OOZS-3932/79/0701AM13$02.00/0
Nouropsychologia. Vol. 17, PP. 413 to 420. 0 Pcrgamon PressLtd 1979.Printed in GreatBritain.
DO CHIMPANZEES RECOGNIZE PHOTOGRAPHS AS REPRESENTATIONS OF OBJECTS? ELLEN WINNER
Psychology Department, Boston College, Chestnut Hill Massachusetts, U.S.A. and Project Zero, Longfellow Hall, Harvard University, Cambridge, Massachusetts, U.S.A. and GEORGE ETTLINGER
Institute of Psychiatry, DeCrespigny Park, London SE5 SAF, England (Received 4 November
1978)
Abstract--Experiment 1: Two chimpanzees were given a matching-to-sample task in which objects were matched with objects and photographs. While subjects were able to match objects, performance decreased significantly when required to match objects to photographs. Experiment 2 : Two chimpanzees were given a cross-modal recognition task in which objects were presented in the dark, objects or their photographs in the light. Subjects succeeded with objects, but failed when cross-modal recognition from objects to photographs and vice-versa was required. These results cast doubt on reports that chimpanzees are able, on first sight, to recognize objects represented photographically.
INTRODUCTION THERE remains a controversy concerning the role of learning in the capacity to perceive photographs as equivalents of objects. On the one hand, a 2 yr-old child was able to recognize photographs without specific training [l]; and infants as young as 6 months can perceive a similarity between objects and their photographic representations [2]. On the other hand, adults in cultures without photographs are initially unable to recognize photographs as representations of objects [3]. However, as HOCHBERG and BROOKS [l] suggest, in the light of the child findings, it is quite possible that such deficiency requires an explanation besides lack of learning.
The reports of K~HLER [4], HAYES [5], and GARDNER and GARDNER [6] suggest that chimpanzees are able to perceive photographs accurately. But these reports are inconclusive since it is possible that the chimpanzees in these studies were taught to perceive photographs as representations by means of repeated association of an object and its representation (or a photograph and its sign, in the case of Gardner and Gardner). However, several researchers have reported that chimpanzees (and rhesus monkeys) are able to recognize an object represented in a photograph at first sight. ZIMMERMAN and HOCHBERG [7] reported that infant rhesus monkeys can transfer learned discriminations between simple geometric objects to both photographs and line drawings of the objects. Two studies of apes, moreover, have suggested that chimpanzees and oranghutangs are immediately capable of recognizing photographs of highly varied and complex objects [8, 91. In these two studies, a cross-modal match-to-sample test was administered in which subjects were 413
414
ELLENWINNERand GEORGEETTLINGER
required to match photographs to novel objects that could be touched but not seen. Results revealed that chimpanzees could immediately match objects to photographs when the photographs were colored, black and white, normal and high contrast, and both full and half size; subjects also succeeded at matching objects and their line drawing representations. However, Davenport et al. did not systematically create pairs of the same size objects: thus, the possibility that subjects succeeded by simply using a strategy such as “choose the bigger one” cannot be ruled out. The present work constitutes an attempt to replicate the findings of DAVENPORT et al. [S, 91 through the use of a unimodal match to sample task and a cross-modal recognition task. We attempted to answer the following questions: (1) Can the finding that chimpanzees perceive photographs as representations of objects without prior learning be replicated? And if so: (2) Is this ability affected by the familiarity of the objects represented? (3) Is this ability dependent upon the use of color photographs? (4) Is this ability dependent on the presence of size cues? That is, can chimpanzees succeed when the two objects (from which one must be chosen as identical to the sample) are exactly the same size (as judged by overall dimensions), so that a strategy such as “choose the bigger one” cannot be employed? EXPERIMENT
1
Method Subjects. The subjects were two juvenile female captive-born chimpanzees (Pan troglodytes verus) who had previously had extensive training in successive match-sample-to tasks with 3-dimensional objects (Ettlinger and Passingham, unpublished). Materials. Phase 1: For the first phase of the experiment, in which subjects were given a match-to-sample task using objects only, 80 test objects (Set A) were drawn from a pool of 440 “junk” objects varying in size, shape, color, and texture. Each object was mounted centrally on a cork (44 mm diameter, 15 mm high). The corks fitted into 3 holes in a board so that the tops of the corks were level with the surface of the board and the objects could be displaced by lifting. The objects were divided into 40 pairs; pairing was determined by a computer program that generated random pairs. Fourteen pairs consisted of objects of approximately the same size (i.e. overall dimensions); the remaining 26 pairs consisted of objects of different size. Phase 2: In order to examine the relative effects of object novelty and familiarity, for the second phase (in which subjects were given a match-to-sample task using objects and photographs), a new set of 40 pairs of objects was constructed (Set B). To construct Set B, 20 pairs from the 40 of Set A were replaced by 20 new, randomly generated pairs. The 20 discarded pairs, as well as the 20 new pairs, each consisted of 7 pairs of same size objects and 13 pairs of different size objects. Thus, the same proportion of different and same size objects was maintained as in Set A. The 20 pairs originally from Set A were familiar items; the 20 new pairs were novel items. Life size photographs were made of each member of Set B. For half of the pairs, color photographs were made; for the remaining half, black and white photographs were prepared. Novel and familiar items were equally divided between color and black and white photographs, as were same and different size pairs. Each photograph was mounted on a piece of cardboard, generally 77 cm wide by 10 cm high, but larger if the height of the object (on its cork) exceeded 10 cm. In general, the object in the photograph filled the space available and the photographs were cut so that no border remained. Procedure. Phase 1: In phase 1, using Set A, subjects were required to match objects in a successive match-to-sample task. The board with three holes for the corks was placed on the shelf of a WGTA and was positioned so that the central hole (which was not aligned with the other 2 holes) was closest to the subjects. The apparatus had a moveable screen between the subject and the materials. A trial was prepared by placing one object (the target object) of a particular pair in the central hole covering a peanut or raisin reward. The screen was raised and the subject was required to lift the object and retrieve the reward. The screen was then lowered and the target object and its paired member were placed in the holes on either side of the baord. The target object again covered a food reward. The screen was then raised and subjects were required to displace one of the objects. If the chosen object matched the
DO CHIMPANZEES
RECOGNIZE
PHOTOGRAPHS
AS REPRESENTATIONS
OF OBJECTS?
