Introduction
Facial expressions represent the primary method to communicate emotions nonverbally. The ability to recognize facial emotion is important for the regulation of interpersonal relationships through understanding the feelings and intentions of others, and to modify behavior adaptively in response (Izard et al., Reference Izard, Fine, Schultz, Mostow, Ackerman and Youngstrom2001; Izard et al., Reference Izard, Woodburn, Finlon, Krauthamer-Ewing, Grossman and Seidenfeld2011). For example, angry facial expressions represent a threat to the viewer and require direct immediate action, while faces that signal distress act as inhibitors of aggression (Blair, Reference Blair2005). The Research Domain Criteria framework recognizes the importance of “reception of facial communication” as a construct within a wider social processes domain that is central for adaptive human social behavior and functioning (see https://www.nimh.nih.gov/research-priorities/rdoc/index.shtml).
Children who demonstrate accurate facial emotion recognition engage in more prosocial behaviors, are liked more by their peers, and show more empathetic responses (Denham, Bassett, Zinsser, & Wyatt, Reference Denham, Bassett, Zinsser and Wyatt2014). Conversely, an impaired ability to identify facial expressions has been linked to maladjustment and implicated in a range of child psychopathology (Collin, Bindra, Raju, Gillberg, & Minnis, Reference Collin, Bindra, Raju, Gillberg and Minnis2013). In particular, behavioral problems – including conduct problems, attention and hyperactivity difficulties, peer problems, and a lack of prosocial behaviors – have been repeatedly linked to impairments recognizing negative facial emotions in later childhood and adolescence (Airdrie, Langley, Thapar, & van Goozen, Reference Airdrie, Langley, Thapar and van Goozen2018; Blair, Colledge, Murray, & Mitchell, Reference Blair, Colledge, Murray and Mitchell2001; Hunnikin, Wells, Ash, & Van Goozen, Reference Hunnikin, Wells, Ash and Van Goozen2020; Kohls et al., Reference Kohls, Baumann, Gundlach, Scharke, Bernhard, Martinelli and Konrad2020; Marsh & Blair, Reference Marsh and Blair2008; van Zonneveld, de Sonneville, Van Goozen, & Swaab, Reference van Zonneveld, de Sonneville, Van Goozen and Swaab2018). Wells, Hunnikin, Ash, and Van Goozen (Reference Wells, Hunnikin, Ash and Van Goozen2019) investigated behavioral problems in children aged 7–11 years old who had been referred for behavioral and/or emotional problems, but had yet no mental health diagnosis, as well as typically developing children. Each child completed a facial emotion recognition task, where they were asked to recognize images of happy, negative (sad, fearful, and angry), and neutral facial expressions. Children with behavioral difficulties scored lower than typically developing children for negative and neutral facial emotion recognition, with no group differences for happy faces. Van Zonneveld et al. (Reference van Zonneveld, de Sonneville, Van Goozen and Swaab2018) similarly reported that children aged 8–12 years at high risk of antisociality performed worse at facial affect recognition compared with typically developing control participants, particularly for sad and fearful faces. This finding suggests that difficulties in emotion recognition associated with behavioral problems are already present in childhood. However, few studies have examined the contribution of wider behavioral problems to impaired facial emotion recognition during early childhood despite this stage representing a key period for the development of emotional understanding, including facial emotion (Chronaki, Hadwin, Garner, Maurage, & Sonuga-Barke, Reference Chronaki, Hadwin, Garner, Maurage and Sonuga-Barke2015; Herba, Landau, Russell, Ecker, & Phillips, Reference Herba, Landau, Russell, Ecker and Phillips2006; Leppänen & Nelson, Reference Leppänen and Nelson2006). Two studies conducted within younger children, specifically preschool-aged children consisting of both typically developing children and those displaying disruptive behavior, found that components of callous–unemotional traits opposed to wider disruptive behavioral difficulties were specifically associated with impaired identification of facial affect, although this pattern varied between a generalized recognition impairment (Kimonis et al., Reference Kimonis, Fanti, Anastassiou-Hadjicharalambous, Mertan, Goulter and Katsimicha2016) and a fear-specific deficit (White et al., Reference White, Briggs-Gowan, Voss, Petitclerc, McCarthy, Blair and Wakschlag2016).
