On-line comprehension of natural speech requires segmenting the acoustic stream into discrete linguistic elements. This process is argued to rely on theta-gamma oscillation coupling, which can parse syllables and encode them in decipherable neural activity. Speech comprehension also strongly depends on contextual cues that help predicting speech structure and content. To explore the effects of theta-gamma coupling on bottom-up/topdown dynamics during on-line syllable identification, we designed a computational model (Precoss-predictive coding and oscillations for speech) that can recognise syllable sequences in continuous speech. The model uses predictions from internal spectro-temporal representations of syllables and theta oscillations to signal syllable onsets and duration. Syllable recognition is best when theta-gamma coupling is used to temporally align spectrotemporal predictions with the acoustic input. This neurocomputational modelling work demonstrates that the notions of predictive coding and neural oscillations can be brought together to account for on-line dynamic sensory processing.

Natural speech perception requires processing the current acoustic input while keeping in mind the preceding one and predicting the next. This complex computational problem could be handled by a multi timescale hierarchical inferential process that coordinates information flow up and down the language hierarchy. While theta and low-gamma neural frequency scales are convincingly involved in bottom-up syllable-tracking and phoneme-level speech encoding, the beta rhythm is more loosely associated with top-down processes without being assigned yet a specific computational function. Here we tested the hypothesis that the beta rhythm drives the precision of states during the speech recognition hierarchical inference process. We used a predictive coding model that recognizes syllables on-line in natural sentences, in which the precision of prediction errors is rhythmically modulated, resulting in alternating bottom-up vs. top-down processing regimes. We show that recognition performance increases with the rate of precision updates, with an optimal efficacy in the beta range (around 20 Hz). The model further performs when prediction errors pertaining respectively to syllable timing and syllable identity oscillate in antiphase. These results suggest that online syllable recognition globally benefits from the alternation of bottom-up and top-down dominant regime at beta rate, and that the gain is stronger when different features are also analyzed in alternation. These results speak to a discontinuous account of inferential operations in speech processing.

Speech comprehension requires segmenting continuous speech to connect it on-line with discrete linguistic neural representations. This process relies on theta-gamma oscillation coupling, which tracks syllables and encodes them in decipherable neural activity. Speech comprehension also strongly depends on contextual cues predicting speech structure and content. To explore the effects of theta-gamma coupling on bottom-up/top-down dynamics during on-line speech perception, we designed a generative model that can recognize syllable sequences in continuous speech. The model uses theta oscillations to detect syllable onsets and align both gamma-rate encoding activity with syllable boundaries and predictions with speech input. We observed that the model performed best when theta oscillations were used to align gamma units with input syllables, i.e. when bidirectional information flows were coordinated, and internal timing knowledge was exploited. This work demonstrates that notions of predictive coding and neural oscillations can usefully be brought together to account for dynamic on-line sensory processing.