415
previous sample, subjects were allowed to retrieve the food. If the wrong object was displaced, correction was not permitted. The screen was then lowered and the next trial prepared. Subjects received all 40 pairs of Set A for 10 consecutive days. The pairs were randomly presented, as determined by Gellerman’s schedule, and a different random order was used each day. The choice of the target object for each pair was randomly determined for each day, but so that each object was a target on 5 days. Placement (right and left) of the correct matching object was randomly varied and evenly balanced over the 40 trials per day. Phase 2: In Phase 2, using Set B, subjects were required both to match objects to their life size photographic replicas, and to match photographs to photographs. The first day of Phase 2 immediately followed the last day of Phase 1; and Phase 2 continued for 16 consecutive days. Subjects received 40 trials per day. Phase 2 consisted of 4 conditions, each lasting for 4 days. For the first 4 days, subjects were required to match a target photograph to its object, presented successively. Subjects were presented with a photograph of one member of an object pair covering a food reward in the central hole. They were allowed to displace the photograph to gain the food. After the screen was lowered, the corresponding object and its paired member were placed in either side hole. Underneath the target object, which corresponded to the previously presented photograph, was a food reward which subjects were allowed to retrieve only upon selection of the object that matched the photograph. For the second 4 days, subjects were required to match a target photograph to its object, presented simultaneously. Subjects first displaced a photograph of one member of an object pair. The photograph covered a reward in the central hole. When the screen was lowered, the target object (corresponding to the photograph) and its paired member were placed in the side holes and the photograph was replaced over the central hole. The target object covered a food reward. For the third 4 days, subjects received a simultaneous object-to-photograph condition. Subjects first displaced one member of an object pair placed over the central hole. After lowering the screen, the object was replaced in the central hole and the target photograph (corresponding to that object) was placed in one side hole and covered a reward; the photograph of the other member of the pair was placed in the other side hole. Again, the animal was required to displace the photograph corresponding to the target object. For the final 4 days, subjects received a successive photograph-to-photograph condition. Subjects first saw the target photograph of a particular pair which covered a reward in the central hole. After the screen was lowered, the target photograph and its corresponding member were placed over the two side holes, the target photograph, as usual, covering a reward. The animal was required to displace the target photograph. As in Phase 1, the pairs in Phase 2 were randomly presented (as determined by Gellerman’s schedule), and a different order was presented each day. The choice of the target object or photograph for each pair was randomly determined for each day, as was placement to the left and right holes, In order to be certain that subjects attended to the photographs, the method of presentation of photographs was systematically varied. Photographs were either placed flat on the board or they were placed on a lectern (which covered the hole) so that they stood at a 45” angle to the board. #en on the lectern, half of the time the photographs were placed behind a clear plexiglass screen covering the lectern; the rest of the time they were placed in front of the screen, with or without spot-light illumination. These three different presentation methods were used in order to be certain that subjects did not fail to attend to the pictures for some irrelevant reason (such as their orientation with respect to the board). As no differences were found among these three methods, they were combined in further analyses. In order to ensure that the matching task itself was not forgotten, subjects received a match-to-sample test, with objects only, at the end of each of the 16 days of Phase 2. The objects used were the 20 pairs discarded from Set A and not used in Set B. Because subjects always performed better than 90% correct on this task, any failure in Phase 2 could not be attributed to a loss of memory for what was required in the match-tosample task.*
RESULTS Figure 1 shows the mean level of performance reached by each chimpanzee (Fanny and on each day of the object matching test of Phase 1. Both subjects achieved scores of 90 % or above by day 4 and scores never fell below this level on any following day. Clearly, thus, subjects quickly recovered the ability to match objects in a successive match-to-sample paradigm. Jane)
*For days 14 and 15, these objects were presented behind clear plexiglass (perspex) shields so that subjects were unable to use tactile recognition cues. While performance deteriorated somewhat on the first day with the shields, it returned to >90% on the next day. The shields were introduced so that failure in Phase 2 could not be attributed to the loss of tactile recognition cues when photographs, rather than objects, had to be recognized.
ELLENWINNERand GEORGE ETIUNGER
416
100 90
-
80
-
4-
70 t ?! 60;; " $
Fanny
5O 40-
OI I
2
3
4
5
6
7
6
Jane
9
IO
Days
FIG. 1.
Figure 2 shows the mean level of performance reached by subjects on each day of Phase 2. A binomial probability test revealed that both subjects performed at chance level on the first day, thus demonstrating that they did not recognize a photograph as a representation of an object on first sight, The experiment was continued for a total of 16 days in order to ascertain whether subjects could learn to recognize representations of objects. A binomial probability test revealed that for both subjects, performance remained at chance level for the first four days. For Fanny, performance rose moderately above chance level on days 5, 10, 11, 13, 14, and 15 (P < 0.05); for Jane, performance moderately above chance was attained on days 7, 8, 9, 11, 13, and 14 (P < 0.05). On all other days, including the last day of object-photograph matching and the last day of photograph-photograph matching, performance remained at chance. ----
Fanny
-
Jane
30 t
I$ , , , , , , , , , , , , , , , 123456789
IO II 12 13 14 I5 Days
FIG. 2.
