tolerance of geometric distortions in infant's face recognition

5
Infant Behavior & Development 37 (2014) 16–20 Contents lists available at ScienceDirect Infant Behavior and Development Tolerance of geometric distortions in infant’s face recognition Wakayo Yamashita a,,1 , So Kanazawa b , Masami K. Yamaguchi c a Graduate School of Science and Engineering, Kagoshima University, 1-21-40 Korimoto, Kagoshima 890-0065, Japan b Department of Psychology, Japan Women’s University, 1-1-1 Nishi-Ikuta, Tama-ku, Kawasaki City, Kanagawa 214-8565, Japan c Department of Psychology, Chuo University, 742-1, Higashinakano, Hachioji, Tokyo 192-0393, Japan a r t i c l e i n f o Article history: Received 1 June 2013 Received in revised form 28 August 2013 Accepted 31 October 2013 Available online 5 December 2013 Keywords: Distortion Face recognition Infant a b s t r a c t The aim of the current study is to reveal the effect of global linear transformations (shear- ing, horizontal stretching, and vertical stretching) on the recognition of familiar faces (e.g., a mother’s face) in 6- to 7-month-old infants. In this experiment, we applied the global lin- ear transformations to both the infants’ own mother’s face and to a stranger’s face, and we tested infants’ preference between these faces. We found that only 7-month-old infants maintained preference for their own mother’s face during the presentation of vertical stretching, while the preference for the mother’s face disappeared during the presenta- tion of shearing or horizontal stretching. These findings suggest that 7-month-old infants might not recognize faces based on calculating the absolute distance between facial fea- tures, and that the vertical dimension of facial features might be more related to infants’ face recognition rather than the horizontal dimension. © 2013 Elsevier Inc. All rights reserved. 1. Introduction For infants, the person with whom they usually have the most frequently contact is their primary caregiver who is almost always their mother. Because infants have many opportunities to look at their mother’s face, an infant’s mother’s face is more special than other faces for that infant. Researchers have demonstrated that infants can distinguish their own mother’s face from strangers’ faces (Bushnell, Sai, & Mullin, 1989; Pascalis, de Schonen, Morton, Deruelle, & Fare-Grenet, 1995; Bushnell, 1982). Bushnell et al. (1989) have reported that even 4-day-olds newborns prefer to look at their mother’s face than at strangers’ faces (i.e., infants show preference for their own mother’s face). The critical cue might be external features such as hair line and facial contour at this age (Pascalis et al., 1995). From the age of 4 months, infants are able to process the internal facial features or configuration in order to differentiate between and recognize faces (Bushnell, 1982). To recognize familiar face, human adults use internal rather than external features that happen to change (Ellis, Shepherd, & Davies, 1979), and process configuration of the internal features (i.e., configural processing) (e.g., Brooks & Kemp, 2007; Mondloch, Le Grand, & Maurer, 2002). Developmental studies have proposed that 7- to 8-month-old infants process faces configurally based on the relationship of facial features (Cohen & Cashon, 2001; Schwarzer & Zauner, 2003). Schwarzer and Zauner (2003) exchanged only single internal features (e.g., eye/mouth) to create a composite face. They found that 8-month-old infants looked longer at a composite face in which either the eyes or mouth had been exchanged than at a familiar face. Face recognition has also been demonstrated to be unaffected or even enhanced by some types of systematic facial distortion (e.g., caricatures). More needs to be known about how recognition is achieved under various circumstances, and Corresponding author. Tel.: +81 99 285 3302; fax: +81 99 285 3302. E-mail addresses: [email protected], [email protected] (W. Yamashita). 1 This experiment was conducted when she belonged to Chuo University. 0163-6383/$ see front matter © 2013 Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.infbeh.2013.10.003

Upload: masami-k

Post on 30-Dec-2016

214 views

Category:

