M227. AN EXPLORATORY STUDY OF AUTOMATED LINGUISTIC ANALYSIS OF SOCIAL NARRATIVES IN SCHIZOPHRENIA

Abstract

BackgroundSchizophrenia has been suggested to be a disorder of social communication, which depends on the way language is used to convey thoughts, beliefs, feelings, and intentions. Everyday language can also reveal personality, emotions, and social skills of the speaker. Extensive past research affirms the central importance of language and thought disorder as diagnostic features of schizophrenia, mostly focused on the neurocognitive aspects of language output collected during clinical interviews, and not on the social nature of language. In this study, we examined narratives written in response to viewing social scenes by individuals with schizophrenia (SZ) and matched controls (CO) using an automated computational linguistics and statistical-based text analysis tool that computes socially-relevant variables.Methods23 individuals with schizophrenia (SZ) and 23 demographically matched controls (CO) were shown paintings of social situations, and were asked to write reflections describing what they thought and felt about these scenes. Two pictures were presented consecutively. There was no time limit. Resulting narratives were analyzed with the Linguistic Inquiry and Word Count program (LIWC; Pennebaker et al, 2015). LIWC computes basic linguistic variables such as the % of self-referring and non-self-referring pronouns, social and emotion words, and cognitive items. LIWC also generates 4 complex variables: formal and logical thinking patterns (“analytic”); social status or confidence (“clout”); authenticity, and emotional tone. Clinical symptoms in SZ were assessed using the Scale for the Assessment of Positive Symptoms (SAPS) and the Scale for the Assessment of Negative Symptoms (SANS). For all participants, the UCLA Loneliness Scale, the National Adult Reading Test (NART), and the Edinburgh Handedness Inventory (EHI) were administered.ResultsThe two groups did not differ in NART or EHI. There was no group difference in the number of words produced. SZ produced greater number of pronouns than CO but this effect was driven by a higher % of self-referring pronouns in SZ than CO, with no group difference in non self-referring pronouns. For complex variables, CO scored significantly higher in authenticity than SZ but no group differences were observed in clout, analytics, or emotional tone. Loneliness was higher in SZ compared with CO. There were no other significant group differences. In SZ, pronoun use was correlated with positive symptoms (especially with ratings of hallucinations, bizarre behavior, delusions of mind-reading and thought broadcasting) and inversely correlated with negative symptoms (avolition, apathy and motor retardation). Social words were inversely correlated with SAPS Thought Disorder. Clout was inversely correlated with SANS Alogia and SAPS Thought Disorder. Authenticity was correlated with SANS Anhedonia and Asociality. In CO, loneliness was correlated with the % negative emotion words and NART was correlated with total number of emotion words.DiscussionWe used an automated linguistic analysis tool to extract information relevant to social communication from written narratives. We found group differences in the use of pronouns and authenticity. We also observed associations of clinical symptoms with certain social aspects of language use in schizophrenia. One advantage of automated text analysis tools is the minimization of implicit biases inherent in ratings of interviews. Limitations of this study include lack of direct social functioning measures and the sample size. Future work will incorporate linguistic text analysis within a social paradigm to directly examine the role of language use in social functioning.