16
DO CHIMPANZEES
RECOGNIZE
PHOTOGRAPHS
AS REPRESENTATIONS
OF OBJECTS?
417
Phase 1 vs Phase 2. A 3-way repeated measures analysis of variance [Chimp (2) by Phase (2) by Size (2)] was performed on scores for the first 8 days of each phase. (Only scores for the 20 objects represented in both Set A and B were used.) No difference was found between the 2 subjects. However, a highly significant main effect for Phase was revealed (F = 23.421, P -=c O$lOl). Subjects performed significantly worse when required to match objects to photographs (the first 8 days of Phase 2). Although no main effect for size was found, there was a significant interaction between Phase and Size (F = 6.519, P = 0.02). When required to match objects, subjects’ performance was facilitated when the objects of a pair differed in size; surprisingly, however, when required to match objects to photographs, performance was actually facilitated when objects of a pair were the same size. Phase 2. In order to examine the effects of the various factors in Phase 2, a 5-way repeated measures analysis of variance was performed on Chimpanzee (2) by Size (2) by Color (2-color vs black and white photographs) by Familiarity (2-familiar items from Set A vs Novel replacement items) by Condition (4) with repeated measures on Size, Color, and Familiarity and with Day as the unit factor. No main effects were found for any of the 5 factors, and no interaction between Chimpanzee and Condition was found. Thus, subjects consistently failed to match-to-sample on all of the conditions of Phase 2, which all required the perception of photographs. Failure occurred whether the objects were familiar or novel, similar or different in size, and whether the photographs were colored or black and white. In order to determine whether performance improved over days, a 4-way analysis of variance was performed on Day (16) nested within Condition (4) by Familiarity (2) by Color (2) by Size (2). No significant effect effect of Day was found. Thus, subjects clearly failed to learn to recognize the representations of objects in photographs despite 12 consecutive days of training; moreover, they failed to learn to match 2 identical photographs in 4 consecutive days of training.
EXPERIMENT
2
In order to attempt to more directly replicate the results obtained by DAVENPORT et al. [8, 93, a second experiment was carried out using a cross-modal procedure. In this experiment, subjects were required to transfer a discrimination between pairs of objects that were felt but not seen to their photographic representations or vice versa. Although not identical to that used by Davenport and his associates, this procedure was a closer approximation of Davenport’s method than that used in Experiment 1. Moreover, this particular crossmodal paradigm was selected because the subjects had already had extensive training in this task using objects [lo]. Finally, unlike in Experiment 1, the subjects were the same age as those used by Davenport; thus, failure could not be attributed to age. METHOD Subjects The subjects were 2 male wild-born chimpanzees approximately 5.5 yr.
(Pan troglodytes verus) weighing 25-28kg. and aged
Materials The test objects were the same object pairs and photographs as those used in Experiment 1. In addition, for object-object problems, pairs were drawn randomly from a collection of 90 further junk objects. Procedure Each subject received 100 problems in which a multi-trial two-choice discrimination
in one modality
418
ELLENWINNERand
GEORGE ETTLINGER
was followed by a single recognition trial in the opposite modality [lo]. Half of the problems tested crossmodal recognition (from vision to touch and vice versa) using 3-dimensional objects. The remaining half tested cross-modal recognition from objects that could be felt (but not seen) to their life size photographic representations and vice versa. Half of the photographs were colored; half were black-and white. Problems with objects only were randomly interspersed with problems with objects and photographs. For a particular problem, one pair of stimuli was presented six times in one modality and subjects were repeatedly rewarded for displacing one member of this pair. Placement (left or right) of the rewarded stimulus was randomly determined using Gellerman’s schedule. These six trials were immediately followed by a single recognition trial in the opposite modality. Performance in the dark was monitored with a video system sensitive to infra-red light.
RESULTS Figure 3 shows the per cent correct for each subject for trials with objects only and for trials in which photographs were introduced. A binomial probability test revealed that subjects responded significantly above chance (P < 0.05) with objects but only at chance level when required to transfer a learned response from a felt object to a photograph or from a photograph to a felt object. No differences between color and black and white photographs were found. 3bjec:s
Photographs
Algy
74%
4 B%
Blue
64%
42%
El
FIG. 3.
DISCUSSION The studies reported here failed to replicate the findings of DAVENPORT et al. [8, 91. In Experiment 1, chimpanzees that readily matched to sample with objects proved unable to match an object to its photographic replica on first sight. Moreover, these subjects failed over the course of a number of days to learn to do so. In Experiment 2, successful performance on a cross-modal recognition task using 3-dimensional stimuli fell to chance level when photographs were introduced. This performance persisted over 100 trials. Several possible explanations suggest themselves. Perhaps the erratic performance in Phase 2 of Experiment 1 was due to the presence of an identical cork at the base of each picture. If subjects looked only at the bottom of the pictures (which was where they tended to touch the pictures), and thus only at the corks, little or no difference between photographs would be perceived. However, it is unlikely that these chimpanzees could not have learned to stop attending to such an irrelevant cue, given their success at matching objects (where the tops of the corks were visible). Their clear distress at their repeated failures indicated a strong motivation to succeed. Moreover, the photographs used by Davenport et al. each contained an almost identical chain from which the object was suspended, yet the chains did not impair photographic recognition. Finally, it is noteworthy that Fanny and Jane also performed at chance on 29 of 30 trials when required to sort pictures of animals (obviously without corks), but performed above chance on 15 of 16 trials with objects differing in shape, size, and color [ 111.
