P34 Pre-target α EEG power predicts intra-individual variability in Digit-in-Noise recognition
Features of spontaneous EEG activity predict individual variability in Speech-in-Noise perception tasks. For instance, resting-state MEG power predicts individual differences in signal-to-noise ratio (SNR) thresholds for correct word recognition of words in noise. Further, mean α power at baseline predicts individual differences in Digit-in-Noise (DiN) reception thresholds. Moreover, properties of the ongoing EEG are predictive of moment-to-moment, within-individual variability in performance on several speech recognition tasks, including syllable discrimination and lexical decision in noise. We set out to investigate the hypothesis that trial-to-trial fluctuations in brain activity during a pre-target, noise-filled period, affect the probability of accurately reporting a target item. To this end, we assessed the effects of pre-stimulus EEG (128-channel EGI HydroCel) power on both recognition and subjective clarity of monosyllabic German digits, at three levels of noise masking (108 trials each), calibrated to individual digit-wise performance levels in a sample of 25 young normally-hearing listeners. First, we computed differences in grand average pre-target spectral power between correct and incorrect trials, and between subjectively clearer and less clear ones. This suggested behavioural relevance of α band power within the. We performed mixed-effects regression of comprehension and clarity on source-reconstructed, parcel-wise power, averaged between 8.5 and 12Hz in 50ms time bins between –300ms and stimulus onset. This revealed a significant (p<.05 after non-parametric cluster-based permutation) interaction between α power and SNR on comprehension in a broad left-lateralised temporo-parietal region: higher α power was associated with increased probability of correct responses at higher SNRs, and decreased probability of correct response at lower SNRs. These results indicate that pre-target α power, in areas known to be involved in various stages of speech processing, predicts moments of relatively improved or decreased DiN comprehension. α band power is known to be associated with decreased cortical excitability and sensitivity to input. Increased α power during the noise-filled pre-target period has been previously interpreted as a noise suppression mechanism, such that α power-associated reduction in the processing of the noise leads to enhanced representation of the signal. In contrast, the trade-off between noise suppression and signal detection observed in the current study supports the more parsimonious view that α power indexes a relatively unspecific gain control mechanism that is engaged to suppress noise, with potentially detrimental consequences for signal detection at very low SNRs.