Objective measures of intelligibility are highly useful for the study of second-language speech because they are easily quantifiable and they allow the researcher to distinguish between the presence of a foreign accent and the measurable effects it has on communication. Two experiments were carried out to examine the relationship between intelligibility and phonemic contrasts in Chinese-accented English. In the first experiment, the speech of two Mandarin Chinese learners of English was recorded and phonetically transcribed, probing initial and final single consonants, consonant clusters, and the stressed vowels of American English, from which a 69-item phonemic error inventory was compiled. A minimal-pairs word test was developed, based on the error inventory.
In the second experiment, 44 native English-speaking listeners were presented with the words from the minimal-pairs test, along with sentences and a passage, spoken by one of eight Mandarin Chinese speakers, or by a native English-speaking control. A feature-based analysis of the minimal-pairs data was performed, with percent-correct scores computed for each of seven consonantal and vocalic features. Sentence and passage intelligibility were scored as percent of words correctly transcribed by listeners.
Correlation analyses revealed substantial differences in both strength and sign of correlation with intelligibility across the seven minimal-pairs variables. In two multiple-regression analyses, performance on the seven minimal-pairs variables accounted for approximately 76% of the variance in sentence intelligibility and 43% of the variance in passage intelligibility. Several sources of evidence suggest that vowel errors are more strongly associated with intelligibility than consonant errors. Most importantly, a strong relationship was found between performance on phonemic contrasts and overall intelligibility.