Introduction
Seventy six percent of the 600,000 children reported to be maltreated in the U.S. experience neglect, making it the most common form of maltreatment [56]. Children with histories of neglect are at increased risk for behavior problems, including both externalizing problems such as aggression and internalizing problems such as anxiety, social withdrawal, and depression [2, 4, 50]. Despite neglect’s negative impact, the specific processes by which neglect may lead to the development of behavior problems is poorly understood.
Emotion knowledge (EK) deficits are hypothesized to mediate the relationship between neglect and poor adjustment [29]. By age 2 years, children exhibit stable individual differences in EK that may be related to differences in children’s self-regulation, verbal ability, and parenting [14, 19]. Emotion knowledge or understanding is characterized by multiple interrelated dimensions including the ability to recognize emotional cues from facial, verbal, or behavioral expressions (emotion or expression recognition); correctly label emotions (expression labeling); and identify situations that elicit specific emotions (situational knowledge) [10, 45]. Neglected children, as a group, exhibit deficits in these skills relative to their same-age peers. Neglected children, for example, exhibit a poorer understanding of and greater difficulty discriminating emotions and emotional expressions [8, 36, 43, 47, 51]. Such EK deficits are associated with both externalizing and internalizing problems [12, 24, 48, 54], further lending support to the hypothesis that EK mediates the relationship between neglect and the development of future behavior problems.
Several potential mechanisms could explain how neglected children develop EK deficits. First, neglectful parents are more likely to be socially isolated, potentially limiting the number of non-parental models and emotions to which their children are exposed [11, 16]. Likewise, neglectful parents tend to exhibit fewer positive emotions, thus providing fewer opportunities for their children to become familiar with these emotions [43, 61]. In addition, neglectful mothers have been found to engage in fewer emotion-focused discussions, provide less support in response to children’s emotional displays, and engage in less elaborative reminiscing about emotional events with their children [18, 57]. Collectively, this relative lack of opportunities may affect young children’s skills in learning to decode and to respond appropriately to emotional signals [43].
Currently, our understanding of how neglect increases risk for behavior problems is limited. As such, research examining EK as a potential mediator may inform the potential etiological pathway between neglect and behavior problems, leading to the development of more focused and effective interventions aimed at preventing the onset and maintenance of behavior problems among neglected children. In examining such mediation models, it is important to consider whether EK mediates the impact of neglect on both externalizing and internalizing problems. While they often co-occur, externalizing and internalizing problems present with somewhat distinct risk factors (e.g., emotion dysregulation vs. self-blame, respectively) among youth with neglect histories [31, 53], raising the possibility that EK may be a stronger mediator for one than the other. For example, lower emotion recognition was found to predict greater internalizing but not externalizing problems in a community sample of children [9].
Developmentally, early childhood is a particularly important period during which to examine these constructs as maltreatment most often occurs during the first five years of life [56]. Moreover, EK skills steadily develop during these years, a time when the association between maltreatment and EK appears to be greatest [36, 60]. Finally, EK skills in preschool-age children may be important in facilitating later functioning as they predict greater behavioral adjustment, social competence, and academic achievement in the elementary school years [14, 15, 55].
The purpose of the current study was to examine the prospective relationship between neglect, EK, and behavior problems. Neglect history was assessed at age 4.0 years, EK at 4.5 years, and behavior problems during the spring of children’s 1st grade school year when they were age 6 to 7 years. The EK battery assessed expression labeling, expression recognition, and situational knowledge. We hypothesized that: a) child neglect history at age 4.0 years would predict poorer EK at 4.5 years; b) child neglect history would predict greater behavior problems during 1st grade; c) poorer EK at 4.5 years would predict greater behavior problems during 1st grade; and d) EK would mediate the relationship between neglect and behavior problems. In addition, we explored whether the hypothesized relationships were of similar strength for each dimension of EK, as well as for both externalizing and internalizing behavior problems in addition to our main outcome, total behavior problems. Teachers were used to assess behavior problems given that maltreating parents may overreport their children’s problems [32].
