An Empirical Study On Predictability Of Writing Form To Second Language Writing Performance

Holistic or analytic scoring has been an eternal topic in the field of testing in recent years. Holistic scoring is the assigning of a single score to a script based on the overall judgment about its quality. In terms of analytic scoring, scripts are rated on several aspects of scoring rubric rather than a single score.The study is to construct the model on predictability of writing form to writing performance. 243 English majors participated in the study. They have been required to write the composition in 50 minutes during the final examination. The students come from Grade Three of different universities in Hebei Province. Namely, Hebei University, Hebei University of Technology and Hebei University of Economics and Business. All of the compositions need to be scored by three raters twice, one with holistic scoring and the other with analytic scoring. Holistic scoring is based on TEM4, whereas analytic scoring adopts the criteria: relevance, explicitness, coherence, sufficiency (Wen, 2007); fluency,accuracy and complexity (Wolfe-Quintero et al.,1998; Liang 2005). After rating, all of the data should be analyzed by SPSS (16.0). AMOS (7.0) is used to establish SEM based on literature and hypothesis. In the study, writing form is indicated by three observed variables: fluency, accuracy and complexity. Writing performance is seen as latent variable. SEM is used to construct and check to what extent these three variables can predict writing performance.The objective of the paper is to provide detailed analysis about writing form and writing performance. The paper attempts to address the following questions: in L2 writing, what is the relationship between writing form and writing performance? As for the second language learners, can fluency, accuracy and complexity predict writing performance? As for high-proficiency learners and low-proficiency learners, how do different understanding of writing form influence writing performance?According to data analysis and interviewees'response, there are three major findings in the study. First, writing form can predict 47.61% variance of writing performance directly. Second, fluency, complexity and accuracy are positively correlated with writing performance. Complexity and fluency can predict more about writing performance than accuracy. Third, different understanding of writing form can affect writing performance directly according to interviewees'response. In sum, the findings are helpful for us to improve the way of assessing L2 writing performance. Meanwhile, fluency, complexity and accuracy can provide more effective feedback to the test participants from different multi-facet, which is helpful to improve the writing performance.