Background
Children with conduct disorder and oppositional defiant disorder exhibit various problem behaviors. According to the DSM-5 [
1], the former is defined by symptoms including aggression towards animals and people, destructing the property of others, deceitfulness or theft, violations of rules and social norms and the latter by angry irritable mood, argumentative/defiant behavior, and vindictiveness. Children with such conduct problems (CP) form a highly heterogeneous group. For example, it is estimated that almost half of the individuals with CP exhibit high levels of callous-unemotional traits (CU-traits) [
2]. CU-traits are related to the affective component of the concept of adult psychopathy and encompass characteristics, such as (a) a lack of empathy, (b) shallow affect, (c) a lack of remorse or guilt, (d) indifference towards one’s own performance [
3].
Children with CP and high levels of CU-traits (CP–CU) and children with CP and low CU-traits (CP-only) have severe social deficits in common. However, children with CP–CU also differ from children with CP-only on certain aspects. They have a higher heritability of antisocial behavior [
4‐
7], show reduced responsiveness to punishment [
8,
9], as well as a reduced amygdala activity [
10] and reduced startle response [
11,
12] to distressing stimuli. These differences indicate that there are several distinct causal pathways, which can lead to the development of CP with and without CU-traits. Thus, understanding these differences is crucial in order to develop meaningful treatment strategies.
Many studies have looked at the social problems of children with CP and CU-traits in light of Dodge’s social information processing theory [
13]. According to this theory, social information is processed in five steps: attention on and encoding of social cues, interpretation of social cues, response search, response evaluation, and enactment. The authors propose that a failure to regularly process either of these steps may lead to aggressive behavior. Even though, the literature has shown that children with CP show irregularities in all five steps [
13‐
16], the vast majority of papers focuses on the encoding aspects of step one [
17‐
20] and less so on the attentional aspects. However, there exists some evidence, which indicates that children with CP with and without CU-traits might, in fact, show differences in the attentional aspects of social information processing. For example, Hodsoll et al. [
21] used an attentional capture task in which the children were presented with three pictures of faces, two female and one male, tilted to either the left or the right side. The faces depicted neutral or emotional (happy, afraid or sad) expressions. The task was to indicate to which side the target face (the male face) was tilted as quickly as possible. They showed that compared to TD children and children with CP-only, children with CP–CU seemed to be less distracted by task-irrelevant emotional facial expressions. Similarly, the interaction of high CU-traits and high aggression significantly predicted a reduced reaction time facilitation in a dot-probe task [
22] and CU-traits were negatively correlated with facilitation effects in a lexical decision task [
23]. Another paradigma, which was used to study attentional biases is the emotional stroop task. In this task, the participants are presented with neutral and aggression type words in different colors. The participants are instructed to ignore the content of the word and name the color as quickly as possible. The difficulty to disengage from a certain stimulus has been interpreted as an indication that these stimuli are more salient to the concerns of the individual as attentional resources seem to be automatically allocated towards these stimuli [
24,
25]. As of yet, no study used the emotional stroop task to investigate attentional biases in children with CP. However, there exist studies looking at individuals with different levels of trait anger [
24,
26,
27]. They found that individuals with high trait anger compared to individuals with low trait anger were significantly slower in naming the color of aggression type words relative to neutral words. Van Honk et al. [
28] further showed that this attentional bias also holds if angry and neutral faces instead of words are presented.
Notably, children with CP might not only show attentional biases but also attributional biases. Thus, children with CP also differ in step two of the social information processing model. Previous studies have shown that aggressive children are more likely to interpret social situations with ambiguous provocative intent as hostile compared to non-aggressive children [
29‐
32]. This tendency is referred to as the hostile attribution bias. A recent review by Martinelli et al. [
14] found a more consistent relationship between reactive aggression and hostile attribution bias compared to proactive aggression and hostile attribution bias. As reactive aggression is commonly found in CP-only as well as CP–CU children [
33] it can be hypothesized that children with CP-only and children with CP–CU both show an increased hostile attribution bias compared to TD children. However, studies, which directly investigated the hostile attribution bias in children with CP and CU-traits, produced inconsistent results. Frick et al. [
31] reported an increased hostile attribution bias in boys with CP-only compared to all other groups, Cima et al. [
32] observed an increased hostile attribution bias in CP–CU boys compared to CP-only boys, and Helseth et al. did not find any group differences [
15]. As Cima et al. [
32] only investigated boys, Frick [
31] only observed group differences in the male portion of their study sample and Helseth et al. [
15] did not observe any group differences in their mixed gender sample, gender might be a potential confounding factor and explain the discrepancies between these results.
