Introduction
Aside from body posture and hand gestures, facial expressions are probably one of the most important aspects of non-verbal communication. Properly decoding facial expressions is crucial in everyday situations and especially in potential conflict situations, as they can inform us about the emotions and feelings of other individuals. Deficits in recognizing negative emotions have been repeatedly linked to various problem behaviors of children with conduct disorder (CD) and oppositional defiant disorder (ODD) [
1]. According to the DSM-5 [
2], children with such conduct problems (CP) show externalizing problem behavior, including aggression, deceitfulness, violations of rules and social norms, argumentative/defiant behavior, and vindictiveness. It is estimated that almost half of the individuals with CP exhibit high levels of callous-unemotional traits (CU-traits) [
3]. CU-traits are characterized by a lack of empathy, shallow affect, a lack of remorse or guilt, and indifference towards one’s performance. High levels of CU-traits also occur in about two to five percent of otherwise healthy individuals [
4]. High levels of CU-traits can also be observed in adults with psychopathy, as they build the affective dimension of the construct of psychopathy [
2]. Psychopathy has been linked to deficits in emotion recognition [
1,
5]. It has been hypothesized that a failure to properly decode fear or sadness in others, disrupts the mechanism which normally inhibits violent behavior upon the detection of distress cues in others (violence inhibition mechanism) [
6,
7]. Interestingly, studies that investigated the relationship between emotion recognition deficits, CP and CU-traits in children have produced mixed results [
8‐
19]. It is yet unclear if emotion recognition deficits are linked to CP [
8,
17,
20], CU-traits [
12,
13], the interaction of CU-traits and CP [
8,
9,
13], if they are not linked at all [
10,
16], and if deficits are limited to specific emotions [
9,
20‐
23] or affect all negative emotions [
5,
15].
Factors Influencing Study Results
Various factors might explain the discrepancies among study results. First, one of these factors could be differences in the experimental paradigms. Studies vary in the stimulus presentation duration (from 0.5 s to unlimited viewing time), the stimulus type (static or dynamic) as well as the response type (button press or verbal response). Longer or even unlimited viewing times could lead to speed-accuracy-trade-offs, meaning that children can compensate for deficits in emotion recognition through longer processing times. The assumption that CU-traits might influence the processing time rather than the accuracy is in line with the finding that emotion processing deficits in adults with psychopathic traits are defined by a longer processing time rather than a general failure to recognize emotions [
24]. As of now, there are only two studies that measured reaction times in addition to error rates when investigating emotion recognition in children with CP and CU-traits [
9,
10]. One of these studies does indicate such a trade-off [
9] as participants with CP compared to typically developing (TD) participants showed a significantly lower number of errors for sadness and fear but significantly longer reaction times.
Second, different compositions of the study samples may affect the study outcome. For example, the majority of the studies with children with CP investigated the relationship between CU-traits and emotion recognition deficits in boys [
10,
12,
21‐
23]. Only one study had a purely female sample [
9] and two studies a mixed-gender study sample [
8,
13]. Interestingly, these were also the only studies that reported a better fear recognition in children with high levels of CU-traits instead of a deficit. Even though these studies do not provide any direct information regarding an interaction effect of gender and CU-traits, the contrasting results indicate that gender should be considered as an influencing factor. This is further supported by the finding that females seem to be better in recognizing emotions compared to males in general [
8].
Another aspect that differs among study samples, is their composition regarding CP diagnosis. Some studies only included children with CD [
8] some studies included children with CD as well as ODD [
9] and some children with subclinical levels of CP [
21‐
23]. This and the heterogeneous nature of CP make it difficult to properly compare study results. Thus, instead of comparing children with and without CP, some studies opted for a dimensional approach and assessed the association between externalizing problem behavior and emotion recognition [
15,
21‐
23]. Most of these studies included community samples and thus children with low or subclinical levels of CP [
21‐
23]. They showed that when controlling for the level of externalizing problem behavior, CU-traits themselves predicted sadness and fear recognition deficits. However, a recent study that included children with various externalizing disorders and assessed externalizing problem behavior dimensionally [
15] showed that there is an association between CU-traits and emotion recognition deficits for various emotions (sadness, fear, anger, disgust). This finding is in line with a meta-analysis, which showed that children with high levels of CU-traits show deficits in recognizing sadness, fear, and anger [
5]. Thus, there is reason to assume that the association between CU-traits and emotion recognition deficits goes beyond sadness and fear but affects all negative emotions.
