Objective Measures Of The English Majors' Syntactic Development

Second language (L2) researchers are trying to understand how and why L2 develops with a focus on three dimensions: fluency, complexity and accuracy (Ellis, 2003; Larsen-Freeman, 2006; Norris & Ortega, 2008; Skehan, 1998a; Skehan, 1998b; Thornbury, 2000; Wolfe-Quintero et al., 1998). In each dimension, a plethora of objective measures has been invented so as to seek for the index of L2 development, a concept borrowed from the universal index of first language development"mean length of utterance". Numerous related investigations have been conducted (Gaies, 1976; Larsen-Freeman, 1978, 1983; Larsen-Freeman & Strom, 1977; Monroe, 1975; Nihalani, 1981), but since second language acquisition is a complicated process, so far no clear indicator has been found to be the index of L2 development. While the effort is still on, another effort starts. The last two decades witnessed more application of the proposed objective measures into pedagogical practice and evaluation in addition to index studies (Frantzen, 1995; Carlisle, 1989; Kepner, 1991; Shang, 2007). Of the studies on the later track, except Lim (1982) and Robb, Ross and Shortreed (1986), most of them did not test the effectiveness of the objective measures they employed, and thus the validity of resulting interpretation might be threatened (Norris & Ortega, 2003: 739). Therefore, greater importance should be attached to objective measures, which need more research in the field of L2 syntactic development.Previous studies on objective measures in written modality were weak in terms of methodology and even kept some fields seldom touched: a) just as Ortega (2003) and Polio (1997) argued, reliability was not calculated or mentioned in most of the studies; b) very few studies utilized two or more statistical methods in the same study except Homburg (1984), Ishikawa (1995) and Oh (2006); c) objective measures were coded and computed manually and therefore potential errors might have arisen in the process of counting; d) the inter-measure relationship was not deeply examined (Wolfe-Quintero et al., 1998: 119); e) measures based on sentences (MLS, C/S, EFS/S, etc.), T-units (MLT, C/T, EFT/T, etc.) and clauses (MLC, DC/C, EFC/C, etc.) were seldom compared to see which sets are better for learners at different developmental levels (Wolfe-Quintero et al., 1998:123).In review of related studies, the present study targets the English majors in China to examine the objective measures as indicators of their syntactic development based on a corpus analysis. In doing so, we hope the result would shed light on the project measuring the linguistic development of EFL learners in National University of Defense Technology. Thus, several research questions are posed:Relationship among the measures1. How are the objective measures correlated with each other?2. What are the relations among fluency, complexity and accuracy measures? Relationship between measures and grade level3. What are the route and rate of the English majors'syntactic development as reflected by the objective measures?4. Which measure(s) is (are) the best in discriminating the English majors at four grade levels in terms of correlation, stability and linearity?5. Among fluency, complexity and accuracy measures, which set(s) can best discriminate the higher and lower levels? Which set(s) is(are) more useful for lower level learners and which set(s) is(are) more useful for the higher level ones?6. Among sentence-based, T-unit-based and clause-based measures, which set(s) can best discriminate the higher and lower levels? Which set(s) is(are) more useful for lower level learners and which set(s) is(are) more useful for the higher level ones?One hundred and twenty-four argumentative compositions were analyzed to answer the questions (32 from freshmen, 30 from sophomores, 32 from juniors and 30 from seniors) from a subcorpus of Written English Corpus of Chinese Learners (WECCL)established by Wen, Wang and Liang (2005). When both inter-rater reliability and intra-rater reliability were guaranteed, tagging of S, T, NT, IC, DC and E in each text was done manually according to the rubrics. Subsequently, a software was designed to search the 18 objective measures recommended by Wolfe-Quintero et al. (1998: 122-123) and Bardovi-Harlig (1992): MLS, MLT, MLC, W/EFS, W/EFT, W/EFC, C/S, C/T, T/S, DC/S, DC/C, DC/T, EFS/S, EFT/T, EFC/C, E/S, E/T and E/C. Thereafter, several statistical treatments were conducted in terms of all the objective measures through SPSS 13.0 and PLS Graph3.0: exploratory factor analysis, partial least-squares regression, ANOVA and post hoc tests, correlational analysis, and discriminant analyses.It is found:1. The objective measures are correlated with each other through three exacted factors, which are totally consistent with the classification by Wolfe-Quintero et al. (1998):MLC, W/EFC, MLT, W/EFT, MLS and W/EFS go with fluency,DC/S, DC/T, DC/C, C/T and C/S, complexityE/C, EFC/C, E/T, EFT/T, E/S and EFS/S, accuracyThus it further validates these objective measures as indicators of L2 syntactic development and Wolfe-Quintero et al.'s categorization.2. Fluency, complexity and accuracy measures are independent on the whole, but negative correlations between fluency and accuracy measures (r=-0.206, p<.05) as well as positive relationship fluency and complexity measures (r=0.179, p<.05) are found significant. No direct relationship between complexity and accuracy measures is revealed (r=0.032, p>.05).3. Fluency and accuracy measures generally exhibit significantly linear developmental paths from grade 1 to grade 4 while complexity measures do not. The 17 objective measures as a whole display a significantly linear developmental trend. From grade 1 to grade 2, learners make substantial progress in fluency and the overall syntactic performance, and then slow down the progress while it takes three years for a significant change to occur in accuracy.4. EFC/C, E/C and W/EFS are the best measures discriminating the English majors at four grade levels in terms of correlation, stability and linearity.5. Fluency measures could best discriminate lower and higher level learners and are also more useful to place lower level learners while complexity and accuracy measures are more useful for higher level learners.6. T-unit-based measures are the best in discriminating the higher and lower level learners. All the three sets of measures based on syntactic units are more useful at lower level.Theoretically, this study multi-examines the objective measures with the English majors in China as the subjects, contributing to the finding of a universal index of L2 development. Besides, the exploration of the relationship among the developmental measures will help us understand the nature of L2 development better. Practically, suggestions in choosing proper measures for specific learners and purposes could help researchers apply the measures according to their needs. Pedagogically, specific instructions such as sentence-combining exercises are suggested to enhance the Chinese learners'L2 syntactic complexity.