Real-world conversations across the schizophrenia spectrum: Implementing passive audio sensing to examine linguistic style matching.

Abstract

Social dysfunction is a hallmark of schizophrenia. Experience sampling has facilitated the investigation of daily socialization to detect dysfunction and identify treatment targets. Yet, poor cognition and insight in schizophrenia interfere with subjective self-report. Passive audio sensing and lexical analysis offer a solution to examine objective markers of real-world functioning. Linguistic style matching (LSM), the tendency to coordinate speech patterns with conversation partners, has been linked to social affiliation and connection in healthy populations. It has been theorized that LSM may be disrupted across the schizophrenia spectrum, indicating underlying psychopathology. Our two-part, proof-of-concept study leverages passive audio sensing to measure LSM during real-life conversations and quantify its relationship with interaction depth coded by trained raters. Data were from 2013 to 2019. Study 1 compared LSM between 31 people high in schizotypy and 26 low in schizotypy, finding no significant differences. In Study 2, we observed that those with schizophrenia (n = 28) exhibited reduced LSM compared to controls (n = 26). Across studies, LSM was associated with real-world interaction depth (i.e., level of self-disclosure during conversation). The results support the feasibility of capturing daily LSM via passive sensing and suggest LSM may reflect real-world interaction depth that can differentiate people with and without schizophrenia. Consequently, LSM may be a useful marker of social dysfunction in psychopathology. However, this does not seem to extend to subclinical schizotypy, where social affiliative deficits are more subtle. Given the automation and ease of lexical analysis, these findings are promising for the future of passive sensing of social interactions. (PsycInfo Database Record (c) 2025 APA, all rights reserved).