Introduction

Young maltreated children, birth to three years, represent the age group with the highest rates of maltreatment in the United States (USDHHS/ACYF 2007). Young maltreated children in the foster care system are less likely to receive mental health services when compared to older maltreated children in the foster care system (Stahmer et al. 2005). Others have advocated for more comprehensive assessment of the developmental needs of young children in foster care (American Academy of Pediatrics 2002; Stahmer et al. 2005). We suggest that these assessments should evaluate the goodness of fit between a child’s emotion regulatory capacities and the parent’s emotional resources; however, there are few available data on early childhood maltreatment and its effects on dyadic regulatory processes and psychopathology to inform necessary intervention.

Many researchers have described emotion regulation as one of the key stage salient issues early in life that a child must resolve for successful socioemotional development (Cicchetti and Toth 1995; Grolnick and Farkas 2002). Emotion regulation is generally defined as the internal and external processes by which the individual manages the occurrence, intensity, and expression of emotions to reach goals or situational demands (Cicchetti and Howes 1991; Eisenberg and Morris 2002; Thompson 1994).

Effortful control has been found to be a measurable key component of early emotion regulation (Eisenberg and Morris 2002). It is the self-regulatory mechanism linked with attention that reflects the ability to suppress a dominant response in order to perform a less salient response (Eisenberg and Spinrad 2004; Rothbart and Bates 1998). Thus, well-regulated children may demonstrate attentional effortful control through persistence in a difficult task, delaying gratification for a prize or treat, turn-taking, and attending to a subdominant stimulus (Kochanska et al. 2000).

Negative reactivity, in contrast to effortful control, represents a child’s biological tendency to react to stressors with high degrees of emotional lability, including anger, irritability, fear, or sadness (Rothbart and Bates 1998). Reactivity refers to the responsivity of the child’s emotional, activation, and arousal systems (Derryberry and Rothbart 1997), rather than a child’s effortful regulation of emotion. Negative reactivity has been measured in previous research as both the intensity of a negative emotion (Hill et al. 2006) and as negative emotional lability (Shields and Cicchetti 1997). Thus, effortful control represents a child’s cognitive control over emotional arousability and reactivity (Buss and Goldsmith 1998; Crockenberg and Leerkes 2006). However, emotion regulation not only involves the management of negative emotions, but also involves the modulation of positive emotions. In most cases, individuals work to diminish negative emotional states and to encourage more positive ones. It has been argued that the measurement of emotion regulation must include evidence of the activation of emotion along with the strategies used to manage the emotional expression (Cole et al. 2004). Thus, the definition of emotion regulation presented here offers a two factor model: the emotional dynamics (emotional intensity) and the control or management of the emotion (cognitive and behavioral strategies).

Emotion regulation is highly modulated by caregiver characteristics and develops from the interaction between the child’s biological maturity and parental responsivity and sensitivity (Eisenberg and Morris 2002). Children’s emotion regulation skills develop within the context of that parent–child relationship, and young children learn to appropriately regulate their affect partially based on their caregiver’s affect and cues, with the aid of a nurturing parent (Campos et al. 1989; Eisenberg and Morris 2002; Fox and Calkins 2003; Morris et al. 2007; Tronick 1989). Individual differences in emotion regulation emerge from the socialization of appropriate emotion-related behaviors by parents through modeling, reinforcement, and discipline (Calkins 1994). Conversely, maternal negative and controlling behavior is related to more emotional arousal, reactivity, and poor physiological regulation in children (Calkins et al. 1998). Furthermore, appropriate family emotion socialization and discipline are associated with more effective strategies to regulate negativity (Garner and Spears 2000) and higher effortful control (Karreman et al. 2008) in preschoolers.

Maltreatment early in development is such an aberration in the adequate caregiving environment that emotion regulation deficits can occur for a variety of reasons such as lack of modeling and support, absence of positive affect, harsh discipline and negative control, inconsistency, and lack of sensitivity. The child within the abusive dyad is expected to manage his or her emotions while there is dyssynchrony of emotional goals between parent and child, such that maltreating caregivers are not able to help their child reduce distress and manage feelings of anger and sadness. Instead, the maltreating parent may indiscriminately ridicule, belittle or neglect their child’s emotional states (Thompson 2001).

