind of con-cept formation is the judgment of conceptual identitywhereby the subject has to judge the relationship be-tween conceptsrsquo samenessdifference at a conceptuallevel (Thompson amp Oden 2000) For example the pairof letters AA is conceptually similar to the pair of lettersBB rather than CD In this case the subjects are match-ing on the basis of the relations between the relations Inthe few available experimental investigations of concep-tual matching human infants (Tyrrell Stauffer amp Snow-man 1991 Tyrrell Zingaro amp Minard 1993) and chim-panzees (Premack 1983 Thompson Oden amp Boysen1997) have succeeded on this task but monkeys haveconsistently failed on this kind of conceptual matchingtask (Grant-Webster Gunderson amp Burbacher 1990Thompson amp Oden 1996 2000 Tomasello amp Call1997)
According to Herrnstein (1990) categorization abilitiesin animals can be described at five levels with increas-ing abstractness including (1) discrimination (2) catego-rization by rote (3) open-ended categorization (4) con-cepts and (5) abstract relations Level 5 of Herrnsteinrsquoscategorization is attained when a subject is able to useabstract relations not only between objects but also be-tween concepts as in conceptual matching or in concep-tual identity
In previous studies using either a response general-ization paradigm (eg DrsquoAmato amp Van Sant 1988) or ahabituationdishabituation paradigm (eg ZuberbuumlhlerCheney amp Seyfarth 1999) monkeys have been shown toperceive items as equivalent or belonging to the samecategory These experiments showed monkeysrsquo ability toform open-ended categories (DrsquoAmato amp Van Sant1988) or to use concepts (Zuberbuumlhler et al 1999) whichcorrespond to Levels 3 and 4 respectively of Herrn-steinrsquos classification scheme however monkeys have notdemonstrated the ability to explicitly judge relationalidentity To our knowledge the only nonhuman animalfor which evidence for conceptual identity has beenfound has been the chimpanzee
The main goal of our study was to test whether mon-keys are able to master conceptual identity of a some-what different typemdashnamely a samedifferent relation-ship between two functional categories In a previousexperiment 4 Olive baboons were trained to categorizeobjects into two different functional classes foods andnonfoods Subsequent generalization tasks demonstratedpositive transfer to novel exemplars of the classes (Bovetamp Vauclair 1998) a Level 4 ability according to Herrn-stein (1990) The same procedure successive simple dis-criminations in a two-alternative forced choice proce-dure is used in the present study In Experiment 1 themonkeys had to judge two physical objects as same ordifferent (perceptual identity) For example they judgedtwo apples as being the same or an apple and a padlockas being different In a crucial test of conceptual identity(Experiment 2) the baboons had to combine their previ-ously acquired skills in order to classify as same two (dif-ferent) objects that belong to the same functional cate-gory (food or nonfood) and apply that learning to newexemplars For example they had to classify as same anapple and a banana or a padlock and a cup and as differ-
Part of the experiments were carried out with a grant from the LillyFoundation to the first author We are grateful to William Hopkins andDavid Washburn for their suggestions on improving the manuscript andto Pierre Lucciani for giving us access to the monkeys at the Station de Pri-matologie of the CNRS at Rousset Correspondence concerning this ar-ticle should be addressed to D Bovet Station Biologique Univer-siteacute de Rennes 1 35380 Paimpont France (e-mail dbovetpaimpontsbpuniv-rennes1fr)
BRIEF REPORTS
Judgment of conceptual identity in monkeys
DALILA BOVETUniversity of Provence Aix-en-Provence France and Georgia State University Atlanta Georgia
and
JACQUES VAUCLAIRUniversity of Provence Aix-en-Provence France
Baboons (Papio anubis) were tested on categorization tasks at two different conceptual levels Themonkeys showed their ability (1) to judge as identical or different the objects belonging to two cate-gories on a perceptual basis and (2) to perform a judgment of conceptual identitymdashthat is to use thesamedifferent relation between two previously learned categories This latter experiment representsthe first demonstration of judgment of conceptual identity in a monkey species
Psychonomic Bulletin amp Review2001 8 (3) 470-475
CONCEPTUAL IDENTITY IN MONKEYS 471
ent an apple and a padlock (see Figure 1) This ability cor-responds to Level 5 of Herrnsteinrsquos classification scheme
GENERAL METHOD
SubjectsThe subjects were 4 laboratory-born adult baboons (Papio anu-
bis)mdash2 males Sylvestre (17 years old) and Balthazar (14 yearsold) and 2 females Esperance (9 years old) and Ida (5 years old)The baboons were housed in social groups (1 male and 5ndash8 fe-males) reared in indoor and outdoor quarters (35 square meterseach) at the Station de Primatologie of the Centre National de laRecherche Scientifique Rousset France The subjects were notfood deprived but they received their daily food ration (fruit mon-key chow and vegetables) at the end of the daily training and test-ing The subjects had been previously trained to categorize objectsinto food or nonfood categories (Bovet amp Vauclair 1998)
ApparatusThe monkeys were individually tested in their home enclosure
with an adapted version of a Wisconsin General Test Apparatus (seeFigure 2) The apparatus was attached to the bars of the enclosurefor the experimental sessions It was made of a vertical board (65 380 cm) consisting of a one-way screen that concealed the testerfrom the baboonsrsquo view during the experiments but allowed the ex-perimenter to see the subjects a horizontal board on which to pre-
sent the stimuli behind a Plexiglas window and two openings forthe manipulanda (two ropes)
StimuliFood and nonfood objects were used as stimuli Food objects
consisted of various types of vegetables plants cereals fruits andsweets Nonfood objects consisted of natural and man-made objectsof various forms and materials (eg glass plastic wood metalpaper textile stone) The objects of both categories were matchedas closely as possible for size and color so that two objects from thesame categories did not resemble each other more than two objectsfrom different categories Prior to each experiment all the stimulusobjects had been left in the baboonsrsquo enclosure for 1 full day Thusthe monkeys had had ample opportunity to manipulate and or eatthe objects prior to testing Manipulation time varied across indi-viduals and objects The food items were generally quickly eatenbut some baboons manipulated the other objects for a long timewhereas others showed indifference toward the nonfood objects
Procedure At the start of each trial the experimenter placed the two objects
on the board while the subjectrsquos view of the board was blocked byan opaque screen The screen was then raised and the subject had5 sec to respond by pulling one of the two ropes according to therelation (same or different) between the presented objects (Experi-ment 1) or between the presented categories (Experiment 2) If the
Figure 1 Tasks successively mastered by the baboons [ belongs to 5 perceptual or conceptual iden-tity F food a apple b banana p padlock c cup
472 BOVET AND VAUCLAIR
subject did not respond during this interval or if he or she did notpull the rope to the end an opaque screen was lowered for 5 secmasking the subjectrsquos view of the board The screen was then raisedagain and the same objects were presented The subjects rarelyfailed to respond within the first 5 sec If the subjectrsquos response wascorrect a food reward (small pieces of apple or grains of corn) wasprovided and another pair of objects was immediately presented(within 5 sec) In the event of an error a 10-second time-out was in-troduced followed by the presentation of a new object The orien-tation of the objects varied over trials Each baboon was trainedtested about 1 h a day 5 days a week which represents an averageof 185 trials per day
EXPERIMENT 1Perceptual Identity
MethodThe monkeysrsquo task was to pull one rope when the two presented
objects were the same and the other rope when the two objects weredifferent The rope on the right side was associated with same for 2of the monkeys and with different for the other 2 monkeys In thisphase when two different objects were presented they were alwaystwo objects belonging to different categories The subjects werefirst trained with two pairs of objects (four possible combinations)and then to assess transfer when they reached a performance cri-terion of 80 correct responses for 100 consecutive trials pairs in-volving new objects were presented
A total of 36 different objects (18 food and 18 nonfood) wereused in 684 possible combinations That is there were 648 (36 X18) trials with different pairs (two different combinations weremade with each pair of different objects according to the relativeposition of the 2 objects) and 36 trials with same pairs presented toeach subject Each new pair of objects was first presented in the 4possible combinations between the 2 objects of the pairs then eachobject was presented in combination with all the preceding objects
Each new combination of objects was tested until a subject com-pleted 8 out of 10 consecutive trials correctly To avoid responsebias trials with novel pairs of objects were interspersed among tri-als involving familiar pairs and trials with different objects wereinterspersed among trials involving pairs of same objects (mostlypairs of familiar objects since most of the new combinations werepairs of different objects) Thus from 1 to 3 familiar pairs of objectswere interspersed with trials involving each novel object Todemonstrate a samendashdifferent judgment we are primarily con-cerned with first-trial data for each novel pair of objects Thereforethe analysis focused on these data
ResultsWe separated the 36 stimuli into nine blocks of four ob-
