Delta and theta neural entrainment during phonological and semantic processing in speech perception

Neural entrainment of acoustic envelopes is important for speech intelligibility in spoken language processing. However, it is unclear how it contributes to processing at different linguistic hierarchical levels. The present EEG study investigated this issue when participants responded to stimuli that dissociated phonological and semantic processing (real-word, pseudo-word and backward utterances). Multivariate Temporal Response Function (mTRF) model was adopted to map speech envelopes from multiple spectral bands onto EEG signals, providing a direct approach to measure neural entrainment. We tested the hypothesis that entrainment at delta (supra-syllabic) and theta (syllabic and sub-syllabic) bands take distinct roles at different hierarchical levels. Results showed that both types of entrainment involve speech-specific processing, but their underlying mechanisms were different. Theta-band entrainment was modulated by phonological but not semantic contents, reflecting the possible mechanism of tracking syllabic- and sub-syllabic patterns during phonological processing. Delta-band entrainment, on the other hand, was modulated by semantic information, indexing more attention-demanding, effortful phonological encoding when higher-level (semantic) information is deficient. Interestingly, we further demonstrated the statistical capacity of mTRFs at the delta-band and theta-band to classify utterances according to their semantic (real-word vs. pseudo-word) and phonological (real-word and pseudo-word vs. backward) contents, respectively. Moreover, analyses on the response weighting of mTRFs showed that delta-band entrainment sustained across neural processing stages up to higher-order timescales (~ 300 ms), while theta-band entrainment occurred mainly at early, perceptual processing stages (