Three-month-olds’ visual preference for faces and its underlying visual processing mechanisms

J. Experimental Child Psychology 90 (2005) 255–273

www.elsevier.com/locate/jecp

Three-month-olds’ visual preference for faces and its underlying visual processing mechanisms

Chiara Turati¤, Eloisa Valenza, Irene Leo, Francesca Simion

Dipartimento di Psicologia dello Svilluppo e della Socializzazione, Università degli Studi di Padova,Padova, Italy

Received 28 April 2004; revised 3 November 2004Available online 15 December 2004

Abstract

This study was aimed at investigating the face preference phenomenon and its underlyingmechanisms at 3 months of age. Using an eye-tracker apparatus, Experiment 1 demonstratedthat 3-month-olds prefer natural face images to unnatural ones, replicating and extending pre-vious evidence obtained with schematic facelike stimuli. Experiments 2 and 3 showed that thegeneral mechanisms that induce face preference in newborns could not explain the same phe-nomenon at 3 months of age, when infants are attracted by perceptual cues more speciWc tofaces. This suggests that signs of a process of cognitive specialization are already present in 3-month-olds’ visual behavior toward faces. 2004 Elsevier Inc. All rights reserved.

Keywords: Face perception; Infant; Cognitive specialization

Introduction

One central issue in cognitive science is how knowledge of speciWc domains is pro-cessed in the human brain. Face processing is an interesting topic in that respect

* Corresponding author. Fax: +39 049 827 6511.E-mail address: [email protected] (C. Turati).

0022-0965/$ - see front matter 2004 Elsevier Inc. All rights reserved.doi:10.1016/j.jecp.2004.11.001

mailto: [email protected]

mailto: [email protected]

256 C. Turati et al. / Journal of Experimental Child Psychology 90 (2005) 255–273

because several lines of evidence suggest that, in adults, face processing rests on ananatomical and/or a functional specialization of cerebral circuits (e.g., Farah, 2000;Kanwisher, 2000). However, what determines this specialization and how this special-ization emerges during development still remain largely unknown. The aim of thisstudy is to examine cognitive specialization during early infancy through the investi-gation of the development of infants’ abilities to process faces.

Three major approaches to the developmental origins of face processing havebeen distinguished (de Haan & Halit, 2001). A Wrst view argues that the developmentof face specialization is an experience-independent process (Farah, Rabinowitz,Quinn, & Liu, 2000). Due to the relevance of faces in our lives, natural selection led,through phylogenesis, to the evolution of innate face-speciWc devices that would beavailable prior to any postnatal experience. Within this perspective, it is suggestedthat the functional and anatomical distinction that characterizes face processing is“explicitly speciWed in the genome” (Farah et al., 2000). A second approach charac-terizes the development of face specialization as an experience-dependent process.Extensive and prolonged experience with faces would gradually render humansexceptionally expert in recognizing individual exemplars belonging to this class ofstimuli. In this approach, general learning processes, which can occur at any timeduring development and might involve any class of visual stimuli, are considered tobe suYcient to explain the gradual emergence of a cortical system for the rapid andeYcient processing of faces (Diamond & Carey, 1986; Gauthier & Logothetis, 2000;Tarr & Gauthier, 2000). Within the frame of a third (more recent) approach, thedevelopment of face processing has been explained in terms of an experience-expec-tant process. In this view, the cortical tissue has gained, through evolutionary pres-sures, the potential to become specialized for face processing. However, thisspecialization would emerge only if the critical type of input is provided within thecrucial time windows (Nelson, 2001, 2003). This approach is contingent on a probabi-listic epigenesis of cognitive development that views interactions among genes, struc-tural brain changes, and psychological functions as bidirectional (Black, Jones,Nelson, & Greenough, 1998; Greenough & Black, 1992). More speciWcally, the partialfunctioning of neural pathways would shape subsequent development of neuralstructures and circuits that are the basis for further functional development. Thisring-shaped process would result in the progressive tuning of certain cerebral tissues,from a broad range of visual information to the speciWc type of information con-veyed by faces (Johnson, 2000; Johnson & de Haan, 2001; Nelson, 2001, 2003).

One way of disentangling the long-standing issue of the development of cognitivespecialization for faces might be to study the developmental time course of a phe-nomenon present a few days from birth—newborns’ visual preference for faces.When presented with face and nonface patterns, newborns spontaneously looklonger at and orient more frequently toward the conWguration that represents a face(Johnson & Morton, 1991; Macchi Cassia, Turati, & Simion, 2004; Valenza, Simion,Macchi Cassia, & Umiltà, 1996).

The well-known model of the development of face processing proposed indepen-dently by Johnson and Morton (1991) and de Schonen and Mathivet (1989)explained newborns’ preference for faces by hypothesizing the presence at birth of a

C. Turati et al. / Journal of Experimental Child Psychology 90 (2005) 255–273 257

speciWc representational bias or mechanism at a lower neural level, that is, in subcor-tical tissue rather than cortical tissue. By ensuring that infants have visual experiencewith faces, this subcortical mechanism would favor the gradual emergence of the spe-cialized cortical circuits that subserve face processing in adults.

However, a diVerent account of newborns’ preference for faces recently suggestedthe possibility that general constraints of the visual system in processing all kinds ofstimuli might be suYcient to explain newborns’ proclivity toward faces (Macchi Cas-sia et al., 2004; Simion, Macchi Cassia, Turati, & Valenza, 2001; Simion, Valenza,Macchi Cassia, Turati, & Umiltà, 2002; Simion, Macchi Cassia, Turati, & Valenza,2003; Turati, 2004; Turati, Simion, Milani, & Umiltà, 2002). More speciWcally, thismodel maintains that newborns’ visual preference for faces reXects a preference for acollection of general structural properties that attract newborns’ gaze. So far, it hasbeen shown that two diVerent general structural properties play a role in determiningnewborns’ preference for faces. The Wrst, termed congruency, is deWned by the pres-ence of a congruent spatial relation between the spatial disposition of the inner fea-tures and the shape of the outer contour, with more features being located in thewidest portion of the conWguration (Macchi Cassia, Valenza, Pividori, & Simion,2002; see also Acerra, Burnod, & de Schonen, 2002). The second, termed up–downasymmetry, is deWned by the presence of more patterning in the upper part of theconWguration than in the lower part (Macchi Cassia et al., 2002; Simion et al., 2002;Turati et al., 2002). Evidence indicates that these nonspeciWc structural propertiesinduce a visual preference in newborns when they are embedded in nonface stimuliand are crucial in eliciting newborns’ preferential response to faces. These outcomessuggest that newborns’ putatively speciWc preference for faces may be explained bythe result of the cumulative eVect of a set of nonspeciWc constraints that stem fromthe general characteristics of the human visuo-perceptual system shortly after birthrather than by an experience-independent subcortical mechanism devoted to detect-ing faces.

