Humans and Macaques Employ Similar Face-Processing Strategies

Share Embed


Descrição do Produto

Current Biology 19, 509–513, March 24, 2009 ª2009 Elsevier Ltd All rights reserved

DOI 10.1016/j.cub.2009.01.061

Report Humans and Macaques Employ Similar Face-Processing Strategies Christoph D. Dahl,1,4,* Christian Wallraven,2,4 Heinrich H. Bu¨lthoff,2 and Nikos K. Logothetis1,3 1Department of Physiology of Cognitive Processes 2Department of Human Perception, Cognition, and Action Max Planck Institute for Biological Cybernetics 72076 Tu¨bingen Germany 3Division of Imaging Science and Biomedical Engineering University of Manchester Manchester, M13 9PL United Kingdom

Summary Primates developed the ability to recognize and individuate their conspecifics by the face. Despite numerous electrophysiological studies in monkeys [1–3], little is known about the face-processing strategies that monkeys employ. In contrast, face perception in humans has been the subject of many studies [4–6] providing evidence for specific face processing that evolves with perceptual expertise [7]. Importantly, humans process faces holistically, here defined as the processing of faces as wholes, rather than as collections of independent features (part-based processing) [8]. The question remains to what extent humans and monkeys share these face-processing mechanisms. By using the same experimental design and stimuli for both monkey and human behavioral experiments, we show that face processing is influenced by the species affiliation of the observed face stimulus (human versus macaque face). Furthermore, stimulus manipulations that selectively reduced holistic and part-based information systematically altered eye-scanning patterns for human and macaque observers similarly. These results demonstrate the similar nature of face perception in humans and monkeys and pin down effects of expert faceprocessing versus novice face-processing strategies. These findings therefore directly contribute to one of the central discussions in the behavioral and neurosciences about how faces are perceived in primates. Results Human Face Perception Twelve adult humans performed a nonreinforced, passive viewing task (see Figure 1 and Experimental Procedures) in which eye movements were recorded. Face stimuli of neutral facial expressions of rhesus macaques and humans were used in an original (upright) presentation and two image manipulations (inverted and blurred) (Figure 1). These manipulations were chosen to selectively disrupt face-processing strategies based on prior human perceptual studies [4, 9]. Holistic face processing develops as our perceptual expertise with faces grows and is characterized by fast and parallel

*Correspondence: [email protected] 4These authors contributed equally to this work

processing of faces, whereas part-based processing is a much slower and more serial process requiring attention to details in the face. More specifically, holistic processing is disrupted by inversion (i.e., turning a face upside down will lead to part-based processing), whereas part-based processing is disrupted by blurring (i.e., reducing high spatial frequency information will leave the holistic percept intact). Additionally, rhesus macaque faces are a stimulus class for which humans have not developed perceptual expertise and therefore should not elicit holistic processing in general [7]. Finally, prior studies have shown that for humans, the eyes play a crucial role in face processing [10–12]. In eye-tracking studies, upright faces elicited less lower face and less random part scanning than did inverted faces [13]. We can therefore take preference for the eye region as an indicator of holistic processing and expect eye-movement patterns to be modulated by the different conditions depending on the faceprocessing strategy afforded by the stimulus. Participants were asked to look at the images as naturally as possible for a total of 12 s of trial duration. We analyzed the viewing time during image presentation and determined the saliency for the variables species and facial parts (eyes, nose, and mouth) as well as for all three manipulation conditions (upright, inverted, blurred). Here, we report statistics involving viewing times (statistical comparisons of number of fixations are fully compatible with viewing times) that showed significant main effects and interactions for our experimental variables. All post-hoc tests were corrected for multiple comparisons. To examine effects of perceptual expertise, viewing times to macaque and human faces were compared. In upright faces, the viewing time was longer for eyes than for nose and mouth when human participants watched human faces (eyes versus nose: F(1,22) = 8.44, p < 0.01; eyes versus mouth: F(1,22) = 19.53, p < 0.001) compared to when they watched macaque faces (eyes versus nose: F(1,22) = 9.37, p < 0.01, while nose > eyes; nose versus mouth: F(1,22) = 19.53, p < 0.001, while nose > mouth) (Figure 2). This indicates a preference for eyes over nose and mouth of human faces, and a higher saliency for human eyes than macaque eyes, respectively. This result is further supported by directly comparing eyes of human and monkey faces, revealing that eyes were looked at significantly longer in human than in macaque faces (F(1,22) = 11.45, p < 0.01). In contrast to upright faces, face inversion led to a drastic loss of eye preference in human faces. Comparisons between facial parts confirm that both human (eyes versus nose: F(1,22) = 0.10, p = 0.75; eyes versus mouth: F(1,22) = 1.12, p = 0.30; nose versus mouth: F(1,22) = 1.18, p = 0.29) and macaque (eyes versus nose: F(1,22) = 1.91, p = 0.18; eyes versus mouth: F(1,22) = 0.28, p = 0.60; nose versus mouth: F(1,22) = 0.54, p = 0.47) faces were treated equally because of inversion. Finally, when reducing part-based information processing by blurring the faces, an upright-like pattern of response was observed: eyes were visited more often than nose (F(1,22) = 7.50, p < 0.01) and mouth (F(1,22) = 28.35, p < 0.001) in human, but not in macaque (eyes versus nose: F(1,22) = 0.26, p = 0.62; eyes versus mouth: F(1,22) = 2.07, p = 0.16) faces, also supported by direct comparisons of eye regions across species (F(1,22) = 17.69, p < 0.001).

