Making Tough Decisions: The Neural Correlates of Categorization in Children with and without Autism

Background: Studies of categorization of objects such as chairs or cats (e.g., Gastgeb et al., 2006, Child Development) or facial expression (Rump et al., 2009, Child Development) indicate that individuals with autism improve at categorization across development but do not reach the level of expertise demonstrated by IQ and age-matched typically developing (TD) participants. In these studies, individuals with autism had particular difficulty categorizing atypical exemplars. This difficulty is thought to be related to the tendency of individuals with autism to use explicit, rule-based strategies for information processing (e.g., Minshew et al., 2002, Neuropsychology); however, this hypothesis is difficult to test using traditional behavioral methods.

Objectives: To examine neurofunction during categorization of ambiguous and nonambiguous items to gain further understanding of the cognitive basis for the behavioral differences observed in autism.

Methods: Currently, fifteen 8- to 15-year-old children with autism (age M = 12.2, FSIQ M = 112.7) and twelve age- and IQ-matched typically developing children (age M = 11.5, FSIQ M = 113.8) have successfully participated in this IRB-approved study.  This is a block-design fMRI study comparing activation differences for photos of items that can be easily assigned a semantic or quantity label (control condition) vs. items that cannot (ambiguity condition). There are four types of pictorial stimuli (two semantic + two quantity).  The semantic stimuli are cued with the question “What is it?” and are of two types: (1) control items with an obvious label (e.g., a car with the choices of car or plane); and (2) ambiguous items that cannot be easily assigned one of the two presented semantic labels (e.g., a liger with a lion’s head and tiger’s body presented with the options of lion or tiger). The quantity stimuli are single photos of groups of items and are cued with the question “How many?” and are of two types: (1) control stimuli with a correct answer presented (e.g., 3 apples presented with “about 3” and “about 4” as options); and (2) ambiguous items with potential responses that are equidistant from the actual number of items (e.g., 3 items with the options of selecting “about 2” or “about 4”). In this way, we can examine semantic and quantitative categorization under both ambiguous and unambiguous circumstances. All children have a minimum score of 70% correct, and there is no performance difference between the groups.

Results: For the children with TD, no activation differences are seen for ambiguous vs. non-ambiguous items; whereas, the group with autism has greater activation in bilateral inferior frontal gyri, medial frontal gyrus, right middle frontal gyrus, and right inferior parietal lobe. Compared to the group with TD, the children with autism have a network of increased activation for the ambiguous items including left inferior triangularis, right middle frontal gyrus, right superior medial frontal gyrus, right insula, and right supramarginal gyrus.

Conclusions: The results indicate that when making challenging categorical decisions, individuals with autism have a neural signature (increased use of frontal processing regions) consistent with use of an explicit, rule-based strategy.