Content uploaded by Sarina Hui-Lin Chien
Author content
All content in this area was uploaded by Sarina Hui-Lin Chien on Jan 11, 2016
Content may be subject to copyright.
中華心理學刊 民101，54卷，1期，95-114
Chinese Journal of Psychology 2012, Vol.54, No.1, 95-114
© 2012 Airiti Press Inc. & Taiwanese Psychology Association
No More Top-Heavy Bias: On Early Specialization Process for
Face and Race in Infants
Sarina Hui-Lin Chien and Hsin Yueh Hsu
Graduate Institute of Neural & Cognitive Sciences, China Medical University
Neonates display reliable visual preferences for human faces and face-like stimuli, which has been taken as strong
evidence for an innate domain-specic bias toward faces. Alternatively, neonatal face preference can be explained by an
innate non-specic top-heavy conguration bias on the basis that faces are inherently top-heavy. This review article aims
to address the early specialization process for face and race in Taiwanese infants based our recent work. In the rst two
sections, we will review the classic ndings on the neonatal face preference research. Three broad theories, the sensory
hypothesis by linear system approach in the late 80s, the innate domain-specic representation hypothesis in the early
90s, and the non-specic top-heavy bias hypothesis in the last decade, will be addressed and compared. In the third
section, we further explored some deeper issues regarding the top-heavy conguration bias hypothesis as well as our
recent follow-up work. Using a forced-choice novelty preference method, we found that 2.5- to 4.5-month-old infants
showed signicant and equal novelty preferences, suggesting a reliable discriminability between the two congurations
and a disappearance of the top-heavy bias. Moreover, using an eye-tracker, we investigated whether the top-heavy bias
is still present in 3- to 5.5-month olds and in adults as a comparison group, and found no evidence for the top-heavy
bias in both infants and adults. In the forth section, we illustrated the idea of an early specialization process for human
face and race. Several recent developmental cognitive neuroscience studies on the infant brain as well as the behavioral
studies on the other-race-effect in Caucasian and Taiwanese infants were reviewed. Taken together, our position on
early face processing is in line with experience-expectant view; this view considers the brain specialization as emerging
gradually from the interaction between small innate constraints and the critical input provided by the species-typical or
race-typical environment.
Keywords: experience-expectant, face processing, innate bias, newborn infants, other-race effect
One of the most important tasks in modern
developmental neuroscience research is to uncover how
domain-specific knowledge is specialized in the human
brain during the course of development. In this respect,
face processing is a particularly appealing topic because
“faces” are biologically significant stimuli, and several
lines of evidence suggest that face processing in adults
rests on an anatomically and/or a functionally specialized
neural circuit (e.g., Farah, 2000; Kanwisher, 2000). Just
a glance at a face, one can be infused with tremendous
Received: March 3, 2011; First Revision: June 25, 2011; Second Revision: August 11, 2011; Accepted: August 19, 2011
Correspondence Author: Sarina Hui-Lin Chien (sarinachien@mail.cmu.edu.tw) No. 91, Hsueh-Shih Rd., Taichung 40402, Taiwan.
Graduate Institute of Neural & Cognitive Sciences, College of Life Sciences, China Medical University
Acknowledgements: This paper was supported in part by National Science Council grants NSC 98-2410-H-039-002 and NSC
99-2410-H-039-003-MY3 to Dr. S. H. L. Chien. We would like to thank President Jong-Tsun Huang of China
Medical University, Dr. Lin Chen of Chinese Academy of Sciences, and to Dr. Su-Ling Yeh of National Taiwan
University for their devotion and enthusiasm for making the special post-conference issue possible. We would also
like to thank the editor and two reviewers for their invaluable comments on the manuscript for the Special Issue:
Highlights from the 2010 Cross-Strait Forum on the Joint Development of Cognitive Science Studies.
08-Chien.indd 95 2012/3/22 下午 02:45:27
96 Sarina Hui-Lin Chien Hsin Yueh Hsu
amount of social, visual, and cognitive information
(e.g., a person’s identity, age, sex, emotional state, or
even intention) that are important for interpersonal
interactions. However, how do these remarkable face-
processing capacities emerge and what might determine
this functional specialization have been hotly debated
for decad es. This review article was divided into four
sections. In the rst section, we started out highlighting
the classic findings on the neonatal face preference
studies. In the second section, we discussed three
broad views, the sensory hypothesis, the configuration
hypothesis, and the non-specific perceptual bias
hypothesis to explain neonatal face preference. In the
third section, we further explored the issues regarding
the d eve lop men tal trajectory a nd the stren gth of the
top-heavy configuration bias. In the forth section, we
elaborated the idea of an early specialization process
for human face and race. Several recent developmental
cognitive neuroscience studies on the infant brain and the
behavioral studies on the other-race-effect in Caucasian
and Taiwanese infants will be reviewed as well.
Newborns Show Looking Preferences
for Schematic and Real Faces
Human neonates’ visual capacity is very limited at
birth; the visual system undergoes rapid development
during the first six months of life and continues to
improve gradually over several years (Atkinson, 1984;
Johnson, 1997; Maurer & Maurer, 1988; Teller, 1997).
Despite of their relatively poor vision at birth, newborn
infants attend to human faces or face-like stimuli within
hours after birth. Using the preferential looking paradigm,
Fantz’s (1963) classic study was the first to report that
newborn infants showed longer fixation for schematic
faces than other stimuli. Subsequent studies supported
and extended such finding; the early face preference
phenomenon was demonstrated with both static and
moving stimuli (Goren, Sarty, & Wu, 1975; Johnson &
Morton, 1991), and with both schematic and veridical
images of faces (Macchi Cassia, Turati, & Simion, 2004;
Valenza, Simion, Macchi Cassia, & Umiltà, 1996). Other
studies revealed that newborns preferred looking at faces
versus other non-face stimuli (Easterbrook, Kisilevsky,
Hains, & Muir, 1999; Maurer & Barrera, 1981) and even
looked longer at attractive faces rated by adults (Slater et
al., 1998).
In addition to schematic faces, newborn infants
look longer at and orient more frequently toward the
familiar faces such as their mothers’ faces. One of the
first studies demonstrating this intriguing phenomenon
revealed that neonates preferred to look at their mother’s
face over a female stranger’s face (Field, Cohen, Garcia,
& Greenberg, 1984). Subsequent studies controlling for
maternal body odors, hair colors, and face brightness also
confirmed Field et al.’s (1984) findings that newborns
preferred their mother ’s face (Bushnell, Sai, & Mullin,
1989; Pascalis, de Schonen, Morton, Deruelle, & Fabre-
Grenet, 1995; Walton, Bower, & Bower, 1992).
What May Account for the Neonatal
Face Preference?
It is unquestionable that humans are born with a
predisposition o r bias to attend to face. However, the
nature of this attentional bias is still in dispute. What
exactly is the information in faces and face-like stimuli
that drives newborns to look longer? Is it because we are
born with an innate face-specific processing mechanism
as Farah (2000) strongly proclaimed? Or, a more general
property of perceptual processing can sufciently trigger
newborns’ attention to faces (see Figure 1)? Three broad
theories, the sensory hypothesis by linear system model in
the late 80s, the innate domain-specic face representation
hypothesis in the early 90s, and the recent non-specic top-
heavy bias hypothesis, will be reviewed in this section.