Based on this literature, it becomes interesting to investigate whether and howinfants’ preferential response to faces becomes speciWc for this stimulus category.Some recent neuropsychological studies that measured scalp-recorded brain electricpotentials (event-related potential, ERP) (Halit, de Haan, & Johnson, 2003) or per-formed positron emission tomography (PET) scans (Tzourio-Mazoyer et al., 2002)suggested that the Wrst signs of cortical specialization for faces can be observed in 2-to 3-month-olds. More speciWcally, Halit and colleagues (2003) identiWed in 3-month-olds an ERP component that diVerentiates human faces from monkey faces andsuggested that “even as young as 3 months of age, infants show some specializationfor human faces,” although not to the same degree as do older infants or adults.Moreover, mapping the brain activity of 2-month-olds with PET while they werelooking at unknown women’s faces, Tzourio-Mazoyer and colleagues (2002) observed“the activation of a distributed network of cortical areas that largely overlapped theadult face processing network” (p. 454). However, to our knowledge, no studies haveinvestigated face specialization in 3-month-olds using a behavioral marker task ratherthan neuropsychological measures. Face preference might be the appropriate candi-date for achieving this purpose. Investigating the developmental time course of the


face preference phenomenon, and of its underlying mechanisms, may contribute toknowledge about the emergence of cognitive specialization for faces.

Paradoxically, because research has been focused on examining whether newbornspossess an innate mechanism devoted to faces, only a few dated studies have investi-gated face preference in infants over 2 months of age (e.g., Haaf, 1974; Haaf & Brown,1976; Johnson, Dziurawiec, Bartrip, & Morton, 1992; Johnson, Dziurawiec, Ellis, &Morton, 1991; Johnson & Morton, 1991; Koopman & Ames, 1968). The resultsobtained in these studies are not fully consistent, mostly because low-level variables,such as amount of contour or spatial frequencies, were not always adequately con-trolled. In addition, although across studies stimuli have ranged from the very simple(e.g., Johnson et al., 1991) to the very complex (e.g., Haaf, Smith, & Smitley, 1983; Kle-iner, 1987; Kleiner & Banks, 1987), only one study recently conducted with 1- to 3-day-olds (Macchi Cassia et al., 2004, Experiment 1) investigated infants’ preference forfaces using veridical face images. Thus, the ecological validity of the results obtainedwith schematic face stimuli throughout infancy is still a matter of question. The Wrstaim of Experiment 1 was to investigate, using photographs of real faces and an eye-tracker apparatus, whether 3-month-olds still manifest a visual preference for faces.

Experiment 1

Experiment 1 was aimed at following the developmental trend of face preference,testing whether this phenomenon is still present at 3 months of age, when a certaindegree of cortical specialization for faces begins to emerge (Halit et al., 2003; Tzou-rio-Mazoyer et al., 2002). The experiment was run using an eye-tracker system thatallows the recording of eye movements’ direction by virtue of an infrared camera andsoftware that automatically determines the position of the eyes with respect to thestimuli. Eye movements were recorded as 3-month-olds observed the photograph ofa canonical upright face and an upside-down version of the same face in which theinner portion of the face was rotated 180° (Fig. 1), as in the classic previous studieswith schematic faces (Johnson et al., 1991; Valenza et al., 1996) and in the study thattested face preference in newborns using photographs of real faces (Macchi Cassiaet al., 2004, Experiment 1).

DiVerent predictions might be advanced regarding possible outcomes of the exper-iment. One possibility is that 3-month-olds, like newborns, might manifest a sponta-neous visual preference for the upright face. This result would extend previouspositive evidence of the presence of the face preference phenomenon obtained withschematic faces (Johnson et al., 1992; Kleiner & Banks, 1987). Moreover, this Wndingwould support the idea that when infants over 2 months of age are tested, the morethe stimulus material is realistic, the more likely it is that a preference for theupright face is found (Johnson & Morton, 1991). Another possibility is that 3-month-olds’ visual behavior would be radically diVerent from that of newborns.More speciWcally, given the extensive visual experience with faces that character-izes infants’ normal development, infants’ gaze at 3 months of age might be moreattracted by the novelty of an upside-down version of the face in which the inner


portion is misoriented than by the highly familiar natural upright face. A Wnalpossibility is that the two conXicting tendencies just described—the spontaneouspreference for the upright face and the novelty preference for the upside-downface—would neutralize each other, determining a lack of any signiWcant preferencefor either visual conWguration.

Method

Participants

Participants were 16 healthy and full-term infants, 3 months of age (mean age D 98days, range D 89–125, 8 males and 8 females). There were 8 other infants tested butnot included in the Wnal sample because they showed a strong position bias (i.e., theylooked to one direction more than 90% of the time, n D 4) or because the eye-trackersystem succeeded in detecting their gaze direction in a reliable way in less than 50%of the trials, n D 4. Infants were tested only if they were awake and in an alert stateafter their parents had given their informed consent.

Stimuli

Five pairs of stimuli were presented, each composed of an upright face and anupside-down face (Fig. 1). Black-and-white photographs of women’s faces wereused. Photographs were modiWed with Adobe Photoshop, so that an upright faceand an upside-down version of the same face, analogous to those employed byMacchi Cassia and colleagues (2004, Experiment 1), were generated. More speciW-

cally, hair was removed from each photograph and, on the upside-down face, onlythe inner portion of the face was rotated 180°. Each face was photographed in afrontal pose and with a neutral expression. Stimuli measured 10 cm in width and15 cm in height and appeared on a black background. The upright face and the

Fig. 1. An example of the pairs of stimuli used in Experiment 1.


upside-down face were shown on a computer screen at a distance of 10 cm fromeach other.