Current Biology Vol 19 No 6 510

A

Figure 1. Example Stimuli and Experimental Design (A) Examples of face images. For human and macaque faces the original faces (upright) were manipulated by inversion (inverted) and blurring (blurred). (B) Participants elicited an image or blank square, alternately, by directing gaze toward the monitor, and terminated a stimulus by looking away. After 12 s of cumulative stimulus display time, a new image was displayed.

B

The current experiment not only supports the findings that eyes are more salient than nose and mouth regions, it also illustrates a systematic modulation of the eye preference resulting from inversion and blurring manipulations. Moreover, humans show a qualitatively different processing during the presentation of conspecific and nonconspecific faces, indicating an effect of perceptual expertise on eye-scanning strategies. Macaque Face Perception Three adult male rhesus macaques (Macaca mulatta) performed the identical passive viewing task by using the same stimuli as in the human experiment. We analyzed the total viewing time during image presentation and determined the saliency for two experimental variables (species, facial parts) as well as for all three manipulation conditions (upright, inverted, blurred). Again, corrected post-hoc tests were applied to the significant effects. First, to determine the impact of perceptual expertise, viewing times to macaque and human faces were compared. In upright faces, the viewing time was longer for eyes than for nose and mouth when the macaques watched macaque faces (eyes versus nose: F(1,70) = 19.50, p < 0.001; eyes versus mouth: F(1,70) = 26.19, p < 0.001) (Figure 2). However, human faces did not elicit a preference for eyes over nose and mouth (eyes versus nose: F(1,70) = 1.00, p = 0.32; eyes versus mouth:

