Linguistic determinants of formal thought disorder in first episode psychosis

Abstract

Aim Thought disorder is a core feature of schizophrenia but assessment of disordered thinking is challenging, which may contribute to the paucity of mechanistic understanding of disorganization in early psychosis. We studied the use of linguistic connectives in relation to clinically quantified dimensions of thought disorder using automated speech analysis in untreated, first episode psychosis (FEPs) and healthy controls (HCs). Methods 39 treatment‐naïve, actively psychotic FEPs and 23 group matched HCs were recruited. Three one‐minute speech samples were induced in response to photographs from the Thematic Apperception Test and speech was analysed using COH‐METRIX software. Five connectives variables from the Coh‐Metrix software were reduced using principle component analysis, resulting in two linguistic connectives factors. Thought disorder was assessed using the Thought Language Index (TLI) and the PANSS‐8. Results Connective factors predicted disorganization, but not impoverishment suggesting aberrant use of connectives is specific to positive thought disorder. An independent t test comparing low and high disorganization FEPs showed higher load of acausal temporal connectives in high disorganization FEPs compared to low disorganization FEPs (mean [SD] in high vs low disorganization FEPs = 0.64 (1.1) vs ‐0.37 (1.02); t = 2.91, P = .006). Acausal‐temporal connectives were not correlated with severity of symptoms or cognition suggesting connective use is a specific index of disorganized thinking rather than overall illness status. Conclusions Clinical assessment of disorganization in psychosis is likely linked to the aberrant use of connectives resulting in an intuitive sense of incoherence. In early psychosis, thought disorder may be reliably quantifiable using automated syntax analysis.

Conceptual disorganization and redistribution of resting-state cortical hubs in untreated first-episode psychosis: A 7T study

“…Nevertheless, our results cannot be taken to represent the neural basis of FTD as such, as we lacked instruments that comprehensively quantify the various aspects of FTD (see Supplementary Note 1). We recommend multidimensional assessment in a future studies to capture the linguistic as well as clinical aspects of FTD, as they may not fully overlap 46 . We had disproportionately fewer female participants; caution is required when generalizing our results in this regard.…”

Natural language processing methods are sensitive to sub-clinical linguistic differences in schizophrenia spectrum disorders

“…Despite affective flattening being a common negative symptom in SSD, use of emotion words were nevertheless similar between SSD and control participants 22 . Cohesion was not significantly reduced in another study comparing first-episode psychosis and healthy control (HC) participants 23 . Multiple levels of acoustic and linguistic analysis were shown to have a poor correlation with clinician ratings for negative symptoms 7 .…”

Natural Language Processing markers in first episode psychosis and people at clinical high-risk

“…We focussed on 12 NLP measures but there are many more that may show significant group differences, e.g. pronoun incidence [ 47 ].…”