- Email: [email protected]
The positive and negative of human expertise in gaze perception
P. Ricciardelli et al. / Cognition 77 (2000) B1±B14
B1
COGNITION Cognition 77 (2000) B1±B14
www.elsevier.com/locate/cognit
Brief article
The positive and negative of human expertise in gaze perception Paola Ricciardelli a,*, Gordon Baylis b, Jon Driver a a
Institute of Cognitive Neuroscience, University College London, 17 Queen Square, London WC1N 3AR, UK b Department of Psychology, University of South Carolina, Columbia, SC, USA Received 1 April 2000; accepted 26 May 2000
Abstract Judging where others look is crucial for many social and cognitive functions. Past accounts of gaze perception emphasize geometrical cues from the seen eye. Human eyes have a unique morphology, with a large white surround (sclera) to the dark iris that may have evolved to enhance gaze processing. Here we show that the contrast polarity of seen eyes has a powerful in¯uence on gaze perception. Adult observers are highly inaccurate in judging gaze direction for images of human eyes with negative contrast polarity (regardless of whether the surrounding face is positive or negative), even though negative images of eyes preserve the geometric properties of positives that are judged accurately. The detrimental effect of negative contrast polarity is much larger for gaze perception than for other directional judgements (e.g. judging which way a head is turned). These results suggest an `expert' system for gaze perception, which always treats the darker region of a seen eye as the part that does the looking. q 2000 Elsevier Science B.V. All rights reserved. Keywords: Gaze; Gaze perception; Social attention
1. Introduction In everyday life, where people look can signal where they are currently attending. Accordingly, many social and cognitive functions depend on the ability to perceive where another person is looking (see Baron-Cohen, 1995). The gaze direction of others can signal sources of possible interest, or of imminent danger. Gaze direction * Corresponding author. Tel.: 144-20-7679-1126; fax: 144-20-7916-8517. E-mail address: [email protected] (P. Ricciardelli). 0010-0277/00/$ - see front matter q 2000 Elsevier Science B.V. All rights reserved. PII: S 0010-027 7(00)00092-5
B2
P. Ricciardelli et al. / Cognition 77 (2000) B1±B14
is also used to regulate many social interactions, ranging from threatening exchanges to turn-taking in polite conversation (e.g. Argyle & Cook, 1976). Gaze perception is also a key element of `joint attention', which allows separate individuals to focus on a common external referent (e.g. Dunham & Moore, 1995). This facilitates communication, and can even enable the initial acquisition of words during language acquisition (e.g. Bruner, 1983). Gaze perception offers a unique window into others' minds, and is thought to be abnormal in pathological conditions such as autism (e.g. see Baron-Cohen, 1995; Frith, 1989). Classic studies show that normal observers are very sensitive to seen gaze direction (e.g. Anstis, Mayhew & Morley, 1969; Cline, 1967; Gibson & Pick, 1963; Ehrlich & Field, 1993). To date, all accounts of this sensitivity have emphasized the purely geometrical cues that are available in seen eyes, concerning the form and shape of the eye region, in which the circular iris deviates towards the direction of gaze. The typical account is that observers judge the position and extent of the iris relative to that of the surrounding sclera (e.g. Anstis et al., 1969); see Langton, Watt and Bruce (2000) for a more contemporary example of such an account, and Ando and Osaka (1998) for related evidence. Human eyes have a unique morphology (Kobayashi & Kohshima, 1997), comprising a widely exposed white sclera that contrasts with the darker iris. This high contrast in luminance is more pronounced than in other primates, as is the wide spatial extent of the contrasting white sclera. Indeed, it has been suggested (e.g. Kobayashi & Kohshima, 1997) that this unique morphology may have evolved to facilitate intraspeci®c gaze perception in humans, and thus to facilitate joint attention in our highly social species. It has often been noted (e.g. Gibson & Pick, 1963) that the high contrast between sclera and iris should facilitate the extraction of those geometric cues that are traditionally considered as essential for perceiving gaze direction. However, the possibility that the actual polarity of this high contrast between iris and sclera may also be critical has never been explicitly considered before. Here we examined whether the perception of seen gaze direction can still proceed accurately for negative images of human eyes. Such images preserve all the geometric properties (and spatial frequencies, etc.) of positive images, but they have the crucial difference that the sclera now appears darker than the iris. If the visual system follows an invariant rule during gaze perception, that the darker part of a seen eye is the part which does the looking, then gaze direction judgements should become inaccurate for negative images of eyes. But if the visual system depends only on the geometric properties of seen eyes, as emphasized in previous accounts, then the usual high accuracy of gaze direction judgements should remain even for negative images. To anticipate, we found that gaze judgements are extraordinarily inaccurate for negative images of eyes. 2. Experiment 1 The ®rst experiment tested whether subjects would make signi®cantly more
P. Ricciardelli et al. / Cognition 77 (2000) B1±B14
B3
mistakes in judging the direction of seen gaze for eyes with a negative contrast polarity. In a three-choice task, people were asked to indicate how they perceived the direction of gaze in a computerized image of a face, i.e. as looking towards their right, their left, or straight at them. For completeness in this initial study, we examined these three possible gaze directions fully crossed with three possible head orientations. 2.1. Method 2.1.1. Subjects The eight subjects (seven female and one male, ranging from 21 to 35 years in age) were paid volunteers. 2.1.2. Stimuli The stimuli comprised different monochrome pictures of the same person (the `looker') gazing at different positions (308 left, 308 right, or straight ahead). Independently of this, the looker's head could face straight, or to the left by 308, or to the right by 308. In half the displays, the two-tone eyes had normal `positive' contrast (that is, white sclera and black iris; see example in Fig. 1a); in the other half, the same pictures now had negative contrast for just the eye region (i.e. black sclera and white iris; see Fig. 1b). These manipulations were effected in Adobe Photoshop. The displays were approximately life-size and viewed from 70 cm on a computer monitor. 2.1.3. Design The two possible eye contrasts (positive or negative) were crossed with the possible gaze directions (left, right, or centre) and possible head directions (left, right or centre) in a 2 £ 3 £ 3 within-subject design. See Fig. 1 for examples of the various stimulus types produced by the full crossing of these factors. 2.1.4. Experimental procedure Prior to the experiment, one example of each condition was shown together with a speci®cation of its correct response. Each subject was then presented with 384 trials divided into six equal blocks, with the 18 conditions being equiprobable in a random sequence within each block. Subjects were informed that some of the gaze stimuli might look rather unusual, but that they were simply to indicate the perceived direction of gaze. On each trial, a central asterisk appeared as a warning signal and ®xation point at the centre of the screen for 448 ms. This was followed by a static image of the looker gazing towards one of the three possible different positions (while facing towards any one of these). This image lasted until response, or for a maximum of 1112 ms. After response, a blank delay of 500 ms preceded the ®xation stimulus for the next trial.
