Baboons (Papio papio) spontaneously process the first-order but not second-order configural properties of faces.код для вставкиСкачать
American Journal of Primatology 70:415–422 (2008) RESEARCH ARTICLE Baboons (Papio papio) Spontaneously Process the First-Order but not Second-Order Configural Properties of Faces CAROLE PARRON AND JOËL FAGOT CNRS, Mediterranean Institute of Cognitive Neurosciences, UMR 6193 University of Méditerranée, Marseille, France A two-alternative forced-choice discrimination task was used to assess whether baboons (N 5 7) spontaneously process qualitative (i.e., first-order) or quantitative (i.e., second-order) variations in the configural arrangement of facial features. Experiment 1 used as test stimuli second-order pictorial faces of humans or baboons in which the mouth and the eyes were rotated upside down relative to the normal face. Baboons readily discriminated two different normal faces but did not discriminate a normal face from its second-order modified version. Experiment 2 used human or baboon faces for which the firstorder configural properties had been distorted by reversing the location of the eyes and mouth within the face. Discrimination was prompt with these stimuli. Experiment 3 replicated some of the conditions and the results of experiment 1, thus ruling out possible effects of learning. It is concluded that baboons are more adept at spontaneously processing first- than second-order configural facial properties, similar to what is known in the human developmental literature. Am. J. Primatol. 70:415–422, 2008. c 2007 Wiley-Liss, Inc. Key words: face; configural processing; nonhuman primate; Thatcher illusion INTRODUCTION Faces are complex stimuli that can be discriminated by a series of ‘‘featural’’ as well as ‘‘configural’’ information [Carey & Diamond, 1977]. Facial features consist of facial parts that can be identified in isolation, such as the nose or the mouth. By contrast, the configural properties of the face refer to the spatial layout of these features; for instance, the distance between the nose and the mouth may differ from one individual to another or may vary with facial expressions. Numerous studies have demonstrated that humans strongly rely on configural information to recognize faces [Bruce et al., 1991; Rhodes et al., 1993], although consideration of facial features may serve the processing of other aspects of the faces, such as their expressed emotions [Deruelle & Fagot, 2005]. Whether or not configural information prevails over featural information in face recognition in animals remains unknown, especially for nonhuman primates. The main source of uncertainty derives from studies in which face recognition was assessed with upright and inverted stimuli. Because features remain unchanged with inversion, effects of inversion in humans are usually interpreted as demonstrating the prevalence of configural over featural cues during face processing [Maurer et al., 2002]. Several studies on nonhuman primates have reported that inversion of facial stimuli has few effects on recognition performance. For instance, Rosenfeld and Van Hoesen  found that inver- r 2007 Wiley-Liss, Inc. sion of the face of conspecifics did not disrupt the recognition performance of rhesus macaques in a forced-choice discrimination task. Also using pictures of conspecifics, Bruce  reported similar results in cynomolgus macaques. Tomonaga et al.  found no effect of stimulus inversion in a chimpanzee (Pan troglodytes) tested on a face discrimination task with familiar faces of humans and chimpanzees. Such findings seem to contradict the vast literature suggesting the prevalence of configural properties in face recognition by humans [e.g., Rhodes et al., 1993]. However, some other primate studies converge with the human literature, finding that upright faces are processed in a different way than upside-down faces. Thus, using a sensory reinforcement procedure, Tomonaga  found that macaques looked longer at upright than inverted photographs. Also, Parr et al.  reported Contract grant sponsor: OMLL Eurocores; Contract grant sponsor: EC NEST SEDSU; Contract grant number: 012984; Contract grant sponsor: CNRS OHLL; Contract grant number: 2004-004. Correspondence to: Joël Fagot, INCM, CNRS, 31 Chemin Joseph Aiguier, 13402 Marseille cedex 20, France. E-mail: email@example.com Received 2 July 2007; revised 24 October 2007; revision accepted 26 October 2007 DOI 10.1002/ajp.20503 Published online 20 November 2007 in Wiley InterScience (www.interscience.wiley.com). 