SEARCH

SEARCH BY CITATION

Abstract

  1. Top of page
  2. Abstract
  3. Method
  4. Results
  5. Discussion
  6. References

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 demonstrate 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 stimulus used in human social communication. Faces provide a wealth of information about other individuals, including identity, sex, age, focus of attention, and emotional state. Most adults are experts at processing faces, and quickly and effortlessly decipher information from faces during social exchanges. Although young infants have rudimentary face processing capabilities (e.g., they can recognize their mother within the first days of life, and can discriminate basic emotions within the first months of life; Field, 1983; Pascalis, de Schonen, Morton, Derulle, & Fabre-Grenet, 1995), children do not show adult-like speed and accuracy in face perception until adolescence. Although the behavioral development of face perception has been well characterized, less is known about the development of the neural systems underlying face perception. Clearly, the origins of behavioral change lie in the development of these neural systems.

Previous research has demonstrated that adults’ face expertise is subserved by a distributed network of brain areas that are preferentially involved in face processing, including the “fusiform face area” within the fusiform gyrus, the superior temporal sulcus, amygdala, and areas of the prefrontal cortex (Adolphs, 2002; Haxby, Hoffman, & Gobbini, 2002; Iidaka et al., 2001; Kanwisher, McDermott, & Chun, 1997). Fewer studies have examined the development of these neural systems, largely due to the methodological limitations involved in using neuroimaging techniques such as functional magnetic resonance imaging with developing populations (although near infrared spectroscopy shows promise for use with infants and young children; see Otsuka et al., 2007; Tzourio-Mazoyer et al., 2002). However, event-related potentials (ERPs), a noninvasive measure of the electrical activity of the brain that occurs in response to discrete stimulus events (Handy, 2005), have been used fruitfully in previous research to investigate the neural correlates of face recognition in infants and children.

Previous studies have focused on several distinct ERP components (deflections in the ongoing electroencephalogram [EEG]) that seem to reflect aspects of face processing in infants and children. The Nc, a negative-going component that occurs over frontocentral regions approximately 300–700 ms following stimulus onset, is larger in amplitude in response to mother’s face compared to a stranger’s face in 6- to 24-month-old infants (Carver et al., 2003; de Haan & Nelson, 1997). The Nc response reverses over the course of development, in that older children between the ages of 45 and 54 months show a larger Nc in response to a stranger’s face compared to their mother’s face (Carver et al., 2003), which probably reflects a developmental change in the social significance of mother’s face compared to stranger’s face. The Nc is also larger in response to fearful faces compared to happy faces (Nelson & de Haan, 1996). Based on these findings, researchers have posited that the Nc reflects attentional allocation (Courchesne, Ganz, & Norcia, 1981; Nelson, 1994; Richards, 2003). Two components located over occipitotemporal regions, the N290 and P400, respond reliably differently to faces compared to other classes of objects in infants, and are thought to reflect face-specific processing. The N290 is larger in response to human faces than monkey faces or nonface stimuli (Halit, Csibra, Volein, & Johnson, 2004; Halit, de Haan, & Johnson, 2003), and the P400 is faster in response to faces than to objects (de Haan & Nelson, 1999). Researchers have hypothesized that the N290 and P400 components are precursors of the adult N170, a face-sensitive component that reflects the structural encoding of a face (Bentin, Allison, Puce, Perez, & McCarthy, 1996). Taylor, Batty, and Itier (2004) have investigated the neural correlates of face processing in 4- to 15-year-old children, and found that both the P1 and N170 components over occipitotemporal regions were present in children across this age range and sensitive to aspects of face processing, including face inversion. The P1 and N170 differed in terms of their response properties, although both showed developmental changes with increasing age. Despite this growing body of knowledge about the neural correlates of face perception during development, little is known about the mechanisms that drive developmental change in the neural system underlying face processing.

Nelson (2001) postulates that the development of face perception is an experience-expectant, activity-dependent process, in which infants’ experiences with faces early in life shape the cortical systems that give rise to expert face processing. Evidence that typically developing infants retain the ability to discriminate faces to which they have been exposed, and lose the ability to discriminate faces to which they have not been exposed, supports this view. For example, Pascalis, de Haan, and Nelson (2002) demonstrated that 6-month-old infants, but not 9-month-old infants or adults, are able to discriminate monkey faces. Thus, between 6 and 9 months of age, infants’ face perception becomes “species-specific,” in that they lose the ability to discriminate monkey faces but retain the ability to discriminate human faces. Infants retain the ability to discriminate monkey faces at 9 months if they have been exposed to monkey faces in the form of a picture book (Pascalis et al., 2005). Interestingly, the “other-species effect” has also been demonstrated in infant monkeys who were selectively exposed to either human or monkey faces following a period of deprivation in which they saw no faces. Following face exposure, the monkeys discriminated only those types of faces (human or monkey) to which they had been exposed (Sugita, 2008).

The “other-race effect” is another example of specific face experiences shaping face processing skills. Many studies have shown that adults are poorer at discriminating and recognizing faces of other races with which they have little experience, compared to faces of their own race (Meissner & Brigham, 2001; Tanaka, Kiefer, & Bukach, 2004; Walker & Tanaka, 2003). This effect has its origin in infancy. Kelly and colleagues (Kelly, Liu, et al., 2007; Kelly et al., 2005) have demonstrated that 3-month-old infants prefer to look at faces of their own race (see Quinn, Yahr, Kuhn, Slater, & Pascalis, 2002, for evidence that 3-month-old infants also show a preference for faces matching the gender of their primary caregiver), and Sangrigoli and de Schonen (2004) have shown that 3-month-old infants are better able to discriminate faces of their own race, although this effect disappears if they are given even a short amount of exposure to other-race faces. By 9 months of age infants appear to lose the ability to discriminate faces of other races (Kelly, Quinn, et al., 2007). However, Sangrigoli, Pallier, Argenti, Ventureyra, and de Schonen (2005) found that adults of Korean origin who had been adopted into European Caucasian families between 3 and 9 years of age showed a “reversed” other-race effect—that is, they were more accurate at identifying Caucasian faces than Asian faces—indicating that the other-race effect remains plastic until at least 9 years of age.

