Category Formation in Autism: Can Individuals with Autism Form Categories of Dot Patterns

Background: Categorization is a critical cognitive ability that reduces demands on memory and allows individuals to focus on important aspects of objects while ignoring irrelevant details.  In fact, it is such a basic ability that infants are able to form categories of dots, objects and faces within the first year of life (e.g., Bomba & Siqueland, 1983; Rubenstein, Kalakanis, & Langlois, 1999; Strauss, 1979; Younger, 1985).  There is a growing amount of evidence suggesting that individuals with autism have difficulty with some aspects of categorization and engage in different categorization processes than do typically developing individuals (e.g., Gastgeb, Strauss, & Minshew, 2006; Klinger & Dawson, 1995; Plaisted, 2000).  A recent study by Gastgeb et al. (2006) found that individuals with autism can readily categorize objects (e.g., cats and chairs) when the task involves simple and typical objects but have difficulty when categorization is more complex or involves less typical objects.  It is not yet known whether these categorization differences are present during category formation.  

Objectives: To examine whether individuals with autism and typically developing individuals differ in their formation of artificially designed categories of dot patterns varying in typicality (i.e., amount of distortion from a prototype).

Methods: High functioning adults, adolescents and children with autism (8-45 years old) and age and IQ matched controls were tested in a category formation task with dot patterns varying in typicality from typical (low distortion of a prototype) to atypical (high distortion of a prototype).  During the learning phase, participants were shown high distortion dot patterns.   After a delay period, participants were presented with new low distortions, new high distortions, and dot patterns from a new/different category (i.e., high distortions of a different dot pattern prototype).  For each dot pattern, participants responded as to whether the dot pattern was a member of the category that they saw earlier or a member of a different/new category.  

Results: The autism group differed from the control group in the typicality structure of the categories they formed and in the strength of their category boundaries.  The control group showed a pattern of more accurate categorization of low (typical) distortions than high (less typical) distortions.  Even though the autism group showed this pattern, they did so to a lesser degree. The control group also evidenced clear category boundaries by accurately excluding the non-category members from the learned category. However, the autism group evidenced more fuzzy boundaries by excluding more high distortion category members and including more non-category members into the learned category.     

Conclusions: Individuals with autism did not seem to abstract typicality structures of the dot patterns and evidenced particular difficulty categorizing members at the boundary of a category (i.e., less typical/highly distorted members).  They also formed less well-defined category boundaries.  The results of this study parallel Gastgeb et al (2006) suggesting that individuals with autism have categorization difficulties with respect to natural and artificially designed categories.