The “Sensory” Hypothesis versus the “Social”
Hypothesis
One of the most influential model explaining
neonatal visual preferences, the linear system model by
Banks and Ginsburg (1985), has successfully illustrated
newborn’s preferences for a variety of visual patterns,
given their limited visual capacities (Banks & Ginsberg,
08-Chien.indd 96 2012/3/22 下午 02:45:27
Early Specialization for Face and Race 97
1985; Banks & Salapatek, 1981). Based on this model, the
sensory hypothesis proposed that infants are attracted to
faces because faces posses low spatial frequency and high
contrast components that are more “visible” to infants
eyes; this idea is contrasted with the social hypothesis
which stated that infants have some built-in sensitivity to
the conguration of faces (Kleiner & Banks, 1987). For
any given visual pattern, two functions may be derived:
the amplitude spectrum (the amplitude and orientation
of the component spatial frequencies) and the ph ase
spectrum (the phases and orientation of the components).
In the lin ear systems mo del, the amplitu de spectrum
of any stimulus pattern, collapsed over orientation, is
filtered through the contrast sensitivity function (CSF)
of the appropriate age group. For a newborn infant, this
effectively removes all information at spatial frequencies
greater than abo ut 2 cycles per degree (cpd). What is
crucial is the extent to which the contrast in the remaining
frequencies exceeds the infant’s contrast threshold. As
newborns are most sensitive to frequencies between
0.2 and 0.5 cpd (Atkinson, Braddick, & French, 1979),
energy in that range will be most effective in attracting
the infant’s attention.
To answer whether the sensory hypothesis works for
faces, Kleiner (1987) designed a clever study in which
she used two stimuli, a schematic face and a lattice (see
Figure 2A and 2B). Using the Fourier transform, she
obtained the amplitude and phase spectrum of each, and
then generated two new images combining the amplitude
Figure 1. Newborn infants show looking preferences for real faces, schematic faces, or face-like patterns. The
three gures in the upper row, from left to right, are the photographed face of Taiwanese celebrity
(Ms. Chi-Ling Lin), the cartoon character (Chibi Maruko), and the face-like pattern (adopted from
Turati, 2004, respectively). To better illustrate that the non-specic “top-heavy” conguration exists
in all three types of gures, the versions in the bottom row are the same ones with the outer contour
and the hairline removed.
08-Chien.indd 97 2012/3/22 下午 02:45:27
98 Sarina Hui-Lin Chien Hsin Yueh Hsu
spectrum of the lattice with the phase spectrum of the
face (see Figure 2C), and vice versa (Figure 2D). If the
sensory hypothesis is correct, that the infants’ preferences
are governed by greater total amplitude of stimulation,
patterns with the amplitude spectrum of the face (Figure
2A and 2D) should be preferred to patterns with the
amplitude spectrum of the lattice (Figure 2B and 2C)
because phase shall be irrelevant. On the contrary, if
the social hypothesis is correct that preferences involve
face structure, patterns with the phase spectrum of the
face (Figure 2A and 2C) should receive most attention.
Results with newborns were mixed. Figure 2D was
preferred to Figure 2C, indicating the irrelevance of
phase (structural) information. However, in trials when
Figure 2A was pitting against Figure 2D (with the same
amplitude spectrum of the face), newborns looked longer
at Figure 2A, indicating a preference for the “phase”
information of the face when the amplitudes were equated
(Kleiner, 1987). Thus, this study demonstrated that indeed
the “visibility” of a pattern is crucial as the sensory
hypothesis proclaimed; however, it is also true that some
built-in sensitivity for the structure of faces is present at
birth.
The Innate Domain-Specic Face
Representation Views
Since the time that Fantz (1963) rst demonstrated
th a t in f ants prefe rred schem atic face than non- f ace
stimuli, it has been a predominant conjecture that neonatal
face perception is driven by an innate bias toward the
configuration or structure of human face (e.g., Farah,
2000; Karmiloff-Smith, 1996). In other words, through
the mechani sm of e volut ion, human n eonates’ ability
A. B.
C. D.
Figure 2. Stimuli used to test phase and amplitude information in infants’ face preference. Panel (A): The
original image of that contains both the amplitude and phase spectra of a schematic face. Panel (B):
The original image of that contains both the amplitude and phase spectra of a lattice. Panel (C): The
combined image of with the amplitude spectrum of the lattice and the phase spectrum of the face.
Panel (D): The combined image of with the amplitude spectrum of the face and the phase spectrum
of the lattice. (adopted from Kellman & Arterberry, 2000).
08-Chien.indd 98 2012/3/22 下午 02:45:28
Early Specialization for Face and Race 99
to process faces can be pre-wired without experience
because faces are highly signicant biological stimuli. In
line with this view, the most inuential model proposed
by Morton and Johnson (1991) hypothesized a two-
stage CONSPEC/CONLERN mechanism for early face
processing. This model argued that infants are born
with some information about the structure of faces. This
information is in the form of a face-detecting subcortical
visuomotor device called CONSPEC (short for
conspecics). The preferential response to face observed
in newborns is mediated primarily (but not exclusively)
by the “CONSPEC,” whereas later developed abilities to
recognize individual faces and facial expressions on the
basis of internal features are inuenced more by cortical
visuomotor pathways, the “CONLERN.” The term
CONLERN is used to describe the cortical systems that
young infants learn about particular faces of members
of their species. CONLERN is simply a system that
acquires and retains specic information about the visual
characteristics of conspecifics (see Johnson & Morton,
1991 for a review).
Principal support for the CONSPEC of Morton and
Johnson’s (1991) model comes from reports that newborn
infants preferentially look at schematic faces over
scrambled and blank faces (Goren et al., 1975; Johnson,
Dziurawiec, Ellis, & Morton, 1991) with a tracking
paradigm (see Figure 3). In these studies, researchers
mounted stimuli on paddles and steadily moved them
across the newborns’ visual eld while recording eye and
head movements. Under these testing conditions, both
eye and head movements were found to be greater for
schematic face stimuli, thus demonstrating that newborns
show more interest in face-like patterns relative to other
stimuli. The finding that newborns only exhibit a face
preference when stimuli are presented in the temporal
visual field, which feeds to subcortical pathways1,
is consistent with the prediction of the subcortical
CONSPEC module. In contrast, when stimuli are
presented in the nasal visual eld, which feeds to cortical
pathways, a face preference is not observed (Simion,
Valenza, Umiltà, & Della Barba, 1995).
Parallel with Johnson and Morton (1991) model,
de Schonen and Mathivet (1989) also proposed the
idea that newbo rns ’ p referenc e for faces is mediate d
by subcortical circuits whose only purpose is to orient
newborns’ gaze toward faces. By ensuring that infants
have visual experience with this type of stimuli, the
subcortical structures support the gradual emergence of
specialized cortical circuits subserving face processing in
adults. This position combines the notion that evolution
adaptively provided newborns with a device specically
tuned to faces with the view that visual experience plays
a crucial role in the normal development of sophisticated
face processing abilities. At a deeper level, Johnson and
Morton’s (1991) and de Schonen and Mathivet’s (1989)
models are rooted in a similar cognitive neuroscience
Face Scrambled Blank
Figure 3. The three stimuli used by Johnson and Morton (1991) in their replication of Goren et al. (1975)
original study: Newborn infants tracked the face in preference to the scrambled and the blank.
08-Chien.indd 99 2012/3/22 下午 02:45:28
100 Sarina Hui-Lin Chien Hsin Yueh Hsu
perspective, in which domain-specic cognitive structures
are thought to emerge gradually from the interaction
between small innate constraints and the vast input
structure provided by the species-typical environment
(Elman et al., 1996).