To date, there are only a few studies examining maternal socialization of emotion within maltreating dyads. Shipman et al. (2005, 2007) have found poor maternal socialization of emotion in maltreated children to be a key link between maltreated children’s emotion regulation skills and psychopathology. In addition, Egeland et al. (2002) found that children physically abused in early childhood experienced more externalizing problems in elementary school; however, the relationship was mediated through the alienation from a caregiver and emotional dysregulation. It is important to note, that Shipman et al. (2007) found maltreating mothers were more likely to be single parents therefore the emotion socialization deficits may be related in part to the absence of a father figure. Shipman et al. (2007) suggest that elements within the maltreating family context such as inter-adult conflict and modeling of ineffective strategies to resolve conflict may be related to maltreated children’s emotion dysregulation. In short, the maltreating parents’ hostile emotional climate, inconsistent emotional demands or emotional neglect may make it difficult for children to successfully develop emotion regulation skills (Cicchetti and Rogosch 2001; Maughan and Cicchetti 2002; Morris et al. 2007).

It is this failure to develop emotion regulation skills that we suggest is related to higher psychopathology among maltreated children. Maltreated children demonstrate less understanding of negative emotions (Shipman et al. 2005), exhibit decreased emotional expression and flexibility (Gaensbauer 1982), use fewer internal state words (Cicchetti and Begley 1987), exhibit more negative affect (Gaensbauer 1982), display more sensitivity to aggressive stimuli (Pollak et al. 2005) and are more emotionally dysregulated (Maughan and Cicchetti 2002; Shipman et al. 2007) compared to non-maltreated children. Behaviorally, maltreated children exhibit less self control and social competence (Fantuzzo et al. 1998; Shields et al. 1994) and more emotional lability, reactivity, and anger (Alessandri 1991) among peers.

In sum, maltreated children’s emotion regulation difficulties have been shown to be related to psychopathology and this relationship may be related to poor parental socialization of emotion, inconsistent emotional interactions, and dyssychrony of emotional goals within the dyad. Particularly important to the long term adjustment of maltreated children is the goodness of fit within their caregiving environment (Zeanah et al. 1996). Therefore understanding the complex “interactive dance” between the child’s emotion regulatory capacities and the parent’s parenting resources is crucial to understanding risk and resiliency among maltreated children.

Despite the harsh treatment maltreated children experience, some do not exhibit difficulties with the development of emotion regulation. Resiliency has been defined as the ability for successful adaptation, competence, or positive functioning in the face of adversity (Masten and Wright 1998). As described earlier, children’s emotion regulation skills play a vital role in their ability to attain successful social interactions and psychological adjustment. Therefore positive emotionality and the ability to successfully regulate or control one’s emotions may be individual level sources of resiliency related to maltreatment and psychopathology (Chang et al. 2003; Cicchetti et al. 1993; Shields et al. 1994). Related to this are temperamental and cognitive factors such as physiological control (Perry et al. 1995) and cognitive ability (Himelein and McElrath 1996) also associated with increased competence in maltreated children.

Because many maltreated children who have been placed in the foster care system are eventually returned to their biological parents, the identification of dyadic factors that protect these children from psychopathology is crucial to interventions with maltreating families. Improved functioning of maltreated children has been associated with improvement in parent–child interactions, family stress levels, maternal functioning, parental support and the emotional quality of the parenting (Crittenden 1996; Egeland et al. 1981). Specifically, Crittenden (1996) found that mothers who were able to improve their dyadic interactions resulted in major developmental skill increases among their children. In addition, secure attachment representations of a caregiver may mitigate some effects of maltreatment (Shields et al. 2001; Toth and Cicchetti 1996a, b). Maternal empathic responding is also related to adaptive emotion regulation (Shipman and Zeman 2001) and lower internalizing symptomatology among school aged maltreated children (Shipman et al. 2005). Taken together, the research suggests that there are individual and parent–child relational factors that may be related to both emotion regulation skill development in maltreated children and the subsequent resilience to the sequelae of maltreatment. The current study will build upon this literature by exploring the specific dyadic factors related to emotion regulation resiliency among young maltreated children and discuss how these data may inform interventions.

Reciprocity within the parent–child relationship and goodness of fit between parent and child emotional capacities is predictive of adjustment in early childhood (Zeanah et al. 1996). The current study expands extant research by specifically investigating the relationship between emotion regulation and parental affect as it relates to individual differences in psychopathology resulting from maltreatment. The current study utilizes observational measures of emotion regulation and parenting during a standardized parent–child interaction procedure. This type of dyadic assessment is crucial to advancing our understanding of the strengths within these relationships and informing parent–child relational interventions for young maltreated children. The current study proposed four main hypotheses: (1) maltreated young children would display more emotion dysregulation compared to non-maltreated children; (2) positive parental affect would be related to better emotion regulation whereas negative parental affect would be related to lower emotion regulation; (3) emotion dysregulation would be related to psychopathology; (4) maltreatment status would moderate the relations between emotion regulation and psychopathology and parenting and psychopathology, such that the effects of emotion dysregulation and negative parenting would be more harmful among maltreated children. Additionally, the role of maltreatment type, multiplicity of maltreatment, and length of time in foster care at assessment were explored to determine if these variables differentially affected emotional development and overall adjustment in this population.