B4
P. Ricciardelli et al. / Cognition 77 (2000) B1±B14
Fig. 1. Examples of stimuli from Experiment 1. (a) Positive eyes, with head and gaze both directed 308 to the observer's right; (b) same as in (a), but with the eyes now in negative polarity. All the stimuli used comprised a full-face picture (as in (a,b)), but the illustrations in (c±j) show just the region around the eyes for brevity. Examples in the left column have positive eyes, while those in the right column have negative eyes: (c,d) eyes-left with the head facing right; (e,f) eyes-straight with the head facing right; (g,h) straight eyes in a straight head; (i,j) eyes-left in a straight face. Left±right re¯ections of the stimulus types shown were also possible for all cases except (g,h).
P. Ricciardelli et al. / Cognition 77 (2000) B1±B14
B5
2.2. Results and discussion The results (see Fig. 2) showed a large in¯uence of contrast polarity, with worse performance for negative eye stimuli (52.3% correct overall versus 93.5% for positives). The data were arcsin transformed and subjected to a three-way repeated measures ANOVA. A main effect of contrast polarity (F 1; 7 28:8, P , 0:001) showed worse performance for negatives overall. The three-way interaction between polarity, gaze direction and head direction was also signi®cant (F 4; 28 3:3, P , 0:05), and had two sources. First, negatives impaired performance for every case except direct gaze in a straight face (central data points in the graph of Fig. 2); note that only in this exceptional case can gaze direction be judged by the symmetry of the image alone (see Fig. 1g,h). Second, negatives produced the largest impairments when the eyes and head were both deviated in the same direction (outer data points in Fig. 2), with many erroneous `straight' responses in this negative condition. The most important result is the much worse performance with negative than positive eyes. Planned comparisons con®rmed that the negative conditions were signi®cantly worse (P , 0:01 or better) than the corresponding positive condition for every case except direct gaze/straight head. In a follow-up study with eight new subjects, we con®rmed that this same pattern of results is found when images of
Fig. 2. Accuracy results for Experiment 1 with standard errors. Filled symbols are for positive-eyes conditions, and open symbols are for negative-eyes conditions.
B6
P. Ricciardelli et al. / Cognition 77 (2000) B1±B14
several different people (®ve women and ®ve men) are used as stimuli. In this follow-up, the effect of contrast polarity was reliable not only across subjects (F 1; 7 27:5, P , 0:001), but also across materials (i.e. the different people used in the photographs; F 1; 9 52:8, P , 0:001). 3. Experiment 2 The next study examined whether the dif®culty of negative eyes in Experiment 1 might have been due to the fact that they always appeared embedded in positive images of faces. Perhaps gaze perception is dif®cult only when the polarity of the eyes does not match that of the surrounding face. To test this, we again used positive and negative gaze stimuli, but now presented each within either positive (see Fig. 1a,b) or negative faces (see Fig. 3a,b). We could thus determine whether performance depends simply on the sign of contrast polarity for the eyes themselves, or instead on whether the polarity of the eyes matches the polarity of the surrounding face. As testing this was the sole purpose of the present experiment, we simpli®ed other aspects of the design, using less permutations of combined eye and head directions, and simplifying the task to a two-choice procedure (is the pictured person looking left or right?). 3.1. Methods 3.1.1. Subjects Eight new subjects (four female and four male) were drawn from a similar age range as Experiment 1. 3.1.2. Experimental design and procedure There were three within-subject factors. First, the eyes in each stimulus had positive or negative contrast polarity. Second, the head faced left or right, and the eyes either gazed in the same direction (`congruent' conditions) or the opposite direction (`incongruent'). Finally, the surrounding face was either shown in positive or negative polarity. These three factors were fully crossed, and the task was now a two-choice (left versus right) gaze direction judgement. The experimental procedure was otherwise as before. 3.2. Results and discussion A three-way ANOVA showed a main effect of eye polarity (F 1; 7 26:7, P , 0:001), but no reliable in¯uence of congruency, and no interaction (see Fig. 4). Thus, gaze judgements were worse for negative than positive eyes, regardless of the polarity of the surrounding face. 4. Experiment 3 The next study examined whether making eye stimuli `unusual' in any respect can
P. Ricciardelli et al. / Cognition 77 (2000) B1±B14
B7
Fig. 3. Example stimuli from the follow-up experiments. (a,b) Positive-eyes and negative-eyes stimuli, respectively, within a surrounding negative face context, as used in Experiment 2. In all experiments, the stimuli comprised a full-face picture (as in (a,b)), but the illustrations in (c±f) show just the region around the eyes for brevity. Examples in the left column have positive eyes, while those in the right column have negative eyes. (c,d) Red-and-green eyes, as used in Experiment 3. (e,f) Eyes with the `highlight' on the iris removed (see arrowed region, and compare with Fig. 1).
impair gaze perception, or whether the substantial impairments observed in the previous study are speci®c to inversions of contrast polarity. We again compared eye stimuli that had positive or negative contrast polarity. However, we now used `unusual' colouring for these eyes in some conditions for both the negative and positive polarity eyes (see Fig. 3c,d). The new stimuli were generated by replacing the black-and-white of the previous eye stimuli with dark-red and light-green, respectively. People do not see red-andgreen eyes in daily life (!), although they may often encounter positive black-andwhite eye stimuli (e.g. in newspaper photographs of faces). If the negative eyes in Experiments 1 and 2 were judged poorly simply because they are `unusual' in some very general sense (as compared with the positive stimuli from those studies), then performance might now become equally poor for positive and negative eyes in the
B8
P. Ricciardelli et al. / Cognition 77 (2000) B1±B14
Fig. 4. Accuracy results for Experiment 2 with standard errors. Filled symbols are for positive-eyes conditions, and open symbols are for negative-eyes conditions.
present study, since both were `unusual' combinations of red and green. But if the positive±negative difference depends speci®cally on contrast polarity instead, then it should be replicated even for the red-and-green stimuli (given that the red was darker than the green). Finally, if gaze perception depends not only upon the sign of contrast polarity for iris versus sclera, but also on the absolute difference in luminance between these two regions of the eye, then the effect of contrast polarity might be somewhat reduced for the red-and-green stimuli. 4.1. Methods 4.1.1. Subjects Eight new subjects (four female and four male) were drawn from a similar age range as Experiment 1. 4.1.2. Experimental design and procedure The design and procedure were as for Experiment 2, except that the factor of eye colour (i.e. black-and-white versus red-and-green eyes) now replaced the previous factor of positive versus negative surrounding faces. All faces were shown in positive grey-scale.
