Linguistic complexity measures of short narrative writings for modeling overall Chinese proficiency

Abstract

This study investigates whether linguistic complexity measures of short narrative writings can model overall proficiency in Chinese as a second language. A sample of native Japanese undergraduate students with intermediate (n = 34) and advanced (n = 43) levels of Chinese proficiency, and a sample of native speakers of Chinese (n = 43) altogether provided 240 short narrative essays in Chinese, with a minimum length of 30 characters and an actual grand mean length of 45 characters. Linguistic complexity measures, as discussed in relevant literature, were identified and coded in the essays. Multivariate analyses of variance tests suggest that learners across proficiency levels performed significantly differently on 11 complexity measures of short narrative writings. Multinomial logistic regression further revealed that six of these measures (lexical diversity, high-difficulty vocabulary, complex noun phrase ratio, average sentence length, error-free T-unit ratio, and the first-person singular pronoun) could collectively distinguish student writers’ general Chinese proficiency levels, with an overall accuracy rate approaching 0.80. The implications of the findings of this study for writing assessments and writing instruction are discussed.