Characterisation of alogia in people with schizophrenia using natural language processing. a comparative study with a...

Ninety participants were recruited for the study, distributed equally in three groups: schizophrenia (SZ), bipolar disorder (BD), and healthy controls (HC). All participants described scenes depicting people interacting with each other in everyday social situations, and their cognitive functioning was assessed using the screen for cognitive impairment in Psychiatry (SCIP). Clinical status was assessed with the Brief Psychiatric Rating Scale (BPRS). Speech samples were explored with metrics that quantify speech production and complexity.

People with schizophrenia exhibited lower verbal production and reduced sentence complexity compared to healthy controls, while the bipolar disorder group showed values largely overlapping with controls. Across the pooled patient sample (SZ + BD), negative symptom severity was associated with multiple linguistic measures, including mean length of utterance (r = -0.384, BF₁₀ = 14.03, 95% CI [-0.571, -0.138]), syntagms, discourse complexity, and lexical-morphological indices, whereas positive symptoms showed little or no association with language metrics. Better cognitive functioning was associated with greater linguistic richness across most measures.

The findings from this study suggest that poverty of speech, as captured by NLP-derived measures, is more prominent in schizophrenia than in bipolar disorder and healthy controls within the present sample. The observed associations with cognitive performance highlight the close relationship between language production and cognition, but do not establish a causal dependence. Rather than prompting a re-examination of alogia's classification among negative symptoms, these results underscore the need for integrative models that account for the well-known covariation between negative symptoms and cognitive functioning.