DO CHIMPANZEES RECOGNIZE PHOTOGRAPHS ASREPRESENTATIONS OF OBJECTS?
419
Given that subjects failed to consistently match photographs to photographs in Experiment 1 (Condition 4), failure in the object-photograph matching task may have been due to the quality of the photographs. However, several judges reported the photographs to be extremely clear. Moreover, although some were clearer than others, subjects performed equally poorly on all stimuli. This suggests that the results are not accounted for by a perceptual failure due to the particular materials but may be explained by a perceptual failure related to photographs in general.* We therefore conclude that our subjects did not realize that a photograph is a visual stimulus that must be “read”. For our chimpanzees, the photographs were meaningless 2-dimensional objects. While the findings of these studies cannot be considered definitive, they are in conflict with previously reported findings that chimpanzees are immediately able to perceive photographs as representations of objects. Further work is needed to clarify the conditions under which photographic recognition does or does not occur in the chimpanzee. Acknowledgements-We thank the Livingston Fund and the Spencer Foundation who supported ELLEN WINNER’Svisit to London; the Grant Foundation of New York who funded the chimpanzee facility at the Institute of Psychiatry; and Dr. D. MALONEfor critical comments on a draft of this paper.
REFERENCES 1. HCMXBERG,J. and BROOKS,V. Pictorial recognition as an unlearned ability: A study of one child’s performance. Am. J. of Psycho/., 75, 625-628, 1962. stimuli. 2. ROSE, S. A. Infants’ transfer of response between two-dimensional and three-dimensional Child Development 48, 1086-1091, 1977. M. J. nie influence of culture on visualperception. Indian3. SEGAL,M., CAMPBELL,D. T. and HERSKOVITS, apolis: Bobbs-Merrill, 1966. 4. K&LER, W. The Mentality of Apes. Translated from the Second Revised Edition by Ella Winter. Harcourt, Brace & Co., London, 1925. 5. HAYES,C. The Ape in our House. Harper, New York, 1951. 6. GARDNER,R. A. and GARDNER,B. T. Teachingsign languageto a chimpanzee. Science 165, 664-672, 1969. of a 7. ZIMMERMAN,R. and HOCHBERG,J. Responses of infant monkeys to pictorial representations learned visual discrimination. Psychonom. Sci. 18, 307-308, 1970. 8. DAVENPORT,R. K. and ROGERS,C. M. Perception of photographs by apes. Behavior 39,318-320,197l. 9. DAVENPORT,R. K., ROGERS,C. M. and RUSSELL,I. S. Cross-modal perception in apes: Altered visual cues and delay. Neuropsychologia 13, 117-120, 1975. 10. JARVIS,M. J. and E~LINGER, G. Cross-modal recognition in chimpanzees and monkeys. Neuropsychologia 15, 4YY-506, 1Y77. Il. GARCHA,H. S. and ETTLINGER,G. Object sorting by chimpanzees and monkeys. Cortex. To be published.
*It has also been suggested to the authors that perhaps the objects themselves were more similar to each other than those used by Davenport et al., in which case the photographs would be less discriminable even with high quality. However, the objects used in the present study differed from each other in color, shape, texture, and, in over half of the pairs, in size. Thus, this objection cannot be maintained.
ELLEN WINNER and GEORGEETTLINGER
420
Experience pariement
jets et avec capable3
1. : On a soumis
2 l'exemple.
leur representation
d'apparier
gnificative
Les objets
dans
les objets;
2 2 chimpanz6s
devaient
Gtre
photoqraphi$ue. il y avait
leur performance
quand
une dpreuve
apparies
avec
Si les sujets
par contre
ils devaient
d'ap-
les ob6taient
une baisse apparier
si-
les ob-
jets aux photoqraphies.
soumise
Experience _2 3 2 chimpanzes,
curits,
les objets
r6ussissaient
avec
de reconnaissance
: Une t&he les objets
intermodale
Btaient
dans
ou leurs photoqraphies les objets
mais
intermodale
Ces r6sultats conclusions
selon
de reconnaitre
lesquelles
pr6sent&s
2 la lumiere.
Bchouaient
quand
(objets-photoqraphies
jettent
mier
regard
de reconnaissance
quelques
doutes
les chimpanzds
des objets
sent
repr6sentes
a et6
l'obs-
Les sujets
il s'aqissait et vice-versa).
sur de prec6dentes capables
dPs le pre-
photographiquement.
Deutschsprachige Zusammenfassung: Experiment I: Zwei Schimpansen wurden einer Zuordnungsaufgabe unterworfen, bei welcher Gegenstande (Objekte) mit Gegenstanden und Photographien tibereinstimmten.Wahrend die Versuchstiere Objekte zuzuordnen vermochten, nahm die Leistung signifikant ab, wenn die Zuordnung von Objekten zu Photographien verlangt wurde. zxperiment II: Zwei Schimpansen wurde eine Aufgabe mit gekreuzt modalem Erkennen dargeboten, in welcher Objekte im Dunkeln gezeigt wurden, sowie Objekte oder ihre Photographien bei Licht. Die Versuchstiere hatten Erfolg mit Objekten, scheiterten jedoch, wenn gekreuzt-modales Erkennen von Objekten auf Photographien und vice-versa verlangt wurde. Diese Ergebnisse werfen Zweifel auf Berichte, daR Ychimpansen auf den ersten Blick in der Lage sind, photographiesh dargestellte Objekte zu erkennen.
