Background
The ability to understand other people’s emotional and other mental states underlies social skills and is a key process in the development of empathy [
1]. The ability to discriminate emotions starts during the first year of life. Infants as young as 10 weeks of age respond differentially to their carer’s emotional states, expressed in both the face and voice [
2]. By 7 months, infants detect incongruence between facial and vocal expressions of emotions [
3]. During their second and third years of life, children start using mental state words in their speech [
4]. Throughout childhood, the accuracy and speed of emotion recognition (ER) improve [
5], children’s emotional vocabulary expands, and they are able to recognize more subtle mental states [
6]. Emotion and mental state recognition skills continue to develop into adolescence and adulthood.
Emotion and mental state recognition are core difficulties in autism spectrum conditions (ASC) [
7-
9]. Most ER studies carried out with individuals with ASC have focused on the recognition of six emotions (happiness, sadness, fear, anger, surprise and disgust). These so-called ‘basic’ emotions are expressed and recognized cross-culturally [
10] and are to some extent neurologically distinct [
11], though it should be noted that the number of emotions that are recognized cross-culturally may exceed six [
12]. In ASC, some studies report difficulties in recognition of basic emotions [
13-
16]. Other studies, however, have found no difficulties in recognition of the basic emotions in children with ASC [
17-
20]. In contrast, studies investigating recognition of
complex emotions and other mental states by children with ASC have shown more conclusive results. Generally, complex emotions involve attributing a cognitive state as well as an emotion and are more context and culture dependent [
11]. They may be belief- rather than situation-based emotions [
21], for example,
disappointed. They may also be self-conscious emotions, for example,
proud or
embarrassed [
22]. Typically developing children start recognizing and verbally labelling complex emotions like embarrassment, pride and jealousy by the age of 7 [
21,
23]. Studies report deficits in complex ER in individuals with ASC on various tasks, including ER from pictures of the eyes [
24], from facial expressions [
25], from linguistic contextual cues [
26,
27] and from holistic, multimodal scenes [
28,
29]. These studies suggest that children with ASC, although initially delayed in the development of
basic ER skills, may achieve this developmental milestone during their school years or successfully compensate for their basic ER difficulties through explicit cognitive, language-based or perceptual mechanisms [
30]. An assessment of ER difficulties in children with ASC therefore needs to address more complex mental states. The current study focuses on recognition of complex emotions to fill a gap in the existing literature and to provide a new test of complex ER using dynamic stimuli.
Among adults with ASC, there is growing evidence for difficulties in the recognition of complex emotions or subtle versions of basic emotions [
31-
34]. However, as mentioned above, there are not many complex ER tasks available for children. Existing tasks have mostly used still pictures [
24]. Those that included faces in motion [
28] have tended to include only a narrow range of complex emotions. As far as we are aware, there has not yet been any study of children testing complex ER in voices alone. Therefore, there is a need for a test that assesses ER in a variety of complex emotions, in both visual and auditory channels, using motion in the visual task, to get closer to the demands of the real world, while using validated stimuli that are standardized and therefore useful for research and clinical purposes.
In this study, we present such a battery: ‘The Cambridge Mindreading Face-Voice Battery for Children’ (or the CAM-C). This is an adaptation of a complex ER battery for adults [
34]. The CAM-C includes nine different complex emotions. The battery provides ER scores for faces and for voices, as well as for the number of emotions correctly recognized. The objectives of the current study were twofold: (a) to compare ER abilities of children with ASC and typically developing controls and (b) to examine the psychometric properties of the CAM-C battery, in terms of reliability, concurrent validity and ability to differentiate between children with ASC and typically developing children in ER skills.
Using this battery, we assessed differences between 8- and 11-year-old children with high-functioning ASC and a typically developing matched control group. We predicted that the ASC group would have lower scores on the battery tasks compared to controls. In addition, we predicted that CAM-C scores would correlate negatively with the level of autistic symptoms [
24,
29,
35] and positively with age [
36] and with IQ [
37,
38]. Correlations with the child version of the ‘Reading the Mind in the Eyes’ (RME) [
39], an existing complex ER task, were also calculated to examine the CAM-C battery’s concurrent validity.
Discussion
The current study tested if there are differences in complex ER between children with ASC and typically developing children. This was examined using the CAM-C, a new battery, testing complex ER in both facial and vocal expressions. As predicted, the ASC group had more difficulties recognizing complex emotions from faces and voices and recognized fewer emotional concepts, compared to the control group, even when controlling for age and verbal IQ. These results support previous findings of difficulties in complex emotion recognition in children with ASC [
25,
27-
29,
49]. The CAM-C battery demonstrated good test-retest reliability and concurrent validity. Scores were positively associated with participants’ age and negatively associated with the level of autistic symptomatology.
