Abstract / Summary
Background. Individuals with schizophrenia (SZ) and bipolar disorder (BD) often misjudge whether social cues, like eye contact, are meant for them, which can fuel symptoms and functional impairment. These difficulties appear to stem from disruptions in how social information is dynamically processed (“evidence accumulation”). We used neurocognitive joint modeling to investigate which neural signals shape evidence accumulation during gaze perception and whether these brain-behavior links help explain disrupted self-referential social processing.Methods. Leveraging EEG data from N=99 SZ, BD, and healthy controls (HC) during a self-referential gaze perception task, we built joint drift-diffusion models characterizing how different trial-level EEG features translated to evidence accumulation and behavior. Joint models were compared against each other (and models using task-averaged EEG or behavior only) for their ability to predict task behavior and relevant clinical dimensions, including paranoia, functioning, and cognition. We examined whether the best-performing model differentiated patients and HC.Results. The best-performing joint model explained evidence accumulation through trial-level fluctuations—not task averages—in two neural signals: CPP and frontal midline theta, implicated in subjective deliberation/certainty and top-down modulation, respectively. Less reliable translation of trial-level CPP signals to evidence accumulation impaired gaze perception in SZ and BD. Mapping how trial-level theta dynamics translated to evidence accumulation captured paranoia and functioning better than behavior or task-averaged neural metrics alone.Conclusions. Model-based approaches that embrace the within-person variability of neural signals may reveal more nuanced, clinically relevant perspectives on the pathophysiology of SZ and BD than those using task-averaged brain activity or behavior alone.