Methods
Participants
Participants were 127 children (56 neglected children [52% female]; 71 comparison children [55% female]) from two longitudinal studies examining adjustment among maltreated children. In each study, children were initially seen at age 4.0 years. The race of children was as follows: 75% Black/African American, 15% Latine, 7% White, and 3% other. Mothers had a mean education level of 12.2 years of schooling (SD = 1.8). Samples from the two studies did not differ significantly in terms of proportion of children with a history of neglect; race; gender; child age at the time of testing; and mothers’ years of education (p > 0.15 for all). Children from foster homes were excluded given the study’s focus on the children’s ongoing emotional development and adjustment within the context of their family of origin. In addition, children with histories of sexual abuse as documented by Child Protective Services (CPS) records (n = 3) and children with known histories of intellectual disability or a significant physical disability (n = 4) were excluded from study enrollment. Children with both neglect and physical abuse allegations (n = 3) were included in the first study sample; physical abuse was an exclusionary criterion in the second sample. All children were recruited for a study on child development through flyers distributed or posted in community settings (e.g., WIC program offices; publicly funded preschool or therapeutic programs; grocery stores; libraries; laundromats) in urban neighborhoods in Philadelphia and central New Jersey.
Measures
Neglect. CPS case records were used to assess neglect history at the time of study intake. History of CPS allegations were confirmed by record review from the Department of Human Services for Philadelphia participants and the Division of Youth and Family Services for New Jersey participants. Neglect was determined by having at least one substantiated CPS report or having an open case record at the time of enrollment. Neglect status was coded as 0 if there was no history of neglect or other types of maltreatment and 1 if there was a history of neglect.
Emotion knowledge (EK). Detailed descriptions and procedures for each task appear in prior work [3, 39] and are briefly described here. Three EK tasks were administered in fixed order: a) expression labeling; b) expression recognition; and c) situational EK (i.e., matching expressions to context).
Expression Labeling. Six black and white photographs of a child of color (“Felicia”) depicting a facial expression (happy, surprise, anger, sad, fear, and disgust) were taken from a standard set of independently verified Facial Action Coding System photographs previously used with children [6]. Children were shown the photographs in a fixed order, beginning with “happy.” The examiner presented each photo and asked: “What kind of face is Felicia making? How does she feel?” Responses were coded as correct if children gave the target label or an acceptable synonym (i.e., “mad” for anger; “scared” or “afraid” for fear; “shocked” for surprise; and “yucky” or “nasty” for disgust). Scores could range from 0 to 6 (the internal consistency (KR-20) coefficient was 0.55).
Expression Recognition. To assess recognition, all six photographs from the Expression Labeling task were randomly placed in front of the child. The examiner then told the child to point to a specific face (e.g., “How does Felicia look when she’s __?”). After assessing all six expressions, the pictures were reshuffled and the task repeated. The recognition score consisted of the number of expressions correctly recognized on both trials, with scores ranging from 0 to 12 (the internal consistency (KR-20) coefficient was 0.65).
Situational Knowledge. To assess children’s ability to match expressions to context, the examiner read 10 brief stories, each illustrated by black and white line drawings in which the facial expressions of the people were left blank (e.g., “It’s dinner time and Felicia’s mommy made her eat something nasty. In this picture it’s spinach and Felicia hates spinach.”). To assess children’s ability to link emotions to contexts, all six expression photographs were again randomly displayed on a table at the beginning of each story. The examiner then asked, for example, “How did Felicia look when she had to eat it? Point to the face that Felicia made.” After the child pointed to a picture, the examiner then said, “Did Felicia make any other faces?” If children said “yes”, the examiner asked them to point to a second face. Selection of the target emotion for each context in response to either query was counted as a correct response. Scores could range from 0 to 10 (the internal consistency (KR-20) coefficient was 0.59).
Composite Emotion Knowledge (EK) score. Raw scores from each of the three EK measures were converted to standardized (Z) scores. EK scores were significantly correlated (r = 0.369 to 0.590; p < 0.001) so a composite score was created by calculating the mean Z score of each (alpha = 0.66).