Additional evidence for the role of experience in shaping the development of face processing comes from studies of atypical development. For example, individuals who are born with cataracts, which block patterned visual input, show rapid improvement in low-level visual abilities such as acuity if the cataracts are removed in the 1st year of life (Lewis & Maurer, 2005). However, even after many years of normal visual input, these individuals show persistent deficits in face perception. Specifically, they have difficulty detecting small changes in the spacing of facial features (i.e., configural processing), which can impair the recognition of a person’s identity (Geldart, Mondloch, Maurer, de Schonen, & Brent, 2002; Le Grand, Mondloch, Maurer, & Brent, 2001, 2003).

Early adverse environments can also affect the development of face perception. For example, children raised in abusive households show altered facial emotion processing, particularly for angry faces, presumably due to their increased exposure to expressions of negative emotion and decreased exposure to expressions of positive emotion. They show a response bias for angry faces, in that they are more likely to match any emotional situation to a picture of an angry face (Pollak, Cicchetti, Hornung, & Reed, 2000). They also overidentify the emotion anger but do not differ from controls in identifying happiness, sadness, and fear (Pollak & Kistler, 2002), and correctly identify facial expressions of anger on the basis of less perceptual information than controls (Pollak & Sinha, 2002). In addition, they attend to angry faces more than controls do and have trouble disengaging from angry faces (Pollak & Tolley-Schell, 2003).

Institutionalization is another example of an adverse early rearing environment that may negatively impact the development of face perception. Institutional care is characterized by psychosocial deprivation; sensory and cognitive stimulation are lacking, and high child-to-caregiver ratios (in some institutions, nearly 20:1) leave children with little social stimulation and almost no opportunity to form stable, emotional attachments to caregivers (Smyke et al., 2007; Zeanah et al., 2003). A wealth of previous research has documented poor physical, cognitive, social, and neurologic outcomes in previously institutionalized children (Fisher, Ames, Chisholm, & Savoie, 1997; Gunnar, 2001; O’Connor, Bredenkamp, & Rutter, 1999; O’Connor & Rutter, 2000; O’Connor, Rutter, Beckett, Keaveney, & Kreppner, 2000), the persistence and severity of which are related to the timing and duration of the institutional experience (Beckett et al., 2006; Rutter et al., 2007).

It is reasonable to suspect that institutional care might adversely affect the development of the neural system underlying face processing specifically. The research reviewed above documents the powerful effects that early experiences can have on the face processing system, and the characteristics of institutional care almost certainly ensure that institutionalized children have qualitatively different experiences with faces compared to family-reared children. Although no studies have directly investigated the amount and kind of exposure to faces children receive in institutions versus family settings, informal observations suggest that children who experience the high child-to-caregiver ratio and high caregiver turnover in institutions likely have less exposure to adult faces in general and less consistent exposure to familiar adult faces but probably more exposure to other children’s faces than family-reared children. Additionally, the broader social context in which faces are experienced clearly differs for institutionalized compared to family-reared children. It seems likely that these qualitative differences in exposure to faces affect the development of the neural system that subserves face processing. Alternatively, it is possible that even the adverse environment of institutionalization provides sufficient perceptual experience with faces to bootstrap the development of the system.

Few studies have directly investigated the effects of institutionalization on the development of face processing, but those that have found differences between family-reared children and children with histories of institutionalization. For example, Parker and Nelson (2005a, 2005b) found that institutionalized children showed cortical hypoarousal, as reflected by dramatically reduced amplitudes of ERP components, compared to noninstitutionalized children in response to pictures of familiar, unfamiliar, and emotional faces. And Wismer Fries and Pollak (2004) found that 4.5-year-old previously institutionalized children showed deficits in identifying primary emotions, such as happiness, sadness, fear, and anger, from pictures of faces. They also showed deficits when matching pictures of emotional faces to emotional scenarios (Wismer Fries & Pollak, 2004).

The general picture that emerges from studies of both typically and atypically developing infants and children is that early experiences are important in determining the developmental course of face processing abilities, although little is known about how these early experiences shape the underlying neural systems responsible for expert face processing. The goal of this study was to investigate further how early experiences shape the neural systems that subserve face perception by studying the neural correlates of face recognition in institutionalized children. This study was part of the Bucharest Early Intervention Project (BEIP), a comprehensive, longitudinal study of the effects of institutionalization on brain and behavioral development (Zeanah et al., 2003). Institutionalized and family-reared children in Bucharest, Romania were assessed on a variety of measures spanning physical, cognitive, social, and neurologic domains of development.

Following a baseline assessment that occurred between 5 and 31 months of age, institutionalized children were randomly assigned to either continued institutional care or placement in high-quality foster care, which was developed and supported by the project itself (see Nelson et al., 2007; Zeanah et al., 2003). Follow-up assessments comparing the institutionalized, foster care, and family-reared children occurred at 18, 30, 42, and 54 months of age. Although the ethical issues raised by this project have been given appropriate consideration elsewhere (Millum & Emanuel, 2007; Zeanah et al., 2006a, 2006b), it is important to address two points regarding children’s placement into caregiving environments during the course of the study. Children who were placed in BEIP foster care were never returned to the institution, and none of the children in this study were restricted in their placement into alternate environments (e.g., reunion with biologic family, placement into state-run foster care, or adoption) that were deemed suitable by the appropriate Romanian authorities.

