Phase alignment of low-frequency neural activity to the amplitude envelope of speech reflects evoked responses to acoustic edges, not oscillatory entrainment

Abstract

The amplitude envelope of speech is crucial for accurate comprehension, and several studies have shown 1 that the phase of neural activity in the theta-delta bands (1 -10 Hz) tracks the phase of the speech 2 amplitude envelope during listening, a process referred to as envelope tracking. However, the 3 mechanisms underlying envelope tracking have been heavily debated. Envelope tracking may reflect 4 either continuous entrainment of endogenous low-frequency oscillations to the speech envelope or the 5 combination of a series of evoked responses to acoustic landmarks within the envelope. To distinguish 6 between these two accounts for envelope tracking, and to identify the acoustic features driving it, we 7 recorded magnetoencephalography (MEG) while participants listened to natural and slowed speech. First, 8 we found that acoustic edges in speech amplitude drove evoked responses and induced phase-resetting in 9 the theta-delta band, supporting the evoked response account. In line with this, phase locking in the theta 10 band was transient and independent of the modulation rate of the speech envelope but matched the 11 temporal extent of the evoked response. Further analysis showed that the magnitude of theta phase-12 locking reflected the slope of amplitude increases. However, although amplitude increases were more 13 gradual in slowed speech, the magnitude of phase-locking did not differ between slow and regular speech, 14 reflecting a normalization for speech rate. Taken together, our results are in line with predictions of the 15 evoked response account of speech envelope tracking and support acoustic edge detection as a flexible 16 mechanism for tracking of temporal dynamics in speech. 17 18

Neural oscillations are a start toward understanding brain activity rather than the end

“…Proposed oscillations, as presented in cognitive science papers, often ignore this point, assuming for the sake of simplicity that each input adjusts the phase by the same amount, or not at all, (e.g., [ 20 ] Fig 1; [ 43 ] Fig 1; [ 44 ] Fig 1A) or that the phase will always be totally reset to 0 (e.g., [ 23 ] Fig 1B; [ 45 ] Fig 1). Many of these examples present oscillators in this fashion as simplifications for didactic purposes.…”

Envelope reconstruction of speech and music highlights stronger tracking of speech at low frequencies

“…In another theory, the neural tracking of the envelope measured with EEG and MEG is represented by its convolution with an evoked response [ 25 , 26 ]. The evoked response has been shown to respond particularly to envelope edges, which mark vowel-nucleus onsets and are highly correlated with syllable onsets in speech [ 20 , 27 ]. Yet it is not clear if the two mechanisms are irreconcilably different.…”

Envelope reconstruction of speech and music highlights unique tracking of speech at low frequencies

“…Given our results, it seems reasonable that theories of evoked responses and phase-related tracking can be parsimoniously explained by frequency-tuned evoked responses (see also Kojima et al, 2020). Further modeling work comparing the theories of evoked tracking and oscillations-based tracking will also be beneficial to reconciling their differences and the observations in EEG (for example, Doelling et al, 2019).…”

“…For speech, the phase consistency of the neural responses observed at theta band (4-8 Hz) and delta band (< 4 Hz) correspond roughly to syllabic and phrase rates respectively. In another theory, the neural tracking of the envelope measured with EEG and MEG is represented by its convolution with an evoked response (Ding and Simon, 2009;Lalor and Foxe, 2010), which has been shown to respond particularly to syllable onsets in speech (Oganian and Chang, 2019;Kojima et al, 2020). Yet it is not clear if the two mechanisms are irreconcilably different.…”

“…Yet it is not clear if the two mechanisms are irreconcilably different. Phase consistency can come about if characteristic evoked responses occur at transient, regular times in the sound 4 (Kojima et al, 2020). Likewise, an evoked response model that captures envelope tracking acts like a filter (de Cheveigné et al, 2018) and, via regularization, focuses on lower frequencies that overlap with the frequencies exhibiting the largest phase consistency.…”