Behavior problems. The Teacher Report Form (TRF) [1] was used to assess behavior problems. The TRF includes 118 items that assess the frequency of problem behaviors or symptoms “now or within the past 2 months” (0 = “not true”, 1 = “somewhat or sometimes true”, 2 = “very often or often true”). The externalizing problem scale includes items from the aggressive behavior and the delinquent behavior subscales, whereas the internalizing problem scale includes items from the anxious/depressed, withdrawn, and somatic complaint subscales. Total behavior problems, the primary outcome, is based on the sum of all 118 items including those in the externalizing and internalizing scales as well as other problems. The TRF has high test–retest reliability, moderately high agreement between teachers, and discriminates children referred vs. not referred for mental health services [1].
Procedures
During enrollment, mothers were informed that the study included the examination of emotional development among children with a history of maltreatment. Mothers signed consent forms permitting the research team to review CPS records for maltreatment allegations involving their child. The study was approved by Institutional Review Boards at the participating universities and written informed consent was obtained from mothers. Mothers and their children came to a research program office to participate. Once enrolled at age 4.0 years, study staff reviewed CPS records to determine if there was a history of maltreatment, including neglect. EK was assessed at the 4.5 year visit by a female research assistant blind to children’s CPS status. Behavior problems were assessed by teacher rating on the Teacher Report Forms (TRF; [1]), which was sent to the child’s teacher during the spring of their 1st grade school year when children were a mean age of 6.7 years (SD = 0.5). Teachers returned the TRF for 87 of the 127 children; no significant differences (p > 0.15) were observed on gender, maternal education, race, age at the time of the EK assessment, or neglect status between participants with vs. without a returned TRF. Families received a gift card at each study visit, as did teachers upon receipt of the completed TRF. Follow-up phone calls were made to teachers who did not initially return the TRF to optimize response rate.
Data Analysis Plan
Bivariate correlations were run to examine associations among all study variables. Overall, 9.6% of data were missing, including 0.2% of predictor variables and 31.5% of outcome variables. Multiple imputation (MI) of 100 data sets was conducted to address missing data among predictor variables in the correlation analyses; outcome variables (behavior problems) were used to create the MI models but not imputed for analyses [59]. Little’s test was consistent with the interpretation that data were missing completely at random [χ2 (18) = 13.93, p = 0.734].
Three mediation models were examined: 1) EK composite as the mediator of total behavior problems (i.e., the most global model); 2) EK composite as a mediator of both externalizing and internalizing problems to examine whether mediation, if present, was similar for both problem types; and 3) each individual EK dimension (i.e., expression recognition, expression labeling, and situational knowledge) as a mediator of total behavior problems to examine whether any resultant mediation was strongest for a particular dimension. Mediation analyses were conducted in [30], which uses structural equation models to estimate direct and indirect effects. Bayesian mediation analyses were conducted because they offer several advantages to conventional frequentist mediation approaches, including the provision of more robust statistical estimates with relatively modest samples such as used in the current study [62]. Bayesian Markov chain Monte Carlo (MCMC) estimation using non-informative prior distributions was employed with four chains, a burn-in of 1000 steps, 10,000 sample iterations, and residual covariances for all variables; convergence was obtained as R-hat values were acceptable for each analysis (all = 1.00; cf. [23]. In Bayesian MCMC modeling, exact p values for significance testing are not provided, as statistical significance is based on estimating a credibility interval (CI) for each parameter and 95% CIs that do not include 0 indicate that a parameter is statistically significant at alpha = 0.05 [58]. In addition, 90% CIs are also reported. Because age was significantly correlated with internalizing problems (r = −0.214, p < 0.05), it was included as a covariate in model 2. In addition, maternal education was included as a covariate in model 3 because a non-significant trend was found between it and expression labeling (r = 0.165, p = 0.064).
Results
Relationship among Study Variables
Table 1 presents descriptive statistics and correlations among all study variables. Of note, externalizing and internalizing problems were only moderately correlated (r = 0.455, p < 0.001), thus suggesting some degree of independence between the two behavior problem dimensions. Findings regarding specific hypotheses follow, with a focus on the most comprehensive measures of EK (i.e., the EK composite) and behavior problems (total problems).