Children with ASC showed specific difficulties in the recognition of six out of the nine complex emotions and mental states tested:
disappointed,
jealous,
nervous,
unfriendly,
bothered and
amused. The grounds for these difficulties are discussed in reference to two main factors characterizing complex emotions [
4,
11]: complexity (that is, combining several basic emotions and mental states) and subtlety (that is, toning down an emotional expression or attempting to conceal it).
Typically developing children have been found to understand and recognize complex emotions such as
jealous,
disappointed and
embarrassed between the ages of 7 and 10 [
36,
50]. Indeed, our findings show that more than 80% of the control group recognized jealousy and disappointment successfully. However, only 60% of the participants in the ASC group recognized the concept
jealous, which includes restrained hostility towards someone as a result of social comparison [
51]. Common errors included mislabelling facial expressions of
jealous as
disappointed, possibly because of focusing on the mouth region of the face, which resembles being unhappy. Relying on the mouth area for ER while disregarding the eyes is characteristic of people with ASC [
47,
52], particularly in complex emotions [
31]. Whereas this may sometimes suffice when interpreting basic emotions (for example,
happy or
sad), configural cues, as well as theory of mind, are required for recognition of complex emotions like
jealous. Voice items for the concept of
jealous were mislabelled as
teasing (‘I can do better than you’) or
bossy (‘I deserve that car more than him’), failing to combine linguistic and paralinguistic components of the verbalizations.
Children with ASC also showed difficulties in the recognition of
disappointment, which involves sadness due to a failed expectation [
53]. Only 53% of the participants in the ASC group correctly recognized this emotion, compared to 84% of the controls. Common errors included mislabelling it as
thinking and
unsure for faces, possibly due to the gaze being directed downwards, away from the camera. Participants may have failed to integrate this cue with the unhappy mouth cue.
Disappointed voice items were commonly mislabelled as
ashamed (‘I should have won’) and
hurt (‘I tried so hard’). Whereas these labels capture the emotion’s negative valence, they do not elicit the failed expectation from the verbalizations.
Interestingly, no group difference was found for the recognition of
embarrassed. Though a larger proportion of controls (44%) recognized this emotion, compared to the ASC group (33%), this difference was not significant. Common errors for face items in both groups included
sad and
jealous. Voice items were mislabelled as
afraid (‘Do you think anyone saw me?’) and
wishful (‘Oh, I wish it hadn’t happened’). Since embarrassment is a complex emotion, dependent on the real (or imagined) presence of others [
54], the correct perception of this emotion would be expected to be facilitated by contextual cues, which were not available in the CAM-C. A task employing holistic situations in context [
29] may be useful to examine the ER of
embarrassment.
As noted, participants in the ASC group had significant difficulties with emotional concepts that form more subtle representations of basic emotions. For example, only 53% of children with ASC (compared to 84% of controls) correctly recognized
bothered, a form of mild anger. Common mistakes included
disbelieving and
bored on the face task, and
unsure (‘What are you doing here?’) and
disbelieving (‘I wish I didn’t have to do it’) on the voice task. These demonstrate how, when emotional cues are more subtle, children with ASC may miss their presence and interpret them as mental states. Another example for difficulties recognizing subtle expressions can be seen in the example of
nervous, a mild expression of fear, recognized by only 40% of the ASC group. Common errors were mislabelling a face item as
annoyed and voice items as
disgusted (‘Don’t put that near me’), or an emotionally neutral option, such as
asking (‘How many people are out there?’). These examples show again how in ASC intonation may be disregarded and verbal content may be used to recognize the speaker’s emotion/mental state. An fMRI study of adults with ASC found that the amygdala, a key brain area underlying the detection of fear in others, does not respond differentially to expressions of subtle fear [
55].
Interestingly, there was no group difference in the recognition of the positive emotion
loving. This is consistent with past research showing specific difficulties to others’ negative emotions in children with ASC [
56,
57]. Nevertheless, the ASC group had difficulties in the recognition of the positive emotion
amused, a form of reflective joy [
58]. Participants with ASC mislabelled it as
interested or
curious on the face task, and as
interested (‘You’ve done it again’) or
excited (‘Imagine that’) on the voice task, relying on the linguistic cues while missing the paralinguistic cues of the speaker’s smile [
59]. These demonstrate that even in the positive emotion domain, as complexity increases, it is harder for children with ASC to integrate the relevant cues, resulting in a misattribution of emotion.