To explore the development of the neural systems underlying face processing in these children, at the baseline, 30-, and 42-month assessments, ERPs were recorded while children passively viewed pictures of a familiar face and an unfamiliar face. Based on previous research with this sample of children (Parker & Nelson, 2005a, 2005b), we predicted that institutionalized children would show general cortical hypoarousal in that the amplitude of their ERP components would be diminished compared to never-institutionalized children. However, it was difficult to predict whether the experience of institutionalization would affect face processing specifically in this sample of children. On the one hand, institutionalized children likely have atypical experiences with faces (especially their primary caregiver’s face) and show indiscriminate behavior and weak attachments to their caregivers (Zeanah, Smyke, Koga, & Carlson, 2005). Consequently, it is quite possible that they might have difficulty recognizing their caregiver’s face or process it differently than family-reared children, leading to differences between institutionalized and never-institutionalized children in their neural responses to familiar and unfamiliar faces. Alternatively, in light of previous research documenting the effects of even brief experience with faces in infancy (Pascalis et al., 2005; Sangrigoli & de Schonen, 2004; Sugita, 2008), it is also plausible that institutionalized children receive sufficient exposure to faces to ensure the typical development of the face processing system, leading to similar neural processing of familiar and unfamiliar faces in institutionalized and never-institutionalized children. Regarding the effects of the intervention, we predicted that children who were placed in foster care would show improvement compared to the children who remained in the institution and that children who were placed in foster care earlier would show greater recovery than children who were placed in foster care later.

Method

  1. Top of page
  2. Abstract
  3. Method
  4. Results
  5. Discussion
  6. References

Participants

Participants in the BEIP included 208 children between 5 and 31 months of age recruited in Bucharest, Romania. At the baseline assessment, the institutionalized group (IG) consisted of 136 children recruited from six institutions within Bucharest, and the never-institutionalized group (NIG) consisted of 72 children who were recruited through birth records from the maternity hospitals where the institutionalized children were born and matched to the IG on age and sex (see Nelson et al., 2007; Zeanah et al., 2003, for a thorough description of this sample). Eleven of the IG children initially recruited for this study retrospectively met exclusion criteria (i.e., genetic syndromes, overt signs of fetal alcohol syndrome, and/or microcephaly) and were subsequently excluded from all analyses. Following the baseline assessment, the 125 remaining children in the IG were randomly assigned to continued institutional care (IG; n = 62) or foster care (FCG; n = 63). An additional 24 participants (IG, n = 11; FCG, n = 1; NIG, n = 12) were not included in the longitudinal sequence of assessments but were recruited for cross-sectional data collection.

ERP data from 121, 99, and 77 participants at the baseline, 30-, and 42-month assessments, respectively, are reported (see Table 1 for a breakdown of participants by group at each assessment). An additional 60 (33%), 56 (36%), and 45 (37%) participants at the baseline, 30-, and 42-month assessments, respectively, were excluded from data analysis due to technical error, having fewer than 10 artifact-free trials per condition, blinking while the picture was on the screen on 25% or more trials, or excessive eye or body movement artifact. Due to the wide range of ages at the baseline assessment (5–31 months), a median split based on age at baseline assessment was performed and included in the baseline analyses. The median age of the IG at baseline was 23.5 months (41 younger, 40 older), whereas the median age of the NIG at baseline was 21.2 months (19 younger, 21 older). The noninterference policy of the BEIP (Zeanah et al., 2003) ensured that many children assigned to the IG and FCG no longer resided in those environments by 42 months of age. However, original group assignments were preserved during data analysis. This intent-to-treat approach probably underestimates any differences among the groups.

Table 1. Breakdown of Participants by Group at Each Assessment
AssessmentGroup
IG n (no. females)FCG n (no. females)NIG n (no. females)
  1. Note. IG = institutionalized group; FCG = foster care group; NIG = never-institutionalized group.

Baseline81 (39)N/A40 (20)
30 months37 (17)42 (20)20 (11)
42 months23 (10)33 (15)21 (13)

Stimuli and Procedure

At each assessment, ERPs were recorded while children viewed two color photographs: their mother’s or caregiver’s face and a stranger’s face. Photographs showed each face in a frontal view posing a neutral expression and were taken under similar lighting conditions. The preferred caregiver for each child in the IG was determined via surveys of the staff working with the children, and each child was subsequently tested with that caregiver’s face. The caregiver face used for children in the FCG was their foster mother. The stranger face used for each participant was the face of a different mother or caregiver. Children sat on their parent’s or caregiver’s lap in front of a computer screen. Black panels surrounding the computer screen blocked children’s view of the room. There were 69 total trials, and the mother or caregiver and stranger faces were presented with equal probability. Each trial consisted of a baseline period (100 ms), stimulus presentation (500 ms), and a poststimulus recording period (800 ms). The intertrial interval varied randomly between 500 and 1,000 ms. During the experiment, an experimenter observed the child from behind the black screen to direct the child’s attention back to the computer screen when necessary and to eliminate trials during which the child looked away. The experiment continued until the child had seen the maximum number of trials or he/she became too fussy or distracted to continue.

Electrophysiologic Recording and Processing

Continuous EEG was recorded from 13 scalp electrodes (Fz, F3, F4, Cz, C3, C4, Pz, P3, P4, T7, T8, O1, and O2) and left and right mastoid electrode sites using a close-fitting lycra cap with sewn-in tin electrodes (Electro-Cap International, Eaton, OH). Cz was the reference electrode during acquisition. Bipolar vertical electrooculogram (EOG) was recorded from electrodes bisecting the midline placed above and below the left eye to record blinks and other eye movements. After cap placement, the scalp was prepared by inserting an abrasive gel into each electrode site and the scalp under each site was gently abraded. A small amount of electrode gel was inserted into each electrode site. Electrode impedances were considered acceptable if at or below 10 kΩ. Signals were acquired using custom bioelectric amplifiers from SA Instrumentation Company (San Diego, CA) and amplified at a 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 recording. All channels were sampled at 512 Hz and digitized onto the hard drive of a PC using a 12-bit A/D converter (±2.5 V input range) and Snap-Master acquisition software (HEM Data Corporation, Southfield, MI).

