Mutual intelligibility of Chinese dialects experimentally tested

We argue that mutual intelligibility testing is an adequate way to determine how different two languages or language varieties are. We tested the mutual intelligibility of 15 Chinese dialects functionally at the level of isolated words (word-intelligibility) and the level of sentences (sentence intelligibility). We collected data for each dialect by playing isolated words and sentences spoken in 15 Chinese dialects to 15 listeners. Word-intelligibility was determined by having listeners perform a semantic categorization task whereby words had to be classified as one of ten different categories such as body part, plant, animal, etc. Sentence intelligibility was estimated by having the listeners translate a target word in each sentence into their own dialect. We obtained 47,250 data (15 × 150 × 15 for the word part and 15 × 60 × 15 for the sentence part). We also had at our disposal structural similarity measures (lexical similarity, phonological correspondence) for each pair of the 15 Chinese dialects published by Cheng (Computational Linguistics & Chinese Language Processing 1997, 2.1, pp. 41–72). Our general conclusion is that the degree of mutual intelligibility can be determined by both opinion and functional tests. These two subjective measures are significantly correlated with one another and can be predicted from objective measures (lexical similarity and phonological correspondence) equally well. However, functional intelligibility measures, especially at the sentence level, better reflect Chinese dialect classifications than opinion scores.