Child Development, July/August 2009, Volume 80, Number 4, Pages 1039–1056

The Effects of Early Experience on Face Recognition: An Event-Related

Potential Study of Institutionalized Children in Romania
Margaret C. Moulson Alissa Westerlund
Massachusetts Institute of Technology Children’s Hospital Boston

Nathan A. Fox Charles H. Zeanah

University of Maryland Tulane University Health Sciences Center

Charles A. Nelson
Children’s Hospital Boston and Harvard Medical School

Data are reported from 3 groups of children residing in Bucharest, Romania. Face recognition in currently
institutionalized, previously institutionalized, and never-institutionalized children was assessed at 3 time
points: preintervention (n = 121), 30 months of age (n = 99), and 42 months of age (n = 77). Children watched
photographs of caregiver and stranger faces while event-related potentials were recorded. Results demon-
strate that institutionalized children show pervasive cortical hypoarousal in response to faces and that foster
care is somewhat effective in remediating this deficit by 42 months of age. All 3 groups of children
distinguished between the familiar and unfamiliar faces. These results have the potential to inform an
understanding of the role of early experience in the development of the neural systems that subserve face
recognition.

Faces are arguably the most important visual stim- perception until adolescence. Although the behav-
ulus used in human social communication. Faces ioral development of face perception has been well
provide a wealth of information about other indi- characterized, less is known about the development
viduals, including identity, sex, age, focus of atten- of the neural systems underlying face perception.
tion, and emotional state. Most adults are experts at Clearly, the origins of behavioral change lie in the
processing faces, and quickly and effortlessly deci- development of these neural systems.
pher information from faces during social Previous research has demonstrated that adults’
exchanges. Although young infants have rudimen- face expertise is subserved by a distributed network
tary face processing capabilities (e.g., they can rec- of brain areas that are preferentially involved in
ognize their mother within the first days of life and face processing, including the ‘‘fusiform face area’’
can discriminate basic emotions within the first within the fusiform gyrus, the superior temporal
months of life; Field, 1983; Pascalis, de Schonen, sulcus, amygdala, and areas of the prefrontal cortex
Morton, Derulle, & Fabre-Grenet, 1995), children do (Adolphs, 2002; Haxby, Hoffman, & Gobbini, 2002;
not show adult-like speed and accuracy in face Iidaka et al., 2001; Kanwisher, McDermott, & Chun,
1997). Fewer studies have examined the develop-
The work reported in this article was supported by funds from ment of these neural systems, largely due to the
the John D. and Catherine T. MacArthur Foundation. We thank methodological limitations involved in using neuro-
Gwen Gordon for assistance in data management; Don Guthrie
for statistical consultation; Sebastian Koga for overseeing the
imaging techniques such as functional magnetic
project in Romania; Hermi Woodward and the MacArthur Foun- resonance imaging with developing populations
dation Research Network on Early Experience and Brain Devel- (although near infrared spectroscopy shows prom-
opment for input regarding the conceptualization, design, and
implementation of this project; the caregivers and children who
ise for use with infants and young children;
participated in this project; the Bucharest Early Intervention Pro- see Otsuka et al., 2007; Tzourio-Mazoyer et al.,
ject staff for their tireless work on our behalf; and our many 2002). However, event-related potentials (ERPs), a
colleagues in Romania who facilitated our work, particularly
Bogdan Simion and Alin Stanescu.
Correspondence concerning this article should be addressed to
Margaret C. Moulson, The MIT Department of Brain and Cogni-  2009, Copyright the Author(s)
tive Sciences, Room 46-4089, 77 Massachusetts Ave, Cambridge, Journal Compilation  2009, Society for Research in Child Development, Inc.
MA 02139. Electronic mail may be sent to [email protected]. All rights reserved. 0009-3920/2009/8004-0008
1040 Moulson, Westerlund, Fox, Zeanah, and Nelson

noninvasive measure of the electrical activity of the Nelson (2001) postulates that the development of
brain that occurs in response to discrete stimulus face perception is an experience-expectant, activity-
events (Handy, 2005), have been used fruitfully in dependent process, in which infants’ experiences
previous research to investigate the neural corre- with faces early in life shape the cortical systems
lates of face recognition in infants and children. that give rise to expert face processing. Evidence
Previous studies have focused on several distinct that typically developing infants retain the ability
ERP components (deflections in the ongoing elec- to discriminate faces to which they have been
troencephalogram [EEG]) that seem to reflect exposed, and lose the ability to discriminate faces
aspects of face processing in infants and children. to which they have not been exposed, supports this
The Nc, a negative-going component that occurs view. For example, Pascalis, de Haan, and Nelson
over frontocentral regions approximately 300– (2002) demonstrated that 6-month-old infants, but
700 ms following stimulus onset, is larger in ampli- not 9-month-old infants or adults, are able to dis-
tude in response to mother’s face compared to a criminate monkey faces. Thus, between 6 and
stranger’s face in 6- to 24-month-old infants (Carver 9 months of age, infants’ face perception becomes
et al., 2003; de Haan & Nelson, 1997). The Nc ‘‘species-specific,’’ in that they lose the ability to
response reverses over the course of development, discriminate monkey faces but retain the ability to
in that older children between the ages of 45 and discriminate human faces. Infants retain the ability
54 months show a larger Nc in response to a stran- to discriminate monkey faces at 9 months if they
ger’s face compared to their mother’s face (Carver have been exposed to monkey faces in the form of
et al., 2003), which probably reflects a developmen- a picture book (Pascalis et al., 2005). Interestingly,
tal change in the social significance of mother’s face the ‘‘other-species effect’’ has also been demon-
compared to stranger’s face. The Nc is also larger strated in infant monkeys who were selectively
in response to fearful faces compared to happy exposed to either human or monkey faces following
faces (Nelson & de Haan, 1996). Based on these a period of deprivation in which they saw no faces.
findings, researchers have posited that the Nc Following face exposure, the monkeys discrimi-
reflects attentional allocation (Courchesne, Ganz, & nated only those types of faces (human or monkey)
Norcia, 1981; Nelson, 1994; Richards, 2003). Two to which they had been exposed (Sugita, 2008).
components located over occipitotemporal regions, The ‘‘other-race effect’’ is another example of
the N290 and P400, respond reliably differently to specific face experiences shaping face processing
faces compared to other classes of objects in infants, skills. Many studies have shown that adults are
and are thought to reflect face-specific processing. poorer at discriminating and recognizing faces of
The N290 is larger in response to human faces than other races with which they have little experience,
monkey faces or nonface stimuli (Halit, Csibra, compared to faces of their own race (Meissner &
Volein, & Johnson, 2004; Halit, de Haan, & Johnson, Brigham, 2001; Tanaka, Kiefer, & Bukach, 2004;
2003), and the P400 is faster in response to faces Walker & Tanaka, 2003). This effect has its origin in
than to objects (de Haan & Nelson, 1999). Research- infancy. Kelly and colleagues (Kelly, Liu, et al.,
ers have hypothesized that the N290 and P400 com- 2007; Kelly et al., 2005) have demonstrated that
ponents are precursors of the adult N170, a face- 3-month-old infants prefer to look at faces of their
sensitive component that reflects the structural own race (see Quinn, Yahr, Kuhn, Slater, & Pascal-
encoding of a face (Bentin, Allison, Puce, Perez, & is, 2002, for evidence that 3-month-old infants also
McCarthy, 1996). Taylor, Batty, and Itier (2004) show a preference for faces matching the gender of
have investigated the neural correlates of face pro- their primary caregiver), and Sangrigoli and de
cessing in 4- to 15-year-old children, and found that Schonen (2004) have shown that 3-month-old
both the P1 and N170 components over occipito- infants are better able to discriminate faces of their
temporal regions were present in children across own race, although this effect disappears if they are
this age range and sensitive to aspects of face pro- given even a short amount of exposure to other-
cessing, including face inversion. The P1 and N170 race faces. By 9 months of age infants appear to
differed in terms of their response properties, lose the ability to discriminate faces of other races
although both showed developmental changes with (Kelly, Quinn, et al., 2007). However, Sangrigoli,
increasing age. Despite this growing body of Pallier, Argenti, Ventureyra, and de Schonen (2005)
knowledge about the neural correlates of face per- found that adults of Korean origin who had been
ception during development, little is known about adopted into European Caucasian families between
the mechanisms that drive developmental change 3 and 9 years of age showed a ‘‘reversed’’ other-
in the neural system underlying face processing. race effect—that is, they were more accurate at
 Early Experience and Face Recognition 1041

