Abnormal Use of Facial Expressions in ASD: A Meta-Analysis

Background: The abnormal use of facial expressions is a noticeable clinical feature of ASD (APA, 2013), yet this specific aspect of nonverbal communication has received little empirical attention. The existing research on this topic has yielded an inconsistent pattern of results. Some studies show that individuals with ASD are less expressive (Loveland et al., 1994; Snow et al., 1987), are less likely to naturally attend to and imitate others’ expressions (McIntosh et al., 2006), and may display confusing, ambiguous or inappropriate facial expressions in which it is difficult to interpret what emotion they are expressing (Brewer et al., 2016; Reddy et al., 2002). However, other studies fail to find group differences in imitative expression (Press et al., 2010) and even show heightened expression in ASD in some circumstances (Capps et al., 1993). Clearly, research is needed to untangle this puzzling pattern of results.

Objectives: The goal of this study is to review and make sense of the existing literature on abnormal facial expressions in ASD. Of special interest is to identify the moderating variables that may be accounting for inconsistent findings in this literature.

Methods: We use meta-analytic techniques to summarize existing effect sizes that compare facial expressions in ASD with matched control participants. We will also examine the influence of several potential moderating variables including study setting (experimental vs. natural), social context of facial expressions (e.g., social vs. nonsocial), task demands (e.g., spontaneous vs. voluntary expression production), and outcome measure (salience vs. quality of expression). More than 30 articles have been retrieved for potential inclusion, although literature searches and data analysis are still in progress.

Results: Thus far, 9 studies have been coded and analyzed yielding 15 separate effect sizes. Across all effect sizes, participants’ facial expressions differed significantly from matched comparison groups, cohen’s d = .230, 95% CI[0.04, 0.42], p = .018. A few interesting patterns are emerging. First, effect sizes appear to be influenced by the specific dependent measure of interest across studies. For example, large group differences were found in quality of expression, cohen’s d = .931, p < .001, k = 5, but not for salience of expression, Cohen’s d = -.104, p = .210, k = 9. Second, the context in which expressions were used influenced the strength of the effect sizes. For example, in naturalistic settings, social smiles (expressions directed towards another person) were much less frequent in the ASD participants, Cohen’s d = 1.14, p < .001, k = 3, compared to non-social expressions which were only marginally different, Cohen’s d = .318, p = .077, k = 5.

Conclusions: Initial results suggest that, compared to neurotypicals, people with ASD may not be particularly less expressive, but are significantly more likely to convey confusing facial expressions that are difficult for others to interpret. Additionally, in naturalistic settings, children with ASD may not be less expressive overall, but are less likely to use facial expressions with communicative intent or to reciprocate others’ expressions. These findings may have important clinical and research implications.