Apparatus

Each infant was placed in an infant seat at a distance of approximately 60cm fromthe computer screen where the stimuli were presented. A system for the automatic regis-tration of eye movements made by the Applied Science Laboratories (ASL) wasemployed. An infrared camera was located at the foot of the computer screen. A stu-dent guided the camera by means of a remote control, so that the left eye of the infantwas always in focus. The image of the left eye on a television monitor rendered this pro-cedure easier. The eye-tracker system was able to detect the position of the pupil andthe corneal reXection of the infrared light-emitting diodes (LEDs) in the eye. Becausethese signals changed in relation to the infant’s gaze direction, the apparatus deter-mined, with a frequency of 50Hz, the x–y coordinates corresponding to the infant’sWxation point during stimulus presentation. Two crosses of diVerent colors correspond-ing to the signals coming from the infant’s pupil and corneal reXection were superim-posed on the images of the stimuli shown, so that the direction of the infant’s gaze withrespect to the stimuli during the experimental session could be videotaped.

Procedure

Each trial began with a calibration phase during which we adapted the parametersfor the detection of the pupil and of the corneal reXection to the characteristics of theeye of each infant and also detected and recorded the signals of the pupil and of thecorneal reXection at the precise moment in which infant gaze was directed towardthree diVerent locations of the computer screen (center, top left, and bottom right).To attract the infant’s gaze toward the predetermined locations, an animated cartoonwith a musical soundtrack was employed. On the basis of the initial calibrationphase, the eye-tracker system was subsequently able to determine the precise direc-tion of the infant’s gaze during the experimental session.

The experimental session began with the presentation, in the middle of the screen,of a central Wxation point particularly attractive for the infant (i.e., a colored movingclown). This central Wxation point was used to attract the infant’s gaze toward thecomputer screen where stimuli were shown and to check that the infant’s gaze wasaligned with the horizontal midline of the screen during the entire experimental ses-sion. As soon as the infant looked at the central Wxation point, the Wrst pair of stimuliautomatically appeared on the computer screen and the central Wxation wasremoved. Each stimulus pair was presented for 8 s. Afterward, the central Wxationpoint reappeared in the middle of the screen and the trial loop started again.

Each of the 5 pairs of stimuli was presented 4 times, in a pseudorandom sequence,for a total of 20 trials. The left versus right positions of the two stimuli within eachpair were counterbalanced. Software allowed us to process the raw data coming fromthe eye-tracker system, calculating infants’ total Wxation time toward the upright andupside-down faces.


Results and discussion

The results are presented in two subsections. In the Wrst subsection (on overalltotal looking time), data parallel those classically registered in a visualpreference paradigm, so that results obtained with 3-month-olds, without theemployment of human coders, could be compared to results obtained in previousstudies with newborns. In addition, by the use of the eye-tracker system, we sepa-rately recorded Wxation times to the upper and lower halves of each stimulus pre-sented. By analyzing the distribution of looking time in each of these selected areas,we could examine which perceptual cues within the stimuli attracted infants’ gaze.These data are reported in the second subsection (on distribution of looking).

Overall total looking time

To investigate whether the upright face was preferred over the upside-down face,a paired-sample t test was performed on infants’ mean Wxation time toward the twoconWgurations. The comparison was statistically signiWcant, t (15) D 2.62, p < .02.Infants looked longer at the upright face (M D 1964 ms, SD D 724) than at theupside-down face (M D 1471 ms, SD D 559). In addition, preference scores (percent-ages) for the upright face over the upside-down face were calculated. Each infant’smean looking time at the upright face was divided by the looking time to both teststimuli and converted into a percentage score. Preference scores for the upright facewere signiWcantly above the chance level of 50% (M D 57.11%, SD D 10.06), one-sample t (15) D 2.82, p < .02. Finally, examination of the data for individual infantsrevealed that 13 of 16 infants looked longer at the upright face (binomial test,p < .009).

These Wndings demonstrate that 3-month-olds exhibit a spontaneous visual pref-erence for an upright image of a real face over an upside-down version of the sameface. This was the case even though, at 3 months of age, the upside-down conWgura-tion might be considered a particularly salient novel stimulus with respect to thehighly familiar image of an upright face. A novelty response toward the upside-down face could be even more likely considering the peculiar characteristics of theupside-down face used; only the inner portion of the face was misoriented and, thus,had an incongruent novel relation with the more external portion of the face. Con-trary to this hypothesis, infants’ gaze was more attracted by the canonical faceimage than by the atypical upside-down face. This outcome extends previousevidence gathered with schematic conWgurations to natural face images, therebyproving the ecological validity of the face preference phenomenon in infants over 2months of age.

Distribution of looking time

To understand which perceptual cues of each stimulus attracted infants’ gaze, aseries of statistical analyses were performed on percentage of looking toward fourselected areas corresponding to the upper and lower halves of each stimulus within


the pair. One-sample t tests were applied to verify whether such percentages of look-ing diVered from a chance level of 25%. The following distribution of looking wasobserved. The 3-month-olds looked at the upper half of the upright face for 1612 ms,corresponding to 47% of total Wxation time (i.e., eyes, M D 46.67%, SD D 16.28),t (15) D 5.33, p < .001, and at the lower half of the face for 352 ms, corresponding to10% of total Wxation time (i.e., mouth, M D 10.43%, SD D 11.44), t (15) D ¡5.09,p < .001. The upper half of the upside-down face was looked at for 887 ms, corre-sponding to 26% of total looking time (i.e., mouth, M D 26.31%, SD D 10.36), t (15) D0.50, p > .60, and the lower half of the upside-down face was Wxated for 584 ms, corre-sponding to 17% of total looking time (i.e., eyes, M D 16.58%, SD D 11.44),t (15) D ¡2.94, p < .02.

These Wndings indicate that the preference for the upright face conWguration isdue mainly to a higher percentage of looking toward the area of the face that corre-sponds to eyes, which attracted nearly half of the total looking time toward thestimuli (46.67%). Interestingly, eyes per se did not attract infants’ gaze when theywere misplaced and located in the lower portion of the face; the percentage of look-ing toward the lower half part of the upside-down face (16.58%) was signiWcantlylower than chance level and comparable to the percentage of looking toward themouth in the upright face. Thus, our Wndings showed that 3-month-olds manifesteda preference for looking at the eyes over other regions of the natural face. How-ever, contrary to the view that an eye direction detector is present from birth(Baron-Cohen, 1995), the perceptual information conveyed by eyes lost its impor-tance when eyes were located in the lower part of the face rather than in the upperpart of the face, as in the upside-down face. Thus, eyes per se do not seem suYcientto attract 3-month-olds’ gaze because infants’ visual preference is modulated bythe position of eyes within the stimulus, that is, by the context in which eyes arelocated.