Nouropsychologia. Vol. 17, PP. 413 to 420. 0 Pcrgamon PressLtd 1979.Printed in GreatBritain.
DO CHIMPANZEES RECOGNIZE PHOTOGRAPHS AS REPRESENTATIONS OF OBJECTS? ELLEN WINNER
Psychology Department, Boston College, Chestnut Hill Massachusetts, U.S.A. and Project Zero, Longfellow Hall, Harvard University, Cambridge, Massachusetts, U.S.A. and GEORGE ETTLINGER
Institute of Psychiatry, DeCrespigny Park, London SE5 SAF, England (Received 4 November
1978)
Abstract--Experiment 1: Two chimpanzees were given a matching-to-sample task in which objects were matched with objects and photographs. While subjects were able to match objects, performance decreased significantly when required to match objects to photographs. Experiment 2 : Two chimpanzees were given a cross-modal recognition task in which objects were presented in the dark, objects or their photographs in the light. Subjects succeeded with objects, but failed when cross-modal recognition from objects to photographs and vice-versa was required. These results cast doubt on reports that chimpanzees are able, on first sight, to recognize objects represented photographically.
INTRODUCTION THERE remains a controversy concerning the role of learning in the capacity to perceive photographs as equivalents of objects. On the one hand, a 2 yr-old child was able to recognize photographs without specific training [l]; and infants as young as 6 months can perceive a similarity between objects and their photographic representations [2]. On the other hand, adults in cultures without photographs are initially unable to recognize photographs as representations of objects [3]. However, as HOCHBERG and BROOKS [l] suggest, in the light of the child findings, it is quite possible that such deficiency requires an explanation besides lack of learning.
The reports of K~HLER [4], HAYES [5], and GARDNER and GARDNER [6] suggest that chimpanzees are able to perceive photographs accurately. But these reports are inconclusive since it is possible that the chimpanzees in these studies were taught to perceive photographs as representations by means of repeated association of an object and its representation (or a photograph and its sign, in the case of Gardner and Gardner). However, several researchers have reported that chimpanzees (and rhesus monkeys) are able to recognize an object represented in a photograph at first sight. ZIMMERMAN and HOCHBERG [7] reported that infant rhesus monkeys can transfer learned discriminations between simple geometric objects to both photographs and line drawings of the objects. Two studies of apes, moreover, have suggested that chimpanzees and oranghutangs are immediately capable of recognizing photographs of highly varied and complex objects [8, 91. In these two studies, a cross-modal match-to-sample test was administered in which subjects were 413
414
ELLENWINNERand GEORGEETTLINGER
required to match photographs to novel objects that could be touched but not seen. Results revealed that chimpanzees could immediately match objects to photographs when the photographs were colored, black and white, normal and high contrast, and both full and half size; subjects also succeeded at matching objects and their line drawing representations. However, Davenport et al. did not systematically create pairs of the same size objects: thus, the possibility that subjects succeeded by simply using a strategy such as “choose the bigger one” cannot be ruled out. The present work constitutes an attempt to replicate the findings of DAVENPORT et al. [S, 91 through the use of a unimodal match to sample task and a cross-modal recognition task. We attempted to answer the following questions: (1) Can the finding that chimpanzees perceive photographs as representations of objects without prior learning be replicated? And if so: (2) Is this ability affected by the familiarity of the objects represented? (3) Is this ability dependent upon the use of color photographs? (4) Is this ability dependent on the presence of size cues? That is, can chimpanzees succeed when the two objects (from which one must be chosen as identical to the sample) are exactly the same size (as judged by overall dimensions), so that a strategy such as “choose the bigger one” cannot be employed? EXPERIMENT
1
Method Subjects. The subjects were two juvenile female captive-born chimpanzees (Pan troglodytes verus) who had previously had extensive training in successive match-sample-to tasks with 3-dimensional objects (Ettlinger and Passingham, unpublished). Materials. Phase 1: For the first phase of the experiment, in which subjects were given a match-to-sample task using objects only, 80 test objects (Set A) were drawn from a pool of 440 “junk” objects varying in size, shape, color, and texture. Each object was mounted centrally on a cork (44 mm diameter, 15 mm high). The corks fitted into 3 holes in a board so that the tops of the corks were level with the surface of the board and the objects could be displaced by lifting. The objects were divided into 40 pairs; pairing was determined by a computer program that generated random pairs. Fourteen pairs consisted of objects of approximately the same size (i.e. overall dimensions); the remaining 26 pairs consisted of objects of different size. Phase 2: In order to examine the relative effects of object novelty and familiarity, for the second phase (in which subjects were given a match-to-sample task using objects and photographs), a new set of 40 pairs of objects was constructed (Set B). To construct Set B, 20 pairs from the 40 of Set A were replaced by 20 new, randomly generated pairs. The 20 discarded pairs, as well as the 20 new pairs, each consisted of 7 pairs of same size objects and 13 pairs of different size objects. Thus, the same proportion of different and same size objects was maintained as in Set A. The 20 pairs originally from Set A were familiar items; the 20 new pairs were novel items. Life size photographs were made of each member of Set B. For half of the pairs, color photographs were made; for the remaining half, black and white photographs were prepared. Novel and familiar items were equally divided between color and black and white photographs, as were same and different size pairs. Each photograph was mounted on a piece of cardboard, generally 77 cm wide by 10 cm high, but larger if the height of the object (on its cork) exceeded 10 cm. In general, the object in the photograph filled the space available and the photographs were cut so that no border remained. Procedure. Phase 1: In phase 1, using Set A, subjects were required to match objects in a successive match-to-sample task. The board with three holes for the corks was placed on the shelf of a WGTA and was positioned so that the central hole (which was not aligned with the other 2 holes) was closest to the subjects. The apparatus had a moveable screen between the subject and the materials. A trial was prepared by placing one object (the target object) of a particular pair in the central hole covering a peanut or raisin reward. The screen was raised and the subject was required to lift the object and retrieve the reward. The screen was then lowered and the target object and its paired member were placed in the holes on either side of the baord. The target object again covered a food reward. The screen was then raised and subjects were required to displace one of the objects. If the chosen object matched the
DO CHIMPANZEES
RECOGNIZE
PHOTOGRAPHS
AS REPRESENTATIONS
OF OBJECTS?
