| Fluency reveals how comfortable a second language writer is with language production. Writing fluency in the present study is represented as a linguistic feature associated with speed, fluidity and accuracy. There have been relatively a number of empirical studies on fluency measures in writing, but the results are too mixed to yield consistent conclusions about how fluency measures function in writing assessment and in writing development across grade levels. Meanwhile, few empirical studies attempted to address the ESL learners in China since fluency is not such a conspicuous phenomenon in L2 writings compared with accuracy, complexity et al. The present quantitative study was designed to investigate the validity of fourteen fluency measures in writing assessment, and most importantly, the development trends of writing fluency across grade levels from a quasi-experimental way. Specifically, the study addresses following questions: a) What is the relationship between fluency measures and the holistic scoring? b) Which fluency measures are valid in predicting the holistic scoring? c) What is the predicting model of fluency measures? d) What are the changes across four grade levels in terms of fluency measures? The subjects were 120 undergraduate English majors randomly selected across four grade levels, from whom writing samples were collected under time constraints. Fourteen fluency measures, counted respectively as frequency measures and ratio measures, were adopted in the analysis of the compositions. In addition, all the compositions were holistically rated based on TOFEL scoring schemes by three independent raters. Correlation analyses were performed to investigate the relationship between fluency measures and the holistic scoring, from which eight measures (W, C, WEFT, WEFC, W/M, W/S, W/T and W/EFT) showed rather strong positive correlations with the writing scores, and negative correlations were found on measures of V, T-unit, S, WCN/T and WCN/C. Results indicate a moderate relationship between these measures and writing quality. The establishing of a predicting model of fluency measures was accomplished using the multiple regression analysis. However, only four fluency measures, WEFT, WEFC, V and W/M, entered in the present model with an ability to account for 51.7% of variances in the holistic scoring. The discrepancies between the results for correlational analysis and the multiple regression analysis were then discussed using a series of inter-correlation analyses among certain fluency measures. ANOVA analyses and/or non-parametric tests were performed to compare the fourteen fluency measures across four grade levels, from which significant linear increase was found on measures of C, WEFT, WEFC, W/S, W/T, W/EFT. In conclusion, measures with discriminating and predicting validity in holistic writing assessment might not function well in the development trends as overall proficiency increases. For example, measures of words and words per minute were good indicators of holistic writing, however, no significant difference were found among four grade levels in terms of words. Other measures (number of clauses, sentence length, T-unit length and error-free T-unit length) showing linear development trends did not discriminate in holistic scoring. The present study discussed the validity of fluency measures in writing assessment, and also the development trends of English majors'L2 writing in China in terms of these measures. Those measures with discriminating ability appear to be yardsticks in writing assessment and development by which researchers could expediently gauge a learner's writing proficiency in a second language. It is suggested that language learner and teachers should pay due attention to writing fluency in writing instruction, especially those measures of discriminating validity. Further qualitative studies are needed to explore writing fluency as psychological and cognitive processes in order to understand the real complexity of language production as a whole.
|
PDF Full Text Request