The Non-specic Perceptual Bias Hypothesis:
Newborns Prefer Top-Heavy Conguration
Within the past ten years, a research group from
Padua, Italy, has established an alternative hypothesis
suggesting a more simplistic account for neonatal face
preference. The main idea is that newborn face perception
is driven by functional properties of the visual system
that is not specic to faces per se. In other words, some
domain-general constraints of the visual system might
be sufficient to explain newborns’ preference toward
faces. They proposed two kinds of non-specic structural
properties that are typically seen in faces. The first,
termed up-down asymmetry, or the top-heavy bias, refers
to the presence of more patterning in the upper than in
the lower part of the configuration (Simion, Valenza,
Macchi Cassia, Turati, & Umiltà, 2002). The second,
termed congruency, refers to the existence of a congruent
spatial relation between the spatial disposition of the
inner features and the shape of the outer contour, with the
greater number of inner elements located in the widest
portion of the configuration (Macchi Cassia, Valenza,
Pividori, & Simion, 2002). Both properties characterized
the face-like patterns used in almost all the experiments
in which newborns’ face preference was demonstrated.
Because faces are up-down asymmetrical (two eyes in the
upper and one mouth in the lower part), thus, neonatal
face preference actually reects a non-specic perceptual
preference for up-down asymmetry (i.e., top-heavy bias),
rather than an innate bias for face-specic representation
(Turati, 2004).
A series of studies have been reported to support the
non-specific bias hypothesis. As illstrated in Figure 4A,
Simion et al. (2002) demonstrated that newborns showed
looki ng pre fer ence fo r t op- hea vy geometr ic pat ter ns
(e.g., T-shape, 甲-shape, ㄇ-shape, see Figure 4A) with
more elements in the upper part than the bottom-heavy
ones. Subsequently, Turati, Simion, Milani, and Umiltà
(2002) conducted a series of experiments using a set of
face-like stimuli with the same number of inner elements
and the same head-shaped contour usually displayed in
face-like configurations (see Figure 4B). They directly
compared infants’ reactions to a face-like configuration
and a top-heavy non-face pattern; the results indicated
that newborns did not show any visual preference
(Figure 4B, middle panel). In another comparison, a
pattern with the inner elements positioned in a face-like
arrangement but globally located lower than in a normal
face was contrasted with a top-heavy pattern in which
the elements were in the upper portion but did not form a
face conguration (Figure 4B, bottom panel). Newborns
preferred the top-heavy pattern (i.e., the right one), even
though it did not represent a face. A very similar pattern
of results has also been replicated with images of real
faces (Macchi Cassia et al., 2004). In summary, this line
of work rmly supports the idea that the innate preference
for some non-specific structural properties may be
sufcient to account for neonatal face preference.
Interim Summary
In su m m ary, the thr e e vi ews on neo n a t a l face
preference described above can be seen as complementary
rather than conicting in some way, as they converge to
describe a highly specialized process that is functional
from the very first moments of life. The concept of
visibility of the sensory hypothesis is important; the
stimuli have to be at least “visible” (above infant’s
threshold) to attract infants’ attention. However, when the
stimuli are equally visible, it is shown that some built-
in sensitivity for face-like configuration is present at
birth. What exactly is there in a face that attracts infants’
attention is the heart of the debate. Is it the “faceness”
itself or something else? The difference between Johnson
and Morton’s CONSPEC/CONLERN model and Turati’s
top-heavy bias hypothesis lies in the nature of the initial
representation. The former hypothesizes a domain-
specic representational bias (i.e., real faces or face-like
stimuli) at the very beginning of life, and its neural basis
rests in the subcortical level rather than the cortical level.
The latter proposed that it is not necessary to assume
08-Chien.indd 100 2012/3/22 下午 02:45:28
Early Specialization for Face and Race 101
the existence of a specific innate cortical or subcortical
representational bias toward faces. The earliest basis
of face specialization appears to rest on the general
functioning of the visual system (something as simple as
a top-heavy configuration), which constrains newborns
to attend to certain broad classes of visual stimuli that
include faces.
It is still an on-going debate whether the initial
representation is domain-specific or non-specific.
Nevertheless, there is at least a certain degree of
consensus that newborn infants demonstrate a visual
preference for both real and schematic human faces over
al mos t a ny othe r c ate gory of stimulus (Fa ntz , 19 63,
Goren et al., 1975; Johnson et al., 1991; Valenza, et al.,
1996). Although this may not provide conclusive evidence
for an innate face processing module, it is suggestive
of a specialized cognitive system operating from very
early in life, and a fast learning process once the system
gains more experience. In the next two sections, we will
review our recent work on further testing the top-heavy
bias hypothesis and on exploring the early sensitivity to
race information in Taiwanese infants to illustrate the fast
learning and specialization process.
No More Top-Heavy Bias at 2.5 Months:
A Fast Learning Process Specically
Tuned to Face Representation
As stated above, series of studies published by
Turati’s group firmly supported the idea that some
A. B.
Geometric gures
in Simion et al. (2001)
Face-like stimuli
In Turati et al. (2002) Total xation time Number of discrete looks
53.86 s vs. 37.62 s
p < .03
10 vs. 8.09
p < .05
34.70 s vs. 41.08 s
p > .20
7.6 vs. 8.3
p > .30
44.15 s vs. 22.89 s
p < .003
10.43 vs. 6.5
p < .01
Figure 4. The main studies that support the non-specic top-heavy perceptual bias hypothesis. Panel (A)
shows the three pairs of non-face geometric gures which were adopted from Simion, Macchi
Cassia, Turati, and Valenza (2001). Newborn infants consistently looked longer at the top-heavy
conguration (the left ones with a star) that more elements are located in the upper part of the
gure. Panel (B) shows the total xation time and the number of discrete looks for the three pairs
of face-like stimuli which were adopted from Turati et al. (2002). Across the three pairs, newborn
infants demonstrated a clear looking preference toward the top-heavy conguration regardless
whether the pattern resembles a face or not.
08-Chien.indd 101 2012/3/22 下午 02:45:28
102 Sarina Hui-Lin Chien Hsin Yueh Hsu
preexisting non-specific structural properties may be
sufficient to account for newborns’ preferences for
real faces and face-like stimuli. Indeed, the idea that a
non-specific top-heavy configuration bias may explain
neonatal facial preference is truly original; nevertheless,
several deeper issues deserve further investigation
(Chien, Hsu, & Su, 2009). The first question is about
the developmental trajectory. How long will such a non-
specific structural bias stay in infancy? Is it a long-
lasting effect or a short-lived phenomenon only present
at birth? The second question is about the strength of the
bias. How strong is the top-heavy bias? Is it as strong
as a reex (i.e., a xed behavior) or can it be eliminated
or enhanced via operant conditioning? The third issue is
about the methodological limitation. As far as the topic
is concerned, there is no direct within-subject evidence
showing that an infant who prefers faces also prefers
top-heavy non-face figures. The final question is about
the physiological basis. What might be the visual neural
substrates in the newborn brain that can account for such
an up-down asymmetry bias? Perhaps the last question is
the hardest; nevertheless, the rst three questions can be
empirically addressed (Chien et al., 2009). Below we will
review our recent findings showing that the top-heavy
bias seems to be extremely short-lived.
