| As one of the four basic language skills and the most difficult to acquire,second language(L2)writing renders the most profound and comprehensive approach to measure students’ language competence.Therefore,it has been taken as a research focus of linguists.And writing text feature has been measured to investigate writing quality.However,through retrospect it is not hard to find that previous researches generally centered on the process,teaching strategies,and influential factors of L2 writing proficiency,etc.Research on text feature analysis only counted a relatively small number.Among these researches a single or a certain types of text features such as cohesion,lexical richness,lexical diversity,etc.were mostly fathomed to evaluate the overall writing quality.Other text features were overlooked from the perspectives of sentence complexity,readability and so on.So analysis of L2 writing text features from a multi-dimensional perspective call for implementation.Advances in the fields of computer science,corpus linguistics and psychology led to the emergence of online text analysis tool Coh-Metrix.The text analysis software can maximize the multi-dimensional and comprehensive extraction and analysis of text features.In this study,90 impromptu argumentative texts of EFL Sophomore English Majors were firstly analyzed with Coh-Metrix 3.0,revealing the most influential factors on the quality of second language writing.The results showed that in the terms of such dimensions of indices as Word length,Text Easability Principle Components Scores,Co-Referential Cohesion,Lexical diversity,Syntactic complexity,Word Information and Readability,significant difference was found between the high and low score groups and significant correlations were found between these indices and composition scores.The linear regression results indicated that Word length,Number of modifiers per noun phrase,Familiarity for content words,Pronoun Incidence and Verb Incidence could respectively account for 12.2%,11%,11.8%,14.6% and 11.8% of the variance of composition scores.The results of Multiple linear regression demonstrated t he standardized regression formulation: Composition scores(15 points)=-0.0288 ? Pronoun Incidence-0.3 ? Verb Incidence-0.231 ? Familiarity for content words).The conclusion of the study had certain reference to the teaching of second language writing.Teachers should focus on the linguistic features significantly correlated with writing proficiency and strengthen training in a targeted manner and improve the L2 writing proficiency gradually.Firstly,synonyms and advanced words should be accumulated and converted to productive vocabulary.Secondly,emphasis should be placed on the difference of composition genres,for example,more static expressions but less verbs should be used in argumentative writing.Thirdly,it is necessary to master the use of various cohesive devices,such as the proper use of pronouns in order to avoid ambiguous anaphora. |