Neural Mechanisms of Social Prediction Errors in Adolescents with ASD

Background: This study investigated the neural mechanisms of social prediction errors using functional magnetic resonance imaging (fMRI) in adolescents with and without autism spectrum disorder (ASD). Prediction errors occur when outcomes differ from expectations and are critical to reward learning. When expectations are violated, predictions are updated to maximize the potential for future rewards (Schultz, 2015). Prediction errors are reflected in neuronal activity in the mesolimbic dopaminergic system when better-than-expected and worse-than-expected rewards are received (Daniel & Pollmann, 2014; Schultz, 2015). Social prediction errors occur when expectations about social outcomes are violated, facilitating the updating of future predictions about social reward (Ruff & Fehr, 2014). It is possible that disruptions in neural mechanisms of social prediction errors contribute, in part, to social interaction deficits of individuals with ASD. Although there is growing functional neuroimaging literature addressing reward processing deficits in ASD (Chevallier, Kohls, Troiani, Brodkin, & Schultz, 2012), most studies have addressed reward anticipation or receipt, but almost none have addressed social prediction errors in ASD. It is critical to examine social prediction errors in ASD, since this neural system is critical for social learning.

Objectives: To compare blood oxygen level dependent (BOLD) signals among adolescents with and without ASD during a social prediction error task.

Methods: The sample included 41 12-to-17-year-olds: 21 with high-functioning ASD and 20 with typical development (TD). Prior to the scan, participants were taught to press a “check” every time they saw a symbol (i.e., “cue 1”) that predicted a clear image and an “x” every time they saw a different symbol (i.e., “cue 2”) that predicted a blurry image. In the scanner, participants saw images of smiling faces (i.e., social reward) or blurry faces (i.e., non-reward) across two 7-minute runs. In 20% of trials, the images violated expectations (i.e., “cue 1” predicted a blurry face, rather than a clear face as expected; “cue 2” predicted a clear rather than blurry face). Analyses focused on group differences with respect to the contrast between expected and unexpected reward outcomes, which are known to elicit prediction error signals in the midbrain, striatum, and prefrontal cortex (Ramnani, Elliott, Athwal, & Passingham, 2004),

Results: Data collection is complete and analysis is ongoing. Preliminary analyses found that the TD group demonstrated increased activation in the right inferior frontal gyrus compared to the ASD group when expectations were violated, whereas the ASD group demonstrated increased activation in the right frontal pole. When outcomes were as expected, the TD group demonstrated increased activation in the right frontal pole, whereas the ASD group exhibited increased activation in the left inferior frontal gyrus. Ongoing analyses will examine functional connectivity differences between groups, as well as relations between brain imaging metrics, task related behavior, and ASD symptom severity.

Conclusions: Preliminary results indicate differential activation in prefrontal cortical regions that are critical for eliciting prediction error signals in adolescents with ASD. These differences may provide a mechanistic account of disrupted social reward learning in ASD and contribute to the literature addressing reward processing deficits in ASD.