The Intrinsic Top-Heavy Bias Is Short-Lived
Few studies have explored whether the “top-heavy”
conguration bias is still present at 3 or 4 months of age,
and the evidence seems inconclusive (Chien & Hsu, 2010;
Chien, Hsu, & Su, 2010; Islam, Lobue, & DeLoache,
2010; Quinn, Tanaka, Lee, Pascalis, & Slater, 2010;
Turati, Valenza, Leo, & Simion, 2005). Using an eye-
tracker, Turati et al. (2005) found that 3-mo-olds preferred
natural face images to unnatural ones, replicating and
extending previous evidence (Exp.1). However, infants
did not show consistent preference for non-face top-heavy
geometric patterns (Exp. 2). Thus, they concluded that
“at 3 months of age, the general mechanism responsible
for infants’ visual preference for top-heavy patterns is
still present but weaker and, thus, is activated only when
up-down asymmetry is highly salient.” To systematically
investigate the developmental trajectory and the strength
of the intrinsic “top-heavy” bias, we (Chien et al.,
2010) tested the basic discriminability between ‘‘top-
heavy” and ‘‘bottom-heavy” geometric patterns in 2- to
4.5- month-old infants, using a modified forced-choice
familiarization/novelty procedure (FNP; Chien, Palmer,
& Teller, 2003). Figure 5 illustrates the procedure the
FNP trials and the main results of Chien et al. (2010). The
key manipulation was that each infant was tested with
three types of top-heavy and bottom-heavy geometric
figures (e.g., T-shape, 甲-shape, ㄇ-shape) and received
both familiarized-to-top-heavy and familiarized-to-
bottom-heavy conditions. The results showed signicant
and equal novelty preferences (i.e., observer’s percent
correct [OPC] scores in this case) in both familiarization
conditions across age and figure types, suggesting a
reliable discriminability between top-heavy and bottom-
heavy congurations and there is no intrinsic bias towards
either conguration at this age (see Figure 5B).
Eye Tracking Study Revealing Looking
Preferences Only for Upright Faces in Infants
and Adults
Chien et al. (2010) found that the top-heavy bias
observed at birth (Simion et al., 2002; Turati, 2004) is
no longer present at 2.5 months of age; however, the
evidence is limited only to the non-face geometric gures.
Therefore, using a free viewing paradigm with an eye
tracker (Tobii T60), we (Chien, 2011) systematically
investigated whether top-heavy configuration bias is
still present for 3 to 5.5 month-old infants and for adults
as a comparison group. To be able to compare with the
main findings in the literature, Chien (2011) adopted
three classes of digitized stimuli including the “top-
heavy” and “bottom-heavy” geometric figures used in
Simion et al. (2002), schematic face-like figures used
in Turati et al. (2002), and photographed faces used in
Macchi Cassia et al. (2004). The key manipulation was
that each infant participant must view all three classes of
stimuli, presented in pairs of top-heavy vs. bottom-heavy
conguration for 20 seconds each.
In Chien’s (2011) study, the infant experiment aimed
to assess whether the top-heavy configuration bias is
08-Chien.indd 102 2012/3/22 下午 02:45:28
Early Specialization for Face and Race 103
still present in 3- to 5.5-mo-old infants. The reason for
choosing this age range is because a major transition from
subcortical to cortical visual pathway takes place at this
time period, marking the starting point where some degree
of cortical specialization for faces begins to emerge (Halit,
de Haan, & Johnson, 2003; Tzourio-Mazoyer et al.,
2002). The adult experiment aimed to assess if there is
any reliable spontaneous looking preference for the same
set of stimuli; the purpose of testing adults was to offer
a comparison end point as human adults’ performance
can be regarded as a mature-state of behavior of visual
information processing. The main predictions were
as follows: If the top-heavy bias was still present and
determines infant’s spontaneous looking preference, it
was expected to obtain reliable and consistent looking
preference for all the top-heavy configurations across
geometric figures, face-like patterns, and photographed
faces in individual participants. On the other hand, if
the top-heavy configuration bias was no longer present,
infant’s looking preference may be specically triggered
by the resemblance of real faces, and there shall be no
consistent looking preference across all kinds of top-
heavy figures except for the ones that were upright
photographed faces.
To reveal whether a consistent looking preference
for the top-heavy conguration across gure types exists
for any individual participants, Chien (2011) computed an
index of top-heavy bias for each gure and conducted a
scatter plot analysis for individual infants and adults. For
A.
“Familiarized-to-top-heavy” “Familiarized-to-bottom-heavy”
Trial #1 Trial #1
Study
Test
Trial #2 Trial #2
Study
Test
Trial #3 Trial #3
Study
Test
08-Chien.indd 103 2012/3/22 下午 02:45:28
104 Sarina Hui-Lin Chien Hsin Yueh Hsu
any particular pair of test stimuli, the index is computed
as the looking time for top-heavy configuration minus
the looking time for bottom-heavy configuration, and
divided by the total looking for the test pair. Therefore,
the value of the index is a number between -1 and 1. A
positive value indicates a preference for the top-heavy
configuration; a negative value indicates a preference
for the bottom-heavy configuration; a value close to
zero indicates no preference for either. These indices
allow us to directly assess whether there is correlated
preference for all top-heavy congurations across gure
types in each individual participant. Figure 6 illustrates
the distribution of the individual infant’s (on the left) and
adult’s (on the right) indices of the top-heavy bias for
eight stimulus pairs in Chien (2011). In each data graph,
the black circles represent the group means for these eight
stimulus pairs. The pink marks are for infants younger
than 20 weeks, while the blue marks are for those older
than 20 weeks. In terms of the degree of resemblance to a
face, the vertical order of the test pairs from top to bottom
follows the subjective facedness rating result. The gure
on the top represents the pattern that was rated least like a
face (i.e., the ㄇ-shape), while the bottom one represents
the pattern that was rated most like a face (i.e., Taiwanese
B.
55.4%
62.4% 58.2%
63.5% 59.0% 59.6%
0%
25%
50%
75%
100%
T-shape 䓚-shape ʁ-shape
Observer’s Percent Correct Across Ages (N = 24)
OPC (boom-heavy figures, familiarized-to-top-heavy)
OPC (top-heavy figures, familiarized-to-boom-heavy)
Figure 5. The experimental design and main ndings in Chien, Hsu, and Su (2010). Panel (A) shows the
experimental conditions and the FNP procedure. The left column shows the familiarized-to-top-
heavy condition and the right column shows the familiarized-to-bottom-heavy condition. Each FNP
trial contains a study phase and a test phase. In the test phase, the observer judges which stimulus
that the infant prefers and makes a forced-choice response about the location of the novel stimulus.
Panel (B): The mean observer’s percent correct (OPC) scores for the three geometric gures in 2.5-
to 4-month-old infants. The abscissa represents the types of gures (T-shape, 甲-shape, ㄇ-shape).
The ordinate depicts the observer’s percent correct (OPC) scores for correctly judging the locations
of the novel stimulus over repeated trials. Error bars represent the standard errors of the group
means. The results indicate signicant and about equal novelty effects for the novel stimuli in both
the familiarized-to-top-heavy and the familiarized-to-bottom-heavy conditions and for all three
gure types.
08-Chien.indd 104 2012/3/22 下午 02:45:29
Early Specialization for Face and Race 105
woman’s face). However, in terms of top-heavy vs.
bottom-heavy configuration, the vertical order of
presentation of the eight stimulus pairs shall not matter at
all, because in any given pair, the top-heavy conguration
is always placed on the right side and the bottom-heavy
one is on the left. If the top-heavy bias is still present and
determines infant’s spontaneous looking preference for all
the top-heavy geometric figures, face-like patterns, and
photographed faces, it is expected that all the data points
shall fall within the positive side. On the other hand, if the
Figure 6. The distribution of the individual infant’s (on the left) and adult’s (on the right) indices of the top-
heavy bias for eight stimulus pairs in Chien (2011). For each graph, the top-heavy gures are on
the right side and the bottom-heavy ones are on the left. The value of the top-heavy bias index is a
number between -1 and 1. A positive value indicates a preference for the top-heavy conguration;
a negative value indicates a preference for the bottom-heavy conguration; a value close to zero
indicates no preference for either. The black circles in both graphs represent the group means for
these eight stimulus pairs. In terms of the degree of resemblance to a face, the order of the test pairs
from top to bottom follows the subjective face-likeness rating result in which the top gure represents
the pattern that is rated least like a face, while the bottom gure represents the pattern that is rated
most like a face. For the infants, the values of the geometric and the face-like patterns were close to
zero, indicating no consistent bias for either the “top-heavy” or the “bottom-heavy” conguration.
