Error-dependent modulation of speech-induced auditory suppression for pitch-shifted voice feedback

Abstract

Background: The motor-driven predictions about expected sensory feedback (efference copies) have been proposed to play an important role in recognition of sensory consequences of self-produced motor actions. In the auditory system, this effect was suggested to result in suppression of sensory neural responses to self-produced voices that are predicted by the efference copies during vocal production in comparison with passive listening to the playback of the identical self-vocalizations. In the present study, event-related potentials (ERPs) were recorded in response to upward pitch shift stimuli (PSS) with five different magnitudes (0, +50, +100, +200 and +400 cents) at voice onset during active vocal production and passive listening to the playback.

Results: Results indicated that the suppression of the N1 component during vocal production was largest for unaltered voice feedback (PSS: 0 cents), became smaller as the magnitude of PSS increased to 200 cents, and was almost completely eliminated in response to 400 cents stimuli.

Conclusions: Findings of the present study suggest that the brain utilizes the motor predictions (efference copies) to determine the source of incoming stimuli and maximally suppresses the auditory responses to unaltered feedback of self-vocalizations. The reduction of suppression for 50, 100 and 200 cents and its elimination for 400 cents pitch-shifted voice auditory feedback support the idea that motor-driven suppression of voice feedback leads to distinctly different sensory neural processing of self vs. non-self vocalizations. This characteristic may enable the audio-vocal system to more effectively detect and correct for unexpected errors in the feedback of self-produced voice pitch compared with externally-generated sounds.

Figure 1

Time course of the ERP responses to pitch-shifted voice feedback at voice onset for 0, 50, 100, 200 and 400 cents stimulus magnitudes. ERPs responses from frontal (Fz) and central (Cz) EEG channels are overlaid for active vocalization (solid) and passive listening (dashed) conditions. The horizontal and vertical dashed lines in each subplot mark the baseline and stimulus onset, respectively.

Figure 2

Topographical scalp distributions of the N1 ERP component in response to 0, 50, 100, 200 and 400 cents pitch shifted voice feedback at vocal onset. The top and bottom rows show the topographical maps of N1 distributions for active vocalization and passive listening conditions, respectively. The maps are calculated for 13 recording sites on the surface of the scalp (C_Z, C₃, C₄, T₇, T₈, F_Z, F₃, F₄, F₇, F₈, P_Z, P₃, P₄).

Figure 3

The bar plot representation of the mean percentage of normalized N1 suppression for 0, +50, +100, +200 and +400 cents upward pitch shift stimulus (PSS) at voice onset. The error bars show the standard error values for each stimulus magnitude, separately.

Figure 4

Schematic of the experimental setup. ERPs were obtained in response to randomly chosen 0, 50, 100, 200 and 400 cents pitch shifts in subjects' voice auditory feedback. The voice signal was fed to a voice onset detector (VOD) module that was used to detect voice onset. The VOD output cued a random generator function to randomly choose a stimulus magnitude and trigger the harmonizer to deliver pitch shifts at voice onset.

Figure 5

An example of the a) sound pressure waveforms and b) frequency spectra of voice output for the vowel sound /a/ at pitch frequency of 175 Hz and its pitch-shifted (200 cents) auditory feedback (196 Hz).