Table 1
Descriptive statistics and correlations between study variables
Mean | N | Neglect | Age | Maternal educ | Emotion knowledge | Behavior problems | ||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
Composite | Expression recog | Expression labeling | Situational knowledge | Total problems | Extern. problems | Intern. problems | ||||||
Neglect (0 = no, 1 = yes) | 0.44 (0.50) | 127 | – | |||||||||
Age at 4.5 year visit | 4.67 (0.25) | 127 | 0.003 | – | ||||||||
Maternal education | 12.10 (1.69) | 127 | −0.276** | 0.238** | – | |||||||
Emotion Knowledge | ||||||||||||
Composite (Z-score) | −0.01 (0.80) | 127 | −0.162 | 0.069 | 0.106 | – | ||||||
Expression recognition | 6.45 (2.87) | 126 | −0.166 | 0.096 | 0.064 | 0.836*** | – | |||||
Expression labeling | 2.11 (1.34) | 127 | −0.142 | 0.056 | 0.165 | 0.817*** | 0.590*** | – | ||||
Situational knowledge | 8.42 (3.16) | 126 | −0.067 | 0.000 | 0.029 | 0.742*** | 0.435*** | 0.369*** | – | |||
Behavior Problems | ||||||||||||
Total behavior problems | 28.67 (29.21) | 87 | 0.227* | −0.094 | 0.007 | −0.383*** | −0.320** | −0.218* | −0.387*** | – | ||
Externalizing problems | 9.08 (11.64) | 87 | 0.196 | −0.020 | 0.005 | −0.371*** | −0.326** | −0.213* | −0.373*** | 0.884*** | – | |
Internalizing problems | 7.98 (9.96) | 87 | 0.122 | −0.214* | 0.015 | −0.212* | −0.139 | −0.115 | −0.245** | 0.745*** | 0.455*** | – |
Hypothesis 1: Neglect predicts future total behavior problems
As hypothesized, children with a history of neglect exhibited more total behavior problems than children without a history of neglect (r = 0.227, p = 0.035).
Hypothesis 2: Neglect predicts poorer emotion knowledge
Neglect did not significantly predict the EK composite in a zero-order correlation (r = −0.162, p = 0.069) nor in the context of the mediation models (i.e., there is not a 95% probability that the true estimate differs from zero although the 90% CI did not contain zero [mean change in EK composite for children with a history of neglect = −0.260 (SD = 0.143); −0.495 to −0.026, model 1; see Table 2)]. Thus, support was not found for the hypothesis that neglect predicts EK deficits.
Table 2
Test of mediation: does emotion knowledge mediate the relationship between neglect and behavior problems?
Posterior | 95% Credible interval | 90% Credible interval | |||||
|---|---|---|---|---|---|---|---|
Mean | Median | SD | Lower | Upper | Lower | Upper | |
1. Model 1 (total behavior problems) | |||||||
Path coefficients | |||||||
a. Neglect to total behavior problems | 11.094 | 11.084 | 4.725 | 1.743 | 20.416 | 3.318 | 18.848 |
b. EK composite to total behavior problems | −10.887 | −10.892 | 3.228 | −17.229 | −4.534 | −16.149 | −5.575 |
c. Neglect to EK composite | −0.260 | −0.259 | 0.143 | −0.542 | 0.020 | −0.495 | −0.026 |
Total effects and total indirect effects | |||||||
a. Total effects | 13.917 | 13.925 | 4.942 | 4.174 | 23.620 | 5.792 | 22.032 |
b. Total indirect effects | 2.823 | 2.632 | 1.817 | −0.212 | 6.912 | 0.224 | 6.091 |
2. Model 2 (externalizing & internalizing problems) | |||||||
Path coefficients | |||||||
a. Neglect to externalizing problems | 2.809 | 2.821 | 2.290 | −1.679 | 7.281 | −0.941 | 6.556 |
b. EK composite to externalizing problems | −4.838 | −4.835 | 1.434 | −7.642 | −1.999 | −7.192 | −2.466 |