F(1,70) = 0.07, p = 0.80) as macaque faces did, suggesting a qualitative difference with respect to species. Direct comparisons of viewing time of the eye region for macaque and human stimuli determined that the saliency of eyes differed across species (F(1,70) = 9.60, p < 0.01). In contrast, inverting the faces, and thus reducing holistic processing, led to a drastic loss of eye preference in macaque faces (eyes versus nose: F(1,74) = 4.11, p < 0.05, while nose > eyes; eyes versus mouth: F(1,74) = 1.82, p = 0.18) as well as in human faces (eyes versus nose: F(1,74) = 1.07, p = 0.31, eyes versus mouth: F(1,74) = 0.67, p = 0.41). Thus, human and macaque faces were treated equally when inverted. Blurring faces, however, and thus reducing access to part-based information, elicited a nearly identical pattern of responses as upright faces: eyes were visited longer than the nose (eyes versus nose: F(1,74) = 4.38, p < 0.05) and mouth (eyes versus mouth: F(1,74) = 19.74, p < 0.001) regions in macaque faces. However, this pattern was not observed for blurred human faces (eyes versus nose: F(1,74) = 0.07, p = 0.78; eyes versus mouth: F(1,74) = 1.55, p = 0.21). Overall, the eyes were more salient in macaque than in human faces (F(1,74) = 6.38, p < 0.05). A few studies have shown that macaques perceive conspecific individuals differently than nonconspecifics. A dishabituation study [14] that used looking time demonstrated with whole-bodied images that macaques perceive their conspecifics on a different categorical level (subordinate level entry point) than nonconspecifics (basic level). Along this line, it has been shown recently that, like humans, rhesus macaques individuate conspecific faces but not nonface category exemplars or nonconspecific faces [15]. Thus, there is indication of a qualitatively different perceptual processing during the presentation of conspecific and nonconspecific faces. Although previous research has provided qualitative evidence for species-specific processing, our results show, for the first time, a clear dissociation in the behavioral characteristics in the macaque’s scan pattern when observing macaque and human faces. Even though the macaque’s visual system is tuned to the same facial features for both macaque and human faces [16], it employs a different oculomotor strategy to inspect macaque and human faces.

Similar Face Processing in Humans and Macaques 511

Figure 2. Viewing Time for Human and Macaque Participants Plotted are viewing times while observers looked at faces of both species (legend) in three different appearance manipulations (upright, inverted, blurred). Viewing times of single parts (eyes, nose, mouth) were normalized to the total viewing time in a trial. The proportion of the area of each facial part relative to the whole image was subtracted from the proportion of data samples in each corresponding facial part and the total number of samples in that trial. It follows that a proportion of viewing time of zero is equivalent with randomly looking at the image or a facial part.

Discussion In the present study, we have shown that the distribution of eye movements in monkeys (Macaca mulatta) and humans is critically affected by whether conspecific versus nonconspecific faces are shown. Additionally, stimulus manipulations (such as blurring and inversion) resulted in similar systematic modulations of eye-scanning patterns for both species. In our view, these findings clearly demonstrate the effect of perceptual expertise, that is, the fact that macaques are experts for macaque faces and that humans are experts for human faces. Even though the same faces were presented to the two participant groups, the difference in eye-movement patterns with respect to the species affiliation of the faces was immense (see Figure 3). There is converging evidence in macaques [15, 17, 18] as well as in humans [13] that eyes are looked at more frequently than any other facial part when faces are presented in a natural (upright) way. This saliency is not due to low-level appearance, but driven by higher-level expectations based on the spatial configuration of the face [19]. Additionally, a high proportion of eye fixations is, to some extent, indicative of holistic face processing [15]. These findings allowed us to directly link the observed changes in gaze behavior under the different image manipulations to face-processing strategies. Because both humans and macaques have access to holistic processing strategies as a result of a high degree of expertise with their own species, the saliency of eyes was maintained for upright and blurred conspecific faces. Conversely, eyes did not receive as much attention for upright and blurred nonconspecific faces, as a result of a relatively low degree of expertise. Furthermore, because holistic processing was not available for inverted conspecific faces, the eyes of inverted faces appeared less salient relative to other facial parts. Moreover, although blurring reduces the information of facial parts, or high-spatial frequency components, it still allows for detection of spatial