Interestingly, little is known about the association between the hostile attribution bias and the attention bias for emotional cues. It seems intuitive to assume that a stronger attention bias towards hostile stimuli is correlated to more hostile interpretations of ambiguous situations. Additionally, considering that the attention towards a stimulus (step one) precedes the interpretation of a stimulus (step two) within the social information processing model [
13], one could conclude that also the attention bias (step one) and the attribution bias (step two) depend on each other. In their integrative cognitive model, Wilkowski et al. [
34] build on the social information processing model and propose that rather than a stepwise process it is an interaction of cognitive biases which leads to aggressive behavior. They suggest that an automatic hostile interpretation of ambiguous situations directly elicits anger but also leads to prolonged attention towards hostile stimuli, which again amplifies the level of anger and thus increases aggressive behavior. Thus, directly comparing the performance of children with CP and CP–CU on tasks involving attributional and attentional aspects of social information processing cannot only help us to uncover differences in two different steps of social information processing among children with CP with and without CU-traits but further provides us with the unique opportunity to investigate the association between attribution and attention biases. Therefore, in this study, we compared boys and girls with CP-only, CP–CU and typically developing children (TD) on two social information processing tasks. The pictorial emotional stroop task, similar to the task conducted by Van Honk et al. [
28], focused on the attentional aspects of social information processing. With regards to the findings of an association between CU-traits and a missing attentional bias [
21‐
23], we hypothesized that the performance of children with CP–CU, compared to children with CP-only or TD, is less affected by aggressive facial expressions. The objective of the second task was to expand our knowledge regarding the relationship between CP, CU-traits and the hostile attribution bias. Therefore we conducted a hostile attribution bias task similar to Frick et al. [
31]. In line with the findings regarding the association between reactive aggression and the hostile attribution bias [
14] and the findings that children with CP–CU as well as children with CP-only show reactive aggression [
33], we hypothesized that children with CP-only as well as children with CP–CU interpret ambiguous situations significantly more often as hostile compared to TD children. In light of the integrative cognitive model [
34], we further hypothesized that there is a positive correlation between the hostile attribution bias and attentional biases towards hostile stimuli. Since the association between the first two steps of the social information processing model has not been studied in children with CP and CU-traits, in an explorative manner we investigated, if the groups show differences in this association.
Discussion
The purpose of the current study was to examine attributional and attentional biases in social information processing via a pictorial emotional stroop task and a hostile attribution bias task in children with CP–CU, CP-only, and TD. We hypothesized that children with CP-only would show a significantly higher attentional bias compared to the control group. We further expected that children with CP–CU would show significantly less attentional interference compared to the TD group and CP-only group.
In contrast to our expectations, the groups did not significantly differ concerning the attentional bias. Instead, we were able to find a significant attentional bias to negative facial expressions for all participants. Thus, irrespective of the group, negative emotional faces interfered with children’s performance on the actual task. Our findings are not in line with previous findings indicating that children with CP–CU are less affected by attentional biases to emotional stimuli [
21‐
23]. In comparison to Hodsoll et al. [
21], our CP–CU sample consisted of children with overall lower levels of CU-traits (
M = 35.36,
SD = 5.86 vs.
M = 49.5,
SD = 7.7). This might indicate that differences in the attentional bias are only observable in children with very high levels of CU-traits. However, it has to be considered that Hodsoll et al. used the teacher version of the ICU, which generally produces higher scores compared to the here applied parent version [
35]. Furthermore, Kimonis et al. [
22] showed that the interaction of CU-traits and aggression negatively predicts the facilitation to distressing cues in a dot-probe task in a sample of detained individuals with only moderate levels of self-reported CU-traits (
M = 23.23,
SD = 7.85). Interestingly, Loney et al. [
23], only reported indications of an association between a lower attentional bias and CU-traits if they controlled for impulsivity. However, reanalyzing our attentional bias data with impulsivity as covariate, did not change our results in a meaningful way.
Furthermore, our findings are also not compatible with the findings of previous studies, which conducted an emotional stroop task with adults with high and low levels of anger and observed a stronger interference for high anger individuals [
24,
28]. In comparison to these previous studies, we did not differentiate between groups according to the level of trait anger, but based on a clinical diagnosis of CP. This could indicate that only individuals with high levels of trait anger but not individuals with high levels of behavioral aggression show increased difficulties to disengage from negative emotional stimuli. However, this assumption needs to be considered with caution. Even though angry irritable mood is a core symptom of oppositional defiant disorder and trait anger is highly predictive of aggressive behavior [
48,
49], we did not assess trait anger in our sample. Future studies should investigate attentional biases including measures of state and trait anger as well as aggressive behavior in order to determine potential differences.
