Faces in Perception [1144] Perception, 1983, volume 12, pages 55-61
A refutation of the hypothesis of the superfidel ity of caricatures relative to photographs
Margaret A Hagen Boston University, Boston, MA 02215, USA
David Perkins Harvard University, Cambridge, MA 02138, USA Received 1 February 1980, in revised form 19 April 1982
Abstract. The experiment was designed to test the hypothesis that caricatures, relative to photographs, are 'superfaithful' carriers of information for facial recognition. Subjects were shown fifteen pictures of people's faces and were then asked to pick those same people out of a set of fifty-four pictures. There were three sets of pictures: caricatures, profile-view photographs, and three-quarter-view photographs. There were nine groups of subjects: for three groups the exposure and test stimuli were in the same medium, for six groups the test stimuli were in one of the media not previously seen. Points were scored for the number of people correctly identified and the number of false positives. Facial recognition within medium was very good, but was seriously disrupted by any medium shift, especially those involving caricatures. It is argued that the superfidelity of caricature may be manifest only when the task involves recognition of actual persons rather than their pictures. 1 Introduction
In general, the theoretical positions established in the field on the functioning of caricature seem to take two forms. The first position, presented most clearly by Gombrich (1969) and Goodman (1968), holds that caricatures succeed as representations because they are highly stylized conventional systems read according to the dictates of culture and the observer's subconscious. They allow for the detection of likeness between subject and picture not on the basis of distinctive facial features preserved in the caricature transformations but on the basis of the culturally prevalent system of pictorial schemata or language. By this argument representational customs and not physiognomy determine the similarity between caricatures and their subjects. The opposite view is that of Gibson (1971), Perkins (1975) , and Hagen (1974). This view holds that caricatures maintain not just fidelity to the subjects they picture, but superfidelity. This superfidelity is a consequence of the preservation and exaggeration of the most salient characteristics of a person's face coupled with the deletion of characteristics irrelevant or unimportant to identification. Perkins (1975) was the first psychologist to begin an analysis of the techniques of caricature within the framework of the distinctive-feature theory described above. He hypothesized that caricature recognition is identical to the process of ordinary facial recognition and that caricatures must therefore contain the same attributes as the caricatured face itself or photographs of the subject. Perkins's hypothesis that feature correspondence between picture and subject is the core of successful facial recognition in caricatures can be approached by the straightforward process of looking at existing caricatures and their subjects. So this is exactly what Perkins did. Through informal observation of caricatures and photographs of Richard M Nixon (President of the USA 1968-1974) and through the deletion of various facial attributes from those pictures Perkins found that four key properties of caricatures of Mr Nixon's face were jowls, a hairline with bays on either side, a box chin, and a long nose. The omission of these properties or contradiction of any one seriously degraded the representative character of the caricature. If Perkins is right about the
Faces in Perception [1145] 56
M A Hagen, D Perkins
critical nature of these facial properties for the success of a caricature, then it should follow that: (a) all artists generally use what Perkins calls 'the rather necessary' key properties and thus are consistent among themselves in the nature of their depictions; (b) an individual artist should be consistent in his depiction across time; and (c) changes over time in the form of the caricature should be a function of changes in the face itself. Recently Goldman and Hagen (1978) undertook an empirical analysis of several of the consistency assumptions which seemed to follow from Perkins's model of the caricature process. They also looked at caricatures of Richard Nixon to facilitate comparisons between their work and that of Perkins. They selected eleven feature ratios which seemed to represent obvious candidates for distinctive features, were easily measured, and contained Perkins's four key facial properties. They looked at the question of consistency of distorted features across seventeen artists within the year 1972, and found a very high degree of concordance among artists in terms of which feature ratios were most and which were least distorted. However, with regard to degree of exaggeration of feature ratios they observed very great discrepancies among artists, ranging from 12% distortion of photographic feature ratios to 86%, with a mean of 53%. Thus it was argued that the choice of what to caricature is determined largely by characteristics of the subject's face, while the degree of caricature is determined by the individual artist's style and bias. Of course, the analysis addressed this important question of artists' styles in a very simplistic fashion; it looked only at degree of distortion as it distinguished one artist from another. It is perfectly clear that there are many other factors which determine the particular style characterizing an artist, but other features of style were beyond the scope of the study. Goldman and Hagen also looked at the question of consistency across time. When features ranked for degree of distortion were compared across 1972-1973, the rank correlation was very high and