relationships between facial parts and therefore does not fully disrupt holistic processing abilities. Accordingly, during the presentation of blurred conspecific faces, the interest in the eyes was enhanced similarly to natural (upright) conditions. Our findings are therefore consistent with previous assumptions of expertise for conspecific, but not for nonconspecific, faces in macaques [15] as well as for non-face-object categories (such as cars or dogs) in humans [20]. Unlike the numerous studies on face perception in humans [5, 6, 21–24], insights into the behavioral abilities of macaque face perception have come almost exclusively from the study of the face inversion effect [25–27]. Most of these studies used explicit reinforcement for some type of discrimination, resulting in idiosyncratic response strategies and sometimes contradictory results, leaving the nature of face inversion in the monkey unclear [15]. To avoid these response strategies in monkeys, we used a nonreinforced paradigm that enabled monkeys to act as naturally as possible. Additionally, it has been shown that face processing relies on a sensitive initial time window that is critical for the identification or classification of a face [28]. Our experimental paradigm, conceptually adapted from Humphrey [14], consists of a passive viewing task that allows the participant to actively set on- and offset of a stimulus via eye gaze ([15], see Experimental Procedures for more details). This not only overcomes a potential effect of habituation while participants view the faces, but also maintains their initial spontaneous viewing strategy, allowing the reproduction of the sensitive time window multiple times during a trial. In our view, this property is crucial in enabling us to use gaze behavior as a sensitive tool to investigate face-processing strategies. In a recent study [16], eye movements of macaques passively viewing faces of different species showed no species-specific effect in contrast to our results. The critical difference lies in the task: by averaging only over the first few fixations in our data, we were not able to find the same modulations in viewing strategies for conspecific versus nonconspecific faces. This clear difference emerged only when averaging over the multiple initial time windows afforded by our task (see Supplemental Data available online). Finally, it remains to be seen how our results generalize to different task contexts. In [15], misaligned and aligned faces were used in a passive viewing task similar to the one used here, showing a strong effect on gaze behavior, whereas

Current Biology Vol 19 No 6 512

Figure 3. Example Stimulus Overlaid by Normalized Fixation Densities Total fixation densities for both participants (macaque and human) while observing conspecific and nonconspecific faces in three stimulus manipulations (upright, inverted, blurred). For reference, fixation densities are superimposed on gray-scale versions of one stimulus exemplar per condition to visualize the effects of face viewing time on species and stimulus manipulations. Fixation densities were spatially normalized to variations of facial part differences across all faces in a particular condition. As described in Figure 2, eye fixations are higher for conspecific faces in upright and blurred presentations than for conspecific face in inverted presentation, as well as for nonconspecific faces in all presentation conditions.

a recent study with an old-new recognition task [29] failed to show effects of face orientation or part alignment on gaze behavior in humans. In light of the clear effects in both species in our study, however, we believe that these discrepancies are due either to instruction-related top-down influences, or to the fact that passive viewing will focus more on the encoding and learning stages of face processing than on the recognition stages as in [29]. In conclusion, our study has shown that macaques possess perceptual expertise for conspecific faces in analogy to that of humans. The observed saliency of the eye region is strongly driven by context, not by structural properties, because the same pixels are interpreted differently by a macaque than by a human. Our study has thus provided clear evidence that both primates have evolved to become perceptual experts in conspecific face processing. Experimental Procedures Three male rhesus macaques (Macaca mulatta, 5–7 years old, 10–13 kg) and 12 human participants (7 females, 20–33 years old) were used in this study. All experiments were conducted in accordance with the guidelines of the

European Community (EU VD 86/609/EEC) for the care and use of laboratory animals under the approval of local authorities (Regierungspraesidium Tuebingen). Informed consent was obtained from all human participants. In total, 40 digital color pictures of neutral rhesus macaque and human faces were used in these experiments. Faces were cut out from their original background, normalized for luminance, and placed on a mid-gray background creating an image of 300 3 300 pixels (13.3 degrees of visual angle). Moreover, three different conditions were generated by keeping the original images unchanged (upright), by inverting the images (inverted), or by blurring the upright images (blurred) via a two-dimensional Gaussian smoothing kernel with a sigma of .035 of image width in frequency space (Figure 1). We used a mid-gray blank square as well as a gray outline marking a frame of the same size as the face stimulus. Eye movements of the macaques were recorded by an iView infrared eyetracking system (SensoMotoric Instruments (SMI), Teltow/Berlin, Germany) sampled at 200 Hz. Human eye movements were recorded by an iView X Hi-Speed infrared eye tracking system sampled at 500 Hz. Stimuli were presented with custom-written software, controlled by the QNX real-time operating system (QNX Software Systems, Ontario, Canada). Macaques were calibrated at the beginning of the session with a 9-point fixation task. During the experiment, juice reward was given during an intertrial interval, regardless of behavior, similar to the procedure described by Humphrey [14]. Humans were calibrated prior to every trial to minimize spatial distortions resulting from head movements. They were financially compensated at the end of the experiment.