identifying Caucasian faces than Asian faces—indi- Kreppner, 2000), the persistence and severity of
cating that the other-race effect remains plastic until which are related to the timing and duration of the
at least 9 years of age. institutional experience (Beckett et al., 2006; Rutter
Additional evidence for the role of experience in et al., 2007).
shaping the development of face processing comes It is reasonable to suspect that institutional care
from studies of atypical development. For example, might adversely affect the development of the neu-
individuals who are born with cataracts, which ral system underlying face processing specifically.
block patterned visual input, show rapid improve- The research reviewed above documents the pow-
ment in low-level visual abilities such as acuity if erful effects that early experiences can have on the
the cataracts are removed in the 1st year of life face processing system, and the characteristics of
(Lewis & Maurer, 2005). However, even after many institutional care almost certainly ensure that insti-
years of normal visual input, these individuals tutionalized children have qualitatively different
show persistent deficits in face perception. Specifi- experiences with faces compared to family-reared
cally, they have difficulty detecting small changes children. Although no studies have directly investi-
in the spacing of facial features (i.e., configural gated the amount and kind of exposure to faces
processing), which can impair the recognition of a children receive in institutions versus family set-
person’s identity (Geldart, Mondloch, Maurer, de tings, informal observations suggest that children
Schonen, & Brent, 2002; Le Grand, Mondloch, who experience the high child-to-caregiver ratio
Maurer, & Brent, 2001, 2003). and high caregiver turnover in institutions likely
Early adverse environments can also affect the have less exposure to adult faces in general and less
development of face perception. For example, chil- consistent exposure to familiar adult faces but prob-
dren raised in abusive households show altered ably more exposure to other children’s faces than
facial emotion processing, particularly for angry family-reared children. Additionally, the broader
faces, presumably due to their increased exposure social context in which faces are experienced clearly
to expressions of negative emotion and decreased differs for institutionalized compared to family-
exposure to expressions of positive emotion. They reared children. It seems likely that these qualita-
show a response bias for angry faces, in that they tive differences in exposure to faces affect the
are more likely to match any emotional situation development of the neural system that subserves
to a picture of an angry face (Pollak, Cicchetti, face processing. Alternatively, it is possible that
Hornung, & Reed, 2000). They also overidentify the even the adverse environment of institutionaliza-
emotion anger but do not differ from controls in tion provides sufficient perceptual experience with
identifying happiness, sadness, and fear (Pollak & faces to bootstrap the development of the system.
Kistler, 2002), and correctly identify facial expres- Few studies have directly investigated the effects
sions of anger on the basis of less perceptual infor- of institutionalization on the development of face
mation than controls (Pollak & Sinha, 2002). In processing, but those that have found differences
addition, they attend to angry faces more than con- between family-reared children and children with
trols do and have trouble disengaging from angry histories of institutionalization. For example, Parker
faces (Pollak & Tolley-Schell, 2003). and Nelson (2005a, 2005b) found that institutional-
Institutionalization is another example of an ized children showed cortical hypoarousal, as
adverse early rearing environment that may nega- reflected by dramatically reduced amplitudes of
tively impact the development of face perception. ERP components, compared to noninstitutionalized
Institutional care is characterized by psychosocial children in response to pictures of familiar, unfa-
deprivation; sensory and cognitive stimulation are miliar, and emotional faces. And Wismer Fries and
lacking, and high child-to-caregiver ratios (in some Pollak (2004) found that 4.5-year-old previously
institutions, nearly 20:1) leave children with little institutionalized children showed deficits in identi-
social stimulation and almost no opportunity to fying primary emotions, such as happiness, sad-
form stable, emotional attachments to caregivers ness, fear, and anger, from pictures of faces. They
(Smyke et al., 2007; Zeanah et al., 2003). A wealth also showed deficits when matching pictures of
of previous research has documented poor physi- emotional faces to emotional scenarios (Wismer
cal, cognitive, social, and neurologic outcomes in Fries & Pollak, 2004).
previously institutionalized children (Fisher, Ames, The general picture that emerges from studies of
Chisholm, & Savoie, 1997; Gunnar, 2001; O’Connor, both typically and atypically developing infants
Bredenkamp, & Rutter, 1999; O’Connor & Rutter, and children is that early experiences are important
2000; O’Connor, Rutter, Beckett, Keaveney, & in determining the developmental course of face
1042 Moulson, Westerlund, Fox, Zeanah, and Nelson

processing abilities, although little is known about show indiscriminate behavior and weak attach-
how these early experiences shape the underlying ments to their caregivers (Zeanah, Smyke, Koga, &
neural systems responsible for expert face process- Carlson, 2005). Consequently, it is quite possible
ing. The goal of this study was to investigate further that they might have difficulty recognizing their
how early experiences shape the neural systems caregiver’s face or process it differently than fam-
that subserve face perception by studying the neu- ily-reared children, leading to differences between
ral correlates of face recognition in institutionalized institutionalized and never-institutionalized chil-
children. This study was part of the Bucharest Early dren in their neural responses to familiar and
Intervention Project (BEIP), a comprehensive, longi- unfamiliar faces. Alternatively, in light of previous
tudinal study of the effects of institutionalization on research documenting the effects of even brief
brain and behavioral development (Zeanah et al., experience with faces in infancy (Pascalis et al.,
2003). Institutionalized and family-reared children 2005; Sangrigoli & de Schonen, 2004; Sugita, 2008),
in Bucharest, Romania were assessed on a variety of it is also plausible that institutionalized children
measures spanning physical, cognitive, social, and receive sufficient exposure to faces to ensure the
neurologic domains of development. typical development of the face processing system,
Following a baseline assessment that occurred leading to similar neural processing of familiar and
between 5 and 31 months of age, institutionalized unfamiliar faces in institutionalized and never-insti-
children were randomly assigned to either contin- tutionalized children. Regarding the effects of the
ued institutional care or placement in high-quality intervention, we predicted that children who were
foster care, which was developed and supported by placed in foster care would show improvement
the project itself (see Nelson et al., 2007; Zeanah compared to the children who remained in the
et al., 2003). Follow-up assessments comparing the institution and that children who were placed in
institutionalized, foster care, and family-reared chil- foster care earlier would show greater recovery
dren occurred at 18, 30, 42, and 54 months of age. than children who were placed in foster care later.
Although the ethical issues raised by this project
have been given appropriate consideration else-
where (Millum & Emanuel, 2007; Zeanah et al., Method
2006a, 2006b), it is important to address two points
Participants
regarding children’s placement into caregiving
environments during the course of the study. Chil- Participants in the BEIP included 208 children
dren who were placed in BEIP foster care were between 5 and 31 months of age recruited in
never returned to the institution, and none of the Bucharest, Romania. At the baseline assessment,
children in this study were restricted in their place- the institutionalized group (IG) consisted of 136
ment into alternate environments (e.g., reunion children recruited from six institutions within
with biologic family, placement into state-run foster Bucharest, and the never-institutionalized group
care, or adoption) that were deemed suitable by the (NIG) consisted of 72 children who were recruited
appropriate Romanian authorities. through birth records from the maternity hospitals
To explore the development of the neural sys- where the institutionalized children were born and
tems underlying face processing in these children, matched to the IG on age and sex (see Nelson et al.,
at the baseline, 30-, and 42-month assessments, 2007; Zeanah et al., 2003, for a thorough description
ERPs were recorded while children passively of this sample). Eleven of the IG children initially
viewed pictures of a familiar face and an unfamiliar recruited for this study retrospectively met exclu-
face. Based on previous research with this sample sion criteria (i.e., genetic syndromes, overt signs of
of children (Parker & Nelson, 2005a, 2005b), we fetal alcohol syndrome, and ⁄ or microcephaly) and
predicted that institutionalized children would were subsequently excluded from all analyses. Fol-
show general cortical hypoarousal in that the lowing the baseline assessment, the 125 remaining
amplitude of their ERP components would be children in the IG were randomly assigned to con-
diminished compared to never-institutionalized tinued institutional care (IG; n = 62) or foster care
children. However, it was difficult to predict (FCG; n = 63). An additional 24 participants (IG,
whether the experience of institutionalization n = 11; FCG, n = 1; NIG, n = 12) were not included
would affect face processing specifically in this in the longitudinal sequence of assessments but
sample of children. On the one hand, institutional- were recruited for cross-sectional data collection.
ized children likely have atypical experiences with ERP data from 121, 99, and 77 participants at the
faces (especially their primary caregiver’s face) and baseline, 30-, and 42-month assessments, respec-
 Early Experience and Face Recognition 1043