416 / Parron and Fagot better discrimination of human and chimpanzee faces by chimpanzees, when these faces were shown right side up. These studies showing variable effects of inversion remain difficult to reconcile and it is unclear if different outcomes are due to subjectrelated factors (e.g., species, past exposure to pictures) or variations in test procedures. Nevertheless, the data minimally suggest that nonhuman primates can potentially process both featural information, when recognition performance survives inversion, and configural information, when performance is disrupted by face inversion. Additional experimental studies are called for to clarify how nonhuman animals process faces. Diamond and Carey  advised researchers to distinguish first- from second-order configural relational properties within a face. First-order properties refer to the global qualitative spatial relations among facial features (e.g., the nose is above the mouth) and provide a means for recognizing faces from other visual stimuli. Second-order relational properties refer to the fine spatial relations among features (e.g., the distance between the nose and the mouth relative to the prototypical face arrangement). Subtle spatial second-order information might be crucial to individuate faces, at least in humans. This study considered this distinction to further examine how a nonhuman primate—the baboon—processes facial stimuli. To our knowledge, the distinction between first- and second-order cues in nonhuman primates’ face processing has been considered so far only by Parr et al. . In a first experiment, those authors used composite faces as stimuli in which the upper and lower half parts of the faces were aligned or misaligned, disorganizing second-order relations. In a second experiment, facial subparts were moved apart keeping the firstorder configural properties constant, or were both moved apart and disorganized to simultaneously disrupt first- and second-order cues. Results from these two experiments converged to show that face recognition was impaired in chimpanzees when second-order cues had been manipulated. Three original experiments are reported here. The first experiment used a ‘‘Thatcher illusion’’ procedure [Thompson, 1980] to specifically test the processing of second-order configural cues by baboons. Examples of ‘‘Thatcherized’’ faces are given in the first column of Figure 1, and the original stimuli from which they derive are shown in the second column of Figure 1. The ‘‘Thatcherized’’ faces were made by rotating by 1801 the mouth and the eyes within a normal face. Normal humans very easily discriminate the normal face from its ‘‘Thatcherized’’ version when both are shown upright. By contrast, discrimination is much harder when the stimuli are both presented upside-down, as illustrated in Figure 1 Block 1a and b. Because the original and ‘‘Thatcherized’’ faces differ only by Am. J. Primatol. second-order configural cues, ready discrimination between the upright normal and ‘‘Thatcherized’’ faces demonstrates that our visual system processes second-order configural properties of the face in that viewing condition [Lewis & Johnston, 1997]. By contrast, our relative blindness to facial differences with upside-down ‘‘Thatcherized’’ faces suggests that a featural processing mode is adopted when viewing inverted faces. Before the current research, we knew of only one published study on the Thatcher illusion in animals [Jitsumori & Yoshihara, 1997]: pigeons showed no Thatcher illusion, suggesting possible human/animal differences in the processing of second-order facial stimuli. The paucity of Thatcher illusion studies in nonhumans along with the apparent human–pigeon differences in the processing of ‘‘Thatcherized faces’’ prompted our first experiment with baboons, whose visual system is more homologous to the human visual system than that of pigeons [Fobes & King, 1982]. The second experiment assessed the processing of first-order facial configural properties by the same baboons. In humans, processing of first-order facial properties emerges earlier in ontogeny than the processing of second-order properties [e.g., Mondloch et al., 2003; Valenza et al., 1996]. This suggests that first-order properties might be relatively more salient, and therefore potentially more accessible to baboons. To assess this hypothesis, baboons were required to discriminate a normal face from a ‘‘firstorder’’ modified face created by shifting the location of the eyes and the mouth (see Fig. 1c). To confirm that the results of experiments 1 and 2 were not due to a learning effect, baboons were retested in experiment 3 using a subset of stimuli used in experiment 1. In addition, familiar