jects Each point on the functions presented in Figure 3represents the percentage of correct responses for thefirst presentation of each of those four objects with allthe preceding objects
On average it took the baboons 12 objects whichmeans 84 tests (6 3 12 different trials 1 12 same trials)to reach a performance criterion of 80 correct re-sponses on the first trials with new pairs of objects Thiscriterion was achieved in an average of 9689 trials perindividual
The baboons performances continued to improve overtrials involving the next 24 objects (Figure 3) At the endof the experiment all 4 subjects were tested on the last 4novel objects presented in combination with all of theprevious stimuli This phase represented 140 novel ob-ject combinations (corresponding to the last point of thecurve on Figure 3) The accuracy levels were 96 [x 2(1)5 120 p lt 001] 91 [x 2(1) 5 92 p lt 001] 91
Figure 2 Test apparatus The background is indicated by dashed lines A wooden board B one-way mir-ror C rope sliding through a hole and terminated by a wooden ball D transparent Plexiglas E board forplacing the stimulus object F food dispenser
CONCEPTUAL IDENTITY IN MONKEYS 473
[x 2(1) 5 92 p lt 001] and 94 [x 2(1) 5 109 p lt 001]respectively
If we focus only on the trials involving new object pairsthe baboonsrsquo performances are still good On the last 12pairs of objects (48 combinations) the baboons achievedscores of 875 for Sylvestre [x 2(1) 5 27 p lt 001]917 for Balthazar [x2(1) 5 33 p lt 001] 896 forEsperance [x 2(1) 5 30 p lt 001] and 854 for Ida[x2(1) 5 24 p lt 001]These results provide evidencethat the baboons were able to class as same or differenttwo physically identical or different objects
EXPERIMENT 2Conceptual Identity
MethodThe subjects were the 2 males involved in the previous experiments
The 2 females gave birth during Experiment 1 and thus became lessmotivated during the test sessions They finished Experiment 1 muchlater than the 2 males and thus were not included in Experiment 2
The task in Experiment 2 was to pull one rope (the right side ropefor Sylvestre the left side rope for Balthazar) when the two pre-sented objects belonged to the same category (two foods or twononfoods) and to pull the other rope when the objects belonged todifferent categories (one food and one nonfood) The baboons werefirst trained with one block of four different objects (shown to themin the six possible combinations) and new blocks of four objectswere introduced after a performance criterion of 80 (out of 100consecutive trials) was reached
Each new block of four objects was first presented in the 6 pos-sible combinations involving the four objects and then each ofthose new objects was presented in combination with each of thepreceding objects A total of 36 different objects (18 food and 18nonfood) was used and all 630 pairwise combinations were pre-sented to each subject As in Experiment 1 eight correct responsesout of 10 consecutive trials were required for a given pair of objectsbefore a novel pair was presented and trials with novel pairs of ob-jects were interspersed among trials involving familiar pairs Heretoo we focused on the first-trial performance during successivetransfer tests for each novel pair of objects
ResultsBoth baboons started this experiment responding as
they did in the first experiment that is by pulling therope for different when two different objects were pre-sented even when those two objects belonged to thesame category Thus their performance was far belowchance at the beginning of the experiment On averageit took 14 objects (91 transfer tests) to reach the criterionof 80 correct on trials involving one or two new ob-jects (see Figure 4) This criterion was reached after anaverage of 14576 trials (17851 for Sylvestre 11300 forBalthazar)
At the end of the experiment the last four objects werepresented in combination with each of the previous ob-jects This allowed for 134 novel combinations Both ba-boons obtained a high score (see Figure 4) 91 forSylvestre [x 2(1) 5 229 p lt 001] and 81 for Balthazar[x 2(1) 5 174 p lt 001] The baboonsrsquo performances areeven better if we consider the trials involving novel ob-jects only over the last six blocks (see Table 1) Such per-centages provide convincing evidence that the monkeysmastered this conceptual identity task
GENERAL DISCUSSION
Categorization is a fundamental component of manycognitive processes and has therefore been the subjectof many animal experiments (see Herrnstein 1990Thompson 1995 Vauclair 1996 and Zayan amp Vauclair1998 for reviews) However very few studies have ex-amined functional categorization in animals Functionalcategorization involving food has been investigated byWatanabe (1997) in pigeons and by Savage-RumbaughRumbaugh Smith and Lawson (1980) in chimpanzeesOther investigators have studied the formation of func-tional categories in pigeons using a discrimination pro-cedure involving repeated reversals (Vaughan 1988)Similarly functional or equivalence class membership
correct responses
100
90
80
70
60
50
40
30
20
10
0
Sylvestre
Balthazar
Espeacuterance
Ida
Number of blocks of 4 objects1 2 3 4 5 6 7 8 9
Figure 3 Percentages of correct responses for the first presen-tation of a novel pair of objects for perceptual identity
Figure 4 Percentages of correct responses for the first presen-tation of a novel pair of objects for conceptual identity
474 BOVET AND VAUCLAIR
has been studied in a California sea lion using a matching-to-sample procedure (Schusterman amp Kastak 1998)
We decided to use this kind of categorization (inwhich the members of the class share a functional prop-erty but do not look alike) because as has been ex-plained by Herrnstein (1990) conceptual categorizationis more abstract than perceptual categorization (forwhich the stimuli share visual properties and are thusphysically similar) Our task involved functional ratherthan perceptual categorization because the only possibleproperty the baboons could have used to classify the ob-jects was their edibility It is of interest to note that thesubjectsrsquo classificatory errors could often be explainedby their food preferences which varied across the indi-viduals Thus uneaten food items (eg marshmallowsfor Sylvestre) were classified in the nonfood category(see Savage-Rumbaugh et al 1980 for similar findings)
In Experiment 2 the hypothesis that the baboonscould master the task by means of a simple discrimina-tion based on the total hedonic value of the display (ifmoderate hedonic value then respond different if highor no hedonic value then respond same) can be dis-missed In fact no errors occurred when one highly val-ued food object was presented with a nonhedonic objector when two low valued food objects or two small foodobjects were presented
The use of three-dimensional (3-D) stimuli (instead ofthe two-dimensional stimuli more typically used in thiskind of research) was chosen to facilitate functional cat-egorization in our previous study (Bovet amp Vauclair1998) It is likely that the use of 3-D stimuli made thefunction of the objects more salient (Bovet amp Vauclair2000) and therefore augmented the rapid generalizationfrom a few training objects to novel food and nonfoodobjects and facilitated the demonstration of conceptualidentity
Numerous training trials were required for the ba-boons to reach the 80 correct criterion in the two ex-periments reported here It is likely that the monkeys re-lied on rote learning of all the combinations in theseearly trials Indeed they required fewer trials to acquirethe perceptual identity task than to acquire the concep-tual identity task This is not surprising given the greaterconceptual complexity of the latter task
For Experiment 1 the identity rule is not the only pos-sible explanation for the baboonsrsquo successful perfor-mance The subjects could also have learned to discrim-inate the stimulus arrays on the basis of symmetry (astimuli configuration consisting of two identical objectsis bilaterally symmetrical whereas a configuration con-sisting of two different objects is not) However in Ex-
periment 2 the configuration could not be used to solvethe task because the two objects presented were alwaysdifferent and consequently the configuration was neverperceptually symmetrical
In Experiment 2 for each block of four objects therewere four different-categories trials and two same-categorytrials The ratio of same to different is thus not equalHowever the results show that the subjects were not re-sponding with a bias toward the different-categories re-sponse because both baboons made slightly more errorsfor the different-categories trials than for the same-categorytrials (Table 1)
The high level of performance attained by the 2 ba-boons at the end of Experiment 2 with totally novel ob-jects (ie objects novel in the task but left in the mon-keysrsquo enclosure before the experiment) demonstratesmastery of the samendashdifferent relation and the ability toconceptually judge as same or different the previouslylearned categories
Moreover contrary to Premackrsquos (1983) contentioncognitive competence comparable to relational matchingdoes not require previous training with explicit tokensand symbols Our subjectsrsquo previous training only in-volved categorizing objects that belonged to one of twocategories and using the samendashdifferent relation betweenobjects (within and between the two categories)
Our experiments suggest that when tested with bio-logically relevant stimuli (grouping objects in food andnonfood categories has obvious ecological significancefor these animals) monkeys are able to judge not onlythe sameness between physical objects but also the same-ness of functional concepts Such a high degree of ab-straction and conceptualization by monkeys has not pre-viously been reported in the literature
REFERENCES
Bovet D amp Vauclair J (1998) Functional categorization of objectsand of their pictures in baboons (Papio anubis) Learning amp Motiva-tion 29 309-322