Method

Participants

Participants included 123 children from ages 12–47 months (Mean age = 32.62 months). All the children were from the parish (county) surrounding a large, urban southern city. The children were 69% African American and 31% non African American (Caucasian, Hispanic, Biracial or Other), 25% higher than the state statistics on percentage of African Americans in the foster care system, but are similar to the demographics of the urban area near where they were recruited. The sample was 47% female and 53% male. Given the limits of sample size and concerns with power by adding a covariate, only mothers were included in the current study. Additional demographic information for all participants included in this study can be found in Table 1. Institutional review board approval was obtained prior to data collection.

Table 1 Demographic information for all participants by group: means, standard deviations, and percentages
Full size table

Data for the 66 maltreated children were collected by an infant mental health team (see Zeanah and Larrieu 1998). These children were referred to the team because they were between the ages of birth and 59 months and had been taken into the custody of child protective services (CPS) for validated abuse or neglect. Consent for participation in the infant team assessments was obtained at the time of the initial clinic visit by the biological parent. Participation in infant team assessments was required as part of the parent’s CPS case plan; however, a separate consent was obtained to use the data for research and training purposes. Parents were informed that the choice not to allow their data to be used in research would not affect their CPS case plan, although the parents were made aware that, either way, findings from these assessments and the subsequent treatment course would be reported to the court system. All mother–child dyads who agreed to the use of their infant team assessments for research purposes were included in the current study. In cases where the maltreating parent had multiple children, only the oldest child who met age criteria was used to avoid violating independence assumptions for analyses.

The maltreated population was coded for type of maltreatment experienced using Barnett et al. (1993) maltreatment subtype definitions (further described in the procedures section). Overlap in the type of maltreatment experienced by participants was consistent with the literature (Howes et al. 2000; Manly et al. 1994). Physical abuse was experienced by 25.7% of the participants, sexual abuse by 2.8%, Neglect/Failure to Provide by 62.3%, Neglect/Lack of Supervision by 87.0% (the largest category), Emotional Maltreatment by 37.1%, Moral/Legal/Educational (involving children in illegal or immoral activities or activities counter to their educational needs) by 4.2%, and Dependency/Abandonment by 16.4%, (added for the current study; see Robinson et al. 2004). The average length of time in foster care at the assessment was 6.87 months.

The 57 non-maltreated children that comprise the comparison group were selected based on age and socio-economic status from two larger studies conducted in the same community. The majority of the non-maltreated children were recruited from Head Start centers using flyers and word of mouth snow ball recruitment techniques. Inclusion criteria were: (a) child age between 1 and 5 years and (b) no history of a traumatic event (this included substantiated or unsubstantiated maltreatment reports and screening for child abuse potential).

Informed consent was given by the caregiver at the time of the laboratory visit and participants were paid $100 for their time. The non-maltreated children were compared to the maltreated children using Analysis of Variance (ANOVA) based on child age, child gender, child ethnicity, maternal age, and maternal education, see Table 1. The maltreated sample and the non-maltreated sample were similar on all demographic variables except maternal education, with the non-maltreated group higher on education. Group analyses revealed maternal education was not significantly correlated to any of the study variables and therefore not controlled for in the analyses.

Procedures

As mentioned previously, the maltreated children were participants in an infant mental health assessment program; however, all data presented here were collected prior to treatment services. A CPS referral was received with the infant team and the family entered the process approximately six weeks after the child had been placed in foster care. The biological parents participated in the parent–child interaction procedure and during a separate clinic visit the foster parent completed the Child Behavior Checklist (CBCL; Achenbach and Rescorla 2000). CBCLs were completed by foster parents rather than biological parents in an attempt to provide a more objective estimate of behavioral symptoms because court-involved biological parents might be biased about child symptomatology (Trickett and Sussman 1988). For the comparison group, participants also came into a clinic laboratory to participate in the parent–child interaction procedure and their biological parents completed the CBCL on the same day.