Only 30% of the participants with ASC correctly recognized the concept
unfriendly. The ASC group mislabelled unfriendly faces as
afraid,
disgusted and
shy. These errors were probably related to the actors moving their faces away from the camera and looking sideways. Failing to recognize a protagonist as unfriendly, as well as mistaking others’ amusement for interest, may be related to the increased risk of teasing and bullying that children with ASC experience [
60,
61].
Two patterns emerge from the results, which may account for the errors made by participants in the ASC group in complex ER. First, the relative difficulty in interpreting gaze, characteristic of individuals with ASC, may underlie the pattern of results found in the
unfriendly,
disappointed and
jealous face task items. Previous studies have shown that individuals with ASC show diminished performance compared to typically developing controls in inferring mental states from the eyes [
24,
62] and atypical eye-gaze processing patterns [
63,
64].
Second, processing of emotion in prosody should be considered in relation to lowered performance of participants with ASC in the voice items. The processing of affective prosody has been found to be impaired among individuals with ASC [
65,
66], who may show overreliance on verbal information on the account of change patterns in prosodic cues such as pitch and volume that may be more relevant for the recognition of emotion.
The positive correlations of all task scores with age, independent of diagnosis, suggest that ER skills continue to develop in both typically developing children and children with ASC. In addition, as predicted, CAM-C scores were negatively correlated with the participants’ level of autism spectrum symptoms. This finding highlights the ER profile as a potential marker of ASC. Furthermore, since the range of CAST scores was quite narrow in both groups, correlations with the level of autistic traits were potentially lower than they could be if the autism spectrum was more fully represented, for example, by including undiagnosed siblings of children with ASC [
67,
68].
As predicted, complex emotion voice task scores were positively correlated with verbal ability. This may be related to the need for integration of the stimuli’s verbal content and intonation, which may depend on verbal ability. It may also demonstrate the compensatory reliance on verbal content, employed by individuals with ASC on emotion recognition tasks [
30,
65], which may be compromised in individuals with poorer verbal abilities. The correlation of verbal ability with the voice task scores may also explain the significant difference between face and voice task scores, over and above group. Indeed, when verbal ability was entered into a MANCOVA as a covariate, the difference between face and voice tasks became non-significant, while the group difference on both tasks remained significant.
Several issues are noteworthy when examining the psychometric properties of the CAM-C. Power calculations for the CAM-C tasks indicated that the battery differentiates well between the two study groups. Test-retest correlations computed for the battery (.74 to .76) suggest that this measure of complex ER is consistent over time. Furthermore, the positive correlations of CAM-C scores with the RME task provide the battery with important measures of external validity. These correlations were significant but moderate (.35 to .40), suggesting they may test different aspects of a common skill. Power levels of the CAM-C show it is sensitive to group differences across all tasks and scores. These data provide support for the CAM-C as a valid and reliable measure of complex ER skills.
Limitations and directions for future research
Several limitations should be noted. In the current study, validation of participants’ clinical diagnosis in the ASC group was based on the CAST, a screener for ASC that is based on parental report. Future studies should validate participants’ diagnosis on the basis of independent standardized measures, such as the ADOS-2 [
69], which could also contribute to the understanding of the association between ASC symptomatology and complex ER deficits. Additional research of the CAM-C is also needed to further investigate its psychometric properties, such as sensitivity and specificity, with a wider age range, a wider range of validation criteria and a larger sample.
Future research may address questions regarding the ability of the CAM-C to differentiate between ASC and other clinical groups, given that it is significantly correlated with the level of autism symptoms. Finally, some studies have examined the question of scan paths in ER using eye tracking [
31]. The application of such a paradigm in the study of the CAM-C might further elucidate the mechanisms underlying the profile found among children with ASC in the recognition of complex emotions from dynamic facial stimuli.
Competing interests
The authors declare that they have no competing interests. The CAM-C is derived from the
Mindreading DVD that is published by Jessica Kingsley Ltd (
www.jkp.com/mindreading), profits from which go to autism research. The CAM-C is available as free download for use in not-for-profit research from
www.autismresearchcentre.com.
Authors’ contributions
OG and SBC designed the study. Data were collected by OG, who undertook all statistical analyses and drafted the manuscript. OG, YSG and SBC contributed to subsequent drafts and the final submitted manuscript. All authors read and approved the final version of the manuscript.