Bovet D amp Vauclair J (2000) Picture recognition in animals andin humans A review Behavioral Brain Research 109 143-165
DrsquoAmato M R amp Van Sant P (1988) The person concept in mon-keys (Cebus apella) Journal of Experimental Psychology AnimalBehavior Processes 14 43-56
Grant-Webster K S Gunderson V M amp Burbacher T M (1990)Behavioural assessment of young non-human primates Perceptual-cognitive development Neurotoxicology amp Teratology 12 543-546
Herrnstein R J (1990) Levels of stimulus control A functional ap-proach Cognition 37 133-166
Premack D (1983) The codes of man and beasts Behavioral amp BrainSciences 6 125-167
Savage-Rumbaugh E S Rumbaugh D M Smith S T amp Law-
Table 1 Percentages of Correct Responses Given for the First Presentation of the Combinations Realized
With a New Block of Four Objects for the Last Six Blocks in Experiment 2
Subject Same Category (n 5 12) x2 Different Categories (n = 24) x2 Total (n 5 36) x2
Sylvestre 1000 12 917 16 944 28Balthazar 917 8dagger 792 8dagger 833 16
p lt 001 daggerp lt 01
CONCEPTUAL IDENTITY IN MONKEYS 475
son J (1980 November) Reference The linguistic essential Sci-ence 210 922-925
Schusterman R J amp Kastak D (1998) Functional equivalence ina California sea lion Relevance to animal social and communicativeinteractions Animal Behaviour 55 1087-1095
Thompson R K R (1995) Natural and relational concepts in animalsIn H Roitblat amp J A Meyer (Eds) Comparative approaches of cog-nitive sciences (pp 175-224) Cambridge MA MIT Press
Thompson R K R amp Oden D L (1996) A profound disparity re-visited Perception and judgment of abstract identity relations bychimpanzees human infants and monkeys Behavioural Processes35 149-161
Thompson R K R amp Oden D L (2000) Categorical perception ampconceptual judgments by nonhuman primates The paleological mon-key and the analogical ape Cognitive Science 24 363-396
Thompson R K R Oden D L amp Boysen S T (1997) Language-naive chimpanzees (Pan troglodytes) judge relations between rela-tions in a conceptual matching-to-sample task Journal of Experi-mental Psychology Animal Behavior Processes 23 31-43
Tomasello M amp Call J (1997) Primate cognition New York Ox-ford University Press
Tyrrell D J Stauffer L B amp Snowman L G (1991) Perception
of abstract identitydifference relationships by infants Infant Behav-ior amp Development 14 125-129
Tyrrell D J Zingaro M C amp Minard K L (1993) Learning andtransfer of identitydifference relationships by infants Infant Behav-ior amp Development 16 43-52
Vauclair J (1996) Animal cognition Recent developments in mod-ern comparative psychology Cambridge MA Harvard UniversityPress
Vaughan W Jr (1988) Formation of equivalence sets in pigeonsJournal of Experimental Psychology Animal Behavior Processes14 36-42
Watanabe S (1997) Visual discrimination of real objects and picturesin pigeons Animal Learning amp Behavior 25 185-192
Zayan R amp Vauclair J (1998) Categories as paradigms for com-parative cognition Behavioural Processes 42 87-99
Zuberbuumlhler K Cheney D L amp Seyfarth R M (1999) Con-ceptual semantics in a nonhuman primate Journal of ComparativePsychology 113 33-42
(Manuscript received March 1 2000 revision accepted for publication August 10 2000)
CONCEPTUAL IDENTITY IN MONKEYS 471
ent an apple and a padlock (see Figure 1) This ability cor-responds to Level 5 of Herrnsteinrsquos classification scheme
GENERAL METHOD
SubjectsThe subjects were 4 laboratory-born adult baboons (Papio anu-
bis)mdash2 males Sylvestre (17 years old) and Balthazar (14 yearsold) and 2 females Esperance (9 years old) and Ida (5 years old)The baboons were housed in social groups (1 male and 5ndash8 fe-males) reared in indoor and outdoor quarters (35 square meterseach) at the Station de Primatologie of the Centre National de laRecherche Scientifique Rousset France The subjects were notfood deprived but they received their daily food ration (fruit mon-key chow and vegetables) at the end of the daily training and test-ing The subjects had been previously trained to categorize objectsinto food or nonfood categories (Bovet amp Vauclair 1998)
ApparatusThe monkeys were individually tested in their home enclosure
with an adapted version of a Wisconsin General Test Apparatus (seeFigure 2) The apparatus was attached to the bars of the enclosurefor the experimental sessions It was made of a vertical board (65 380 cm) consisting of a one-way screen that concealed the testerfrom the baboonsrsquo view during the experiments but allowed the ex-perimenter to see the subjects a horizontal board on which to pre-
sent the stimuli behind a Plexiglas window and two openings forthe manipulanda (two ropes)
StimuliFood and nonfood objects were used as stimuli Food objects
consisted of various types of vegetables plants cereals fruits andsweets Nonfood objects consisted of natural and man-made objectsof various forms and materials (eg glass plastic wood metalpaper textile stone) The objects of both categories were matchedas closely as possible for size and color so that two objects from thesame categories did not resemble each other more than two objectsfrom different categories Prior to each experiment all the stimulusobjects had been left in the baboonsrsquo enclosure for 1 full day Thusthe monkeys had had ample opportunity to manipulate and or eatthe objects prior to testing Manipulation time varied across indi-viduals and objects The food items were generally quickly eatenbut some baboons manipulated the other objects for a long timewhereas others showed indifference toward the nonfood objects
Procedure At the start of each trial the experimenter placed the two objects
on the board while the subjectrsquos view of the board was blocked byan opaque screen The screen was then raised and the subject had5 sec to respond by pulling one of the two ropes according to therelation (same or different) between the presented objects (Experi-ment 1) or between the presented categories (Experiment 2) If the
Figure 1 Tasks successively mastered by the baboons [ belongs to 5 perceptual or conceptual iden-tity F food a apple b banana p padlock c cup
472 BOVET AND VAUCLAIR
subject did not respond during this interval or if he or she did notpull the rope to the end an opaque screen was lowered for 5 secmasking the subjectrsquos view of the board The screen was then raisedagain and the same objects were presented The subjects rarelyfailed to respond within the first 5 sec If the subjectrsquos response wascorrect a food reward (small pieces of apple or grains of corn) wasprovided and another pair of objects was immediately presented(within 5 sec) In the event of an error a 10-second time-out was in-troduced followed by the presentation of a new object The orien-tation of the objects varied over trials Each baboon was trainedtested about 1 h a day 5 days a week which represents an averageof 185 trials per day
EXPERIMENT 1Perceptual Identity
MethodThe monkeysrsquo task was to pull one rope when the two presented
objects were the same and the other rope when the two objects weredifferent The rope on the right side was associated with same for 2of the monkeys and with different for the other 2 monkeys In thisphase when two different objects were presented they were alwaystwo objects belonging to different categories The subjects werefirst trained with two pairs of objects (four possible combinations)and then to assess transfer when they reached a performance cri-terion of 80 correct responses for 100 consecutive trials pairs in-volving new objects were presented
A total of 36 different objects (18 food and 18 nonfood) wereused in 684 possible combinations That is there were 648 (36 X18) trials with different pairs (two different combinations weremade with each pair of different objects according to the relativeposition of the 2 objects) and 36 trials with same pairs presented toeach subject Each new pair of objects was first presented in the 4possible combinations between the 2 objects of the pairs then eachobject was presented in combination with all the preceding objects
Each new combination of objects was tested until a subject com-pleted 8 out of 10 consecutive trials correctly To avoid responsebias trials with novel pairs of objects were interspersed among tri-als involving familiar pairs and trials with different objects wereinterspersed among trials involving pairs of same objects (mostlypairs of familiar objects since most of the new combinations werepairs of different objects) Thus from 1 to 3 familiar pairs of objectswere interspersed with trials involving each novel object Todemonstrate a samendashdifferent judgment we are primarily con-cerned with first-trial data for each novel pair of objects Thereforethe analysis focused on these data
ResultsWe separated the 36 stimuli into nine blocks of four ob-
jects Each point on the functions presented in Figure 3represents the percentage of correct responses for thefirst presentation of each of those four objects with allthe preceding objects
On average it took the baboons 12 objects whichmeans 84 tests (6 3 12 different trials 1 12 same trials)to reach a performance criterion of 80 correct re-sponses on the first trials with new pairs of objects Thiscriterion was achieved in an average of 9689 trials perindividual
The baboons performances continued to improve overtrials involving the next 24 objects (Figure 3) At the endof the experiment all 4 subjects were tested on the last 4novel objects presented in combination with all of theprevious stimuli This phase represented 140 novel ob-ject combinations (corresponding to the last point of thecurve on Figure 3) The accuracy levels were 96 [x 2(1)5 120 p lt 001] 91 [x 2(1) 5 92 p lt 001] 91
Figure 2 Test apparatus The background is indicated by dashed lines A wooden board B one-way mir-ror C rope sliding through a hole and terminated by a wooden ball D transparent Plexiglas E board forplacing the stimulus object F food dispenser
CONCEPTUAL IDENTITY IN MONKEYS 473
