Language-based inference generation under working memory load: the role of schizotypal traits in jumping to conclusions

Background and hypothesis: Individuals with schizophrenia often exhibit language abnormalities and impairments in Theory of Mind (ToM). These difficulties may stem from underlying cognitive processes such as a tendency to jump to conclusions (JTC), making decisions without sufficient external evidence, and disruptions in circular inference, which can produce atypical beliefs, impaired probabilistic decision-making, and heightened perceptions of visual ambiguity. However, it remains unclear whether similar impairments occur in healthy individuals who display non-clinical schizotypal personality traits. Study design: The present study examined JTC through inference generation and its association with schizotypal traits. A total of 532 participants completed the Schizotypal Personality Questionnaire-Brief Revised (SPQ-BR) and were audio-recorded while narrating a nine-frame comic strip. A between-subjects working memory (WM) manipulation was used to assess the effect of cognitive load. Inference generation was evaluated by independent raters who manually annotated the speech transcripts. Study results: WM load reliably increased the number of inferred events produced, whereas schizotypal traits alone were not significantly associated with inferred events. Instead, WM load moderated this relationship: disorganized traits predicted more inferred events only under WM load. For visual events, disorganized traits demonstrated a quadratic association, and WM load again moderated this pattern, with quadratic effects emerging only when WM load was absent. Overall, WM load played a central role in shaping how disorganized schizotypal traits related to both inferred and visual event production. Conclusions: These findings indicate that increased JTC tendencies can emerge even in the absence of clinically significant schizophrenia symptoms and that disorganized traits may contribute to disruptions in circular inference mechanisms. The results also underscore the utility of analyzing speech production as a method for investigating inference generation in future research.