For the adults, the indices for the geometric and face-like patterns were also close to zero except for
the T-shape gure and the ones that had higher ratings on facedness scale. Moreover, the indices for
photographed upright faces were positive in both infants and adults, indicating a clear preference for
upright natural faces over inverted or unnatural ones.
08-Chien.indd 105 2012/3/22 下午 02:45:29
106 Sarina Hui-Lin Chien Hsin Yueh Hsu
top-heavy conguration bias is no longer present, infant’s
looking preference may be specifically triggered by the
resemblance of real faces, the data points for the non-face
stimuli shall fall randomly around the zero line.
As revealed in Figure 6, the top-heavy bias is no
longer present in infants of this age range. Majority
of the data points for the top-heavy and bottom-heavy
geometric and face-like patterns fall randomly on either
side of the zero line; while the data points for the last two
photographed face pairs clearly shift towards the positive
side. For the infants (the left graph), the values of the
geometric and the face-like patterns are close to zero,
indicating no consistent bias for either the “top-heavy”
or the “bottom-heavy” conguration. For the adults (the
right graph), the indices for the geometric and face-like
patterns are also close to zero except for the T-shape
gure and the ones that had higher ratings on facedness
scale. Moreover, the indices for photographed upright
faces were positive in both infants and adults, indicating a
clear preference for upright natural faces over inverted or
unnatural ones. To conclude this session, we support the
notion that the non-specific innate bias for “top-heavy”
configuration seems to vanish quickly at about 2 to 3
months of age. From this age on, infants’ preference for
faces is likely to reect a fast learning process specically
tuned to the representations of faces, and not the top-
heavy conguration per se.
Early Specialization Process for Natural
Faces of Own Race
As shown in Figure 6, it is the photographed
upright faces, and not any other kinds of stimuli, the
infants showed a clear preference for. What might be the
implications for that? We consider this is evidence for
a rather fast learning process that is specifically tuned
to upright, natural, or realistic face representation in
early infancy. Several recent studies on developmental
cognitive neuroscience that performed positron emission
tomo g r a p h y (P E T ) scans (Tzourio - M a z o y e r et al.,
2002) or measured event-related potentials (ERPs; Halit
et al., 2003) suggested that the first signs of cortical
specialization for faces can be observed in 2- to 3-month-
old infants. Tzourio-Mazoyer and colleagues (2002)
observed that when looking at unknown women’s faces,
2-month-old infants activated a distributed network
of cortical areas that largely overlapped the adult face
processing network. Moreover, Halit et al. (2003) even
found an ERP component that differentiates human faces
from monkey faces in 3-month-old infants. In other
words, they suggested that “even as young as 3 months of
age, infants show some specialization for human faces,”
although not to the same degree as do older infants or
adults.
In fact, the early face learning process can be
extended to race-specific identity information. Human
adults can better discriminate own-race faces than other-
race faces; this is referred to as the other-race effect (ORE;
Meissner & Brigham, 2001). An early onset of ORE has
been observed in 3- or 4-month-old Caucasian infants
(Hayden, Bhatt, Joseph, & Tanaka, 2007; Kelly et al.,
2005; Sangrigoli & de Schonen, 2004) and in similarly
aged Asian infants (Chien & Hsu, 2010; Hsu & Chien,
2011; Kelly, Liu, et al., 2007). This indicates a very early
influence of the race-typical stimuli in the environment
on face processing (Elman et al., 1996). For example,
Sangrigoli and de Schonen (2004) used the novelty
preference paradigm to assess Caucasian 3-month-old
infants’ recognition for Caucasian faces and Asiatic faces.
They found that infants showed a better recognition for
Caucasian than for Asiatic faces. Kelly et al. (2005) used
spontaneous preference paradigm found that 3-month-
old infants had a preference for their own-race face,
but newborn infants did not. Kelly, Quinn, et al. (2007)
further used visual-paired comparison (VPC) method to
test 3-, 6-, and 9-month-old Caucasian infants. They found
that 3-month-olds can discriminate both own- and other-
race faces while 9-month-olds can only discriminate their
own-race faces. However, the results of discrimination
capability at a younger age seems to contradict with other
studies showing an own race advantage or preference
(Hayden et al., 2007; Kelly et al., 2005; Sangrigoli & de
Schonen, 2004).
In a recent study investigating more rened aspects
of face processing, Hsu and Chien (2011) explored
08-Chien.indd 106 2012/3/22 下午 02:45:29
Early Specialization for Face and Race 107
whether the other-race effect exists in Taiwanese infants
aged between 4 and 9 months when the visual system is
still maturing. As shown in Figure 7A, the face stimuli
were collected from three races (Asian, Caucasian, and
African). There are three levels of difficulty in the face
discrimination task: easy, medium and hard conditions.
In the easy condition, the novel face was changed to a
different person’s face. In the medium condition, the
novel face was the same one but with eyes changed and
mouth moved up. In the hard condition, the novel face
was the same one but with mouth moved up. The visual
paired-comparison (VPC) task was adopted to assess
A.
Level
Race
Familiarized face
Easy
(change identity)
Medium
(eyes changed,
mouth moved up)
Hard
(mouth moved up)
Asian
Caucasian
African
B.
Own-race Other-race
Asian Caucasian African
Easy Medium Hard Easy Medium Hard Easy Medium Hard
4-mo P
6-mo PPP
9-mo P P P P P
Figure 7. The own-race and other-race faces and the main ndings in Hsu and Chien (2011). Panel (A): The
stimuli contained female/male faces of three ethnic groups (Asian, Caucasian, and African) only
male faces are shown here. Three levels of difculty in stimulus discriminability are adopted (i.e.,
easy, medium and hard). (Note that in Panel (A) two faces are shown in mosaic for image privacy
protection). Panel (B): The pattern of discriminability for own-race and other-race faces in 4-, 6-,
and 9-month-old infants. The cells with check marks represent positive results of discriminability.
4-month-old infants can only discriminate the Asian “easy” faces; 6-month-old infants can
discriminate all three race faces in the easy conditions; 9-month-old infants can further discriminate
Asian and Caucasian faces in the medium conditions, but stay at the easy condition for African faces.
08-Chien.indd 107 2012/3/22 下午 02:45:29
108 Sarina Hui-Lin Chien Hsin Yueh Hsu
4-, 6-, and 9-month-old infants’ discriminability for the
familiar/novel faces through recording infant’s looking
behavior (e.g., total l ookin g time, eye gazes). Figu re
7B illustrates the main results for the three age groups.
Contrary to Kelly, Quinn, et al. (2007) main ndings, Hsu
and Chien (2011) found that 4-month-old infants can only
discriminate Asian “easy” faces, 6-month-old infants can
discriminate “easy” faces of all three ethnic groups, and
9-month-old infants can further discriminate “medium”
Asian and Caucasian faces but not “medium” African
faces. In conclusion, we found that own-race advantage
emerges around 4 months of age, while other-race effect
may take place between 6 and 9 months. Taken together,
these ndings suggest a mixture of general improvement
in face processing ability as well as a specic tuning and
facilitative effect by the own-race experience in the rst
year of life.
