Correlates of Emotion Recognition Task Performance in Autism: A Meta-Analysis

Background: Emotion recognition is thought to be an important skill necessary for successful social interaction (Harms et al., 2010), a core area of difficulty in ASD. Yet, few studies have examined the degree to which emotion recognition ability in ASD is associated with other theoretically related constructs that would provide support for the construct validity of existing measures of emotion recognition. Hall et al. (2009), for instance, conducted a meta-analysis to examine the extent to which interpersonal sensitivity (i.e., perceiving others accurately) in typically developing individuals correlates with psychosocial variables (e.g., personality traits, indicators of mental health, and social and work-related competencies) of the perceiver. The present meta-analysis takes a similar approach, focusing on correlates of emotion recognition in participants with ASD. To date, no meta-analysis has examined the convergent validity of emotion recognition tasks used in autism research. 

Objectives:  The present meta-analysis seeks to examine the convergent validity of measures of emotion recognition in ASD.

Methods:  Studies were extracted from searches in PsycINFO, Web of Science, PubMed and PsycARTICLES using keyword combinations including, autism, Asperger syndrome, pervasive developmental disorders, emotion recognition, facial affect, and emotion perception. Additional articles were extracted from reference lists of recently published review articles. Articles were included if they a) were accessed for free and published in English, b) contained an experiment that examined emotion recognition performance from faces, vocal tones or body cues, and c) reported a Pearson’s r correlation coefficient for associations between task performance and psychosocial variables. Given the paucity of research in this area, we were only able to identify three main categories of psychosocial correlates for which at least five effect sizes for one outcome measure were obtained: (1) ASD symptomology, (2) caregiver ratings of social skills, and (3) self-reported difficulties in identifying and describing one’s own emotional experiences (variations in alexithymia, Bagby et al., 1994). Data was analyzed using Comprehensive Meta-analysis Software (CMA). Data collection and variable coding are still in progress. Attempts will be made to code for methodological variables, such as stimulus and task characteristics, which may moderate the relationships between task performance and psychosocial correlates. 

Results:  Although data collection and analyses are still in progress, preliminary results are provided in Table 1. 

Conclusions:  Preliminary results indicate relatively modest significant correlations in the expected directions. Emotion recognition performance was negatively correlated with ASD symptomology; positively correlated with caregiver-reported social competence; and negatively correlated with alexithymia. Preliminary results provide support for the convergent validity of emotion recognition tasks, although the strength of correlations may be exaggerated by publication bias. In addition, the relatively small number of studies that examined relationships between emotion recognition and theoretically related constructs suggests more research is needed to examine the convergent validity of measures of emotion recognition used in autism research.