P. Ricciardelli et al. / Cognition 77 (2000) B1±B14
B9
4.2. Results and discussion A three-way ANOVA showed a main effect of eye polarity (F 1; 7 42:6, P , 0:001), with more errors for negative eyes again (see Fig. 5). There was a signi®cant effect of eye colour (F 1; 7 27:2, P , 0:001), with judgements of red-and-green eyes being somewhat more accurate. The interaction of eye polarity and eye colour was also signi®cant (F 1; 7 33:5, P , 0:001), with colour affecting judgements only for negative eyes (F 1; 7 53:9, P , 0:001) and not for positive eyes (F 1; 7 0:95, NS). Although the effect of contrast was somewhat reduced for red-and-green eyes (see Fig. 5), it nevertheless remained highly signi®cant for them alone (F 1; 7 35:9, P , 0:001). Thus, judgements of gaze direction can remain extremely accurate even for highly `unusual' red-and-green eyes, provided that the iris remains darker than the surrounding sclera. The slight reduction in the polarity effect for strangely coloured eyes suggests that this effect is somewhat stronger with more extreme differences in luminance between bright and dark regions (as for black-and-white versus red-andgreen).
Fig. 5. Accuracy results for Experiment 3 with standard errors. Filled symbols are for positive-eyes conditions, and open symbols are for negative-eyes conditions.
B10
P. Ricciardelli et al. / Cognition 77 (2000) B1±B14
5. Experiment 4 The ®nal study tested whether inverting contrast polarity disrupts judgements of gaze direction more than other directional judgements, and speci®cally, more than for judgements of the direction in which a seen head is facing. In a pilot study, we used face stimuli just like those from the previous studies, presenting the whole image either in negative polarity (e.g. Fig. 3a) or in positive polarity (e.g. Fig. 1a), to test whether negative polarity would impair judgements of whether the seen head was turned 308 to the right, to the left, or was facing the viewer. In fact there was no such in¯uence of polarity; head orientation judgements were better than 90% accurate in all conditions. However, one might argue that the head direction judgement was simply too easy to be affected by polarity (although note that gaze judgements are also very `easy' and highly accurate for positive stimuli). We followed this up by using a more dif®cult head orientation discrimination (58 to left or right of centre), and by testing many subjects in this task, to assess whether any impact of contrast polarity whatsoever could be found in a more demanding head orientation task that should bring performance off ceiling to provide a sensitive measure. 5.1. Methods 5.1.1. Subjects Thirty-four new subjects (22 female and 12 male) were again drawn from a similar age range. 5.1.2. Experimental stimuli, design and procedure The stimuli again each comprised a pictured face, with the head now facing straight at the viewer, 58 to the left, or 58 to the right. Gaze was now always in the same direction that the head faced. The three-choice task was to judge whether the head faced left, straight or right. Each subject underwent ®ve blocks of 120 trials, with the six possible stimulus types (head facing left, right or straight; all crossed with positive versus negative polarity for the whole image) being equally likely in each block. 5.2. Results and discussion A two-way ANOVA (head direction £ polarity) showed no effect of polarity (F 1; 33 0:002), a signi®cant effect of head direction (F 2; 66 25:2, P , 0:001, with best performance for straight heads), and no interaction (F 2; 66 1:4, NS). In contrast to the dramatic effects of contrast polarity on gaze direction judgements, head direction judgements were unin¯uenced by contrast polarity (see Fig. 6), even when many subjects were tested in a demanding head orientation task.
P. Ricciardelli et al. / Cognition 77 (2000) B1±B14
B11
Fig. 6. Accuracy results for Experiment 4 with standard errors. Filled symbols are for positive faces, and open symbols are for negative faces.
6. General discussion Our experiments show that perception of gaze direction is dramatically impaired for eyes seen in negative contrast polarity. This applies regardless of whether the surrounding face is in positive or negative polarity; it holds across images of different people, and applies even for eyes shown with bizarre colour schemes. This effect of polarity arises even though negatives share all the `geometric' properties of positive eyes which have been emphasized in previous accounts of gaze perception (e.g. Anstis et al., 1969; Cline, 1967; Gibson & Pick, 1963). The dif®culty of perceiving the direction of gaze for negative eyes remains even for sophisticated observers (as readers can con®rm for themselves by inspection of Figs. 1 and 3). This effect on gaze perception cannot simply be reduced to previous known in¯uences on face processing. Although at ®rst glance it may appear reminiscent of previous ®ndings that negative images of faces are harder to recognize as known individuals than positive images (e.g. Galper, 1970; Philips, 1972), in fact the latter face effect cannot explain the present gaze effect. All the faced used here were unknown to the subjects, and no face recognition was required by the task. Moreover, the effect on gaze perception remains even when just the eyes are shown alone (as can be con®rmed by suitable inspection of Figs. 1 and 3). Furthermore, the present gaze effect is found regardless of the polarity of the surrounding face (see Experiment 2). Finally, the dif®culty of recognizing familiar faces when shown in
B12
P. Ricciardelli et al. / Cognition 77 (2000) B1±B14
negative polarity is commonly attributed to a disruptive effect on the interpretation of shadow cues to the 3D structure of a face (e.g. Kemp, Pike, White & Musselman, 1996). There were some lines immediately surrounding the eye in the images used here, which may have been due to small shadows, and which reversed from black to white with the change in contrast polarity (see Figs. 1 and 3). Nevertheless, it seems highly unlikely that these lines provided 3D cues used to judge gaze direction. We propose that the effect of contrast polarity on gaze processing arises because the visual system follows an in¯exible contrast rule for gaze perception, invariably treating the dark part of the eye image as the part that does the looking. 1 Evidently this `rule' cannot be overridden for negative stimuli, even though the geometry of the image is just the same as for the positives which are accurately perceived (so that, in principle, negatives might be judged just as accurately based on these geometric cues). Even if the surrounding face is also presented in negative contrast, this extremely obvious cue evidently cannot be used to modify the usual gaze perception rule. The great dif®culty with negatives thus suggests the involvement of a dedicated `expert' system, applying an obligatory rule in the processing of gaze stimuli. `Expertise' is implied here in a similar sense to that often inferred from the effects of inversion on face processing. Inversion disrupts recognition more for faces than for other classes of object (e.g. Yin, 1969). Analogously, reversal of contrast polarity disrupts directional judgements more for eyes than for other classes of stimuli (e.g. judgements of head orientation are unaffected; Experiment 4). Furthermore, just as evidence from neuroscience and neuropsychology has documented the existence of specialized neural systems involved in the processing of faces (e.g. Gross, RochaMiranda & Bender, 1972; Kanwisher, McDermott & Chun, 1997; Perrett & Mistlin, 1990), so there is now evidence for such specialization in the processing of gaze, within somewhat different neural areas (e.g. Hoffman & Haxby, 2000; Perrett et al., 1990). There has long been controversy over whether specialized processing of faces is pre-programmed genetically, or is the consequence of acquired `expertise' during extensive exposure to faces, or re¯ects some speci®c combination of nature and nurture (e.g. Diamond & Carey, 1986; Johnson & Morton, 1991; Gauthier, Skudlarski, Gore & Anderson, 2000). Similar issues arise for the contrast polarity speci®city uncovered here for gaze perception. Since even young babies are highly sensitive to gaze direction (at least in `positive' stimuli; e.g. Hood, Willen & Driver, 1998; Maurer, 1985), developmental work with the stimuli introduced here could 1 It could perhaps be suggested that the `specular highlight' often present in seen eyes might become treated as a small iris or pupil when becoming dark in a negative contrast polarity eye stimulus, and thus could be erroneously treated as the `part that does the looking'. To test this directly, we ran a further study with the same design as Experiment 2, but with a manipulation of the presence versus absence of specular highlights now replacing the previous factor of face polarity (see Fig. 3e,f, noting the arrowed region where the previous highlight has been removed; compare with Fig. 1c,d). All faces were shown in positive grey-scale. The results from eight new subjects showed the usual strong effect of polarity (F 1; 7 37:1, P , 0:001) but no effect of presence/absence of the highlight (F 1; 7 0:4), nor any interaction of highlight with polarity (F 1; 7 0:1). Thus, the specular highlight on the eye is not responsible for the observed dif®culty in gaze perception with negative eyes.
