Mutual intelligibility of Chinese, Dutch, and American speakers of English

TitleMutual intelligibility of Chinese, Dutch, and American speakers of English
Publication TypePresentation
Year of Publication2003
Conference NameDag van de Fonetiek 2003
AuthorsHongyan, Wang, and Vincent van Heuven
PublisherNederlandse Vereniging voor Fonetische Wetenschappen
Conference LocationUtrecht, The Netherlands

Very little is known about the loss of intelligibility that is incurred by L2 speakers when they communicate with native L1 listeners. Even less is known about the differences in intelligibility among L2 English speakers of diverse national back­grounds, such as Chinese-accented speakers of English versus Dutch-accented speakers. The first aim of our study is to test the hypotheses that (i) Dutch English is more intelligible to native listeners of English than Chinese English, and (ii) both foreign-accented varieties are less intelligible to L1 English listeners than native English. Hypothesis (i) follows from a contrastive analysis of the sound systems of the languages involved, showing that Dutch and English are much more similar in their sound structure than Chinese and English. Our third hypothesis relates to the relative intelligibility of the three types of English for non-native listeners. Two hypotheses are plausible here: (iii) L1 English is always more intelligible to listeners of any nationality, since it optimally conforms to the norm the foreign speaker/listeners were taught to adhere to, or (iv) Dutch English is more intelligible to Dutch listeners, and Chinese English to Chinese listeners, as these varieties embody precisely the interference phenomena that the L2 speakers are used to.

We recorded a male and a female speaker of (American) English, of Dutch English and of Chinese English. Speakers were young adults, studying at the university level with no specialisation in English. Five types of English materials were recorded for each speaker: (1) vowel test: a list of word containing the 20 vowels in identical /hVd/ contexts, (2) consonant test: a list of nonsense words /aCa/ containing 24 intervocalic single consonants, (3) cluster test: a list of 20 CC or CCC clusters in /aCC(C)a/ clusters, (4) SUS-test: 30 Semantically Unpredictable Sentences with high-frequency words occurring in syntactically correct but semantically nonsense sentences, and (5) SPIN test: 50 short sentences, with a contextually predictable or unpredictable target word in final position. The entire set of materials was then presented in perceptual identification and recognition tests three groups of listeners belonging to the same population as the speakers.

For each test, hypotheses (i), (ii) and (iv) but not (iii) were supported. In our talk we will present the confusion structure in the vowel, consonant, and cluster data, and show how intelligibility at the sentence level can be predicted through regression analysis from the phoneme-identification results.