Listeners vary in how gradiently and consistently they assign speech sounds to phonetic categories. Using a multimodal neuroimaging approach, several experiments assessed how structural and functional properties of the central auditory pathway predict differences in acoustic-phonetic categorization and speech-in-noise (SIN) perception. We recorded brainstem frequency-following responses (FFRs) and cortical event-related potentials (ERPs) from young, normal-hearing listeners while they labeled vowel sounds along an acoustic-phonetic continuum during active perceptual tasks. We measured consistency (i.e., trial-to-trial stability) of speech coding from FFRs and ERPs. We also measured SIN understanding to relate speech sound categorization to higher-level speech processing. To further assess how volumetric and structural connectivity properties of the auditory-linguistic brain relate to perceptual differences in categorization, we collected structural MRI scans and diffusion-weighted imaging (DWI).

Across the various studies, functional EEG and behavioral results collectively suggest i) more perceptually gradient/consistent listeners have stronger neural encoding of speech that is less degraded by noise and thus achieve better SIN perception, ii) speech representations are more gradient in brainstem but more categorical in cortex, and iii) listeners’ perceptual consistency in phonetic labeling is predicted by neural consistency of their speech-evoked potentials. MRI/DWI further revealed i) more gradient perception relates to greater surface area of frontal language regions, while increased perceptual consistency relates to thicker auditory cortical regions, and ii) more gradient listeners have denser white matter pathways in left hemisphere auditory-language tracts. Taken together, these studies reveal important structure-function-behavior relationships across the auditory system that explain variance in categorization abilities and their relation to SIN performance.