Note: Most Internet Explorer 8 users encounter issues playing the presentation videos. Please update your browser or use a different one if available.

Probabilistic Reinforcement Learning in Young Adults with Autism Spectrum Disorders Reflects Cognitive Control Deficits

Thursday, 2 May 2013: 09:00-13:00
Banquet Hall (Kursaal Centre)
M. Solomon1, M. J. Frank2, A. C. Smith3, J. D. Ragland4,5, M. J. Minzenberg3, T. A. Niendam6,7, J. H. Yoon5,8, T. A. Lesh6,9 and C. S. Carter5,6, (1)Department of Psychiatry, MIND Institute, Imaging Research Center, Sacramento, CA, (2)Brown University, Providence, RI, (3)UC Davis, Sacramento, CA, (4)Psychiatry, UC Davis, Sacramento, CA, (5)Imaging Research Center, Sacramento, CA, (6)Department of Psychiatry & Behavioral Sciences, University of California-Davis, Sacramento, CA, Sacramento, CA, (7)Imaging Research Center, Sacramentp, CA, (8)UC Davis, Psychiatry, Sacramento, CA, (9)Psychiatry, UC Davis, Sacramentp, CA
Background: Individuals with autism spectrum disorders (ASD) display a unique pattern of learning strengths and challenges. They show intact (or even enhanced) lower-level learning of stimulus response associations, of single items of information, of facts, of habits, and of information learned implicitly. However, they display deficits in generalizing (or transferring) what they have learned during training to new similar situations. Generalization problems have a profound impact on the academic, social, and adaptive functioning of persons with ASD, and have not been well studied.

Objectives: The goal of the current research is to illuminate the neural mechanisms of learning differences in persons with ASD using functional magnetic resonance neuroimaging (fMRI), and to translate findings into clinically relevant insights.

Methods: Participants included young adults aged 18-40 years with diagnosed with ASD, who were largely medication free and evaluated using gold standard diagnostic measures (n=22), and age, IQ, and gender matched young adults with typical development (TYP; n=25). They completed a probabilistic selection task where they were trained to choose the correct stimulus in three different stimulus pairs (AB, CD, EF) presented with feedback that was valid 80%, 70%, and 60% of the time, respectively. Participants had to determine which was the rewarded stimulus from this relatively unpredictable feedback. Whole brain voxel-wise analyses were conducted using Bayesian state-space learning curves and prediction errors as parametric modulators. Regions of interest in the striatum, prefrontal cortex (PFC), and medial temporal lobes, were interrogated and functional connectivity analyses using these regions as seeds were conducted.    

Results:  As in our previous behavioral study, individuals with ASD learned the task to comparable rates as TYP, but were slower to learn. The ASD group exhibited greater striatal and medial temporal lobe activation during early learning that was related to task performance. There was less activation in prefrontal regions throughout learning in the ASD group, and unlike the TYP group, activation in the orbito-frontal cortex (OFC) was not related to the probability of having learned. Activation of the striatum during early learning was positively associated with restricted interests and repetitive behaviors in the ASD group.

Conclusions: Overall, results suggest that those with ASD have cognitive control related learning deficits. They use prefrontal brain regions including the OFC less, and rely on striatal, and medial temporal lobe regions to a greater extent, suggesting the use of these brain regions may be compensatory. Consequently they rely on a rote learning-based strategy as opposed to a more flexible one that can incorporate rapid updating of reward contingencies, and integrate this information in the service of goal directed behavior. There may be a relationship between repetitive behavior symptoms and their learning style. This interpretation is supported by the systems-level computational modeling work of Frank et al. (2004, 2005, 2006).

| More