faces of human or baboon models were systematically used as stimuli in experiments 1–3. According to Diamond and Carey , the ability to exploit second-order relational properties that individuate members of classes is promoted by a stimulus familiarity or expertise. Use of familiar models in our experiments was aimed at promoting consideration of secondorder cues by the baboons. EXPERIMENT 1: PROCESSING OF SECONDORDER RELATIONAL PROPERTIES Participants The participants were seven 18-year-old Guinea baboons (Papio papio), five males (B03, B07, B09, B11, B15) and two females (B04, B08), living in social groups of two to four individuals within the primate facility of the CNRS (Marseille). They were already familiar with the two-alternative forced-choice discrimination task procedures [e.g., Deruelle et al., 2000]. They were fed normally after completion of daily training or testing sessions. Configural Processing by Baboons / 417 Fig. 1. Illustration of some of the face stimuli used in blocks 1–3 and their modified version. (a) ‘‘Thatcherized’’ version of the stimuli, (b) normal faces, (c) first-order modified stimuli. All facial stimuli were presented either upright or inverted to two groups of baboons. Human faces published with written consent. Apparatus The baboons were tested in an experimental booth (68 50 72 cm) comprising a food dispenser, a view port and two hand ports that provided free access to an analogue joystick controlling isomorphic displacements of a cursor on a 17-in color monitor screen located 49 cm away from the view port. The food dispenser delivered 190-mg banana-flavored food pellets inside the test apparatus in accordance with the prevailing reinforcement contingency. Inversion of the eyes and mouth involved an average of only 7.97% (SE 5 1.17) of the total number of pixels composing the original faces. The four human models were all familiar to the baboons, being the caretakers and the first author. The two baboon models were one familiar male and one familiar female, both nonparticipants. All stimuli were shown on a black background with a 1024 768 definition. They subtended approximately 8 81 of visual angle. Procedure Stimuli The stimuli were color digitized frontal views of four human and two baboon faces, (see Fig. 1b)and their ‘‘Thatcherized’’ versions created by turning the eyes and the mouth upside-down (see Fig. 1a). One group of baboons (i.e., B03, B04, B08 and B09) was tested with upright faces and the other one (i.e., B07, B11 and B15) with the same but inverted faces. Each group received three consecutive blocks of test sessions differing only in the models Am. J. Primatol. 418 / Parron and Fagot presented as stimuli. To control for any effect of block order, and to investigate potential differences in the processing of human and baboon faces, blocks 1 and 3 used two different pairs of human faces as stimuli whereas block 2 used pictures of baboons’ faces (see Fig. 1). Using this design, any effect of species should be revealed by comparing the performance on block 2 (pictures of baboons) and blocks 1 and 3 (pictures of humans). By contrast, learning effects should be revealed by a performance increment from blocks 1 to 3, independent of the stimulus species. Each block was divided in two phases aimed at identifying processing differences between normal and ‘‘Thatcherized’’ faces. The first phase used only normal faces. It consisted of six sessions of 20 randomly ordered trials, in which the baboons had to select the face assigned as positive (S1) from among two left–right counterbalanced normal faces. Responses consisted of moving the joystick so as to place the cursor on the S1. There was no time limit for responding. The second phase used ‘‘Thatcherized’’ faces. It consisted of six additional sessions of the 20 trials-test in which the previous S1 stimulus (i.e., a normal face) continued as S1, but was now presented against its ‘‘Thatcherized’’ (S) version. For all the participants, the second phase followed the first phase of the same block. In all blocks and phases, correct responses were food-reinforced, whereas wrong responses triggered a 3-sec time-out green screen and no reward delivery. A 6-sec inter-trial interval separated two consecutive trials, irrespective of their outcome, and therefore followed the time-out in case of an error. Both scores and response times were retained for statistical analyses. The participants completed 1–3 sessions in a day. Results Scores were analyzed with a four-way group (upright, inverted) by block (1, 2 and 3), by phase (phase 1 vs. phase 2), by session (1–6) analysis of variance (ANOVA). This analysis showed that baboons tested with the upright faces behaved indifferently from baboons tested with inverted faces (mean upright 5 73.8%, SD 5 3; mean inverted 5 73.9%, SD 5 2.94, F(1,5) 5 .002, P 5 .96). In addition, none of the possible two- or three-way interactions involving the group factor was even close to significance (all P4.18). The effect of phase was significant, F(1,5) 5 74.32, Po.001. As shown in Figure 2, performance was significantly greater on average in phase 1 using normal faces (mean 5 88.26% correct, SD 5 4.18%) than in phase 2 using ‘‘Thatcherized’’ faces (mean 5 59.2% correct, SD 5 3.33%). This finding indicates a poor sensitivity of the baboons to the configural variations distinguishing a normal face Am. J. Primatol. Fig. 2. Mean percentage of correct responses for each session and with the normal (phase 1) and ‘‘Thatcherized’’ faces (phase 2) in experiment 1. Bars correspond to standard errors. from its ‘‘Thatcherized’’ version. Also reliable was the main effect of block, F(2,10) 5 7.2, Po.03. Tukey honestly significance post hoc tests (Po.05) provided the following pattern of results: block 1 (human faces: 67.2%) oblock 2 (baboon faces: 76.5%) 5 block 3 (human faces: 77.8%). Clearly, the overall performance increment appearing from blocks 1–3 was independent of presenting human or baboon faces as stimuli. Finally, a reliable main effect of the session emerged, F(5,25) 5 24.30, Po.001, attributable to the significant phase (phase 1, phase 2) by session (1–6) interaction of a higher order, F(5,25) 5 16.91, Po.001 (see Fig. 2). Monkeys scored 88% correct on average (SD 5 0.79%) in the second session of phase 1 (normal faces), all blocks confounded. Interestingly, the performance in the final session of phase 2, with ‘‘Thatcherized’’ faces, remained poor (i.e., 62.1% correct) with no clear evidence of learning. D primes (d0 ) were computed to evaluate the discrimination performance of individual baboons. D0 is a measure from the signal detection theory that considers the rate of correct hits and false-alarm responses to evaluate discriminability of the stimuli. A significant d0 indicates that the signal is readily detected, or, in our context, that the feature is processed and that the baboons perceived the difference between a normal and a Thatcherized face. D0 s were calculated for each subject and session (1–6) and for phases 1 and 2. A Bonferroni correction was applied in view of the large number of comparisons (N 5 84). No baboon had a reliable d0 in the first session of phase 1, when they were facing a novel discrimination problem. However, all baboons demonstrated a reliable d0 in sessions 2–6 of phase 1, except for baboon B07 that demonstrated a reliable d0 in sessions 3–6. By contrast, no baboon discriminated reliably in phase 2, in any session. Response times for correct trials were also analyzed, omitting session as a factor owing to the large number of errors per session limiting the data set size. A group (upright-inverted) by block (blocks Configural Processing by Baboons / 419 1, 2, 3), by phase (phases 1 and 2) ANOVA revealed only a significant effect of phase, F(1,5) 5 26.76, Po.001: response times were longer on average in phase 2 (mean 5 1056 ms, SD 5 300) than in phase 1 (mean 5 787 ms, SD 5 217 ms). This result confirms that the discrimination of the Thatcherized faces from their normal faces (phase 2) was more difficult for the baboons than the discrimination of the faces of two different individuals (phase 1). Discussion Two conclusions may be drawn from the first experiment. First, baboons can readily learn to discriminate the faces of two different humans or baboons, but struggle to tell apart a normal face from its ‘‘Thatcherized’’ version. Second, this difficulty emerged independently of the orientation of the stimuli, that is, upright or inverted. These findings are in sharp contrast with what has been reported in the human literature, which shows that the Thatcher illusion emerges without training and with upright faces only [e.g., Thompson, 1980]. Lack of discrimination between the normal and ‘‘Thatcherized’’ faces suggests that the baboons did not spontaneously process the second-order relational properties of the facial attributes. EXPERIMENT 2: PROCESSING OF ‘‘FIRSTORDER’’ RELATIONAL PROPERTIES Given that experiment 1 failed to demonstrate the processing of second-order relational properties by baboons, experiment 2 assessed their ability to process first-order spatial relations, which, in humans, emerge earlier in development [in newborn infants: e.g., Goren et al., 1975; Valenza et al., 1996] than