[x 2(1) 5 92 p lt 001] and 94 [x 2(1) 5 109 p lt 001]respectively
If we focus only on the trials involving new object pairsthe baboonsrsquo performances are still good On the last 12pairs of objects (48 combinations) the baboons achievedscores of 875 for Sylvestre [x 2(1) 5 27 p lt 001]917 for Balthazar [x2(1) 5 33 p lt 001] 896 forEsperance [x 2(1) 5 30 p lt 001] and 854 for Ida[x2(1) 5 24 p lt 001]These results provide evidencethat the baboons were able to class as same or differenttwo physically identical or different objects
EXPERIMENT 2Conceptual Identity
MethodThe subjects were the 2 males involved in the previous experiments
The 2 females gave birth during Experiment 1 and thus became lessmotivated during the test sessions They finished Experiment 1 muchlater than the 2 males and thus were not included in Experiment 2
The task in Experiment 2 was to pull one rope (the right side ropefor Sylvestre the left side rope for Balthazar) when the two pre-sented objects belonged to the same category (two foods or twononfoods) and to pull the other rope when the objects belonged todifferent categories (one food and one nonfood) The baboons werefirst trained with one block of four different objects (shown to themin the six possible combinations) and new blocks of four objectswere introduced after a performance criterion of 80 (out of 100consecutive trials) was reached
Each new block of four objects was first presented in the 6 pos-sible combinations involving the four objects and then each ofthose new objects was presented in combination with each of thepreceding objects A total of 36 different objects (18 food and 18nonfood) was used and all 630 pairwise combinations were pre-sented to each subject As in Experiment 1 eight correct responsesout of 10 consecutive trials were required for a given pair of objectsbefore a novel pair was presented and trials with novel pairs of ob-jects were interspersed among trials involving familiar pairs Heretoo we focused on the first-trial performance during successivetransfer tests for each novel pair of objects
ResultsBoth baboons started this experiment responding as
they did in the first experiment that is by pulling therope for different when two different objects were pre-sented even when those two objects belonged to thesame category Thus their performance was far belowchance at the beginning of the experiment On averageit took 14 objects (91 transfer tests) to reach the criterionof 80 correct on trials involving one or two new ob-jects (see Figure 4) This criterion was reached after anaverage of 14576 trials (17851 for Sylvestre 11300 forBalthazar)
At the end of the experiment the last four objects werepresented in combination with each of the previous ob-jects This allowed for 134 novel combinations Both ba-boons obtained a high score (see Figure 4) 91 forSylvestre [x 2(1) 5 229 p lt 001] and 81 for Balthazar[x 2(1) 5 174 p lt 001] The baboonsrsquo performances areeven better if we consider the trials involving novel ob-jects only over the last six blocks (see Table 1) Such per-centages provide convincing evidence that the monkeysmastered this conceptual identity task
GENERAL DISCUSSION
Categorization is a fundamental component of manycognitive processes and has therefore been the subjectof many animal experiments (see Herrnstein 1990Thompson 1995 Vauclair 1996 and Zayan amp Vauclair1998 for reviews) However very few studies have ex-amined functional categorization in animals Functionalcategorization involving food has been investigated byWatanabe (1997) in pigeons and by Savage-RumbaughRumbaugh Smith and Lawson (1980) in chimpanzeesOther investigators have studied the formation of func-tional categories in pigeons using a discrimination pro-cedure involving repeated reversals (Vaughan 1988)Similarly functional or equivalence class membership
correct responses
100
90
80
70
60
50
40
30
20
10
0
Sylvestre
Balthazar
Espeacuterance
Ida
Number of blocks of 4 objects1 2 3 4 5 6 7 8 9
Figure 3 Percentages of correct responses for the first presen-tation of a novel pair of objects for perceptual identity
Figure 4 Percentages of correct responses for the first presen-tation of a novel pair of objects for conceptual identity
474 BOVET AND VAUCLAIR
has been studied in a California sea lion using a matching-to-sample procedure (Schusterman amp Kastak 1998)
We decided to use this kind of categorization (inwhich the members of the class share a functional prop-erty but do not look alike) because as has been ex-plained by Herrnstein (1990) conceptual categorizationis more abstract than perceptual categorization (forwhich the stimuli share visual properties and are thusphysically similar) Our task involved functional ratherthan perceptual categorization because the only possibleproperty the baboons could have used to classify the ob-jects was their edibility It is of interest to note that thesubjectsrsquo classificatory errors could often be explainedby their food preferences which varied across the indi-viduals Thus uneaten food items (eg marshmallowsfor Sylvestre) were classified in the nonfood category(see Savage-Rumbaugh et al 1980 for similar findings)
In Experiment 2 the hypothesis that the baboonscould master the task by means of a simple discrimina-tion based on the total hedonic value of the display (ifmoderate hedonic value then respond different if highor no hedonic value then respond same) can be dis-missed In fact no errors occurred when one highly val-ued food object was presented with a nonhedonic objector when two low valued food objects or two small foodobjects were presented
The use of three-dimensional (3-D) stimuli (instead ofthe two-dimensional stimuli more typically used in thiskind of research) was chosen to facilitate functional cat-egorization in our previous study (Bovet amp Vauclair1998) It is likely that the use of 3-D stimuli made thefunction of the objects more salient (Bovet amp Vauclair2000) and therefore augmented the rapid generalizationfrom a few training objects to novel food and nonfoodobjects and facilitated the demonstration of conceptualidentity
Numerous training trials were required for the ba-boons to reach the 80 correct criterion in the two ex-periments reported here It is likely that the monkeys re-lied on rote learning of all the combinations in theseearly trials Indeed they required fewer trials to acquirethe perceptual identity task than to acquire the concep-tual identity task This is not surprising given the greaterconceptual complexity of the latter task
For Experiment 1 the identity rule is not the only pos-sible explanation for the baboonsrsquo successful perfor-mance The subjects could also have learned to discrim-inate the stimulus arrays on the basis of symmetry (astimuli configuration consisting of two identical objectsis bilaterally symmetrical whereas a configuration con-sisting of two different objects is not) However in Ex-
periment 2 the configuration could not be used to solvethe task because the two objects presented were alwaysdifferent and consequently the configuration was neverperceptually symmetrical
In Experiment 2 for each block of four objects therewere four different-categories trials and two same-categorytrials The ratio of same to different is thus not equalHowever the results show that the subjects were not re-sponding with a bias toward the different-categories re-sponse because both baboons made slightly more errorsfor the different-categories trials than for the same-categorytrials (Table 1)
The high level of performance attained by the 2 ba-boons at the end of Experiment 2 with totally novel ob-jects (ie objects novel in the task but left in the mon-keysrsquo enclosure before the experiment) demonstratesmastery of the samendashdifferent relation and the ability toconceptually judge as same or different the previouslylearned categories
Moreover contrary to Premackrsquos (1983) contentioncognitive competence comparable to relational matchingdoes not require previous training with explicit tokensand symbols Our subjectsrsquo previous training only in-volved categorizing objects that belonged to one of twocategories and using the samendashdifferent relation betweenobjects (within and between the two categories)
Our experiments suggest that when tested with bio-logically relevant stimuli (grouping objects in food andnonfood categories has obvious ecological significancefor these animals) monkeys are able to judge not onlythe sameness between physical objects but also the same-ness of functional concepts Such a high degree of ab-straction and conceptualization by monkeys has not pre-viously been reported in the literature
REFERENCES
Bovet D amp Vauclair J (1998) Functional categorization of objectsand of their pictures in baboons (Papio anubis) Learning amp Motiva-tion 29 309-322
Bovet D amp Vauclair J (2000) Picture recognition in animals andin humans A review Behavioral Brain Research 109 143-165
DrsquoAmato M R amp Van Sant P (1988) The person concept in mon-keys (Cebus apella) Journal of Experimental Psychology AnimalBehavior Processes 14 43-56
Grant-Webster K S Gunderson V M amp Burbacher T M (1990)Behavioural assessment of young non-human primates Perceptual-cognitive development Neurotoxicology amp Teratology 12 543-546
Herrnstein R J (1990) Levels of stimulus control A functional ap-proach Cognition 37 133-166
Premack D (1983) The codes of man and beasts Behavioral amp BrainSciences 6 125-167
Savage-Rumbaugh E S Rumbaugh D M Smith S T amp Law-
Table 1 Percentages of Correct Responses Given for the First Presentation of the Combinations Realized
With a New Block of Four Objects for the Last Six Blocks in Experiment 2
Subject Same Category (n 5 12) x2 Different Categories (n = 24) x2 Total (n 5 36) x2
Sylvestre 1000 12 917 16 944 28Balthazar 917 8dagger 792 8dagger 833 16
p lt 001 daggerp lt 01
CONCEPTUAL IDENTITY IN MONKEYS 475
son J (1980 November) Reference The linguistic essential Sci-ence 210 922-925