General Discussions
Despite of the relatively poor vision at birth, human
infants are born with a predisposition to attend to human
faces or face-like stimuli. In this review article, we started
out highlighting the main findings for the neonatal face
preference studies, and then discussed three broad views,
the sensory hypothesis, the conguration hypothesis, and
the non-specific perceptual bias hypothesis to explain
neonatal face preference. The concept of visibility of the
sensory hypothesis is indispensible as the stimuli have to
be at least “visible” to attract infants’ attention. However,
beyond considering the subjective visibility, some built-
in preference for face-like structure or configuration is
also present at birth. Whether the preference is driven by
the “faceness” itself or something more general in the
face-like structure is the heart of the debate. The core
difference between Johnson and Morton’s model and
Turati’s top-heavy bias hypothesis lies in the nature of
initial representation. The former hypothesizes a domain-
specic bias towards face representation, while the latter
proclaimed that a non-specic perceptual bias is sufcient
to attract newborns to attend to various classes of visual
stimuli including faces per se.
If Top-Heavy Bias Is Short-Lived, Why Is It
Present at the First Place?
It is really not the purpose of this review to settle the
dispute about whether the initial representation is domain-
specific (face-specific) or non-specific at birth. Rather,
what we try to paint is the larger picture showing that a
specialized cognitive system seems to operate at the very
beginning of life, and a fast learning process develops
when the visual system starts to gain more experiences.
Thus in the third section, we illustrated the deeper issues
regarding the developmental trajectory and the strength of
the top-heavy configuration bias. Using a forced-choice
novelty preference method, we (Chien et al., 2010) found
that 2.5- to 4.5-month-old infants showed significant
and equal novelty preferences, suggesting a reliable
discriminability between the two configurations, which
indicates that the so-called top-heavy perceptual bias is no
longer present at 2.5 months of age. Moreover, using an
eye-tracker, we (Chien, 2011) further investigated whether
the top-heavy bias is still present in 3- to 5.5-month
olds and in adults as a comparison group, and found no
evidence for the top-heavy bias in both infants and adults.
The absence of top-heavy bias plus a clear preference for
photographed upright faces in infants seems to suggest a
very early cognitive specialization process toward face
representation.
However, one may ask such a question: If the top-
heavy bias is short-lived, why is it present at the first
place? In our view, the theoretical implication for the
top-heavy bias account for neonatal face preference
is profound. This hypothesis needs not to assume the
existence of a specific innate subcortical or cortical
representational bias toward faces; it provides a simpler
explanation, considering the Occam’s Razor principle
of model comparison. In addition, newborns’ vision is
very poor; the presence of the top-heavy bias at birth
serves as the earliest basis of face preference. This bias
rests on general functioning of the visual system and
directs newborns to attend to visual stimuli containing
such configuration including faces. What is more
important, however, once infants start to gain more visual
experiences, the top-heavy bias is quickly replaced by
a true preference towards the more refined real-face
08-Chien.indd 108 2012/3/22 下午 02:45:29
Early Specialization for Face and Race 109
representation. In other words, the top-heavy bias is only
suited for the very beginning stage, which is somewhat
like the role of radial glia cells as a to-be-discarded
precursor in neurogenesis.
The Early Face and Race Specialization
Supports the Experience-Expectant
Perspective
In the forth section, we highlighted a few
developmental cognitive neuroscience studies showing
evidence for early cortical specialization for detecting
faces. Moreover, young infants’ preference for up-right
real faces is not only identity-specific, but even race-
sensitive. When shown own-race faces paired with other-
race faces, newborn infants demonstrated no spontaneous
preference for faces from their own ethnic group.
However, at about 3 months of age, infants displayed
a spontaneous looking preference for own-race faces,
a finding that applies to both Caucasian and Chinese
infants (Kelly et al., 2005; Kelly, Liu, et al., 2007).
Using more elaborated face-discrimination task, Hsu and
Chien (2011) found 4-month-old Taiwanese infants can
only discriminate own-race “easy” faces and not other-
race “easy” faces, indicating an early emergence of own-
race advantage. Moreover, 6-month-old infants can
discriminate “easy” faces of all three ethnic groups, while
9-month-old infants can further progress to discriminate
“medium” Asian and Caucasian faces but not “medium”
African faces. Taken together, these findings suggest
a mixture of general improvement in face processing
ability from non-specic to specic (Slater et al., 2010),
as well as a tuning and facilitative effect by the own-
race experience (Hsu & Chien, 2011). Well, what might
all these race-specic effects tell us about the process of
early face development?
In a recent review chapter, de Haan and Halit (2001)
broadly distinguished three major approaches regarding
the developmental origins of face processing. The first
view argues that the development of face specialization
is an experience-independent (Farah, Rabinowitz, Quinn,
& Liu, 2000). Owing to the high relevance of faces in
human lives, natural selection has led to the evolution
of innate face-specific devices that are available prior
to a ny postnatal perceptual experience. Accord ing t o
this view, the functional and anatomical distinction that
characterizes face processing is supposed to be “explicitly
specied in the genome.” The second view characterizes
such process as experience-dependent (e.g., Diamond &
Carey, 1986; Gauthier & Logothetis, 2000). For human
beings, extensive and prolonged experience with faces
would gradually make humans exceptional experts in
recognizing individual exemplars (faces). According
to this approach, general learning processes, which can
occur at any time during development and might involve
any class of visual stimuli, are sufficient to explain the
emergence of a cortical system that processes faces
efficiently. In other words, an innate device for face
processing is not necessary at all.
In fact, we have shown solid evidence for neonatal
face preference as well as a particular tuning effect by
own-race face experience in early infancy; our position is
neither the experience-independent, nor the experience-
dependent approach, but the so called experience-
expectant view (Nelson, 2001, 2003). This approach
is neither extremely innate nor completely acquired; it
considers that the development of face processing shall
be explained as an experience-expectant process which
emphasizes the interactions between the innate propensity
and the proper environmental stimulation at the right
time. Through evolutionary pressures, the cortical tissue
has gained the potential to become specialized for face
processing. However, this specialization would not
emerge unless the critical inputs (e.g., human faces)
are provided within the crucial time windows. More
specifically, the partial functioning of neural pathways
would shape subsequent development of neural circuits,
resulting in the progressive tuning of certain cerebral
tissues to be specialized in processing faces (Johnson &
de Haan, 2001; Nelson, 2001, 2003). As infants build
on more and more specic visual experiences relating to
faces, these cortical issues may further differentiate into
subareas for refined processes of identity information
such as race, gender, and age.
08-Chien.indd 109 2012/3/22 下午 02:45:29
110 Sarina Hui-Lin Chien Hsin Yueh Hsu
Conclusion and Future Directions
The multifaceted task of face recognition is
intriguing, complex, and undoubtedly important in
everyday living. In adults, convergent evidence from
behavioral, electrophysiological, and imaging studies
provide strong evidence for a highly specialized neural
architecture that is dedicated to processing faces. In
infants, a body of research demonstrates that newborns
show a significant preference for face and face-like
stimuli. Regardless of whether the initial representation is
face-specic or non-specic at birth, the studies described
in this review converge to reveal that a specialized
cognitive system for processing and learning naturalistic,
race-typical faces quickly unfold as infants gain more
visual experience.
Presently, it remains unclear that what kinds of face
processing mechanisms (for example, the configural or
featural processing) that infants at different ages may
utilize and how might them differ from those of adults.