the processing of second-order relations. Stimuli Experiment 2 used the same original faces as in experiment 1 (see Fig. 1b). However, these faces were contrasted with their first-order modified versions, created by shifting the position of the eyes and the mouth within the face (see Fig. 1c). These shifts involved an average of only 8.17% (SE 5 1) of the total number of pixels comprising the original faces. The similar magnitude of pixel change in this experiment and in experiment 1 (i.e., 7.97%) validates direct comparisons between these experiments. All stimuli were shown on a black background with a 1,024 768 definition. They subtended approximately 8 81 of visual angle. conditions would prevent overtraining with upright or inverted faces. All other aspects of the procedure were identical to experiment 1. The only difference was that phase 2 now used as negative stimuli the first-order modified versions of the faces instead of their ‘‘Thatcherized’’ versions. Results Scores were analyzed with a four-way group (upright, inverted) by block (1, 2 and 3) by phase (phase 1 vs. phase 2) by session (1–6) ANOVA. Baboons in the upright group performed similarly to those in the inverted group (mean upright: 84.3%, SD 5 2.6; mean inverted: 79.6%, SD 5 2.6), F(1,5) 5 .385, P 5 .56. A reliable main effect of session, F(5,25) 5 11.38, Po.001, indicated a performance increment. There was also a reliable effect of phase, F(1,5) 5 12.06, Po.05, indicating a better performance with normal faces (phase 1: mean 5 95.2% correct, SD 5 7.4%) than with firstorder modified faces (phase 2: mean phase 2 5 82% correct, SD 5 42.33%). Finally, interactions were not significant with the exception of phase (1 vs. 2) by session (1–6) two-ways interaction, F(5,25) 5 4.21, Po.05. As shown in Figure 3, repeated testing had greater effects on performance in phase 2 than in phase 1. Calculation of d0 s indicated that all baboons correctly discriminated the two stimuli in phase 1, from as early as the first test session (all d0 s were significant for that phase after a Bonferroni correction). Most d0 s (i.e., 34/42) were also significant in phase 2, considering the group or each individual baboon, indicating that the baboons readily discriminated the normal face from its first-order version. In fact, most of the nonreliable d0 s concerned the first sessions of phase 2, before the baboons had learned the discrimination. Response times on correct trials were analyzed following the same procedure as in experiment 1. The group (upright, inverted) by phase (1, 2), by block (1, 2, 3) ANOVA revealed reliable effects of Procedure Baboons tested in experiment 1 with upright faces were now tested with inverted faces, and vice versa. Although no group effect was obtained in experiment 1, it was considered that changing Fig. 3. Mean percentage of correct responses for each session and with the normal (phase 1) and ‘‘first-order modified’’ faces (phase 2) in experiment 2. Bars correspond to standard errors. Am. J. Primatol. 420 / Parron and Fagot both phase blocks; the former indicated longer mean response times in phase 2 (mean 5 786 ms, SD 5 140 ms) than in phase 1 (mean 5 1,030 ms, SD 5 159 ms; F(1,5) 5 18.06, Po.01). This finding confirmed accuracy results, suggesting that the face stimuli were more easily discriminated in phase 1. The main effect of block indicated decreasing mean response times with repeated testing (mean block 1 5 990 ms, SD 5 219 ms; mean block 2 5 885 ms, SD 5 207 ms; mean block 3 5 849 ms, SD 5 120 ms; F(1,5) 5 13.41, Po.001). None of the interactions reached significance, but the phase by block interaction approached significance (F(2,10) 5 3.83, P 5 .063). Inspection of response times suggests that the effect of block was more pronounced in phase 2 (mean block 1 5 1,136 ms; mean block 2 5 1,053 ms; mean block 3 5 900 ms) than in phase 1 (mean block 1 5 842 ms; mean block 2 5 717 ms; mean block 3 5 798 ms). This result is taken as additional evidence that the processing of first-order configural cues by baboons benefited from repeated testing. Discussion This experiment has shown that baboons are capable of discriminating first-order relational properties within a face, even after only a few trials. Comparison of experiments 1 and 2 moreover suggests that first-order spatial properties are more easily processed by baboons than second-order ones. However, it remains to be demonstrated that the different results in experiments 1 and 2 are not due to an effect of learning or test order. To assess this possibility, in experiment 3 all baboons were retested in some of the same conditions