Effects of Sentence Predictability on EFL Learners’ Speech-in-noise Recognition

The present study investigates how the presence of background noise and sentence predictability affect Japanese EFL learners’ listening comprehension. The following two research questions were defined: (a) How does the level of noise affect Japanese EFL learners’ listening comprehension? and (b) How does the predictability of words affect Japanese EFL learners’ listening comprehension in the presence of background noise? The speech-perception-in-noise test devised by Kalikow et al. (1977) was used. This test includes standardized sentences in which the last word in each sentence is either a high-predictability word or a low-predictability word. There were five noise conditions: no noise, signal-to-noise ratio (SNR) = 15, SNR = 10, SNR = 5, and SNR = 0. A total of 119 Japanese university students did the listening test and rated their confidence level with regard to their listening comprehension at the end of each noise condition. Two-way ANOVA was used to analyze the scores. The results of the listening test and the questionnaire indicated that both noise level and sentence predictability affected EFL learners’ listening comprehension, and that both factors had interactive effects. The contextual information improved their listening comprehension in the SNR 15 and SNR 5 conditions. However, in the no-noise condition, the contextual information did not affect their listening comprehension, as the listening scores were similar for low- and high-predictability sentences. In the SNR 5 and SNR 0 conditions, the learners’ listening comprehension was affected by the noise, and in the SNR 0 condition, the level of noise was so high that the scores on the listening test were the lowest under this condition. Thus, the findings of this study imply that the effects of contextual information on the participants’ listening comprehension varied depending on the noise level.