A probe into discourse structure of english relative clauses of Iranian EFL learners’ reading and writing performance: Centering theory in focus

Abstract

The purpose of this study was to explore the function of English restrictive and non-restrictive relative clauses in discourse cohesion and also intended to examine the possible differences between comprehension and production of restrictive relative clauses and non-restrictive relative clauses and their sub-types. Centering Theory was adopted as a framework to evaluate the coherence of discourse. To this end, 200 Iranian EFL language learners (112 females and 88 males) majoring in English as a foreign language, selected based on convenience sampling from Marvdasht and Shiraz Islamic Azad universities. They did two researcher-made tests, namely a 20-item reading comprehension test and a writing test. The reading comprehension test required the participants to read sentences containing a main and subordinate relative clause and choose the correct choice from 3 proposed answers. Likewise, to analyze their writing and productions in terms of relative clauses, the participants were asked to read 20 sentences consisting of a main clause followed by a blank in which they needed to create a sentence using a relative clause. algorithm was utilized to analyze the data. The results suggested that the participants understood and produced restrictive relative clauses better than non-restrictive relative clauses. Moreover, Identifying restrictive relative clauses and Continuative non-restrictive relative clauses were understood and produced better than Classifying restrictive relative clauses, Subjectivity, and Relevance non-restrictive relative clauses, respectively. The results of this study carry crucial pedagogical implications for teachers, language learners, and syllabus designers.

Keywords:

• centering theory
• english relative clauses
• restrictive relative clauses
• non-restrictive relative clauses
• discourse structure
• EFL learners

PUBLIC INTEREST STATEMENT

Producing intelligible and coherent discourse lies at the heart of effective communication. Given the syntactic complexity of relative clauses to language learners and their salient role in maintaining connected discourse, it is important to explore how language learners process relative clauses in reading comprehension and writing tasks. This study thus set out to examine the function of English relative clauses in discourse and to investigate what sub-types of restrictive relative and non-restrictive relative clauses are understood and produced more appropriately. Building on Centering theory as a framework to evaluate coherence, we found that EFL learners understood and produced restrictive relative clauses better than non- restrictive relative clauses in English. Likewise, identifying restrictive relative clauses and continuative non- restrictive relative clauses were understood and produced better than Classifying restrictive relative clauses, Subjectivity, and Relevance non-restrictive relative clauses, respectively. EFL teachers are suggested to address various subtypes of relative clauses in their instruction.

1. Introduction

Producing well-organized and comprehensible discourse is one of the main aspects of communication. Wang and Guo (2014) define discourse as a coherent combination of utterances arising from interacted contact between speakers, listeners, writers, or readers. Besides, a discourse denotes written or spoken communications intended to convey a social purpose (Tofiloski, 2009). It possesses different criteria, among which, according to Renkema (2004), coherence and cohesion are instrumental in the study of discourse structure and they are two salient linguistic notions which make a discourse easy to understand. While the former refers to the unity and purposefulness of a text, the latter forms connections between different parts of the text to create a coherent text. Cohesive devices hold the relations between lexical items and grammatical structures and connect clauses (Hoey, 2001). Coherence is like a glue that binds the structure of discourse. The most straightforward and central unit of meaning is the word, the larger unit is a phrase, then a clause, and the next is the sentence, and above the sentence is a text. The primary purpose of coherence is to combine these units to make a meaningful text (Castro, 2004; Halliday & Hasan, 1976). Grosz and Sidner (1986) divided discourse into two levels: global and local. While the former refers to the coherence over the whole text, the latter concerns coherence between adjacent sentences. Centering Theory (henceforth, CT) is one of the most dominant theories of coherence, a theory of local discourse structure focusing on the relationship between local coherence i.e., coherence between utterances in a segment and the choice of referring expression. It identifies how two utterances possess local cohesion (Grosz et al., 1995). Likewise, Joshi and Weinstein (1981) viewed CT as a model of inferential complexity. Grosz (1977) and Grosz and Sidner (1986) stated that “attentional state” is the fundamental part of a discourse that imparts essential information about the entities of utterances. If the attention does not shift to new entities from one utterance to the next, the discourse will be more coherent, but if the attention shifts to a new entity, the discourse will be less coherent.