The Role of Attention
It is not only unclear whether externalizing behavior or CU-traits are associated with specific or general emotion recognition deficits, but also the cause of these deficits is unknown. One cause that recent research has focused on is an irregularity in the reactivity of the amygdala [
25‐
27]. It has been hypothesized that similar to patients with amygdala lesions [
28‐
30], emotion recognition deficits in individuals with CU-traits and CP might be due to aberrant attention. As of now, four studies in children provide evidence that CP, as well as CU-traits, might be related to a deficit in attending towards the eyes of emotional faces [
8,
12,
21,
22]. Dadds and colleagues [
22] observed that high levels of CU-traits not only predicted deficits in fear recognition but also a lower number and duration of fixations on the eyes of fearful faces. In a previous study Dadds and colleagues [
21] showed, that in a community sample of boys, CU-traits were negatively correlated with the accuracy in recognizing fearful facial expressions (r = − 0.36). However, once the children were instructed to focus their attention on the eyes of the face, this association disappeared (r = 0.05) and reappeared when the children were instructed to look at the mouth (r = − 0.24). The authors concluded that automatic misdirected attention towards the mouth instead of the eyes, in other words, a decreased eye-preference, can explain the emotion recognition deficit. To date, these results could not be replicated [
8,
12]. Furthermore, the authors did not directly test the association between eye-preference and emotion recognition accuracy. For their conclusion to hold, it needs to be shown that the eye-preference level substantially mediates the relationship between CU-traits and fear recognition deficits.
To our knowledge, there are only two studies that directly investigated these associations in children with CP and CU-traits [
8,
12]. Neither of these studies reported a significant association between CU-traits and fear recognition. However, Billeci and colleagues [
12] found that eye-preference levels significantly mediated the association between CU-traits and sadness recognition deficits if they only analyzed children with CP, but not if they analyzed TD children. In contrast to these findings, Martin-Key et al. (2018) [
8] reported that not CU-traits but CP were related to lower eye-preference levels as well as to higher emotion recognition deficits for fearful and sad facial expressions. Even though they also found that eye-preference levels were a significant predictor for the emotion recognition accuracy in their regression models, the degree to which eye-preference level explained the variance in emotion recognition accuracy was low (
f2 < 0.05). The authors concluded that the predictive power of the level of attention towards the eyes is not sufficient to explain the emotion recognition deficits in children with CP, and thus other mechanisms must be underlying this deficit.
The Present Study
In summary, the investigation of the relationship between CP or CU-traits with emotion recognition deficits has produced somewhat inconsistent results. Some of these inconsistencies might be due to gender differences, CP diagnosis or differences in the applied experimental paradigms. Thus, in the current study, we conducted two different emotion recognition paradigms in boys and girls with various levels of CU-traits and externalizing problem behavior. To ensure sufficient variance in the level of CU-traits and externalizing behavior, our sample consisted of TD children as well as children with a CP diagnosis. In line with previous studies, we analyzed our data dimensionally [
15,
21‐
23]. In the first paradigm, similar to the paradigm employed by Golan and colleagues [
31], children had to recognize a target emotion among three different emotional expressions. To increase participants’ motivation to answer correctly, they received auditory feedback for wrong answers and could only proceed once they chose the correct stimulus. To investigate possible speed-accuracy-trade-offs, we collected reaction times as well as error rates. The second paradigm was similar to previously conducted paradigms in which the children had to categorize the emotion of a single facial expression and were instructed to answer as fast and as correctly as possible [
11‐
13]. During the presentation of the stimuli in this paradigm, we further collected eye-tracking data to investigate if eye-preference levels mediate the association between CU-traits and emotion recognition deficits.
In line with the meta-analytic findings [
5] indicating a unique relationship between CU-traits and deficits in recognizing various emotions, we expected that higher levels of CU-traits would be associated with stronger negative emotion recognition deficits independent of the level of externalizing behavior. However, as some findings also indicated that the combination of externalizing behavior and CU-traits might influence emotion recognition [
8,
9,
13], we further explored the association between the interaction of CU-traits and externalizing behavior and emotion recognition. We expected that any observed associations between CU-traits and emotion recognition deficits would be mediated by the level of attention towards the eyes.