Documents


2 download

TRANSCRIPT

Infant Behavior & Development 37 (2014) 16– 20

Contents lists available at ScienceDirect

Infant Behavior and Development

Tolerance of geometric distortions in infant’s face recognition

Wakayo Yamashitaa,∗,1, So Kanazawab, Masami K. Yamaguchic

a Graduate School of Science and Engineering, Kagoshima University, 1-21-40 Korimoto, Kagoshima 890-0065, Japanb Department of Psychology, Japan Women’s University, 1-1-1 Nishi-Ikuta, Tama-ku, Kawasaki City, Kanagawa 214-8565, Japanc Department of Psychology, Chuo University, 742-1, Higashinakano, Hachioji, Tokyo 192-0393, Japan

a r t i c l e i n f o

Article history:Received 1 June 2013Received in revised form 28 August 2013Accepted 31 October 2013Available online 5 December 2013

Keywords:DistortionFace recognitionInfant

a b s t r a c t

The aim of the current study is to reveal the effect of global linear transformations (shear-ing, horizontal stretching, and vertical stretching) on the recognition of familiar faces (e.g.,a mother’s face) in 6- to 7-month-old infants. In this experiment, we applied the global lin-ear transformations to both the infants’ own mother’s face and to a stranger’s face, and wetested infants’ preference between these faces. We found that only 7-month-old infantsmaintained preference for their own mother’s face during the presentation of verticalstretching, while the preference for the mother’s face disappeared during the presenta-tion of shearing or horizontal stretching. These findings suggest that 7-month-old infantsmight not recognize faces based on calculating the absolute distance between facial fea-tures, and that the vertical dimension of facial features might be more related to infants’face recognition rather than the horizontal dimension.

© 2013 Elsevier Inc. All rights reserved.

1. Introduction

For infants, the person with whom they usually have the most frequently contact is their primary caregiver who is almostalways their mother. Because infants have many opportunities to look at their mother’s face, an infant’s mother’s face is morespecial than other faces for that infant. Researchers have demonstrated that infants can distinguish their own mother’s facefrom strangers’ faces (Bushnell, Sai, & Mullin, 1989; Pascalis, de Schonen, Morton, Deruelle, & Fare-Grenet, 1995; Bushnell,1982). Bushnell et al. (1989) have reported that even 4-day-olds newborns prefer to look at their mother’s face than atstrangers’ faces (i.e., infants show preference for their own mother’s face). The critical cue might be external features suchas hair line and facial contour at this age (Pascalis et al., 1995). From the age of 4 months, infants are able to process theinternal facial features or configuration in order to differentiate between and recognize faces (Bushnell, 1982).

To recognize familiar face, human adults use internal rather than external features that happen to change (Ellis, Shepherd,& Davies, 1979), and process configuration of the internal features (i.e., configural processing) (e.g., Brooks & Kemp, 2007;Mondloch, Le Grand, & Maurer, 2002). Developmental studies have proposed that 7- to 8-month-old infants process facesconfigurally based on the relationship of facial features (Cohen & Cashon, 2001; Schwarzer & Zauner, 2003). Schwarzerand Zauner (2003) exchanged only single internal features (e.g., eye/mouth) to create a composite face. They found that8-month-old infants looked longer at a composite face in which either the eyes or mouth had been exchanged than at a

familiar face.

Face recognition has also been demonstrated to be unaffected or even enhanced by some types of systematic facialdistortion (e.g., caricatures). More needs to be known about how recognition is achieved under various circumstances, and

∗ Corresponding author. Tel.: +81 99 285 3302; fax: +81 99 285 3302.E-mail addresses: [email protected], [email protected] (W. Yamashita).

1 This experiment was conducted when she belonged to Chuo University.

0163-6383/$ – see front matter © 2013 Elsevier Inc. All rights reserved.http://dx.doi.org/10.1016/j.infbeh.2013.10.003