Integral to CT are centers, in which semantic entities are found in the discourse model of each utterance. The most salient entity within an utterance is the preferred center, abbreviated as Cp. A list of forward-looking centers (Cf) is ranked according to the discourse salience. Walker et al. (1998) listed entities based on the grammatical role shown as Subject>object(s)>other.

Cp is the most essential item in the utterance, which has a strong influence on the selection of the next utterance. According to Grosz et al. (1983), backward-looking center (Cb) is the main item in the utterance that holds the discourse together. Cp and Cb can be the same or different. Based on the different arrangements of the Cp and Cb, there can be four transitions between two utterances: Continue, Retain, Smooth-shift, and Rough-shift. Tofiloski (2009, p. 27) defines these transitions as follows:

Continue: When the Cb of the current utterance is the same as the current Cp.

Retain: While there is a new preferred center and the previously preferred center is in the current utterance.

Smooth-shift: When Cb of the two utterances is not the same.

Rough-shift: When the Cb of the utterances is not the same, and the preferred center was not the highest-ranked entity in the previous utterance.

According to Walker et al. (1998, p. 3), the constraints of CT are as follows:

1. Each utterance must have one backward-looking center.

2. The most salient element of forward-looking is the center.

Walker et al. (1998, p. 4) also stated that Centering Theory has two main rules:

1. If Cf in an utterance is a pronoun, so the Cb must be a pronoun.

2. Centering transition is ordered according to their salience.

CONTINUE> RETAIN > SMOOTH SHIFT >ROUGH-SHIFT

According to the Pronoun Rule and the Centering transitions, the hearer better comprehends sentences that require minimal processing effort. If the discourse contains continue transitions, the hearer or writer comprehends it effortlessly, since the Cb and Cp of utterances are the same (Miltsakaki, 2003).

Centering is a theory which best explains anaphora resolution defined as how anaphors refer to their antecedent in the discourse which contributes to a coherent sequence of utterances (Miltsakaki, 2003). There is a relationship between the attentional state and different kinds of referring expressions in different languages, and Cf-ranking and Rule 2 of CT help pronoun resolution (Frossard et al., 2001; Fais, 2004; Hedberg & Dueck, 1999; Iida, 1998; Kim et al., 1999; Miltsakaki, 2001, 2002; Prasad & Strube, 2000; Roh & Lee, 2006; Taboada, 2002, 2005; Taboada & Hadic Zabala, 2008; Tetreault, 2001).

Furthermore, based on the considerable research conducted on the unit of a discourse in CT at the local level, the clause is the central unit of discourse analysis (Miltsakaki, 2002, 2003, 2005; Poesio et al., 2000; Taboada & Hadic Zabala, 2008; Thompson & Couper-Kuhlen, 2005). Given the syntactic complexity of Relative clauses (RCs) to EFL learners, it is useful to perform research as to how language learners process RCs in reading comprehension as well as writing tasks. Within relative clause taxonomy, a distinction is made between the restrictive and nonrestrictive relative clauses. De Haan (1987) distinguished between identifying restrictive relative clauses and classifying restrictive relative clauses (RRCs) within RCs. To him, identifying RRCs have definite head nouns, while classifying relative clauses provide new subclasses for the reference or the head noun. Loock (2005, 2006, 2007, 2010a, 2010b)), and Loock and O’Connor (2013) made a distinction between three different types of non-restrictive relative clauses (NRRCs): Continuative NRRCs, Relevance NRRCs, and Subjectivity NRRCs.

English sentences containing embedded clauses are difficult to comprehend for EFL learners since they need to understand the embedded clauses and track the main clause (Kuno, 1972). Iranian EFL learners have difficulty comprehending and producing relative clauses in English. As a theory of local coherence in the discourse, Centering Theory can assist EFL learners to produce well-organized and comprehensible discourse by using centering rules and transitions of CT. Further, EFL teachers can profit from using the centering algorithm to analyze and score learners’ essays. Given their crucial effects on discourse processing, considerable research has been conducted on complex sentences in Centering Theory. Hence, there is a need for research to shed light on the understanding and production of English RCs, which contribute to the cohesion of discourse and referring expressions, including pronouns within clauses. Notwithstanding a plethora of research on centering theory as a well-known model of discourse processing with implications on the pronominal resolution, a review of the literature indicates that no studies have thus far investigated EFL learners’ understanding and producing of English RCs. Therefore, the current study seeks to explore how RCs contribute to cohesion and address the incoherence found in language learners’ writing and reading skills.