Another factor that is important to the development of children's emotional understanding is their emotional environment. Parents and caregivers influence their child's ability to recognize emotions through their explicit teaching of emotion labels and emotion knowledge, via their own ability to recognize emotions, and by sharing with their children their own beliefs about emotions (Castro, Halberstadt, Lozada, & Craig, Reference Castro, Halberstadt, Lozada and Craig2015; Hunnikin & van Goozen, Reference Hunnikin and van Goozen2019). This suggests that a negative emotional environment may adversely contribute to children's emotional development. The “stress acceleration hypothesis” presents an alternative perspective that a negative emotional environment can lead to selective emotional advantages as an adaptive survival strategy through the premature activation and acceleration of core circuitry important for emotional learning and reactivity (Callaghan & Tottenham, Reference Callaghan and Tottenham2016). This perspective is supported by evidence that early adversity can lead to heightened threat sensitivity in animals (Callaghan & Richardson, Reference Callaghan and Richardson2011, Reference Callaghan and Richardson2012; Moriceau, Shionoya, Jakubs, & Sullivan, Reference Moriceau, Shionoya, Jakubs and Sullivan2009). Conversely, studies have found that children who have been exposed to adverse emotional environments, such as those who have experienced maltreatment, neglect, and/or abuse in childhood, show difficulties recognizing facial emotions compared with matched control children suggesting impaired performance (da Silva Ferreira, Crippa, & de Lima Osório, Reference da Silva Ferreira, Crippa and de Lima Osório2014; Pollak, Cicchetti, Hornung, & Reed, Reference Pollak, Cicchetti, Hornung and Reed2000; Pollak & Kistler, Reference Pollak and Kistler2002; Pollak & Sinha, Reference Pollak and Sinha2002), although these children demonstrated faster detection of angry facial expressions consistent with hypervigilance (Pollak & Kistler, Reference Pollak and Kistler2002; Pollak & Sinha, Reference Pollak and Sinha2002). It is evident that the child's emotional environment plays an important role in the development of their emotional understanding with adverse emotional environments linked with impaired facial emotion recognition.
However, no study to date has explored the role of children's emotional environment and childhood behavioral problems concurrently in relation to facial emotion recognition, despite research suggesting that children with behavioral problems experience more negative family environments (Beck, Daley, Hastings, & Stevenson, Reference Beck, Daley, Hastings and Stevenson2004; Caspi et al., Reference Caspi, Moffitt, Morgan, Rutter, Taylor, Arseneault and Polo-Tomas2004; Hastings, Daley, Burns, & Beck, Reference Hastings, Daley, Burns and Beck2006). In addition, there is evidence that children at risk of psychological problems may only evidence emotion recognition impairments (or evidence greater deficits) when this underlying risk is combined with maladaptive parenting (Kujawa et al., Reference Kujawa, Dougherty, Durbin, Laptook, Torpey and Klein2014). Although Kujawa et al. (Reference Kujawa, Dougherty, Durbin, Laptook, Torpey and Klein2014) investigated children at risk of internalizing difficulties, specifically children of mothers with depression, a similar effect may be present in children at risk of externalizing difficulties and highlights the importance of exploring the role of parenting factors in conjunction with childhood behavioral problems in relation to emotion recognition.
Current study
This study examined variables that influence facial emotion recognition impairments in young children by focusing on the severity of childhood behavioral problems and parental expressed emotion (EE). We were interested to test the explanatory power of these variables for children's facial emotion recognition, given the importance of this ability as a transdiagnostic function. Parental EE is a measure of a family's emotional climate and the nature of family interactions (Jenkins & Karno, Reference Jenkins and Karno1992; Weston, Hawes, & Pasalich, Reference Weston, Hawes and Pasalich2017). Parental EE is defined by increased negative comments, reduced positive comments, a lack of warmth, and greater negativity towards the child and is associated with a range of negative mental and physical outcomes in children (Peris & Miklowitz, Reference Peris and Miklowitz2015; Sher-Censor, Reference Sher-Censor2015). Our study was conducted in a relatively large sample of young children (aged 5–7 years old) who were referred by their schoolteachers for behavioral problems and had three objectives: (a) to examine whether childhood behavioral problems were related to emotion recognition impairments; (b) to investigate the relationship between parental EE and facial emotion recognition, and (c) to test whether there is a combined adverse influence of childhood behavioral problems and parental EE for children's facial emotion recognition. We predicted that children with behavioral difficulties would demonstrate impaired recognition of negative facial expressions, but no deficit for happy facial expressions, consistent with findings previously outlined in older children and adolescents with disruptive behavioral problems. We expected that parental EE would be related to impaired facial emotion recognition, reflecting the importance of the family environment on subsequent facial emotion recognition (da Silva Ferreira et al., Reference da Silva Ferreira, Crippa and de Lima Osório2014). Importantly, following findings within other at-risk child populations (Kujawa et al., Reference Kujawa, Dougherty, Durbin, Laptook, Torpey and Klein2014), we predicted that there would be a combined interactive effect of child behavioral problems and parental EE for explaining impaired negative facial emotion recognition.
