In this study we examined factors influencing Japanese speakers’ discrimination of can and can’t in a spoken English sentence. Drawing on theories of phonemes and L2 speech perception, we prepared 36 sentences containing can or can’t that were read by 7 native speakers of American English. Japanese university students (N = 148) listened to a recording of 252 sentences in random order and marked which word they perceived. Bivariate regression analyses showed that subsequent sound, allophones of /t/, and sentence position were significant predictors of identification. Multiple regression analyses of significant phonemic factors and additional nonphonemic factors (the TOEIC IP score and the speaker) indicate that articulation of the /t/ in can’t, coarticulatory effects, and individual speaker effects are the most significant predictors of comprehension.
本研究では148人の日本人の大学生を対象に、センテンスレベルのcanとcan’tの聞き分けの難しさの要因について音素の理論に基づき分析を実施した。7人のアメリカ人のネィティブスピーカーによって音読されたcanあるいはcan’tを含む252のセンテンスを利用し、音素と非音素の要因（TOEIC IPテストの結果とスピーカーによる違いの要因など）に関する重回帰分析を実施した。それらの統計的な分析により、can’tの /t/ の調音、調音結合の影響、スピーカーの相違の3要因が最も有意な予測変数であることを示唆する結果を得た。