The findings of this study might prove fruitful to EFL teachers as well as language learners. EFL teachers can use Centering Theory as a model to make language learners attend to specific parts of essays where topic discontinuity occurs. They can also use Centering transitions, constraints, or algorithms to evaluate language learners’ essays. Likewise, EFL teachers can translate Centering transitions and rules into instructive feedback for language learners. Informed of these rules and transition, language learners can produce more coherent writing productions.

Researchers (Miltsakaki (2002, 2003, 2005), Poesio et al. (2000), Poesio et al. (2004a), and (2004b))) believe that to understand the roles of English relative clauses (ERCs) in discourse processing is of great importance. Following their lines of research related to these constructions, this study, as its main goal, investigates whether restrictive and non-restrictive clauses are understood and produced by our EFL participants appropriately. Given the emphasis on understanding the role of RCs in discourse cohesion, this study examines the function of English relative clauses in discourse. Different sub-types of restrictive (identifying, Classifying) and non-restrictive (subjectivity, relevance, and continuative) relative clauses have been identified. Another objective is to investigate what sub-types of restrictive relative clauses (RRCs) (Identifying or Classifying), and what sub-types of non-restrictive relative clauses (NRRCs) (relevance, subjectivity, or continuative) are understood and produced more appropriately. The results obtained from RRCS and their sub-types are compared to those of NRRCS and their subtypes to examine if there exists a significant difference among them or not as another goal of this study. More specifically, the following research questions were addressed:

1. Do Iranian EFL learners understand and produce restrictive relative clauses better than non- restrictive relative clauses in English?

2. What sub-types of restrictive relative clauses i.e., identifying or classifying, are understood or produced better by Iranian EFL learners?

3. What sub-types of non-restrictive relative clauses, including subjectivity, continuative, or relevance, are comprehended or produced better by Iranian EFL learners?

2. Literature review

Extensive research has been carried out to find the relationship between the attentional state and referring expressions in different languages (e.g., Di Eugenio, 1998; Fais, 2004; Frossard et al., 2001; Iida, 1998; Miltsakaki, 2001, 2002, 2007; Prasad & Strube, 2000; Roh & Lee, 2006; Strube & Hahn, 1996; Taboada, 2002, 2005; Taboada & Hadic Zabala, 2008; Tetreault, 2001).

Miltsakaki and Kukich (2004) used an e-rater essay scoring system to evaluate the local coherence by measuring Rough-Shift transitions to examine whether it could help evaluate essays. They concluded that Rough-Shift transitions contributed to the accuracy of computer-generated essay scores and that the e-rater essay scoring system could be beneficial for both students and instructors. Moreover, they discovered that students could use the Rough-shift measure to identify their weak points and textual discontinuities in writing essays. Additionally, Poesio et al. (2000) and Poesio et al., 2004a, Poesio et al., 2004b) investigated different forms of utterances in Centering Theory such as the sentence, the finite clause, and the non-finite clause. In their research, they focused on Rule 1 and Constraint 1 of Centering. Based on their observations, sentences were preferred over the finite clause concerning Constraint 1, whereas, the sentence could not be the unit of analysis in terms of Rule 1. Further, Miltsakaki (2005) investigated the salience of entities in the main and relative clauses in English and Greek. She proposed, “if entities evoked in relative clause were of equal salience status as entities evoked in main clauses, then a single clause condition should yield more ‘coherent’ transition than the complex sentence condition” (p. 183). Nevertheless, her results showed that was not true. She also concluded that “subjects of relative clauses do not always warrant pronominal reference even in the absence of more correct competing antecedents” (p. 183). Likewise, she evaluated the hypothesis of whether non-restrictive relative clauses could be separate utterances or units of segmentation in Centering Theory. Analyzing 200 NRRCs with the Centering algorithm, she concluded that thirteen utterances were more cohesive based on Centering transitions, whereas 46 utterances were less cohesive.