A digital lowpass filter of 30 Hz was applied using the ERP Analysis Systems from James Long Company (Caroga Lake, NY). Subsequent processing and analysis of the EEG signal was carried out using the ERP32 analysis software package (New Boundary Technologies, Minneapolis, MN). ERP trials were extracted that consisted of a 100-ms baseline period, 500-ms stimulus presentation, and 800-ms poststimulus recording period. Scalp channels with excessive artifact (i.e., EEG signals > ±200 μV) were rejected, and the entire trial was rejected if more than two scalp channels exceeded this threshold. Data were referenced to an average mastoids configuration. Eye movement-related artifact was corrected (Gratton, Coles, & Donchin, 1983) and individual averages were constructed for each stimulus type (familiar face, unfamiliar face) using 100 ms prior to stimulus onset for baseline correction. A minimum of 10 trials per condition average were required for inclusion in the final sample. A separate grand mean was created for each condition by averaging the individual subject averages together.

Grand means were inspected to identify components of interest. At all three assessments, three occipital components (P1, N170, and P400) and two frontocentral components (P250 and Nc) were analyzed. Time windows that captured the components of interest were identified separately for each assessment, as previous research has shown that there are maturational changes in the latency of various components across age (Batty & Taylor, 2006; Taylor et al., 2004; Webb, Long, & Nelson, 2005). Thus, the time windows for some components differed in duration and/or starting latency across assessments.

Results

  1. Top of page
  2. Abstract
  3. Method
  4. Results
  5. Discussion
  6. References

At each assessment, the peak amplitude and latency of each of the five components were analyzed. For occipital components, 2 (condition: caregiver, stranger) × 2 (hemisphere: left, right) × 3 (group: IG, FCG, NIG) repeated measures omnibus analyses of variance (ANOVAs) were carried out. For frontocentral components, data at frontal and central leads were averaged to create composite frontocentral variables, and 2 (condition: caregiver, stranger) × 3 (laterality: left, central, right) × 3 (group: IG, FCG, NIG) repeated measures ANOVAs were conducted. At the baseline assessment, the between-subjects factor group had only two levels (IG and NIG) as random assignment had not occurred yet, and an additional between-subjects factor, age (younger, older), was included to take into account the wide range of ages of the children at the baseline assessment. When the omnibus ANOVAs revealed significant main effects, post hoc comparisons were carried out using t tests with a Bonferroni correction for multiple comparisons. When the omnibus ANOVAs revealed significant interaction effects, independent samples t tests, paired-sample t tests, or one-way ANOVAs were conducted.

Baseline Assessment

Occipital Components

P1 (90–200 ms).  There was a main effect of group for peak amplitude, F(1, 114) = 12.62, p = .001, ηp2 = .10, indicating that the P1 was larger in the NIG than the IG (Figure 1). There was also a Hemisphere × Age interaction for the peak amplitude, F(1, 114) = 8.41, p = .004, ηp2 = .07. Follow-up t tests revealed that older children showed larger amplitudes than younger children over the right hemisphere only, t(116) = −2.62, p = .01. Latency analyses revealed a Group × Age interaction, F(1, 114) = 4.12, p = .045, ηp2 = .04. Follow-up t tests revealed that older NIG children showed significantly faster latency to the P1 than older IG children, 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).

image

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 (μV). Note. ERP = event-related potential; IG = institutionalized group; NIG = never-institutionalized group.

Download figure to PowerPoint

image

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.

Download figure to PowerPoint

N170 (150–300 ms).  For peak amplitude, there was a main effect of age, F(1, 114) = 5.43, p = .022, ηp2 = .04, in that younger children showed a significantly larger N170 than older children. There were no main effects or interactions for the N170 latency.

P400 (250–500 ms).  There was a significant Condition × Lead × Age × Group interaction for peak amplitude, F(1, 114) = 4.32, p = .04. Due to the complexities involved in interpreting four-way interactions, no follow-up tests were conducted. For the P400 latency, there was a main effect of age, F(1, 114) = 3.96, p = .049, ηp2 = .03, which was qualified by a significant Condition × Age × Group interaction, F(1, 114) = 5.55, p = .02, ηp2 = .05 (Figure 3). Follow-up analyses suggested that there were no significant differences for P400 latency in the IG as a function of condition or age. However, in the NIG, younger children showed faster P400 latency in response to caregiver’s face compared to stranger’s face, t(17) = −3.30, p = .004, whereas older children did not, t(19) = 0.61, p > .10. There was also a significant Condition × Hemisphere × Group interaction, F(1, 114) = 4.71, p = .032, ηp2 = .04; however, there were no significant differences upon follow-up.

image

Figure 3.  Condition × Age × Group interaction for P400 latency at the baseline assessment. Error bars represent standard error of the mean.

Download figure to PowerPoint

Frontocentral Components

P250 (175–375 ms).  For peak amplitude, there was a main effect of laterality, F(2, 232) = 5.76, p = .005, ηp2 = .05. Post hoc comparisons revealed that the amplitude of the P250 was larger over midline than left leads. There were no significant main effects or interactions for P250 latency.

Nc (350–650 ms).  There was a main effect of laterality for peak amplitude, F(2, 232) = 5.60, p = .004, ηp2 = .05, indicating that Nc amplitude was larger over midline than left leads. There were no main effects or interactions for Nc latency.