significant, indicating great consistency across time in choice of features to be exaggerated. However, when degree of distortion across time was measured, it was found that real physical changes in the subject's face could not account for the significant increase in the degree of distortion observed from 1972 to 1973. The authors argued that this increase in distortion was largely a function of the increased negativity of the political climate surrounding Mr Nixon in 1973 during the unfolding of the Watergate situation. In sum, then, Goldman and Hagen found substantial evidence that the degree of distortion present in a caricature is a function of political climate, personal bias and style of the artist, and will vary significantly across time and artist. At the same time, however, they also found strong support for Perkins's analysis of the process of caricature as primarily a function of true physiognomy. In general, research on facial perception in photographs is becoming increasingly more sophisticated as is evidenced by the recent work by Pittenger and Shaw ( l 975a, 1975b) on the geometric determinants of relative and absolute age in photographs, and by the work of Hochberg and Galper (1967) and of Langberg (1971) on the recognition of faces, and in the continuing interest of the neuropsychologists in the localization of facial-recognition functions (eg Warrington and James 1967; Moscovitch et al 1976). Yet the field lacks any explicit test of the effects of medium on information transmission in pictures of faces. The only study which comes close is that of Ryan and Schwartz (1956), who investigated not caricatures but cartoons. They ran an experiment to test the effects of different media on the rapid pickup of pictorial information for such items as the position of hands, the position of a switch,
etc. They compared a black-and-white photograph, an ink-and-wash shaded drawing, a high-fidelity line drawing, and cartoons. The order of pickup from fastest to slowest was as follows: (a) cartoons; (b) photographs and shaded drawints; and (c)
Faces in Perception [1146] Refutation of superfidelity of caricatures
57
line drawings. Although the relationship between cartoons and caricatures in general has never been empirically determined, the cartoons of Ryan and Schwartz were basically drawings which exaggerated the distinctive features of the subjects and deleted the inessentials. Thus it is reasonable to argue that, like these cartoons, caricatures should be the most efficient pictorial medium for depicting information essential to facial recognition. If this analysis is correct, then caricatures should be superior to photographs in facilitating recognition accuracy and in the speed with which they transmit salient information. The specific purpose of the present study was to test this analysis and to determine if, indeed, caricatures are more efficient pictorial transmitters of information crucial to facial recognition than are photographs. 2 Methods 2.1 Subjects One hundred and eighty men and women from introductory psychology classes served as subjects in this experiment. There were twenty subjects in each of the nine conditions. About half of each group were male and half were female.
2.2 Stimuli The single greatest difficulty in the way of investigating caricature in psychological studies is the problem of acquiring adequate caricatures. The simplest solution, and that taken by Perkins (1975) and Goldman and Hagen (1978), is to take caricatures of public figures from the journalistic media of newspapers and newsmagazines. But there is a special difficulty with this solution. Public figures have faces which are both overlearned and overcaricaturized. With such subjects, it is extremely difficult to disentangle the roles of distinctive features, conventions of depiction, and familiarity. So it seems more reasonable to use photographs and caricatures of unfamiliar faces to avoid this difficulty. But such caricatures generally do not exist, so it is necessary to create them. This presents a serious problem for the investigator. Successful caricatures are the products of very talented people who have nowhere managed to systematize their techniques into rules available to the general public. Ordinary principles of drafting and perspective simply do not apply, and while the geometric-transformation system suggested by Pittenger and Shaw (1975a, 1975b) looks promising, it is far too early to apply it to a specific generative task. Thus the investigator must find a caricaturist to create stimuli, and this is what we have done. For the project the artist created fifty-four caricatures working from the profile view and three-quarter-view black-and-white photographs of fifty-four different individuals. Thus for fifty-four people we had available for testing three different pictures of each: one three-quarter-view caricature, one three-quarter-view black-and white photograph, and one profile-view photograph. The pictures were all printed on 5 in x 7 in white cards. These fifty-four sets of three pictures each were the stimuli for the investigation. Two examples of such sets are given in figure 1. 2.3 Procedure Two types of experiment were designed to explore the problem. In the first type of experiment, subjects were shown either fifteen black-and-white photographs or fifteen caricatures. These fifteen were then randomly combined with thirty-nine more of the same type of stimulus item; ie photographs or caricatures. In the second type of experiment the subjects were also shown fifteen black-and-white photographs or caricatures. However, these subjects were then shown fifty-four stimulus items in the medium which they had not seen before; ie photographs or caricatures. In both types of experiment, the subjects were asked to identify the people whom they had seen before.
