Not Created Equal: Identifying Subtypes of Toddlers with Autism Based on Their Attentional Patterns

Background:  

Regulating attention to complex visual scenes and abstracting task-relevant information is essential for development of cognition.  Compared to typical peers, toddlers with ASD experience attentional difficulties and atypical visual scanning patterns, yet their performance is highly heterogeneous.  Identifying sources of such heterogeneity would advance our grasp of the underlying mechanisms and inform design of optimal intervention strategies.

Objectives:  

As part of a study examining contextual factors on visual scanning strategies in ASD, we exposed a large group of toddlers with autism (83.3% male) (n=84, age=1.93 yrs, sd=0.61) to a 1-min video with an actress speaking to the camera simulating bid for dyadic attention.  Four salient toys were placed in each corner of the screen.  The purpose of this analysis was to identify consistent subtypes within the autism group based on their looking patterns. 

Methods:  

Gaze behavior was recorded with an eye tracker. Proportion of looking at specific regions of interest were examined using hierarchical clustering analysis using Ward’s algorithm. Distances between pairs of observations were calculated using percentage values of overall time spent looking at the scene, the speaker, toys, and background. To examine cluster stability, a set of bootstrap samples was obtained from the data by resampling with replacement, and the same hierarchical clustering procedure was applied to each such sample.

Results:  

Three clusters were identified; all between-cluster comparisons were significant at least at p<.01 level. Cluster#1: Toddlers who had difficulty attending to the scene in general, and when they did, spent less time looking at the speaker and more at toys; they had low verbal and nonverbal developmental quotient (DQ).  Cluster #2: Toddlers who showed very good attention to the scene, but spent less time on the speaker and more on the toys; they had higher verbal and nonverbal DQ. Cluster#3: Toddlers showed distribution of attention similar to our typical toddler sample (data not reported here):  very good attention to the scene in general and to the speaker in particular and limited attention to toys; this group was as high functioning as those in Cluster#2. There were no differences between clusters with regard to symptom severity score on the ADOS.  Bootstrapping analysis revealed very high cluster stability.  Preliminary analysis comparing performance of a subset of toddlers at 2 and 3 years suggests excellent stability of clusters over time.

Conclusions:

Performance of toddlers with autism differed with regard in their ability to regulate attention to the visual scene as well as their interest in the speaker and objects within the scene.  There was considerable stability in cluster membership at 2 and from 2 to 3 years. Cluster membership could not be accounted for by differences in symptom severity and were only moderately associated with verbal and nonverbal functioning.  Considering the importance of regulating attention in general, as well as selecting context-relevant aspects of a scene for learning and adaptive functioning, the properties of visual attention and scanning in the second year are likely to be highly consequential for future development of social and nonsocial cognition and predictive of outcome.