Summary

Never-institutionalized children showed significantly larger amplitudes for the P1 than institutionalized children, and among the older children, never-institutionalized children showed significantly faster P1 latencies than institutionalized children. Additionally, only the never-institutionalized children processed the caregiver and stranger faces differently; specifically, they showed faster P400 latency for the caregiver face versus the stranger face, whereas the institutionalized children did not. Surprisingly, there were no group or condition effects over frontocentral electrodes.

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, ηp2 = .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, ηp2 = .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 main effect of hemisphere, F(1, 90) = 6.58, p = .012, ηp2 = .07, in that the latency to the P1 was faster over the right than the left hemisphere. There was also a significant Condition × Group interaction, F(2, 90) = 5.26, p = .007, ηp2 = .10. Follow-up t tests revealed that the P1 latency was faster in response to the caregiver face than the stranger face in the IG only, t(32) = −2.54, p = .016 (Figure 5). Neither the NIG nor the FCG showed different P1 latencies for caregiver versus stranger (all ps > .10).

image

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 (μV). Note. ERP = event-related potential; IG = institutionalized group; FCG = foster care group; NIG = never-institutionalized group.

Download figure to PowerPoint

image

Figure 5.  Condition × Group interaction for P1 latency at the 30-month assessment. Error bars represent standard error of the mean. Note. IG = institutionalized group; FCG = foster care group; NIG = never-institutionalized group.

Download figure to PowerPoint

N170 (175–325 ms).  There was a main effect of hemisphere for the peak amplitude, F(1, 90) = 4.87, p = .03, ηp2 = .05, that was qualified by a significant Condition × Hemisphere interaction, F(1, 90) = 4.09, p = .046, ηp2 = .04, but no Condition × Group interaction, p > .10. Follow-up analyses revealed that N170 amplitude was larger over the left than the right hemisphere for the caregiver condition only, t(92) = −3.12, p = .002. There were no hemispheric differences for the stranger condition, t(92) = −1.14, p > .10. For the N170 latency, there was a significant Hemisphere × Group interaction, F(2, 90) = 3.21, p = .045, ηp2 = .07. Follow-up analyses revealed that the NIG had significantly faster latencies than the IG over the right hemisphere only, F(2, 94) = 4.51, p = .013 (Figure 6). The FCG was not significantly different from either the NIG or IG in terms of N170 latency.

image

Figure 6.  Hemisphere × Group interaction for N170 latency at the 30-month assessment. Error bars represent standard error of the mean. Note. IG = institutionalized group; FCG = foster care group; NIG = never-institutionalized group.

Download figure to PowerPoint

P400 (250–500 ms).  For peak amplitude, there was a main effect of hemisphere, F(1, 90) = 4.77, p = .032, ηp2 = .05, indicating that the P400 was larger over the right than left hemisphere. There were no main effects or interactions for P400 latency.

Frontocentral Components

P250 (185–350 ms).  There was a main effect of laterality for peak amplitude, F(2, 186) = 9.40, p < .001, ηp2 = .09, indicating that the P250 was larger over midline than left or right leads. For P250 latency, there was a main effect of condition, F(1, 93) = 11.68, p = .001, ηp2 = .11, but no Condition × Group interaction (p > .10) indicating that the response to the stranger’s face was significantly faster than the response to the caregiver’s face in all three groups (Figure 7). There was also a main effect of laterality, F(2, 186) = 3.36, p = .039, ηp2 = .04. Post hoc comparisons revealed that the latency to the P250 was faster over midline that left leads.

image

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 (μV). Note. ERP = event-related potential; IG = institutionalized group; FCG = foster care group; NIG = never-institutionalized group.

Download figure to PowerPoint

Nc (350–550 ms).  There was a main effect of condition for peak amplitude, F(1, 93) = 7.31, p = .008, ηp2 = .07 and latency, F(1, 93) = 9.62, p = .003, ηp2 = .09, but no Condition × Group interactions (all ps > .10) indicating that the response to stranger’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, ηp2 = .11, indicating that the Nc, like the P250, was larger over midline than left or right leads.

Summary

Similar to the baseline assessment, the never-institutionalized children showed significantly larger amplitudes for the P1 than institutionalized children at the 30-month assessment. Children who had been placed in foster care showed some recovery, in that their P1 amplitudes were intermediate between the institutionalized and never-institutionalized children. There was also a difference among groups for N170 latency; never-institutionalized children showed faster N170 latencies than institutionalized children over the right hemisphere only. In terms of condition differences, only the institutionalized children showed differential processing of caregiver and stranger faces over occipital electrodes (P1 latency was faster for caregiver vs. stranger). Surprisingly, however, the three groups of children showed identical condition differences for the frontocentral components P250 and Nc. The Nc was larger in amplitude, and both the P250 and the Nc were faster in latency, in response to stranger compared to caregiver in all three groups of children.

42-Month Assessment

Occipital Components

P1 (80–210 ms).  There was a main effect of group for the peak amplitude, F(2, 72) = 3.47, p = .037, ηp2 = .09. Post hoc comparisons revealed that the NIG showed significantly larger P1 amplitudes than the IG, whereas the FCG did not differ from either group (Figure 8). This was qualified by a significant Condition × Group interaction, F(2, 72) = 3.40, p = .039, ηp2 = .09. Follow-up tests revealed that the NIG showed larger P1 amplitudes than the IG for the stranger condition, F(2, 72) = 4.91, p = .01, but not the caregiver condition, F(2, 72) = 1.53, p > .10 (Figure 9). Additionally, P1 amplitude was larger over the right than the left hemisphere, F(1, 72) = 6.54, p = .013, ηp2 = .08. There was also a significant Condition × Group interaction for P1 latency, F(2, 72) = 3.62, p = .032, ηp2 = .09; however, there were no significant differences upon follow-up.

image

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 (μV). Note. ERP = event-related potential; IG = institutionalized group; FCG = foster care group; NIG = never-institutionalized group.