wResbro(cWtssv

otiftBea

2

2

aidn

2

ao4so

2

vTf

ncvwta

sd

W. Yamashita et al. / Infant Behavior & Development 37 (2014) 16– 20 17

hat kinds of distortion can and cannot be tolerated by the processes underlying recognition. Hole, George, Eaves, andasek (2002) examined the effects of different types of distortion on familiar face recognition with adult participants. Theyxamined the effect of distortion by using three global linear transformations (shearing, horizontal stretching, and verticaltretching). Their results showed that recognition was impaired by the shearing or horizontal stretching of familiar faces,ut unaffected by the vertical stretching of faces to twice their normal height. For infants’ recognition of familiar faces, noesearchers had previously examined the effect of these global linear transformations. In this study, we examined the effectf the global linear transformations (shearing, horizontal stretching, vertical stretching) on the recognition of familiar facemother’s face) in 6- to 7-month-old infants that have configural processing face perception. To begin this experiment, weonducted a simple test to determine whether infants discriminate between their own mother’s face and a stranger’s face.

e presented the facial images side by side, and examined the preference for the infant’s own mother’s face by measuringhe infant’s looking time. Considering Hole et al.’s results (2002), if global linear transformation by shearing or horizontaltretching affects the infant’s face recognition, and if vertical stretching does not, the presentation of sheared or horizontallytretched images should decrease the infant’s preference for his/her own mother’s face. Moreover, the presentation ofertically stretched images should maintain the infant’s preference for his/her own mother’s face.

To examine whether the effects described above were face-specific, we tested each infant’s preference for the presentationf inverted face images. Recognition of inverted faces is surprisingly poor compared to upright faces. Yin (1969) reportedhat the effect of inversion is larger for the recognition of faces than for the recognition of objects. This effect exists even innfancy. Turati, Sangrigoli, Ruely, & de Schonen (2004) demonstrated that 4-month-olds process faces differently when theaces are upright as opposed to when they are inverted. Infants recognized upright faces better than inverted faces whenhe faces were shown in a variety of poses, and when the infants were required to recognize a new pose of the same face.hatt, Bertin, Hayden, and Reed (2005) also showed that 5-month-olds could detect changes in the spacing between theyes when a face was shown upright but not when it was inverted. These reports imply that our subjects would not showny preference when faces are presented inverted.

. Methods

.1. Participants

Participants consisted of twelve 6 month-old infants (mean age = 183.1 days, range = 167–191 days, 5 females, 7 males),nd twelve 7 month-old infants (mean age = 209.8 days, range = 195–224 days, 6 females, 6 males). All were healthy Japanesenfants who had a birth weight greater than 2500 g. An additional 21 infants were tested, but were excluded from the analysisue to fussiness (1) or no preference for his/her mother’s face (20). The participants were recruited using advertisements inewspapers.

.2. Apparatus

All stimuli were displayed on a 22-in. CRT monitor (Mitsubishi Diamond Pro 2070SB) controlled by a computer. The infantnd the CRT monitor were located inside an enclosure which was made of iron poles and covered with cloth. Each infant satn his or her parent’s lap in front of the CRT monitor. The infant’s viewing distance from the monitor was approximately0 cm. There were two loudspeakers, one on either side of the CRT monitor. There was a CCD camera just below the monitorcreen. Throughout the experiment, the infant’s behavior was videotaped through this camera. The experimenter couldbserve the infant’s behavior via a TV monitor connected to the CCD camera.

.3. Stimuli

We used full-color photographs of each infant’s own mother’s faces. Each mother’s face was photographed in a frontaliew against a white background under the same lighting conditions using a color digital camera (Canon IXY digital 920IS,okyo, Japan). These photographs were controlled for neutrality (no smile or frown) and gaze direction. For the stranger’sace, we used a different mother’s face.

We set face images in four conditions (Fig. 1): normal, shearing, horizontal stretching, and vertical stretching. For theormal condition, original face images were used. These original images were transformed into a trapezoid whose acuteorners were at angles of approximately 45◦ for the shearing condition. The original images were stretched horizontally orertically to twice their original width or height for the horizontal stretching or vertical stretching condition, respectively,hile preserving their original height or width. The horizontally and vertically stretched images were scaled down to 71%

o make them approximately same size as the normal condition’s face images. We presented each condition both upright

nd inverted.