One promising solution in the near future would be to
design experiments that can directly compare infants’
performance with adults’ performance in sensible ways. A
second promising research direction would be to study the
atypical development of face processing such as infants
who may be at risk of autism spectrum disorder (ASD). In
the recent years, increasingly more evidence has revealed
that face processing ability in ASD infants seems to
impair, and which can be a hint for an early detection of
ASD infants before the age of language explosion. This
is a new focus with great clinical value. As the ability to
attend and respond properly to information contained in
human faces forms the very foundation for later social
and cognitiv e develo pment, i t is of great importa nce
to understand the normal as well as the abnormal
development of face processing in infancy.
Endnote
1. It is known that the development of nasal hemifield lagged behind
that of the temporal hemield in infants. Lewis and Maurer (1992)
proposed that the temporal-nasal asymmetry was due to a later
maturation of projections from cortex to the nucleus of the optic tract,
which governs nasally directed eye movement. An earlier maturation
of direct colliculular input would be associated with temporally
directed eye movements. Thus, a temporal-nasal asymmetry of rapid
orienting to faces or face-like stimuli (Simion et al., 1995; Tomalski,
Johnson, & Csibra, 2009) have been reported and attributed to
rel at ively matu re su bcort ic al vi sual route (proje ction to su perio r
colliculus) at birth.
References
Atkinson, J. (1984). Human visual development over
the rst six months of life. A review and a hypothesis.
Human Neurobiology, 3, 61-74.
Atkinson, J., Braddick, O., & French, J. (1979). Contrast
sensitivity of the human neonate measured by the
visual evoked potential. Investigative Ophthalmology
& Visual Science, 18, 210-213.
Banks, M. S., & Ginsburg, A. P. (1985). Infant visual
preferences: A review and new theoretical treatment.
In H. W. Reese (Ed.), Advances in child development
and behavior (Vol. 19, pp. 207-246). Orlan do, F L:
Academic Press.
Banks, M. S., & Salapatek, P. (1981). Infant pattern
vision: A new approach based on the contrast sensory
function. Journal of Experimental Child Psychology,
31, 1-45.
Bushnell, I. W. R., Sai, F., & Mullin, J. T. (1989).
Neonatal recognition of the mother’s face. British
Journal of Developmental Psychology, 7, 3-15.
Chien, S. H. L. (2011). No more top-heavy bias: Infants
and adults prefer upright faces, but not top-heavy
geometric or face-like patterns. Journal of Vision,
11(6), 1-14.
Chien, S. H. L., & Hsu, H. Y. (2010, August). Emergence
of the own-race advantage in Taiwanese infants:
Re think ing perc eptu al na r ro wing in a deve lopin g
visual system. Symposium talk at the 7th International
Conference on Cognitive Science, Beijing, China.
Chien, S. H. L., Hsu, H. Y., & Su, B. H. (2009, August).
Revisit the non-specic “top-heavy” structural bias in
young infants. Paper presented at the 2009 European
Conference on Visual Perception Annual Meeting,
Regensburg, Germany.
Chien, S. H. L., Hsu, H. Y., & Su, B. H. (2010).
Discriminating “top-heavy” versus “bottom-heavy”
08-Chien.indd 110 2012/3/22 下午 02:45:29
Early Specialization for Face and Race 111
geometric patterns in 2- to 4.5-month-old infants.
Vision Research, 50, 2029-2036.
Chien, S. H. L., Palmer, J., & Teller, D. Y. (2003). Infant
lightness perception: Do 4-month-old infants follow
Wallach’s ratio rule? Psychological Science, 14, 291-
295.
de Haan, M., & Halit, H. (2001). Neural bases and
development of face recognition during infancy. In A.
F. Kalverboer & A. Gramsbergen (Eds.), Handbook of
brain and behaviour in human development (pp. 921-
938). Dordrecht, The Netherlands: Kluwer Academic
Publishers.
de Schonen, S., & Mathivet, E. (1989). First come,
first served: A scenario about the development of
hemispheric specialization in face recognition during
infancy. Current Psychology of Cognition, 9, 3-44.
Diamond, R., & Carey, S. (1986). Why faces are and
are not special: An effect of expertise. Journal of
Experimental Psychology: General, 115, 107-117.
Easterbrook, M. A., Kisilevsky, B. S., Hains, S. M., &
Muir, D. M. (1999). Faceness or complexity: Evidence
from newborn visual tracking of facelike stimuli. Infant
Behavior and Development, 22, 17-35.
Elman, J. L., Bates, E. A., Johnson, M. H., Karmiloff-
Smith, A., Parisi, D., & Plunkett, K. (1996).
Rethinking innateness: A connectionist perspective on
development. Cambridge, MA: MIT Press.
Fantz, R. L. (1963). Pattern vision in newborn infants.
Science, 140, 296-297.
Farah, M. J. (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.
Field, T. M., Cohen, D., Garcia, R., & Greenberg, R.
(1984). Mother-stranger face discrimination by the
newborn. Infant Behavior and Development, 7, 19-26.
Gauthier, I., & Logothetis, N. K. (2000). Is face
recognition not so unique after all? Cognitive
Neuropsychology, 17, 125-142.
Goren, C. C., Sarty, M., & Wu, P. Y. K. (1975). Visual
following and pattern discrimination of face-like
stimuli by newborn infants. Pediatrics, 56, 544-549.
Halit, H., de Haan, M., & Johnson, M. H. (2003). Cortical
specialization for face processing: Face-sensitive event-
related potential components in 3- and 12-month-old
infants. NeuroImage, 19, 1180-1193.
Hayden, A., Bhatt, R. S., Joseph, J. E., & Tanaka, J. W.
(2007). The other-race effect in infancy: Evidence
using a morphing technique. Infancy, 12, 95-104.
Hsu, H. Y., & Chien, S. H. L. (2011). Exploring the other-
race effect in Taiwanese infants and adults. Chinese
Journal of Psychology, 53, 35-57.
Islam, N., Lobue, V., & DeLoache, J. S. (2010, March).
Not just an ordinary face: Infants’ Perception of face-
like objects. Poster presented at the 17th Biennial
Meeting of International Conference on Infant Studies,
Baltimore, MD.
Johnson, M. H. (1997). Developmental cognitive
neuroscience: An introduction. Oxford, UK: Blackwell.
Johnson, M. H., & de Haan, M. (2001). Developing
cortical specialization for visual-cognitive function:
Th e c ase of fac e re cog nition. In J. L. McC lelland
& R. S. Siegler (Eds.), Mechanisms of cognitive
development: Behavioral and neural perspectives (pp.
253-270). Mahwah, NJ: Lawrence Erlbaum Associates.
Johnson, M. H., Dziurawiec, S., Ellis, H., & Morton, J.
(1991). Newborns’ preferential tracking of face-like
stimuli and its subsequent decline. Cognition, 40, 1-19.
Johnson, M. H., & Morton, J. (1991). Biology and
cognitive development: The case of face recognition.
Oxford, UK: Blackwell.
Kanwisher, N. (2000). Domain specificity in face
perception. Nature Neuroscience, 3, 759-763.
Karmiloff-Smith, A. (1996). Beyond modularity: A
developmental perspective on cognitive science.
Cambridge, MA: MIT Press.
Kellman, P. J., & Arterberry, M. E. (2000). The cradle
of knowledge: Development of perception in infancy.
Cambridge, MA: MIT press.
