再者，以使用的時機而言，高校英語學生常使用關系從句來陳述研究背景、目的和定義專門用語;關系從句也會用來闡述相關的概念、表達額外的信息和限制論述的主題;在研究方法的部分，則使用關系從句來說明研究對象的背景資料、限制研究對象、說明研究資料的來源和測驗研究對象的方法以及說明資料分析所采用的理論結構和分析的重點。在表達研究結果及結語的部分，關系從句則是用來確定資料的分析來源。

　　Thus, it was suggested that instructors introduce the concepts of grounding and description in relative clauses to help students better understand the restrictive and non-restrictive functions and teach them the academically common used forms of relative clauses in the class of academic writing; students can benefit from instructors’ introduction, and practicing using the academically used forms of relative clauses. Based on our findings, we offer pedagogical implications with the aim of hoping that students can learn the discourse functions of English restrictive relative clauses.

　　研究結果發現，在大學生的寫作中，限制性關系從句的先行詞常是泛指，而且限制性關系從句的內容常常是舊有的信息。因此， ;除了課堂上的外，學生也可透過自身的練習，來體會并改善關系從句的用法。

　　Based on the findings of this study, it was obvious that Chinese college students still have difficulty in distinguishing restrictive and non-restrictive relative clauses and overuse restrictive relative clauses in their proposals.

　　Finally, we provide a preliminary analysis of the discourse functions of Restrictive Relative Clauses in the middle and final positions of discourse units. Restrictive Relative Clauses in the final are likely to occur with given head NPs and serve proposition-linking grounding function. The entire NPs have the function of wrapping up the main points in the discourse. On the other hand, Restrictive Relative Clauses in the middle tend to provide description about the head NPs which themselves are not important referents in the discourse.

　　In this chapter, we examine English Restrictive Relative Clauses in the expository and argumentative compositions by Chinese junior English majors. The compositions were collected from twelve juniors of English Department in xxx [p7] University. All of them have taken required courses in their second and third years: English Writing I and II. So, we suppose that their proficient level of writing ability is comparable. The total compositions collected were 16 and the total number of the Restrictive Relative Clauses collected was 32. Like the analytical processes of native English speakers’ data, first, we examine the occurrences of the Restrictive Relative Clauses in discourse units in the students’ writings. Then, we classify these Restrictive Relative Clauses into different syntactic types: S-relatives, A-relatives and O-relatives in order to investigate the main discourse function of each syntactic type in the students’ writings. Also, we make an attempt to compare the strategies used by native speakers with those by the college students. This comparison aims to uncover the difference between them and the unobserved discourse functions of Restrictive Relative Clauses in the students’ writings.

　　In this section, we examine the positions of English Restrictive Relative Clauses in discourse units and the distribution of Restrictive Relative Clauses’ syntactic types in the students’ writings. We found that the number of Restrictive Relative Clauses in the beginning of discourse units is 17 which takes 55% of the total Restrictive Relative Clauses.

　　In Table 5, we can see that 55 % (17 out of 32) of the Restrictive Relative Clauses used by the college students occur in the beginning of discourse units and 45% (15 out of 32) of the Restrictive Relative Clauses occur in the middle and final positions of discourse units. Then, we divide the Restrictive Relative Clauses into different syntactic types, as sketched in Table 6.

　　In the students’ compositions, S-relatives and A-relatives have preponderance over O-relatives: S-relatives have the highest frequency, followed by A-relatives. O-relatives constitute the smallest group.

　　Then, we investigate the main discourse function of each syntactic type of Restrictive Relative Clauses. The remainder of this chapter is a detailed discussion on how each syntactic type of Restrictive Relative Clauses performs their discourse function in the beginning position of discourse units in the students’ argumentative and expository writings. In particular, we want to see if the factors and strategies involved in the native English speakers’ writings are also at work in the students’ compositions.

　　In the analysis of the S-relatives employed by the college students in the beginning of the discourse units, we found that the majority of the S-relatives occur with the generic head NPs in the subject position of sentences. The discourse functions of the S-relatives used by the students are summarized in Table 7.