Parent–Child Interaction Procedure

The parent–child interaction procedure is an assessment of the parent–child relationship in which the dyad is both stressed and allowed opportunity for fun (Crowell and Feldman 1988; Heller et al. 1998). This type of assessment procedure has been used extensively in clinical settings to identify parent and child strengths within their interactions and identify points of intervention (Zeanah and Larrieu 1998). The parent–child dyad is instructed to complete seven different tasks: free play, cleanup, bubbles, and four teaching tasks of increasing developmental difficulty. The procedure lasts 30–45 min and is videotaped. The first segment of the structured interaction is a ten minute free play period where the parent and child are instructed to play as they normally do at home. The second segment is the cleanup task in which the child is asked to clean up his or her free play toys while the parent gives help if needed. The cleanup is variable in length (with a time limit of 5 min) depending upon how the parent and child handle the task. The third segment is two to three minutes in which the parent asks the child to pop bubbles as she blows the bubbles. The final four segments are 4 teaching tasks, such as pop-up toys and puzzles, of increasing developmental difficulty. The first tasks which are selected to be below and at the child’s developmental level last between two and four minutes. The last two tasks, which are selected to be above the child’s level, take between three and five minutes depending on the child’s age and skill with the task. The parent is instructed to give help if the child needs it and continue until the researcher calls.

The procedure is later coded on three affect scales for parent and child (positive affect, withdrawal/depression, and anger). In the current study, the child and parent withdrawal/depression scales were dropped due to low variability. Child task effortful control is also coded. All scales described range from one to seven where one is low in the construct of interest. Effortful control was coded as the mean score across the cleanup and the most difficult teaching task because these tasks require the most effortful control. Affect intensity was measured as the mean across all tasks.

In the original study using the parent–child interaction procedure, Crowell and Feldman (1988) found 93% discriminant validity for predicting clinical or non-clinical groups. The coding scheme used here is the one adapted for younger maltreated children (Heller et al. 1998). As reported in Crowell and Feldman (1988), scores were considered reliable if agreement was within one point. The primary coder completed reliability training for this parent–child interaction procedure and received a coding reliability score of over 0.75 for one-point percentage score agreement with expert coder on each scale; correlations with expert scores ranged from .70 to 1.00. The same primary and secondary coders were used for both the maltreatment and comparison groups. Additionally, over 25% (n = 35) of the tapes were double-coded for inter-rater reliability. Inter-rater percent agreement scales, within one point, for the mean score of the remaining scales ranged from 71 to 84%.

Measures

Emotion Regulation

Two components of children’s emotion regulation (child affect intensity and use of effortful control) were measured using the parent–child interaction procedure. Child affect intensity was measured using the mean of all the individual segment scores from the parent–child interaction procedure (free play, bubbles, cleanup, and the four teaching tasks). Affect intensity was calculated for two emotions for the child: positive affect and anger. Child positive affect was a measure of the intensity and frequency of smiles, laughter, and joyful expressions toward the parent. Child anger was a measure of the intensity and frequency of irritability, angry withdrawal, and hostility expressed toward the parent. A score of one for each emotion would indicate absence of the construct and a score of seven would indicate an extremely high level of the construct.

Child use of effortful control was measured by the level of persistence and frequency of on task behavior during the parent–child interaction procedure. One is equal to no effortful control, “the child actively tries to avoid the task. The child seems to want no part in this problem-solving exercise and spends very little time doing the task at all,” and a score of seven is equal to extreme or very high effortful control, “the child’s motivation to master the task appears to come from the child not the parent” (Heller et al. 1998, p. 26). The effortful control score was coded as the mean score across the cleanup and the most difficult teaching task because these tasks require the most effortful control. A high score would therefore indicate a considerable use of effortful control.

Parenting

Additionally, the parent–child interaction task was used to measure parenting variables: parental positive affect intensity and parental anger intensity. Parent affect intensity was measured using the mean of all the individual segment scores from the parent–child interaction procedure (free play, bubbles, cleanup, and the four teaching tasks). Parent positive affect was measured as the intensity and frequency of parental displays of joy or pleasure toward the child (such as smiling and laughter). Parent irritability/anger intensity was measured as the intensity and frequency of hostility, angry expressions, threatening behaviors, and frustration displayed toward the child. As in the child scales, a score of one on each construct would indicate absence of the construct and a score of seven would indicate an extremely high level of the construct.

Symptoms of Psychopathology

Symptoms of psychopathology were measured using the Child Behavior Checklist, CBCL (Achenbach and Rescorla 2000). Both internalizing and externalizing broad band t-scores (M = 50, SD = 10) scales were used; they are normed by age and gender. The CBCL is a 100-item checklist, completed by the child’s current caregiver, which gives information about symptoms of psychopathology. The CBCL has been validated on large, nationally representative samples and is shown to be stable across time (Achenbach and Rescorla 2000). Test-retest reliability coefficients over one month averaged .90 for the broad bands (internalizing and externalizing).

