The Effects of Stimulus Complexity and Social Realism on Emotion Recognition in Children with ASD.

Background: The use of visual supports is a common strategy for facilitating social cognition and the understanding of emotions in children with autism spectrum disorder (ASD). Yet, while the processing of visual information is considered a relative strength in ASD, the deficits in processing social information are well-documented. Furthermore, very little is known about the stimulus characteristics that support (or impede) the understanding of social imagery. The purpose of this study was to examine the effect of stimulus complexity and social realism on emotion recognition in ASD and typical development.

Objectives: This study compared the ability of typically developing (TD) children and children with ASD (ages 6 – 12 years) to identify emotions (happy, sad, mad, scared, disgust, surprise) when presented with face stimuli that systematically varied in complexity (i.e., the amount of detail and the intricacy of the lines, patterns, and colors in the image) and social realism (i.e., how realistic an image). Our exploratory research questions were:

1) Is there a difference between TD and ASD groups in the ability to identify emotions when complexity and social realism vary?

2) Which stimulus characteristics predict better emotion recognition for each group?

Methods: Nineteen TD children and 18 age- and gender-matched children with ASD participated. Children viewed face stimuli (presented in counterbalanced order) on a video monitor. Face stimuli were carefully manipulated using photoshop to make incremental adjustments in stimulus complexity and social realism: two levels (low, high) for each dimension were developed using a 2x2 approach. Each image was presented for 4 seconds and children were asked to “tell me what emotion you see”.

Results: Data for each emotion were submitted to six 2 (group) x 2 (social realism) X 2 (complexity) mixed model ANOVAs. A main effect of social realism was found such that higher social realism was associated with more accurate emotion recognition. When verbal ability was controlled, this main effect disappeared. A group X emotion interaction approached significance for ‘mad’ faces such that the ASD group performed best under conditions of low realism and the TD group continued to perform best under conditions of high realism.

Conclusions: Better performance by both groups in conditions of high social realism could be explained by higher ecological validity. Alternatively, our ‘social realism’ condition may actually have captured ‘prototypicality’ or how well an image represents its category. Although the ability to extract prototypes is thought to be disrupted in ASD, some researchers have argued that it is not the ability to extract the prototype but to apply it flexibly. Thus, both the ASD and TD samples may have experienced an emotion recognition advantage under conditions of higher prototypicality in this experiment. Finally, ‘mad’ faces may be particularly dependent on configural processing and there is an atypical tendency in ASD toward part- rather than whole-processing.