Schusterman R J amp Kastak D (1998) Functional equivalence ina California sea lion Relevance to animal social and communicativeinteractions Animal Behaviour 55 1087-1095
Thompson R K R (1995) Natural and relational concepts in animalsIn H Roitblat amp J A Meyer (Eds) Comparative approaches of cog-nitive sciences (pp 175-224) Cambridge MA MIT Press
Thompson R K R amp Oden D L (1996) A profound disparity re-visited Perception and judgment of abstract identity relations bychimpanzees human infants and monkeys Behavioural Processes35 149-161
Thompson R K R amp Oden D L (2000) Categorical perception ampconceptual judgments by nonhuman primates The paleological mon-key and the analogical ape Cognitive Science 24 363-396
Thompson R K R Oden D L amp Boysen S T (1997) Language-naive chimpanzees (Pan troglodytes) judge relations between rela-tions in a conceptual matching-to-sample task Journal of Experi-mental Psychology Animal Behavior Processes 23 31-43
Tomasello M amp Call J (1997) Primate cognition New York Ox-ford University Press
Tyrrell D J Stauffer L B amp Snowman L G (1991) Perception
of abstract identitydifference relationships by infants Infant Behav-ior amp Development 14 125-129
Tyrrell D J Zingaro M C amp Minard K L (1993) Learning andtransfer of identitydifference relationships by infants Infant Behav-ior amp Development 16 43-52
Vauclair J (1996) Animal cognition Recent developments in mod-ern comparative psychology Cambridge MA Harvard UniversityPress
Vaughan W Jr (1988) Formation of equivalence sets in pigeonsJournal of Experimental Psychology Animal Behavior Processes14 36-42
Watanabe S (1997) Visual discrimination of real objects and picturesin pigeons Animal Learning amp Behavior 25 185-192
Zayan R amp Vauclair J (1998) Categories as paradigms for com-parative cognition Behavioural Processes 42 87-99
Zuberbuumlhler K Cheney D L amp Seyfarth R M (1999) Con-ceptual semantics in a nonhuman primate Journal of ComparativePsychology 113 33-42
(Manuscript received March 1 2000 revision accepted for publication August 10 2000)
472 BOVET AND VAUCLAIR
subject did not respond during this interval or if he or she did notpull the rope to the end an opaque screen was lowered for 5 secmasking the subjectrsquos view of the board The screen was then raisedagain and the same objects were presented The subjects rarelyfailed to respond within the first 5 sec If the subjectrsquos response wascorrect a food reward (small pieces of apple or grains of corn) wasprovided and another pair of objects was immediately presented(within 5 sec) In the event of an error a 10-second time-out was in-troduced followed by the presentation of a new object The orien-tation of the objects varied over trials Each baboon was trainedtested about 1 h a day 5 days a week which represents an averageof 185 trials per day
EXPERIMENT 1Perceptual Identity
MethodThe monkeysrsquo task was to pull one rope when the two presented
objects were the same and the other rope when the two objects weredifferent The rope on the right side was associated with same for 2of the monkeys and with different for the other 2 monkeys In thisphase when two different objects were presented they were alwaystwo objects belonging to different categories The subjects werefirst trained with two pairs of objects (four possible combinations)and then to assess transfer when they reached a performance cri-terion of 80 correct responses for 100 consecutive trials pairs in-volving new objects were presented
A total of 36 different objects (18 food and 18 nonfood) wereused in 684 possible combinations That is there were 648 (36 X18) trials with different pairs (two different combinations weremade with each pair of different objects according to the relativeposition of the 2 objects) and 36 trials with same pairs presented toeach subject Each new pair of objects was first presented in the 4possible combinations between the 2 objects of the pairs then eachobject was presented in combination with all the preceding objects
Each new combination of objects was tested until a subject com-pleted 8 out of 10 consecutive trials correctly To avoid responsebias trials with novel pairs of objects were interspersed among tri-als involving familiar pairs and trials with different objects wereinterspersed among trials involving pairs of same objects (mostlypairs of familiar objects since most of the new combinations werepairs of different objects) Thus from 1 to 3 familiar pairs of objectswere interspersed with trials involving each novel object Todemonstrate a samendashdifferent judgment we are primarily con-cerned with first-trial data for each novel pair of objects Thereforethe analysis focused on these data
ResultsWe separated the 36 stimuli into nine blocks of four ob-
jects Each point on the functions presented in Figure 3represents the percentage of correct responses for thefirst presentation of each of those four objects with allthe preceding objects
On average it took the baboons 12 objects whichmeans 84 tests (6 3 12 different trials 1 12 same trials)to reach a performance criterion of 80 correct re-sponses on the first trials with new pairs of objects Thiscriterion was achieved in an average of 9689 trials perindividual
The baboons performances continued to improve overtrials involving the next 24 objects (Figure 3) At the endof the experiment all 4 subjects were tested on the last 4novel objects presented in combination with all of theprevious stimuli This phase represented 140 novel ob-ject combinations (corresponding to the last point of thecurve on Figure 3) The accuracy levels were 96 [x 2(1)5 120 p lt 001] 91 [x 2(1) 5 92 p lt 001] 91
Figure 2 Test apparatus The background is indicated by dashed lines A wooden board B one-way mir-ror C rope sliding through a hole and terminated by a wooden ball D transparent Plexiglas E board forplacing the stimulus object F food dispenser
CONCEPTUAL IDENTITY IN MONKEYS 473
[x 2(1) 5 92 p lt 001] and 94 [x 2(1) 5 109 p lt 001]respectively
If we focus only on the trials involving new object pairsthe baboonsrsquo performances are still good On the last 12pairs of objects (48 combinations) the baboons achievedscores of 875 for Sylvestre [x 2(1) 5 27 p lt 001]917 for Balthazar [x2(1) 5 33 p lt 001] 896 forEsperance [x 2(1) 5 30 p lt 001] and 854 for Ida[x2(1) 5 24 p lt 001]These results provide evidencethat the baboons were able to class as same or differenttwo physically identical or different objects
EXPERIMENT 2Conceptual Identity
MethodThe subjects were the 2 males involved in the previous experiments
The 2 females gave birth during Experiment 1 and thus became lessmotivated during the test sessions They finished Experiment 1 muchlater than the 2 males and thus were not included in Experiment 2
The task in Experiment 2 was to pull one rope (the right side ropefor Sylvestre the left side rope for Balthazar) when the two pre-sented objects belonged to the same category (two foods or twononfoods) and to pull the other rope when the objects belonged todifferent categories (one food and one nonfood) The baboons werefirst trained with one block of four different objects (shown to themin the six possible combinations) and new blocks of four objectswere introduced after a performance criterion of 80 (out of 100consecutive trials) was reached
Each new block of four objects was first presented in the 6 pos-sible combinations involving the four objects and then each ofthose new objects was presented in combination with each of thepreceding objects A total of 36 different objects (18 food and 18nonfood) was used and all 630 pairwise combinations were pre-sented to each subject As in Experiment 1 eight correct responsesout of 10 consecutive trials were required for a given pair of objectsbefore a novel pair was presented and trials with novel pairs of ob-jects were interspersed among trials involving familiar pairs Heretoo we focused on the first-trial performance during successivetransfer tests for each novel pair of objects
ResultsBoth baboons started this experiment responding as
they did in the first experiment that is by pulling therope for different when two different objects were pre-sented even when those two objects belonged to thesame category Thus their performance was far belowchance at the beginning of the experiment On averageit took 14 objects (91 transfer tests) to reach the criterionof 80 correct on trials involving one or two new ob-jects (see Figure 4) This criterion was reached after anaverage of 14576 trials (17851 for Sylvestre 11300 forBalthazar)
At the end of the experiment the last four objects werepresented in combination with each of the previous ob-jects This allowed for 134 novel combinations Both ba-boons obtained a high score (see Figure 4) 91 forSylvestre [x 2(1) 5 229 p lt 001] and 81 for Balthazar[x 2(1) 5 174 p lt 001] The baboonsrsquo performances areeven better if we consider the trials involving novel ob-jects only over the last six blocks (see Table 1) Such per-centages provide convincing evidence that the monkeysmastered this conceptual identity task
GENERAL DISCUSSION
Categorization is a fundamental component of manycognitive processes and has therefore been the subjectof many animal experiments (see Herrnstein 1990Thompson 1995 Vauclair 1996 and Zayan amp Vauclair1998 for reviews) However very few studies have ex-amined functional categorization in animals Functionalcategorization involving food has been investigated byWatanabe (1997) in pigeons and by Savage-RumbaughRumbaugh Smith and Lawson (1980) in chimpanzeesOther investigators have studied the formation of func-tional categories in pigeons using a discrimination pro-cedure involving repeated reversals (Vaughan 1988)Similarly functional or equivalence class membership
correct responses
100
90
80
70
60
50
40
30
20
10
0
Sylvestre
Balthazar
Espeacuterance
Ida
Number of blocks of 4 objects1 2 3 4 5 6 7 8 9
Figure 3 Percentages of correct responses for the first presen-tation of a novel pair of objects for perceptual identity