Method
Participants
One hundred and thirty-five children (44 girls) aged 60–95 months (M = 77.28 months, SD = 9.68) were referred to the Neurodevelopment Assessment Unit (NDAU; https://neurodevelopment.cardiff.ac.uk/) at Cardiff University by their schools for socio-emotional and/or behavioral difficulties. The NDAU is an assessment service available to schools with concerns or questions about a pupil's functioning and the sample therefore demonstrated a heterogenous range of difficulties, including children with low through to high levels of emotional and/or behavioral problems. Written informed consent was obtained from a parent or caregiver for each child. Each child and their parent/caregiver attended two assessment sessions for approximately 2–3 hr each, where the child completed a range of tasks. All experimental procedures were approved by Cardiff University (EC.16.10.11.4592GR). Socio-economic status (SES) was assessed using the Welsh Index of Multiple Deprivation (WIMD), which is a ranked measure of relative deprivation (from 1–1,909 where lower numbers indicate greater deprivation) for areas within Wales based on several indices including income, employment, health, and education (Welsh Government, 2019). There was a large range of WIMD ranks within the sample (range: 12–1,902) with a sample mean of 909.28 (SD = 616.34). We note that we were unable to establish a WIMD rank for seven participants as they lived outside of Wales and therefore their address was not covered by the WIMD. Each child's verbal IQ was assessed using the Lucid Ability assessment (Version 5.15; GL Assessment, 2014). There are two tasks that assess verbal ability dependent on the child's age: Younger children (aged under 7 years old) were assessed using a picture vocabulary task where they identified a picture that matched a word that was read aloud to them; older children (aged 7 years) were assessed using a conceptual similarities task where they identified a word that linked two images. Children received an age-standardized score and the sample data indicated a range of abilities within the sample (range: 64–190) with a mean of 105.95 (SD = 14.56).
Strengths and difficulties questionnaire
The child's teacher completed the Strengths and Difficulties Questionnaire (SDQ; Goodman, Reference Goodman1997) prior to the child's assessment session. The teacher-informant version of the SDQ was used, as the children were referred to the current study by the teachers for difficulties at school and teachers (compared with parents) are generally considered to be more accurate reporters of children's behavioral difficulties (Goodman, Ford, Simmons, Gatward, & Meltzer, Reference Goodman, Ford, Simmons, Gatward and Meltzer2000; Verhulst, Koot, & Van der Ende, Reference Verhulst, Koot and Van der Ende1994). The SDQ is a 25-item screening tool to assess the child's functioning across emotional problems, conduct problems, hyperactivity/inattention, and peer relationships problems, as well as examining prosocial behaviors. Missing SDQ-item scores were calculated based on the mean scores for the remaining items and rounded to the nearest whole number. Each subscale demonstrated acceptable internal consistency (Cronbach's α: emotional problems = .74, conduct problems = .73, hyperactivity = .80, peer problems = .67, prosocial behaviour = .82).
We observed a high degree of correlation between the conduct problems, hyperactivity, peer problems, and reversed prosocial behavior subscales in the current sample, consistent with their conceptual overlap (Caspi et al., Reference Caspi, Houts, Belsky, Goldman-Mellor, Harrington, Israel and Poulton2014; Martel et al., Reference Martel, Pan, Hoffmann, Gadelha, do Rosário, Mari and Bressan2017; Parker, Rubin, Erath, Wojslawowicz, & Buskirk, Reference Parker, Rubin, Erath, Wojslawowicz, Buskirk, Cicchetti and Cohen2006). By contrast, the emotional problems subscale showed little association with each of the other subscales (see Table 1) and the magnitude of these correlations were statistically smaller than the overlapping relationships between the conduct problems, hyperactivity, peer problems, and reversed prosocial behavior subscales (ps < .027) using Dunn and Clark (Reference Dunn and Clark1969)'s z as recommended by Hittner, May, and Silver (Reference Hittner, May and Silver2003). This statistical analysis was conducted using Cocor (Diedenhofen & Musch, Reference Diedenhofen and Musch2015) in R Studio (version 1.2.1335). We therefore defined a “behavioral problems” dimension for the children as the composite summed score across the conduct problems, hyperactivity, peer problems, and reversed prosocial behavior subscales following Burley & van Goozen (Reference Burley and van Goozen2020). The behavioral problems dimension was used throughout our analyses to index child behavioral difficulties. The distinction across SDQ subscales was again confirmed using a principal components factor analysis with orthogonal varimax rotation where two factors were extracted (eigenvalues > 1.0) explaining 69.02% of the variance. The conduct problems, hyperactivity, peer problems, and prosocial behaviors (reversed) loaded onto the first factor (factor loadings ranged from .71 to .85) – the “behavioral problems” dimension – while emotional problems alone loaded onto a separate second factor (factor loading .97) and did not contribute to “behavioral problems”.
Table 1. Pearson's correlations (r) between subscales of the teacher-rated strengths and difficulties questionnaire (SDQ)
* p < .05, **p < .01, ***p < .001
N = 135
Five-minute speech sample
Parental EE was assessed using the 5-minute speech sample (FMSS; Magaña et al., Reference Magaña, Goldstein, Karno, Miklowitz, Jenkins and Falloon1986). During this task, parents (126 mothers, eight fathers, one grandmother as the primary caregiver) were asked to describe their child's personality and their relationship for 5 minutes. Parents were not interrupted unless they were struggling to talk, in which instance they were given semistructured probes by the researcher such as, “how would you describe X's personality?” Each parent speech sample was recorded and later transcribed. Independent coders rated speech based on the content of the sample and the tone of voice used based on the guidelines described in Caspi et al. (Reference Caspi, Moffitt, Morgan, Rutter, Taylor, Arseneault and Polo-Tomas2004). Scores for parental EE were derived on four components: negative comments and positive comments, respectively made by the parent about the child (both coded as a frequency count); warmth of the parent when describing their child as a rating from 0 (no warmth) to 5 (high warmth); negativity of the parent when describing their child as a rating from 0 (no negativity) to 5 (high negativity and hostile). Two separate coders rated the data samples (Rater 1 completed 54.01% and Rater 2 completed 45.90% of the samples). Interrater reliability was found to be high (Cronbach's α: negative comments = .96, positive comments = .96, warmth = .91, negativity = .93) across a subset of 14 samples (10.37% of total samples). There was no difference across any of the FMSS measures dependent on whether the task was completed by the child's mother or father (ps > .628).