c. Age to externalizing problems | 1.976 | 1.966 | 1.937 | −1.812 | 5.773 | −1.206 | 5.172 |
d. Neglect to EK composite | −0.253 | −0.253 | 0.145 | −0.539 | 0.031 | −0.493 | −0.014 |
e. Age to EK composite | 0.186 | 0.185 | 0.281 | −0.365 | 0.738 | −0.277 | 0.647 |
f. Neglect to internalizing problems | 3.842 | 3.842 | 2.064 | −0.224 | 7.872 | 0.444 | 7.241 |
g. EK composite to internalizing problems | −1.979 | −1.980 | 1.274 | −4.474 | 0.552 | −4.076 | 0.113 |
h. Age to internalizing problems | −0.739 | −0.748 | 1.879 | −4.417 | 2.957 | −3.814 | 2.371 |
Total effects and indirect effects | |||||||
a. Total—neglect on externalizing problems | 4.037 | 4.055 | 2.358 | −0.586 | 8.623 | 0.157 | 7.884 |
b.– Neglect on internalizing problems | 4.348 | 4.353 | 2.048 | 0.310 | 8.375 | 0.981 | 7.701 |
c.– Age on externalizing problems | 1.077 | 1.088 | 2.397 | −3.658 | 5.757 | −2.872 | 5.001 |
d.– Age on internalizing problems | −1.105 | −1.105 | 1.991 | −5.015 | 2.771 | −4.383 | 2.159 |
e. Indirect—neglect on externalizing problems | 1.228 | 1.143 | 0.821 | −0.144 | 3.052 | 0.051 | 2.696 |
f.– Neglect on internalizing problems | 0.505 | 0.420 | 0.476 | −0.189 | 1.638 | −0.080 | 1.395 |
g.– Age on externalizing problems | −0.899 | −0.809 | 1.445 | −4.016 | 1.826 | −3.378 | 1.327 |
h.– Age on internalizing problems | −0.367 | −0.242 | 0.702 | −2.040 | 0.865 | −1.669 | 0.590 |
3. Model 3 (EK dimensions & total behavior problems) | |||||||
Path coefficients | |||||||
a. Neglect to total behavior problems | 9.327 | 9.319 | 4.787 | −0.145 | 18.710 | 1.418 | 17.224 |
b. Expression recognition to total behavior problems | −4.347 | −4.340 | 3.147 | −10.513 | 1.857 | −9.480 | 0.819 |
c. Expression production to total behavior problems | −1.091 | −1.080 | 3.114 | −7.245 | 5.026 | −6.244 | 4.020 |
d. Situational knowledge to total behavior problems | −7.379 | −7.386 | 2.854 | −12.983 | −1.740 | −12.084 | −2.649 |
e. Maternal education to total behavior problems | 1.805 | 1.812 | 0.784 | 0.266 | 3.345 | 0.512 | 3.094 |
f. Maternal education to expression production | 0.072 | 0.072 | 0.043 | −0.014 | 0.157 | 0.001 | 0.144 |
g. Neglect to expression recognition | −0.336 | −0.336 | 0.179 | −0.690 | 0.015 | −0.630 | −0.042 |
h. Neglect to expression production | −0.238 | −0.238 | 0.183 | −0.599 | 0.120 | −0.539 | 0.063 |
i. Neglect to situational knowledge | −0.173 | −0.172 | 0.181 | −0.529 | 0.181 | −0.470 | 0.124 |
Total effects and indirect effects | |||||||
a. Total—neglect on total behavior problems | 12.323 | 12.345 | 5.165 | 2.173 | 22.465 | 3.792 | 20.844 |
b.– Maternal education on total behavior problems | 1.725 | 1.728 | 0.794 | 0.155 | 3.277 | 0.421 | 3.030 |
c. Indirect–neglect on total behavior problems | 2.996 | 2.858 | 2.264 | −1.133 | 7.874 | −0.470 | 6.929 |
d.– Maternal education on total behavior problems | −0.080 | −0.048 | 0.266 | −0.689 | 0.429 | −0.553 | 0.315 |
Hypothesis 3: Emotion knowledge predicts future behavior problems
The EK composite predicted fewer total behavior problems as hypothesized (r = −0.383, p < 0.001). In the context of mediation models, the EK composite likewise predicted fewer total behavior problems (M = −10.887 [SD = 3.228]; 95% CI: −17.229 to −4.534, model 1; see Table 2).