P. Ricciardelli et al. / Cognition 77 (2000) B1±B14
B13
reveal whether the contrast rule re¯ects learned or innate expertise in gaze processing. Our stimuli could also be used to test whether contrast-speci®c expertise in gaze perception is lacking in individuals who exhibit (or go on to show) dysfunctional social attention, as in autism (e.g. Baron-Cohen, 1995; Frith, 1989). Acknowledgements P.R. and J.D. were supported by European TMR grant ERBFMBICT971928, and by the Human Frontiers Science Program; G.B. was supported by grant SBR 9616555 from the National Science Foundation, and by intramural funding from the University of South Carolina. Thanks to Charlotte Russell for being a looker and to Steffan Kennett for his help in data analysis. References Ando, S., & Osaka, N. (1998). Blood shot illusion: luminance affects perceived gaze direction. Investigative Ophthalmology & Visual Science, 39, s/72. Anstis, S. M., Mayhew, J. W., & Morley, T. (1969). The perception of where a Tv portrait is looking. American Journal of Psychology, 82, 474±489. Argyle, M., & Cook, M. (1976). Gaze and mutual gaze. Cambridge: Cambridge University Press. Baron-Cohen, 1995. Mindblindness: an essay on autism and theory of mind. Cambridge: MIT Press. Bruner, J. (1983). Child's talk: learning to use language. New Tork: Norton. Cline, M. G. (1967). The perception of where a person is looking. American Journal of Psychology, 80, 41±50. Diamond, R., & Carey, S. (1986). Why faces are and are not special: an effect of expertise. Journal of Experimental Psychology: General, 115, 107±117. Dunham, P. J., & Moore, C. (1995). Themes in research on joint attention. In C. Moore, & P. J. Dunham (Eds.), Joint attention: its origins and role in development. Hillsdale, NJ: Erlbaum. Ehrlich, S., & Field, D. (1993). Why is the acuity for the direction of gaze so good? Investigative Ophthalmology & Visual Science, 34 (4), 778. Frith, U. (1989). Autism: explaining the enigma. Oxford: Blackwell. Galper, R. E. (1970). Recognition of faces in photographic negative. Psychonomic Science, 19, 207±208. Gauthier, I., Skudlarski, P., Gore, J. C., & Anderson, A.W. (2000). Expertise for cars and birds recruits brain areas involved in face recognition. Nature Neuroscience, 3 (2), 191±197. Gibson, J. J., & Pick, A. D. (1963). Perception of another person's looking behaviour. American Journal of Psychology, 76, 386±394. Gross, C. G., Rocha-Miranda, C. E., & Bender, D. B. (1972). Visual properties of neurons in inferotemporal cortex of the macaque. Journal of Neurophysiology, 35 (1), 96±111. Hoffman, E. A., & Haxby, J. V. (2000). Distinct representations of eye gaze and identity in the distributed human neural system for face perception. Nature Neuroscience, 3 (1), 80±84. Hood, B. M., Willen, J. D., & Driver, J. (1998). Adult's eyes trigger shifts of visual attention in human infants. Psychological Science, 9 (2), 131±134. Johnson, M. H., & Morton, J. (1991). Biology and cognitive development: the case of face recognition. Oxford: Blackwell. Kanwisher, N., McDermott, J., & Chun, M. M. (1997). The fusiform face area: a module in human extrastriate cortex specialized for face perception. The Journal of Neuroscience, 17 (11), 4302±4311. Kemp, R., Pike, G., White, P., & Musselman, A. (1996). Perception and recognition of normal and negative faces: the role of shape from shading and pigmentation cues. Perception, 25, 37±52. Kobayashi, H., & Kohshima, S. (1997). Unique morphology of the human eye. Nature, 387, 767±768.
B14
P. Ricciardelli et al. / Cognition 77 (2000) B1±B14
Langton, S.R.H., Watt, R.J., & Bruce, V. (2000). Do the eyes have it? Cues to the direction of social attention. Trends in Cognitive Sciences. 4 (2), 50±59. Maurer, D. (1985). Infant's perception of faceness. In T. Field, & M. Fox (Eds.), Social perception in infants. Norwood, NJ: Ablex. Perrett, D., & Mistlin, A. (1990). Perception of facial characteristics by monkeys. In W. Stebbins, & M. Berkely (Eds.), Comparative perception (Vol. 2). New York: Wiley. Perrett, D., Harries, M., Mistlin, A., Hietanen, J., Benson, P., Bevan, R., Thomas, S., Oram, M., Ortega, J., & Brierley, K. (1990). Social signals analysed at the single cell level: someone is looking at me, something touched me, something moved! International Journal of Comparative Psychology, 4, 25± 55. Philips, R. J. (1972). Why are faces hard to recognize in photographic negative? Perception and Psychophysics, 12, 425±426. Yin, R. K. (1969). Looking at upside-down faces. Journal of Experimental Psychology, 81, 141±145.
