Modeling human auditory evoked brainstem responses to speech syllables

Abstract

Auditory evoked brainstem responses (ABR) are routinely used to assess the neural encoding of sound. Various types of stimuli have been historically considered, and there is a current increasing trend towards the use of syllables, speech and complex (non-speech) sounds. Despite the peripheral origin of ABRs, the nonlinear processing within the cochlea and brainstem makes interpreting responses to these new stimuli a challenge. A recent model was developed [Ro/nne et al. (2012)] to simulate ABRs to transient sounds such as tone pulses, clicks and rising chirps as a function of stimulus level. The present study extends this model to simulate synthetic /ba/, /da/ and /ga/ syllable-evoked ABRs, where the stimuli only differ in their spectral energy of the second formant, f2, within their first 60 ms. The model takes the convolution of the instantaneous discharge rates using a “humanized” nonlinear auditory-nerve model of Zilany et al. (2009) and an empirically derived unitary response function, assumed to reflect contributions from different cell populations within the auditory brainstem. The ABR model was used to explore the physiological basis and spectro-temporal characteristics of key features observed in syllable-evoked ABRs in the literature [Skoe et al. (2011), Hornickel et al. (2009)].