Conducting a study on a 450-utterance contextualized corpus, Loock (2007) suggested a taxonomy of ARC classifying NRRCs into Continuative NRRCs, Relevance NRRCs, and Subjectivity NRRCs. To him, Continuative NRRCs show that the information conveyed by the ARC is semantically dependent on the information provided by the main clause. The second type of NRRCs in his taxonomy is Relevance NRRCs that the speaker uses to give additional information. The last category of NRRCs taxonomy is the subjectivity appositive relative clause. The speaker uses it to convey an opinion, a judgment, or a comment. He proved that Continuative NRRCs are too difficult to understand. In a similar vein, Loock and O’Connor (2013) confirmed that Subjectivity NRRCS and Relevance NRRCs are produced and understood easier than Continuative NRRCs.

As stated, in almost all of the studies, the corpus-based study was used to investigate the discourse relations, coherence of the text, and anaphora resolution. Likewise, other studies attempted to seek the best way to segment the discourse. Taboada and Hadic Zabala (2008) evaluated different kinds of discourse segmentation, including Miltsakaki’s (2003) sentence-based approach, Kameyama’s (1998) hierarchical clause-based approach, and Poesio et al.’s (2000) clause-based approach. Although they favored clause-based segmentation, they stated that the sentence was a better unit of discourse.

Attempting to extend Centering Theory to measure the coherence of a text, Tofiloski (2009) pointed out that the leading role of Centering Theory is to contribute to anaphora resolution. He also claimed that CT measures coherence more accurately by considering all entities within an utterance.

Wiesemann (2009) explores the best method to segment relative clauses in the English and Spanish languages based on Centering Theory. The focus of her study was to divide discourse into smaller units of analysis. Attempting to measure the degree of cohesion between an utterance containing an RC and its immediate co-text, as modeled by Centering Theory, she concluded that using different kinds of RCs in a text could improve its cohesion.

Doing a corpus study to investigate the discourse relations and entity coherence of Wall Street Journal texts, Louis and Nenkova (2010) realized that discourse relations and switches of the focus of attention between discourse entities were integral to local coherence because they could connect adjacent sentences in a text. They also found that the two types of coherence i.e., relational and entity, were complementary to each other and account for most local coherence.

As evident from the studies reviewed above, extensive studies have been undertaken to explore the role of CT in producing a coherent text. The researchers attempted to examine the relationship between the attentional state and the form of referring expressions in various languages. However, no studies have investigated EFL learners’ understanding and production of English RCs. What distinguishes the present study from the ones reviewed above is that these studies evaluated CT employing a corpus-based approach. However, the current study assessed EFL learners’ comprehension and production of ERCs within the framework of CT.

2.1. Method

The current study adopted a quantitative approach to answer the research questions. Quantitative research involves the utilization and analysis of numerical data using specific statistical techniques to quantify and analyze variables to obtain the intended results (Creswell, 2003; Leedy & Ormrod, 2001).

2.2. Participants

A total of 200 (112 females and 88 males) Iranian M.A. students at Islamic Azad University, Marvdasht, and Shiraz branches selected based on convenience sampling participated in the current study. The participants were majoring in English in their first and third semesters of the academic year 2017–2018 with their ages ranging from 23 to 38 and Farsi as their first language. Noteworthy to mention is that before the test administrations, the participants filled out the consent form for participation in the study. They were also assured that their data would be kept confidential and their test scores have no effect on their end of the term grades.

3. Instruments

Two different instruments, including a reading comprehension task and a writing test were deployed in this study.

3.1. Reading comprehension test

The reading comprehension test (see Appendix A) consisting of 20 multiple-choice items was constructed by the main researcher. Each item included a passage of connected discourse with certain relative clauses, including both restrictive and non-restrictive relative clauses followed by three alternatives. De Haan’s (1987) taxonomy of RRCs, Identifying RRCs and Classifying RRCs, and Loock’s (2007) classification of NRRCs, Continuative, Relevance, and Subjectivity were adopted to construct the reading comprehension test.