To summarize, the obtained results demonstrate that 3-month-olds, like new-borns, manifest a visual preference for faces. However, it is still not clear what visualprocessing mechanisms underlie infants’ response to faces at 3 months of age. Exper-iments 2 and 3 were aimed at exploring this issue.

Experiment 2

Recently, a preference for up–down asymmetrical nonface patterns with moreelements in the upper part was demonstrated in newborns (Simion et al., 2002).Because faces are up–down asymmetrical, some authors have maintained that facepreference at birth actually reXects a preference for up–down asymmetry in thedistribution of the inner elements (Macchi Cassia et al., 2004; Simion et al., 2001,2003; Turati, 2004; Turati et al., 2002). Experiment 2 investigated whether up–down asymmetry is capable of producing a preferential response with non-facelikestimuli, even in 3-month-olds. More speciWcally, it tested whether 3-month-olds,like newborns, would manifest a preference for nonface patterns with a greater


number of high-contrast areas in the upper part of the conWguration than in thelower part.

Method

Participants

In this experiment, 38 infants were tested. However, 4 of them were not included inthe Wnal sample because the eye-tracker system was able to detect infants’ gaze direc-tion in a reliable manner in less than 50% of the trials. The Wnal sample was com-posed of 34 healthy and full-term infants, 3 months of age (mean age D 92 days,range D 84–102), randomly assigned to one of two groups (16 to Group 1 and 17 toGroup 2). Infants were tested only if they were awake and in an alert state. Parentswere informed and gave their consent.

Stimuli

Each stimulus was composed of Wve black elements that were located in a whiterectangle that was 10 cm in width and 15 cm in height. Two of the three types of stim-uli used in the study by Simion and colleagues (2001) were selected (Fig. 2). Each pairwas composed of a stimulus in which more high-contrast areas (i.e., black elements)were placed in the upper part of the conWguration than in the lower part (i.e., top-heavy conWguration) and a stimulus in which more high-contrast areas were locatedin the lower part of the conWguration than in the upper part (i.e., bottom-heavy con-Wguration). In the Wrst type of stimuli, which was shown to Group 1, the black ele-ments were placed in the correct locations to form two T shapes: one upright (i.e., topheavy, with three elements in the upper half of the conWguration) and the otherupside down (i.e., bottom heavy, with three elements in the lower half of the conWgu-ration). In the second type of stimuli, which was shown to Group 2, the top-heavystimulus consisted of four elements in the upper part of the conWguration and onlyone element in the lower part. The bottom-heavy stimulus was composed of four ele-ments in the lower part of the conWguration and only one element in the upper part.

For each type of stimuli, Wve diVerent pairs were generated. Stimulus pairs diVeredby virtue of the form of inner black elements, which could be stars, squares, hexa-gons, four-leaf clovers, or circles.

Apparatus and procedure

The apparatus and procedure were identical to those used in Experiment 1.


As in Experiment 1, the report of results is organized into two subsections, withthe Wrst focusing on overall total looking time and the second focusing on distribu-tion of looking time.



To test whether 3-month-olds exhibited a visual preference for both types of top-heavy patterns presented, a 2 £ 2 mixed analysis of variance (ANOVA), with Stimu-lus (top heavy or bottom heavy) as a within-subjects factor and Group (1 or 2) as abetween-subjects factor, was performed. No main eVect reached statistical signiW-cance. However, the analysis revealed a signiWcant Stimulus £ Group interaction,F (1, 31) D 4.36, p < .05. Post hoc t tests indicated that mean looking times toward thetop-heavy and bottom-heavy conWgurations were not signiWcantly diVerent forinfants assigned to Group 1 (top-heavy M D 1854 ms, SD D 575; bottom-heavyM D 1978 ms, SD D 476), t (15) D 0.72, p > .45. In contrast, infants assigned to Group 2looked longer at the top-heavy conWguration (M D 2265 ms, SD D 512) than at thebottom-heavy conWguration (M D 1964 ms, SD D 564), t(16) D 2.68, p < .02.

A preference score was also computed as in Experiment 1. When computed irre-spective of the group to which infants belonged, the mean preference score was51.03% (SD D 7.61) and did not diVer signiWcantly from the chance level of 50%,t(32) D 0.78, p > .40. However, when two separate one-sample t tests were applied (one



for each group), preference scores for the top-heavy pattern were signiWcantly abovechance for Group 2 (four elements in the upper part, M D 53.93%, SD D 5.94),t(16) D 2.73, p < .02, but not for Group 1 (three elements in the upper part,M D 47.95%, SD D 8.15). A t test for independent samples revealed that the meanpreference scores for the two groups diVered signiWcantly, t(31) D 2.42, p < .03.Finally, examination of data of individual infants revealed that 13 of 17 infantsassigned to Group 2 (binomial test, p < .03), but only 6 of 16 assigned to Group 1,looked longer at the top-heavy pattern (binomial test, p > .05).

Thus, statistical analyses converge to indicate that 3-month-olds’ responses varied asa function of the type of up–down asymmetrical stimuli that were presented. More spe-ciWcally, when infants were shown T-shaped patterns (Group 1), their visual behaviorappeared to be radically diVerent from that exhibited by newborns (Simion et al., 2002)because the visual preference for the stimulus with more patterning in the upper por-tion of the conWguration (i.e., the top-heavy upright T shape) disappeared. However,when the other type of stimuli, with four elements in the upper or lower half of the con-Wguration (Group 2), was considered, a diVerent pattern of results emerged; that is, 3-month-olds’ behavior paralleled that shown by newborns (Simion et al., 2002), with thetop-heavy conWguration being preferred over the bottom-heavy conWguration. TheseWndings suggest that, at 3 months of age, the determinants of infants’ preference fornonface top-heavy patterns are still active but are less powerful than in newborns.