415
previous sample, subjects were allowed to retrieve the food. If the wrong object was displaced, correction was not permitted. The screen was then lowered and the next trial prepared. Subjects received all 40 pairs of Set A for 10 consecutive days. The pairs were randomly presented, as determined by Gellerman’s schedule, and a different random order was used each day. The choice of the target object for each pair was randomly determined for each day, but so that each object was a target on 5 days. Placement (right and left) of the correct matching object was randomly varied and evenly balanced over the 40 trials per day. Phase 2: In Phase 2, using Set B, subjects were required both to match objects to their life size photographic replicas, and to match photographs to photographs. The first day of Phase 2 immediately followed the last day of Phase 1; and Phase 2 continued for 16 consecutive days. Subjects received 40 trials per day. Phase 2 consisted of 4 conditions, each lasting for 4 days. For the first 4 days, subjects were required to match a target photograph to its object, presented successively. Subjects were presented with a photograph of one member of an object pair covering a food reward in the central hole. They were allowed to displace the photograph to gain the food. After the screen was lowered, the corresponding object and its paired member were placed in either side hole. Underneath the target object, which corresponded to the previously presented photograph, was a food reward which subjects were allowed to retrieve only upon selection of the object that matched the photograph. For the second 4 days, subjects were required to match a target photograph to its object, presented simultaneously. Subjects first displaced a photograph of one member of an object pair. The photograph covered a reward in the central hole. When the screen was lowered, the target object (corresponding to the photograph) and its paired member were placed in the side holes and the photograph was replaced over the central hole. The target object covered a food reward. For the third 4 days, subjects received a simultaneous object-to-photograph condition. Subjects first displaced one member of an object pair placed over the central hole. After lowering the screen, the object was replaced in the central hole and the target photograph (corresponding to that object) was placed in one side hole and covered a reward; the photograph of the other member of the pair was placed in the other side hole. Again, the animal was required to displace the photograph corresponding to the target object. For the final 4 days, subjects received a successive photograph-to-photograph condition. Subjects first saw the target photograph of a particular pair which covered a reward in the central hole. After the screen was lowered, the target photograph and its corresponding member were placed over the two side holes, the target photograph, as usual, covering a reward. The animal was required to displace the target photograph. As in Phase 1, the pairs in Phase 2 were randomly presented (as determined by Gellerman’s schedule), and a different order was presented each day. The choice of the target object or photograph for each pair was randomly determined for each day, as was placement to the left and right holes, In order to be certain that subjects attended to the photographs, the method of presentation of photographs was systematically varied. Photographs were either placed flat on the board or they were placed on a lectern (which covered the hole) so that they stood at a 45” angle to the board. #en on the lectern, half of the time the photographs were placed behind a clear plexiglass screen covering the lectern; the rest of the time they were placed in front of the screen, with or without spot-light illumination. These three different presentation methods were used in order to be certain that subjects did not fail to attend to the pictures for some irrelevant reason (such as their orientation with respect to the board). As no differences were found among these three methods, they were combined in further analyses. In order to ensure that the matching task itself was not forgotten, subjects received a match-to-sample test, with objects only, at the end of each of the 16 days of Phase 2. The objects used were the 20 pairs discarded from Set A and not used in Set B. Because subjects always performed better than 90% correct on this task, any failure in Phase 2 could not be attributed to a loss of memory for what was required in the match-tosample task.*
RESULTS Figure 1 shows the mean level of performance reached by each chimpanzee (Fanny and on each day of the object matching test of Phase 1. Both subjects achieved scores of 90 % or above by day 4 and scores never fell below this level on any following day. Clearly, thus, subjects quickly recovered the ability to match objects in a successive match-to-sample paradigm. Jane)
*For days 14 and 15, these objects were presented behind clear plexiglass (perspex) shields so that subjects were unable to use tactile recognition cues. While performance deteriorated somewhat on the first day with the shields, it returned to >90% on the next day. The shields were introduced so that failure in Phase 2 could not be attributed to the loss of tactile recognition cues when photographs, rather than objects, had to be recognized.
ELLENWINNERand GEORGE ETIUNGER
416
100 90
-
80
-
4-
70 t ?! 60;; " $
Fanny
5O 40-
OI I
2
3
4
5
6
7
6
Jane
9
IO
Days
FIG. 1.
Figure 2 shows the mean level of performance reached by subjects on each day of Phase 2. A binomial probability test revealed that both subjects performed at chance level on the first day, thus demonstrating that they did not recognize a photograph as a representation of an object on first sight, The experiment was continued for a total of 16 days in order to ascertain whether subjects could learn to recognize representations of objects. A binomial probability test revealed that for both subjects, performance remained at chance level for the first four days. For Fanny, performance rose moderately above chance level on days 5, 10, 11, 13, 14, and 15 (P < 0.05); for Jane, performance moderately above chance was attained on days 7, 8, 9, 11, 13, and 14 (P < 0.05). On all other days, including the last day of object-photograph matching and the last day of photograph-photograph matching, performance remained at chance. ----
Fanny
-
Jane
30 t
I$ , , , , , , , , , , , , , , , 123456789
IO II 12 13 14 I5 Days
FIG. 2.
