Speech intelligibility in speech noise: A perceptual training study.

Abstract

The ability to understand speech in noisy conditions is not well predicted by the traditional audiogram or by an individual’s ability to understand speech in quiet. Furthermore, listeners with different language backgrounds (e.g., native versus non-native listeners) are affected differentially by the presence of noise, and listeners of a given native language experience more interference from background noise in their native language versus in a foreign language. Given these observations, it is hypothesized that the capacity to tune into target speech and tune out interfering speech noise is subject to experience-related modification and as such should be trainable. The objective of this study is to identify training parameters than can best enhance speech-in-speech intelligibility. To do this, we employ a multiple-talker training paradigm and three between-subjects training conditions: training in speech-shaped noise (a high energetic masker), in English two-talker babble (a high informational masker), and in Mandarin two-talker babble (a lower-energetic/lower-informational masker). By administering a common speech-in-speech post-test, the comparison of these training conditions allows us to begin to isolate the most trainable processes involved in speech-in-speech intelligibility. This approach also allows us to investigate whether training can mitigate the previously observed detrimental effects of native-language noise versus foreign-language noise.