As in Experiment 1, to understand which portions of each stimulus attractedinfants’ gaze, a series of one-sample t tests were run comparing the chance level of25% with the percentages of looking toward four selected areas corresponding to theupper and lower halves of each stimulus within a pair. For the pair of stimuli com-posed of the upright and upside-down T (Group 1), the following distribution oflooking was observed. The 3-month-olds looked at the upper half of the upright Tfor 1426 ms, corresponding to 37% of total Wxation time (M D 37.13%, SD D 10.38),t (15) D 4.68, p < .001. They looked at the lower half of the upright T for 429 ms,corresponding to 11% of total Wxation time (M D 10.82%, SD D 7.93), t (15) D 7.15,p < .001. The upper half of the upside-down T was looked at for 960 ms, correspond-ing to 25% of total looking time (M D 25.15%, SD D 10.11), t (15) D 0.06, p > .90.Finally, the lower half of the upside-down T was Wxated for 1018 ms, correspondingto 27% of total looking time (M D 26.89%, SD D 10.13), t (15) D 0.74, p > .45.

A similar pattern of looking was observed for the pair of stimuli shown to Group2. The upper half of the top-heavy stimulus was Wxated for 1468 ms, corresponding to34% of total Wxation time (M D 34.40%, SD D 12.56), t(16) D 3.08, p < .008. The lowerhalf of the top-heavy pattern was Wxated for 797 ms, corresponding to 19.53% oftotal looking time (M D 19.53%, SD D 12.06), t(16) D 1.87, p < .01. The upper halfof the bottom-heavy conWguration was looked at for 1030 ms, corresponding to 23%of total looking time (M D 23.42%, SD D 10.57), t(16) D 0.62, p > .50. Finally, thelower half of the bottom-heavy stimulus was Wxated for 934 ms, corresponding to23% of total looking time (M D 22.66%, SD D 11.71), t(16) D 0.82, p > .40.


To summarize, infants’ distributions of looking toward the two top-heavy andbottom-heavy patterns were quite similar. In both groups, the percentage of lookingtoward the two selected areas of the bottom-heavy pattern did not diVer from chance.More important, in both groups, the upper portion of the top-heavy conWgurationwas looked at signiWcantly more than chance level. Infants’ focus of interest towardthe top part of the top-heavy conWguration is of particular interest for our purposesbecause in Experiment 1 with upright faces, a higher percentage of looking towardthe area of the face that corresponds to eyes was observed. However, it is worth men-tioning that a t test comparison between infants tested in Experiment 1 and infantstested in Experiment 2 on percentage of looking toward the upper portion of the top-heavy conWguration (i.e., face in Experiment 1 and geometric patterns in Experiment2) revealed a signiWcant diVerence between means (face: M D 46.67%, SD D 35.73;geometric patterns: M D 35.73%, SD D 11.46), t(47) D ¡2.72, p < .01. Based on theseWndings, it seems that when 3-month-olds are shown with top-heavy patterns (facesor geometric stimuli), they explore the upper part of the stimuli longer. Nevertheless,this tendency appears to be signiWcantly strengthened in the case of faces. This is veryinteresting because the number of elements and the amount of contrast contained inthe upper part of geometric stimuli are undoubtedly greater than those contained inthe upper part of top-heavy geometric stimuli.

Overall, evidence obtained in Experiment 2 leads us to conclude that a tendency toexplore more extensively conWgurations that display more patterning in the upperpart is still present, although weaker, in 3-month-olds. However, it remains unclearwhether this tendency might explain 3-month-olds’ preference for faces. Experiment3 was aimed at investigating this issue more directly.

Experiment 3

The weak preference for up–down asymmetrical patterns obtained in Experiment2 raises the possibility that up–down asymmetry plays a role in determining 3-month-olds’ preference for faces. This possibility motivated us to explore whether face pref-erence at 3 months of age might be ascribed, as in the case of newborns, to a nonspe-ciWc perceptual bias toward top-heavy patterns or whether, during the Wrst months oflife, infants’ preferential response becomes more speciWc to the face category.

To achieve this purpose, we directly compared up–down asymmetry and facednessin inducing a visual preference. An image of a natural face was presented togetherwith a top-heavy scrambled face in which the number of features placed in the upperportion of the conWguration was paired to that present in the natural face. Twoopposite predictions may be advanced. If face preference at 3 months of age derivesfrom a nonspeciWc bias toward top-heavy patterns, infants would not be expected tomanifest any preference because the natural face and the top-heavy scrambled faceare paired for up–down asymmetry in the distribution of the inner features. Alterna-tively, if 3-month-olds’ face preference emerges by virtue of the speciWc spatial rela-tions between face features, infants should prefer the natural face that preserves thisrelation.


Method

Participants

In this experiment, 14 healthy and full-term infants, 3 months of age (meanage D 96.7 days, range D 89–98 days), were recruited. Of these infants, 4 were removedfrom the study because of experimenter errors. The Wnal sample consisted of 10infants (6 males and 4 females) who met the screening criteria of normal delivery.Informed consent was obtained from their parents.

Stimuli

The same high-quality black-and-white photographs of Wve women’s faces used inExperiment 1 were employed. Each photograph was digitally modiWed, manipulatingthe position of the inner face features so that a top-heavy scrambled version of eachface was generated (Fig. 3). Thus, Wve pairs of natural and top-heavy scrambled faceswere created. Each infant was presented with 5 pairs of stimuli, each consisting of anatural face and its corresponding top-heavy scrambled version.

Apparatus and procedure

The apparatus and procedure were identical to those used in Experiment 1.



Infants’ mean Wxation times toward the upright face and the top-heavy scrambledface were examined Wrst. Mean looking times toward the two conWgurations were



signiWcantly diVerent, paired-sample t test, t(9) D 2.60, p < .03. The average Wxationtime toward the upright face was greater (M D 2400 ms, SD D 830) than the averageWxation time toward the top-heavy scrambled face (M D 1900 ms, SD D 773). In addi-tion, mean percentages of looking time for the upright face over the top-heavy scram-bled face were calculated for each infant as in Experiments 1 and 2. Preferencesscores for the upright face were signiWcantly above the chance level of 50%(M D 56.18, SD D 7.63), t(9) D 2.56, p < .035. Finally, 8 of the 10 infants tested lookedlonger at the natural face (binomial test, p < .045).

These Wndings suggest that the face arrangement of the inner features in the natu-ral face aVects infants’ performance, attracting infants’ gaze even when face conWgu-rations are paired for stimuli equated with up–down asymmetry in the distribution ofinner features. Therefore, infants’ face preference does not rely, as in the case of new-borns, on the amount of information appearing in the upper portion of the stimuli asopposed to the lower portion; rather, it relies on the presence of perceptual informa-tion speciWc to faces.