Participants in this study received one of two versions, CBCL/1.5–5 years (Achenbach and Rescorla 2000) or CBCL/2–3 years (Achenbach 1992) based on their age and the date of administration. Both CBCL versions utilize the same coding scheme and scores are adjusted for age and gender norms. Most of the children in this study received the CBCL/1.5–5 years. The CBCL/1.5–5 years version consists of the internalizing and externalizing scores and seven narrow band syndrome scores (withdrawn, somatic complaints, anxious/depressed, attention problems, emotionally reactive, sleep problems, and aggressive behavior), which additionally yields one total score. The CBCL/2–3 years version generates the same two broad band scores and six narrow band scores, five of which are the same as the CBCL/1.5–5 years version (withdrawn, somatic complaints, anxious/depressed, sleep problems, and aggressive behavior) and one which is unique to the CBCL/2–3 years version (destructive behavior). Both the internalizing and externalizing broad band scales have found to be highly correlated across the two versions, r = .86 and r = .93 respectively (Achenbach and Rescorla 2000). T scores were used because they are normed and therefore may be more comparable across versions.

Type of Maltreatment

Child protective services (CPS) validated reports of maltreatment and clinical case files on the maltreated children were reviewed and evaluated using Barnett et al.’s (1993) classification system. First, CPS court documents (CPS investigative summary, adjudication reports, hearing minutes) and other legal/medical documents (police records, hospital forensic reports) were reviewed. Then all the clinical assessments (interviews, Partner Violence Inventory adapted from Straus 1979, parent–child dyadic observations) were reviewed to fill in any unclear information. Finally, all cases were compared to clinical update reports to determine any changes in maltreatment status. Over sixty percent of the maltreated cases (n = 35) were double coded for inter-coder reliability. Exact percent agreement scores for maltreatment types ranged from 74–100%.

Results

Descriptive Analyses

Means and standard deviations for all the major variables are presented in Table 1 by group membership. The average length of time in foster care at the assessment was somewhat variable, range of 1.22–26.57 months, however length of time in care did not correlate significantly with any of the major variables nor did it moderate the relations between constructs. Inter-correlations for all variables are found by group membership in Table 2. As mentioned previously, the two samples were not significantly different on any demographic variable except maternal education. Despite a significant difference, both groups had low maternal education levels with the control group having a mean of 12 years and the maltreated group a mean of 10 years. Furthermore, when maternal education was partialed from the zero-order correlations for the entire sample, relations were highly similar. Thus, maternal education did not attenuate the overall pattern of relations and was not controlled for in the following analyses.

Table 2 Inter-correlations of major variables, non-maltreated above the diagonal and maltreated below the diagonal
Full size table

Group Differences in Emotion Regulation and Adjustment

Results partially supported hypothesis one that emotion regulation is associated with maltreatment. No differences were found among maltreatment types, but differences were found comparing maltreated versus non-maltreated children. Multivariate Analysis of Variance (MANOVA) revealed significant differences between non-maltreated and maltreated children for the major study variables F(1,117) = 7.45, p < .001. Post-hoc univariate analyses revealed lower child positive affect intensity, higher child anger intensity, lower parent positive affect intensity, higher parent anger intensity, and higher child internalizing for the maltreated group (see Table 1).

Linear Relations Among Constructs

As hypothesis two proposed, parenting was associated with emotion regulation for both the maltreated and non-maltreated groups in the expected directions. Pearson correlations for the entire sample revealed that positive parental affect intensity was associated with higher child positive affect (r = .65, p < .001) and marginally associated with higher child effortful control (r = .18, p < .10). Positive parental affect was also related to lower child anger (r = −.28, p < .001). Conversely, parent anger was associated with lower child positive affect (r = −.52, p < .001) and lower effortful control (r = −.32, p < .001). Parent anger was associated with higher child anger (r = .60, p < .001). Parenting was also significantly associated with symptoms of child psychopathology.

Correlations for the entire sample revealed positive parent affect was related to lower internalizing symptoms (r = −.29, p < .05), whereas, parent anger was associated with higher internalizing symptoms (r = .30, p < .05). Correlations for the maltreated group revealed that positive parental affect was associated with higher child positive affect and marginally associated with lower child anger (see Table 2). Conversely, parent anger was associated with lower child positive affect, lower effortful control, lower parental positive affect, and higher child anger. For the non-maltreated group, positive parental affect was associated with higher child positive affect and effortful control (see Table 2). For the non-maltreated group, parent anger was associated with lower child positive affect, lower effortful control, lower parental positive affect, and higher child anger. These findings suggest that parenting was associated with positive and negative affect and effortful control in the parent–child dyad.

In terms of parenting and psychopathology, there were no significant correlations for the nonmaltreated group but parent anger was associated with higher internalizing symptoms (r = .41, p < .05) among the maltreated group. This suggests that parent affect, specifically negative affect, is related to internalizing symptoms and this may be more salient among maltreated children.