B1
COGNITION Cognition 77 (2000) B1±B14
www.elsevier.com/locate/cognit
Brief article
The positive and negative of human expertise in gaze perception Paola Ricciardelli a,*, Gordon Baylis b, Jon Driver a a
Institute of Cognitive Neuroscience, University College London, 17 Queen Square, London WC1N 3AR, UK b Department of Psychology, University of South Carolina, Columbia, SC, USA Received 1 April 2000; accepted 26 May 2000
Abstract Judging where others look is crucial for many social and cognitive functions. Past accounts of gaze perception emphasize geometrical cues from the seen eye. Human eyes have a unique morphology, with a large white surround (sclera) to the dark iris that may have evolved to enhance gaze processing. Here we show that the contrast polarity of seen eyes has a powerful in¯uence on gaze perception. Adult observers are highly inaccurate in judging gaze direction for images of human eyes with negative contrast polarity (regardless of whether the surrounding face is positive or negative), even though negative images of eyes preserve the geometric properties of positives that are judged accurately. The detrimental effect of negative contrast polarity is much larger for gaze perception than for other directional judgements (e.g. judging which way a head is turned). These results suggest an `expert' system for gaze perception, which always treats the darker region of a seen eye as the part that does the looking. q 2000 Elsevier Science B.V. All rights reserved. Keywords: Gaze; Gaze perception; Social attention
1. Introduction In everyday life, where people look can signal where they are currently attending. Accordingly, many social and cognitive functions depend on the ability to perceive where another person is looking (see Baron-Cohen, 1995). The gaze direction of others can signal sources of possible interest, or of imminent danger. Gaze direction * Corresponding author. Tel.: 144-20-7679-1126; fax: 144-20-7916-8517. E-mail address: [email protected] (P. Ricciardelli). 0010-0277/00/$ - see front matter q 2000 Elsevier Science B.V. All rights reserved. PII: S 0010-027 7(00)00092-5
B2
P. Ricciardelli et al. / Cognition 77 (2000) B1±B14
is also used to regulate many social interactions, ranging from threatening exchanges to turn-taking in polite conversation (e.g. Argyle & Cook, 1976). Gaze perception is also a key element of `joint attention', which allows separate individuals to focus on a common external referent (e.g. Dunham & Moore, 1995). This facilitates communication, and can even enable the initial acquisition of words during language acquisition (e.g. Bruner, 1983). Gaze perception offers a unique window into others' minds, and is thought to be abnormal in pathological conditions such as autism (e.g. see Baron-Cohen, 1995; Frith, 1989). Classic studies show that normal observers are very sensitive to seen gaze direction (e.g. Anstis, Mayhew & Morley, 1969; Cline, 1967; Gibson & Pick, 1963; Ehrlich & Field, 1993). To date, all accounts of this sensitivity have emphasized the purely geometrical cues that are available in seen eyes, concerning the form and shape of the eye region, in which the circular iris deviates towards the direction of gaze. The typical account is that observers judge the position and extent of the iris relative to that of the surrounding sclera (e.g. Anstis et al., 1969); see Langton, Watt and Bruce (2000) for a more contemporary example of such an account, and Ando and Osaka (1998) for related evidence. Human eyes have a unique morphology (Kobayashi & Kohshima, 1997), comprising a widely exposed white sclera that contrasts with the darker iris. This high contrast in luminance is more pronounced than in other primates, as is the wide spatial extent of the contrasting white sclera. Indeed, it has been suggested (e.g. Kobayashi & Kohshima, 1997) that this unique morphology may have evolved to facilitate intraspeci®c gaze perception in humans, and thus to facilitate joint attention in our highly social species. It has often been noted (e.g. Gibson & Pick, 1963) that the high contrast between sclera and iris should facilitate the extraction of those geometric cues that are traditionally considered as essential for perceiving gaze direction. However, the possibility that the actual polarity of this high contrast between iris and sclera may also be critical has never been explicitly considered before. Here we examined whether the perception of seen gaze direction can still proceed accurately for negative images of human eyes. Such images preserve all the geometric properties (and spatial frequencies, etc.) of positive images, but they have the crucial difference that the sclera now appears darker than the iris. If the visual system follows an invariant rule during gaze perception, that the darker part of a seen eye is the part which does the looking, then gaze direction judgements should become inaccurate for negative images of eyes. But if the visual system depends only on the geometric properties of seen eyes, as emphasized in previous accounts, then the usual high accuracy of gaze direction judgements should remain even for negative images. To anticipate, we found that gaze judgements are extraordinarily inaccurate for negative images of eyes. 2. Experiment 1 The ®rst experiment tested whether subjects would make signi®cantly more
P. Ricciardelli et al. / Cognition 77 (2000) B1±B14
B3
mistakes in judging the direction of seen gaze for eyes with a negative contrast polarity. In a three-choice task, people were asked to indicate how they perceived the direction of gaze in a computerized image of a face, i.e. as looking towards their right, their left, or straight at them. For completeness in this initial study, we examined these three possible gaze directions fully crossed with three possible head orientations. 2.1. Method 2.1.1. Subjects The eight subjects (seven female and one male, ranging from 21 to 35 years in age) were paid volunteers. 2.1.2. Stimuli The stimuli comprised different monochrome pictures of the same person (the `looker') gazing at different positions (308 left, 308 right, or straight ahead). Independently of this, the looker's head could face straight, or to the left by 308, or to the right by 308. In half the displays, the two-tone eyes had normal `positive' contrast (that is, white sclera and black iris; see example in Fig. 1a); in the other half, the same pictures now had negative contrast for just the eye region (i.e. black sclera and white iris; see Fig. 1b). These manipulations were effected in Adobe Photoshop. The displays were approximately life-size and viewed from 70 cm on a computer monitor. 2.1.3. Design The two possible eye contrasts (positive or negative) were crossed with the possible gaze directions (left, right, or centre) and possible head directions (left, right or centre) in a 2 £ 3 £ 3 within-subject design. See Fig. 1 for examples of the various stimulus types produced by the full crossing of these factors. 2.1.4. Experimental procedure Prior to the experiment, one example of each condition was shown together with a speci®cation of its correct response. Each subject was then presented with 384 trials divided into six equal blocks, with the 18 conditions being equiprobable in a random sequence within each block. Subjects were informed that some of the gaze stimuli might look rather unusual, but that they were simply to indicate the perceived direction of gaze. On each trial, a central asterisk appeared as a warning signal and ®xation point at the centre of the screen for 448 ms. This was followed by a static image of the looker gazing towards one of the three possible different positions (while facing towards any one of these). This image lasted until response, or for a maximum of 1112 ms. After response, a blank delay of 500 ms preceded the ®xation stimulus for the next trial.