As in Experiments 1 and 2, to obtain a clear picture of the Wxation patterns, wecomputed the cumulative amount of looking time within four areas corresponding tothe upper and lower halves of each stimulus within the pair. Four one-sample t testswere run on percentages of looking time toward the four selected areas, comparingpercentages of looking with the chance value of 25%. A large portion of looking timewas devoted to the upper half of the natural face; infants looked at this area for 1958ms, corresponding to 45% of total Wxation time (i.e., eyes, M D 45.04%, SD D 10.43),t(9) D 6.07, p < .001. Also, the upper half of the scrambled face attracted a consider-able amount of infants’ gaze, although to a lesser extent than the corresponding por-tion of the natural face; infants’ eyes Wxated on this area for 1552 ms, correspondingto 36% of total Wxation time (M D 35.62%, SD D 7.81), t(9) D 4.30, p < .003. The lowerpart of the natural face and the lower part of the scrambled face were observed for441 ms, corresponding to 11% of total looking time (i.e., mouth, M D 11.14%,SD D 8.33), t(9) D 5.26, p < .002, and 345 ms, corresponding to 8% of total lookingtime (M D 8.20%, SD D 4.33), t(9) D 4.30, p < .003, respectively. As revealed by one-sample t tests, both percentages of looking were lower than chance value.

The current Wndings, when linked to those obtained in Experiment 1, are interest-ing with respect to the issue of what perceptual cues render faces so attractive for 3-month-olds. The presence of more features in the upper part does not represent asuYcient property to explain 3-month-olds’ preference for faces because the top-heavy nonface conWguration attracted infants’ gaze to a lesser extent than did theface. One might propose that eyes per se are the key factor in determining 3-month-olds’ preference for faces. Indeed, evidence suggests that the upper part of the naturalface (i.e., eyes) plays a critical role in attracting infants’ gaze. However, this interpre-tation is not consistent with the results obtained in Experiment 1; that is, when eyesare misplaced and located in the lower part of a face, they lose part of their salience.Therefore, two remaining possibilities may be suggested to explain the obtained


pattern of evidence. One might maintain that the eyes are not preferred per se butrather are preferred only when they are properly located in the upper part of the faceconWguration. So, eyes appear to be a more salient feature of a face than is its nose ormouth. However, it would not be suYcient that the eyes are present anywhere on theface because where those eyes are located determines whether the 3-month-olds’visual preference emerges.

An alternative possibility is that infants were not responding to some local fea-tures but rather were responding to the arrangement of the inner features relative toeach other. In this view, eyes would not be predominant in inducing face preference,but they would play a role, together with the other face features, in deWning the spe-ciWc geometry of the face. The 3-month-olds’ gaze would be triggered by the interre-lation between the shape and position of the various facial features (e.g., eyes, nose,mouth). In other words, due to the extensive visual experience with faces that charac-terizes infants’ normal development, infants might have formed a kind of face repre-sentation or prototype that rendered faces attractive visual stimuli. In agreementwith this hypothesis, 3-month-olds’ face preference might be explained as the resultof an expertise-based eVect. Such a conclusion is reinforced by a recent study thatobserved a preference for female faces over male faces at 3 months of age, probablydue to the infants’ greater familiarity with female individuals (Quinn, Yahr, Kuhn,Slater, & Pascalis, 2002). In any case, whatever the correct interpretation, our crucialpoint here is that the mechanisms responsible for 3-month-olds’ preference for facesdiVer from those responsible for the same phenomenon in newborns.

General discussion and conclusions

Two diVerent purposes motivated the current study. The Wrst was to investigatewhether a visual preference for faces is still present at 3 months of age when veridicalface images are used. Our Wndings were clear in demonstrating that natural faces arestill preferred over unnatural faces at 3 months of age. Thus, apparently, 3-month-olds’ behavioral response toward faces appears to be highly similar to that shown by3-day-olds with comparable procedure and stimulus material.

The second aim of this study was to explore which visuo-perceptual processingmechanisms underlie face preference at 3 months of age. Do they overlap the generalprocesses recently indicated as the determinants of newborns’ face preference (Mac-chi Cassia et al., 2004; Simion et al., 2001, 2002; Turati, 2004)? Or, do they show anincreased level of speciWcity for faces given the countless opportunities that theseinfants have had to experience this category of visual stimuli?

Because recent evidence has suggested the possibility that a nonspeciWc structuralproperty, up–down asymmetry, plays a crucial role in determining newborns’ prefer-ence for faces (Simion et al., 2001, 2002), Experiment 2 explored whether this generalproperty still attracts 3-month-olds. The Wndings did not give a deWnite answer to thisquestion because the two types of top-heavy visual conWgurations used induceddiVerent responses in infants, with one being preferred and the other not. A possibleinterpretation of these outcomes might suggest that, at 3 months of age, the general


mechanism responsible for infants’ visual preference for top-heavy patterns is stillpresent but weaker and, thus, is activated only when up–down asymmetry is highlysalient. Following this line of reasoning, the stimulus with four grouped elements inthe upper half would still be able to activate the up–down asymmetry detector. Incontrast, the top-heavy T shape that has only three elements in the upper half wouldno longer be able to trigger a preferential response. In this view, preference for up–down asymmetrical patterns would not disappear abruptly but rather would do soprogressively during the Wrst months of life.

Therefore, 3-month-olds’ visual behavior did not entirely overlap that shown bynewborns with geometric up–down asymmetrical stimuli. Newborns exhibited avisual preference for any stimulus that was up–down asymmetrical, with more pat-terning in the upper portion of the conWguration than in the lower portion (MacchiCassia et al., 2004; Simion et al., 2001, 2002; Turati, 2004; Turati et al., 2002). But at 3months of age, at least in the case of one stimulus type (i.e., the top-heavy T shape),up–down asymmetry failed to attract infants’ gaze.

The absence of any visual preference for the top-heavy T shape is very interestingbecause this type of top-heavy geometric stimulus is the one that, of the two used inExperiment 2, most resembled a face given that the disposition of its inner elementsroughly paralleled that of the inner face features. This suggests that the determinantsof 3-month-olds’ preference for faces might be diVerent from those that induced thepreference for the top-heavy patterns with four elements in the upper portion of theconWguration because the top-heavy T shape, which may be considered the up-downasymmetrical stimulus most similar to faces, was not able to elicit a preferentialresponse.

