PRELIMINARY INVESTIGATION OF THE UTILITY OF THE N400 FOR PASSIVELY ESTIMATING SPEECH HEARING ABILITY IN NOISE

Abstract

Recommended procedures for hearing assessment include the evaluation of speech hearing in background noise. Speech in noise assessments rely heavily on behaviour and attention, making them unsuitable for many patient populations. Electrophysiological measures are increasingly being used by audiologists to circumvent the limits of behavioural testing. This project looked at the possibility of using a well- studied event-related potential, the N400, as an objective measure of speech comprehension in noise. The N400 is associated with the processing of meaningful stimuli, and has been of particular interest in the study of written and spoken language comprehension. N400 amplitude varies with several factors including the degree to which a word is expected based on the surrounding context. The effect of varying levels of speech-frequency background noise on the N400 as elicited by semantically anomalous spoken sentences in the absence of attention was investigated in eleven adults with normal hearing. It was hypothesized that the magnitude of the difference in N400 amplitude between congruent and incongruent trials (known as the N400 effect) would vary systematically with intelligibility, decreasing in more adverse listening conditions. Five signal to noise ratios relative to the behavioural threshold (-2 dB, threshold, +1 dB, +2 dB, and +4 dB) were tested as well as a quiet condition. The amplitude of the N400 effect did vary with intelligibility however this effect was nonlinear, with the +1 dB and -2 dB conditions having significantly smaller N400 effect amplitudes than the other conditions as a group, as determined through a partial least squares (PLS) analysis. This effect appeared to be driven by changes in the responses to incongruent, rather than congruent, stimuli. The results are discussed in the context of related literature on the N400 in adverse listening conditions. Because of the complex nature of the effect of noise on the N400 to sentence-level expectancy violations, this paradigm does not appear to be immediately useful for application in clinical speech in noise audiometry.