Human cortical activity measured with magnetoencephalography (MEG) has been shown to track the temporal regularity of linguistic information in connected speech. In the current study, we investigate the underlying neural sources of these responses and test the hypothesis that they can be directly modulated by changes in speech intelligibility. MEG responses were measured to natural and spectrally degraded (noise-vocoded) speech in 19 normal hearing participants. Results showed that cortical coherence to “abstract” linguistic units with no accompanying acoustic cues (phrases and sentences) were lateralized to the left hemisphere and changed parametrically with intelligibility of speech. In contrast, responses coherent to words/syllables accompanied by acoustic onsets were bilateral and insensitive to intelligibility changes. This dissociation suggests that cerebral responses to linguistic information are directly affected by intelligibility but also powerfully shaped by physical cues in speech. This explains why previous studies have reported widely inconsistent effects of speech intelligibility on cortical entrainment and, within a single experiment, provided clear support for conclusions about language lateralization derived from a large number of separately conducted neuroimaging studies. Since noise-vocoded speech resembles the signals provided by a cochlear implant device, the current methodology has potential clinical utility for assessment of cochlear implant performance.

No abstract

Providing a plausible neural substrate of speech processing and language comprehension, cortical activity has been shown to track different levels of linguistic structure in connected speech (syllables, phrases and sentences), independent of the physical regularities of the acoustic stimulus. In the current study, we investigated the effect of speech intelligibility on this brain activity as well as the underlying neural sources. Using magnetoencephalography (MEG), brain responses to natural speech and noise-vocoded (spectrally-degraded) speech in nineteen normal hearing participants were measured. Results showed that cortical MEG coherence to linguistic structure changed parametrically with the intelligibility of the speech signal. Cortical responses coherent with phrase and sentence structures were lefthemisphere lateralized, whereas responses coherent to syllable/word structure were bilateral. The enhancement of coherence to intelligible compared to unintelligible speech was also left lateralized and localized to the parasylvian cortex. These results demonstrate that cortical responses to higher level linguistics structures (phrase and sentence level) are sensitive to speech intelligibility. Since the noise-vocoded sentences simulate the auditory input provided by a cochlear implant, such objective neurophysiological measures have potential clinical utility for assessment of cochlear implant performance.