| English writing development is an important element of a student's educational career and mirrors learners'linguistic development. As part of linguistic proficiency, writing skills can be represented by textual features. Previous researches have focused their attention on what internal textual features distinguish essays ranked high and low in overall quality (Witte & Faigley, 1981), i.e., textual features like cohesive devices, lexical proficiency and syntactic complexity have had predictive powers for the function of writing proficiency. The previous general studies have focused on a singular textual feature or a certain type of features, for example, cohesion, lexical proficiency, in predicting the overall writing proficiency. As there are multiple textual features in writing, it is in great urgency for language teachers, second language researchers as well as computational linguists to study the relationships between cohesive devices, lexical diversity, syntactic complexity and writing proficiency.The developments in computational algorithms allow for the quick and accurate assessment of a multiple of textual features. The computational tool Coh-Metrix, reports on a wide range of textual features, affording analyses of writing at various levels of text structure, viz., basic textual features, cohesion, lexical diversity, syntactic complexity, word frequency as well as readability. Recent studies have examined textual features of writing proficiency related to co-reference, connectives, syntactic complexity, lexical diversity, and word frequency. However, the examination of the relationships between cohesive devices, lexical and syntactic features with writing proficiency using these computational indices is still in infancy, and almost all studies using Coh-Metrix have compared native speakers of different writing proficiency or compared and contrasted between native speakers and L2 speakers, no studies has been conducted in view of L2 learners, let alone studying L2 learners in China. It will enlarge researches assessing L2 learners'writing proficiency and extracting the multiple textual features using Coh-Metrix.The goal of this study is to better understand textual feature differences in L2 writings of high- and low-proficiency, the relationships between textual features and human holistic writing evaluation, and the predictive powers of these textual features for writing proficiency through independent samples t test, correlation and multiple regression in light of 441 second-year non-English majors'timed argumentative writings, by using the computational tool Coh-Metrix2.1.Results suggest that the combination of three indices have had a 25.6% predictive power for human holistic evaluation of writing proficiency, viz., number of words have had 19.8%, readability 4.7% and word frequency 1.1% of the variance of the holistic writing evaluation. No indices of cohesion, lexical diversity or syntactic complexity have showed the predictive ability for L2 writing proficiency of the targeted students. These results indicate that writings judged higher proficiency are likely to have more words, more infrequent content words, which are associated with text difficulty.Empirically, this study aims to tap into the relationships between textual features and second language writing proficiency in light of L2 learners through a Coh-Metrix2.1 analysis, of which the results have also testified the accessibility and reliability of this computational tool. Practically, attention should be drawn to textual features in English writing. More detailed information in regard to textual features and L2 writing will provide researchers and teachers with statistical feedback for the betterment of L2 writing research and writing pedagogy. In addition, according to Coh-Metrix2.1 output, students can modify their writing strategies and styles based on the strengths and weaknesses in their writing process and product, thus to increase their writing proficiency. |
PDF Full Text Request