as in experiment 1. EXPERIMENT 3: ASSESSMENT OF LEARNING EFFECT Experiment 3 replicated the test condition of experiment 1, block 1. It was expected that performance in this test would reach that of experiment 2 using first-order pairs, if baboons have learned to process configural properties of the stimuli between experiments 1 and 2. By contrast, absence of learning effects between these two experiments should be indicated in experiment 3 by chance level performance with these stimuli, as already obtained in experiment 1. Stimuli and Procedure One block of trials was run for each individual in this experiment. This block was strictly identical to block 1 of experiment 1, and therefore used the second-order faces shown in Figure 1a and b. Results and Discussion All the baboons were comfortably above chance in phase 1, with the average performance ranging Am. J. Primatol. Fig. 4. Mean percentage of correct responses for each session, with the ‘‘Thatcherized’’ faces (phases 2 of experiments 1 and 3). Bars correspond to standard errors. from 93 to 97% correct, pooling over all sessions. Phase 2 data from the first block of experiment 2 (using a first-order modified face) were directly compared with the data obtained in phase 2 of experiment 3 (second-order modified face) by way of an experiment (experiments 2, 3) session (1–6) ANOVA. This analysis revealed a reliable effect of the experiment, F(1,6) 5 11.07; Po.05, corresponding to a better performance in experiment 2 (mean 5 76.31% correct) than in experiment 3 (mean 5 62.5%). The ANOVA also revealed a significant effect of session, F(5,30) 5 2.62; Po.05, but experiment by session interaction was also significant, as shown in Figure 4, (F(5,30) 5 3.49, Po.05). Response times were not analyzed due to the large number of errors limiting the data set size. Calculation of d0 s indicated that all baboons continued to fail to discriminate the original face from its ‘‘Thatcherized’’ version in experiment 3 (all d0 s below Bonferroni corrected significance level), in spite of intervening training in experiment 2. This final experiment therefore confirmed that baboons remain more capable of discriminating first-order facial properties than second-order properties, even accounting for learning effects. GENERAL DISCUSSION Experiment 1 showed that baboons did not experience the Thatcher illusion in a human-like manner. First, unlike humans [Rhodes et al., 1993], baboons behaved indistinctively when presented with upright and inverted faces. Second, while baboons readily discriminated the faces of two different individuals, they had serious difficulties discriminating normal faces from their ‘‘Thatcherized’’ versions, even when these faces were presented upright. In experiment 1, the lack of accurate discrimination in phase 2 cannot be due to a Configural Processing by Baboons / 421 misunderstanding of the task, as the baboons were successful in phase 1, which used the same general procedure. Similarly, poor discrimination in phase 2 cannot be explained by a lack of expertise or familiarity with the stimuli, as only pictures of familiar human or baboon models were shown. In humans at least, increased familiarity via prolonged exposure to specific stimuli enhances the encoding of second-order relational information [Diamond & Carey, 1986]. Our use of familiar models should have facilitated the processing of second-order relational properties, but this turned out to be very difficult for the baboons. In experiment 1, poor performance in phase 2 rather suggests that secondorder relational variations distinguishing the normal face from its ‘‘Thatcherized’’ version were not readily perceived by the baboons, which contrasts with reports in the human literature [e.g., Lewis & Johnston, 1997]. Experiment 2 provided additional information on the processing of configural cues by baboons. Accurate discrimination of normal faces from their first-order transformations was impressive and rapid. The baboons’ failure again to discriminate second-order pairs in experiment 3 rules out the possible effects of learning as an explanation for their discrimination of the first-order face pairs in experiment 2. Furthermore, the similar amount of pixels differing between the normal and the test faces (Thatcherized in experiment 1 and first-order transformed in experiment 2) rules out the possibility that discrimination of the first-order stimuli reflects a quantitatively inequitable alteration of the stimuli. Considered together, the results of experiments 1–3 strongly suggest that the baboons were much more skillful at discriminating first-order configural