Reliability and generalizability of neural speech tracking in younger and older adults

Neural tracking of continuous, spoken speech is increasingly used to examine how the brain encodes speech and is considered a potential clinical biomarker, for example, for age-related hearing loss. A biomarker must be reliable (intra-class correlation [ICC] >0.7), but the reliability of neural-speech tracking is unclear. In the current study, younger and older adults (different genders) listened to stories in two separate sessions while electroencephalography (EEG) was recorded in order to investigate the reliability and generalizability of neural speech tracking. Neural speech tracking was larger for older compared to younger adults for stories under clear and background noise conditions, consistent with a loss of inhibition in the aged auditory system. For both age groups, reliability for neural speech tracking was lower than the reliability of neural responses to noise bursts (ICC >0.8), which we used as a benchmark for maximum reliability. The reliability of neural speech tracking was moderate (ICC [~]0.5-0.75) but tended to be lower for younger adults when speech was presented in noise. Neural speech tracking also generalized moderately across different stories (ICC [~]0.5-0.6), which appeared greatest for audiobook-like stories spoken by the same person. This indicates that a variety of stories could possibly be used for clinical assessments. Overall, the current data provide results critical for the development of a biomarker of speech processing, but also suggest that further work is needed to increase the reliability of the neural-tracking response to meet clinical standards. Significance statementNeural speech tracking approaches are increasingly used in research and considered a biomarker for impaired speech processing. A biomarker needs to be reliable, but the reliability of neural speech tracking is unclear. The current study shows in younger and older adults that the neural-tracking response is moderately reliable (ICC [~]0.5-0.75), although more variable in younger adults, and that the tracking response also moderately generalize across different stories (ICC [~]0.5-0.6), especially for audiobook-like stories spoken by the same person. The current data provide results critical for the development of a biomarker of speech processing, but also suggest that further work is needed to increase the reliability of the neural-tracking response to meet clinical standards.