Findings obtained in Experiment 3 conWrmed that the preference for faces at 3months of age could not be explained, as in newborns, as the result of the asymmetri-cal distribution of the inner features within the face. The face conWguration prevailedover the nonface up–down asymmetrical conWguration in triggering infants’ atten-tion. Because the two stimuli were equated with the number of features appearing inthe upper and lower portions of the conWguration, up–down asymmetry was not thecrucial factor in inducing 3-month-olds’ preference for faces. Such a pattern of evi-dence diVers from that observed at birth, where faces are no longer preferred if theyare paired with conWgurations equated with up–down asymmetry (Macchi Cassia etal., 2004; Turati et al., 2002). Thus, it seems that face preference in 3-month-olds is theresult of perceptual cues speciWc to faces rather than of general properties that facesshare with other visual conWgurations. However, the nature of such speciWc percep-tual cues remains unclear and might be the focus of future studies in this Weld. Evi-dence suggested that a major role is played by the eyes, but only when they arelocated in their natural arrangement within the face.

To summarize, whereas at birth face preference seems to depend on the existence ofgeneral biases that orient newborns’ attention toward certain structural properties thatfaces share with other visual stimuli (Simion et al., 2001, 2002; Turati et al., 2002), at 3months of age a modiWcation in the determinants of infants’ face preference seems tooccur in that the mechanisms at the basis of face preference diVer from those involvedin visual preferences for nonface patterns. At 3 months of age, face preference appears


to be the product of more speciWc mechanisms that respond more selectively to thoseperceptual characteristics that distinguish faces from other stimulus categories.

The whole picture of our results is consistent with two diVerent models of thedevelopment of the face processing system during infancy. A Wrst interpretationmight revise the well-known Johnson and Morton (1991) model, rejecting thehypothesis of the existence at birth of a speciWc mechanism devoted solely to orient-ing newborns’ gaze toward faces but preserving the idea that a qualitative change inthe mechanisms that process faces occurs during the Wrst months of life. In line withthis view, some subcortical and nonspeciWc visual processing mechanisms at birthwould induce newborns to prefer a large category of stimuli that include faces. At 2months of age, a diVerent cortical mechanism would begin to emerge and would formthe basis of the adult system specialized in face processing.

An alternative interpretation of the same pattern of results might place greateremphasis on developmental continuity in the emergence of the face processing systemduring infancy, thereby contradicting the idea of the existence of marked discontinu-ities in the mechanisms that mediate face processing at birth and at 3 months of age(Nelson, 2003). More speciWcally, in line with a computational model recently imple-mented by Bednar and colleagues (Bednar, 2003; Bednar & Miikkulainen, 2003), thedevelopment of the face processing system might be viewed as the gradual and pro-gressive emergence of a single, increasingly complex system rather than as the succes-sive appearance of novel and independent mechanisms. In particular, it may besuggested that a portion of those nonspeciWc visual processing mechanisms thatdetermine a visual preference for those stimuli that are up–down asymmetrical in thedisposition of their inner elements become progressively specialized for the process-ing of a narrower category of visual stimuli with which infants have extensive visualexperience, that is, faces.

To conclude, the evidence presented in this study appears to be relevant to the issueof how face processing emerges as a specialized ability during development. Data pre-sented are consistent with an experience-expectant model of face processing that con-siders brain specialization and speciWcity as emerging gradually from the interactionbetween tiny innate constraints and the critical input provided by the species-typicalenvironment (Elman et al., 1996; Johnson, 1993; Nelson, 2001, 2003). To develop in itsadultlike expert form, the face processing system might not require a highly speciWc apriori input (i.e., a face-speciWc bias at birth). Rather, a number of general biases mightwork together to provide the minimal information that would be suYcient to set thesystem. Combined with the prolonged exposure to human faces typical of infants’ envi-ronments, these general biases drive the system toward the emergence of the functionaland neural specialization that can be observed later in development. Signs of this pro-cess of cognitive specialization seem to be already present in 3-month-olds.

Acknowledgments

This research project was supported by a grant from the Ministero dell’Uni-versità e della Ricerca ScientiWca e Tecnologica (No. 2003112997-004). The


authors are deeply indebited to C. Umiltà and V. Macchi Cassia for their com-ments on preliminary data and earlier versions of the article, S. Massaccesi forhis great help in assembling the eye-tracker apparatus, and L. Gava for her helpin testing babies.

References

Acerra, F., Burnod, I., & de Schonen, S. (2002). Modeling aspects of face processing in early infancy. Devel-opmental Science, 5, 98–117.

Baron-Cohen, S. (1995). Mindblindness: An essay on Autism and theory of mind. Cambridge, MA: MIT Press.Bednar, J. A. (2003). The role of internally generated neural activity in newborn and infant face prefer-

ences. In O. Pascalis & A. Slater (Eds.), The development of face processing in infancy and early child-hood: Current perspectives (pp. 133–142). New York: Nova Science.

Bednar, J. A., & Miikkulainen, R. (2003). Learning innate face preferences. Neural Computation, 15, 1525–1557.

Black, J. E., Jones, T. A., Nelson, C. A., & Greenough, W. T. (1998). Neuronal plasticity and the developingbrain. In N. E. Alessi, J. T. Coyle, S. I. Harrison, & S. Eth (Eds.), Handbook of child and adolescent psy-chiatry, Vol. 6: Basic psychiatric science and treatment (pp. 31–53). New York: John Wiley.

de Haan, M., & Halit, H. (2001). Neural bases and development of face recognition during infancy. In A.Kalverboer & A. Gramsbergen (Eds.), Handbook of brain and behaviour in human development (pp.921–938). Dordrecht, Netherlands: Kluwer Academic.

de Schonen, S., & Mathivet, E. (1989). First come, Wrst served: A scenario about the development of hemi-spheric specialization in face recognition during infancy. European Bulletin of Cognitive Psychology, 9,3–44.

Diamond, R., & Carey, S. (1986). Why faces are not special: An eVect of expertise?. Journal of ExperimentalPsychology: General, 115, 107–117.

Elman, J. L., Bates, E. A., Johnson, M. H., KarmiloV-Smith, A., Parisi, D., & Plunkett, K. (1996). Rethink-ing innateness: A connectionist perspective on development. Cambridge, MA: MIT Press.