16
DO CHIMPANZEES
RECOGNIZE
PHOTOGRAPHS
AS REPRESENTATIONS
OF OBJECTS?
417
Phase 1 vs Phase 2. A 3-way repeated measures analysis of variance [Chimp (2) by Phase (2) by Size (2)] was performed on scores for the first 8 days of each phase. (Only scores for the 20 objects represented in both Set A and B were used.) No difference was found between the 2 subjects. However, a highly significant main effect for Phase was revealed (F = 23.421, P -=c O$lOl). Subjects performed significantly worse when required to match objects to photographs (the first 8 days of Phase 2). Although no main effect for size was found, there was a significant interaction between Phase and Size (F = 6.519, P = 0.02). When required to match objects, subjects’ performance was facilitated when the objects of a pair differed in size; surprisingly, however, when required to match objects to photographs, performance was actually facilitated when objects of a pair were the same size. Phase 2. In order to examine the effects of the various factors in Phase 2, a 5-way repeated measures analysis of variance was performed on Chimpanzee (2) by Size (2) by Color (2-color vs black and white photographs) by Familiarity (2-familiar items from Set A vs Novel replacement items) by Condition (4) with repeated measures on Size, Color, and Familiarity and with Day as the unit factor. No main effects were found for any of the 5 factors, and no interaction between Chimpanzee and Condition was found. Thus, subjects consistently failed to match-to-sample on all of the conditions of Phase 2, which all required the perception of photographs. Failure occurred whether the objects were familiar or novel, similar or different in size, and whether the photographs were colored or black and white. In order to determine whether performance improved over days, a 4-way analysis of variance was performed on Day (16) nested within Condition (4) by Familiarity (2) by Color (2) by Size (2). No significant effect effect of Day was found. Thus, subjects clearly failed to learn to recognize the representations of objects in photographs despite 12 consecutive days of training; moreover, they failed to learn to match 2 identical photographs in 4 consecutive days of training.
EXPERIMENT
2
In order to attempt to more directly replicate the results obtained by DAVENPORT et al. [8, 93, a second experiment was carried out using a cross-modal procedure. In this experiment, subjects were required to transfer a discrimination between pairs of objects that were felt but not seen to their photographic representations or vice versa. Although not identical to that used by Davenport and his associates, this procedure was a closer approximation of Davenport’s method than that used in Experiment 1. Moreover, this particular crossmodal paradigm was selected because the subjects had already had extensive training in this task using objects [lo]. Finally, unlike in Experiment 1, the subjects were the same age as those used by Davenport; thus, failure could not be attributed to age. METHOD Subjects The subjects were 2 male wild-born chimpanzees approximately 5.5 yr.
(Pan troglodytes verus) weighing 25-28kg. and aged
Materials The test objects were the same object pairs and photographs as those used in Experiment 1. In addition, for object-object problems, pairs were drawn randomly from a collection of 90 further junk objects. Procedure Each subject received 100 problems in which a multi-trial two-choice discrimination
in one modality
418
ELLENWINNERand
GEORGE ETTLINGER
was followed by a single recognition trial in the opposite modality [lo]. Half of the problems tested crossmodal recognition (from vision to touch and vice versa) using 3-dimensional objects. The remaining half tested cross-modal recognition from objects that could be felt (but not seen) to their life size photographic representations and vice versa. Half of the photographs were colored; half were black-and white. Problems with objects only were randomly interspersed with problems with objects and photographs. For a particular problem, one pair of stimuli was presented six times in one modality and subjects were repeatedly rewarded for displacing one member of this pair. Placement (left or right) of the rewarded stimulus was randomly determined using Gellerman’s schedule. These six trials were immediately followed by a single recognition trial in the opposite modality. Performance in the dark was monitored with a video system sensitive to infra-red light.
RESULTS Figure 3 shows the per cent correct for each subject for trials with objects only and for trials in which photographs were introduced. A binomial probability test revealed that subjects responded significantly above chance (P < 0.05) with objects but only at chance level when required to transfer a learned response from a felt object to a photograph or from a photograph to a felt object. No differences between color and black and white photographs were found. 3bjec:s
Photographs
Algy
74%
4 B%
Blue
64%
42%
El
FIG. 3.
DISCUSSION The studies reported here failed to replicate the findings of DAVENPORT et al. [8, 91. In Experiment 1, chimpanzees that readily matched to sample with objects proved unable to match an object to its photographic replica on first sight. Moreover, these subjects failed over the course of a number of days to learn to do so. In Experiment 2, successful performance on a cross-modal recognition task using 3-dimensional stimuli fell to chance level when photographs were introduced. This performance persisted over 100 trials. Several possible explanations suggest themselves. Perhaps the erratic performance in Phase 2 of Experiment 1 was due to the presence of an identical cork at the base of each picture. If subjects looked only at the bottom of the pictures (which was where they tended to touch the pictures), and thus only at the corks, little or no difference between photographs would be perceived. However, it is unlikely that these chimpanzees could not have learned to stop attending to such an irrelevant cue, given their success at matching objects (where the tops of the corks were visible). Their clear distress at their repeated failures indicated a strong motivation to succeed. Moreover, the photographs used by Davenport et al. each contained an almost identical chain from which the object was suspended, yet the chains did not impair photographic recognition. Finally, it is noteworthy that Fanny and Jane also performed at chance on 29 of 30 trials when required to sort pictures of animals (obviously without corks), but performed above chance on 15 of 16 trials with objects differing in shape, size, and color [ 111.