Faces in Perception [1147] 58
M A Hagen D Perkins
Figure 1. Two examples of picture sets: three-quarter-view photograph (top), profile-view photograph (center), caricature (bottom). Artist: Andras Goldinger.
Faces in Perception [1148] 59
Refutation of superfidelity of caricatures
In all of the nine conditions (see table I) the procedure was the same. Each observer was shown fifteen stimulus items. Each item was shown for 3 s. Then the subject was shown fifty-four items and asked to identify the people she or he had seen before. This procedure yielded two scores per subject. The first score was the number the subject correctly identified as having seen before. The second score was the number of items she or he thought had been seen before but indeed had not been, ie the number of false alarms. Table 1. Means
(x ) and standard deviations (sd) for same-media and crossed-media conditions.
Condition3
Correct responses (15 possible)
False alarms (39 possible)
X
sd
X
sd
13·00 11·90 11·55
2·27 1·69 2·09
1·70 3·10 5·00
2·70 2·45 4·19
9·75 8·47 7·15 6·25 5·75 5·00
3·01 2·95 2·11 2·22 2·46 2·73
5·25 5·55 9·75 9·70 11·80 6·90
3·89 4·63 3·96 4·71 5·65 4·32
Same-media 3 3 4-4
profile-proflle caricature-caricature Crossed-media
�-proflle profile-� i -caricature caricature-i caricature -proflle proflle-caricature _
3 i stands for three-quarter-view photograph; proflle stands for proflle-view photograph. *Values within a box do not differ significantly from each other.
3 Results For each of the nine conditions a mean score and standard deviation were calculated both for the number of correct responses and for the number of false alarms. Table 1 presents these means and standard deviations. The total number of correct identifications that was possible was fifteen and the total number of possible false alarms was thirty-nine. An analysis of variance was performed on the number of correct responses and on the number of false alarms to examine: (i) the relationship among the three same-media conditions; (ii) the relationship among the six crossed media conditions and; (iii) the differences between the same-media and crossed-media conditions. Sex of subject was also examined as a possible variable. No significant differences were found among the same-media conditions with regard to the number of correct responses, but for the number of false alarms a significant 5 ·47, p < 0·007). Duncan multiple-range a posteriori diffference was found (F comparison revealed that significantly more false alarms were generated in the caricature conditions than in the three-quarter-view or profile-view photograph conditions. For the crossed-media conditions a significant difference was found among conditions both for the number correct and for the number of false alarms (F 7·87, p < 0·001, and 5 ·55, p < 0 ·00 I, respectively). Duncan multiple-range a posteriori comparisons F showed that for the number of correct responses each crossed-media condition did not differ from its reversal, ie three-quarter-view-caricature (7·15) did not differ from caricature-three-quarter-view (6 25), etc; and, for the number of false alarms, each crossed-media condition did not differ from its reversal, with the single exception that the caricature-profile-view condition did generate more false alarms (11·80) than the profile-view-caricature condition (6 ·90). Overall the four conditions which involved caricatures generated fewer correct responses than did the two conditions in =
=
=
·
Faces in Perception [1149] M A Hagen, D Perkins
60
which only photographs were used, again with the single exception that the three quarter-view-caricature condition (7·15) did not generate significantly fewer correct responses than the profile-view-three-quarter-view condition (8 ·47). It was also generally true that conditions with caricatures produced more false alarms than conditions with photographs, with the exception that the profile-view-caricature condition (6 ·90) did not differ significantly from either the three-quarter-view profile-view condition (5 25) or its reversal (5 ·55). In the combined analysis testing for differences among all nine conditions it was found that the main effect for condition was significant both for the number of correct responses and for the number of false alarms (F 28·61, p < 0·0001, and F 11·34, p < 0·0001, respectively). Duncan multiple-range a posteriori comparisons showed that all of the same-media conditions generated more correct identifications than did the crossed-media conditions (p < 0·05). In addition, all of the same-media conditions generated fewer false alarms than did the crossed-media conditions, except for the caricature-caricature condition (p < 0·05). In a posteriori testing for sex differences, no sex difference was consistent across conditions. ·
=
=