Hypothesis 4: Emotion knowledge mediates the relationship between neglect and behavior problems
As shown in Table 2, the indirect effect was not significant (M = 2.823 [SD = 1.817]; 95% CI: −0.212 to 6.912, model 1), indicating that the EK composite did not mediate the relationship between neglect and greater total behavior problems. Of note, the 90% CI did not contain 0 for the indirect effect (0.224 to 6.091).
Do results vary for externalizing and internalizing behavior problems?
Examining correlations, neglect was not a significant predictor of either externalizing problems (r = 0.196, p = 0.068) or internalizing problems (r = 0.122, p = 0.260; see Table 1). Each of the three EK dimensions predicted fewer externalizing problems (expression recognition: r = −0.326, p = 0.002; expression labeling: r = −0.213, p < 0.047; situational knowledge: r = −0.373, p < 0.001). In contrast, situational knowledge was the only dimension to predict fewer internalizing problems (r = −0.245, p = 0.022).
Examining mediation models, no significant indirect effects were found for the 95% CI, although zero was not included in the 90% CI for indirect effects for the model including neglect and externalizing problems (M = 1.228 [SD = 0.821]; 0.051 to 2.696; see model 2, Table 2).
Do results vary by emotion knowledge dimension?
Neglect did not predict any of the individual EK dimensions. Each of the three EK dimensions, however, were associated with fewer total behavior problems; the relationship was greatest for situational knowledge (r = −0.387, p < 0.001). Examining mediation models, the 95% CI for indirect effects contained zero, suggesting no significant mediation (see model 3, Table 2). The path coefficient from situational knowledge to total behavior problems was significant (M = −7.379 [SD = 2.854]; −12.983 to −1.740). None of the paths from neglect to an EK dimension were significant at the 95% CI, although the 90% CI for neglect to expression recognition did not contain zero (M = −0.336 [SD = 0.179]; −0.630 to −0.042).
Discussion
The current study is the first to our knowledge to examine EK as a mediator of the prospective association between neglect and behavior problems in children. Neglect and EK both predicted future behavior problems, consistent with prior research [2, 20, 25, 41, 50, 54]. EK, however, was not a significant mediator of the relationship between neglect and total behavior problems, nor of the relationship between neglect and either externalizing or internalizing problems.
Situational knowledge was the EK dimension to most strongly and consistently predict total behavior problems, externalizing problems, and internalizing problems. Situational knowledge of emotions involves a relatively advanced set of skills that develop after those of expression recognition and labeling [13, 36]. Importantly, because situational knowledge requires the ability to take another person’s perspective [22], it may be the dimension most closely related to empathy, a lack of which has been associated with both neglect and greater externalizing but not internalizing problems [5, 22]. Likewise callous-unemotional traits, characterized by a lack of empathy, are consistently associated with externalizing problems among preschool-age children [34]. Collectively, this suggests that one potential mechanism by which situational EK may predict behavior problems, and externalizing problems in particular, is through its association with empathy.
A second mechanism by which EK may predict externalizing problems focuses on the role of emotion regulation. Emotion dysregulation, like EK in the current study, is more strongly and consistently associated with children’s externalizing than internalizing problems [28, 31], with some indication that emotion regulation skills precede the development of EK skills [35]. In addition, emotion regulation has been found to mediate the relationship between EK and adjustment in young children [17, 38]. Accordingly, it is possible that emotional dysregulation and EK deficits amplify each other in bidirectional relationships, leading to increased risk of externalizing problems among young children.