Hypothesis three, emotion dysregulation would be related to psychopathology, was supported for the entire sample and among the maltreated group. Correlations for the entire sample revealed child positive affect intensity was related to lower internalizing symptoms (r = −.25, p < .05) and marginally to lower externalizing symptomatology (r = −.18, p < .10). Child effortful control was related to lower externalizing symptomatology (r = −.22, p < .05) and marginally to lower internalizing symptomatology (r = −.18, p < .10). As seen in Table 2, correlations for the maltreated group indicate that internalizing symptomatology was associated with lower child positive affect and higher child anger intensity. Externalizing symptomatology was marginally associated with lower effortful control for the maltreated group. There were no significant correlations for emotion regulation and psychopathology among the non-maltreated group.

Correlations for maltreated and non-maltreated groups were compared by transforming correlations to z scores and dividing the difference by the standard error. Several correlations were significantly different by group. The difference between the correlations of child anger and parent anger, were significant at the p < .05 level. The difference between the correlations of effortful control and internalizing were significant at the p < .01 level. Overall, findings for children’s emotion regulation and psychopathology suggest that observed affect and effortful control is more related to symptoms of psychopathology among maltreated children compared to non-maltreated children.

Relations Between Emotion Regulation, Parenting, Maltreatment Status, and Psychopathology

It was expected that maltreatment status would moderate the relation between emotion regulation and psychopathology and the relation between parenting and psychopathology. Baron and Kenny’s (1986) procedure was used to test for moderation. Emotion regulation and parenting variables and their interactions were centered (M = 0) prior to inclusion in the regression equations in order to minimize multicollinearity (Aiken and West 1991). Holmbeck’s (2002) procedure for testing the significance of the slopes for maltreatment status was then used.

Significant interactions were found for child anger with parent and internalizing (see Table 3). Among maltreated children, increased anger with their parent was significantly related to increased levels of internalizing behaviors. In contrast, among non-maltreated children, anger with their parent was not significantly related to internalizing behaviors. There were no significant interactions found for the moderation of parenting and psychopathology by maltreatment status.

Table 3 Regression analyses predicting internalizing from effortful control and anger, maltreatment status, and their interactions
Full size table

Relations Between Emotion Regulation and Maltreatment Experience

Exploratory analyses were also run in the maltreated group comparing type of maltreatment, number of different types of maltreatment, and developmental timing of maltreatment. There were no differences found for type, timing or number of maltreatment types for any of the emotion regulation variables, or for internalizing and externalizing symptomatology. Moreover analyses were run comparing length of time in foster care on emotion regulation variables. No significant differences were found for length of time in foster care and any of the emotion regulation variables or psychopathology.

Discussion

Even very young children are sophisticated social partners. By using a standardized assessment procedure and then coding interactive behavior using an emotion regulation framework, we were able to see differences between how maltreated children under age 4 respond to caregivers and vice versa when compared to non-maltreated children. It is no surprise that what has been called the “interactive dance” (cf., Cooper et al. 2005) between a young child and his or her caregiver might be altered in the context of maltreatment. Nevertheless, these data suggest that standardized assessments of dyadic emotion regulation are clinically useful and even suggest that such assessments reveal particular patterns of interaction which may be predictive of later child psychopathology in young maltreated children.

We did not take for granted that the interactive dance between children and the mothers with whom they lived when maltreated would be different than the dance between non-maltreated children and their mothers. Parents in the maltreatment group were aware that they were being videotaped and aware that data from the assessments would be shared with the courts in helping to determine placement status. Moreover, maltreated children in this study were living separately from their parents at the time of the assessment. Even in a context where such parents might be expected to “be on their best behavior,” our raters, who were unaware of maltreatment status, observed lower parent positive affect intensity and higher parent anger intensity in the maltreatment group. Interestingly, the maltreated children also showed lower positive affect intensity and higher anger intensity with their mothers. Some reports have shown that maltreated children, even preschoolers, dampen their negative affect in the presence of an angry maltreating caregiver, a syndrome which has been labeled “compulsive compliance” (Crittenden and Dillala 1988; Koenig et al. 2000). It is hypothesized that compulsive compliance is a protective interactive behavior among young maltreated children who fear eliciting parental anger. However, compulsive compliance appears uncommon and other studies are consistent with ours in revealing that many maltreated children match their parents’ negative affect step for step, sometimes even setting up coercive cycles of negative behavior (Dodge et al. 1990; Erickson et al. 1989). It is possible that many of the maltreated children in our sample were maltreated by others in the family rather than by their mothers and so fear of the caregiver might have been less salient than in other samples. In fact, the most common subtype of maltreatment experienced was neglect (rather than the comparatively far less common physical or sexual abuse) in our maltreated sample. The fear underlying “compulsive compliance” would be less likely following neglect than physical abuse though the anger we observed maltreated children to show might be predictable following either neglect or other types of maltreatment. However, exploratory analyses did not reveal differences on psychopathology or emotion variables by type of maltreatment.