Similar Face Processing in Humans and Macaques 513

Participants could actively control the on- and off-set of the stimuli by entering and leaving the central image frame, thus allowing us to reproduce the initial time window for face perception multiple times during a trial (see Figure 1). The order of trials was intermixed such that no more than three consecutive trials showed pictures of the same species. For the macaque experiments, we used five stimulus sets with given predetermined stimulus order. The macaques did 144 upright, 152 inverted, and 152 blurred trials, split up into 8 to 10 days of experimental testing per macaque and condition. Statistics were calculated across sessions. For humans, we used two stimulus sets with given predetermined stimulus order. The human participants did 120 trials of each condition, and statistics were obtained across participants. The total number of fixations and the viewing time were determined. Fixation periods were extracted as a function of velocity, including eye movement samples that were not faster than 20 deg/s within a time period of at least 100 ms. The average position of samples containing one fixation period was taken as the final eye position of that fixation period. To statistically evaluate the fixation frequency and density of single facial parts (eyes, nose, and mouth), number of fixations and viewing time of single parts were normalized to the total number of fixations and the viewing time in that trial. Furthermore, the proportion of the area of a particular facial part relative to the whole image was subtracted from the proportion of data samples in a particular facial part and the total number of samples in that trial. Any difference in viewing time from zero means that this particular facial area was looked at more or less than predicted by a uniform looking strategy.

Supplemental Data Supplemental Data include Supplemental Experimental Procedures and one figure and can be found with this article online at http://www. current-biology.com/supplemental/S0960-9822(09)00679-4. Acknowledgments We thank Christopher I. Petkov, Isabelle Bu¨lthoff, Regine Armann, and three anonymous reviewers for helpful comments on the manuscript. This work was supported by the Max Planck Society. The authors have declared that no financial conflicts of interest exist. Received: October 21, 2008 Revised: January 29, 2009 Accepted: January 29, 2009 Published online: February 26, 2009 References 1. Desimone, R., Albright, T.D., Gross, C.G., and Bruce, C. (1984). Stimulus-selective properties of inferior temporal neurons in the macaque. J. Neurosci. 4, 2051–2062. 2. Gross, C.G., Rocha-Miranda, C.E., and Bender, D.B. (1972). Visual properties of neurons in inferotemporal cortex of the Macaque. J. Neurophysiol. 35, 96–111. 3. Perrett, D.I., Rolls, E.T., and Caan, W. (1982). Visual neurones responsive to faces in the monkey temporal cortex. Exp. Brain Res. 47, 329–342. 4. Maurer, D., Grand, R.L., and Mondloch, C.J. (2002). The many faces of configural processing. Trends Cogn. Sci. 6, 255–260. 5. Tanaka, J.W., and Farah, M.J. (1993). Parts and wholes in face recognition. Q. J. Exp. Psychol. A 46, 225–245. 6. Young, A.W., Hellawell, D., and Hay, D.C. (1987). Configurational information in face perception. Perception 16, 747–759. 7. Schwarzer, G., Zauner, N., and Jovanovic, B. (2007). Evidence of a shift from featural to configural face processing in infancy. Dev. Sci. 10, 452–463. 8. Le Grand, R., Mondloch, C.J., Maurer, D., and Brent, H.P. (2004). Impairment in holistic face processing following early visual deprivation. Psychol. Sci. 15, 762–768. 9. Schwaninger, A., Lobmaier, J., Collishaw, S., and Wallraven, C. (2009). Two routes to face perception: evidence from psychophysics and computational modeling. Cogn. Sci., in press. 10. Emery, N.J. (2000). The eyes have it: the neuroethology, function and evolution of social gaze. Neurosci. Biobehav. Rev. 24, 581–604.

