Psychoacoustic research has highlighted the fundamental role of temporal modulations for speech perception in noisy environments. The present study seeks for a relationship between AM processing (using a behavioural task and an electroencephalography (EEG) measure) and speech perception in noise. The AM following response (AMFR, or envelope following response), an auditory potential, reflecting the brain activity following the modulation frequency of amplitude modulated tones, can be recorded at the scalp level with EEG. It was hypothesized that higher AM processing (higher magnitudes of AMFR and better abilities in AM perception behavioural task) would positively correlate with lower (better) speech-in-noise thresholds. Moreover, this relationship was expected to differ as a function of AM rates, as speech information is mainly conveyed by slow AM cues.

Fifty young adults with normal hearing (18-30 years) completed two experiments: 1) an EEG session measuring AMFR at two AM rates (8 vs 40 Hz using a pure tone centred at 1024 Hz sinusoidally modulated at 100%), and 2) a behavioural measure estimating consonant-identification thresholds in noise. Thirty-four adults of this group completed a third experiment: 3) a behavioural assessment of AM detection thresholds at 8 Hz. For the EEG experiment, Fourier transform (FFT) was performed on the averaged waveforms in each AM rate condition. The maximum magnitude value at 8 and 40 Hz was estimated individually and corrected by the EEG noise for each participant. For the speech-in-noise adaptive task, syllables of the form /aCa/ were presented. Six phonetic conditions were designed presenting a minimal change in place of articulation, voice or manner of articulation for fricatives or stop consonants. Syllables were presented in a XAB task to assess consonant identification thresholds within a steady speech-shaped noise. In the AM detection adaptive task, participants had to identify the modulated sound among two different sounds, which only one was modulated in amplitude. Every time they succeeded the modulation depth of the AM sound decreased. In one condition of this task the carrier frequency of sound is a 500 Hz pure tone and in the other condition it is a narrow-band noise. This allow us to evaluated the masking effect (masking effect = score in pure tone condition - score in noise's one).

AMFRs were observed at each modulation rate. Thresholds for consonant-identification in noise were comprised between -11 and -19 dB SNR for all phonetic conditions averaged. Correlation between these three experiments (n=34) revealed that higher magnitude of AMFR only at slow AM rate (8 Hz) is positively correlated with lower (better) speech-in-noise threshold for stop consonants varying in place and for fricative consonants varying in voicing. Magnitude of AMFR at slow AM rate (8 Hz) is also correlated with masking effect. This study suggests that AM processing relates to some extent to consonant processing in noise.