Facial emotion recognition task
Emotion recognition was examined using the facial emotion recognition task (Hunnikin et al., Reference Hunnikin, Wells, Ash and Van Goozen2020; Wells et al., Reference Wells, Hunnikin, Ash and Van Goozen2019) consisting of 40 faces chosen from the Radboud Faces Database (Langner et al., Reference Langner, Dotsch, Bijlstra, Wigboldus, Hawk and van Knippenberg2010) depicting happy, sad, fearful, angry, and neutral facial expressions. Varying intensity versions of each affective expression were created by merging the target expression (100%) with a neutral expression (0%) to create expressions that varied between 90–35% and were validated by independent raters (Hunnikin et al., Reference Hunnikin, Wells, Ash and Van Goozen2020). Each facial expression was presented for 3 s before the emotion category labels appeared and the child was asked to identify the facial expression. The participants had no time limit to identify the facial emotion. The presentation order of facial expressions was pseudorandomized across two task versions. The task was presented on a laptop, although six children who had difficulties sitting still during the task completed the task on paperFootnote 1. Recognition accuracy was determined as a percentage of the number of correct answers provided for each facial expression valence.
Data analysis
A repeated measures analysis of variance was run to examine recognition across facial emotion (happy, neutral, sad, fearful, and angry), which was followed up by pairwise contrasts to examine specific comparisons between facial expressions. Emotion recognition accuracy for negative facial expression was calculated by taking the mean score of sad, fearful, and angry faces. We ran Pearson's correlational analysis to explore the relationship between SDQ “behavioral problems”, as well as each subscale of the SDQ, and facial emotion recognition. We similarly explored the association between each dimension of parental EE and facial emotion recognition. Stepwise regression analyses were used to examine whether there was a combined interactive effect of child behavioral problems and parental EE for explaining facial emotion recognition. All analyses were run in IBM SPSS Statistics (version 25).
Facial emotion recognition for each expression did not differ by participant gender (ps > .231) and was not associated with family SES, as measured via WIMD (ps > .307), and therefore neither gender nor family SES was considered further in analysis. Verbal IQ was positively associated with recognition of happy faces, r(133) = .25, p = .003, but there were no relationships with neutral, sad, fear, or angry faces (ps > .120). Participant age was related with recognition of neutral facial expressions, r(135) = .26, p = .003, and showed a trending relationship for happy faces, r(135) = .15, p = .086, but was not related to recognition of sad, fear, or angry expressions (ps > .195). The pattern of findings did not differ when verbal IQ or participant age were entered as covariates and so results are presented without controlling for either variable to maintain greater statistical power. An alpha level of p < .05 was applied to determine statistical significance.
Results
A repeated measures analysis of variance revealed that there was a main effect of emotion, F (3.66, 390.80) = 45.09, p < .001, $\eta _p^2 = .25$
. As can be seen in Table 2, happy facial expressions were recognized most accurately, followed by neutral, angry, sad, and fearful facial expressions; pairwise comparisons showed these differences to be significant (ps < .05) apart from the difference in recognition accuracy between angry and sad facial expressions (p = .811).
Table 2. Mean scores and standard deviation across the facial emotion recognition (FER) task, the 5-minute speech sample (FMSS), and the subscales of the strengths and difficulties questionnaire (SDQ)
† High risk for the SDQ subscales was defined as scores within the “high” or “very high” categorization and “low” or “very low” for the SDQ prosocial subscales (Meltzer et al., Reference Meltzer, Gatward, Goodman and Ford2003)
FMSS negative comments and FMSS positive comments are scored as a frequency count; FMSS parental warmth and FMSS parental negativity scale are scored across a range of 0–5; SDQ subscales are scored across a range of 0–10 and SDQ total difficulties are scored across a range of 0–40
Behavioral problems and facial emotion recognition
Table 2 presents participant means scores across the SDQ subscales and the percentage of children within the high and very high problem categorization bands based on UK normed data (Meltzer, Gatward, Goodman, & Ford, Reference Meltzer, Gatward, Goodman and Ford2003). The SDQ scores indicate a high overall level of difficulties within the sample across each subscale and for total difficulties: 63% of the sample showed total difficulty scores in the high or very high problem range, although the sample demonstrated a range of scores across each subscale.
