The current study investigated the influence of body posture on facial expression perception in autistic and non-autistic adults. We found that body posture has a significant influence on facial expression perception in both groups, but autistic individuals showed a greater influence of body posture compared with non-autistic individuals. We also found reduced precision of isolated facial expression representations, but no difference in the precision of isolated body posture representations, in autistic individuals compared with non-autistic individuals. Within each group, the precision of isolated facial expression representations was related to the body context effect, with lower precision leading to an increased influence of body posture. Taken together, our results show that perception of facial expressions in individuals with Autism is more strongly influenced by body posture compared to non-autistic individuals, and that a reduced precision in facial expression representations is likely a key factor driving this effect.
Our finding of a significant influence of body posture on facial expression categorisation in non-autistic adults, as well as in autistic adults, is consistent with the findings of previous research (Aviezer et al., 2008; Brewer et al., 2017; Meeren et al., 2005; Teufel et al., 2019). However, the comparison between the autistic and non-autistic groups differed sharply with the only previous study on this topic (Brewer et al., 2017). Contrasting with the current work, Brewer and colleagues found no difference in the influence of body context between the autistic group and the non-autistic group, as well as no difference in the precision of facial expression representations between the groups. Importantly, however, from a Bayesian perspective both sets of results are consistent. Specifically, a Bayesian account would predict that the biasing influence of body context is related to the reliability or precision with which isolated facial expression cues are represented. Our data indicate that this prediction holds for both the autistic and non-autistic groups. In both groups, the body context effect is larger in individuals with less reliable facial expression representations compared to those with more reliable representations. The difference in the body context effect between the groups might thus be a consequence of the difference in the reliability of facial expression representations: as predicted, the autistic group showed less reliable representations than the non-autistic group and might therefore be more strongly influenced by body context. In the study by Brewer and colleagues (2017), the two groups do not differ in how reliably they represented facial expressions. The fact that these authors did not find a difference in the body context effect between groups is thus expected from a Bayesian perspective. In summary, the contrasting results between our work and the only previous study focussing on the influence of body context on facial expression perception in autism are both consistent with a Bayesian interpretation of the body context effect.
Research on facial expression recognition in autistic individuals has produced conflicting results, with some studies finding difficulties in autistic compared to non-autistic individuals and others finding no differences (For reviews, see: Keating & Cook, 2020; Uljarevic & Hamilton, 2013). One potential explanation for these mixed findings is the genuine heterogeneity of facial expression perception in autism, with different subgroups showing different extents of difficulties in facial expression perception (Meyer-Lindenberg et al., 2022). Our autistic sample was more than twice the size of the autistic sample in Brewer and colleagues (2017), increasing the probability of including individuals from subgroups with pronounced difficulties in facial expression perception.
Another potential factor contributing to the inconsistencies in the literature may be the type of facial expression stimuli used. Studies that rely on stereotypical 100% facial expressions may not be sensitive enough to detect subtle differences between autistic and neurotypical groups due to ceiling effects associated with using these stimuli (Keating & Cook, 2020). In our study, the use of facial expression morphs allowed us to characterise facial expression recognition in a more sensitive manner. The reduced precision of facial expression representations that we observed in autistic individuals is consistent with the results of a previous study by Wang and Adolphs (2017), who also used facial expression morphs and found a similarly reduced precision of facial expression representations in autistic individuals.
As mentioned above, in both groups, we found a relationship between the precision of facial expression representations and the influence of body posture. This link between facial expression processing and body context effect has also been found in children and adolescents (Ward et al., 2023), with younger children showing reduced precision in facial expression representations and a greater reliance on body posture in making judgments of facial expression, and vice versa for older children. This consistent pattern of findings across typical development (Ward et al., 2023), non-autistic adults, and autistic adults provides support for the notion that shared mechanisms drive the integration of emotion information from face and body across these groups. The extent to which such principles apply more widely to the integration of other social signals from face and body, or indeed other contextual signals, remains an open and interesting question.
With respect to the difference in body context effects between the two groups, it is important to note that this finding cannot be explained by avoidance of the face and/or eye region in the autistic group and judgments based purely on body context. Avoidance of eye contact is a key diagnostic feature of Autism (American Psychiatric Association, 2013). However, the high consistency in performance of autistic individuals across the isolated facial expression and face-in-context tasks suggests that autistic individuals were discriminating facial expressions in the face-in-context task, rather than focussing on the body posture only. Additionally, based on our exclusion criteria, any participants, who simply responded according to body posture, would have been removed from the analysis due to an inability to fit psychometric functions to the data.
A limitation of our study is the focus on the emotional expressions of anger and disgust only. These emotions were chosen due to a robust effect of body context compared to other emotions (Aviezer et al., 2008). A robust influence of body posture was important for maximising sensitivity to individual differences within and between our two groups of observers. Furthermore, disgust and anger can be conveyed by simple, clearly identifiable body postures. By contrast, other postures like sadness and happiness are less easily conveyed clearly via a simple posture (Lopez et al., 2017). However, there is no reason to believe that the integration of other emotions wouldn’t be governed by similar underlying principles.
In summary, our results demonstrate both similarities and differences between autistic and non-autistic individuals regarding the perception of facial expressions, body postures and facial expressions in the context of a body posture. A significant reduction was found in the precision of facial expression representations in autistic individuals, relative to non-autistic individuals, but there was no difference in the precision of body posture representations between groups. Autistic individuals also showed a greater reliance on body posture in making their judgements of facial expression than non-autistic individuals. Across each group, the influence of body posture on facial expression perception was related to the precision of facial expression representations. Our results highlight that similar principles govern the integration of facial expression and body posture information in autistic and non-autistic individuals.