Table 1 blocked children’s view of the room. There were 69

Breakdown of Participants by Group at Each Assessment total trials, and the mother or caregiver and stran-
ger faces were presented with equal probability.
Group
Each trial consisted of a baseline period (100 ms),
IG FCG NIG stimulus presentation (500 ms), and a poststimulus
Assessment n (no. females) n (no. females) n (no. females) recording period (800 ms). The intertrial interval
varied randomly between 500 and 1,000 ms. During
Baseline 81 (39) N⁄A 40 (20)
the experiment, an experimenter observed the child
30 months 37 (17) 42 (20) 20 (11)
from behind the black screen to direct the child’s
42 months 23 (10) 33 (15) 21 (13)
attention back to the computer screen when neces-
Note. IG = institutionalized group; FCG = foster care group; sary and to eliminate trials during which the child
NIG = never-institutionalized group. looked away. The experiment continued until the
child had seen the maximum number of trials or
tively, are reported (see Table 1 for a breakdown of he ⁄ she became too fussy or distracted to continue.
participants by group at each assessment). An addi-
tional 60 (33%), 56 (36%), and 45 (37%) participants
Electrophysiologic Recording and Processing
at the baseline, 30-, and 42-month assessments,
respectively, were excluded from data analysis due Continuous EEG was recorded from 13 scalp
to technical error, having fewer than 10 artifact-free electrodes (Fz, F3, F4, Cz, C3, C4, Pz, P3, P4, T7, T8,
trials per condition, blinking while the picture was O1, and O2) and left and right mastoid electrode
on the screen on 25% or more trials, or excessive eye sites using a close-fitting lycra cap with sewn-in tin
or body movement artifact. Due to the wide range of electrodes (Electro-Cap International, Eaton, OH).
ages at the baseline assessment (5–31 months), a Cz was the reference electrode during acquisition.
median split based on age at baseline assessment Bipolar vertical electrooculogram (EOG) was
was performed and included in the baseline recorded from electrodes bisecting the midline
analyses. The median age of the IG at baseline was placed above and below the left eye to record
23.5 months (41 younger, 40 older), whereas the blinks and other eye movements. After cap place-
median age of the NIG at baseline was 21.2 months ment, the scalp was prepared by inserting an abra-
(19 younger, 21 older). The noninterference policy of sive gel into each electrode site and the scalp under
the BEIP (Zeanah et al., 2003) ensured that many each site was gently abraded. A small amount of
children assigned to the IG and FCG no longer electrode gel was inserted into each electrode site.
resided in those environments by 42 months of age. Electrode impedances were considered acceptable
However, original group assignments were if at or below 10 kW. Signals were acquired using
preserved during data analysis. This intent-to-treat custom bioelectric amplifiers from SA Instrumenta-
approach probably underestimates any differences tion Company (San Diego, CA) and amplified at a
among the groups. gain of 5,000 for scalp leads and 2,500 for EOG. The
bandpass filter setting was 0.1–100 Hz, and the
amplifier was calibrated before each participant’s
Stimuli and Procedure
recording. All channels were sampled at 512 Hz
At each assessment, ERPs were recorded while and digitized onto the hard drive of a PC using a
children viewed two color photographs: their 12-bit A ⁄ D converter (±2.5 V input range) and
mother’s or caregiver’s face and a stranger’s face. Snap-Master acquisition software (HEM Data Cor-
Photographs showed each face in a frontal view poration, Southfield, MI).
posing a neutral expression and were taken under A digital lowpass filter of 30 Hz was applied
similar lighting conditions. The preferred caregiver using the ERP Analysis Systems from James Long
for each child in the IG was determined via surveys Company (Caroga Lake, NY). Subsequent process-
of the staff working with the children, and each ing and analysis of the EEG signal was carried out
child was subsequently tested with that caregiver’s using the ERP32 analysis software package (New
face. The caregiver face used for children in the Boundary Technologies, Minneapolis, MN). ERP tri-
FCG was their foster mother. The stranger face als were extracted that consisted of a 100-ms base-
used for each participant was the face of a different line period, 500-ms stimulus presentation, and 800-
mother or caregiver. Children sat on their parent’s ms poststimulus recording period. Scalp channels
or caregiver’s lap in front of a computer screen. with excessive artifact (i.e., EEG signals > ±200 lV)
Black panels surrounding the computer screen were rejected, and the entire trial was rejected if
1044 Moulson, Westerlund, Fox, Zeanah, and Nelson

more than two scalp channels exceeded this thresh- FCG, NIG) repeated measures omnibus analyses of
old. Data were referenced to an average mastoids variance (ANOVAs) were carried out. For frontocen-
configuration. Eye movement-related artifact was tral components, data at frontal and central leads
corrected (Gratton, Coles, & Donchin, 1983) and were averaged to create composite frontocentral
individual averages were constructed for each stim- variables, and 2 (condition: caregiver, stranger) · 3
ulus type (familiar face, unfamiliar face) using (laterality: left, central, right) · 3 (group: IG, FCG,
100 ms prior to stimulus onset for baseline correc- NIG) repeated measures ANOVAs were conducted.
tion. A minimum of 10 trials per condition average At the baseline assessment, the between-subjects
were required for inclusion in the final sample. A factor group had only two levels (IG and NIG) as
separate grand mean was created for each condition random assignment had not occurred yet, and an
by averaging the individual subject averages additional between-subjects factor, age (younger,
together. older), was included to take into account the wide
Grand means were inspected to identify compo- range of ages of the children at the baseline assess-
nents of interest. At all three assessments, three ment. When the omnibus ANOVAs revealed signifi-
occipital components (P1, N170, and P400) and two cant main effects, post hoc comparisons were
frontocentral components (P250 and Nc) were carried out using t tests with a Bonferroni correction
analyzed. Time windows that captured the compo- for multiple comparisons. When the omnibus
nents of interest were identified separately for each ANOVAs revealed significant interaction effects,
assessment, as previous research has shown that independent samples t tests, paired-sample t tests,
there are maturational changes in the latency of or one-way ANOVAs were conducted.
various components across age (Batty & Taylor,
2006; Taylor et al., 2004; Webb, Long, & Nelson,
2005). Thus, the time windows for some components Baseline Assessment
differed in duration and ⁄ or starting latency across
Occipital Components
assessments.
P1 (90–200 ms). There was a main effect of group
for peak amplitude, F(1, 114) = 12.62, p = .001,
gp2 = .10, indicating that the P1 was larger in the
Results
NIG than the IG (Figure 1). There was also a Hemi-
At each assessment, the peak amplitude and latency sphere · Age interaction for the peak amplitude,
of each of the five components were analyzed. For F(1, 114) = 8.41, p = .004, gp2 = .07. Follow-up t
occipital components, 2 (condition: caregiver, stran- tests revealed that older children showed larger
ger) · 2 (hemisphere: left, right) · 3 (group: IG, amplitudes than younger children over the right

Figure 1. Grand averaged ERP waveforms of the IG and NIG at the baseline assessment over the right occipital electrode (collapsed
across conditions). The x-axis represents latency in milliseconds (ms), and the y-axis represents amplitude in microvolts (lV).
Note. ERP = event-related potential; IG = institutionalized group; NIG = never-institutionalized group.
 Early Experience and Face Recognition 1045

Figure 2. Group · Age interaction for P1 latency at the baseline

assessment. Error bars represent standard error of the mean.
Note. IG = institutionalized group; NIG = never-institutionalized
group.