4 Conclusions and implications Quite clearly the above results have not supported Gibson's (1971) and Perkins's (1975) hypothesis of the superfidelity of caricatures. There is no evidence to support the notion that caricatures are a more efficacious communication device than photographs for the specification of facial identity. Indeed, caricatures generate more false alarms even than profile photographs, contrary to any simple distinctive feature hypothesis of facial recognition. The consistent result overall was that facial recognition within medium was very good, but was seriously impaired by any medium shift including those involving caricature. Of course, facial recognition can also be considered to be like any other pattern-recognition task where the identity judgment may often be made on the basis of single features rather than clusters of distinctive features specifying particular persons. In that case caricatures should enjoy no advantage over photographs, even profile-views. We argue that the superiority of caricatures may therefore only be manifested in tasks which require real 'person identity', like picking the subject from a set of live models, as in a police lineup, or in tasks where exposure time is extremely limited, like that used by Ryan and Schwartz (1956). Surely it is reasonable to argue that the speed of processing all available information increases as the amount of information available decreases; it is reasonable, too, to argue that the clarity of presentation of information increases as the amount of information available decreases. Thus, a Ryan and Schwartz design might show the superiority of caricatures, again not for 'person perception' but simply for clarity of presented features. In any case the present method of testing identification with 3 s exposure times most certainly did not support the hypothesized superiority of caricatures relative to photographs. Both for the number of correct responses and for the number of false alarms three out of four of the conditions with caricatures differed significantly from the two conditions, with photographs, and those differences were not in the predicted directions. Acknowledgements. The authors wish to thank the artist, Andras Goldinger, for his invaluable
help with this project. References
Gibson J J, 1971 "The information available in pictures" Leonardo 4 27-35 Goldman M, Hagen M A, 1978 "The forms of caricature: Physiognomy and political bias" Studies on the Anthropology of Visual Communication 6 30-36 Gombrich E H, 1969 Art and Illusion: A Study in the Psychology of Pictorial Representation (Princeton, NJ: Princeton University Press)
Faces in Perception [1150] Refutation of superfidelity of caricatures
Goodman N, 1968 Languages of Art: An Approach to a Theory of Symbols (New York: Hobbs Merrill) Hagen M A, 1974 "Picture perception: Toward a theoretical model" Psychological Bulletin 81 471-479 Hochberg J, Galper R, 1967 "Recognition of faces: I. An explanatory study" Psychonomic Science 9 619-620 Laughery K R, 1971 "Recognition of human faces: Effects of target exposure time, target position, pose position and type of photograph" Journal of Applied Psychology 55 477-483 Moscovitch M, Scullion D, Christie D, 1976 "Early versus late stages of processing and their relation to functional hemispheric asymetries in face recognition" Journal of Experimental Psychology: Human Perception and Performance 2 401-416 Perkins D, 1975 "A definition of caricature and recognition" Studies in the Anthropology of Visual Communication 2(1 ), 1-24 Pittenger J B, Shaw R E, 1975a "Ageing faces as viscal-elastic events: Implications for a theory of nonrigid shape perception" Journal of Experimental Psychology: Human Perception and Performance 1 374-382 Pittenger J B, Shaw R E, 1975b "Perception of relative and absolute age in facial photog raphs" Perception & Psychophysics 18 137-143 Ryan T A, Schwartz C B, 1956 "Speed of perception as a function of mode of representation" American Journal of Psychology 69 193- 199 Warrington E K, James M, 1967 "An experimental investigation of facial recognition in patients with unilateral cerebral lesions" Cortex 3 317-326
61
A refutation of the hypothesis of the superfidel ity of caricatures relative to photographs
Margaret A Hagen Boston University, Boston, MA 02215, USA
David Perkins Harvard University, Cambridge, MA 02138, USA Received 1 February 1980, in revised form 19 April 1982
Abstract. The experiment was designed to test the hypothesis that caricatures, relative to photographs, are 'superfaithful' carriers of information for facial recognition. Subjects were shown fifteen pictures of people's faces and were then asked to pick those same people out of a set of fifty-four pictures. There were three sets of pictures: caricatures, profile-view photographs, and three-quarter-view photographs. There were nine groups of subjects: for three groups the exposure and test stimuli were in the same medium, for six groups the test stimuli were in one of the media not previously seen. Points were scored for the number of people correctly identified and the number of false positives. Facial recognition within medium was very good, but was seriously disrupted by any medium shift, especially those involving caricatures. It is argued that the superfidelity of caricature may be manifest only when the task involves recognition of actual persons rather than their pictures. 1 Introduction