DO CHIMPANZEES RECOGNIZE PHOTOGRAPHS ASREPRESENTATIONS OF OBJECTS?
419
Given that subjects failed to consistently match photographs to photographs in Experiment 1 (Condition 4), failure in the object-photograph matching task may have been due to the quality of the photographs. However, several judges reported the photographs to be extremely clear. Moreover, although some were clearer than others, subjects performed equally poorly on all stimuli. This suggests that the results are not accounted for by a perceptual failure due to the particular materials but may be explained by a perceptual failure related to photographs in general.* We therefore conclude that our subjects did not realize that a photograph is a visual stimulus that must be “read”. For our chimpanzees, the photographs were meaningless 2-dimensional objects. While the findings of these studies cannot be considered definitive, they are in conflict with previously reported findings that chimpanzees are immediately able to perceive photographs as representations of objects. Further work is needed to clarify the conditions under which photographic recognition does or does not occur in the chimpanzee. Acknowledgements-We thank the Livingston Fund and the Spencer Foundation who supported ELLEN WINNER’Svisit to London; the Grant Foundation of New York who funded the chimpanzee facility at the Institute of Psychiatry; and Dr. D. MALONEfor critical comments on a draft of this paper.
REFERENCES 1. HCMXBERG,J. and BROOKS,V. Pictorial recognition as an unlearned ability: A study of one child’s performance. Am. J. of Psycho/., 75, 625-628, 1962. stimuli. 2. ROSE, S. A. Infants’ transfer of response between two-dimensional and three-dimensional Child Development 48, 1086-1091, 1977. M. J. nie influence of culture on visualperception. Indian3. SEGAL,M., CAMPBELL,D. T. and HERSKOVITS, apolis: Bobbs-Merrill, 1966. 4. K&LER, W. The Mentality of Apes. Translated from the Second Revised Edition by Ella Winter. Harcourt, Brace & Co., London, 1925. 5. HAYES,C. The Ape in our House. Harper, New York, 1951. 6. GARDNER,R. A. and GARDNER,B. T. Teachingsign languageto a chimpanzee. Science 165, 664-672, 1969. of a 7. ZIMMERMAN,R. and HOCHBERG,J. Responses of infant monkeys to pictorial representations learned visual discrimination. Psychonom. Sci. 18, 307-308, 1970. 8. DAVENPORT,R. K. and ROGERS,C. M. Perception of photographs by apes. Behavior 39,318-320,197l. 9. DAVENPORT,R. K., ROGERS,C. M. and RUSSELL,I. S. Cross-modal perception in apes: Altered visual cues and delay. Neuropsychologia 13, 117-120, 1975. 10. JARVIS,M. J. and E~LINGER, G. Cross-modal recognition in chimpanzees and monkeys. Neuropsychologia 15, 4YY-506, 1Y77. Il. GARCHA,H. S. and ETTLINGER,G. Object sorting by chimpanzees and monkeys. Cortex. To be published.
*It has also been suggested to the authors that perhaps the objects themselves were more similar to each other than those used by Davenport et al., in which case the photographs would be less discriminable even with high quality. However, the objects used in the present study differed from each other in color, shape, texture, and, in over half of the pairs, in size. Thus, this objection cannot be maintained.
ELLEN WINNER and GEORGEETTLINGER
420
Experience pariement
jets et avec capable3
1. : On a soumis
2 l'exemple.
leur representation
d'apparier
gnificative
Les objets
dans
les objets;
2 2 chimpanz6s
devaient
Gtre
photoqraphi$ue. il y avait
leur performance
quand
une dpreuve
apparies
avec
Si les sujets
par contre
ils devaient
d'ap-
les ob6taient
une baisse apparier
si-
les ob-
jets aux photoqraphies.
soumise
Experience _2 3 2 chimpanzes,
curits,
les objets
r6ussissaient
avec
de reconnaissance
: Une t&he les objets
intermodale
Btaient
dans
ou leurs photoqraphies les objets
mais
intermodale
Ces r6sultats conclusions
selon
de reconnaitre
lesquelles
pr6sent&s
2 la lumiere.
Bchouaient
quand
(objets-photoqraphies
jettent
mier
regard
de reconnaissance
quelques
doutes
les chimpanzds
des objets
sent
repr6sentes
a et6
l'obs-
Les sujets
il s'aqissait et vice-versa).
sur de prec6dentes capables
dPs le pre-
photographiquement.
Deutschsprachige Zusammenfassung: Experiment I: Zwei Schimpansen wurden einer Zuordnungsaufgabe unterworfen, bei welcher Gegenstande (Objekte) mit Gegenstanden und Photographien tibereinstimmten.Wahrend die Versuchstiere Objekte zuzuordnen vermochten, nahm die Leistung signifikant ab, wenn die Zuordnung von Objekten zu Photographien verlangt wurde. zxperiment II: Zwei Schimpansen wurde eine Aufgabe mit gekreuzt modalem Erkennen dargeboten, in welcher Objekte im Dunkeln gezeigt wurden, sowie Objekte oder ihre Photographien bei Licht. Die Versuchstiere hatten Erfolg mit Objekten, scheiterten jedoch, wenn gekreuzt-modales Erkennen von Objekten auf Photographien und vice-versa verlangt wurde. Diese Ergebnisse werfen Zweifel auf Berichte, daR Ychimpansen auf den ersten Blick in der Lage sind, photographiesh dargestellte Objekte zu erkennen.