Discussion
With the current study, we tested whether externalizing problem behavior and CU-traits are associated with emotion recognition deficits of negative emotional faces, using two different experimental paradigms. We further investigated whether found associations between emotion recognition deficits and CU-traits would be mediated by the level of attention participants pay to the eyes of an emotional stimulus. One aim of the current study was to take different factors into account which might have influenced previous results. Thus, we used two different experimental paradigms. One paradigm, which included additional measures to increase participants’ motivation to answer correctly and allowed for an analysis of potential trade-off effects between processing time and error rate and another paradigm without such measures. Furthermore, to account for differences in group composition, gender was added as a covariate in the analysis and we included children with and without CP but analyzed the data dimensionally to reduce the influence of the heterogeneity of CP and to account for subclinical levels of externalizing behavior.
We observed different results in the two different emotion recognition paradigms. As expected, CU-traits were associated with increased reaction times to fearful, sad and angry stimuli in the emotion recognition task. Thus, in line with previous studies [
5,
15] CU-traits seem to be associated with emotion processing deficits for various emotions and not just sadness and fear. However, neither in the emotion recognition task nor the categorization task did CU-traits significantly predict the error rates. This could be due to various reasons.
First, the number of mistakes may have decreased in exchange for an increase in reaction time. This assumption would be in line with the findings indicating that the emotion recognition deficit in adults with psychopathic traits lies in a longer processing time rather than a general failure to recognize emotions [
24]. Trade-offs between processing time and error rates could also be one of the reasons for discrepancies among studies with differing stimulus presentation times. Dadds and colleagues [
15] for example presented the stimuli for only 500 ms and, in keeping with our findings, reported a relationship between CU-traits and a recognition deficit for all negative emotions. The participants in the study by Woodworth and Waschbusch [
13] on the other hand, had an unlimited viewing time, opening the possibility for a trade-off between processing time and error rates. Such a trade-off could, for example, explain their finding of better fear recognition in children with high levels of CU-traits. As suggested by Vitale and colleagues [
24], the problem of a slower emotion processing, in contrast to a general problem in recognizing emotions in others, might be susceptible to treatment. Thus, emotion recognition training targeting the processing speed of children with high levels of CU-traits might help them to encompass their insufficiencies.
Second, the tasks might have been too easy to detect a relationship between CU-traits and error rates. Even though previous studies conducted similar tasks with children of the same age [
11‐
13] and reported deficits in emotion recognition, the children of our study made very few mistakes in general. Nonetheless, they made considerably fewer mistakes in the recognition task compared to the categorization task (relative to trial number 36 vs. 10). Furthermore, there was a trend (p = 0.052) for an association between CU-traits and more mistakes in recognizing fear in the emotion categorization task. Thus, the emotion recognition task may have been easier than the emotion categorization task. In general, our results indicate that reaction times might be a more sensitive method to detect differences in the emotion recognition ability than error rates.
Even though there was no significant relationship between error rates and CU-traits in either of the emotion recognition tasks, we observed a significant relationship between the interaction of CU-traits and externalizing problems and the error rate for angry stimuli in the second paradigm, indicating that CU-traits were associated with a higher error rate if the level of externalizing problems was high as well. This result is rather surprising as previous studies have only reported a relationship between having CP and a reduced ability to recognize angry facial expressions but no interaction effects [
8,
14,
20]. Interestingly, there was also an interaction effect of CU-traits and externalizing problems in the first paradigm. However, this interaction indicated that CU-traits were associated with longer reaction times to angry stimuli if the level of externalizing behavior was low. Thus, this could indicate that children with high levels of CU-traits and high levels of externalizing behavior have a general problem with recognizing angry stimuli and children with high levels of CU-traits and low levels of externalizing behavior have deficits in their processing speed. Due to the differences among the paradigms, this assumption needs to be considered with caution. As previously mentioned, the paradigms might have differed regarding their level of difficulty. Also, the emotion recognition task may have been less influenced by motivational aspects. In contrast to the categorization task, wrong choices in the recognition task were followed by a buzzing sound and the task only continued once a correct choice was made. Thus, the children might have been more motivated to make the right choice in the emotion recognition task compared to the categorization task.
