Prior Trial Effects on Working Memory in Schizophrenia, Bipolar Disorder, and Healthy Control Individuals

Importance: Healthy individuals typically show attractive serial dependence in working memory tasks, with responses biased toward recent inputs, whereas people with schizophrenia (SZ) reliably show the opposite-repulsive biases. Determining whether individuals with bipolar disorder (BD) and a history of psychosis exhibit similar or distinct serial-dependence patterns is essential for clarifying shared vs disorder-specific cognitive mechanisms across the psychosis spectrum. Objective: To determine whether people with BD exhibit the repulsive bias seen in people with SZ or the attractive bias seen in healthy control (HC) individuals. Design, Setting, and Participants: This case-control study included individuals who met DSM-5 criteria for SZ or BD with a history of psychosis, along with matched HC participants recruited across 5 US sites. Clinical samples were recruited from outpatient psychiatric clinics and day programs, and HC participants were recruited from the community. A working memory task was used in which participants remembered the orientation of a teardrop-shaped object and reproduced it using a computer mouse after a variable delay. The Brief Psychiatric Rating Scale was used to quantify symptom severity. The study was conducted from September 2023 to April 2025. Main Outcomes and Measures: The primary outcome measure was the bias index, used to evaluate the extent to which current-trial responses were biased toward or away from previous-trial targets. Exposures: Diagnostic group and the orientation of the previous trial's target during a spatial working memory paradigm. Results: A total of 41 participants were included in BD group (21 [51%] male; mean [SD] age, 37.02 [11.53] years), 30 in the SZ group (18 [60%] male; mean [SD] age, 37.5 [9.04] years), and 27 in the HC group (16 [59%] male; mean [SD] age, 39.98 [11.01] years). Overall bias differed across groups, with people with SZ (mean, -1.99; 95% CI, -2.74 to 1.23) demonstrating significantly greater repulsive bias than both people with BD (mean, 0.44; 95% CI, -0.25 to 1.13) and HC individuals (mean, 1.82; 95% CI, 1.31 to 2.32). Fifteen of 27 HC individuals exhibited attractive bias, with none showing repulsive bias, whereas 17 of 30 people with SZ exhibited repulsive bias, with none showing attractive bias. Attractive bias was common in people with BD (13 of 41), but some exhibited a repulsive bias (7 of 41). Conclusions and Relevance: In this case-control study, attractive and repulsive serial biases differed between HC individuals and people with SZ, with some people with BD showing the HC pattern and few showing the SZ pattern. These results suggest that serial bias may serve as a valuable biomarker for pathophysiology and precision psychiatry.