Section Structure
1. Clinical context and rationale
2. Study design and methods
3. Main findings
4. Mechanistic and translational interpretation
5. Clinical relevance and limitations
6. Conclusion
7. Funding, registration, and citation
8. References
Clinical context and rationale
Cognitive impairment is a core feature of schizophrenia and is also common in bipolar disorder, particularly in individuals with psychosis. Among the most reproducible impairments in schizophrenia is reduced working memory precision, but cognitive deficits are not limited to simple capacity or accuracy measures. Increasingly, psychiatry is turning toward computationally informed behavioral markers that may capture how the brain integrates information over time. One such phenomenon is serial dependence, the tendency for responses on a current trial to be biased by stimuli encountered on the immediately preceding trial.
In healthy observers, serial dependence during perception and working memory is typically attractive: the current response is drawn toward the recent past. This is thought to promote perceptual continuity and stabilize representations in noisy environments. In schizophrenia, prior work has shown the opposite pattern in visual working memory tasks: repulsive serial dependence, in which current responses are pushed away from the previous stimulus. That reversal is theoretically important because it may reflect altered short-term synaptic plasticity, abnormal adaptation, or disturbed integration of past and present information.
The unresolved question is where bipolar disorder with psychosis fits within this framework. Bipolar disorder and schizophrenia share some clinical, genetic, and neurocognitive features, yet they also diverge in course, phenomenology, and likely neurobiology. If bipolar disorder with psychosis showed the same repulsive bias as schizophrenia, serial dependence could be viewed as a transdiagnostic psychosis-related process. If, instead, bipolar disorder more closely resembled healthy individuals, serial dependence might help dissociate disorder-specific mechanisms within the psychosis spectrum. The study by Bansal and colleagues directly addresses this issue.
Study design and methods
This was a multisite case-control study conducted across 5 US sites from September 2023 to April 2025. The investigators enrolled three groups: individuals with schizophrenia meeting DSM-5 criteria, individuals with bipolar disorder who had a history of psychosis, and matched healthy control participants recruited from the community. Clinical participants were recruited from outpatient psychiatric clinics and day programs.
The final sample included 41 participants in the bipolar disorder group, 30 in the schizophrenia group, and 27 healthy control individuals. The bipolar disorder group included 21 men (51%) and had a mean age of 37.02 years (SD, 11.53). The schizophrenia group included 18 men (60%) and had a mean age of 37.5 years (SD, 9.04). The healthy control group included 16 men (59%) and had a mean age of 39.98 years (SD, 11.01).
Participants completed a spatial working memory task involving the orientation of a teardrop-shaped object. After a variable delay, they reproduced the remembered orientation using a computer mouse. The key analytic variable was the bias index, which quantified whether responses on the current trial were drawn toward the previous trial’s target orientation or pushed away from it. Positive values indicate attractive bias, whereas negative values indicate repulsive bias. Symptom severity was measured using the Brief Psychiatric Rating Scale.
Methodologically, the study is notable because it does not treat working memory performance solely as a matter of error magnitude. Instead, it interrogates the directionality of error relative to previous stimuli. This approach is potentially more informative mechanistically because attractive and repulsive biases can arise from different neural processes, including persistent activity, adaptation, and synaptic carryover effects.
Main findings
The central result was a clear group difference in serial dependence. Participants with schizophrenia showed a significantly more repulsive bias than both the bipolar disorder group and healthy controls. Mean bias was -1.99 in schizophrenia (95% CI, -2.74 to -1.23), compared with 0.44 in bipolar disorder (95% CI, -0.25 to 1.13) and 1.82 in healthy controls (95% CI, 1.31 to 2.32).
The healthy control group demonstrated the expected normative pattern. Fifteen of 27 healthy participants exhibited attractive bias, and none showed repulsive bias. At the group level, this supports the idea that recent inputs ordinarily stabilize working memory representations rather than distort them in a direction away from prior experience.
The schizophrenia group showed a markedly different profile. Seventeen of 30 participants exhibited repulsive bias, and none showed attractive bias. This striking polarity reversal replicates and extends prior evidence that schizophrenia is associated with an abnormal influence of recent trial history on working memory. Importantly, the effect is not simply “more noise” or poorer accuracy. It is a directional distortion, suggesting a qualitatively different computation.
