1. 研究目的与意义（文献综述包含参考文献）

2. literature review

2.1 the relationship between sentence combining and syntactic complexity

2. 研究的基本内容、问题解决措施及方案

3. Methodology

3.1 Research questions

The purpose of the study is aimed at making comparison of sentence combining ability through one of his main measurement indexes-syntactic complexity between different learning stages: the beginning of a semester and the end of it. Thus, the following questions need to be resolved step by step:

1. Is there any intra-group difference in syntactic maturity in sentence combining at the same level but between EFL learners#8217; of different learning stages?

1.1 Is there any intra-group difference in syntactic maturity in sentence combining in the high-proficiency group between EFL learners at the beginning or at the end of the semester?

1.2 Is there any intra-group difference in syntactic maturity in sentence combining in the low-proficiency group between EFL learners at the beginning or at the end of the semester?

2. Is there any intra-group difference in syntactic maturity in the sentence combining at the same learning stage between different proficiency groups?

2.1 Is there any intra-group difference in syntactic maturity in the sentence combining between the high- and low-proficiency groups at the beginning of the semester?

2.2 Is there any intra-group difference in syntactic maturity in the sentence combining between the high- and low-proficiency groups at the end of the semester?

3.2 Variables and operational definitions

The methods applied in doing this research are paired samples t-tests for the first one question and independent sample t-test for the second one.

3.2.1 Paired sample t-test

With regard to paired sample t-test, it is a statistical technique aiming to compare two population means in one sample that is considered to be correlated used when ”before-after#8217; studies are covered, or when the samples are matched pairs or a control study is being researched. For instance, in this case, we want to know whether or not the syntactic complexity in sentence combining of the same level has something to do with different learning stages.

Paired samples t-test is carried out in four steps: setting up hypothesis, selecting the level of significance, calculating the parameter, and testing of hypothesis or decision making. In this study, we should firstly set up our two hypotheses; say, the null hypothesis and alternative hypothesis. Our null hypothesis is that the means of these two samples from different learning stages are equal while the alternative hypothesis assumes the means are not equal. After making hypothesis, this study determines its significance level to be 5% and a calculation of the parameter with the following formula will be made:

And an alternative formula:

In these formulas, d bar is the mean difference of the same sample between different learning stages, ssup2; is the sample variance, n is the sample size and t is a paired sample t-test with n-1 degrees of freedom.

When the calculation is completed, a comparison between calculated value and the table value will be made. We will reject the null hypothesis if the calculated value is bigger that the table value and vice verse. Thus we decide to accept one of our hypotheses made above after doing this research, say, there is some intra-group difference in syntactic maturity in sentence combining at the same level but between EFL learners#8217; of different learning stages or there is not.

3.2.2 Independent sample t-test

Independent sample t-test, taken into consideration, is a hypothesis testing procedure that uses separate samples for each treatment condition(between subjects design), used when the population mean and standard deviation are unknown, and 2 separate groups are being compared. For instance, in this case, we want to know whether or not the syntactic complexity in sentence combining of the same learning stage at different learning stages has any correlation.

Below are assumptions for the Independent t-Test :

#8226; Independence: Observations within each sample must be independent (they don#8217;t

influence each other)

#8226; Normal Distribution: The scores in each population must be normally distributed

#8226; Homogeneity of Variance: The two populations must have equal variances (the degree to which the distributions are spread out is approximately equal)

Independent sample t-test is implement in five steps: firstly, State the hypotheses H0: micro;1-micro;2=0 (micro;1=micro;2) H1: micro;1-micro;2≠0 (micro;1≠micro;2). This is a two-tailed test (no direction is predicted). Secondly, set the criterion, such as df=n1 n2-2. Thirdly, Collect sample data, calculate X and S. Fourthly, Compute the t-statistic. Fifthly, make a decision about the hypotheses. the following formula will be made:

Here, ，are sample mean of two different proficiency respectively;

， are sample variance of the two samples separately；

is the correlation index of correlation samples.

When the calculation is completed, a comparison between calculated value and the table value will be made. We will reject the null hypothesis if the calculated value is bigger that the table value and vice verse. Thus we decide to accept one of our hypotheses made above after doing this research, say, there is some intra-group difference in syntactic maturity in sentence combining at the same learning stage between EFL learners#8217; of different proficiency levels or there is not.

3.3 Data collection

The experiment was administered to 30 juniors of English major from one intact class of Nanjing University of Technology.

Subjects were required to finish a sentence-combining exercise within 20 minutes among the same junior students group, respectively from different learning stages: one at the beginning of the term; the other at the end of it. Here the same exercise Aluminum was assigned to subjects in order to remove variations in syntactic structure which might be due to differences in contents. Thus features of subjects#8217; syntactic complexity may be more clearly visible and measurable (Hunt, 1970). To ensure the accuracy and reliability of the data, the data of sentence-combining were transcribed by the same four undergraduates, who had received a special training before they started their job. Altogether 60 sentence-combining samples were derived from the experiment.

According to scores of TEM-4, the subjects were divided into 2 equal groups. Group one with its scores ranging from 51 to 69, was known as low-proficiency group. Group two with its scores ranging from 70 to 82, was known as high-proficiency group.

3.4 Data analysis

Syntactic complexity here is defined as great length and subordination of T-unit. The approach to syntactic complexity in this study is clauses per T-unit (C/T). The formula is shown as follows:

Notes: T= T-units; C=clauses

According to Hunt (1970), a T-unit is seen as one main or independent clause either plus whatever subordinate or dependent clauses attached to or embedded with it or none. While a clause is operationalized as a structure with an overt subject and a finite verb (Hunt, 1965), which includes independent clauses, as well as three types of dependent clauses: adverbial clauses, adjective/relative clauses and nominal clauses.

In previous studies, C/T measure increased with proficiency level no matter how proficiency was defined. In some cases, the middle levels were undifferentiated (Monroe, 1975; Cooper, 1976; Gui Bao, 2009), but there was a general progression upward from low to intermediate levels.

The data thus collected were processed though the following steps. First of all, the researcher tagged T-units and clauses in each text, and then a computation of C/T in each text was made according to the corresponding formulae. The methods applied here are respectively paired sample and independent sample T-test.

After attaining the syntactic complexity indices of the data, the researcher applied SPSS 13.0 to conduct analyses to answer the research questions. By comparing the data from different groups and different learning stages, the researcher can derive some new findings about EFL learners#8217; syntactic complexity in sentence combining.