properties of the faces than their secondorder properties. It could also be proposed that our procedure facilitated a featural mode of processing in phase 1, and that this processing strategy led to a poor discrimination performance in phase 2 of experiment 1, when the two faces (i.e., the normal face and its Thatcherized version) had all their features in common but differed in configural second-order cues. Although this cannot entirely be ruled out, also given the use of a small set of very different images as stimuli, this does not account for the high levels of performance in phase 2 of experiment 2. Indeed, use of a featural mode of processing should have led to a poorer discrimination of the faces in that phase, but this did not occur. Our demonstration that nonhuman primates considered first-order relational properties in facial stimuli converges with earlier reports. Dittrich  trained longtailed macaques to recognize schematic line drawings of a macaque face depicting different expressions. He then tested post-learning recognition using jumbled faces in which the firstorder properties of the facial features were disorga- nized. Results revealed that recognition strongly depended on the integrity of the first-order relational properties. In another study, Myowa-Yamakoshi and Tomonaga  tested an infant gibbon’s discrimination of face-like and nonface-like drawings. The stimuli consisted of two kinds of schematic faces, one representing a normal configuration of a face and another showing a first-order modified version in which the configuration of the facial features was disordered (one eye above the mouth, and the nose above the other eye). The gibbon clearly preferred looking at normal schematic faces. Similar results— suggesting processing of first-order configural relations—were obtained in another experiment in the same study in which a real pictorial face and its scrambled first-order version were presented. Parr et al.  conducted three experiments on the contribution of first- and second-order configural cues in the processing of faces by chimpanzees. Their first and third experiments manipulated second-order cues by disaligning the upper or lower facial parts of the stimuli (experiment 1) or by blurring the facial features (experiment 3). Results suggested that second-order cues at least partly controlled face discrimination in these two tasks. Their second experiment was even more relevant for the current research because it assessed the respective contribution of first- and second-order cues during face discrimination. In that experiment, the inner features of a chimpanzee’s face were extracted and split into several subparts, with gaps between them, to disrupt the second-order relational properties of the faces. In another condition, these same facial features were moved apart and spatially rearranged, thus simultaneously disrupting first- and second-order relational properties of the faces. The overall performance of the chimpanzees was similar under these two conditions (second-order manipulation: 56% correct, first plus second-order manipulation: 59.2% correct).1 Lack of a significant difference along with the generally poor performance of the chimpanzees leave it unclear as to whether the chimpanzees prioritize first- or secondorder configural cues. Another interesting finding of this study is that baboons tested with upright or inverted stimuli provided identical results. It is tempting to propose that the orientation of the face is not as important in baboons as it is in humans. However, such an interpretation is not compelling, because baboons are quadrupedal, terrestrial primates, and conspecifics (or humans in the laboratory) are mostly likely to be perceived in a human-like canonical upright orientation. In addition, facial expressions are 1 Note that these means are different from those of Parr et al. [2006; Table 1]. This difference reflects a mistake in Parr et al.’s  Table 1 in which the means of the ‘‘split’’ (second-order) and ‘‘rearranged’’ (first-order) conditions were inverted. Am. J. Primatol. 