In general, the theoretical positions established in the field on the functioning of caricature seem to take two forms. The first position, presented most clearly by Gombrich (1969) and Goodman (1968), holds that caricatures succeed as representations because they are highly stylized conventional systems read according to the dictates of culture and the observer's subconscious. They allow for the detection of likeness between subject and picture not on the basis of distinctive facial features preserved in the caricature transformations but on the basis of the culturally prevalent system of pictorial schemata or language. By this argument representational customs and not physiognomy determine the similarity between caricatures and their subjects. The opposite view is that of Gibson (1971), Perkins (1975) , and Hagen (1974). This view holds that caricatures maintain not just fidelity to the subjects they picture, but superfidelity. This superfidelity is a consequence of the preservation and exaggeration of the most salient characteristics of a person's face coupled with the deletion of characteristics irrelevant or unimportant to identification. Perkins (1975) was the first psychologist to begin an analysis of the techniques of caricature within the framework of the distinctive-feature theory described above. He hypothesized that caricature recognition is identical to the process of ordinary facial recognition and that caricatures must therefore contain the same attributes as the caricatured face itself or photographs of the subject. Perkins's hypothesis that feature correspondence between picture and subject is the core of successful facial recognition in caricatures can be approached by the straightforward process of looking at existing caricatures and their subjects. So this is exactly what Perkins did. Through informal observation of caricatures and photographs of Richard M Nixon (President of the USA 1968-1974) and through the deletion of various facial attributes from those pictures Perkins found that four key properties of caricatures of Mr Nixon's face were jowls, a hairline with bays on either side, a box chin, and a long nose. The omission of these properties or contradiction of any one seriously degraded the representative character of the caricature. If Perkins is right about the
Faces in Perception [1145] 56
M A Hagen, D Perkins
critical nature of these facial properties for the success of a caricature, then it should follow that: (a) all artists generally use what Perkins calls 'the rather necessary' key properties and thus are consistent among themselves in the nature of their depictions; (b) an individual artist should be consistent in his depiction across time; and (c) changes over time in the form of the caricature should be a function of changes in the face itself. Recently Goldman and Hagen (1978) undertook an empirical analysis of several of the consistency assumptions which seemed to follow from Perkins's model of the caricature process. They also looked at caricatures of Richard Nixon to facilitate comparisons between their work and that of Perkins. They selected eleven feature ratios which seemed to represent obvious candidates for distinctive features, were easily measured, and contained Perkins's four key facial properties. They looked at the question of consistency of distorted features across seventeen artists within the year 1972, and found a very high degree of concordance among artists in terms of which feature ratios were most and which were least distorted. However, with regard to degree of exaggeration of feature ratios they observed very great discrepancies among artists, ranging from 12% distortion of photographic feature ratios to 86%, with a mean of 53%. Thus it was argued that the choice of what to caricature is determined largely by characteristics of the subject's face, while the degree of caricature is determined by the individual artist's style and bias. Of course, the analysis addressed this important question of artists' styles in a very simplistic fashion; it looked only at degree of distortion as it distinguished one artist from another. It is perfectly clear that there are many other factors which determine the particular style characterizing an artist, but other features of style were beyond the scope of the study. Goldman and Hagen also looked at the question of consistency across time. When features ranked for degree of distortion were compared across 1972-1973, the rank correlation was very high and significant, indicating great consistency across time in choice of features to be exaggerated. However, when degree of distortion across time was measured, it was found that real physical changes in the subject's face could not account for the significant increase in the degree of distortion observed from 1972 to 1973. The authors argued that this increase in distortion was largely a function of the increased negativity of the political climate surrounding Mr Nixon in 1973 during the unfolding of the Watergate situation. In sum, then, Goldman and Hagen found substantial evidence that the degree of distortion present in a caricature is a function of political climate, personal bias and style