B4
P. Ricciardelli et al. / Cognition 77 (2000) B1±B14
Fig. 1. Examples of stimuli from Experiment 1. (a) Positive eyes, with head and gaze both directed 308 to the observer's right; (b) same as in (a), but with the eyes now in negative polarity. All the stimuli used comprised a full-face picture (as in (a,b)), but the illustrations in (c±j) show just the region around the eyes for brevity. Examples in the left column have positive eyes, while those in the right column have negative eyes: (c,d) eyes-left with the head facing right; (e,f) eyes-straight with the head facing right; (g,h) straight eyes in a straight head; (i,j) eyes-left in a straight face. Left±right re¯ections of the stimulus types shown were also possible for all cases except (g,h).
P. Ricciardelli et al. / Cognition 77 (2000) B1±B14
B5
2.2. Results and discussion The results (see Fig. 2) showed a large in¯uence of contrast polarity, with worse performance for negative eye stimuli (52.3% correct overall versus 93.5% for positives). The data were arcsin transformed and subjected to a three-way repeated measures ANOVA. A main effect of contrast polarity (F 1; 7 28:8, P , 0:001) showed worse performance for negatives overall. The three-way interaction between polarity, gaze direction and head direction was also signi®cant (F 4; 28 3:3, P , 0:05), and had two sources. First, negatives impaired performance for every case except direct gaze in a straight face (central data points in the graph of Fig. 2); note that only in this exceptional case can gaze direction be judged by the symmetry of the image alone (see Fig. 1g,h). Second, negatives produced the largest impairments when the eyes and head were both deviated in the same direction (outer data points in Fig. 2), with many erroneous `straight' responses in this negative condition. The most important result is the much worse performance with negative than positive eyes. Planned comparisons con®rmed that the negative conditions were signi®cantly worse (P , 0:01 or better) than the corresponding positive condition for every case except direct gaze/straight head. In a follow-up study with eight new subjects, we con®rmed that this same pattern of results is found when images of
Fig. 2. Accuracy results for Experiment 1 with standard errors. Filled symbols are for positive-eyes conditions, and open symbols are for negative-eyes conditions.
B6
P. Ricciardelli et al. / Cognition 77 (2000) B1±B14
several different people (®ve women and ®ve men) are used as stimuli. In this follow-up, the effect of contrast polarity was reliable not only across subjects (F 1; 7 27:5, P , 0:001), but also across materials (i.e. the different people used in the photographs; F 1; 9 52:8, P , 0:001). 3. Experiment 2 The next study examined whether the dif®culty of negative eyes in Experiment 1 might have been due to the fact that they always appeared embedded in positive images of faces. Perhaps gaze perception is dif®cult only when the polarity of the eyes does not match that of the surrounding face. To test this, we again used positive and negative gaze stimuli, but now presented each within either positive (see Fig. 1a,b) or negative faces (see Fig. 3a,b). We could thus determine whether performance depends simply on the sign of contrast polarity for the eyes themselves, or instead on whether the polarity of the eyes matches the polarity of the surrounding face. As testing this was the sole purpose of the present experiment, we simpli®ed other aspects of the design, using less permutations of combined eye and head directions, and simplifying the task to a two-choice procedure (is the pictured person looking left or right?). 3.1. Methods 3.1.1. Subjects Eight new subjects (four female and four male) were drawn from a similar age range as Experiment 1. 3.1.2. Experimental design and procedure There were three within-subject factors. First, the eyes in each stimulus had positive or negative contrast polarity. Second, the head faced left or right, and the eyes either gazed in the same direction (`congruent' conditions) or the opposite direction (`incongruent'). Finally, the surrounding face was either shown in positive or negative polarity. These three factors were fully crossed, and the task was now a two-choice (left versus right) gaze direction judgement. The experimental procedure was otherwise as before. 3.2. Results and discussion A three-way ANOVA showed a main effect of eye polarity (F 1; 7 26:7, P , 0:001), but no reliable in¯uence of congruency, and no interaction (see Fig. 4). Thus, gaze judgements were worse for negative than positive eyes, regardless of the polarity of the surrounding face. 4. Experiment 3 The next study examined whether making eye stimuli `unusual' in any respect can
P. Ricciardelli et al. / Cognition 77 (2000) B1±B14
B7
Fig. 3. Example stimuli from the follow-up experiments. (a,b) Positive-eyes and negative-eyes stimuli, respectively, within a surrounding negative face context, as used in Experiment 2. In all experiments, the stimuli comprised a full-face picture (as in (a,b)), but the illustrations in (c±f) show just the region around the eyes for brevity. Examples in the left column have positive eyes, while those in the right column have negative eyes. (c,d) Red-and-green eyes, as used in Experiment 3. (e,f) Eyes with the `highlight' on the iris removed (see arrowed region, and compare with Fig. 1).
impair gaze perception, or whether the substantial impairments observed in the previous study are speci®c to inversions of contrast polarity. We again compared eye stimuli that had positive or negative contrast polarity. However, we now used `unusual' colouring for these eyes in some conditions for both the negative and positive polarity eyes (see Fig. 3c,d). The new stimuli were generated by replacing the black-and-white of the previous eye stimuli with dark-red and light-green, respectively. People do not see red-andgreen eyes in daily life (!), although they may often encounter positive black-andwhite eye stimuli (e.g. in newspaper photographs of faces). If the negative eyes in Experiments 1 and 2 were judged poorly simply because they are `unusual' in some very general sense (as compared with the positive stimuli from those studies), then performance might now become equally poor for positive and negative eyes in the
B8
P. Ricciardelli et al. / Cognition 77 (2000) B1±B14
Fig. 4. Accuracy results for Experiment 2 with standard errors. Filled symbols are for positive-eyes conditions, and open symbols are for negative-eyes conditions.
present study, since both were `unusual' combinations of red and green. But if the positive±negative difference depends speci®cally on contrast polarity instead, then it should be replicated even for the red-and-green stimuli (given that the red was darker than the green). Finally, if gaze perception depends not only upon the sign of contrast polarity for iris versus sclera, but also on the absolute difference in luminance between these two regions of the eye, then the effect of contrast polarity might be somewhat reduced for the red-and-green stimuli. 4.1. Methods 4.1.1. Subjects Eight new subjects (four female and four male) were drawn from a similar age range as Experiment 1. 4.1.2. Experimental design and procedure The design and procedure were as for Experiment 2, except that the factor of eye colour (i.e. black-and-white versus red-and-green eyes) now replaced the previous factor of positive versus negative surrounding faces. All faces were shown in positive grey-scale.