Figure 4 Percentages of correct responses for the first presen-tation of a novel pair of objects for conceptual identity
474 BOVET AND VAUCLAIR
has been studied in a California sea lion using a matching-to-sample procedure (Schusterman amp Kastak 1998)
We decided to use this kind of categorization (inwhich the members of the class share a functional prop-erty but do not look alike) because as has been ex-plained by Herrnstein (1990) conceptual categorizationis more abstract than perceptual categorization (forwhich the stimuli share visual properties and are thusphysically similar) Our task involved functional ratherthan perceptual categorization because the only possibleproperty the baboons could have used to classify the ob-jects was their edibility It is of interest to note that thesubjectsrsquo classificatory errors could often be explainedby their food preferences which varied across the indi-viduals Thus uneaten food items (eg marshmallowsfor Sylvestre) were classified in the nonfood category(see Savage-Rumbaugh et al 1980 for similar findings)
In Experiment 2 the hypothesis that the baboonscould master the task by means of a simple discrimina-tion based on the total hedonic value of the display (ifmoderate hedonic value then respond different if highor no hedonic value then respond same) can be dis-missed In fact no errors occurred when one highly val-ued food object was presented with a nonhedonic objector when two low valued food objects or two small foodobjects were presented
The use of three-dimensional (3-D) stimuli (instead ofthe two-dimensional stimuli more typically used in thiskind of research) was chosen to facilitate functional cat-egorization in our previous study (Bovet amp Vauclair1998) It is likely that the use of 3-D stimuli made thefunction of the objects more salient (Bovet amp Vauclair2000) and therefore augmented the rapid generalizationfrom a few training objects to novel food and nonfoodobjects and facilitated the demonstration of conceptualidentity
Numerous training trials were required for the ba-boons to reach the 80 correct criterion in the two ex-periments reported here It is likely that the monkeys re-lied on rote learning of all the combinations in theseearly trials Indeed they required fewer trials to acquirethe perceptual identity task than to acquire the concep-tual identity task This is not surprising given the greaterconceptual complexity of the latter task
For Experiment 1 the identity rule is not the only pos-sible explanation for the baboonsrsquo successful perfor-mance The subjects could also have learned to discrim-inate the stimulus arrays on the basis of symmetry (astimuli configuration consisting of two identical objectsis bilaterally symmetrical whereas a configuration con-sisting of two different objects is not) However in Ex-
periment 2 the configuration could not be used to solvethe task because the two objects presented were alwaysdifferent and consequently the configuration was neverperceptually symmetrical
In Experiment 2 for each block of four objects therewere four different-categories trials and two same-categorytrials The ratio of same to different is thus not equalHowever the results show that the subjects were not re-sponding with a bias toward the different-categories re-sponse because both baboons made slightly more errorsfor the different-categories trials than for the same-categorytrials (Table 1)
The high level of performance attained by the 2 ba-boons at the end of Experiment 2 with totally novel ob-jects (ie objects novel in the task but left in the mon-keysrsquo enclosure before the experiment) demonstratesmastery of the samendashdifferent relation and the ability toconceptually judge as same or different the previouslylearned categories
Moreover contrary to Premackrsquos (1983) contentioncognitive competence comparable to relational matchingdoes not require previous training with explicit tokensand symbols Our subjectsrsquo previous training only in-volved categorizing objects that belonged to one of twocategories and using the samendashdifferent relation betweenobjects (within and between the two categories)
Our experiments suggest that when tested with bio-logically relevant stimuli (grouping objects in food andnonfood categories has obvious ecological significancefor these animals) monkeys are able to judge not onlythe sameness between physical objects but also the same-ness of functional concepts Such a high degree of ab-straction and conceptualization by monkeys has not pre-viously been reported in the literature
REFERENCES
Bovet D amp Vauclair J (1998) Functional categorization of objectsand of their pictures in baboons (Papio anubis) Learning amp Motiva-tion 29 309-322
Bovet D amp Vauclair J (2000) Picture recognition in animals andin humans A review Behavioral Brain Research 109 143-165
DrsquoAmato M R amp Van Sant P (1988) The person concept in mon-keys (Cebus apella) Journal of Experimental Psychology AnimalBehavior Processes 14 43-56
Grant-Webster K S Gunderson V M amp Burbacher T M (1990)Behavioural assessment of young non-human primates Perceptual-cognitive development Neurotoxicology amp Teratology 12 543-546
Herrnstein R J (1990) Levels of stimulus control A functional ap-proach Cognition 37 133-166
Premack D (1983) The codes of man and beasts Behavioral amp BrainSciences 6 125-167
Savage-Rumbaugh E S Rumbaugh D M Smith S T amp Law-
Table 1 Percentages of Correct Responses Given for the First Presentation of the Combinations Realized
With a New Block of Four Objects for the Last Six Blocks in Experiment 2
Subject Same Category (n 5 12) x2 Different Categories (n = 24) x2 Total (n 5 36) x2
Sylvestre 1000 12 917 16 944 28Balthazar 917 8dagger 792 8dagger 833 16
p lt 001 daggerp lt 01
CONCEPTUAL IDENTITY IN MONKEYS 475
son J (1980 November) Reference The linguistic essential Sci-ence 210 922-925
Schusterman R J amp Kastak D (1998) Functional equivalence ina California sea lion Relevance to animal social and communicativeinteractions Animal Behaviour 55 1087-1095
Thompson R K R (1995) Natural and relational concepts in animalsIn H Roitblat amp J A Meyer (Eds) Comparative approaches of cog-nitive sciences (pp 175-224) Cambridge MA MIT Press
Thompson R K R amp Oden D L (1996) A profound disparity re-visited Perception and judgment of abstract identity relations bychimpanzees human infants and monkeys Behavioural Processes35 149-161
Thompson R K R amp Oden D L (2000) Categorical perception ampconceptual judgments by nonhuman primates The paleological mon-key and the analogical ape Cognitive Science 24 363-396
Thompson R K R Oden D L amp Boysen S T (1997) Language-naive chimpanzees (Pan troglodytes) judge relations between rela-tions in a conceptual matching-to-sample task Journal of Experi-mental Psychology Animal Behavior Processes 23 31-43
Tomasello M amp Call J (1997) Primate cognition New York Ox-ford University Press
Tyrrell D J Stauffer L B amp Snowman L G (1991) Perception
of abstract identitydifference relationships by infants Infant Behav-ior amp Development 14 125-129
Tyrrell D J Zingaro M C amp Minard K L (1993) Learning andtransfer of identitydifference relationships by infants Infant Behav-ior amp Development 16 43-52
Vauclair J (1996) Animal cognition Recent developments in mod-ern comparative psychology Cambridge MA Harvard UniversityPress
Vaughan W Jr (1988) Formation of equivalence sets in pigeonsJournal of Experimental Psychology Animal Behavior Processes14 36-42
Watanabe S (1997) Visual discrimination of real objects and picturesin pigeons Animal Learning amp Behavior 25 185-192
Zayan R amp Vauclair J (1998) Categories as paradigms for com-parative cognition Behavioural Processes 42 87-99
Zuberbuumlhler K Cheney D L amp Seyfarth R M (1999) Con-ceptual semantics in a nonhuman primate Journal of ComparativePsychology 113 33-42
(Manuscript received March 1 2000 revision accepted for publication August 10 2000)
CONCEPTUAL IDENTITY IN MONKEYS 473
[x 2(1) 5 92 p lt 001] and 94 [x 2(1) 5 109 p lt 001]respectively
If we focus only on the trials involving new object pairsthe baboonsrsquo performances are still good On the last 12pairs of objects (48 combinations) the baboons achievedscores of 875 for Sylvestre [x 2(1) 5 27 p lt 001]917 for Balthazar [x2(1) 5 33 p lt 001] 896 forEsperance [x 2(1) 5 30 p lt 001] and 854 for Ida[x2(1) 5 24 p lt 001]These results provide evidencethat the baboons were able to class as same or differenttwo physically identical or different objects
EXPERIMENT 2Conceptual Identity
MethodThe subjects were the 2 males involved in the previous experiments
The 2 females gave birth during Experiment 1 and thus became lessmotivated during the test sessions They finished Experiment 1 muchlater than the 2 males and thus were not included in Experiment 2
The task in Experiment 2 was to pull one rope (the right side ropefor Sylvestre the left side rope for Balthazar) when the two pre-sented objects belonged to the same category (two foods or twononfoods) and to pull the other rope when the objects belonged todifferent categories (one food and one nonfood) The baboons werefirst trained with one block of four different objects (shown to themin the six possible combinations) and new blocks of four objectswere introduced after a performance criterion of 80 (out of 100consecutive trials) was reached
Each new block of four objects was first presented in the 6 pos-sible combinations involving the four objects and then each ofthose new objects was presented in combination with each of thepreceding objects A total of 36 different objects (18 food and 18nonfood) was used and all 630 pairwise combinations were pre-sented to each subject As in Experiment 1 eight correct responsesout of 10 consecutive trials were required for a given pair of objectsbefore a novel pair was presented and trials with novel pairs of ob-jects were interspersed among trials involving familiar pairs Heretoo we focused on the first-trial performance during successivetransfer tests for each novel pair of objects