Download figure to PowerPoint

image

Figure 9.  Condition × Group interaction for P1 amplitude at the 42-month assessment. Error bars represent standard error of the mean. Note. IG = institutionalized group; FCG = foster care group; NIG = never-institutionalized group.

Download figure to PowerPoint

N170 (160–300 ms).  For peak amplitude, there was a main effect of condition, F(1, 72) = 3.97, p = .05, ηp2 = .05, in that the response to the stranger’s face was significantly larger overall than the response to the caregiver’s face. However, this finding was qualified by a significant Condition × Group interaction, F(2, 72) = 3.29, p = .043, ηp2 = .08, which revealed that the IG showed a significantly larger N170 for the stranger face compared to the caregiver face, F(1, 22) = 10.23, p = .004 (Figure 10), whereas the FCG and NIG did not (all ps > .10). There were no main effects or interactions for N170 latency.

image

Figure 10.  Condition × Group interaction for N170 amplitude at the 42-month assessment. Error bars represent standard error of the mean. Note. IG = institutionalized group; FCG = foster care group; NIG = never-institutionalized group.

Download figure to PowerPoint

P400 (250–500 ms).  For peak amplitude, there was a main effect of condition, F(1, 69) = 4.64, p = .035, ηp2 = .06, but no Condition × Group interaction (p > .10), indicating that the response to the caregiver’s face was significantly larger than the response to the stranger’s face in all three groups. There were no main effects or interactions for the P400 latency.

Frontocentral Components

P250 (185–350 ms).  For peak amplitude, there was a main effect of laterality, F(2, 138) = 9.69, p < .001, ηp2 = .12. Post hoc comparisons indicated that the P250 was larger over midline (M = 10.35, SD = 4.88) than left (M = 9.33, SD = 4.56) or right (M = 9.23, SD = 4.38) leads. There were no significant main effects or interactions for P250 latency.

Nc (350–550 ms).  There were main effects of condition for both peak amplitude, F(1, 69) = 6.49, p = .013, ηp2 = .09 and latency, F(1, 69) = 14.38, p < .001, ηp2 = .17, but no Condition × Group interactions (all ps > .10), indicating that the response to stranger’s face was both larger and faster than the response to caregiver’s face in all three groups (Figure 11). Additionally, there was a main effect of laterality for latency, F(2, 138) = 3.37, p = .04, ηp2 = .05, revealing that the latency to the Nc was faster over right than midline leads.

image

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 (μV). Note. ERP = event-related potential; IG = institutionalized group; FCG = foster care group; NIG = never-institutionalized group.

Download figure to PowerPoint

Summary

Similar to both the baseline and 30-month assessments, never-institutionalized children showed larger amplitudes for the P1 than institutionalized children at the 42-month assessment. Foster care children showed some recovery, in that their P1 amplitudes were intermediate between the other two groups. Institutionalized children, but not never-institutionalized or foster care children, showed a condition difference for the N170 component; specifically, the N170 was larger in response to the stranger face compared to the caregiver face. Surprisingly, all three groups showed identical condition differences for the P400 over occipital electrodes and the Nc over frontocentral electrodes. The P400 was larger in response to the caregiver face compared to the stranger face; conversely, the Nc was both larger and faster in response to the stranger face compared to the caregiver face.

Analysis of Intervention Effects

The unique design of the BEIP provided an opportunity to examine the effects of the timing and duration of the foster care intervention on ERP outcomes. Because children were placed in foster care at varying ages depending on the age at which they were recruited into the study, which occurred at anytime between 5 and 31 months of age, we were able to compare the outcomes of children who 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 occipital components (P1, N170, and P400) in the IG and FCG at 42 months of age (the NIG was used for comparison purposes only). Perhaps surprisingly, independent sample t tests revealed no significant differences between the IG and FCG at 42 months of age. However, the wide range of ages at which children in the FCG were placed in foster care could have obscured any effects of the foster care intervention. It is possible that children placed in foster care earlier showed greater improvements in their ERP outcomes than children placed in foster care later.

To evaluate whether the timing of placement into foster care, and correspondingly its duration, affected ERP outcomes at 42 months of age, two analysis strategies were used. First, the FCG was divided into an early-placed group and a late-placed group based on a median split at 25.5 months of age, and one-way ANOVAs comparing the IG, early-placed FCG, and late-placed FCG were conducted for the amplitudes and latencies of the occipital components at 42 months of age (see Table 2). Second, Pearson correlations between age at placement and the amplitudes and latencies of the occipital components were conducted for the FCG only. One-way ANOVAs revealed no significant differences among the early-placed FCG, late-placed FCG, and IG for any of the occipital components. Correlational analyses revealed only one significant correlation, between age at placement and N170 amplitude, r = −.39, p = .03. As the N170 is a negative-going component, the negative correlation here indicates that children who were older when placed in foster care tended to have larger (i.e., more negative) amplitudes for the N170.

Table 2. Descriptive Statistics for Occipital Components for the IG, Early- and Late-Placed FCG, and NIG at 42-Month Assessment
ComponentMeasureGroup
IGEarly-placed FCGLate-placed FCGNIG
  1. Note. Amplitude is reported in microvolts (μV) and latency is reported in milliseconds (ms). Values are given as M (SD). IG = institutionalized group; FCG = foster care group; NIG = never-institutionalized group.

