A single composite index of semantic behavior tracks symptoms of psychosis over time

Semantic variables automatically extracted from spontaneous speech characterize anomalous semantic associations generated by groups with schizophrenia spectrum disorders (SSD). However, with the use of different language models and numerous aspects of semantic associations that could be tracked, the semantic space has become very high-dimensional, challenging both theoretical understanding and practical applications. This study aimed to summarize this space into a single composite semantic index and to test whether it can track diagnosis and symptom profiles over time at an individual level. The index was derived from a principal component analysis (PCA) yielding a linear combination of 117 semantic variables. It was tested in discourse samples of English speakers performing a picture description task, involving a total of 103 individuals with SSD and 36 healthy controls (HC) compared across four time points. Results showed that the index distinguished between SSD and HC groups, identified transitions from acute psychosis to remission and stabilization, predicted the sum of scores of the Thought, Language and Communication (TLC) index as well as subscores, capturing 65 % of the variance in the sum of TLC scores. These findings show that a single indicator meaningfully summarizes a shift in semantic associations in psychosis and tracks symptoms over time, while also pointing to variance unexplained, which is likely covered by other semantic and non-semantic factors.