11. Farroni, T., Csibra, G., Simion, F., and Johnson, M.H. (2002). Eye contact detection in humans from birth. Proc. Natl. Acad. Sci. USA 99, 9602–9605. 12. Bentin, S., Allison, T., Puce, A., Perez, E., and McCarthy, G. (1996). Electrophysiological studies of face perception in humans. J. Cogn. Neurosci. 8, 551–565. 13. Barton, J.J., Radcliffe, N., Cherkasova, M.V., Edelman, J., and Intriligator, J.M. (2006). Information processing during face recognition: the effects of familiarity, inversion, and morphing on scanning fixations. Perception 35, 1089–1105. 14. Humphrey, N.K. (1974). Species and individuals in the perceptual world of monkeys. Perception 3, 105–114. 15. Dahl, C.D., Logothetis, N.K., and Hoffman, K.L. (2007). Individuation and holistic processing of faces in rhesus monkeys. Proc. Biol. Sci. 274, 2069–2076. 16. Guo, K., Robertson, R.G., Mahmoodi, S., Tadmor, Y., and Young, M.P. (2003). How do monkeys view faces? A study of eye movements. Exp. Brain Res. 150, 363–374. 17. Keating, C.F., and Keating, E.G. (1982). Visual scan patterns of rhesus monkeys viewing faces. Perception 11, 211–219. 18. Nahm, F.K.D., Perret, A., Amaral, D.G., and Albright, T.D. (1997). How do monkeys look at faces? J. Cogn. Neurosci. 9, 611–623. 19. Guo, K. (2007). Initial fixation placement in face images is driven by topdown guidance. Exp. Brain Res. 181, 673–677. 20. Tanaka, J.W., and Taylor, M. (1991). Object categories and expertise: is the basic level in the eye of the beholder? Cognit. Psychol. 23, 457–482. 21. Yin, R.K. (1969). Looking at upside-down faces. J. Exp. Psychol. 81, 141–145. 22. De Renzi, E., and Saetti, M.C. (1997). Associative agnosia and optic aphasia: qualitative or quantitative difference? Cortex 33, 115–130. 23. Haxby, J.V., Hoffman, E.A., and Gobbini, M.I. (2000). The distributed human neural system for face perception. Trends Cogn. Sci. 4, 223–233. 24. Kanwisher, N., McDermott, J., and Chun, M.M. (1997). The fusiform face area: a module in human extrastriate cortex specialized for face perception. J. Neurosci. 17, 4302–4311. 25. Tomonaga, M. (1994). How laboratory-raised Japanese monkeys (Macaca fuscata) perceive rotated photographs of monkeys: evidence of an inversion effect in face perception. Primates 35, 155–165. 26. Rosenfeld, S.A., and Van Hoesen, G.W. (1979). Face recognition in the rhesus monkey. Neuropsychologia 17, 503–509. 27. Perrett, D.I., Mistlin, A.J., Chitty, A.J., Smith, P.A., Potter, D.D., Broennimann, R., and Harries, M. (1988). Specialized face processing and hemispheric asymmetry in man and monkey: evidence from single unit and reaction time studies. Behav. Brain Res. 29, 245–258. 28. Rossion, B., Gauthier, I., Tarr, M.J., Despland, P., Bruyer, R., Linotte, S., and Crommelinck, M. (2000). The N170 occipito-temporal component is delayed and enhanced to inverted faces but not to inverted objects: an electrophysiological account of face-specific processes in the human brain. Neuroreport 11, 69–74. 29. de Heering, A., Rossion, B., Turati, C., and Simion, F. (2008). Holistic face processing can be independent of gaze behavior: evidence from the face composite effect. J. Neuropsychol. 2, 183–195.

Lihat lebih banyak...

Comentários

Copyright © 2017 DADOSPDF Inc.