Altered Reinforcement Learning from Reward and Punishment in Anorexia Nervosa: Evidence from Computational Modeling

Area of Science:

• Neuroscience
• Psychiatry
• Behavioral Science

Background:

• Anorexia nervosa (AN) is linked to altered reward and punishment sensitivity.
• Cognitive inflexibility and decision-making deficits are common in AN.
• Deficient reinforcement learning may perpetuate maladaptive behaviors in AN.

Purpose of the Study:

• To investigate learning processes in individuals with AN using a probabilistic associative learning task.
• To compare learning from reward versus punishment in AN and healthy controls (HCs).
• To explore the relationship between learning parameters and body mass index (BMI) at discharge.

Main Methods:

• A probabilistic associative learning task was used to assess learning from positive and negative outcomes.
• Computational reinforcement learning models identified prediction errors (PEs) in AN and HCs.
• Linear regression analyzed the predictive power of learning parameters on discharge BMI.

Main Results:

• Individuals with AN demonstrated lower learning rates than HCs following both positive and negative PEs.
• AN participants were less likely to utilize learned information.
• Negative PEs on punishment trials predicted lower discharge BMI in AN.

Conclusions:

• This study reveals diminished learning rates in AN for both positive and negative outcomes.
• Impaired punishment learning correlates with reduced weight gain, suggesting a role for negative expectancy in persistent food avoidance.
• Interventions targeting negative expectancy modification may be beneficial for treating AN.