hemisphere only, t(116) = )2.62, p = .01. Latency

analyses revealed a Group · Age interaction, F(1,
114) = 4.12, p = .045, gp2 = .04. Follow-up t tests
revealed that older NIG children showed signifi-
cantly faster latency to the P1 than older IG chil-
dren, t(58) = 2.94, p = .005, whereas younger NIG
and IG children showed similar latencies to the P1,
t(56) = )0.25, p > .10 (Figure 2).
N170 (150–300 ms).. For peak amplitude, there Figure 3. Condition · Age · Group interaction for P400 latency
was a main effect of age, F(1, 114) = 5.43, p = .022, at the baseline assessment. Error bars represent standard error of
gp2 = .04, in that younger children showed a signif- the mean.
icantly larger N170 than older children. There were
no main effects or interactions for the N170 latency.
P400 (250–500 ms). There was a significant Con- p = .005, gp2 = .05. Post hoc comparisons revealed
dition · Lead · Age · Group interaction for peak that the amplitude of the P250 was larger over mid-
amplitude, F(1, 114) = 4.32, p = .04. Due to the com- line than left leads. There were no significant main
plexities involved in interpreting four-way interac- effects or interactions for P250 latency.
tions, no follow-up tests were conducted. For the Nc (350–650 ms).. There was a main effect of
P400 latency, there was a main effect of age, F(1, laterality for peak amplitude, F(2, 232) = 5.60,
114) = 3.96, p = .049, gp2 = .03, which was qualified p = .004, gp2 = .05, indicating that Nc amplitude
by a significant Condition · Age · Group interac- was larger over midline than left leads. There
tion, F(1, 114) = 5.55, p = .02, gp2 = .05 (Figure 3). were no main effects or interactions for Nc
Follow-up analyses suggested that there were no latency.
significant differences for P400 latency in the IG as
a function of condition or age. However, in the
Summary
NIG, younger children showed faster P400 latency
in response to caregiver’s face compared to stran- Never-institutionalized children showed signifi-
ger’s face, t(17) = )3.30, p = .004, whereas older cantly larger amplitudes for the P1 than institution-
children did not, t(19) = 0.61, p > .10. There was alized children, and among the older children,
also a significant Condition · Hemisphere · Group never-institutionalized children showed signifi-
interaction, F(1, 114) = 4.71, p = .032, gp2 = .04; cantly faster P1 latencies than institutionalized chil-
however, there were no significant differences upon dren. Additionally, only the never-institutionalized
follow-up. children processed the caregiver and stranger faces
differently; specifically, they showed faster P400
latency for the caregiver face versus the stranger
Frontocentral Components
face, whereas the institutionalized children did not.
P250 (175–375 ms).. For peak amplitude, there Surprisingly, there were no group or condition
was a main effect of laterality, F(2, 232) = 5.76, effects over frontocentral electrodes.
1046 Moulson, Westerlund, Fox, Zeanah, and Nelson

Figure 4. Grand averaged ERP waveforms of the IG, FCG, and NIG at the 30-month assessment over the right occipital electrode
(collapsed across conditions). The x-axis represents latency in milliseconds (ms), and the y-axis represents amplitude in microvolts (lV).
Note. ERP = event-related potential; IG = institutionalized group; FCG = foster care group; NIG = never-institutionalized group.

30-Month Assessment
Occipital Components
P1 (80–190 ms).. For peak amplitude, there was a
main effect of hemisphere, F(1, 90) = 8.95, p = .004,
gp2 = .09, indicating that P1 amplitude was larger
over the right than left hemisphere. There was also a
marginal main effect of group, F(2, 90) = 2.97,
p = .057, gp2 = .06. Post hoc comparisons revealed
that the NIG showed significantly larger amplitudes
than the IG, whereas the FCG was not different from
either group (Figure 4). Latency analyses revealed a Figure 5. Condition · Group interaction for P1 latency at the 30-
main effect of hemisphere, F(1, 90) = 6.58, p = .012, month assessment. Error bars represent standard error of the
mean.
gp2 = .07, in that the latency to the P1 was faster Note. IG = institutionalized group; FCG = foster care group;
over the right than the left hemisphere. There was NIG = never-institutionalized group.
also a significant Condition · Group interaction,
F(2, 90) = 5.26, p = .007, gp2 = .10. Follow-up t tests was a significant Hemisphere · Group interaction,
revealed that the P1 latency was faster in response F(2, 90) = 3.21, p = .045, gp2 = .07. Follow-up analy-
to the caregiver face than the stranger face in the IG ses revealed that the NIG had significantly faster
only, t(32) = )2.54, p = .016 (Figure 5). Neither the latencies than the IG over the right hemisphere
NIG nor the FCG showed different P1 latencies for only, F(2, 94) = 4.51, p = .013 (Figure 6). The FCG
caregiver versus stranger (all ps > .10). was not significantly different from either the NIG
N170 (175–325 ms).. There was a main effect of or IG in terms of N170 latency.
hemisphere for the peak amplitude, F(1, 90) = 4.87, P400 (250–500 ms).. For peak amplitude, there
p = .03, gp2 = .05, that was qualified by a significant was a main effect of hemisphere, F(1, 90) = 4.77,
Condition · Hemisphere interaction, F(1, 90) = p = .032, gp2 = .05, indicating that the P400 was
4.09, p = .046, gp2 = .04, but no Condition · Group larger over the right than left hemisphere. There
interaction, p > .10. Follow-up analyses revealed were no main effects or interactions for P400 latency.
that N170 amplitude was larger over the left than
the right hemisphere for the caregiver condition
Frontocentral Components
only, t(92) = )3.12, p = .002. There were no hemi-
spheric differences for the stranger condition, P250 (185–350 ms). There was a main effect of
t(92) = )1.14, p > .10. For the N170 latency, there laterality for peak amplitude, F(2, 186) = 9.40,
 Early Experience and Face Recognition 1047

(all ps > .10) indicating that the response to stran-

ger’s face was both larger in amplitude and shorter
in latency than the response to caregiver’s face in
all three groups (Figure 7). There was also a main
effect of laterality for peak amplitude, F(2,
186) = 11.60, p < .001, gp2 = .11, indicating that the
Nc, like the P250, was larger over midline than left
or right leads.

Summary
Figure 6. Hemisphere · Group interaction for N170 latency at Similar to the baseline assessment, the never-
the 30-month assessment. Error bars represent standard error of
institutionalized children showed significantly lar-
the mean.
Note. IG = institutionalized group; FCG = foster care group; ger amplitudes for the P1 than institutionalized chil-
NIG = never-institutionalized group. dren at the 30-month assessment. Children who had
been placed in foster care showed some recovery, in
that their P1 amplitudes were intermediate between
p < .001, gp2 = .09, indicating that the P250 was the institutionalized and never-institutionalized
larger over midline than left or right leads. For children. There was also a difference among groups
P250 latency, there was a main effect of condition, for N170 latency; never-institutionalized children
F(1, 93) = 11.68, p = .001, gp2 = .11, but no Condi- showed faster N170 latencies than institutionalized
tion · Group interaction (p > .10) indicating that children over the right hemisphere only. In terms of
the response to the stranger’s face was significantly condition differences, only the institutionalized chil-
faster than the response to the caregiver’s face in all dren showed differential processing of caregiver
three groups (Figure 7). There was also a main and stranger faces over occipital electrodes (P1
effect of laterality, F(2, 186) = 3.36, p = .039, gp2 = latency was faster for caregiver vs. stranger). Sur-
.04. Post hoc comparisons revealed that the latency prisingly, however, the three groups of children
to the P250 was faster over midline that left leads. showed identical condition differences for the
Nc (350–550 ms). There was a main effect of con- frontocentral components P250 and Nc. The Nc was
dition for peak amplitude, F(1, 93) = 7.31, p = .008, larger in amplitude, and both the P250 and the Nc
gp2 = .07 and latency, F(1, 93) = 9.62, p = .003, were faster in latency, in response to stranger com-
gp2 = .09, but no Condition · Group interactions pared to caregiver in all three groups of children.

Figure 7. Grand averaged ERP waveforms of the IG, FCG, and NIG at the 30-month assessment over the middle frontal electrode
(collapsed across groups). The x-axis represents latency in milliseconds (ms), and the y-axis represents amplitude in microvolts (lV).
Note. ERP = event-related potential; IG = institutionalized group; FCG = foster care group; NIG = never-institutionalized group.
1048 Moulson, Westerlund, Fox, Zeanah, and Nelson

Figure 8. Grand averaged ERP waveforms of the IG, FCG, and NIG at the 42-month assessment over the right occipital electrode
(collapsed across conditions). The x-axis represents latency in milliseconds (ms), and the y-axis represents amplitude in microvolts
(lV).
Note. ERP = event-related potential; IG = institutionalized group; FCG = foster care group; NIG = never-institutionalized group.

