Efficacy of a Facial Affect Recognition Training Tool for Children with Autism Spectrum Disorders

Background: A great deal of human communication is nonverbal. As such, recognizing facial expressions is critical to understanding and communicating with others. Children with autism spectrum disorders (ASD) often struggle to recognize emotions from facial expressions (facial affect), hindering their social interactions.

Objectives: We hypothesized that computerized facial affect recognition (FAR) training, paired with coaching, will help children with ASD (1) to improve their accuracy and speed at identifying facial affect by increasing their attention to critical facial features and (2)  to more effectively communicate emotion in their own facial expressions with imitation training.

Methods: Children with ASD (8-14 years old) who had an existing FAR deficit were eligible for this study. To establish a pre-training baseline, children completed several direct assessments of FAR. Parent, teacher, and child self-report questionnaires of social functioning were also collected. During the intervention, trained coaches guided children through the MiX computerized training program (Humintell.com). The program utilizes audio-visual instruction with adult faces portraying seven emotions. Faces changed from neutral to expressive faces, reflecting particular emotions. Coaches modified the screens to direct children to important facial features associated with each emotion and also assisted children with imitation trials and practice tests. Individual training sessions (1 hour) were held twice weekly. Sessions were video-recorded to monitor treatment fidelity and to allow for multiple coders to rate imitations. The presentation rate of facial stimuli varied systematically across sessions and according to the child’s previous success rate. Training continued until the child was accurate at a presentation rate 1/5 second across three consecutive sessions. Both immediately upon completion and then approximately one month later, direct assessments of FAR and social functioning questionnaires were re-administered.

Results: Each child progressed rapidly through the training, reaching criterion within six weeks. Improvements in recognition and processing speed were demonstrated by progression through the program and performance on the MiX post-test. Improvements also generalized to other FAR measures. Specifically, children performed at a higher level on experimental FAR measures that utilized adult faces via a similar labeling paradigm and a novel matching paradigm. A third experimental measure that utilized children’s faces did not show consistent improvement across children. A composite FAR score, including all experimental measures, showed that children performed below age-expected levels before training and improved to at or above age-expected levels afterwards. Children also demonstrated modest imitation improvements as observed when coding the video recordings. Improvements persisted at maintenance testing. Further analyses are planned to evaluate extended effects on social functioning as rated by the child, parent, and teacher.

Conclusions: These data show that coach-assisted computerized training with imitation exercises successfully alleviated FAR deficits in children with ASD. Although effects persisted one month after training, future studies should investigate whether “boosters” are necessary to maintain the skill long-term. Additional work is planned to isolate which components individually or together results in the greatest training effect and to better evaluate behavioral improvements in peer interactions after training.