The approximate size of viewed stimuli in visual angle were 7.9 × 10.6◦ in the normal condition, 14.8 × 13.0◦ in thehearing condition, 11.3 × 7.5◦ in the horizontal stretching condition, and 5.7 × 15.1◦ in the vertical stretching condition. Theistance between the left and right images was about 14.5◦. The background was a homogeneous white field.

18 W. Yamashita et al. / Infant Behavior & Development 37 (2014) 16– 20

Fig. 1. Examples of the manipulations in our experiment: (A) normal, (B) shear, (C) horizontal stretch and (D) vertical stretch.

2.4. Procedure

To test each infant’s discrimination of his/her own mother’s face from a stranger’s face, a preferential–looking paradigmwas used. Prior to each trial, a cartoon fixation figure, accompanied by a short sound, was presented at the center of themonitor. When the infant looked at the fixation figure, the experimenter initiated the trial.

In each trial, stimuli were presented for 10 s. The presentation time of the stimuli was fixed regardless of whether theinfants looked or not. For each trial, the infant’s own mother’s face image and a stranger’s face image were presented sideby side. We defined the infant’s own mother’s face image as the target stimulus. The positions of the two face images werecounterbalanced across participants in the first trial and reversed in the second trial. Every infant underwent two trials foreach of the eight combinations of the condition and orientation types. The order of presentation of the eight combinationswas randomly exchanged across infants.

One observer, who was unaware of the stimulus identity, measured the infants’ fixations to the left and right sides ofthe display based on video recordings. Only the infants’ looking behavior was visible in the video. Although the observercould not see the stimulus, he or she was aware of the timing of the beginning and the end of each trial from the beep soundthat was presented at those times. To compute the inter-observer agreement, a second observer’s measurement of infant’slooking time was obtained from 25% of the total data. Inter-observer agreement was r = 0.927 across this study.

3. Results

For the 6-month-olds, the mean total looking times over the two trials of each transformation were 16.3 s (SE = 0.8) in the upright normal condition, 16.6 s(SE = 0.8) in the upright shearing, 14.6 s (SE = 1.3) in the upright horizontal stretching, 16.4 s (SE = 0.8) in the upright vertical stretching, 15.9 s (SE = 0.8) in theinverted normal condition, 17.1 s (SE = 0.6) in the inverted shearing, 14.4 s (SE = 1.0) in the inverted horizontal stretching, and 15.9 s (SE = 1.1) in the invertedvertical stretching. For the 7-month-olds, the mean total looking times over the two trials of each transformation were 17.2 s (SE = 0.7) in the upright normalcondition, 16.6 s (SE = 0.7) in the upright shearing, 17.2 s (SE = 0.7) in the upright horizontal stretching, 15.9 s (SE = 0.9) in the upright vertical stretching, 17.0 s(SE = 0.9) in the inverted normal condition, 15.9 s (SE = 0.7) in the inverted shearing, 16.2 s (SE = 0.8) in the inverted horizontal stretching, and 15.9 s (SE = 0.7)in the inverted vertical stretching. To examine the effect of age groups, conditions and orientations on the total looking time, a three-way ANOVA of age (6months/7 months) × condition (normal/shear/horizontal stretching/vertical stretching) × orientation (upright/inverted) did not revealed any main effects(age: F(1, 22) = 0.471, p = .500, condition: F(3, 66) = 1.306, p = .280, orientation: F(1, 22) = 1.556, p = .225) or interactions (age × condition × orientation: F(3,66) = 0.702, p = .555, age × condition: F(3, 66) = 2.362, p = .079, age × orientation: F(1, 22) = 0.366, p = .568, condition × orientation: F(3, 66) = 0.228, p = .877).We calculated individual preference scores, which is the ratio of the looking time at the infant’s own mother’s face to the total looking time over the twotrials in each experimental condition, for each infant.

