A recent model for subcortical neural coding of speech by neural fluctuations (NFs) is robust over the range of sound levels used in conversational speech and also in background noise. The responses of AN fibers tuned near spectral peaks (e.g., formants, or the broad high-frequency spectral peaks associated with fricatives) have small NFs, and AN fibers tuned in spectral valleys or on slopes have deep NFs. Thus, the spectral envelope is represented by the profile of NF depths along the tonotopic axis. The NF code is facilitated by cochlear tuning and saturating peripheral nonlinearities that result in capture (or dominance) of inner-hair-cell (IHC) and AN responses by energy near spectral peaks. The NF code is translated to a rate code at the level of the midbrain by fluctuation-sensitive neurons in the IC.

The operating points of the saturating peripheral nonlinearities depend upon cochlear sensitivity. Therefore, decreased cochlear gain due to sensorineural hearing loss (SNHL) reduces capture and decreases the contrast in the NF profile. As a result, SNHL degrades the representation of the spectrum at the level of the IC. We will illustrate subcortical speech responses for normal hearing and SNHL using a computation model for the AN that includes nonlinear cochlear tuning and saturating nonlinearities associated with IHC transduction and the IHC-AN synapse. We hypothesize that efferent cochlear gain-control pathways are critical for maintaining and enhancing NF contrast over a wide range of sound levels and of signal-to-noise ratios. We are exploring this hypothesis using a model with efferent pathways that convey both level-driven signals from the cochlear nucleus and fluctuation-driven signals from the inferior colliculus to the medial olivocochlear cochlear (MOC) gain-control system. Key parameters of the efferent model that are currently under investigation are the bandwidths of the MOC-to-cochlea projections that are driven by energy vs. fluctuations.

Funding: Supported by NIH-DC-R01010813.