08-Chien.indd 111 2012/3/22 下午 02:45:29
112 Sarina Hui-Lin Chien Hsin Yueh Hsu
Kelly, D. J., Liu, S., Ge, L., Quinn, P. C., Slater, A. M.,
Lee, K., et al. (2007). Cross-race preferences for same-
race faces extend beyond the African versus Caucasian
contrast in 3-month-old infants. Infancy, 11, 87-95.
Kelly, D. J., Quinn, P. C., Slater, A. M., Lee, K., Ge, L.,
& Pascalis, O. (2007). The other-race effect develops
duri ng infan cy: Evid ence of perce ptual narrowing.
Psychological Science, 18, 1084-1089.
Kelly, D. J., Quinn, P. C., Slater, A. M., Lee, K., Gibson,
A., Smith, M., et al. (2005). Three-month-olds, but
not newborns, prefer own-race faces. Developmental
Science, 8, F31-F36.
Kleiner, K. A. (1987). Amplitude and phase spectra as
indices of infants’ pattern preferences. Infant Behavior
and Development, 10, 49-59.
Kleiner, K. A., & Banks, M. S. (1987). Stimulus energy
does not account for 2-month olds’ face preferences.
Journal of Experimental Psychology: Human
Perception and Performance, 13, 594-600.
Lewis, T. L., & Maurer, D. (1992). The development of
the temporal and nasal visual fields during infancy.
Vision Research, 32, 903-911.
Macchi Cassia, V., Turati, C., & Simion, F. (2004). Can
a nonspecific bias toward top-heavy patterns explain
newborns’ face preference? Psychological Science, 15,
379-383.
Macchi Cassia, V., Valenza, E., Pividori, D., & Simion,
F. (2002, April). Facedness vs. non-specic structural
properties: What is crucial in determining face
preference at birth? Poster presented at the 13th
Biennial Meeting of International Conference on Infant
Studies, Toronto, Canada.
Maurer, D., & Barrera, M. (1981). Infants’ perception of
natural and distorted arrangements of a schematic face.
Child Development, 52, 196-202.
Maurer, D., & Maurer, C. (1988). The world of the
newborn. New York: Basic Books.
Meissner, C. A., & Brigham, J. C. (2001). Thirty years of
investigating the own-race bias in memory for faces: A
meta-analytic review. Psychology, Public Policy, and
Law, 7, 3-35.
Morton, J., & Johnson, M. H. (1991). CONSPEC and
CONLERN: A two-process theory of infant face
recognition. Psychological Review, 98, 164-181.
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 reflects an experience-expectant and
activity dependent process. In O. Pascalis, & A. Slater
(Eds.), The development of face processing in infancy
and early childhood: Current perspectives (pp. 79-97).
New York: Nova Science.
Pascalis, O., de Schonen, S., Morton, J., Deruelle, C., &
Fabre-Grenet, M. (1995). Mother’s face recognition by
neonates: A replication and extension. Infant Behavior
and Development, 18, 79-85.
Quinn, P. C., Tanaka, J. W., Lee, K., Pascalis, O., &
Slater, A. (2010, March). Are faces special to young
infants? Configural and upper-region processing
advantages for houses in 3- to 4-month-olds. Poster
presented at the 17th Biennial Meeting of International
Conference on Infant Studies, Baltimore, MD.
Sangrigoli, S., & de Schonen, S. (2004). Recognition
of own-race and other-race faces by three-month-old
infants. Journal of Child Psychology and Psychiatry
and Allied Disciplines, 45, 1219-1227.
Simion, F., Macchi Cassia, V., Turati, C., & Valenza, E.
(2001). The origins of face perception: Specic vs. non-
specific mechanisms. Infant and Child Development,
10, 59-65.
Simion, F., Valenza, E., Macchi Cassia, V., Turati, C., &
Umiltà, C. (2002). Newborns’ preference for up-down
asymmetrical congurations. Developmental Science, 5,
427-434.
Simion, F., Valenza, E., Umiltà, C., & Della Barba, B.
(1995). Inhibition of return in newborns is temporo-
nasal asymmetrical. Infant Behavior and Development,
18, 189-194.
Slater, A., Quinn, P. C., Kelly, D. J., Lee, K., Longmore,
C. A., McDonald, P. R., et al. (2010). The shaping of
the face space in early infancy: Becoming a native face
processor. Child Development Perspectives, 4, 205-211.
08-Chien.indd 112 2012/3/22 下午 02:45:29
Early Specialization for Face and Race 113
Slater, A., von der Schulenburg, C., Brown, E., Badenoch,
M., Butterworth, G., Parsons, S., et al. (1998). Newborn
infants prefer attractive faces. Infant Behavior and
Development, 21, 345-354.
Teller, D. Y. (1997). First glances: The vision of infants.
The friedenwald lecture. Investigative Ophthalmology
& Vision Science, 38, 2183-2203.
Tomalski, P., Johnson, M. H., & Csibra, G. (2009).
Temporal-nasal asymmetry of rapid orienting to face-
like stimuli. NeuroReport, 20, 1309-1312.
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.
Turati, C., Valenza, E., Leo, I., & Simion, F. (2005).
Three-month-olds’ visual preference for faces and its
underlying visual processing mechanisms. Journal of
Experimental Child Psychology, 90, 255-273.
Tzourio-Mazoyer, N., de Schonen, S., Crivello, F.,
Reutter, B., Aujard, Y., & Mazoyer, B. (2002). Neural
correlates 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
Experimental Psychology: Human Perception and
Performance, 22, 892-903.
Walton, G. E., Bower, N. J. A., & Bower, T. G. R. (1992).
Recognition of familiar faces by newborns. Infant
Behavior and Development, 15, 265-269.
08-Chien.indd 113 2012/3/22 下午 02:45:29
114 Sarina Hui-Lin Chien Hsin Yueh Hsu
上重偏好不再：嬰兒的早期臉孔與種族專化歷程
簡惠玲 許馨月
中國醫藥大學神經科學與認知科學研究所
本文主要目的在回顧嬰兒對於臉孔和種族的知覺發展的早期專化歷程。我們首先回顧探討新生兒臉孔偏好的
幾篇經典研究，接著進一步介紹解釋此臉孔偏好的三大理論取向，包括八零年代的「線性模式取向的感官假說」、
九零年代的嬰兒對「特定臉孔表徵的先天性偏好理論」，以及近十年內Turati 等人主張的「非特定的知覺偏好理
論」。本文的第三部分深入討論Turati等人的「上重偏好假說」的一些限制與相關的後續研究，利用強迫選擇新奇
偏好注視法和眼動追蹤儀，我們重新檢驗此假說並發現二、三個月大的嬰兒並未顯現出上重偏好；表示此非特定的
上重結構的偏好可能只是個很短暫的趨力，當嬰兒開始累積視覺經驗，他們會轉而傾向去注意到真正的臉孔表徵。
本文的第四大段闡述生命早期的臉孔學習與專化歷程，除了回顧來自於神經生理方面的支持證據外，也探討嬰兒對
於臉孔的種族訊息的敏感度如「他種族效應」的研究。「他種族效應」的研究結果發現，三個月大的嬰兒較偏好自
己種族的臉孔，而我們研究台灣地區四到九個月大的嬰兒辨識不同種族臉孔的能力，發現嬰兒且辨識自己種族臉孔
的能力會最早出現，且會隨著年齡增長而有所精進；而辨認他種族的能力會較晚出現且進步有限。整體而言，綜觀
我們所回顧的西方文獻與本土的研究，本論文支持早期臉孔知覺發展與專化的「經驗─期待」觀點。
關鍵詞：他種族效應、先天性偏好、新生兒、經驗─期待、臉孔處理
08-Chien.indd 114 2012/3/22 下午 02:45:30