of the artist, and will vary significantly across time and artist. At the same time, however, they also found strong support for Perkins's analysis of the process of caricature as primarily a function of true physiognomy. In general, research on facial perception in photographs is becoming increasingly more sophisticated as is evidenced by the recent work by Pittenger and Shaw ( l 975a, 1975b) on the geometric determinants of relative and absolute age in photographs, and by the work of Hochberg and Galper (1967) and of Langberg (1971) on the recognition of faces, and in the continuing interest of the neuropsychologists in the localization of facial-recognition functions (eg Warrington and James 1967; Moscovitch et al 1976). Yet the field lacks any explicit test of the effects of medium on information transmission in pictures of faces. The only study which comes close is that of Ryan and Schwartz (1956), who investigated not caricatures but cartoons. They ran an experiment to test the effects of different media on the rapid pickup of pictorial information for such items as the position of hands, the position of a switch,
etc. They compared a black-and-white photograph, an ink-and-wash shaded drawing, a high-fidelity line drawing, and cartoons. The order of pickup from fastest to slowest was as follows: (a) cartoons; (b) photographs and shaded drawints; and (c)
Faces in Perception [1146] Refutation of superfidelity of caricatures
57
line drawings. Although the relationship between cartoons and caricatures in general has never been empirically determined, the cartoons of Ryan and Schwartz were basically drawings which exaggerated the distinctive features of the subjects and deleted the inessentials. Thus it is reasonable to argue that, like these cartoons, caricatures should be the most efficient pictorial medium for depicting information essential to facial recognition. If this analysis is correct, then caricatures should be superior to photographs in facilitating recognition accuracy and in the speed with which they transmit salient information. The specific purpose of the present study was to test this analysis and to determine if, indeed, caricatures are more efficient pictorial transmitters of information crucial to facial recognition than are photographs. 2 Methods 2.1 Subjects One hundred and eighty men and women from introductory psychology classes served as subjects in this experiment. There were twenty subjects in each of the nine conditions. About half of each group were male and half were female.
2.2 Stimuli The single greatest difficulty in the way of investigating caricature in psychological studies is the problem of acquiring adequate caricatures. The simplest solution, and that taken by Perkins (1975) and Goldman and Hagen (1978), is to take caricatures of public figures from the journalistic media of newspapers and newsmagazines. But there is a special difficulty with this solution. Public figures have faces which are both overlearned and overcaricaturized. With such subjects, it is extremely difficult to disentangle the roles of distinctive features, conventions of depiction, and familiarity. So it seems more reasonable to use photographs and caricatures of unfamiliar faces to avoid this difficulty. But such caricatures generally do not exist, so it is necessary to create them. This presents a serious problem for the investigator. Successful caricatures are the products of very talented people who have nowhere managed to systematize their techniques into rules available to the general public. Ordinary principles of drafting and perspective simply do not apply, and while the geometric-transformation system suggested by Pittenger and Shaw (1975a, 1975b) looks promising, it is far too early to apply it to a specific generative task. Thus the investigator must find a caricaturist to create stimuli, and this is what we have done. For the project the artist created fifty-four caricatures working from the profile view and three-quarter-view black-and-white photographs of fifty-four different individuals. Thus for fifty-four people we had available for testing three different pictures of each: one three-quarter-view caricature, one three-quarter-view black-and white photograph, and one profile-view photograph. The pictures were all printed on 5 in x 7 in white cards. These fifty-four sets of three pictures each were the stimuli for the investigation. Two examples of such sets are given in figure 1. 2.3 Procedure Two types of experiment were designed to explore the problem. In the first type of experiment, subjects were shown either fifteen black-and-white photographs or fifteen caricatures. These fifteen were then randomly combined with thirty-nine more of the same type of stimulus item; ie photographs or caricatures. In the second type of experiment the subjects were also shown fifteen black-and-white photographs or caricatures. However, these subjects were then shown fifty-four stimulus items in the medium which they had not seen before; ie photographs or caricatures. In both types of experiment, the subjects were asked to identify the people whom they had seen before.