42-Month Assessment over the right than the left hemisphere, F(1,
72) = 6.54, p = .013, gp2 = .08. There was also a
Occipital Components
significant Condition · Group interaction for P1
P1 (80–210 ms).. There was a main effect of group latency, F(2, 72) = 3.62, p = .032, gp2 = .09; however,
for the peak amplitude, F(2, 72) = 3.47, p = .037, there were no significant differences upon follow-
gp2 = .09. Post hoc comparisons revealed that the up.
NIG showed significantly larger P1 amplitudes than N170 (160–300 ms).. For peak amplitude, there
the IG, whereas the FCG did not differ from either was a main effect of condition, F(1, 72) = 3.97,
group (Figure 8). This was qualified by a signifi- p = .05, gp2 = .05, in that the response to the stran-
cant Condition · Group interaction, F(2, 72) = 3.40, ger’s face was significantly larger overall than the
p = .039, gp2 = .09. Follow-up tests revealed that the response to the caregiver’s face. However, this find-
NIG showed larger P1 amplitudes than the IG for ing was qualified by a significant Condi-
the stranger condition, F(2, 72) = 4.91, p = .01, but tion · Group interaction, F(2, 72) = 3.29, p = .043,
not the caregiver condition, F(2, 72) = 1.53, p > .10 gp2 = .08, which revealed that the IG showed a
(Figure 9). Additionally, P1 amplitude was larger

Figure 9. Condition · Group interaction for P1 amplitude at the Figure 10. Condition · Group interaction for N170 amplitude at
42-month assessment. Error bars represent standard error of the the 42-month assessment. Error bars represent standard error of
mean. the mean.
Note. IG = institutionalized group; FCG = foster care group; Note. IG = institutionalized group; FCG = foster care group;
NIG = never-institutionalized group. NIG = never-institutionalized group.
 Early Experience and Face Recognition 1049

significantly larger N170 for the stranger face com- gp2 = .05, revealing that the latency to the Nc was
pared to the caregiver face, F(1, 22) = 10.23, faster over right than midline leads.
p = .004 (Figure 10), whereas the FCG and NIG did
not (all ps > .10). There were no main effects or
Summary
interactions for N170 latency.
P400 (250–500 ms).. For peak amplitude, there Similar to both the baseline and 30-month assess-
was a main effect of condition, F(1, 69) = 4.64, ments, never-institutionalized children showed
p = .035, gp2 = .06, but no Condition · Group inter- larger amplitudes for the P1 than institutionalized
action (p > .10), indicating that the response to the children at the 42-month assessment. Foster care
caregiver’s face was significantly larger than the children showed some recovery, in that their P1
response to the stranger’s face in all three groups. amplitudes were intermediate between the other
There were no main effects or interactions for the two groups. Institutionalized children, but not
P400 latency. never-institutionalized or foster care children,
showed a condition difference for the N170
component; specifically, the N170 was larger in
Frontocentral Components
response to the stranger face compared to the
P250 (185–350 ms).. For peak amplitude, there caregiver face. Surprisingly, all three groups
was a main effect of laterality, F(2, 138) = 9.69, showed identical condition differences for the P400
p < .001, gp2 = .12. Post hoc comparisons indicated over occipital electrodes and the Nc over frontocen-
that the P250 was larger over midline (M = 10.35, tral electrodes. The P400 was larger in response to
SD = 4.88) than left (M = 9.33, SD = 4.56) or right the caregiver face compared to the stranger face;
(M = 9.23, SD = 4.38) leads. There were no signifi- conversely, the Nc was both larger and faster in
cant main effects or interactions for P250 latency. response to the stranger face compared to the care-
Nc (350–550 ms).. There were main effects of con- giver face.
dition for both peak amplitude, F(1, 69) = 6.49,
p = .013, gp2 = .09 and latency, F(1, 69) = 14.38,
Analysis of Intervention Effects
p < .001, gp2 = .17, but no Condition · Group inter-
actions (all ps > .10), indicating that the response to The unique design of the BEIP provided an
stranger’s face was both larger and faster than the opportunity to examine the effects of the timing
response to caregiver’s face in all three groups and duration of the foster care intervention on ERP
(Figure 11). Additionally, there was a main effect of outcomes. Because children were placed in foster
laterality for latency, F(2, 138) = 3.37, p = .04, care at varying ages depending on the age at which

Figure 11. Grand averaged ERP waveforms of the IG, FCG, and NIG at the 42-month assessment over the middle frontal electrode
(collapsed across groups). The x-axis represents latency in milliseconds (ms), and the y-axis represents amplitude in microvolts (lV).
Note. ERP = event-related potential; IG = institutionalized group; FCG = foster care group; NIG = never-institutionalized group.
1050 Moulson, Westerlund, Fox, Zeanah, and Nelson

Table 2
Descriptive Statistics for Occipital Components for the IG, Early- and Late-Placed FCG, and NIG at 42-Month Assessment

Group

Component Measure IG Early-placed FCG Late-placed FCG NIG

P1 Amplitude 16.53 (7.37) 19.41 (7.80) 17.19 (6.81) 22.69 (9.91)

Latency 134.67 (24.13) 136.81 (17.55) 127.78 (16.39) 128.74 (22.40)
N170 Amplitude )3.67 (5.37) )3.27 (3.33) )5.97 (3.75) )3.57 (8.10)
Latency 231.62 (34.89) 232.52 (25.52) 221.03 (22.55) 231.72 (28.74)
P400 Amplitude 13.07 (4.41) 12.80 (4.57) 12.83 (2.54) 15.36 (6.23)
Latency 370.22 (44.53) 378.30 (43.64) 363.27 (41.52) 354.88 (57.48)

Note. Amplitude is reported in microvolts (lV) and latency is reported in milliseconds (ms). Values are given as M (SD).
IG = institutionalized group; FCG = foster care group; NIG = never-institutionalized group.

they were recruited into the study, which occurred children who were older when placed in foster care
at anytime between 5 and 31 months of age, we tended to have larger (i.e., more negative) ampli-
were able to compare the outcomes of children who tudes for the N170.
were placed in foster care early versus late. First, to
evaluate the overall effect of the intervention, we
compared the amplitudes and latencies of the
Discussion
occipital components (P1, N170, and P400) in the
IG and FCG at 42 months of age (the NIG was used This study used ERPs to examine the neural
for comparison purposes only). Perhaps surpris- correlates of face recognition in three groups of
ingly, independent sample t tests revealed no sig- children: currently institutionalized, previously
nificant differences between the IG and FCG at institutionalized (subsequently placed in foster
42 months of age. However, the wide range of ages care), and never institutionalized. We hypothe-
at which children in the FCG were placed in foster sized that the psychosocial deprivation associated
care could have obscured any effects of the foster with institutional care would lead to abnormalities
care intervention. It is possible that children placed in the neural response to faces among institution-
in foster care earlier showed greater improvements alized children, and that foster care would be
in their ERP outcomes than children placed in fos- at least somewhat effective in remediating any
ter care later. deficits in previously institutionalized children.
To evaluate whether the timing of placement The results of this study not only supported these
into foster care, and correspondingly its duration, hypotheses but also revealed some unexpected
affected ERP outcomes at 42 months of age, two findings.
analysis strategies were used. First, the FCG was The first major finding was that institutionalized
divided into an early-placed group and a late- children showed strikingly smaller amplitudes for
placed group based on a median split at the P1 component than never-institutionalized chil-
25.5 months of age, and one-way ANOVAs com- dren across all three assessments. Although signifi-
paring the IG, early-placed FCG, and late-placed cant group differences were not found consistently
FCG were conducted for the amplitudes and laten- for the other occipital components, inspection of
cies of the occipital components at 42 months of the occipital ERP waveforms (Figures 1, 4, and 8)
age (see Table 2). Second, Pearson correlations suggests that this pattern of decreased amplitude in
between age at placement and the amplitudes and the institutionalized children continued across the
latencies of the occipital components were con- entire ERP epoch. Institutionalized children in this
ducted for the FCG only. One-way ANOVAs sample also showed decreased amplitudes for the
revealed no significant differences among the early- P1 and P400 occipital components at all three
placed FCG, late-placed FCG, and IG for any of the assessments during a facial emotion processing
occipital components. Correlational analyses task, in which ERPs were recorded while they
revealed only one significant correlation, between watched pictures of emotional faces (Moulson, Fox,
age at placement and N170 amplitude, r = ).39, Zeanah, & Nelson, 2009). These findings suggest
p = .03. As the N170 is a negative-going compo- that institutionalized children show cortical hypoa-
nent, the negative correlation here indicates that rousal while perceiving faces. Because ERPs were
 Early Experience and Face Recognition 1051

