Computational Modeling of Reversal Learning Impairments in Schizophrenia and Bipolar Disorder Reveals Shared Failure to Exploit Rewards

Abstract

The distinction between the concepts of schizophrenia and bipolar disorder is fundamental to the Kraepelinian tradition in psychiatry. One mechanism undergirding this distinction, a difference in reward sensitivity, has been championed by a number of scholars. As part of the Cognitive Neuroscience Test Reliability and Clinical applications for Serious mental illnesses consortium, 225 participants including people with schizophrenia (n = 69), schizoaffective disorder (n = 55), and bipolar affective disorder (n = 53) performed a probabilistic reversal learning task. This task switches the rewarded stimulus at various times throughout the task. Our analyses leveraged a Hidden Markov Model to examine trial-by-trial decisions of participants to reveal the differences between patient groups in their response to reward feedback. Whereas no patient group showed difficulty reversing their preferred categories after a switch in the task’s contingencies and bipolar patient performance was spared in some other ways, all patient groups made more errors throughout the task because of a greater tendency to shift away from rewarded categories (i.e., win-switching). Furthermore, patients’ cognitive ability is specifically related to this aspect of the task. Rather than validating a Kraepelinian dichotomy, these findings suggest that a failure to exploit rewards may reflect a mechanistic deficit common across both affective and nonaffective psychoses related to cognitive impairments in patients.