Probabilistic reinforcement learning in adults with autism spectrum disorders

Area of Science:

• Neuroscience
• Cognitive Psychology
• Developmental Psychology

Background:

• Autism spectrum disorders (ASDs) are increasingly viewed as learning disorders.
• However, experimental research from this perspective remains limited.

Purpose of the Study:

• To investigate probabilistic reinforcement learning in adults with ASDs.
• To compare learning performance between individuals with ASDs and typically developing adults.

Main Methods:

• A probabilistic reinforcement learning task involving Japanese characters and varying feedback probabilities (80%, 70%, 60%).
• Involved 28 adults with ASDs and 30 typically developing adults.
• Employed univariate and Bayesian state-space data analytic methods.

Main Results:

• Both groups learned the task, but differences emerged in early learning phases.
• Individuals with ASDs showed comparable acquisition for the 80% reinforced pair but poorer acquisition for the 70% pair.
• ASD group outperformed controls on the 60% pair and showed deficits in using positive feedback for exploitation.

Conclusions:

• Findings suggest individuals with ASDs may be slower learners.
• A potential explanation involves impaired flexible updating of reinforcement history, possibly mediated by the orbito-frontal cortex, with intact basal ganglia function.
• This hypothesis can be further investigated using functional magnetic resonance imaging (fMRI).