not recorded in response to stimuli other than faces, aroused brains. The underlying neural cause of this
it is unclear whether this cortical hypoarousal deficit remains unknown. One might speculate,
extends to the processing of other stimuli. EEG and however, that one source of this hypoactivation
neuroimaging findings from children with histories may be an error in pruning, specifically, that the
of institutionalization do suggest that this cortical lack of normative experience may lead to over-
hypoarousal may be pervasive, rather than specific pruning of neurons or synapses, both of which
to children’s responses to faces. However, it is also would lead to reductions in brain activity and a
possible that certain neural systems, including corresponding reduction in head circumference
those that underlie face perception, are particularly (Johnson, 2006; Smyke et al., 2007).
vulnerable to the early adverse experience of insti- The second major finding of this study was
tutionalization. that children who were placed in foster care
Marshall and Fox (2004) found that institutional- following early institutionalization showed some
ized children in the BEIP showed increased low evidence of improvement. The amplitude of the
frequency (theta) EEG activity and decreased P1 component among the foster care children fell
high-frequency (alpha and beta) EEG activity com- between the P1 amplitudes of the institutionalized
pared to never-institutionalized children at the and never-institutionalized children at the 30- and
baseline assessment. A study with previously insti- 42-month assessments. Although the foster care
tutionalized children found that 18-month-old tod- group was not significantly different from either
dlers who had been adopted following early group, this result suggests that children placed in
institutional care also had increased low-frequency foster care do, to some extent, recover from the
EEG power and decreased high-frequency EEG psychosocial deprivation that they experienced
power compared to children raised with their bio- early in life. Surprisingly, the timing of placement
logical families (Tarullo, 2008), which replicated the into foster care was unrelated to the children’s
findings in the BEIP sample (Marshall & Fox, 2004). ERP outcomes at 42 months of age, a result that
Chugani et al. (2001), using positron emission is consistent with the results of the facial emotion
tomography, found reduced metabolic activity in ERP study conducted with this sample of children
limbic (i.e., medial temporal lobe) and paralimbic (Moulson et al., 2009). It seems that even
(i.e., areas of prefrontal cortex) regions of the brain relatively late placement in foster care was
in a group of previously institutionalized children sufficient to initiate improvements in neurologic
compared to a group of family-reared children. The functioning.
previously institutionalized children had spent an In contrast, many previous studies with postin-
average of 5.5 years in their adoptive homes, sug- stitutionalized children have demonstrated that ear-
gesting that the brain dysfunction associated with lier adoption leads to more favorable behavioral,
early institutional care persists long after placement cognitive, and social outcomes than later adoption
into a typical environment. (Beckett et al., 2006; O’Connor & Rutter, 2000;
Children who live in noninstitutional high-risk O’Connor et al., 2000; Rutter, 1998; Rutter et al.,
environments in North America also show signs of 2007). That there was no relation between the
cortical hypoarousal. In a study of Mexican chil- timing of intervention and ERP outcomes in this
dren, Otero and colleagues (Otero, 1997; Otero, sample of children was particularly unexpected
Pliego-Rivero, Fernandez, & Ricardo, 2003) showed because previous findings from the BEIP also indi-
that 4- to 6-year-old children living in families of cate that earlier intervention leads to better out-
low socioeconomic status (SES) displayed increased comes. Children who were placed in foster care
low-frequency power and decreased high- before 2 years of age had significantly better cogni-
frequency power compared to children living in tive outcomes at 42 and 54 months than children
low-risk families. And in a study of 7- to 12-year- placed after 2 years of age (Nelson et al., 2007).
old children in California, Kishiyama, Jimenez, Children who were placed before 2 years of age
Perry, Boyce, and Knight (2007) found that low-SES also showed more mature patterns of EEG activity
children showed smaller amplitudes for ERP com- (i.e., higher EEG alpha power and lower EEG
ponents at central and occipital leads compared to short-distance coherence) than children who
high-SES children in an oddball paradigm. Thus, remained in the institution, whereas children
the results of this study bolster the existing evi- placed after 2 years of age did not (Marshall, Reeb,
dence that children who have experienced early Fox, Nelson, & Zeanah, 2008).
deprivation associated with institutional or other- In light of this previous research, it is difficult to
wise inadequate care have chronically under- interpret the current finding and its implications
1052 Moulson, Westerlund, Fox, Zeanah, and Nelson

for our understanding of sensitive periods in the The third major finding of this study was that
development of face processing. Although there are all three groups of children showed differing neu-
several plausible interpretations, distinguishing ral responses to the caregiver and stranger faces.
among them is impossible because we cannot deter- Surprisingly, the results also suggest that all three
mine based on the current data whether children groups of children processed the faces similarly,
placed in foster care will ever recover fully (i.e., despite the disparity in the quality of their early
show ERP components that do not differ in ampli- caregiving environments. At the 30- and 42-month
tude, latency, or morphology from those of children assessments, all three groups showed a larger
who have never experienced institutionalization). amplitude and shorter latency response to the
Additionally, the lack of a nonface control stimulus stranger face compared to the caregiver face for
in this study makes it difficult to determine the Nc over frontocentral electrodes, a finding
whether any conclusions drawn about sensitive that is consistent with previous research. Carver
periods based on these findings are specific to the et al. (2003) documented a developmental progres-
domain of face perception or relevant for our under- sion in the response to mother and stranger faces
standing of brain development more generally. in typically developing infants, which they attrib-
If children placed in foster care do eventually uted to the changing significance of mother’s face
catch up to never-institutionalized children, it and strangers’ faces over the first years of life.
would suggest either that there is no sensitive per- They found that younger infants (< 24 months of
iod, or that the sensitive period remains open age) showed a larger Nc response to mother’s
beyond the age at which these children were placed face than stranger’s face, whereas older infants
in foster care (i.e., beyond 30 months). The former (> 45 months of age) showed a larger Nc response
possibility is unlikely given the findings that early to stranger’s face than mother’s face. Infants
visual deprivation results in long-term deficits in between 24 and 45 months of age showed no dif-
face perception (Geldart et al., 2002; Le Grand ferences in their Nc responses to mother and
et al., 2001, 2003); however, the latter possibility is stranger faces, suggesting that it is a time of tran-
consistent with previous research on the reversibil- sition in terms of the social significance of famil-
ity of the other-race effect that suggests that the iar and unfamiliar faces.
face processing system remains plastic at least into The children in this study showed no differences
middle childhood (Sangrigoli et al., 2005). How- in the amplitude or latency of the Nc or P250 at the
ever, another possible interpretation of this finding baseline assessment, which suggests that either (a)
is that children’s initial improvement is unrelated the baseline assessment captured the transition
to their age at intervention, but further improve- stage reported by Carver et al. (2003), or (b) the
ment depends on having been placed in foster care range of ages at the baseline assessment (5–
early. If this were the case, we would predict that 31 months) obscured any developmental progres-
both earlier and later placed children would sion in processing familiar and unfamiliar faces
improve initially, but only earlier placed children (although the children were split into younger and
would continue on an accelerated developmental older groups for the baseline analyses, the children
trajectory and eventually catch up to family-reared within these groups still ranged widely in age). By
children. This pattern of results would suggest that 30 and 42 months of age, however, children in this
there is a sensitive period that narrows early, but study consistently showed larger amplitude and
the effects of that sensitive period are not realized shorter latency frontocentral components in
until later in development (see Maurer, Mondloch, response to stranger. Although this is somewhat
& Lewis, 2007, for a discussion of such ‘‘sleeper earlier than Carver et al. reported, the overall pat-
effects’’ following visual deprivation). Of course, it tern of results suggests that there is a developmen-
is also possible that none of the children who expe- tal progression among all three groups of children
rienced early deprivation followed by intervention in this study in terms of their neural response to
will show a full recovery, which would suggest caregiver and stranger faces.
early closure of the sensitive period for the neural There were also condition effects for occipital
system underlying face perception. Again, the lack components, and some evidence of differential
of data beyond 42 months of age in this study processing of caregiver and stranger faces among
makes it impossible to distinguish among these the groups. At the baseline assessment, the fam-
possibilities or draw firm conclusions regarding the ily-reared children showed a faster latency to the
existence and timing of sensitive periods in this P400 for the caregiver face compared to the
developmental domain. stranger face, whereas the institutionalized chil-
 Early Experience and Face Recognition 1053