P. Ricciardelli et al. / Cognition 77 (2000) B1±B14
B9
4.2. Results and discussion A three-way ANOVA showed a main effect of eye polarity (F 1; 7 42:6, P , 0:001), with more errors for negative eyes again (see Fig. 5). There was a signi®cant effect of eye colour (F 1; 7 27:2, P , 0:001), with judgements of red-and-green eyes being somewhat more accurate. The interaction of eye polarity and eye colour was also signi®cant (F 1; 7 33:5, P , 0:001), with colour affecting judgements only for negative eyes (F 1; 7 53:9, P , 0:001) and not for positive eyes (F 1; 7 0:95, NS). Although the effect of contrast was somewhat reduced for red-and-green eyes (see Fig. 5), it nevertheless remained highly signi®cant for them alone (F 1; 7 35:9, P , 0:001). Thus, judgements of gaze direction can remain extremely accurate even for highly `unusual' red-and-green eyes, provided that the iris remains darker than the surrounding sclera. The slight reduction in the polarity effect for strangely coloured eyes suggests that this effect is somewhat stronger with more extreme differences in luminance between bright and dark regions (as for black-and-white versus red-andgreen).
Fig. 5. Accuracy results for Experiment 3 with standard errors. Filled symbols are for positive-eyes conditions, and open symbols are for negative-eyes conditions.
B10
P. Ricciardelli et al. / Cognition 77 (2000) B1±B14
5. Experiment 4 The ®nal study tested whether inverting contrast polarity disrupts judgements of gaze direction more than other directional judgements, and speci®cally, more than for judgements of the direction in which a seen head is facing. In a pilot study, we used face stimuli just like those from the previous studies, presenting the whole image either in negative polarity (e.g. Fig. 3a) or in positive polarity (e.g. Fig. 1a), to test whether negative polarity would impair judgements of whether the seen head was turned 308 to the right, to the left, or was facing the viewer. In fact there was no such in¯uence of polarity; head orientation judgements were better than 90% accurate in all conditions. However, one might argue that the head direction judgement was simply too easy to be affected by polarity (although note that gaze judgements are also very `easy' and highly accurate for positive stimuli). We followed this up by using a more dif®cult head orientation discrimination (58 to left or right of centre), and by testing many subjects in this task, to assess whether any impact of contrast polarity whatsoever could be found in a more demanding head orientation task that should bring performance off ceiling to provide a sensitive measure. 5.1. Methods 5.1.1. Subjects Thirty-four new subjects (22 female and 12 male) were again drawn from a similar age range. 5.1.2. Experimental stimuli, design and procedure The stimuli again each comprised a pictured face, with the head now facing straight at the viewer, 58 to the left, or 58 to the right. Gaze was now always in the same direction that the head faced. The three-choice task was to judge whether the head faced left, straight or right. Each subject underwent ®ve blocks of 120 trials, with the six possible stimulus types (head facing left, right or straight; all crossed with positive versus negative polarity for the whole image) being equally likely in each block. 5.2. Results and discussion A two-way ANOVA (head direction £ polarity) showed no effect of polarity (F 1; 33 0:002), a signi®cant effect of head direction (F 2; 66 25:2, P , 0:001, with best performance for straight heads), and no interaction (F 2; 66 1:4, NS). In contrast to the dramatic effects of contrast polarity on gaze direction judgements, head direction judgements were unin¯uenced by contrast polarity (see Fig. 6), even when many subjects were tested in a demanding head orientation task.
P. Ricciardelli et al. / Cognition 77 (2000) B1±B14
B11
Fig. 6. Accuracy results for Experiment 4 with standard errors. Filled symbols are for positive faces, and open symbols are for negative faces.
6. General discussion Our experiments show that perception of gaze direction is dramatically impaired for eyes seen in negative contrast polarity. This applies regardless of whether the surrounding face is in positive or negative polarity; it holds across images of different people, and applies even for eyes shown with bizarre colour schemes. This effect of polarity arises even though negatives share all the `geometric' properties of positive eyes which have been emphasized in previous accounts of gaze perception (e.g. Anstis et al., 1969; Cline, 1967; Gibson & Pick, 1963). The dif®culty of perceiving the direction of gaze for negative eyes remains even for sophisticated observers (as readers can con®rm for themselves by inspection of Figs. 1 and 3). This effect on gaze perception cannot simply be reduced to previous known in¯uences on face processing. Although at ®rst glance it may appear reminiscent of previous ®ndings that negative images of faces are harder to recognize as known individuals than positive images (e.g. Galper, 1970; Philips, 1972), in fact the latter face effect cannot explain the present gaze effect. All the faced used here were unknown to the subjects, and no face recognition was required by the task. Moreover, the effect on gaze perception remains even when just the eyes are shown alone (as can be con®rmed by suitable inspection of Figs. 1 and 3). Furthermore, the present gaze effect is found regardless of the polarity of the surrounding face (see Experiment 2). Finally, the dif®culty of recognizing familiar faces when shown in
B12
P. Ricciardelli et al. / Cognition 77 (2000) B1±B14
negative polarity is commonly attributed to a disruptive effect on the interpretation of shadow cues to the 3D structure of a face (e.g. Kemp, Pike, White & Musselman, 1996). There were some lines immediately surrounding the eye in the images used here, which may have been due to small shadows, and which reversed from black to white with the change in contrast polarity (see Figs. 1 and 3). Nevertheless, it seems highly unlikely that these lines provided 3D cues used to judge gaze direction. We propose that the effect of contrast polarity on gaze processing arises because the visual system follows an in¯exible contrast rule for gaze perception, invariably treating the dark part of the eye image as the part that does the looking. 1 Evidently this `rule' cannot be overridden for negative stimuli, even though the geometry of the image is just the same as for the positives which are accurately perceived (so that, in principle, negatives might be judged just as accurately based on these geometric cues). Even if the surrounding face is also presented in negative contrast, this extremely obvious cue evidently cannot be used to modify the usual gaze perception rule. The great dif®culty with negatives thus suggests the involvement of a dedicated `expert' system, applying an obligatory rule in the processing of gaze stimuli. `Expertise' is implied here in a similar sense to that often inferred from the effects of inversion on face processing. Inversion disrupts recognition more for faces than for other classes of object (e.g. Yin, 1969). Analogously, reversal of contrast polarity disrupts directional judgements more for eyes than for other classes of stimuli (e.g. judgements of head orientation are unaffected; Experiment 4). Furthermore, just as evidence from neuroscience and neuropsychology has documented the existence of specialized neural systems involved in the processing of faces (e.g. Gross, RochaMiranda & Bender, 1972; Kanwisher, McDermott & Chun, 1997; Perrett & Mistlin, 1990), so there is now evidence for such specialization in the processing of gaze, within somewhat different neural areas (e.g. Hoffman & Haxby, 2000; Perrett et al., 1990). There has long been controversy over whether specialized processing of faces is pre-programmed genetically, or is the consequence of acquired `expertise' during extensive exposure to faces, or re¯ects some speci®c combination of nature and nurture (e.g. Diamond & Carey, 1986; Johnson & Morton, 1991; Gauthier, Skudlarski, Gore & Anderson, 2000). Similar issues arise for the contrast polarity speci®city uncovered here for gaze perception. Since even young babies are highly sensitive to gaze direction (at least in `positive' stimuli; e.g. Hood, Willen & Driver, 1998; Maurer, 1985), developmental work with the stimuli introduced here could 1 It could perhaps be suggested that the `specular highlight' often present in seen eyes might become treated as a small iris or pupil when becoming dark in a negative contrast polarity eye stimulus, and thus could be erroneously treated as the `part that does the looking'. To test this directly, we ran a further study with the same design as Experiment 2, but with a manipulation of the presence versus absence of specular highlights now replacing the previous factor of face polarity (see Fig. 3e,f, noting the arrowed region where the previous highlight has been removed; compare with Fig. 1c,d). All faces were shown in positive grey-scale. The results from eight new subjects showed the usual strong effect of polarity (F 1; 7 37:1, P , 0:001) but no effect of presence/absence of the highlight (F 1; 7 0:4), nor any interaction of highlight with polarity (F 1; 7 0:1). Thus, the specular highlight on the eye is not responsible for the observed dif®culty in gaze perception with negative eyes.