Fig. 2 shows the mean preference for the infant’s own mother’s face in each experimental condition. To determine whether infants preferred their ownmother’s faces, we conducted two-tailed one-sample t-tests (versus chance) with a Bonferroni correction ( ̨ level of .05/8 = .006) on the preference for theinfants’ own mother’s face. For the 6-month-olds, the results revealed that the preference scores in the upright normal condition were significantly greaterthan the chance level [t(11) = 4.55, p = .0008, d = 1.86]; however, no differences were shown in the other conditions [upright shearing: t(11) = 0.11, uprighthorizontal stretching: t(11) = −0.20, upright vertical stretching: t(11) = −0.37]. For the 7-month-olds, the results revealed that the preference scores both inthe upright normal and vertical stretching conditions were significantly greater than the chance level [upright normal: t(11) = 4.29, p = .0013, d = 1.75, uprightvertical stretching: t(11) = 4.21, p = .0015, d = 1.72]; however, no differences were shown in the other conditions [upright shearing: t(11) = 0.67, uprighthorizontal stretching: t(11) = 0.31]. For 6- and 7-month-olds, no differences were shown in any inverted conditions [6-month-olds, normal: t(11) = 0.47;shearing: t(11) = −1.43; horizontal stretching: t(11) = −0.19; vertical stretching: t(11) = −0.67, 7-month-olds, normal: t(11) = 1.86; shearing: t(11) = −0.81;horizontal stretching: t(11) = −1.24; vertical stretching: t(11) = 1.17]. These results showed that for the only 7 month-old infants, the presentation ofshearing or horizontal stretching images decreased of the infant’s preference for their own mother’s face and the presentation of vertical stretching

images maintained the infant’s preference for their own mother’s face, and our subjects showed no preferences at the presentation of inverted faceimages.

To examine the effect of age groups, conditions and orientations on preference for the infants’ own mother’s face, we conducted a three-way ANOVA ofage (6 months/7 months) × condition (normal/shear/horizontal stretching/vertical stretching) × orientation (upright/inverted) to the preference scores. TheANOVA revealed significant main effects of condition (F(3, 66) = 6.505, p = .0006) and orientation (F(1, 22) = 6.870, p = .0156). No other effect or interactionapproached significance (age: F(1, 22) = 2.829, p = .107, age × condition × orientation: F(3, 66) = 0.552, p = .649, age × condition: F(3, 66) = 1.730, p = .169,age × orientation: F(1, 22) = 0.231, p = .636, condition × orientation: F(3, 66) = 0.985, p = .405).

W. Yamashita et al. / Infant Behavior & Development 37 (2014) 16– 20 19

0.40

0.45

0.50

0.55

0.60

0.65

ver�cal

stretch

horizontal

stretchshearoriginal

0.40

0.45

0.50

0.55

0.60

0.65

ver�cal

stretch

horizontal

stretchshearoriginal

7 month-old infants6 month-old infants

Pref

eren

ce fo

r ow

n m

othe

r’ s f

ace * **

Fi*

4

tvsstiocrwYhim

tptfsifh

A

nYrf

R

B

B

BB

ig. 2. Mean preference score for infants’ own mother’s face. The black bar denotes the preference score for the mother’s face in the presentation of uprightmages, and the gray bar denotes the preference score for the mother’s face in the presentation of inverted images. Error bars show mean standard errors.p < .05 (after Bonferroni’s correction for multiple comparison).

. Discussion

In this study, we examined whether 6- to 7-month-old infants preferred to their own mother’s face to stranger’s faces usinghe preferential looking technique, even if the infants’ mother’s face was transformed (via shearing, horizontal stretching orertical stretching). Considering Hole et al.’s results (2002), we hypothesized that the presentation of shearing or horizontaltretching images would decrease the infants’ preference for their own mother’s face, and that the presentation of verticaltretching images would maintain the infants’ preference for their own mother’s face. The results of our experiments showedhat while infants preferred their own mother’s face to stranger’s faces for the presentation of normal images, 6-month-oldnfants did not prefer them for the presentation of any transformed images. However, 7-month-old infants preferred theirwn mother’s face to stranger’s faces only for the presentation of vertical stretching. Our results for 7-month-old infants areonsistent with Hole et al.’s study (2002). This suggests that at around 7 months of age, infant might develop adult-like faceecognition which includes the tolerance of image transformation. In our experiment, to examine whether this mechanismas face-specific, we tested each infant’s preference for his/her own mother’s faces using inverted presentations. Consideringin’s study (1995), we hypothesized that the presentation of inverted face images would decrease the infant’s preference foris/her own mother’s face. Our results showed that 6–7 month-old infants had no preference for their own mother’s faces