Exploratory analyses examined relations between maltreatment experience, emotion regulation, and psychopathology. The lack of findings is surprising given the extensive data that has found relations between maltreatment variables, emotional development, and psychopathology (Bolger and Patterson 2001; Cicchetti and Barnett 1991; Kaufman and Cicchetti 1989; Pinderhughes 1998; Toth Cicchetti et al. 2000). The description of maltreatment experience in the current study reflects a comprehensive review of substantiated reports from court records and clinical evaluation; however, it is important to note there is a pervasive problem within maltreatment research of poor records and inconsistent legal documentation. In a study of over 800 children aged four to eight years, Hussey et al. (2005) found that there were no differences on child behavioral or developmental outcomes between substantiated and unsubstantiated maltreatment reports. Therefore it may not be surprising that parenting variables and not elements of the maltreatment experience were associated with child emotion regulation and psychopathology given that maltreatment experiences may be even more complex than we were able to document. Moreover, there was significant overlap in maltreatment type, almost three-quarters of the sample experienced more than one type of maltreatment, and the developmental timing of the maltreatment was a relatively small range because all children were recruited for the study having experienced maltreatment prior to the age of four years. Future work should further explore maltreatment experience with a larger sample by examining differences between substantiated and unsubstantiated reports along with including observational measures of parenting within the home environment.

Our most surprising finding was that both observed parental anger and observed child anger was associated with higher internalizing (rather than externalizing) behavior in children. In fact, the relations between parental anger and internalizing symptoms and between child anger and internalizing symptoms held only among maltreated children. On one hand, it seems counter-intuitive that the kind of dance that is characterized by high levels of parent and child anger and negative affect would not generalize to a child showing the oppositional or aggressive behaviors in other settings. On the other hand, it is not surprising that children who repeatedly encounter parental anger might become both angry and sad or withdrawn. The fact that our maltreated sample had a high proportion of children who were neglected may explain why parental anger (in the context of neglect rather than other types of maltreatment) was associated with internalizing symptoms. Previous research has found that physical abuse in early childhood is highly related to externalizing symptomatology (Bennett et al. 2005; Egeland et al. 2002; Manly et al. 2001), whereas neglect is more highly related to internalizing symptomatology (Bolger and Patterson 2001; Hildyard and Wolfe 2002; Manly et al. 2001). Still, our moderational analysis was interesting in that only for the maltreated children in our sample was more parental anger linked to more internalizing symptoms. For clinicians observed anger (either parental or child) in the context of maltreatment should trigger concerns about internalizing disorders. Internalizing disorders can be under-recognized in young children because quiet children sometimes don’t get the attention of teachers or parents (Dwyer et al. 2006); such disorders are nonetheless highly impairing (Kovacs and Devlin 1998).

Our data also show that low levels of effortful control among maltreated children may portend externalizing disorders at least in maltreated children; here, young maltreated children who seem less able to delay gratification or shift their attention when working on tasks with a caregiver were the ones rated to show more externalizing symptoms.

The fact that standardized assessments using an emotion regulation framework yielded interesting differences among groups of children and even predicted parental ratings of psychopathology led us to think about implications of these data for treatment. The data from the current study suggest that parents whose children have been maltreated could benefit from training on being better, more responsive “interactive dance” partners, through instruction on modeling, emotion labeling, and demonstrating appropriate affect in response to their children’s cues (Gottman et al. 1996). Because maltreating parent–child interactions appear to be characterized by more anger and less positive emotions, such approaches are critical for maltreating families. Parent management training approaches, which are designed to teach parents how to change their children’s behaviors, have demonstrated some success in improving parenting skills and ameliorating risk among parents at high risk for child maltreatment (Barth 1991; Timmer et al. 2005; Wolfe et al. 1999). Moreover interventions which specifically train parents how to follow their infants’ lead have been able to improve young children’s emotion regulation and symptoms of psychopathology (Cohen et al. 2002; Knapp et al. 2007). One model, the Circle of Security (COS, Marvin et al. 2002), goes a step further by teaching parents not only how to read their child’s cues (after giving them a “roadmap” for understanding the attachment system) but also helping them identify issues from their own past which might inhibit them from seeing and responding appropriately to their child’s need for comfort or support (Hoffman et al. 2006). Videotaped review of the parent–child interaction is central to the Circle of Security model. Taken together, these interventions suggest that working with parents to attend in new ways to the parent–child dance is effective in affecting children’s emotion regulation abilities and symptoms of psychopathology. Identifying the key elements in any intervention is a next step for the field though it appears that effective interventions can be relatively short in length especially if such interventions focus on parental responsivity and emotional communication, practice parenting techniques (Kaminiski et al. 2008), are family-based, and use video feedback as an intervention tool (Bakermans-Kraneburg et al. 2003).