P. Ricciardelli et al. / Cognition 77 (2000) B1±B14
B13
reveal whether the contrast rule re¯ects learned or innate expertise in gaze processing. Our stimuli could also be used to test whether contrast-speci®c expertise in gaze perception is lacking in individuals who exhibit (or go on to show) dysfunctional social attention, as in autism (e.g. Baron-Cohen, 1995; Frith, 1989). Acknowledgements P.R. and J.D. were supported by European TMR grant ERBFMBICT971928, and by the Human Frontiers Science Program; G.B. was supported by grant SBR 9616555 from the National Science Foundation, and by intramural funding from the University of South Carolina. Thanks to Charlotte Russell for being a looker and to Steffan Kennett for his help in data analysis. References Ando, S., & Osaka, N. (1998). Blood shot illusion: luminance affects perceived gaze direction. Investigative Ophthalmology & Visual Science, 39, s/72. Anstis, S. M., Mayhew, J. W., & Morley, T. (1969). The perception of where a Tv portrait is looking. American Journal of Psychology, 82, 474±489. Argyle, M., & Cook, M. (1976). Gaze and mutual gaze. Cambridge: Cambridge University Press. Baron-Cohen, 1995. Mindblindness: an essay on autism and theory of mind. Cambridge: MIT Press. Bruner, J. (1983). Child's talk: learning to use language. New Tork: Norton. Cline, M. G. (1967). The perception of where a person is looking. American Journal of Psychology, 80, 41±50. Diamond, R., & Carey, S. (1986). Why faces are and are not special: an effect of expertise. Journal of Experimental Psychology: General, 115, 107±117. Dunham, P. J., & Moore, C. (1995). Themes in research on joint attention. In C. Moore, & P. J. Dunham (Eds.), Joint attention: its origins and role in development. Hillsdale, NJ: Erlbaum. Ehrlich, S., & Field, D. (1993). Why is the acuity for the direction of gaze so good? Investigative Ophthalmology & Visual Science, 34 (4), 778. Frith, U. (1989). Autism: explaining the enigma. Oxford: Blackwell. Galper, R. E. (1970). Recognition of faces in photographic negative. Psychonomic Science, 19, 207±208. Gauthier, I., Skudlarski, P., Gore, J. C., & Anderson, A.W. (2000). Expertise for cars and birds recruits brain areas involved in face recognition. Nature Neuroscience, 3 (2), 191±197. Gibson, J. J., & Pick, A. D. (1963). Perception of another person's looking behaviour. American Journal of Psychology, 76, 386±394. Gross, C. G., Rocha-Miranda, C. E., & Bender, D. B. (1972). Visual properties of neurons in inferotemporal cortex of the macaque. Journal of Neurophysiology, 35 (1), 96±111. Hoffman, E. A., & Haxby, J. V. (2000). Distinct representations of eye gaze and identity in the distributed human neural system for face perception. Nature Neuroscience, 3 (1), 80±84. Hood, B. M., Willen, J. D., & Driver, J. (1998). Adult's eyes trigger shifts of visual attention in human infants. Psychological Science, 9 (2), 131±134. Johnson, M. H., & Morton, J. (1991). Biology and cognitive development: the case of face recognition. Oxford: Blackwell. Kanwisher, N., McDermott, J., & Chun, M. M. (1997). The fusiform face area: a module in human extrastriate cortex specialized for face perception. The Journal of Neuroscience, 17 (11), 4302±4311. Kemp, R., Pike, G., White, P., & Musselman, A. (1996). Perception and recognition of normal and negative faces: the role of shape from shading and pigmentation cues. Perception, 25, 37±52. Kobayashi, H., & Kohshima, S. (1997). Unique morphology of the human eye. Nature, 387, 767±768.
B14
P. Ricciardelli et al. / Cognition 77 (2000) B1±B14
Langton, S.R.H., Watt, R.J., & Bruce, V. (2000). Do the eyes have it? Cues to the direction of social attention. Trends in Cognitive Sciences. 4 (2), 50±59. Maurer, D. (1985). Infant's perception of faceness. In T. Field, & M. Fox (Eds.), Social perception in infants. Norwood, NJ: Ablex. Perrett, D., & Mistlin, A. (1990). Perception of facial characteristics by monkeys. In W. Stebbins, & M. Berkely (Eds.), Comparative perception (Vol. 2). New York: Wiley. Perrett, D., Harries, M., Mistlin, A., Hietanen, J., Benson, P., Bevan, R., Thomas, S., Oram, M., Ortega, J., & Brierley, K. (1990). Social signals analysed at the single cell level: someone is looking at me, something touched me, something moved! International Journal of Comparative Psychology, 4, 25± 55. Philips, R. J. (1972). Why are faces hard to recognize in photographic negative? Perception and Psychophysics, 12, 425±426. Yin, R. K. (1969). Looking at upside-down faces. Journal of Experimental Psychology, 81, 141±145.