P1Amplitude16.53 (7.37)19.41 (7.80)17.19 (6.81)22.69 (9.91)
Latency134.67 (24.13)136.81 (17.55)127.78 (16.39)128.74 (22.40)
N170Amplitude−3.67 (5.37)−3.27 (3.33)−5.97 (3.75)−3.57 (8.10)
Latency231.62 (34.89)232.52 (25.52)221.03 (22.55)231.72 (28.74)
P400Amplitude13.07 (4.41)12.80 (4.57)12.83 (2.54)15.36 (6.23)
Latency370.22 (44.53)378.30 (43.64)363.27 (41.52)354.88 (57.48)

Discussion

  1. Top of page
  2. Abstract
  3. Method
  4. Results
  5. Discussion
  6. References

This study used ERPs to examine the neural correlates of face recognition in three groups of children: currently institutionalized, previously institutionalized (subsequently placed in foster care), and never institutionalized. We hypothesized that the psychosocial deprivation associated with institutional care would lead to abnormalities in the neural response to faces among institutionalized children, and that foster care would be at least somewhat effective in remediating any deficits in previously institutionalized children. The results of this study not only supported these hypotheses but also revealed some unexpected findings.

The first major finding was that institutionalized children showed strikingly smaller amplitudes for the P1 component than never-institutionalized children across all three assessments. Although significant group differences were not found consistently for the other occipital components, inspection of the occipital ERP waveforms (Figures 1, 4, and 8) suggests that this pattern of decreased amplitude in the institutionalized children continued across the entire ERP epoch. Institutionalized children in this sample also showed decreased amplitudes for the P1 and P400 occipital components at all three assessments during a facial emotion processing task, in which ERPs were recorded while they watched pictures of emotional faces (Moulson, Fox, Zeanah, & Nelson, 2009). These findings suggest that institutionalized children show cortical hypoarousal while perceiving faces. Because ERPs were not recorded in response to stimuli other than faces, it is unclear whether this cortical hypoarousal extends to the processing of other stimuli. EEG and neuroimaging findings from children with histories of institutionalization do suggest that this cortical hypoarousal may be pervasive, rather than specific to children’s responses to faces. However, it is also possible that certain neural systems, including those that underlie face perception, are particularly vulnerable to the early adverse experience of institutionalization.

Marshall and Fox (2004) found that institutionalized children in the BEIP showed increased low frequency (theta) EEG activity and decreased high-frequency (alpha and beta) EEG activity compared to never-institutionalized children at the baseline assessment. A study with previously institutionalized children found that 18-month-old toddlers who had been adopted following early institutional care also had increased low-frequency EEG power and decreased high-frequency EEG power compared to children raised with their biological families (Tarullo, 2008), which replicated the findings in the BEIP sample (Marshall & Fox, 2004). Chugani et al. (2001), using positron emission tomography, found reduced metabolic activity in limbic (i.e., medial temporal lobe) and paralimbic (i.e., areas of prefrontal cortex) regions of the brain in a group of previously institutionalized children compared to a group of family-reared children. The previously institutionalized children had spent an average of 5.5 years in their adoptive homes, suggesting that the brain dysfunction associated with early institutional care persists long after placement into a typical environment.

Children who live in noninstitutional high-risk environments in North America also show signs of cortical hypoarousal. In a study of Mexican children, Otero and colleagues (Otero, 1997; Otero, Pliego-Rivero, Fernandez, & Ricardo, 2003) showed that 4- to 6-year-old children living in families of low socioeconomic status (SES) displayed increased low-frequency power and decreased high-frequency power compared to children living in low-risk families. And in a study of 7- to 12-year-old children in California, Kishiyama, Jimenez, Perry, Boyce, and Knight (2007) found that low-SES children showed smaller amplitudes for ERP components at central and occipital leads compared to high-SES children in an oddball paradigm. Thus, the results of this study bolster the existing evidence that children who have experienced early deprivation associated with institutional or otherwise inadequate care have chronically under-aroused brains. The underlying neural cause of this deficit remains unknown. One might speculate, however, that one source of this hypoactivation may be an error in pruning, specifically, that the lack of normative experience may lead to over-pruning of neurons or synapses, both of which would lead to reductions in brain activity and a corresponding reduction in head circumference (Johnson, 2006; Smyke et al., 2007).

The second major finding of this study was that children who were placed in foster care following early institutionalization showed some evidence of improvement. The amplitude of the P1 component among the foster care children fell between the P1 amplitudes of the institutionalized and never-institutionalized children at the 30- and 42-month assessments. Although the foster care group was not significantly different from either group, this result suggests that children placed in foster care do, to some extent, recover from the psychosocial deprivation that they experienced early in life. Surprisingly, the timing of placement into foster care was unrelated to the children’s ERP outcomes at 42 months of age, a result that is consistent with the results of the facial emotion ERP study conducted with this sample of children (Moulson et al., 2009). It seems that even relatively late placement in foster care was sufficient to initiate improvements in neurologic functioning.

In contrast, many previous studies with postinstitutionalized children have demonstrated that earlier adoption leads to more favorable behavioral, cognitive, and social outcomes than later adoption (Beckett et al., 2006; O’Connor & Rutter, 2000; O’Connor et al., 2000; Rutter, 1998; Rutter et al., 2007). That there was no relation between the timing of intervention and ERP outcomes in this sample of children was particularly unexpected because previous findings from the BEIP also indicate that earlier intervention leads to better outcomes. Children who were placed in foster care before 2 years of age had significantly better cognitive outcomes at 42 and 54 months than children placed after 2 years of age (Nelson et al., 2007). Children who were placed before 2 years of age also showed more mature patterns of EEG activity (i.e., higher EEG alpha power and lower EEG short-distance coherence) than children who remained in the institution, whereas children placed after 2 years of age did not (Marshall, Reeb, Fox, Nelson, & Zeanah, 2008).