The bipolar disorder group occupied an intermediate position, but not merely as a simple average between the other two groups. Mean bias in bipolar disorder was slightly positive and the confidence interval crossed zero, implying heterogeneity rather than a uniform pattern. Attractive bias was common in bipolar disorder, seen in 13 of 41 participants, but repulsive bias was also present in 7 of 41. This mixed distribution is clinically and theoretically important. It suggests that bipolar disorder with psychosis is not fully aligned with schizophrenia on this measure, yet neither is it entirely indistinguishable from healthy individuals.
From a translational standpoint, this heterogeneity may be the most informative finding in the paper. It raises the possibility that serial dependence could map onto biologically meaningful subgroups within bipolar disorder, perhaps related to psychosis burden, mood state history, medication exposure, cognitive reserve, or underlying network physiology. The abstract does not indicate that these subgroup analyses were resolved, but it strongly motivates future work.
Mechanistic and translational interpretation
Why would healthy individuals show attraction to the recent past, whereas schizophrenia is associated with repulsion? One influential framework proposes that attractive serial dependence reflects the brain’s attempt to exploit temporal continuity in the environment. Since objects and contexts usually change gradually rather than randomly, leaning slightly toward recent inputs can improve stability and reduce the effect of momentary noise. This has been described across visual features such as orientation, numerosity, and face identity.
Repulsive effects, by contrast, are often linked to sensory adaptation or gain changes that push current representations away from recently encoded features. In normal cognition, adaptation and attractive serial dependence may coexist over different timescales, with the net observed effect depending on task demands and delay. In schizophrenia, the balance may be altered. Prior work has suggested weaker short-term synaptic potentiation or abnormal maintenance-related dynamics in cortical circuits, potentially allowing adaptation-like repulsion to dominate over attraction.
This study’s findings in bipolar disorder with psychosis suggest a more complex landscape. The presence of both attractive and repulsive patterns implies that the disorder may encompass multiple cognitive phenotypes rather than one unitary mechanism. Some individuals with bipolar disorder may preserve the normative temporal smoothing seen in healthy controls, while a subset may share aspects of the aberrant computation observed in schizophrenia. This aligns with broader literature showing that cognitive dysfunction in bipolar disorder is real but variable, and often less severe and less uniform than in schizophrenia.
The idea of serial dependence as a biomarker is appealing, but it should be interpreted carefully. A useful biomarker in psychiatry should ideally be reliable, mechanistically grounded, and linked to clinically relevant outcomes. This study strengthens the mechanistic case, particularly for schizophrenia, but does not yet establish prognostic value, treatment responsiveness, or diagnostic utility at the individual level. The bipolar disorder findings, precisely because they are heterogeneous, may ultimately be more valuable for precision psychiatry than for traditional case-control classification.
Clinical relevance and limitations
For clinicians, the immediate implication is not that serial dependence testing is ready for office-based diagnosis. Rather, the study provides evidence that subtle trial-history effects in working memory may capture aspects of pathophysiology not visible on conventional neuropsychological testing. If replicated and refined, such measures could complement symptom-based classification and help identify biologically coherent subgroups across psychotic disorders.
The study has several strengths. It used a clearly defined behavioral paradigm, included healthy and disease comparison groups, recruited across multiple US sites, and focused on a theoretically informative primary outcome rather than exploratory data mining. The findings are also internally coherent: healthy controls showed attraction, schizophrenia showed repulsion, and bipolar disorder showed a mixed distribution.
At the same time, several limitations deserve emphasis. First, the sample size was modest, especially for subgroup analyses or for determining whether medications, illness duration, psychosis history, or symptom severity account for the heterogeneity in bipolar disorder. Second, the bipolar disorder sample required a history of psychosis, which makes the findings especially relevant to psychotic bipolar disorder but may limit generalizability to bipolar disorder without psychotic features.