dren did not. At the 30-month assessment, the dren. These findings have implications both for
institutionalized children showed a faster latency our understanding of the development of the
to the P1 for the caregiver face compared to the neural systems that subserve face processing and,
stranger face, whereas the family-reared and fos- more importantly, for the millions of children
ter care children did not. And at the 42-month around the world who are reared in environ-
assessment, only the institutionalized children ments characterized by psychosocial adversity
showed a larger N170 for the stranger face and deprivation.
compared to the caregiver face. The lack of con-
sistency across assessments makes it difficult to
interpret the extent of occipital processing differ-
References
ences among the three groups of children, but it
is noteworthy that there was evidence of differ- Adolphs, R. (2002). Recognizing emotion from facial
ential processing of the familiar and unfamiliar expressions: Psychological and neurological mecha-
faces over occipital leads. In contrast, the facial nisms. Behavioral and Cognitive Neuroscience Reviews,
emotion ERP task conducted with these children 1(1), 21–62.
Batty, M., & Taylor, M. J. (2006). The development of
revealed no differential processing of emotion
emotional face processing during childhood. Develop-
over occipital leads, although there was differen-
mental Science, 9, 207–220.
tial processing of emotion over frontocentral Beckett, C., Maughan, B., Rutter, M., Castle, J., Colvert,
leads (see Moulson et al., 2009). This suggests E., Groothues, C., et al. (2006). Do the effects of early
that the neural correlates of processing facial severe deprivation on cognition persist into early ado-
identity and facial emotion are at least somewhat lescence? Findings from the English and Romanian
separable in young children, which is consistent Adoptees Study. Child Development, 77, 696–711.
with the adult neuroimaging literature (Haxby Bentin, S., Allison, T., Puce, A., Perez, E., & McCarthy,
et al., 2002) and findings from congenital proso- G. (1996). Electrophysiological studies of face percep-
pagnosia (Humphreys, Avidan, & Behrmann, tion in humans. Journal of Cognitive Neuroscience, 8,
2007). 551–565.
Carver, L. J., Dawson, G., Panagiotides, H., Meltzoff, A.
It is surprising that the neural processing of
N., McPartland, J., Gray, J., et al. (2003). Age-related
familiar and unfamiliar faces is remarkably simi-
differences in neural correlates of face recognition dur-
lar in institutionalized and family-reared children. ing the toddler and preschool years. Developmental Psy-
Based on the adverse early rearing environment chobiology, 42, 148–159.
that institutionalized children suffer, we expected Chugani, H. T., Behen, M. E., Muzik, O., Juhasz, C.,
that they might show no differences at all in the Nagy, F., & Chugani, D. C. (2001). Local brain func-
neural processing of familiar and unfamiliar tional activity following early deprivation: A study of
faces. The fact that institutionalized children postinstitutionalized Romanian orphans. Neuroimage,
show processing differences that are similar to 14, 1290–1301.
family-reared children suggests that the atypical Courchesne, E., Ganz, L., & Norcia, A. M. (1981). Event-
experience that they receive over the first years related potentials to human faces in infants. Child
Development, 52, 804–811.
of life is sufficient to set up the neural architec-
de Haan, M., & Nelson, C. A. (1997). Recognition of the
ture underlying at least the perceptual discrimi-
mother’s face by six-month-old infants: A neurobehav-
nation of faces. It is possible, however, that these ioral study. Child Development, 68, 187–210.
children may show deficits on more subtle face de Haan, M., & Nelson, C. A. (1999). Brain activity differ-
processing tasks later in life, such as processing entiates face and object processing in 6-month-old
faces configurally (a hallmark of expert face pro- infants. Developmental Psychology, 35, 1113–1121.
cessing) or identifying complex emotions Field, T. M. (1983). Discrimination and imitation of facial
expressed in faces. expressions by term and preterm neonates. Infant
In summary, the psychosocial deprivation Behavior and Development, 6, 485–489.
experienced by institutionalized children leads Fisher, L., Ames, E. W., Chisholm, K., & Savoie, L. (1997).
to pervasive cortical hypoarousal compared to Problems reported by parents of Romanian orphans
adopted to British Columbia. International Journal of
family-reared children. Despite this cortical
Behavioral Development, 20, 67–82.
hypoarousal, institutionalized children processed
Geldart, S., Mondloch, C. J., Maurer, D., de Schonen, S.,
familiar and unfamiliar faces similarly to family- & Brent, H. P. (2002). The effect of early visual depriva-
reared children. Placement in high-quality foster tion on the development of face processing. Develop-
care led to a partial amelioration of the cortical mental Science, 5, 490–501.
hypoarousal in previously institutionalized chil-
1054 Moulson, Westerlund, Fox, Zeanah, and Nelson

Gratton, G., Coles, M. G., & Donchin, E. (1983). A new Le Grand, R., Mondloch, C. J., Maurer, D., & Brent, H. P.
method for off-line removal of ocular artifact. Electro- (2003). Expert face processing requires visual input to
encephalography and Clinical Neurophysiology, 55, 468– the right hemisphere during infancy. Nature Neurosci-
484. ence, 6, 1108–1112.
Gunnar, M. R. (2001). Effects of early deprivation: Find- Lewis, T. L., & Maurer, D. (2005). Multiple sensitive peri-
ings from orphanage-reared infants and children. In M. ods in human visual development: Evidence from visu-
Luciana & C. A. Nelson (Eds.), Handbook of developmen- ally deprived children. Developmental Psychobiology, 46,
tal cognitive neuroscience (pp. 617–630). Cambridge, MA: 163–183.
MIT Press. Marshall, P. J., & Fox, N. A. (2004). A comparison of the
Halit, H., Csibra, G., Volein, A., & Johnson, M. H. (2004). electroencephalogram between institutionalized and
Face-sensitive cortical processing in early infancy. Jour- community children in Romania. Journal of Cognitive
nal of Child Psychology and Psychiatry, 45, 1228–1234. Neuroscience, 16, 1327–1338.
Halit, H., de Haan, M., & Johnson, M. H. (2003). Cortical Marshall, P. J., Reeb, B. C., Fox, N. A., Nelson, C. A., &
specialisation for face processing: Face-sensitive event- Zeanah, C. H. (2008). Effects of early intervention on
related potential components in 3- and 12-month-old EEG power and coherence in previously institutiona-
infants. Neuroimage, 19, 1180–1193. lized children in Romania. Development and Psycho-
Handy, T. C. (Ed.). (2005). Event-related potentials: A meth- pathology, 20, 861–880.
ods handbook. Cambridge, MA: MIT Press. Maurer, D., Mondloch, C. J., & Lewis, T. L. (2007). Sleeper
Haxby, J. V., Hoffman, E. A., & Gobbini, M. I. (2002). effects. Developmental Science, 10, 40–47.
Human neural systems for face recognition and social Meissner, C. A., & Brigham, J. C. (2001). Thirty years of
communication. Biological Psychiatry, 51, 59–67. investigating the own-race bias in memory for faces: A
Humphreys, K., Avidan, G., & Behrmann, M. (2007). A meta-analytic review. Psychology, Public Policy, and Law,
detailed investigation of facial expression processing 7, 3–35.
in congenital prosopagnosia as compared to acquired Millum, J., & Emanuel, E. J. (2007). The ethics of interna-
prosopagnosia. Experimental Brain Research, 176, 356– tional research with abandoned children. Science, 318,
373. 1874–1875.
Iidaka, T., Omori, M., Murata, T., Kosaka, H., Yonekura, Moulson, M. C., Fox, N. A., Zeanah, C. H., & Nelson,
Y., Tomohisa, O., et al. (2001). Neural interaction of the C. A. (2009). Early adverse experiences and the neuro-
amygdala with the prefrontal and temporal cortices in biology of facial emotion processing. Developmental
the processing of facial expressions as revealed by Psychology, 45, 17–30.
fMRI. Journal of Cognitive Neuroscience, 13, 1035–1047. Nelson, C. A. (1994). Neural correlates of recognition
Johnson, D. E. (2006). Physical growth in children: Effects of memory in the first postnatal year of life. In G. Dawson
early deprivation. Paper presented at the American Asso- & K. Fischer (Eds.), Human behavior and the developing
ciation for the Advancement of Science, St. Louis, MO. brain (pp. 269–313). New York: Guilford.
Kanwisher, N., McDermott, J., & Chun, M. M. (1997). The Nelson, C. A. (2001). The development and neural bases
fusiform face area: A module in human extrastriate of face recognition. Infant and Child Development, 10(1–
cortex specialized for face perception. Journal of Neuro- 2), 3–18.
science, 17, 4302–4311. Nelson, C. A., & de Haan, M. (1996). Neural correlates
Kelly, D. J., Liu, S., Ge, L., Quinn, P. C., Slater, A., Lee, of infants’ visual responsiveness to facial expressions
K., et al. (2007). Cross-race preferences for same-race of emotion. Developmental Psychobiology, 29(7), 577–
faces extend beyond the African versus Caucasian con- 595.
trast in 3-month-old infants. Infancy, 11, 87–96. Nelson, C. A., Zeanah, C. H., Fox, N. A., Marshall, P. J.,
Kelly, D. J., Quinn, P. C., Slater, A., Lee, K., Ge, L., & Pas- Smyke, A., & Guthrie, D. (2007). Cognitive recovery in
calis, O. (2007). The other-race effect develops during socially deprived young children: The Bucharest Early
infancy: Evidence of perceptual narrowing. Psychologi- Intervention Project. Science, 318, 1937–1940.
cal Science, 18, 1084–1089. O’Connor, T. G., Bredenkamp, D., & Rutter, M. (1999).
Kelly, D. J., Quinn, P. C., Slater, A. M., Lee, K., Gibson, Attachment disturbances and disorders in children
A., Smith, M., et al. (2005). Three-month-olds, but not exposed to early severe deprivation. Infant Mental
newborns, prefer own-race faces. Developmental Science, Health Journal, 20, 10–29.
8(6), F31–F36. O’Connor, T. G., & Rutter, M. (2000). Attachment disor-
Kishiyama, M. M., Jimenez, A. M., Perry, L. M., Boyce, der behavior following early severe deprivation:
W. T., & Knight, R. T. (2007). Social disparity affects pre- Extension and longitudinal follow-up. Journal of the
frontal development in children: An ERP study. Paper pre- American Academy of Child and Adolescent Psychiatry,
sented at the annual meeting of the Cognitive 39, 703–712.
Neuroscience Society, New York. O’Connor, T. G., Rutter, M., Beckett, C., Keaveney, L., &
Le Grand, R., Mondloch, C. J., Maurer, D., & Brent, H. P. Kreppner, J. M. (2000). The effects of global severe pri-
(2001). Early visual experience and face processing. vation on cognitive competence: Extension and longitu-
Nature, 410, 890. dinal follow-up. Child Development, 71, 376–390.
 Early Experience and Face Recognition 1055