ResultsBoth baboons started this experiment responding as
they did in the first experiment that is by pulling therope for different when two different objects were pre-sented even when those two objects belonged to thesame category Thus their performance was far belowchance at the beginning of the experiment On averageit took 14 objects (91 transfer tests) to reach the criterionof 80 correct on trials involving one or two new ob-jects (see Figure 4) This criterion was reached after anaverage of 14576 trials (17851 for Sylvestre 11300 forBalthazar)
At the end of the experiment the last four objects werepresented in combination with each of the previous ob-jects This allowed for 134 novel combinations Both ba-boons obtained a high score (see Figure 4) 91 forSylvestre [x 2(1) 5 229 p lt 001] and 81 for Balthazar[x 2(1) 5 174 p lt 001] The baboonsrsquo performances areeven better if we consider the trials involving novel ob-jects only over the last six blocks (see Table 1) Such per-centages provide convincing evidence that the monkeysmastered this conceptual identity task
GENERAL DISCUSSION
Categorization is a fundamental component of manycognitive processes and has therefore been the subjectof many animal experiments (see Herrnstein 1990Thompson 1995 Vauclair 1996 and Zayan amp Vauclair1998 for reviews) However very few studies have ex-amined functional categorization in animals Functionalcategorization involving food has been investigated byWatanabe (1997) in pigeons and by Savage-RumbaughRumbaugh Smith and Lawson (1980) in chimpanzeesOther investigators have studied the formation of func-tional categories in pigeons using a discrimination pro-cedure involving repeated reversals (Vaughan 1988)Similarly functional or equivalence class membership
correct responses
100
90
80
70
60
50
40
30
20
10
0
Sylvestre
Balthazar
Espeacuterance
Ida
Number of blocks of 4 objects1 2 3 4 5 6 7 8 9
Figure 3 Percentages of correct responses for the first presen-tation of a novel pair of objects for perceptual identity
Figure 4 Percentages of correct responses for the first presen-tation of a novel pair of objects for conceptual identity
474 BOVET AND VAUCLAIR
has been studied in a California sea lion using a matching-to-sample procedure (Schusterman amp Kastak 1998)
We decided to use this kind of categorization (inwhich the members of the class share a functional prop-erty but do not look alike) because as has been ex-plained by Herrnstein (1990) conceptual categorizationis more abstract than perceptual categorization (forwhich the stimuli share visual properties and are thusphysically similar) Our task involved functional ratherthan perceptual categorization because the only possibleproperty the baboons could have used to classify the ob-jects was their edibility It is of interest to note that thesubjectsrsquo classificatory errors could often be explainedby their food preferences which varied across the indi-viduals Thus uneaten food items (eg marshmallowsfor Sylvestre) were classified in the nonfood category(see Savage-Rumbaugh et al 1980 for similar findings)
In Experiment 2 the hypothesis that the baboonscould master the task by means of a simple discrimina-tion based on the total hedonic value of the display (ifmoderate hedonic value then respond different if highor no hedonic value then respond same) can be dis-missed In fact no errors occurred when one highly val-ued food object was presented with a nonhedonic objector when two low valued food objects or two small foodobjects were presented
The use of three-dimensional (3-D) stimuli (instead ofthe two-dimensional stimuli more typically used in thiskind of research) was chosen to facilitate functional cat-egorization in our previous study (Bovet amp Vauclair1998) It is likely that the use of 3-D stimuli made thefunction of the objects more salient (Bovet amp Vauclair2000) and therefore augmented the rapid generalizationfrom a few training objects to novel food and nonfoodobjects and facilitated the demonstration of conceptualidentity
Numerous training trials were required for the ba-boons to reach the 80 correct criterion in the two ex-periments reported here It is likely that the monkeys re-lied on rote learning of all the combinations in theseearly trials Indeed they required fewer trials to acquirethe perceptual identity task than to acquire the concep-tual identity task This is not surprising given the greaterconceptual complexity of the latter task
For Experiment 1 the identity rule is not the only pos-sible explanation for the baboonsrsquo successful perfor-mance The subjects could also have learned to discrim-inate the stimulus arrays on the basis of symmetry (astimuli configuration consisting of two identical objectsis bilaterally symmetrical whereas a configuration con-sisting of two different objects is not) However in Ex-
periment 2 the configuration could not be used to solvethe task because the two objects presented were alwaysdifferent and consequently the configuration was neverperceptually symmetrical
In Experiment 2 for each block of four objects therewere four different-categories trials and two same-categorytrials The ratio of same to different is thus not equalHowever the results show that the subjects were not re-sponding with a bias toward the different-categories re-sponse because both baboons made slightly more errorsfor the different-categories trials than for the same-categorytrials (Table 1)
The high level of performance attained by the 2 ba-boons at the end of Experiment 2 with totally novel ob-jects (ie objects novel in the task but left in the mon-keysrsquo enclosure before the experiment) demonstratesmastery of the samendashdifferent relation and the ability toconceptually judge as same or different the previouslylearned categories
Moreover contrary to Premackrsquos (1983) contentioncognitive competence comparable to relational matchingdoes not require previous training with explicit tokensand symbols Our subjectsrsquo previous training only in-volved categorizing objects that belonged to one of twocategories and using the samendashdifferent relation betweenobjects (within and between the two categories)
Our experiments suggest that when tested with bio-logically relevant stimuli (grouping objects in food andnonfood categories has obvious ecological significancefor these animals) monkeys are able to judge not onlythe sameness between physical objects but also the same-ness of functional concepts Such a high degree of ab-straction and conceptualization by monkeys has not pre-viously been reported in the literature
REFERENCES
Bovet D amp Vauclair J (1998) Functional categorization of objectsand of their pictures in baboons (Papio anubis) Learning amp Motiva-tion 29 309-322
Bovet D amp Vauclair J (2000) Picture recognition in animals andin humans A review Behavioral Brain Research 109 143-165
DrsquoAmato M R amp Van Sant P (1988) The person concept in mon-keys (Cebus apella) Journal of Experimental Psychology AnimalBehavior Processes 14 43-56
Grant-Webster K S Gunderson V M amp Burbacher T M (1990)Behavioural assessment of young non-human primates Perceptual-cognitive development Neurotoxicology amp Teratology 12 543-546
Herrnstein R J (1990) Levels of stimulus control A functional ap-proach Cognition 37 133-166
Premack D (1983) The codes of man and beasts Behavioral amp BrainSciences 6 125-167
Savage-Rumbaugh E S Rumbaugh D M Smith S T amp Law-
Table 1 Percentages of Correct Responses Given for the First Presentation of the Combinations Realized
With a New Block of Four Objects for the Last Six Blocks in Experiment 2
Subject Same Category (n 5 12) x2 Different Categories (n = 24) x2 Total (n 5 36) x2
Sylvestre 1000 12 917 16 944 28Balthazar 917 8dagger 792 8dagger 833 16
p lt 001 daggerp lt 01
CONCEPTUAL IDENTITY IN MONKEYS 475
son J (1980 November) Reference The linguistic essential Sci-ence 210 922-925
Schusterman R J amp Kastak D (1998) Functional equivalence ina California sea lion Relevance to animal social and communicativeinteractions Animal Behaviour 55 1087-1095
Thompson R K R (1995) Natural and relational concepts in animalsIn H Roitblat amp J A Meyer (Eds) Comparative approaches of cog-nitive sciences (pp 175-224) Cambridge MA MIT Press
Thompson R K R amp Oden D L (1996) A profound disparity re-visited Perception and judgment of abstract identity relations bychimpanzees human infants and monkeys Behavioural Processes35 149-161
Thompson R K R amp Oden D L (2000) Categorical perception ampconceptual judgments by nonhuman primates The paleological mon-key and the analogical ape Cognitive Science 24 363-396
Thompson R K R Oden D L amp Boysen S T (1997) Language-naive chimpanzees (Pan troglodytes) judge relations between rela-tions in a conceptual matching-to-sample task Journal of Experi-mental Psychology Animal Behavior Processes 23 31-43
Tomasello M amp Call J (1997) Primate cognition New York Ox-ford University Press
Tyrrell D J Stauffer L B amp Snowman L G (1991) Perception
of abstract identitydifference relationships by infants Infant Behav-ior amp Development 14 125-129
Tyrrell D J Zingaro M C amp Minard K L (1993) Learning andtransfer of identitydifference relationships by infants Infant Behav-ior amp Development 16 43-52
Vauclair J (1996) Animal cognition Recent developments in mod-ern comparative psychology Cambridge MA Harvard UniversityPress
Vaughan W Jr (1988) Formation of equivalence sets in pigeonsJournal of Experimental Psychology Animal Behavior Processes14 36-42
Watanabe S (1997) Visual discrimination of real objects and picturesin pigeons Animal Learning amp Behavior 25 185-192
Zayan R amp Vauclair J (1998) Categories as paradigms for com-parative cognition Behavioural Processes 42 87-99
Zuberbuumlhler K Cheney D L amp Seyfarth R M (1999) Con-ceptual semantics in a nonhuman primate Journal of ComparativePsychology 113 33-42
(Manuscript received March 1 2000 revision accepted for publication August 10 2000)
