Objective:

To investigate the dynamic temporal trajectory of contextual coherence breakdown in disorganized speech among patients with schizophrenia using autoregressive language models, highlighting the significance of this approach.

Key Findings:

• Patients exhibited progressively divergent surprisal trajectories as discourse unfolded, indicating a breakdown in contextual predictability.
• Deviations in surprisal intensified between tokens 50 to 70, reflecting rapid deterioration in sustaining global contextual constraints.
• Higher overall mean surprisal was detected in patients, particularly during unstructured narrative tasks, suggesting greater difficulty in maintaining coherence.

Interpretation:

The findings provide quantitative evidence that language anomalies in schizophrenia are dynamic phenomena that rapidly unfold within a short discourse, aligning with existing literature on cognitive processing in schizophrenia.

Limitations:

• The study focused on a specific linguistic feature and may not capture all aspects of disorganized speech.
• The sample was limited to Korean-speaking individuals, which may affect generalizability and introduce potential biases in selection.

Conclusion:

This research highlights the utility of temporal natural language processing metrics in elucidating the psychopathology of schizophrenia, with potential applications in clinical assessment and intervention.