We explored the relationship between the “behavioral problems” composite dimension (combined scores across the conduct problems, hyperactivity, peer problems, and reversed prosocial behavior SDQ subscales), as well as each of the teacher-rated SDQ subscales, and children's facial emotion recognition (see Table 3). Facial emotion recognition was examined for negative (combined scores across sad, fearful, and angry expressions), happy, and neutral expressions. Behavioral problems were inversely related to the recognition of negative faces, as well as specifically to the identification of sad expressions, with trending relationships for fearful and angry faces (ps < .10). At the subscale level, increased hyperactivity and decreased prosocial behavior SDQ scores were associated with impaired recognition of negative facial expressions, and specifically reduced identification of sad and fearful facial emotions. Increased SDQ peer problems were also associated with reduced recognition of sad facial expressions. No associations emerged for the recognition of happy facial expressions and only SDQ emotional problems were associated with greater recognition of neutral faces. Overall, these results highlight that behavioral difficulties in young children were related to impaired recognition of negative facial expressions, particularly the identification of sad faces.
Table 3. Pearson's correlations (r) assessing the relationship between teacher-rated strengths and difficulties questionnaire (SDQ), parental expressed emotion from the 5-minute speech sample (FMSS) and facial emotion recognition (FER)
Note: “Behavioral problems” represents a composite dimension based on combined scores across each of the conduct problems, hyperactivity, peer problems, and reversed prosocial behavior SDQ subscales
* p < .05, **p < .01, ***p < .001
Behavioral problems was not related to a bias for reporting a particular facial expression overall (ps > .111) or alternatively in response to neutral facial expression (ps > .462).
Parental EE and facial emotion recognition
Mean parental EE scores are presented in Table 2. We examined the relationship between parental EE and children's facial emotion recognition (see Table 3). Increased parental negative comments and negativity were related to children's reduced recognition of negative facial expressions, and specifically lower recognition of sad and angry facial expressions. Parental warmth was related to better recognition of negative facial emotions and specifically to increased recognition of sad facial expressions. No associations emerged for the recognition of happy or neutral faces. These results highlight that negative parental EE was related to impaired recognition of negative facial expressions, with specific associations for sad and angry facial expressions, whereas parental warmth was associated with improved recognition of sad facial expressions.
No associations emerged between parental EE and SDQ scores (ps > .115), apart from a positive correlation between parental warmth and children's prosocial behavior, r(135) = .27, p = .002, suggesting that parents were not more negative about children rated as having more severe difficulties or challenging behavior.
The interaction between behavioral problems and parental EE on facial emotion recognition
We conducted multiple stepwise regression analyses to examine whether there was an interactive effect of child behavioral problems and parental EE predicting children's recognition of negative facial emotion. We conducted this analysis for the dimensions of parental EE (negative comments, warmth, negativity) that were previously associated with impaired recognition of negative facial expressions. This analysis focused specifically on the recognition of negative facial expressions (opposed to happy or neutral faces) given that the previous associations identified were specific to the recognition of negative facesFootnote 2. All variables were standardized within this analysis.
We tested three stepwise regression models explaining negative facial emotion recognition, with behavioral Problems and the dimension of parental EE (negative comments, warmth, negativity) entered as predictor variables at the first step, with their interaction entered at the second step. The three regression models accounted for a significant amount of the variance for negative facial emotion recognition with 9% of the variance explained at Step 1 for each model and 10% to 12% explained when their interaction term was included at Step 2 (see Tables 4–6).
Table 4. Multiple linear regression predicting children's negative facial emotion recognition from children's behavioral problems based on the teacher-rated strengths and difficulties questionnaire (SDQ) and parental negative comments from the 5-minute speech sample
Note: The overall model was significant at each step: Step 1, (R2 = .09), F (2, 132) = 6.72, p = .002; Step 2, (R2 = .12), F (3, 131) = 5.94, p = .001
The interaction term accounted for a significant increase in the proportion of variance explained, ΔR2 = .03, ΔF (1, 131) = 4.07, p < .05
“Behavioral problems” represents a composite dimension based on combined scores across each of the conduct problems, hyperactivity, peer problems, and reversed prosocial behavior SDQ subscales
Table 5. Multiple linear regression predicting children's negative facial emotion recognition from children's behavioral problems based on the teacher-rated strengths and difficulties questionnaire (SDQ) and parental warmth from the 5-minute speech sample
Note: The overall model was significant at each step: Step 1, (R2 = .09), F (2, 132) = 6.28, p = .002; Step 2, (R2 = .11), F (3, 131) = 5.29, p = .002
The interaction term accounted for a significant increase in the proportion of variance explained, ΔR 2 = .02, ΔF (1, 131) = 3.10, p = .08
“Behavioral problems” represents a composite dimension based on combined scores across each of the conduct problems, hyperactivity, peer problems, and reversed prosocial behavior SDQ subscales
Table 6. Multiple linear regression predicting children's negative facial emotion recognition from children's behavioral problems based on the teacher-rated strengths and difficulties questionnaire (SDQ) and parental negativity from the 5-minute speech sample
Note: The overall model was significant at each step: Step 1, (R2 = .09), F (2, 132) = 6.47, p = .002; Step 2, (R2 = .10), F (3, 131) = 4.98, p = .003
The interaction term accounted for a significant increase in the proportion of variance explained, ΔR2 = .01, ΔF (1, 131) = 1.91, p = .17
“Behavioral problems” represents a composite dimension based on combined scores across each of the conduct problems, hyperactivity, peer problems, and reversed prosocial behavior SDQ subscales
Negative comments
Behavioral problems and parental negative comments were both unique inverse predictors of negative facial emotion across each step of the regression model (see Table 4). There was a significant interaction between behavioral problems and parental negative comments for negative facial emotion recognition. The inclusion of this interaction term accounted for a significant 3% increase in the proportion of explained variance for negative facial emotion recognition above both variables independently. To explore this interaction further, the sample was divided into three groups based on the number of parental negative comments (high number of negative comments, mean = 4.78, n = 37; medium, mean = 2.45, n = 51; low, mean = 0.57, n = 47). Behavioral problems were related to reduced recognition of negative facial expression for the high parental negative comments group, r(37) = −.43, p = .008, but not within the medium, r(51) = −.22, p = .13, or low parental negative comments group, r(47) = −.12, p = .44 (see Figure 1).