474 BOVET AND VAUCLAIR
has been studied in a California sea lion using a matching-to-sample procedure (Schusterman amp Kastak 1998)
We decided to use this kind of categorization (inwhich the members of the class share a functional prop-erty but do not look alike) because as has been ex-plained by Herrnstein (1990) conceptual categorizationis more abstract than perceptual categorization (forwhich the stimuli share visual properties and are thusphysically similar) Our task involved functional ratherthan perceptual categorization because the only possibleproperty the baboons could have used to classify the ob-jects was their edibility It is of interest to note that thesubjectsrsquo classificatory errors could often be explainedby their food preferences which varied across the indi-viduals Thus uneaten food items (eg marshmallowsfor Sylvestre) were classified in the nonfood category(see Savage-Rumbaugh et al 1980 for similar findings)
In Experiment 2 the hypothesis that the baboonscould master the task by means of a simple discrimina-tion based on the total hedonic value of the display (ifmoderate hedonic value then respond different if highor no hedonic value then respond same) can be dis-missed In fact no errors occurred when one highly val-ued food object was presented with a nonhedonic objector when two low valued food objects or two small foodobjects were presented
The use of three-dimensional (3-D) stimuli (instead ofthe two-dimensional stimuli more typically used in thiskind of research) was chosen to facilitate functional cat-egorization in our previous study (Bovet amp Vauclair1998) It is likely that the use of 3-D stimuli made thefunction of the objects more salient (Bovet amp Vauclair2000) and therefore augmented the rapid generalizationfrom a few training objects to novel food and nonfoodobjects and facilitated the demonstration of conceptualidentity
Numerous training trials were required for the ba-boons to reach the 80 correct criterion in the two ex-periments reported here It is likely that the monkeys re-lied on rote learning of all the combinations in theseearly trials Indeed they required fewer trials to acquirethe perceptual identity task than to acquire the concep-tual identity task This is not surprising given the greaterconceptual complexity of the latter task
For Experiment 1 the identity rule is not the only pos-sible explanation for the baboonsrsquo successful perfor-mance The subjects could also have learned to discrim-inate the stimulus arrays on the basis of symmetry (astimuli configuration consisting of two identical objectsis bilaterally symmetrical whereas a configuration con-sisting of two different objects is not) However in Ex-
periment 2 the configuration could not be used to solvethe task because the two objects presented were alwaysdifferent and consequently the configuration was neverperceptually symmetrical
In Experiment 2 for each block of four objects therewere four different-categories trials and two same-categorytrials The ratio of same to different is thus not equalHowever the results show that the subjects were not re-sponding with a bias toward the different-categories re-sponse because both baboons made slightly more errorsfor the different-categories trials than for the same-categorytrials (Table 1)
The high level of performance attained by the 2 ba-boons at the end of Experiment 2 with totally novel ob-jects (ie objects novel in the task but left in the mon-keysrsquo enclosure before the experiment) demonstratesmastery of the samendashdifferent relation and the ability toconceptually judge as same or different the previouslylearned categories
Moreover contrary to Premackrsquos (1983) contentioncognitive competence comparable to relational matchingdoes not require previous training with explicit tokensand symbols Our subjectsrsquo previous training only in-volved categorizing objects that belonged to one of twocategories and using the samendashdifferent relation betweenobjects (within and between the two categories)
Our experiments suggest that when tested with bio-logically relevant stimuli (grouping objects in food andnonfood categories has obvious ecological significancefor these animals) monkeys are able to judge not onlythe sameness between physical objects but also the same-ness of functional concepts Such a high degree of ab-straction and conceptualization by monkeys has not pre-viously been reported in the literature
REFERENCES
Bovet D amp Vauclair J (1998) Functional categorization of objectsand of their pictures in baboons (Papio anubis) Learning amp Motiva-tion 29 309-322
Bovet D amp Vauclair J (2000) Picture recognition in animals andin humans A review Behavioral Brain Research 109 143-165
DrsquoAmato M R amp Van Sant P (1988) The person concept in mon-keys (Cebus apella) Journal of Experimental Psychology AnimalBehavior Processes 14 43-56
Grant-Webster K S Gunderson V M amp Burbacher T M (1990)Behavioural assessment of young non-human primates Perceptual-cognitive development Neurotoxicology amp Teratology 12 543-546
Herrnstein R J (1990) Levels of stimulus control A functional ap-proach Cognition 37 133-166
Premack D (1983) The codes of man and beasts Behavioral amp BrainSciences 6 125-167
Savage-Rumbaugh E S Rumbaugh D M Smith S T amp Law-
Table 1 Percentages of Correct Responses Given for the First Presentation of the Combinations Realized
With a New Block of Four Objects for the Last Six Blocks in Experiment 2
Subject Same Category (n 5 12) x2 Different Categories (n = 24) x2 Total (n 5 36) x2
Sylvestre 1000 12 917 16 944 28Balthazar 917 8dagger 792 8dagger 833 16
p lt 001 daggerp lt 01
CONCEPTUAL IDENTITY IN MONKEYS 475
son J (1980 November) Reference The linguistic essential Sci-ence 210 922-925
Schusterman R J amp Kastak D (1998) Functional equivalence ina California sea lion Relevance to animal social and communicativeinteractions Animal Behaviour 55 1087-1095
Thompson R K R (1995) Natural and relational concepts in animalsIn H Roitblat amp J A Meyer (Eds) Comparative approaches of cog-nitive sciences (pp 175-224) Cambridge MA MIT Press
Thompson R K R amp Oden D L (1996) A profound disparity re-visited Perception and judgment of abstract identity relations bychimpanzees human infants and monkeys Behavioural Processes35 149-161
Thompson R K R amp Oden D L (2000) Categorical perception ampconceptual judgments by nonhuman primates The paleological mon-key and the analogical ape Cognitive Science 24 363-396
Thompson R K R Oden D L amp Boysen S T (1997) Language-naive chimpanzees (Pan troglodytes) judge relations between rela-tions in a conceptual matching-to-sample task Journal of Experi-mental Psychology Animal Behavior Processes 23 31-43
Tomasello M amp Call J (1997) Primate cognition New York Ox-ford University Press
Tyrrell D J Stauffer L B amp Snowman L G (1991) Perception
of abstract identitydifference relationships by infants Infant Behav-ior amp Development 14 125-129
Tyrrell D J Zingaro M C amp Minard K L (1993) Learning andtransfer of identitydifference relationships by infants Infant Behav-ior amp Development 16 43-52
Vauclair J (1996) Animal cognition Recent developments in mod-ern comparative psychology Cambridge MA Harvard UniversityPress
Vaughan W Jr (1988) Formation of equivalence sets in pigeonsJournal of Experimental Psychology Animal Behavior Processes14 36-42
Watanabe S (1997) Visual discrimination of real objects and picturesin pigeons Animal Learning amp Behavior 25 185-192
Zayan R amp Vauclair J (1998) Categories as paradigms for com-parative cognition Behavioural Processes 42 87-99
Zuberbuumlhler K Cheney D L amp Seyfarth R M (1999) Con-ceptual semantics in a nonhuman primate Journal of ComparativePsychology 113 33-42
(Manuscript received March 1 2000 revision accepted for publication August 10 2000)
CONCEPTUAL IDENTITY IN MONKEYS 475
son J (1980 November) Reference The linguistic essential Sci-ence 210 922-925
Schusterman R J amp Kastak D (1998) Functional equivalence ina California sea lion Relevance to animal social and communicativeinteractions Animal Behaviour 55 1087-1095
Thompson R K R (1995) Natural and relational concepts in animalsIn H Roitblat amp J A Meyer (Eds) Comparative approaches of cog-nitive sciences (pp 175-224) Cambridge MA MIT Press
Thompson R K R amp Oden D L (1996) A profound disparity re-visited Perception and judgment of abstract identity relations bychimpanzees human infants and monkeys Behavioural Processes35 149-161
Thompson R K R amp Oden D L (2000) Categorical perception ampconceptual judgments by nonhuman primates The paleological mon-key and the analogical ape Cognitive Science 24 363-396
Thompson R K R Oden D L amp Boysen S T (1997) Language-naive chimpanzees (Pan troglodytes) judge relations between rela-tions in a conceptual matching-to-sample task Journal of Experi-mental Psychology Animal Behavior Processes 23 31-43
Tomasello M amp Call J (1997) Primate cognition New York Ox-ford University Press
Tyrrell D J Stauffer L B amp Snowman L G (1991) Perception
of abstract identitydifference relationships by infants Infant Behav-ior amp Development 14 125-129
Tyrrell D J Zingaro M C amp Minard K L (1993) Learning andtransfer of identitydifference relationships by infants Infant Behav-ior amp Development 16 43-52
Vauclair J (1996) Animal cognition Recent developments in mod-ern comparative psychology Cambridge MA Harvard UniversityPress
Vaughan W Jr (1988) Formation of equivalence sets in pigeonsJournal of Experimental Psychology Animal Behavior Processes14 36-42
Watanabe S (1997) Visual discrimination of real objects and picturesin pigeons Animal Learning amp Behavior 25 185-192
Zayan R amp Vauclair J (1998) Categories as paradigms for com-parative cognition Behavioural Processes 42 87-99
Zuberbuumlhler K Cheney D L amp Seyfarth R M (1999) Con-ceptual semantics in a nonhuman primate Journal of ComparativePsychology 113 33-42
(Manuscript received March 1 2000 revision accepted for publication August 10 2000)
