Temporo-Spatial Gaze Patterns in Autistic Children and Adults While Viewing Video Stories

Background: Eye-tracking has been used to investigate gaze behavior in individuals with autism spectral disorder (ASD) while viewing socially salient stimuli like faces, but results are not consistent across studies. The discrepancy might be due to difference in age groups and limitation of the fixation time analysis that has been used in most studies so far. We hypothesized that differences are most likely to be detected by taking sub-second temporo-spatial gaze patterns into account, because social salience in a scene dynamically changes from one place to another over time.

Objectives: To clarify sub-second characteristics of gaze behavior in children and adults with ASD, we examined temporo-spatial gaze patterns while they viewed ecological video stimuli.

Methods: Children with ASD (2-8 y.o., n=26), normal children (1-7 y.o., n=26), normal adults (>21 y.o., n = 27), and adults with ASD (>18 y.o., n=27) participated. Gaze positions were recorded with an eye-tracker, while they viewed a short video stimulus with sound, 77 s long. The video consisted of 12 short clips, each of which was taken from a film or a TV program for young children, and involved one, two, three, or more human characters. To quantify differences in temporal patterns of gaze movement, we calculated distance between gaze positions in every pair of subjects (₁₀₆C₂) in every frame (2327 frames), and applied multidimensional scaling (MDS) to visualize similarity of the temporo-spatial gaze patterns among 106 subjects on a two-dimensional plane (MDS plane). We further analyzed each frame whether there was any discrepancy in gaze distributions across groups.

Results: Normal adults and normal children formed two distinct clusters near the center of the MDS plane. Groups of autistic children and adults distributed in the periphery, each surrounding its age-matched control group. The results indicate that temporo-spacial gaze patterns were highly conserved within normal children, within normal adults, but varied in ASD subjects. Significant across-group differences in gaze distribution appeared, for example, when the subjects viewed two children that talked in turn, and when the subjects viewed a single child that announced her name. In the former case, gazes in normally developing children and adults moved between the two children in a highly conserved manner, but those in children and adults with ASD did not. In the latter case, normally developing children spent much more time in viewing the mouth than in viewing the eye, in marked contrast to the normal adults that concentrated on the eyes. In children and adults with ASD, there was no such preference to the mouth or to the eyes in viewing the scene.

Conclusions: We conclude that normally developing children share temporo-spatial gaze patterns that grew into those shared among normal adults. On the other hand, temporo-spatial gaze patterns were varied from subject to subject in both children and adults with ASD. Diagnostic application of the present temporo-spatial analysis using MSD merits further investigation in both age groups.