Finally, a methodological explanation deserves consideration for the more consistent relationships between EK and externalizing than internalizing problems. Specifically, teacher ratings were used in the current study because parents with a history of maltreatment may over-report their children’s behavior problems [32]. Yet, while teachers as a group may be better at identifying externalizing problems, they may be less aware of children’s internalizing problems [7, 26, 49]. As such, it is possible that teacher ratings of externalizing problems may have greater validity than teacher ratings of internalizing problems.
The hypothesized mediation of the relationship between neglect and behavior problems by EK was not statistically significant when examining 95% CIs. Of note, 89% CIs were a typical default in Bayesian statistics until fairly recently, as they are more stable [37]. Examining model 1, the current study did find a greater than 90% chance that indirect effects were non-zero, raising the possibility of partial mediation that would be identified by a larger sample in future research. In addition to testing models with a larger sample, the potential role of supportive parenting also needs to be considered. Mothers’ supportive presence at age 4 years, for example, predicts better behavioral adjustment at age 10, a relationship mediated by children’s own emotion labeling skills at ages 4 to 8 [44].
Given that EK predicted greater behavior problems, early screening of EK skills could be useful for young children. While lower EK is associated with both externalizing and internalizing problems, it is noteworthy that the negative effects of EK deficits on externalizing (but not internalizing) problems increase during the preadolescent and adolescent years [54], further highlighting the potential importance of early intervention aimed at enhancing EK in preventing externalizing problems. Accordingly, professionals working with young children should consider interventions aimed at enhancing EK skills as one component to support their emotional, behavioral, and social development (cf. [21, 40, 46, 54].
While the current study has multiple strengths including the use of longitudinal data and the testing of novel mediation models, several limitations deserve mention. First, CPS records are an imperfect measure of neglect as they may underreport the number of children who have experienced neglect [27]. Second, as noted above, teacher ratings were the sole source of behavior problems. Future studies examining EK as a mediator of neglect and behavioral adjustment should include ratings from additional sources (e.g., caregivers and children themselves at older ages). Third, mediation models are most informative regarding potential causality when measures of the independent, mediating, and outcome variables are assessed at each time point and when baseline measures of the mediator and outcome can be controlled for [33, 42]. The current study design did not allow for this possibility as many children were not in a school setting in which teacher ratings of behavior problems could be collected at the time of the 4.0 and 4.5-year assessments. Of note, CPS records were tracked throughout the study period and none of the participants had new CPS allegations between the time of the 4.0 visit and teacher ratings at age 6 or 7 years. Fourth, although no significant differences were observed in characteristics of children whose teachers returned vs. did not return behavior problem ratings, TRF data was missing for some children. Fifth, internal consistency scores for the EK measures approached but were under the generally accepted 0.70 value for Cronbach’s alpha or KR-20; greater internal consistency might enhance model prediction in future research. Finally, great heterogeneity exists among children with histories of neglect (e.g., [36]. As such, our findings do not necessarily generalize to children of other ages, ethnicities, or SES groups nor to children who have experienced forms of maltreatment in addition to or other than neglect.
The development of behavior problems among neglected children involves a complex set of interacting mechanisms. As such, it will be important for future research to examine how relationships between neglect, EK, emotion regulation, empathy, parental support, and behavior problems change over time. In examining prospective models, it will also be important to consider the potential for bidirectional relationships between EK and behavior problems, as externalizing and internalizing problems at 18 months have been found to predict EK at 36 months [14, 52]. Likewise, factors such as children’s social cognition, neuroendocrine function, epigenetic processes and gene expression need to be considered when identifying children most at risk for developing behavior problems following neglect [29]. Such research will enhance our understanding of how children exposed to neglect during early childhood develop behavior problems and, subsequently, may help to prevent the onset or worsening of such problems.
Acknowledgements
The authors gratefully acknowledge the assistance of the Philadelphia Department of Human Services and the New Jersey Division of Youth and Family Services.
Declarations
Conflict of interest
The authors have no relevant financial or non-financial interests to disclose.
Ethical Approval
Informed consent was obtained from mothers and assent from adolescents for all study procedures, which were approved by the Institutional Review Boards of Drexel University and the University of Medicine and Dentistry of New Jersey (now Rutgers University).
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.