Developmental changes in face visual scanning in autism spectrum disorder as assessed by data-based analysis

Abstract

Individuals with autism spectrum disorder (ASD) present reduced visual attention to faces. However, contradictory conclusions have been drawn about the strategies involved in visual face scanning due to the various methodologies implemented in the study of facial screening. Here, we used a data-driven approach to compare children and adults with ASD subjected to the same free viewing task and to address developmental aspects of face scanning, including its temporal patterning, in healthy children, and adults. Four groups (54 subjects) were included in the study: typical adults, typically developing children, and adults and children with ASD. Eye tracking was performed on subjects viewing unfamiliar faces. Fixations were analyzed using a data-driven approach that employed spatial statistics to provide an objective, unbiased definition of the areas of interest. Typical adults expressed a spatial and temporal strategy for visual scanning that differed from the three other groups, involving a sequential fixation of the right eye (RE), left eye (LE), and mouth. Typically developing children, adults and children with autism exhibited similar fixation patterns and they always started by looking at the RE. Children (typical or with ASD) subsequently looked at the LE or the mouth. Based on the present results, the patterns of fixation for static faces that mature from childhood to adulthood in typical subjects are not found in adults with ASD. The atypical patterns found after developmental progression and experience in ASD groups appear to remain blocked in an immature state that cannot be differentiated from typical developmental child patterns of fixation.

Keywords: autism; development; eye tracking; face; face perception; spatial statistic.

FIGURE 1

Experimental task and recording methodology. (A) Sequence of the different displays presented during the task. The task was sequentially incremented by presentation of up 20 different photographs (here shown only for the four first slides). (B) Fixations and visual trajectories for a typical adult subject (B1) note the stereotyped triangular pattern of fixation and a child with ASD (B2). Each blue point corresponds to a single fixation.

FIGURE 2

Spatial normalization of stimulus faces. (A1) For each face the following anatomical landmarks were identified: the two pupils, tip of the nose, lip commissures, and the ear tragus. (A2) Example of a normalized prototypic face. (B) Face resulting from the average of 22 stimulus faces. The colored dots (one color per face) represent the fixations performed by a single typical subject for each stimulus face presented. (C) Fixations on the calibration map. (C1) Each subject had to stare at the various targets. Each color dot corresponds to one subject. (C2) Statistically significant clusters of fixation (red tiles) detected using Dirichlet tessellation.

FIGURE 3

Comparison between a priori and a posteriori methods. (A,A1) Prototypic face with superimposed fixations performed by tested typical adults (N = 13 subjects) for all neutral non-familiar faces (N = 22 presented faces). Each colored dot corresponds to one subject. The regions of interests (ROIs) were centered on core features (eyes, mouth). (A2) From the pattern of superimposed fixations presented in (A1), statistically significant clusters of fixation (red tiles) were detected using Dirichlet tessellation. Large ROIs, not centered on core features (see yellow spots) were circumscribed. (B) Bar graphs presenting the mean number of fixations (B1) and the mean delay to the first fixation (B2) per subject in the four ROIs. M, mouth; LE, left eye; RE, right eye. ^∗ indicates a significant variation.

FIGURE 4

Comparison for the four groups of total time spent on the eyes and mouth. TD-A, typical adult; TD-C, typically developing children; ADL/ASD, adults with autism spectrum disorders; ASD-C, children with ASD. Stars on the graph indicate significance between groups for each ROI as for all four groups the time spent in each ROI was significantly different (see text). ^∗ indicates a significant variation.

FIGURE 5

Characteristics of visual fixations in the four groups. (A) mean total time spent on fixation: TD-A (LE > RE > M p < 0.01); TD-C [LE > (M = RE) p < 0.01]; ADL/ASD (LE > M > RE p < 0.01); ASD-C (LE > M > RE p < 0.01). (B) Delay to first fixation in TD-A was shorter than in TD-C or both ADS groups. (C) Summary schematic comparing data for the four groups. The size of each area is proportional to the mean total fixation time and the number beside each area indicates the sequence of fixation. TD-A, typical adult; TD-C, typically developing children; ASD-A, adults with ASD; ASD-C, children with ASD. M, mouth; LE, left eye; RE, right eye. ^∗ indicates a significant variation.