Figure 1. The relationship between child behavioral problems and recognition of negative facial expressions for children of parents who expressed a high, medium, or low number of negative comments about them. * indicates a significant relationship, p < .05.
We examined the interaction between behavioral problems and parental negative comments by conducting similar regression analysis for each negative emotion (sad, fearful, angry). We found that the interaction between behavioral problems and parental negative comments was specific to the recognition of sad facial expression, t (131) = −2.76, β = −.23, p = .007, with this interaction term accounting for a significant 5% increase in the explained variance, ΔR2 = .05, ΔF (1, 131) = 7.64, p = .007. When we examined our high, medium, and low parental negative comments groups, behavioral problems were related to reduced recognition of sad facial expression for the high parental negative comments group, r(37) = −.60, p < .001, but not within the medium, r(51) = −.25, p = .08, or low parental negative comments groups, r(47) = −.08, p = .61 (see Figure 2).The interaction between behavioral problems and parental negative comments was not predictive of the recognition of either fearful, t (131) = −0.87, β = −.08, p = .385, or angry facial expressions, t (131) = −1.15, β = −.10, p = .254.
Figure 2. The relationship between child behavioral problems and recognition of sad facial expressions for children of parents who expressed a high, medium, or low number of negative comments about them. * indicates a significant relationship, p < .05.
Warmth
Behavioral problems and parental warmth were unique predictors (inversely for behavioral problems and positively for warmth) of negative facial emotion recognition across each step, although warmth was only predictive at a trend level at Step 1 (see Table 5). The interaction between behavioral problems and parental warmth was trending but did not reach statistical significance for predicting recognition of negative facial emotion.
Negativity
Behavioral problems and parental negativity were unique inverse predictors of negative facial emotion recognition, although negativity was only trending once the interaction term between these two variables was included (see Table 6). There was no interaction between behavioral Problems and parental negativity for predicting negative facial emotion recognition.
Discussion
Facial emotion recognition is an important underlying process linked to effective interpersonal functioning, while impairments are associated with a range of neurodevelopmental difficulties. The current study investigated whether severity of behavioral problems in young children (aged 5–7 years) and parental EE were related to impaired facial emotion recognition. In line with expectations, child behavioral problems were related to impaired recognition of negative facial expressions consisting of sad, fearful, and angry expressions. Parental EE, as measured by negative comments, lower warmth, and negativity, was also associated with impaired recognition of negative facial expressions as predicted. In addition, there was a specific interaction between child behavioral problems and parental negative comments in explaining impaired negative facial emotion recognition (and specifically sad faces). That is, the relationship between behavioral problems and reduced recognition of negative faces was strongest when combined with adverse parental EE as measured by high parental negative comments. These results highlight the importance of a multidimensional approach to understand children's emotion recognition and provide further insights into variables that are crucial for understanding children's impairments in facial emotion recognition. Neither child behavioral problems nor parental EE were related to the recognition of happy or neutral faces.
The current study is the first to demonstrate that behavioral problems in young children aged 5–7 years old are associated with impaired recognition of negative facial expressions, consistent with research in older children and adolescents at risk of antisocial behavior (van Zonneveld et al., Reference van Zonneveld, de Sonneville, Van Goozen and Swaab2018; Wells et al., Reference Wells, Hunnikin, Ash and Van Goozen2019). Facial emotion processing develops throughout preschool, middle childhood, and reaches near-adult accuracy by age 11 years (Chronaki et al., Reference Chronaki, Hadwin, Garner, Maurage and Sonuga-Barke2015; Herba et al., Reference Herba, Landau, Russell, Ecker and Phillips2006) and the current study indicates that impairments in recognizing facial emotions have already emerged by age 7 years for children with behavioral problems, highlighting the importance of early intervention. We note that this deficit was specific for identifying negative facial expressions, especially sad faces (and fear to a lesser extent), rather than happy or neutral faces. This finding within young children is consistent with the integrated emotion systems model (Blair, Reference Blair2005) that argues that individuals displaying antisocial behavior demonstrate an insensitivity to distress cues. This insensitivity is proposed to impair the ability of the individual to use distress cues as an aversive social reference to learn to avoid aggressive behaviors that cause others harm. The integrated emotion systems theory also indicates that these individuals are impaired in the formation of stimulus–punishment associations/reduced representations of aversive stimuli meaning that they do not create sufficient associations between a victim's distress, their own affect, and their behavior. These processes lead to the development and persistence of negative disruptive behavior (Hunnikin & van Goozen, Reference Hunnikin and van Goozen2019). The extent to which antisocial development is caused by impaired formation of stimulus–reward associations/reduced representation of positive stimuli is less clear, but the current findings suggests that the adverse effects of behavioral problems within young children are limited to specific recognition impairments for negative facial expressions.
