Patients with schizophrenia showed a repulsive serial bias in working memory, while patients with bipolar disorder and a history of psychosis showed a mixed pattern that, on average, leaned toward the attractive bias seen in healthy controls, according to a case-control study published in JAMA Psychiatry.
Researchers evaluated 98 participants across five US sites, including 30 patients with schizophrenia, 41 patients with bipolar disorder and a history of psychosis, and 27 healthy controls. Participants completed a working memory task in which they viewed a teardrop-shaped stimulus, retained its orientation across delays of 0, 2, 4, or 8 seconds, and then reproduced that orientation using a computer mouse. The primary outcome was a bias index measuring whether responses shifted toward or away from the previous trial.
Overall bias differed across groups. Healthy controls showed an attractive bias on average, patients with bipolar disorder showed a weaker attractive bias, and patients with schizophrenia showed a repulsive bias.
At the individual level, patterns diverged sharply. Among patients with schizophrenia, 17 of 30 (58.6%) showed repulsive bias and none showed attractive bias. Among healthy controls, 15 of 27 showed attractive bias and none showed repulsive bias. Patients with bipolar disorder showed a mixed distribution, with 13 of 41 showing attractive bias and 7 of 41 showing repulsive bias, while the remainder were indeterminate.
The bipolar disorder group differed clearly from the schizophrenia group in bias patterns. In contrast, differences between the bipolar disorder group and healthy controls did not reach statistical significance, although the results suggested a trend toward weaker attraction in bipolar disorder.
Researchers also examined working memory precision. Patients with schizophrenia had lower precision than both patients with bipolar disorder and healthy controls and showed a steeper drift rate, indicating faster decline in memory precision over time. Patients with bipolar disorder did not differ significantly from healthy controls on either overall precision or rate of decline.
Exploratory analyses suggested that greater repulsive bias in the bipolar disorder group was associated with higher mania scores. In the schizophrenia group, greater repulsion was associated with more severe positive symptoms, although that association did not remain significant after correction for multiple comparisons. Other symptom relationships were not robust.
Analyses also found no meaningful association between antipsychotic medication dose and serial bias, suggesting the observed differences were unlikely to be driven solely by medication effects. In the bipolar disorder group, repulsive bias occurred in patients both taking and not taking antipsychotics, further supporting this finding.
Beyond these descriptive findings, the researchers proposed a mechanistic framework to explain the opposing bias patterns. In healthy individuals, attractive bias is thought to reflect short-term plasticity in prefrontal networks, which pulls current representations toward recent experience. Repulsive bias, in contrast, may arise from neuronal adaptation in the visual cortex, in which prior stimuli suppress responses to similar inputs. The authors suggest that schizophrenia may involve a shift in balance between these processes, with adaptation outweighing plasticity.
This framework may also help explain the heterogeneity observed in bipolar disorder. The presence of a subgroup with repulsive bias similar to schizophrenia raises the possibility that some patients with bipolar disorder share underlying neurobiological features with schizophrenia, rather than differing only in severity.
The findings also point to potential clinical applications. The researchers suggested that serial bias could serve as a biomarker to guide treatment development, with interventions targeting repulsive bias potentially benefiting patients who exhibit that pattern. Such an approach would align with precision medicine strategies by focusing on biologically defined subgroups rather than diagnostic categories alone.
The study had several limitations. Researchers did not collect extensive cognitive measures, limiting their ability to characterize subgroups in greater detail. All patients with schizophrenia were receiving antipsychotic treatment, leaving open questions about medication-naive populations. Larger samples and alternative analytic approaches may provide more detailed insight in future studies.
“Opposite direction serial biases appear to be relatively specific to people with SZ and index a fundamental difference in how recent experience impacts working memory in people with SZ relative to HC individuals and people with BD,” wrote lead study researcher Sonia Bansal, PhD, of the University of Maryland School of Medicine, and colleagues.
The study was supported by National Institute of Mental Health grants and an International Brain Research Organization Rising Star Award. One researcher reported consulting fees and external funding; no other major conflicts were disclosed.
Source: JAMA Psychiatry