More than a feeling: Expressive style influences cortical speech tracking in subjective cognitive decline

A research team led by Matthew King-Hang Ma has published a study demonstrating a potential AI-powered method for detecting early signs of cognitive decline. The study, involving 60 cognitively normal older adults, used EEG to measure cortical speech tracking (CTS)—how well the brain follows speech—while participants listened to speech in different expressive styles (scrambled, descriptive, dialogue, exciting). The key finding was that a greater degree of subjective cognitive decline (SCD), a self-reported sense of worsening cognition that doubles dementia risk, was associated with weaker brain tracking of higher-level linguistic features, but not basic acoustic ones.

Crucially, this weakened tracking was most pronounced when listening to prosodically 'flat' speech, such as scrambled or descriptive monologues. The researchers built encoding models that mapped three types of speech representations—acoustic, subsyllabic, and phonotactic features—to the EEG signals. Models using linguistic features significantly outperformed those using only acoustic data. This suggests the brain's difficulty in processing complex language structure, especially without expressive vocal cues, could be an early neural signature of decline.

The study, published on arXiv, proposes that the strength of cortical tracking for linguistic features during flat speech could serve as a non-invasive biomarker. This approach moves beyond subjective questionnaires and could enable much earlier intervention. The method's reliance on naturalistic listening and EEG makes it a practical candidate for future clinical screening tools.