High parental EE is another risk factor for a range of childhood psychopathology and negative outcomes (Peris & Miklowitz, Reference Peris and Miklowitz2015; Sher-Censor, Reference Sher-Censor2015). The current study is the first to demonstrate the association between parental EE – specifically negative comments, lack of warmth, and negativity – and facial emotion recognition, supporting the idea that the family emotional environment is important for facial emotion recognition in young children (Castro et al., Reference Castro, Halberstadt, Lozada and Craig2015; da Silva Ferreira et al., Reference da Silva Ferreira, Crippa and de Lima Osório2014; Pollak et al., Reference Pollak, Cicchetti, Hornung and Reed2000; Pollak & Kistler, Reference Pollak and Kistler2002; Pollak & Sinha, Reference Pollak and Sinha2002). A negative parental emotional climate may fail to provide children with an “emotional teacher” who has the ability or belief system to help them to understand emotions (Castro et al., Reference Castro, Halberstadt, Lozada and Craig2015). We note that parental EE was not associated with a selective advantage for negative facial expressions, including fearful faces, in contrast to the “stress acceleration hypothesis”. This may reflect that we measured overall recognition rates for facial emotions, which may be less sensitive to subtle differences in vigilance than measures such as time to recognition or the amount of sensory information required to identify facial expressions that have previously been indexed to identify the relationship between child adversity and hypervigilance (Pollak & Kistler, Reference Pollak and Kistler2002; Pollak & Sinha, Reference Pollak and Sinha2002).
Importantly, we found an interactive relationship between childhood behavioral problems and parental negative comments in explaining impaired negative facial emotion recognition (and specifically sad facial expressions). The severity of childhood behavioral problems was increasingly associated with reduced recognition of negative facial emotions when the frequency of parental negative comments was high, but the same relationship was not observed (or at least weaker) when the number of parental negative comments was lower. This highlights that it is the combination of both childhood behavioral problems and high parental negative comments that is associated with the greatest risk for impaired negative facial expression recognition within children. The same interaction was not observed for parental warmth or negativity, indicating that the quantity of negative comments that a parent makes about their child may be a more sensitive indicator of the influence of parental EE in their child's emotional functioning.
The current study has identified the associative risk of childhood behavioral problems and parental EE – specifically parental negative comments – in the reduced recognition of negative facial expressions and, in particular, the recognition of sad faces. Learning to cope with stress is an important aspect of healthy child development that leads to the development of an adaptive stress response system. However, toxic stressors that are strong, frequent, or involve prolonged activation of the body's stress response system during early, sensitive periods can disrupt neural development and may lead to enduring alterations to brain architecture (National Scientific Council on the Developing Child, 2005/2014; Sapolsky, Reference Sapolsky2000). Certain individuals may have a greater underlying vulnerability to the deleterious effects of toxic stressors, reflecting an elevated risk that adverse early environments lead to enduring neurobiological difficulties (National Scientific Council on the Developing Child, 2005/2014, 2010). Peris and Miklowitz (Reference Peris and Miklowitz2015) proposed parental EE as a form of chronic toxic family stressor that interacts with the child's vulnerability to psychopathology, such as those at risk for early behavioral problems, that exacerbates emerging symptoms and interferes with the child's neural development. This disrupted neural development could lead to abnormal functioning of underlying processes, such as impaired emotional processing. This interpretation could perhaps be extended to the present study in so far as certain children (i.e., those with greater severity of behavioral problems) may be more vulnerable to the adverse effects of a negative emotional environment contributing to impaired facial emotion recognition, as found by Kujawa et al. (Reference Kujawa, Dougherty, Durbin, Laptook, Torpey and Klein2014). Future longitudinal research can establish the temporal order of these processes that lead to impaired facial emotion recognition.