n the inverted presentation. These results suggest that infants’ recognition of face images with global linear transformationight be a special response to faces.According to Hole et al.’s study (2002), the tolerance of spatial distortions in human adult’s face recognition suggests

hat configural information is used as a basis for face recognition and is unlikely to involve information about the absoluteosition of facial features relative to each other. The vertical/horizontal stretching images in our experiments were stretchedo only twice their original height/width. If infants could recognize faces based on simply calculating the distance betweenacial features, infants would not prefer their own mother’s face to a stranger’s faces for both the vertical and horizontaltretching images. The results showed that infants had no preference only for the presentation of the horizontal stretchingmages. Our results indicate that infants might not recognize faces based on calculating the absolute distance betweenacial features, and that the vertical dimension of facial features might be more related to face recognition rather than theorizontal dimension.

cknowledgements

This research was supported by a Grant-in-Aid for Scientific Research on Innovative Areas, “Face Perception and Recog-ition” from MEXT KAKENHI (20119002), and a Grant-in-Aid for Scientific Research (A) (21243041) and a Grant-in-Aid foroung Scientists (B) (23700323) from Japan Society for the Promotion of Science. We thank to Hiroko Ichikawa, Aki Tsu-uhara, Emi Nakato, Megumi Kobayashi, Jiale Yang, Yuka Yamazaki, Yuna Inada, Kazuki Sato, Yuiko Sakuta and Ahn Jooheeor their help in collecting the data.

eferences

hatt, R. S., Bertin, E., Hayden, A., & Reed, A. (2005). Face processing in infancy: Developmental changes in the use of different kinds of relational information.

Child Development: 76., (1), 169–181.

rooks, K. R., & Kemp, R. I. (2007). Sensitivity to feature displacement in familiar and unfamiliar faces: Beyond the internal/external feature distinction.Perception: 36., (11), 1646–1659.

ushnell, I. W. R., Sai, F., & Mullin, J. T. (1989). Neonatal recognition of the mother’s face. British Journal of Developmental Psychology: 7., (1), 3–15.ushnell, I. W. R. (1982). Discrimination of faces by young infants. Journal of Experimental Child Psychology: 33., (2), 298–308.

20 W. Yamashita et al. / Infant Behavior & Development 37 (2014) 16– 20

Cohen, L. B., & Cashon, C. H. (2001). Do 7-month-old infants process independent features or facial configurations? Infant and Child Development: 10., (1/2),83–92.

Ellis, H. D., Shepherd, J. W., & Davies, G. M. (1979). Identification of familiar and unfamiliar faces from internal and external features: Some implications fortheories of face recognition. Perception: 8., (4), 431–439.

Hole, G. J., George, P. A., Eaves, K., & Rasek, A. (2002). Effects of geometric distortions on face-recognition performance. Perception: 31., (10), 1221–1240.Mondloch, C. J., Le Grand, R., & Maurer, D. (2002). Configural face processing develops more slowly than featural face processing. Perception: 31., (5), 553–566.Pascalis, O., de Schonen, S., Morton, J., Deruelle, C., & Fare-Grenet, M. (1995). Mother’s face recognition by neonates: A replication and an extension. Infant

Behavior and Development: 18., 79–85.Schwarzer, G., & Zauner, N. (2003). Face processing in 8-month-old infants: Evidence for configural and analytical processing. Vision Research: 43., (26),

2783–2793.Turati, C., Sangrigoli, S., Ruely, J., & de Schonen, S. (2004). Evidence of the face inversion effect in 4-month-old infants. Infancy: 6., (2), 275–297.Yin, R. K. (1969). Looking at upside-down faces. Journal of Experimental Psychology: 81., (1), 141–145.