In the second part of our study, we investigated whether the found associations between CU-traits and emotion recognition deficits are mediated by the participant’s level of attention on the eyes. Similar to Han and colleagues [
26] and in line with the assumption that the eyes are the most salient feature for the recognition of fear [
52], we observed a better recognition for fear and anger over the whole sample, when only the eyes were presented compared to trials in which only the mouth was visible. Given the finding that the eyes compared to the mouth receive the most attention in sad stimuli [
53] our finding of a higher number of mistakes in the eye-compared to the mouth-condition for sad faces is surprising. In accordance with the finding that the relation between CU-traits and fear recognition deficits seems to decrease, once children are instructed to look at the eyes of fearful faces [
21], we expected a similar result when we presented the children with only the eyes of the fearful stimuli. However, putting the eyes in the focus of attention did not decrease the relationship between CU-traits and a fear recognition deficit. Consistent with this finding, our eye-tracking results do not provide any evidence suggesting that the eye-preference level mediates the emotion recognition deficits in children with high levels of CU-traits. Comparable to Martin-Key and colleagues [
8] we did not observe a significant association between eye-preference and CU-traits for either of the three emotions. Thus, the longer processing time in children with high levels of CU-traits does not seem to be associated with a lack of attention to the eyes of emotional faces. In line with Martin-Key et al. [
8] and previous studies in healthy adults [
54], we also observed significant gender effects with a higher eye-preference level for girls compared to boys. Interestingly, Billeci and colleagues [
12] who only investigated boys, observed a significant mediation of eye-preference on CU-traits and the recognition of sad facial expressions in their CP-group. Thus, it is possible that eye-preference levels only contribute to the emotion recognition deficits in boys with high levels of CU-traits and CP but not in girls. As of yet, all the studies that investigated the relationship of CU-traits, emotion recognition and attention to the eyes used a cross-sectional study design [
8,
12,
21]. Even though, Dadds and colleagues [
21] established the causal relationship of attention to the eyes and emotion recognition deficits by showing that emotion recognition deficits can be reduced through instructing the children to look at the eyes of an emotional facial expression, the causal relationship between CU-traits and attention to the eyes remains unclear. As causal inferences in mediation analysis can only be withdrawn if the temporal ordering of the variables in a mediation model is correct [
48,
55], longitudinal studies would be needed to establish whether CU-traits are a result of or a cause of aberrant attention to the eyes.
Strengths and Limitations
The present study is the first study to investigate emotion recognition deficits using two different paradigms and one of few studies to directly investigate the influence of eye-preference levels on the relationship between CU-traits and emotion recognition deficits in a mixed study sample. Thus, our study not only replicates and extends previous results on the relationship between CU-traits, externalizing problem behavior, emotion recognition, and attention but further delivers valuable information about different aspects that influence the results of emotion recognition tasks, which might explain discrepancies among previous study results.
Including the measurement of reaction times in our emotion recognition task allowed us to ensure appropriate task performance and reduce the influence of impulsive or inattentive behavior. We believe that the assessment of reaction times, the buzzing sound signaling a wrong choice, as well as the trial repetition until the correct choice was made contributed to an increased willingness to properly participate in the task and thus led to a reduction of the influence of motivational aspects.
Aside from these strengths, several limitations need to be mentioned. Six participants with CP took medication on the day of the experiment, however, we did not have sufficient power to compare performance among children with and without medication. Thus, we do not know in how far medication might have influenced children’s performance. Additionally, we cannot tell whether our attempts to increase the children’s motivation to follow task instructions in the emotion recognition task was successful, as we did not have an objective measure of the level of motivation. However, in light of the fact that children with externalizing problems have problems with motivation [
9,
56,
57] and as indifference to one's performance is a symptom characterizing CU-traits, future studies should consider the influence the motivation to follow task instructions has on the study results.
To properly assess eye-movements, we had to use a fixed stimulus presentation time in the emotion categorization task, which is why we could not assess reaction times in this paradigm. The inclusion of the measurement of reaction times in this paradigm would have been important to support our findings of the emotion recognition task indicating that CU-traits are associated with longer processing times rather than deficits in recognizing emotions. Before each trial of the categorization task, we presented a fixation cross to refocus the attention of the participants on the center of the screen. This fixation cross was positioned slightly closer to the mouth AOI than the eye AOI, hence, we had to exclude the first fixation from our analysis and were not able to investigate potential differences in the first initial shift. According to a study in patients with amygdala lesion [
30], emotion recognition deficits may be related to a lack of the first automatic shift towards the eyes instead of differences in the total fixation duration on the eyes. However, others who investigated CU-traits and initial eye-preference did not observe meaningful relations [
8]. Furthermore, due to the use of static facial expressions, our study has a rather low ecological validity. Future studies would benefit from including dynamic facial expressions in the investigation of a potential failure to attend to the eyes.
As indicated by a priori power analyses, the mediation analysis would have required at least 130 participants. As the final analysis only included 73 participants, we cannot exclude the possibility that we were unable to observe a significant mediation effect of attention on the association between CU-traits and emotion recognition due to insufficient power. However, post hoc power analyses showed that with a medium effect size settled at f
2 = 0.15 [
5] and p = 0.05 we achieved a power > 0.8 for all our other analyses.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.