Data-driven classification of narrative speech characteristics in stroke aphasia distinguishes neurological and...

Narrative speech deficits are common in post-stroke aphasia, resulting in negative influences on social participation and quality of life. Speech rate, complexity, and informativeness deficits all contribute to narrative speech. Research studies typically (implicitly) assume that these aspects of narrative speech production are a result of cognitive/neurological impairment, but they may also result from strategic choices made as individuals with aphasia attempt to produce narrative speech. Here, we used data-driven methods to classify aphasic narrative speech patterns and evaluated their predictability from lesion patterns. 76 stroke aphasia patients completed 11 narrative speech production tasks. Quantitative Production Analysis (QPA) and Correct Information Unit (CIU) analysis were used to measure their structural and functional properties. Based on prior work, we selected QPA measures of speech rate (words per minute) and complexity (mean sentence length, inflection index, and auxiliary index) and four CIU measures of informativeness (#CIU, CIU/min, %CIU, #nonCIU). These measures produced two orthogonal dimensions with four orthogonal participant clusters. Comprehensive comparison between clusters revealed that speech rate and complexity were strongly associated with general aphasia severity and total lesion volume, and were predicted by frontoparietal grey matter and dorsal pathway white matter damage. In contrast, informativeness was independent of other behavioral and neurological deficits, and was not predictable from lesion patterns, suggesting that it reflects communication strategy rather than specific neurological impairment. These results provide an important step toward distinguishing neurological and strategic aspects of narrative speech deficits in post-stroke aphasia, with potential implications for treatment approaches that target communication strategies.