Faces in Perception [1147] 58
M A Hagen D Perkins
Figure 1. Two examples of picture sets: three-quarter-view photograph (top), profile-view photograph (center), caricature (bottom). Artist: Andras Goldinger.
Faces in Perception [1148] 59
Refutation of superfidelity of caricatures
In all of the nine conditions (see table I) the procedure was the same. Each observer was shown fifteen stimulus items. Each item was shown for 3 s. Then the subject was shown fifty-four items and asked to identify the people she or he had seen before. This procedure yielded two scores per subject. The first score was the number the subject correctly identified as having seen before. The second score was the number of items she or he thought had been seen before but indeed had not been, ie the number of false alarms. Table 1. Means
(x ) and standard deviations (sd) for same-media and crossed-media conditions.
Condition3
Correct responses (15 possible)
False alarms (39 possible)
X
sd
X
sd
13·00 11·90 11·55
2·27 1·69 2·09
1·70 3·10 5·00
2·70 2·45 4·19
9·75 8·47 7·15 6·25 5·75 5·00
3·01 2·95 2·11 2·22 2·46 2·73
5·25 5·55 9·75 9·70 11·80 6·90
3·89 4·63 3·96 4·71 5·65 4·32
Same-media 3 3 4-4
profile-proflle caricature-caricature Crossed-media
�-proflle profile-� i -caricature caricature-i caricature -proflle proflle-caricature _
3 i stands for three-quarter-view photograph; proflle stands for proflle-view photograph. *Values within a box do not differ significantly from each other.
3 Results For each of the nine conditions a mean score and standard deviation were calculated both for the number of correct responses and for the number of false alarms. Table 1 presents these means and standard deviations. The total number of correct identifications that was possible was fifteen and the total number of possible false alarms was thirty-nine. An analysis of variance was performed on the number of correct responses and on the number of false alarms to examine: (i) the relationship among the three same-media conditions; (ii) the relationship among the six crossed media conditions and; (iii) the differences between the same-media and crossed-media conditions. Sex of subject was also examined as a possible variable. No significant differences were found among the same-media conditions with regard to the number of correct responses, but for the number of false alarms a significant 5 ·47, p < 0·007). Duncan multiple-range a posteriori diffference was found (F comparison revealed that significantly more false alarms were generated in the caricature conditions than in the three-quarter-view or profile-view photograph conditions. For the crossed-media conditions a significant difference was found among conditions both for the number correct and for the number of false alarms (F 7·87, p < 0·001, and 5 ·55, p < 0 ·00 I, respectively). Duncan multiple-range a posteriori comparisons F showed that for the number of correct responses each crossed-media condition did not differ from its reversal, ie three-quarter-view-caricature (7·15) did not differ from caricature-three-quarter-view (6 25), etc; and, for the number of false alarms, each crossed-media condition did not differ from its reversal, with the single exception that the caricature-profile-view condition did generate more false alarms (11·80) than the profile-view-caricature condition (6 ·90). Overall the four conditions which involved caricatures generated fewer correct responses than did the two conditions in =
=
=
·
Faces in Perception [1149] M A Hagen, D Perkins
60
which only photographs were used, again with the single exception that the three quarter-view-caricature condition (7·15) did not generate significantly fewer correct responses than the profile-view-three-quarter-view condition (8 ·47). It was also generally true that conditions with caricatures produced more false alarms than conditions with photographs, with the exception that the profile-view-caricature condition (6 ·90) did not differ significantly from either the three-quarter-view profile-view condition (5 25) or its reversal (5 ·55). In the combined analysis testing for differences among all nine conditions it was found that the main effect for condition was significant both for the number of correct responses and for the number of false alarms (F 28·61, p < 0·0001, and F 11·34, p < 0·0001, respectively). Duncan multiple-range a posteriori comparisons showed that all of the same-media conditions generated more correct identifications than did the crossed-media conditions (p < 0·05). In addition, all of the same-media conditions generated fewer false alarms than did the crossed-media conditions, except for the caricature-caricature condition (p < 0·05). In a posteriori testing for sex differences, no sex difference was consistent across conditions. ·
