Aberrant reward processing to positive versus negative outcomes across psychotic disorders

Abstract

Several studies of reward processing in schizophrenia have shown reduced sensitivity to positive, but not negative, outcomes although inconsistencies have been reported. In addition, few studies have investigated whether patients show a relative deficit to social versus nonsocial rewards, whether deficits occur across the spectrum of psychosis, or whether deficits relate to negative symptoms and functioning. This study examined probabilistic implicit learning via two visually distinctive slot machines for social and nonsocial rewards in 101 outpatients with diverse psychotic disorders and 48 community controls. The task consisted of two trial types: positive (optimal to choose a positive vs. neutral machine) and negative (optimal to choose a neutral vs. negative machine), with two reward conditions: social (faces) and nonsocial (money) reward conditions. A significant group X trial type interaction indicated that controls performed better on positive than negative trials, whereas patients showed the opposite pattern of better performance on negative than positive trials. In addition, both groups performed better for social than nonsocial stimuli, despite lower overall task performance in patients. Within patients, worse performance on negative trials showed significant, small-to-moderate correlations with motivation and pleasure-related negative symptoms and social functioning. The current findings suggest reward processing disturbances, particularly decreased sensitivity to positive outcomes, extend beyond schizophrenia to a broader spectrum of psychotic disorders and relate to important clinical outcomes.

Keywords: Reinforcement learning; Reward learning; Reward sensitivity; Schizophrenia; Social rewards.

Fig. 1.

A schematic diagram of the One-Armed Bandit Task. (A) There are 3 types of slot machines. A “good” slot machine had an 80% probability of a positive outcome and a 20% probability of a neutral outcome; a “bad” slot machine had an 80% probability of a negative outcome and a 20% probability of a neutral outcome; and a “neutral” slot machine had a one-third probability of each positive, neutral, and negative outcomes. (B) Each reward condition (nonsocial vs. social) with an identical trial structure consisted of the 2 trial types: a high-payout trial and a low-payout trial. For the high-payout trial, a good slot machine was paired with a neutral slot machine. For the low-payout trial, a bad slot machine was paired with a neutral slot machine.

Fig. 2.

Reward processing of positive and negative outcomes in patients with psychotic disorders and controls on the One-Armed Bandit Task, collapsing across condition types (i.e., social vs nonsocial stimuli). The y-axis represents the proportion of trials with the optimal outcome (i.e., choosing a neutral machine over a bad machine during the low-payout trials or choosing a good machine over a neural machine during the high payout trials). Error bars ± 1 SE.