Semantic Memory Structure in Children and Adolescents with Autism Spectrum Disorder

Background: Learning and memory abilities are fundamental to the application and generalization of skills acquired in Autism Spectrum Disorder (ASD) treatment programs. One important aspect of learning and memory is semantic categorization, whereby related concepts are grouped and linked to form an organized semantic memory system. Categorization processes occur hierarchically across three different levels: superordinate (‘Animal’), basic (‘Dog’), and subordinate (‘Beagle’; Rosch & Lloyd, 1978). Typically developing individuals (TD) identify objects fastest at the basic level (Jolicoeur et al., 1984), indicating it as the usual entry point into semantic memory. In addition, TD categorize living objects with different proficiency than non-living objects (McMullen & Purdy, 2006). ASD children appear to categorize according to concrete rather than abstract features (Shulman et al., 1995), and have trouble forming category prototypes (Klinger & Dawson, 2001). The Weak Central Coherence (WCC) theory, which postulates a bias to process detail, may help to explain categorization differences in ASD. 

Objectives: 1) To determine if and how categorization differs in ASD compared to TD youths. Based on WCC theory, we expected ASD participants, relative to controls, to demonstrate faster subordinate responses (i.e., most specific) and slower superordinate responses (i.e., most general). 2) To examine relationships between categorization and central coherence performance.

Methods: To date, our sample includes 19 high-functioning youths (aged 8 to 18 years) with ASD and 16 age-, sex-, and IQ-matched TD participants. Participants were shown word and line drawing pairs of living (dogs, bugs, birds) and non-living (cars, boats, aircrafts) objects to match at the three levels of categorization. The Embedded Figures Test (EFT) was employed as a measure of central coherence.

Results: A Level (basic, subordinate, superordinate) x Group (TD, ASD) mixed Analysis of Variance of correct response times to match trials, revealed an effect of Level, F(2,64) = 13.48, p < 0.001, only. For the ASD, better performance on the EFT (i.e., weaker central coherence) was related to slower responses at basic (r(19)= - 0.53, p < 0.05) and superordinate (r(19) = -0.65, p < 0.01) levels. EFT performance was not correlated with responses at any level for the TD. Separate analyses of responses to living and non-living objects provided preliminary support for our hypothesis with living objects: ASD categorized slower at the superordinate level (d = 0.24) and faster at the subordinate level (d = 0.29) than the TD. ASD categorized non-living objects slower than the TD at both basic (d = 0.28) and subordinate levels (d = 0.24). 

Conclusions: Consistent with WCC theory, results provide preliminary evidence that youths with ASD categorize living objects slower than their TD peers at the superordinate (most general) level and faster at the subordinate (most specific) level. Weaker central coherence was associated with slower categorization at more general levels. In contrast, ASD participants categorized non-living objects slower than TD youths at the subordinate and basic levels. Results indicate that semantic memory organization and processing differs between ASD and TD youths and that the living /non-living distinction is an important direction for future studies.