　　In Table 7, we can see that most S-relatives in the students’ writings primarily perform proposition-linking grounding: 78% of the S-relatives perform proposition-linking grounding function whereas only 22% of the S-relatives serve description function. It is found that S-relatives tend to have generic NPs as their heads and perform proposition-linking grounding function.

　　According to the knowledge we gain from the analysis of native speakers’ writings: Restrictive Relative Clauses serve to make the head NPs become persistent and important referents in the ongoing discourse, it follows that the generic head NP people should be persistent in the following discourse. Thus, in this example, the pronoun they should occur in the discourse unit. But the author does not have the head NP further developed. Instead, she/he digresses to talk about the other issue. Besides, it is found that two S-relatives serve description function, providing new information for their new heads, as can be seen from Table 9. However, we found that these two S-relatives do not introduce the new head NPs as the topics in the following discourse.

　　(24) [1]Students in China are different from those in other countries. Because they need to go to cram schools after school. Their parents send them to cram schools to learn as much as possible, not only academic subjects but also music and art. Parents in China expect their children to be good at everything.

　　[2]Therefore, it is a trend that students in China need to take part in various afterschool programs and that will causes students live under too much pressure physically and mentally which is harmful to their learning process. Cram schools are not a normal educational system. It is like a getting-high-scores machine emphasizing only on efficiency. In that way, students will lose their interest in any subjects eventually.

　　In example (24), the S-relative provides new information for its new head NP pressure. However, the content of the S-relative being harmful to their learning process and the head NP pressure are not further developed in the discourse unit. As we can see from the remaining sentences of discourse unit [2], the entire new information pressure which is harmful to their learning process is not further developed. Instead, the author digresses to mention that cram schools have a negative influence on students’ learning. It is likely that the author has not fully understood that when a RRC provides new information for a new head NP in the beginning of a new discourse unit, the information coded in the RRC and its new head NP should be further developed in the subsequent discourse.
　　In Table 8, we can see that the A-relatives tend to function as proposition-linking grounding. 80 % of the A-relatives perform proposition-linking grounding function. In addition, the head NPs of the A-relatives serving proposition-linking grounding function tends to be generic NPs. Consider the following example.

　　(25) [1]In other countries, however, same-sex marriage has still not been legalized. In China, for instance, even if a same-sex couple have been together for a long time and decide to live a family life, they will soon find that their marriage is not guaranteed by the law and thus they cannot have the same right as heterosexual couple have in marriage, such as the right to adopt children and the right to inherit property.

　　[2]Besides, most people in China think that same-sex marriages have bad influence on the whole society; therefore, they strongly object to same-sex marriages. Same-sex couples are thus often not respected and are deprived of their rights by most people.

　　[3]People who oppose same-sex marriages say that a gay/lesbian couple will have bad influence on children because their family lacks a father or a mother. Children raised by homosexual couples may lack tender love and care and possibly their sexual identity may be confused.

　　In (25), the A-relative provides proposition-linking grounding function for the generic head NP people since the information coded in the A-relative opposing same-sex marriages has already been mentioned in the prior discourse unit.

　　The A-relative thus serves grounding function for the generic head NP people to delimit the domain of referents referred to by the generic head NP. As we can see from text (25), the content of the A-relatives and the generic head NP form given information which can be referred back to the content of discourse unit [2]. Moreover, the weakness of the use of the A-relative in text (25) is that the RRC does not ground the generic head NP people for subsequent development.

　　O-relatives constitute the smallest group of Restrictive Relative Clauses in the college students’ expository and argumentative compositions. Three O-relatives we found all occur with new head NPs. As we have mentioned, the purpose of using O-relatives is to ground the new head NPs by linking the new referents to the human subjects in O-relatives. Given human referents in O-relatives make new head NPs a relevant contextual part in the discourse and further introduce them as the topics in the subsequent. However, in students’ data, http://ukthesis.org/Thesis_Tips/ we found that only two O-relatives were used to develop the new head NPs as the topics in the subsequent discourse whereas the other one was not used as a vehicle for topic construction.