The current study has many strengths. First, this study used observational data to investigate emotion regulation within the context of the parent–child dyad rather than relying on parent report. Indeed, emotion regulation in early childhood is best conceptualized within the parent–child relationship (NICHD 2004) because young children’s environments are often controlled and shaped by their parents. Second, the use of foster parent report of symptomatology for the maltreated group may have reduced parental bias and provided validity for the observations with the biological parent. Third, this study is one of few (Erickson et al. 1989) that has investigated emotion regulation in the maltreating parent–child dyad with very young children. Finally, this study also explored the heterogeneity of maltreatment through multiple forms of substantiated verification by examining differences in emotion regulation in terms of type of maltreatment and multiplicity of maltreatment types.

A limitation of the present study was that emotion regulation was measured in the parent–child context and with mothers only; however, the parent–child context has the most ecological validity because most emotion regulation development at this age occurs primarily within this context. Mothers were chosen for this sample because they were most often the primary caregivers; nonetheless, the maltreatment literature has severely neglected the role of the fathers in emotional development and future work must include fathers. The maltreated children in the current study were assessed with the family where maltreatment occurred and following a separation; therefore, these data may be less generalizable to other potentially less emotional contexts. In addition, the maltreated children had not lived with their biological parents for six weeks at the time of assessment. Thus, their reactions within the dyad may be more severe than would normally occur with this parent. Another limitation is that visitation with a biological parent while in foster care varied by CPS case plan and parental motivation. Visitation level might be an important mediator of maltreatment effects, for example consistent biological parent contact could have made the parent–child assessment procedure less anxiety-provoking and provided less unpredictability in interaction. In contrast, biological parent visitation on the same day as the completion of the CBCL could have lead to increased reporting of symptomatology by the foster parent due to the disruptiveness of the visitation. In general, biological parent visitation was generally encouraged and lack of visitation may simply offer further information about the quality of the parent–child relationship. Unfortunately, the reality of child maltreatment research is that disruption of placement, the stress of visitation, and the maltreatment itself are all confounded. Finally, foster parents were chosen to report on child symptomatology in an attempt to provide a less biased estimate of adjustment. The mean CBCL scores reported by foster parents in the current study were very similar to scores reported by biological parents of young maltreated children in another study (Dombrowski et al. 2005). It should be recognized, however, that foster parents may be biased in their reporting in an attempt to receive more support or present a “rosier” picture of the child to reflect their parenting skills. In addition, there was also some missing CBCL data and it is possible that only the foster parents with the most child behavior problems completed the CBCL in attempt to receive services. Both these cases could be equally true for biological parents. A final limitation is that four of the maltreated children had been placed with their foster parents for slightly more than a month and the CBCL requires parents to report on behaviors over the past two months. It is possible the data that these foster parents report may be less accurate than data for children placed longer; however, Heflinger et al. (2000) found their CBCL data reported by foster parents of children placed at least 2 months was comparable to data reported by foster parents of children placed at least one month (Marcus 1991). Future investigations would be enhanced by the observation of the parent–child dyad outside the laboratory setting as well, measuring parental visitation and using multiple reporters to capture the most comprehensive information on dyadic regulation and adjustment.

One challenge in planning the analyses had to do with whether to control for maternal education. Previous research (Casady and Lee 2002; Kotch et al. 1995) has found lower maternal education to be a risk factor for maltreatment. We elected to focus on the results without controlling for maternal education, as doing so would limit the power of the current study and decrease the ecological validity of the project because maternal education is rarely something intervened with in interventions. However, when maternal education was controlled in post-hoc analyses in the comparison of maltreated and non-maltreated children only internalizing symptomatology became non-significant. Additionally, the moderation of the relations between child anger and internalizing symptomatology by maltreatment status remained unchanged. These results suggest that perhaps maternal education is related to child psychopathology but not through an effect on child emotion regulation.

In sum, our data suggest that maltreated children experience difficulties in emotion regulation during interactions with caregivers, and higher levels of behavioral symptomatology may result. Assessments which focus on the parent–child interactive dance and use an emotion regulation framework should be further developed and tested so that the large numbers of maltreated children in the U.S. can be placed on a trajectory of healthy development.