Implications
The current study has increased our understanding of the underlying multidimensional processes that contribute to facial emotion recognition impairments in young children; this is important as deficient emotion identification is associated with a range of negative developmental outcomes (Collin et al., Reference Collin, Bindra, Raju, Gillberg and Minnis2013). Our findings specifically suggest that future studies interested in understanding facial emotion recognition impairments should account for the interaction between child behavioral problems and parental EE rather than exploring these factors independently. Given that emotion recognition impairments are already evident at a young age, this supports the importance of targeted interventions for children at risk of behavioral difficulties and early in development when brain processes underlying socio-emotional functioning are not yet fully matured (Leppänen & Nelson, Reference Leppänen and Nelson2006). Interventions, such as the Cardiff Emotion Recognition Training (Hunnikin & van Goozen, Reference Hunnikin and van Goozen2019; Hunnikin, Wells, Ash, & Van Goozen, Reference Hunnikin, Wells, Ash and van Goozen2021), have been developed that not only target and improve facial emotion recognition through addressing the underlying neuropsychological impairments (i.e., by directing attention towards key facial features, and encouraging empathy and facial mimicry) (Hunnikin, Wells, Ash, & Van Goozen, Reference Hunnikin, Wells, Ash and van Goozen2021), but also improve behavior (Hubble, Bowen, Moore, & Van Goozen, Reference Hubble, Bowen, Moore and van Goozen2015; Penton-Voak et al., Reference Penton-Voak, Thomas, Gage, McMurran, McDonald and Munafò2013) and mental health ratings (Wells, Hunnikin, Ash, & van Goozen, Reference Wells, Hunnikin, Ash and Van Goozen2020) . It is important that these training interventions are now applied to younger children with behavioral problems at a time when emotional processing is developing and may be more responsive to intervention (Herba et al., Reference Herba, Landau, Russell, Ecker and Phillips2006). In addition, given the current study's findings regarding the role of parental EE in shaping emotion recognition in children with behavioral problems, it may be beneficial to concurrently target both children's and parent's emotion recognition abilities. This concurrent approach to improving facial emotion recognition is supported by studies that highlight direct links between parent and child emotion recognition abilities (Castro et al., Reference Castro, Halberstadt, Lozada and Craig2015).
Limitations
The identification of happy facial expression was high and a lack of findings in relation to happy faces may be explained by a ceiling effect. However, we also observed no effect for less intense and therefore more difficult to recognize happy facial expressions, so this explanation seems unlikely. There was similarly no relationship between child behavioral problems and recognition of neutral facial expression, and no biased pattern of responding to neutral faces including no evidence of a “hostility bias” (Dodge, Reference Dodge1980), which was somewhat unexpected but in line with other studies using similar measures (Airdrie et al., Reference Airdrie, Langley, Thapar and van Goozen2018; Hunnikin et al., Reference Hunnikin, Wells, Ash and Van Goozen2020). We also note that the children in the current study were categorizing facial expressions from static images, which are less arousing compared with dynamic stimuli, and therefore considered to be less naturalistic (Alves, Reference Alves2013; Burley, Gray, & Snowden, Reference Burley, Gray and Snowden2017). It could be speculated that the same impairments would not be observed in response to dynamic stimuli for children with behavioral difficulties, although equally it could be argued that associated deficits could be underestimated using static images. Evidence in adolescents suggests similar emotion recognition deficits in relation to behavioral difficulties regardless of the use of static or dynamic stimuli (Martin-Key, Graf, Adams, & Fairchild, Reference Martin-Key, Graf, Adams and Fairchild2018). We note that we found no relationship between parental EE and children's behavioral difficulties as rated by their teachers, other than parental warmth being related to fewer prosocial problems. This lack of relationship contrasts with Caspi et al. (Reference Caspi, Moffitt, Morgan, Rutter, Taylor, Arseneault and Polo-Tomas2004) who reported that parental EE was related to children's behavioral problems, although associations based on teacher ratings were lower than when assessed using parent ratings. It may be that the current study did not detect this small relationship owing to limited participant size compared with Caspi et al. (Reference Caspi, Moffitt, Morgan, Rutter, Taylor, Arseneault and Polo-Tomas2004). Finally, the current study did not examine the children's recognition of surprised and disgusted facial expressions and so no conclusions can be drawn in relation to these emotions.
Conclusion
The current study found that both childhood behavioral difficulties and parental EE were associated with impairments in recognizing negative facial affect by age 7 years, highlighting the importance of early intervention. Importantly, we identified an interactive effect between childhood behavioral problems and parental negative comments on negative facial expression recognition, and in particular sad faces, which suggests that the combination of both factors was associated with the greatest risk for impaired recognition of negative facial emotion. This finding may indicate that children with behavioral problems are more susceptible to the adverse effects of a negative emotional environment at an important time of socio-emotional development, specifically when developing the ability to recognize and understand the emotions of others. This study improves our understanding of crucial multidimensional factors that offer explanatory power for understanding impairments in facial emotion recognition within young children.
Acknowledgments
We are extremely grateful to the children, families, and schools that took part in our research. We would like to thank Dr Rosanna Stenner, Dr Angie Wigford, Dr Owen Barry, Holly Howe-Davies, Claire Bowsher-Murray, Kate Anning, Katie Larkins, Eleri Jones, Louise Page, Lowri Adams, Judith Ogunkoya, Ffion Williams, George Metaxa, Jessica Brown, Lucie Mathison, and Kirsty Firman for their assistance with this study and their help with recruitment.
Funding Statement
This work was supported by The Waterloo Foundation (grant number 511633) awarded to SHMvG.
Conflicts of Interest
None.