Farah, M. (2000). The cognitive neuroscience of vision. Oxford, UK: Blackwell.Farah, M. J., Rabinowitz, C., Quinn, G. E., & Liu, G. T. (2000). Early commitment of neural substrates for

face recognition. Cognitive Neuropsychology, 17, 117–123.Gauthier, I., & Logothetis, N. K. (2000). Is face recognition not so unique after all?. Cognitive Neuropsy-

chology, 17, 125–142.Greenough, W. T., & Black, J. E. (1992). Induction of brain structure by experience: Substrates for cogni-

tive development. In M. Gunnar & C. A. Nelson (Eds.), Minnesota Symposia on Child Psychology (Vol.24). Behavioral developmental neuroscience (pp. 35–52). Hillsdale, NJ: Lawrence Erlbaum.

Haaf, R. (1974). Complexity and facial resemblance as determinants of response to face-like stimuli by 5-and 10-week-old infants. Journal of Experimental Child Psychology, 18, 480–487.

Haaf, R., & Brown, C. (1976). Infants’ response to face-like patterns: Developmental changes between 10and 15 weeks of age. Journal of Experimental Child Psychology, 22, 155–160.

Haaf, R., Smith, P. H., & Smitley, S. (1983). Infant response to facelike patterns under Wxed trial andinfant-control procedures. Child Development, 54, 172–177.

Halit, H., de Haan, M., & Johnson, M. H. (2003). Cortical specialization for face processing: Face sensitiveevent-related potential components in 3- and 12-month-old infants. Neuroimage, 19, 1180–1193.

Johnson, M. H. (1993). Constraints on cortical plasticity. In M. J. Johnson (Ed.), Brain development andcognition: A reader (pp. 703–721). Cambridge, MA: Blackwell.

Johnson, M. H. (2000). Functional brain development in infants: Elements of an interactive specializationframework. Child Development, 71, 75–81.

Johnson, M. H., & de Haan, M. (2001). Developing cortical specialization for visual-cognitive function:The case of face recognition. In J. L. McClelland & R. S. Siegler (Eds.), Mechanisms of cognitive devel-opment: Behavioral and neural perspectives (pp. 253–270). Mahwah, NJ: Lawrence Erlbaum.


Johnson, M. H., Dziurawiec, S., Bartrip, J., & Morton, J. (1992). The eVects of movement of internal fea-tures on infants’ preferences for face-like stimuli. Infant Behavior and Development, 15, 129–136.

Johnson, M. H., Dziurawiec, S., Ellis, H., & Morton, J. (1991). Newborns’ preferential tracking of face-likestimuli and its subsequent decline. Cognition, 90, 1–19.

Johnson, M. H., & Morton, J. (1991). Biology and cognitive development: The case of face recognition.Oxford, UK: Blackwell.

Kanwisher, N. (2000). Domain speciWcity in face perception. Nature Neuroscience, 3, 759–763.Kleiner, K. A. (1987). Amplitude and phase spectra as indices of infants’ pattern preferences. Infant Behav-

ior and Development, 10, 49–59.Kleiner, K. A., & Banks, M. S. (1987). Stimulus energy does not account for 2-month-olds’ face prefer-

ences. Journal of Experimental Psychology: Human Perception and Performance, 13, 594–600.Koopman, P. R., & Ames, E. W. (1968). Infants’ preferences for facial arrangements: A failure to replicate.

Child Development, 39, 481–487.Macchi Cassia, V., Turati, C., & Simion, F. (2004). Can a nonspeciWc bias toward top-heavy patterns

explain newborns’ face preference?. Psychological Science, 15, 379–383.Macchi Cassia, V., Valenza, E., Pividori, D., & Simion, F. (2002). Facedness vs. non-speciWc structural prop-

erties: What is crucial in determining face preference at birth? Poster presented at the International Con-ference on Infant Studies, Toronto.

Nelson, C. A. (2001). The development and neural bases of face recognition. Infant and Child Development,10, 3–18.

Nelson, C. A. (2003). The development of face recognition reXects an experience-expectant and activity-dependent process. In O. Pascalis & A. Slater (Eds.), The development of face processing in infancy andearly childhood: Current perspectives (pp. 79–97). New York: Nova Science.

Quinn, P. C., Yahr, J., Kuhn, A., Slater, A. M., & Pascalis, O. (2002). Representation of the gender ofhuman faces by infants: A preference for female. Perception, 31, 1109–1121.

Simion, F., Macchi Cassia, V., Turati, C., & Valenza, E. (2001). The origins of face perception: SpeciWc vs.non-speciWc mechanisms. Infant and Child Development, 10, 59–65.

Simion, F., Macchi Cassia, V., Turati, C., & Valenza, E. (2003). Non-speciWc perceptual biases at the ori-gins of face processing. In O. Pascalis & A. Slater (Eds.), The development of face processing in infancyand early childhood: Current perspectives (pp. 13–25). New York: Nova Science.

Simion, F., Valenza, E., Macchi Cassia, V., Turati, C., & Umiltà, C. (2002). Newborns’ preference for up-down asymmetrical conWgurations. Developmental Science, 5, 427–434.

Tarr, M. J., & Gauthier, I. (2000). FFA: A Xexible fusiform area for subordinate-level visual processingautomatized by expertise. Nature Neuroscience, 8, 764–769.

Turati, C. (2004). Why faces are not special to newborns: An alternative account of the face preference.Current Directions in Psychological Science, 13, 5–8.

Turati, C., Simion, F., Milani, I., & Umiltà, C. (2002). Newborns’ preference for faces: What is crucial?Developmental Psychology, 38, 875–882.

Tzourio-Mazoyer, N., de Schonen, S., Crivello, F., Reutter, B., Aujard, Y., & Mazoyer, B. (2002). Neuralcorrelates of woman face processing by 2-month-old infants. Neuroimage, 15, 454–461.

Valenza, E., Simion, F., Macchi Cassia, V., & Umiltà, C. (1996). Face preference at birth. Journal of Exper-imental Psychology: Human Perception and Performance, 22, 892–903.

Three-month-olds’ visual preference for faces and its underlying visual processing mechanisms

Documents

Transcript of Three-month-olds’ visual preference for faces and its underlying visual processing mechanisms