Intact Prototype Formation but Impaired Generalization in Autism

Abstract

Cognitive processing in autism has been characterized by a difficulty with the abstraction of information across multiple stimuli or situations and subsequent generalization to new stimuli or situations. This apparent difficulty leads to the suggestion that prototype formation, a process of creating a mental summary representation of multiple experienced stimuli that go together in a category, may be impaired in autism. Adults with high functioning autism and a typically developing comparison group matched on age and IQ completed a random dot pattern categorization task. Participants with autism demonstrated intact prototype formation in all four ways it was operationally defined, and this performance was not significantly different from that of control participants. However, participants with autism categorized dot patterns that were more highly distorted from the category prototypes less accurately than did control participants. These findings suggest, at least within the constraints of the random dot pattern task, that although prototype formation may not be impaired in autism, difficulties may exist with the generalization of what has been learned about a category to novel stimuli, particularly as they become less similar to the category's prototype.

Figure 1

The prototype pattern for one of the three categories and one each of the medium-, low-, and high-distortion patterns generated from the prototype.

Figure 2

Mean proportion of correct categorization responses for control (dark bars) and autism (light bars) participants for the stimuli presented in the final-test phase, including the combined training and repeated-transfer medium-distortion stimuli (Medium) and each of the three final-transfer stimulus types: prototype, low-distortion (Low), and high-distortion (High). Means and standard errors are represented. Asterisks indicate those distortion levels for which means differed significantly between groups at p < .05.

Figure 3

Mean proportion of “old” responses for control (dark bars) and autism (light bars) participants for the stimuli presented in the final-test phase. Along the x-axis, below the label for each distortion level, is a second label of “old” or “new” to indicate the correct response for that stimulus type.

Figure 4

Increased functional connectivity to primary visual cortex in subjects with improved generalizability. Images show correlation coefficient between connectivity to primary visual cortex and generalizability score for 16 subjects. Colored regions satisfied p<0.05, uncorrected. Areas for which connectivity to visual cortex was most associated with subjects showing high generalizability included attentional regions: bilateral intraparietal sulcus, left frontoinsula, and supplementary motor area.