In light of this previous research, it is difficult to interpret the current finding and its implications for our understanding of sensitive periods in the development of face processing. Although there are several plausible interpretations, distinguishing among them is impossible because we cannot determine based on the current data whether children placed in foster care will ever recover fully (i.e., show ERP components that do not differ in amplitude, latency, or morphology from those of children who have never experienced institutionalization). Additionally, the lack of a nonface control stimulus in this study makes it difficult to determine whether any conclusions drawn about sensitive periods based on these findings are specific to the domain of face perception or relevant for our understanding of brain development more generally.

If children placed in foster care do eventually catch up to never-institutionalized children, it would suggest either that there is no sensitive period, or that the sensitive period remains open beyond the age at which these children were placed in foster care (i.e., beyond 30 months). The former possibility is unlikely given the findings that early visual deprivation results in long-term deficits in face perception (Geldart et al., 2002; Le Grand et al., 2001, 2003); however, the latter possibility is consistent with previous research on the reversibility of the other-race effect that suggests that the face processing system remains plastic at least into middle childhood (Sangrigoli et al., 2005). However, another possible interpretation of this finding is that children’s initial improvement is unrelated to their age at intervention, but further improvement depends on having been placed in foster care early. If this were the case, we would predict that both earlier and later placed children would improve initially, but only earlier placed children would continue on an accelerated developmental trajectory and eventually catch up to family-reared children. This pattern of results would suggest that there is a sensitive period that narrows early, but the effects of that sensitive period are not realized until later in development (see Maurer, Mondloch, & Lewis, 2007, for a discussion of such “sleeper effects” following visual deprivation). Of course, it is also possible that none of the children who experienced early deprivation followed by intervention will show a full recovery, which would suggest early closure of the sensitive period for the neural system underlying face perception. Again, the lack of data beyond 42 months of age in this study makes it impossible to distinguish among these possibilities or draw firm conclusions regarding the existence and timing of sensitive periods in this developmental domain.

The third major finding of this study was that all three groups of children showed differing neural responses to the caregiver and stranger faces. Surprisingly, the results also suggest that all three groups of children processed the faces similarly, despite the disparity in the quality of their early caregiving environments. At the 30- and 42-month assessments, all three groups showed a larger amplitude and shorter latency response to the stranger face compared to the caregiver face for the Nc over frontocentral electrodes, a finding that is consistent with previous research. Carver et al. (2003) documented a developmental progression in the response to mother and stranger faces in typically developing infants, which they attributed to the changing significance of mother’s face and strangers’ faces over the first years of life. They found that younger infants (< 24 months of age) showed a larger Nc response to mother’s face than stranger’s face, whereas older infants (> 45 months of age) showed a larger Nc response to stranger’s face than mother’s face. Infants between 24 and 45 months of age showed no differences in their Nc responses to mother and stranger faces, suggesting that it is a time of transition in terms of the social significance of familiar and unfamiliar faces.

The children in this study showed no differences in the amplitude or latency of the Nc or P250 at the baseline assessment, which suggests that either (a) the baseline assessment captured the transition stage reported by Carver et al. (2003), or (b) the range of ages at the baseline assessment (5–31 months) obscured any developmental progression in processing familiar and unfamiliar faces (although the children were split into younger and older groups for the baseline analyses, the children within these groups still ranged widely in age). By 30 and 42 months of age, however, children in this study consistently showed larger amplitude and shorter latency frontocentral components in response to stranger. Although this is somewhat earlier than Carver et al. reported, the overall pattern of results suggests that there is a developmental progression among all three groups of children in this study in terms of their neural response to caregiver and stranger faces.

There were also condition effects for occipital components, and some evidence of differential processing of caregiver and stranger faces among the groups. At the baseline assessment, the family-reared children showed a faster latency to the P400 for the caregiver face compared to the stranger face, whereas the institutionalized children did not. At the 30-month assessment, the institutionalized children showed a faster latency to the P1 for the caregiver face compared to the stranger face, whereas the family-reared and foster care children did not. And at the 42-month assessment, only the institutionalized children showed a larger N170 for the stranger face compared to the caregiver face. The lack of consistency across assessments makes it difficult to interpret the extent of occipital processing differences among the three groups of children, but it is noteworthy that there was evidence of differential processing of the familiar and unfamiliar faces over occipital leads. In contrast, the facial emotion ERP task conducted with these children revealed no differential processing of emotion over occipital leads, although there was differential processing of emotion over frontocentral leads (see Moulson et al., 2009). This suggests that the neural correlates of processing facial identity and facial emotion are at least somewhat separable in young children, which is consistent with the adult neuroimaging literature (Haxby et al., 2002) and findings from congenital prosopagnosia (Humphreys, Avidan, & Behrmann, 2007).

It is surprising that the neural processing of familiar and unfamiliar faces is remarkably similar in institutionalized and family-reared children. Based on the adverse early rearing environment that institutionalized children suffer, we expected that they might show no differences at all in the neural processing of familiar and unfamiliar faces. The fact that institutionalized children show processing differences that are similar to family-reared children suggests that the atypical experience that they receive over the first years of life is sufficient to set up the neural architecture underlying at least the perceptual discrimination of faces. It is possible, however, that these children may show deficits on more subtle face processing tasks later in life, such as processing faces configurally (a hallmark of expert face processing) or identifying complex emotions expressed in faces.

In summary, the psychosocial deprivation experienced by institutionalized children leads to pervasive cortical hypoarousal compared to family-reared children. Despite this cortical hypoarousal, institutionalized children processed familiar and unfamiliar faces similarly to family-reared children. Placement in high-quality foster care led to a partial amelioration of the cortical hypoarousal in previously institutionalized children. These findings have implications both for our understanding of the development of the neural systems that subserve face processing and, more importantly, for the millions of children around the world who are reared in environments characterized by psychosocial adversity and deprivation.

References

  1. Top of page
  2. Abstract
  3. Method
  4. Results
  5. Discussion
  6. References