How far down does the top-down control of speech processing go?

The perception of speech sounds in humans requires a fast and accurate integration of both bottom-up and top-down information carried by the auditory afferent and efferent pathways respectively. However, it remains poorly understood whether during speech perception the top-down control reaches the level of the brainstem or even the cochlea via the auditory efferent pathway. There is evidence that this efferent control may play a role in extracting signals from noise (Giraud et al., 1997; Maison et al., 2001; Garinis et al., 2011) and the detection of target sounds (Scharf et al., 1997; Tan et al., 2008). Here we evaluated the role of attention on the auditory efferent control of the brainstem (using auditory brainstem responses; ABRs) and the cochlear gain (using otoacoustic emissions; OAEs) during passive listening and a task of variable difficulty. We hypothesised that both the ABRs and OAEs would be modulated by attention and that the degree of suppression of OAEs (relative to the passive condition) would increase with increasing task difficulty. Thirty normal-hearing, English speakers between 18- 35 years old were assessed using lists of monosyllabic words and non-words presented in a natural or noise-vocoded (less-intelligible) conditions. Participants were tested during active and passive listening conditions. In the active condition, the participants were instructed to press a button every time they heard a non-word. A synchronised 64 channels-EEG and click evoked-OAE (CEOAEs) recording system allowed the measurement of speech onset auditory event-related potentials (ERPs), click-auditory brainstem responses (ABRs), CEOAEs and behavioral responses. Click evoked-OAE were obtained from the contralateral ear to the speech stimuli with a probe in the external ear canal. Preliminary results suggest that auditory attention modulates the activity of the cochlear activity, brainstem and cortical responses. Click evoked-OAE amplitudes were inhibited during lexical decision tasks relative to passive listening, in contrast, ABR-wave V amplitude increased and latency decreased in the active listening condition. Behavioral results showed that the accuracy (d’) decreased as the number of vocoder channels decreased. This task-difficulty effect was reflected in the early ERP components, as well as in the CEOAE data where the less intelligible condition showed stronger inhibition. These data suggest that the accurate classification of speech sounds engage not only bottom-up processing but also top-down control mechanisms that influences the auditory brainstem and the cochlear gain via the auditory efferent system.