Functional Neuroimaging of Cognitive Flexibility in Autism

Background: Individuals with ASD demonstrate impairments in flexibly shifting to new adaptive response strategies. This cognitive impairment may underlie aspects of behavioral rigidity and insistence on sameness evident in some individuals with ASD.  Animal and healthy adult neuroimaging studies indicate that frontostriatal circuitry is involved in cognitive flexibility, but few studies have directly examined the neural circuitry involved in cognitive flexibility impairments in ASD. 

Objectives: To examine the neural systems supporting cognitive flexibility using reversal learning in individuals with ASD.

Methods: To date, eight individuals with ASD (3 females; mean age 16.4, SD 6.3) and eight age-matched healthy controls (3 females; mean age 15.6, SD 5.7) have performed 2- and 4-choice spatial reversal learning tests during fMRI. Recruitment of additional subjects is ongoing. Individuals selected the correct stimulus location from identical stimuli (2 or 4 boxes), and continued selecting it over repeated 3 second trials.  The correct location changed after 4, 5 or 6 consecutive correct responses, at which point individuals received feedback that their choice was now incorrect, indicating that they needed to choose a new correct location. Thus, in the 2-choice condition individuals knew what the new correct location would be, but in the 4-choice condition they needed to learn which of the 3 alternative locations was correct.

Results: Control subjects engaged bilateral premotor cortex, anterior cingulate cortex, posterior parietal cortex (PPC), thalamus and caudate when shifting to a new response (P<.05 corrected). This replicates results in an independent pilot study of 15 healthy controls who performed this task. In contrast, individuals with ASD showed significant activation only in PPC. Neither group showed activation in frontostriatal circuitry in the 2-choice task. No behavioral differences in the number of errors or reaction times were found for individuals with ASD compared to that of controls. 

Conclusions: These preliminary data indicate that there is reduced frontostriatal network activation during reversal learning in ASD compared to that of age-matched healthy individuals.  These results were specific to the 4-choice condition, in which individuals needed to flexibly search for a new response strategy. The lack of frontostriatal activation during reversal learning between unknown alternatives may reflect a distinct pattern of neural system dysfunction associated with behavioral rigidity in ASD.