Reimann · Autism research : official journal of the International Society for Autism Research 2021 · case-control study · n=69

Gauging facial feature viewing preference as a stable individual trait in autism spectrum disorder.

Level 4 - case-series / case-control

Case-control observational study comparing individuals with ASD to typically developing controls

PubMed 34008916 · doi:10.1002/aur.2540 · record verified 2026-08-26

What was done

Researchers used eye tracking to evaluate gaze stability across varying amounts of dynamic video data in 33 high-functioning individuals with autism spectrum disorder (ASD) and 36 age- and IQ-matched typically developing individuals (ages 12 to 30 years). Participants viewed 22 Hollywood movie clips (14 seconds each, 7,410 total frames) depicting social interactions. An artificial neural network segmented the video stimuli, and dwell times on core facial features were assessed across subsets of one, three, five, eight, and 11 clips to measure stability within subjects, within groups, and between groups.

What was found

Gaze dwell times on core facial features became progressively more stable as the number of movie clips analyzed increased. Using an adequate volume of data, participants with ASD exhibited significantly less face-centered gaze (centralized on the nose; p < 0.001) compared to typically developing controls. There were no statistically significant group differences in looking times directed at the eyes or mouth. Specific numerical dwell times, percentages, and confidence intervals were not provided in the abstract.

Why it matters

This study shows that facial feature viewing preferences can serve as stable individual traits in ASD research, provided sufficient stimulus data are analyzed to avoid measurement error. It also demonstrates an automated machine learning workflow for segmenting naturalistic social video stimuli.

Limits

The sample is modest (n = 69) and restricted to high-functioning individuals aged 12 to 30 years, limiting generalizability to younger children or individuals with lower cognitive functioning. The abstract omits precise dwell time metrics, effect sizes, and variance measures.