Subcortical rather than cortical sources of the frequency-following response (FFR) relate to speech-in-noise perception in normal-hearing listeners

Abstract

Scalp-recorded frequency-following responses (FFRs) reflect a mixture of phase-locked activity across the auditory pathway. FFRs have been widely used as a neural barometer of complex listening skills, especially speech-in noise (SIN) perception. Applying individually optimized source reconstruction to speech-FFRs recorded via EEG (FFR_EEG), we assessed the relative contributions of subcortical [auditory nerve (AN), brainstem/midbrain (BS)] and cortical [bilateral primary auditory cortex, PAC] source generators with the aim of identifying which source(s) drive the brain-behavior relation between FFRs and SIN listening skills. We found FFR strength declined precipitously from AN to PAC, consistent with diminishing phase-locking along the ascending auditory neuroaxis. FFRs to the speech fundamental (F0) were robust to noise across sources, but were largest in subcortical sources (BS > AN > PAC). PAC FFRs were only weakly observed above the noise floor and only at the low pitch of speech (F0≈100 Hz). Brain-behavior regressions revealed (i) AN and BS FFRs were sufficient to describe listeners' QuickSIN scores and (ii) contrary to neuromagnetic (MEG) FFRs, neither left nor right PAC FFR_EEG related to SIN performance. Our findings suggest subcortical sources not only dominate the electrical FFR but also the link between speech-FFRs and SIN processing in normal-hearing adults as observed in previous EEG studies.

Keywords: Auditory brainstem response (ABR); Auditory event-related potentials (ERPs); Cocktail party scenario; Cortical FFR; Noise-degraded speech perception.

Figure 1:. FFR (sensor-level) waveforms and scalp topographies as a function of noise.

(a) Electrode recordings at channel Fpz across noise levels. Gray trace= stimulus waveform. FFRs appear as phase-locked potentials that mirror the acoustic periodicity of speech. (b) FFR topographies. Maps reflect the voltage distribution on the scalp, averaged across the periodic “wavelets” of the FFR [i.e., 25 most prominent positive peaks; see 5] in the response time window (0–350 ms; yellow shading). Maximal FFR amplitude near frontocentral sites (e.g., Fpz) and polarity inversion at the mastoids are consistent with deep midbrain sources that point obliquely in an anterior orientation to the vertex (parallel to the brainstem) [5, 6]. Red/blue shading = +/− voltage.

Figure 2:. Source-level FFR waveforms and spectra along the ascending auditory neuroaxis.

Grand average response spectra and waveforms (insets) at each dipole source of the FFR [head model; 6]. Waveforms and FFRs reflect responses to clean speech. Inset bar charts show noise-related changes in FFR F0. Gray shading = spectral noise floor measured in the pre-stimulus interval. FFRs show strong phase-locking at the speech F0 frequency (~100 Hz) in both subcortical and cortical sources. Subcortical sources (AN, BS) show additional response energy at higher harmonics of speech (H2 and H3). errorbars = ±1 s.e.m.

Figure 3:. Subcortical speech coding is sufficient to explain SIN listening skills.

(a) QuickSIN scores. (b) Subcortical and cortical sources of the FFR account for significant variance in behavioral QuickSIN scores. Bilateral sources (PAC, AN) are pooled across hemispheres. Statistical flags mark significant regressors in the GLME model. Subcortical sources (AN, BS) are associated with SIN performance, whereas cortical FFRs (PAC) do not relate to behavior. (c) Bootstrapped p-values for each source regressor. Only BS and AN reach significance among N=250 resamples. Error bars = 95% CI. *p < 0.05; **p < 0.01, ***p < 0.001