Using a continuous listening task, we evaluated the coupling between the listener's cortical activity and the temporal envelopes of different sounds in a multitalker auditory scene using magnetoencephalography and corticovocal coherence analysis. Neuromagnetic signals were recorded from 20 right-handed healthy adult humans who listened to five different recorded stories (attended speech streams), one without any multitalker background (No noise) and four mixed with a "cocktail party" multitalker background noise at four signal-to-noise ratios (5, 0, Ϫ5, and Ϫ10 dB) to produce speech-in-noise mixtures, here referred to as Global scene. Coherence analysis revealed that the modulations of the attended speech stream, presented without multitalker background, were coupled at ϳ0.5 Hz to the activity of both superior temporal gyri, whereas the modulations at 4 -8 Hz were coupled to the activity of the right supratemporal auditory cortex. In cocktail party conditions, with the multitalker background noise, the coupling was at both frequencies stronger for the attended speech stream than for the unattended Multitalker background. The coupling strengths decreased as the Multitalker background increased. During the cocktail party conditions, the ϳ0.5 Hz coupling became left-hemisphere dominant, compared with bilateral coupling without the multitalker background, whereas the 4 -8 Hz coupling remained right-hemisphere lateralized in both conditions. The brain activity was not coupled to the multitalker background or to its individual talkers. The results highlight the key role of listener's left superior temporal gyri in extracting the slow ϳ0.5 Hz modulations, likely reflecting the attended speech stream within a multitalker auditory scene.

During connected speech listening, brain activity tracks speech rhythmicity at delta (∼0.5 Hz) and theta (4-8 Hz) frequencies. Here, we compared the potential of magnetoencephalography (MEG) and high-density electroencephalography (EEG) to uncover such speech brain tracking. Ten healthy right-handed adults listened to two different 5-min audio recordings, either without noise or mixed with a cocktail-party noise of equal loudness. Their brain activity was simultaneously recorded with MEG and EEG. We quantified speech brain tracking channel-by-channel using coherence, and with all channels at once by speech temporal envelope reconstruction accuracy. In both conditions, speech brain tracking was significant at delta and theta frequencies and peaked in the temporal regions with both modalities (MEG and EEG). However, in the absence of noise, speech brain tracking estimated from MEG data was significantly higher than that obtained from EEG. Furthemore, to uncover significant speech brain tracking, recordings needed to be ∼3 times longer in EEG than MEG, depending on the frequency considered (delta or theta) and the estimation method. In the presence of noise, both EEG and MEG recordings replicated the previous finding that speech brain tracking at delta frequencies is stronger with attended speech (i.e., the sound subjects are attending to) than with the global sound (i.e., the attended speech and the noise combined). Other previously reported MEG findings were replicated based on MEG but not EEG recordings: 1) speech brain tracking at theta frequencies is stronger with attended speech than with the global sound, 2) speech brain tracking at delta frequencies is stronger in noiseless than noisy conditions, and 3) when noise is added, speech brain tracking at delta frequencies dampens less in the left hemisphere than in the right hemisphere. Finally, sources of speech brain tracking reconstructed from EEG data were systematically deeper and more posterior than those derived from MEG. The present study demonstrates that speech brain tracking is better seen with MEG than EEG. Quantitatively, EEG recordings need to be ∼3 times longer than MEG recordings to uncover significant speech brain tracking. As a consequence, MEG appears more suited than EEG to pinpoint subtle effects related to speech brain tracking in a given recording time.

In multitalker backgrounds, the auditory cortex of adult humans tracks the attended speech stream rather than the global auditory scene. Still, it is unknown whether such preferential tracking also occurs in children whose speech-in-noise (SiN) abilities are typically lower compared with adults. We used magnetoencephalography (MEG) to investigate the frequency-specific cortical tracking of different elements of a cocktail party auditory scene in 20 children (age range, 6-9 years; 8 females) and 20 adults (age range, 21-40 years; 10 females). During MEG recordings, subjects attended to four different 5 min stories, mixed with different levels of multitalker background at four signal-to-noise ratios (SNRs; noiseless, ϩ5, 0, and Ϫ5 dB). Coherence analysis quantified the coupling between the time courses of the MEG activity and attended speech stream, multitalker background, or global auditory scene, respectively. In adults, statistically significant coherence was observed between MEG signals originating from the auditory system and the attended stream at Ͻ1, 1-4, and 4-8 Hz in all SNR conditions. Children displayed similar coupling at Ͻ1 and 1-4 Hz, but increasing noise impaired the coupling more strongly than in adults. Also, children displayed drastically lower coherence at 4-8 Hz in all SNR conditions. These results suggest that children's difficulties to understand speech in noisy conditions are related to an immature selective cortical tracking of the attended speech streams. Our results also provide unprecedented evidence for an acquired cortical tracking of speech at syllable rate and argue for a progressive development of SiN abilities in humans.