Spectrotemporal-modulation (STM) sensitivity is a strong predictor of speech recognition in noise, yet the neural mechanisms underlying this predictive power remain unclear. We investigate how the peripheral auditory system encodes STM stimuli by analyzing single-unit responses from chinchilla auditory-nerve (AN) fibers and simulating responses using a computational AN model. A central question is which cues listeners rely on when listening to this stimulus and how these cues are preserved or degraded in the auditory periphery following various forms of sensorineural hearing loss (SNHL). We analyze spike-train data to quantify envelope coding, TFS coding, and short-term place coding. To overcome the limitations of CF sampling in physiological recordings, we apply the Spectro-Temporal Manipulation Procedure (STMP), which simulates a population response while recording from a single unit by varying the stimulus sampling rate. Species-specific stimulus design is also explored, as chinchillas have broader cochlear tuning than humans. The computational model supports experimental design and parameter selection, allowing for a broader exploration of STM parameters than is practical experimentally. Preliminary modeling and physiology guide future efforts to understand how various SNHL subtypes affect STM coding and how those effects may explain the predictive power of STM sensitivity for speech-in-noise perception in individual listeners.
Farhadi et al. (Wed,) studied this question.