Otero, G. A. (1997). Poverty, cultural disadvantage and Rutter, M., Colvert, E., Kreppner, J. M., Beckett, C., Cas-
brain development: A study of preschool children in tle, J., Groothues, C., et al. (2007). Early adolescent out-
Mexico. Electroencephalography and Clinical Neurophysiol- comes for institutionally-deprived and non-deprived
ogy, 102, 512–516. adoptees. I: Disinhibited attachment. Journal of Child
Otero, G. A., Pliego-Rivero, F., Fernandez, T., & Ricardo, Psychology and Psychiatry, 48, 17–30.
J. (2003). EEG development in children with sociocul- Sangrigoli, S., & de Schonen, S. (2004). Recognition of
tural disadvantages: A follow-up study. Clinical Neuro- own-race and other-race faces by three-month-old
physiology, 114, 1918–1925. infants. Journal of Child Psychology and Psychiatry, 45,
Otsuka, Y., Nakato, E., Kanazawa, S., Yamaguchi, 1219–1227.
M. K., Watanabe, S., & Kakigi, R. (2007). Neural acti- Sangrigoli, S., Pallier, C., Argenti, A.-M., Ventureyra,
vation to upright and inverted faces in infants mea- V. A. G., & de Schonen, S. (2005). Reversibility of the
sured by near infrared spectroscopy. Neuroimage, 34, other-race effect in face recognition during childhood.
399–406. Psychological Science, 16, 440–444.
Parker, S. W., & Nelson, C. A. (2005a). The impact of Smyke, A. T., Koga, S. F., Johnson, D. E., Fox, N. A., Mar-
early institutional rearing on the ability to discriminate shall, P. J., Nelson, C. A., et al. (2007). The caregiving
facial expressions of emotion: An Event-Related Poten- context in institution-reared and family-reared infants
tial study. Child Development, 76, 54–72. and toddlers in Romania. Journal of Child Psychology and
Parker, S. W., & Nelson, C. A. (2005b). An event-related Psychiatry, 48(2), 210–218.
potential study of the impact of institutional rearing on Sugita, Y. (2008). Face perception in monkeys reared with
face recognition. Development and Psychopathology, 17, no exposure to faces. Proceedings of the National Academy
621–639. of Sciences of the United States of America, 105, 394–398.
Pascalis, O., de Haan, M., & Nelson, C. A. (2002). Is face Tanaka, J. W., Kiefer, M., & Bukach, C. M. (2004). A holis-
processing species-specific during the first year of life? tic account of the own-race effect in face recognition:
Science, 296(5571), 1321–1323. Evidence from a cross-cultural study. Cognition, 93, B1–
Pascalis, O., de Schonen, S., Morton, J., Derulle, C., & B9.
Fabre-Grenet, M. (1995). Mother’s face recognition in Tarullo, A. R. (2008). Neurobiological development and
neonates: A replication and an extension. Infant Behav- social communication in the wake of early social depri-
ior & Development, 18, 79–85. vation: Electrophysiological activity, joint attention,
Pascalis, O., Scott, L. S., Kelly, D. J., Shannon, R. W., and language in post-institutionalized toddlers. Disser-
Nicholson, E., Coleman, M., et al. (2005). Plasticity of tation Abstracts International: Section B: The Sciences and
face processing in infancy. Proceedings of the National Engineering, 69(6-B), 3881.
Academy of Sciences of the United States of America, 102, Taylor, M. J., Batty, M., & Itier, R. J. (2004). The faces of
5297–5300. development: A review of early face processing over
Pollak, S. D., Cicchetti, D., Hornung, K., & Reed, A. childhood. Journal of Cognitive Neuroscience, 16, 1426–
(2000). Recognizing emotion in faces: Developmental 1442.
effects of child abuse and neglect. Developmental Psy- Tzourio-Mazoyer, N., De Schonen, S., Crivello, F., Reut-
chology, 36, 679–688. ter, B., Aujard, Y., & Mazoyer, B. (2002). Neural corre-
Pollak, S. D., & Kistler, D. J. (2002). Early experience is lates of woman face processing by 2-month-old infants.
associated with the development of categorical repre- Neuroimage, 15, 454–461.
sentations for facial expressions of emotion. Proceedings Walker, P. M., & Tanaka, J. W. (2003). An encoding
of the National Academy of Sciences of the United States of advantage for own-race versus other-race faces. Percep-
America, 99, 9072–9076. tion, 32, 1117–1125.
Pollak, S. D., & Sinha, P. (2002). Effects of early experi- Webb, S. J., Long, J. D., & Nelson, C. A. (2005). A longitu-
ence on children’s recognition of facial displays of emo- dinal investigation of visual event-related potentials in
tion. Developmental Psychology, 38, 784–791. the first year of life. Developmental Science, 8(6), 605–616.
Pollak, S. D., & Tolley-Schell, S. A. (2003). Selective atten- Wismer Fries, A. B., & Pollak, S. D. (2004). Emotion
tion to facial emotion in physically abused children. understanding in post-institutionalized Eastern Euro-
Journal of Abnormal Psychology, 112, 323–338. pean children. Development and Psychopathology, 16,
Quinn, P. C., Yahr, J., Kuhn, A., Slater, A. M., & Pascalis, 355–369.
O. (2002). Representation of the gender of human faces Zeanah, C. H., Koga, S. F., Simion, B., Stanescu, A., Taba-
by infants: A preference for female. Perception, 31, caru, C. L., Fox, N. A., et al. (2006a). Ethical consider-
1109–1121. ations in international research collaboration: The
Richards, J. E. (2003). Attention affects the recognition of Bucharest Early Intervention Project. Infant Mental
briefly presented visual stimuli in infants: An ERP Health Journal, 27, 558–576.
study. Developmental Science, 6, 312–328. Zeanah, C. H., Koga, S. F., Simion, B., Stanescu, A., Taba-
Rutter, M. (1998). Developmental catch-up, and delay, caru, C. L., Fox, N. A., et al. (2006b). Response to com-
following adoption after severe global early privation. mentary: Ethical dimensions of the BEIP. Infant Mental
Journal of Child Psychology and Psychiatry, 39, 465–476. Health Journal, 27, 581–583.
1056 Moulson, Westerlund, Fox, Zeanah, and Nelson

Zeanah, C. H., Nelson, C. A., Fox, N. A., Smyke, A. T., Zeanah, C. H., Smyke, A. T., Koga, S. F., & Carlson,
Marshall, P., Parker, S. W., et al. (2003). Designing E. (2005). Attachment in institutionalized and com-
research to study the effects of institutionalization on munity children in Romania. Child Development, 76,
brain and behavioral development: The Bucharest 1015–1028.
Early Intervention Project. Development and Psychopa-
thology, 15(4), 885–907.