Effects of speaker variability and noise on identifying Mandarin fricatives by native and non-native listeners

Abstract

A fundamental issue in speech perception is how listeners with different characteristics deal with various sources of acoustic variability. This study examined how speaker variability and noise affected native and nonnative identification of the place contrast in voiceless Mandarin fricatives. Monosyllabic Mandarin words fa, sa, xia, and sha produced by three female and three male speakers were mixed with five levels of speech-shaped noise. The stimuli were presented either blocked by speaker or mixed across speakers to 40 native Mandarin listeners and 52 English-speaking listeners with various amounts of Mandarin experience. It was predicted that the mixed presentation and noise would affect nonnative identification to a greater extent because of the imperfect knowledge of the target language. The results showed that nonnative performance was compromised to a greater extent by noise but not by speaker variability. These results were compared to previous studies which compared native and nonnative perception of segmental and suprasegmental contrasts in other adverse conditions. It was concluded that not all sources of acoustic variability affect native and nonnative speech perception similarly.