422 / Parron and Fagot commonly used by baboons to convey social information [Altmann, 1980] and faces are thus ecologically relevant stimuli for them. We suggest caution in interpreting the lack of an inversion effect in our task. First, absence of an effect leaves open the possibility of a type II error. Second, our results were collected using a between- rather than a withinsubjects design, and moreover with a relatively small number of subjects, and therefore may have been influenced by intersubject variability. The possible effects of inversion require further testing. In closing, we would like to emphasize that our results do not imply that baboons are completely blind to second-order information available in faces. The processing of both first- and second-order configural properties may serve adaptive functions in natural settings. An analysis of first-order relational cues may permit the animal to quickly discriminate facial from nonfacial stimuli, and therefore potentially animals from nonanimals. Second-order cues may help to refine this initial rough analysis, and to individuate different conspecifics, or to infer their emotional state. Following this reasoning, we presume that baboons are capable of processing second-order relational cues, even though the ability has not been demonstrated here. This assumption is supported by previous research by our group showing that baboons can process second-order relational properties contained in nonfacial stimuli if the task forces them to do so [e.g., evaluate the distance between a line and a dot; Dépy et al., 1998]. The main interest of our findings, in this context, is that the baboons seem to prioritize first-order relational properties in absence of extended training. This conclusion is in line with the human developmental literature showing that infants or children process first-order cues well before second-order cues [e.g., Mondloch et al., 2003; Valenza et al., 1996]. ACKNOWLEDGMENTS The use and care of animals used in this research was fully approved by the ‘‘Comité d’Ethique Régional de la Région Provence Alpes Côtes d’Azur’’. All pictures employed for stimuli were used with the written consent of the people concerned. REFERENCES Altmann J. 1980. Baboon mothers and infants. Harvard University Press. Cambridge, Massachussetts. Am. J. Primatol. Bruce C. 1982. Face recognition by monkeys: absence of an inversion effect. Neuropsychologia 20:515–522. Bruce V, Doyle T, Dench N, Burton M. 1991. Remembering facial configurations. Cognition 38:109–144. Carey S, Diamond R. 1977. From piecemeal to configurational representation of faces. Science 195:312–314. Dépy D, Fagot J, Vauclair J. 1998. Comparative assessment of distance processing and hemispheric specialization in humans and baboons (Papio papio). Brain Cogn 38:165–182. Deruelle C, Fagot J. 2005. Categorizing facial identities, emotions, and genders: attention to high and low spatial frequencies by children and adults. J Exp Child Psychol 90:172–184. Deruelle C, Barbet I, Dépy D, Fagot J. 2000. Perception of partly occluded figures by baboons. Perception 29:1483–1497. Diamond R, Carey S. 1986. Why faces are and are not special: an effect of expertise. J Exp Psychol Gen 115:107–117. Dittrich W. 1990. Representation of faces in longtailed macaques (Macaca fascicularis). Ethology 85:265–278. Fobes JL, King JE. 1982. Primate behavior. New York: Academic Press. Goren CC, Sarty M, Wu P. 1975. Visual following and pattern discrimination of face-like stimuli by newborn infants. Pediatrics 56:544–549. Jitsumori M, Yoshihara M. 1997. Categorical discrimination of human facial expressions by pigeons: a test of the linear feature model. Q J Exp Psychol B 50:253–268. Lewis MB, Johnston RA. 1997. The Thatcher illusion as a test of configural disruption. Perception 26:225–227. Maurer D, Le Grand R, Mondloch CJ. 2002. The many faces of configural processing. Trends Cogn Sci 6:255–260. Mondloch CJ, Geldart S, Maurer D, Le Grand R. 2003. Developmental changes in face processing skills. J Exp Child Psychol 86:67–84. Myowa-Yamakoshi M, Tomonaga M. 2001. Development of face recognition in an infant gibbon. Inf Behav Dev 24: 215–227. Parr LA, Dove T, Hopkins D. 1998. Why faces may be special: evidence of the inversion effect in chimpanzees. J Cogn Neurosci 10:615–622. Parr LA, Heintz M, Akamagwuna U. 2006. Three studies on configural face processing by chimpanzees. Brain Cogn 62: 30–42. Rhodes G, Brake S, Atkinson AP. 1993. What’s lost in inverted faces? Cognition 47:25–57. Rosenfeld SA, Van Hoesen GW. 1979. Face recognition in the rhesus monkey. Neuropsychologia 17:503–509. Thompson P. 1980. Margaret Thatcher: a new illusion. Perception 9:483–484. Tomonaga M. 1994. How laboratory-raised Japanese monkeys (Macaca fuscata) perceive rotated photographs of monkeys: evidence for an inversion effect in face perception. Primates 35:55–165. Tomonaga M, Itakura S, Matsuzawa T. 1993. Superiority of conspecific faces and reduced inversion effect in face perception by a chimpanzee. Folia Primatol 61:110–114. Valenza E, Simion F, Cassia MV, Umilta C. 1996. Face preference at birth. J Exp Psychol Hum Percept Perform 22:892–903.