Third, medication effects are an important possible confounder in studies of psychosis-spectrum cognition. Antipsychotics, mood stabilizers, benzodiazepines, and antidepressants could plausibly alter attention, working memory precision, or adaptation-like processes. The abstract does not detail medication-adjusted analyses. Fourth, because this is a case-control design, the study cannot determine whether repulsive serial dependence predates illness onset, changes across mood or psychotic states, or responds to treatment.
There is also a conceptual limitation worth noting. Serial dependence is influenced by task structure, delay duration, stimulus domain, and analytic approach. A biomarker based on this phenomenon will need strong test-retest reliability and cross-paradigm robustness before it can be translated into broader clinical use. Nonetheless, the distinct pattern in schizophrenia, especially the absence of attractive bias in that group, is compelling.
These results also sit within a wider cognitive-neuroscience literature. Prior studies from Gold, Stein, and colleagues have reported repulsive serial dependence in schizophrenia and have linked it to altered working memory mechanisms. More broadly, foundational work by Fischer and Whitney and by Bliss and colleagues established attractive serial dependence as a normative property of perception and memory. The present study extends that literature by positioning bipolar disorder with psychosis between healthy control and schizophrenia groups rather than assuming a single psychosis-spectrum signature.
Conclusion
Bansal and colleagues provide clinically relevant evidence that prior-trial effects on working memory differ across schizophrenia, bipolar disorder with psychosis, and healthy individuals. Healthy controls showed the expected attractive serial dependence, schizophrenia showed robust repulsive bias, and bipolar disorder exhibited a heterogeneous mix of both patterns. The findings argue against a simple one-size-fits-all cognitive abnormality across psychotic disorders and instead support a more differentiated model of psychosis-spectrum neurobiology.
The most important translational message is that serial dependence may be useful not merely as a group difference but as a candidate precision marker. In schizophrenia, it appears to index a relatively consistent abnormality in how recent information influences current working memory. In bipolar disorder, the mixed pattern may help reveal subgroups that are obscured by traditional diagnoses. The next steps should include longitudinal studies, medication-aware analyses, state-versus-trait assessment, integration with neurophysiology and imaging, and evaluation of whether these biases relate to functional outcomes or treatment response.
For now, this study strengthens the emerging view that the direction of memory error matters. In psychosis research, not all inaccuracies are equivalent. Whether cognition is pulled toward the recent past or pushed away from it may offer a window into disease mechanisms that standard accuracy scores alone cannot provide.
Funding, registration, and citation
Funding information was not provided in the source abstract. No ClinicalTrials.gov registration number was reported in the abstract, which is not unusual for a behavioral case-control study that is not an interventional clinical trial.
Citation: Bansal S, Bae GY, Barch DM, Erickson M, Carter CS, MacDonald AW, Luck SJ, Gold JM. Prior Trial Effects on Working Memory in Schizophrenia, Bipolar Disorder, and Healthy Control Individuals. JAMA Psychiatry. 2026 Jun 1;83(6):630-639. PMID: 41984461. URL: https://pubmed.ncbi.nlm.nih.gov/41984461/
References
1. Bansal S, Bae GY, Barch DM, Erickson M, Carter CS, MacDonald AW, Luck SJ, Gold JM. Prior Trial Effects on Working Memory in Schizophrenia, Bipolar Disorder, and Healthy Control Individuals. JAMA Psychiatry. 2026;83(6):630-639. PMID: 41984461.
2. Fischer J, Whitney D. Serial dependence in visual perception. Nat Neurosci. 2014;17(5):738-743. doi:10.1038/nn.3689.
3. Bliss DP, Sun JJ, D’Esposito M. Serial dependence is absent at the time of perception but increases in visual working memory. Sci Rep. 2017;7:14739. doi:10.1038/s41598-017-15199-7.
4. Stein H, Barbosa J, Rosa-Justicia M, et al. Reduced serial dependence suggests deficits in synaptic potentiation in anti-NMDAR encephalitis and schizophrenia. Nat Commun. 2020;11(1):4250. doi:10.1038/s41467-020-18033-3.
5. Barch DM, Sheffield JM. Cognitive impairments in psychotic disorders: common mechanisms and measurement. World Psychiatry. 2017;16(1):32-40. doi:10.1002/wps.20372.