=
=
4 Conclusions and implications Quite clearly the above results have not supported Gibson's (1971) and Perkins's (1975) hypothesis of the superfidelity of caricatures. There is no evidence to support the notion that caricatures are a more efficacious communication device than photographs for the specification of facial identity. Indeed, caricatures generate more false alarms even than profile photographs, contrary to any simple distinctive feature hypothesis of facial recognition. The consistent result overall was that facial recognition within medium was very good, but was seriously impaired by any medium shift including those involving caricature. Of course, facial recognition can also be considered to be like any other pattern-recognition task where the identity judgment may often be made on the basis of single features rather than clusters of distinctive features specifying particular persons. In that case caricatures should enjoy no advantage over photographs, even profile-views. We argue that the superiority of caricatures may therefore only be manifested in tasks which require real 'person identity', like picking the subject from a set of live models, as in a police lineup, or in tasks where exposure time is extremely limited, like that used by Ryan and Schwartz (1956). Surely it is reasonable to argue that the speed of processing all available information increases as the amount of information available decreases; it is reasonable, too, to argue that the clarity of presentation of information increases as the amount of information available decreases. Thus, a Ryan and Schwartz design might show the superiority of caricatures, again not for 'person perception' but simply for clarity of presented features. In any case the present method of testing identification with 3 s exposure times most certainly did not support the hypothesized superiority of caricatures relative to photographs. Both for the number of correct responses and for the number of false alarms three out of four of the conditions with caricatures differed significantly from the two conditions, with photographs, and those differences were not in the predicted directions. Acknowledgements. The authors wish to thank the artist, Andras Goldinger, for his invaluable
help with this project. References
Gibson J J, 1971 "The information available in pictures" Leonardo 4 27-35 Goldman M, Hagen M A, 1978 "The forms of caricature: Physiognomy and political bias" Studies on the Anthropology of Visual Communication 6 30-36 Gombrich E H, 1969 Art and Illusion: A Study in the Psychology of Pictorial Representation (Princeton, NJ: Princeton University Press)
Faces in Perception [1150] Refutation of superfidelity of caricatures
Goodman N, 1968 Languages of Art: An Approach to a Theory of Symbols (New York: Hobbs Merrill) Hagen M A, 1974 "Picture perception: Toward a theoretical model" Psychological Bulletin 81 471-479 Hochberg J, Galper R, 1967 "Recognition of faces: I. An explanatory study" Psychonomic Science 9 619-620 Laughery K R, 1971 "Recognition of human faces: Effects of target exposure time, target position, pose position and type of photograph" Journal of Applied Psychology 55 477-483 Moscovitch M, Scullion D, Christie D, 1976 "Early versus late stages of processing and their relation to functional hemispheric asymetries in face recognition" Journal of Experimental Psychology: Human Perception and Performance 2 401-416 Perkins D, 1975 "A definition of caricature and recognition" Studies in the Anthropology of Visual Communication 2(1 ), 1-24 Pittenger J B, Shaw R E, 1975a "Ageing faces as viscal-elastic events: Implications for a theory of nonrigid shape perception" Journal of Experimental Psychology: Human Perception and Performance 1 374-382 Pittenger J B, Shaw R E, 1975b "Perception of relative and absolute age in facial photog raphs" Perception & Psychophysics 18 137-143 Ryan T A, Schwartz C B, 1956 "Speed of perception as a function of mode of representation" American Journal of Psychology 69 193- 199 Warrington E K, James M, 1967 "An experimental investigation of facial recognition in patients with unilateral cerebral lesions" Cortex 3 317-326
61