Regarding our second task, we expected both CP groups to show a higher hostile attribution bias compared to the TD group. Although the overall group effect became only marginally significant, we observed a significantly higher hostile attribution bias in the CP–CU group compared to the TD group. The comparison between the CP-only group and TD group almost approached significance. As both children with CP–CU as well as children with CP-only show increased levels of reactive aggression [
33] our findings of a tendency for a higher hostile attribution bias in both of these groups fit the results of a meta-analysis, showing that reactive aggression is associated with higher levels of the hostile attribution bias [
14]. Nonetheless, our findings contradict the previous finding of a stronger hostile attribution bias in CP-only compared to CP–CU children [
31], the findings of a higher hostile attribution bias in CP–CU compared to CP-only children [
32] and the findings of no group differences at all [
15]. Our study overcame several shortcomings of these previous studies. Due to their small sample size, Helseth et al. [
15] were not able to consider potential gender effects. Cima et al. [
32] similarly failed to consider possible gender confounds as their study sample consisted of exclusively male participants. In line with Frick et al. [
31], we observed an association between gender and the hostile attribution bias, with boys of the CP–CU group showing a significantly higher hostile attribution bias compared to girls. Interestingly, we only observed this gender difference within the CP–CU group but not the CP-only group. These results could indicate that children with CP–CU might benefit from gender-specific interventions.
As expected, our analysis revealed that a higher attention bias towards angry stimuli is associated with a higher hostile attribution bias as well as a higher anger reactivity score. Thus, children who interpreted ambiguous situations as hostile more often also showed higher interference when presented with angry facial expressions. We do not have a ready explanation of the correlation between attentional biases to happy stimuli and the hostile attribution bias. However, this correlation was rather small and only was significant when investigating the whole sample and not in the individual groups. Notably, a significant association between the hostile attribution bias and the attention bias could only be observed for anger stimuli in the CP-only group. Furthermore, this association was significantly stronger within the CP-only group compared to the TD group. Thus, it seems that even though the children with CP-only did not show a higher attentional bias or a higher hostile attribution bias compared to the TD participants, this group of children has an especially strong connection between attributional and attentional biases. This could indicate, that the interaction of these biases is what leads to the aggressive behavior in children with CP-only, as the integrative cognitive model [
34] proposes. As children with CP–CU did not show a significant correlation between the attribution bias and attention bias and did not significantly differ in their association compared to the TD group, it is likely that their aggressive behavior results from a different processing pathway. However, due to the explorative manner of this investigation and the fact that the strength of the association between the CP-only and the CP–CU group did not significantly differ, this interpretation needs to be considered with caution.
Strengths and limitations
The current study is, to our knowledge, the first study to compare children with CP with and without CU-traits on social information processing tasks involving two different cognitive biases. Investigating both attentional and attributional aspects of social information processing can help us in precisely identifying where anomalies in the processing pathway occur and thus provide us with valuable information to guide the development of successful treatments. In contrast to most previous studies, we investigated a mixed study sample and thus took the possible influence of gender into account. Aside from the previously mentioned limitations, the following considerations should be noted. According to post hoc power analysis using G*power [
50], our study sample might be too small to detect significant group differences. As we could only rely on one study to estimate the effect size for the emotional stroop task [
28] we used the reported effect size (f = 0.37) as well as a more conservative estimation (f = 0.25). Similarly, we used a more liberal (f = 0.35) and a more conservative estimation (f = 0.25) for the hostile attribution bias task, as a meta-analysis of the hostile attribution bias [
51] indicated that effect sizes across studies are highly variable and depend on various factors. Thus, with an alpha level of 0.05, our sample size achieved a power between 0.67 and 0.91 to detect group differences in the emotional stroop task and a power between 0.59 and 0.88 to detect group differences in the hostile attribution bias task.
We did not counterbalance the order of the tasks, as we did not expect the tasks to influence each other. All children first performed the emotional stroop task and then the hostile attribution bias task. This order was chosen as the hostile attribution bias task does not require fast reactions and is thus less affected by a possible drop in concentration.
Similarly to most studies including children with CP, we cannot completely rule out that motivational factors may have influenced the results. However, the CP-only, as well as CP–CU children did not differ from the TD children in their general reaction time, in the number of mistakes in the pictorial emotional stroop task and they did not show any answer-patterns in the hostile attribution bias task that would indicate that e.g. they simply pressed the same button to answer each question. We did not assess reactive aggression within our sample, thus it is possible that the CP–CU group and CP-only group slightly differed on their level of reactive aggression. A slightly lower level of reactive aggression in the children with CP-only, for example, could explain why the difference to the TD group in the hostile attribution bias task did not become significant. As we were primarily interested in the investigation of subgroups of children with CP, we did not include a CU-only group in our study. However, future studies should consider including a CU-only group when investigating social information processing in order to assess in how far CU-traits influence social information processing in the absence of CP.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.