The reliability of the measurement instrument was assessed using Cronbach’s alpha. Table 1 depicts the reliability indexes of the reading comprehension test.

Table 1. Reliability analysis of the comprehension test

Different Taxonomy of RRCs and NRRCs	Cronbach’s Alpha
Relevance	.701
Continuative	.721
Subjectivity	.708
Identifying	.741
Classifying	.731

The Cronbach’s alpha values for the variables turned out to be more than 0.7, which are acceptable (Salkind, 2007), confirming the reliability of the questions.

A panel of three EFL experts was asked to check the reading comprehension tests concerning content validity, ambiguity, and appropriateness. Their feedback on the wording of the questions and their suggestions were taken into account. To ensure the content validity of the questions, the questions and multiple choices were examined carefully and critically by the same panel of experts. Accordingly, some questions were either discarded or modified.

To measure the construct validity, internal consistency was used. The internal consistency was performed by SPSS 21. As Table 2 shows, there are significant relationships between each item and the total grade. Therefore, the validity of the reading comprehension test was confirmed.

Table 2. Construct validity analysis of the comprehension test

Items	Classifying	Identifying	Subjectivity	Continuative	Relevance
1 2 3					0/531 0710 0/517
4 5				0/551 0/727
6 7 8 9			0/467 0/628 0/554 0/531
10		0/526
11 12					0/531 0/536
13 14 15		0/432 0/412 0/512
16	0/631
17 18 19	0/710 0/517 0/551
20		0/621

3.2. Writing test

The second test was the researcher-made writing test (see Appendix B) consisting of 20 open-ended questions. The participants were asked to read 20 sentences consisting of a main clause followed by a blank. They were required to produce a sentence using a second clause (a relative clause). Here again, De Haan’s (1987) taxonomy of RRCs, Identifying RRCs and Classifying RRCs, and Loock’s (2007) classification of NRRCs, Continuative, Relevance, and Subjectivity were used to design the writing test. To construct this test, all transitions and rules of Centering Theory were considered. Three EFL experts were also asked to check the writing test concerning content validity, ambiguity, and appropriateness.

The reliability of the writing test was assessed using Cronbach’s alpha. Table 3 shows the reliability indexes of the writing test.

Table 3. Reliability analysis of the writing test

Different Taxonomy of RRCs and NRRCs	Cronbach’s Alpha
Relevance	.721
Continuative	.702
Subjectivity	.800
Identifying	.764
Classifying	.730

Since the Cronbach’s alpha values for the variables turned out to be more than 0.7, the reliability of the test was confirmed.

Moreover, the construct validity of the writing test was measured by internal consistency whose results as depicted in Table 4 demonstrate a significant relationship between the items of the writing test and the total grade, thereby confirming the construct validity of the test.

Table 4. Construct validity analysis of writing test

Items	Classifying	Identifying	Subjectivity	Continuative	Relevance
1 2		0/710 0/517
3	0/551
4				0/727
5				0/467
6 7 8					0/682 0/554 0/531
9 10 11			0/526 0/531 0/536
12					0/424
13			0/412
14 15				0/512 0/779
16		0/654
17		0/629
18 19 20	0/504 0/517 0/623

3.3. Data collection procedure

The data collection lasted ten months from October 2018 to July 2019. In the first session, the reading comprehension test was administered. The participants were asked to read sentences containing a main and subordinate relative clause and choose the correct choice from 3 proposed alternatives. It took the language learners 40 minutes to respond to the test.

The second test was the writing test, whose items involved a connected discourse in which the participants were required to generate a sentence in the blanks using a relative clause. The stem provided EFL learners with sufficient context so that the participants could complete the blanks with an appropriate relative clause. It took them 60 minutes to undertake the task. To decrease the memory factor, a two-week interval period was observed between the administrations of the two tests.

3.4. Scoring the tests

Choosing the correct answer in the comprehension test, the participants received a score of one. The participants choosing the wrong choice received a score of zero.

Concerning the writing test, Miltsakaki’s (2003) algorithm of anaphora resolution was employed to analyze and score the participants’ answers. First, the last identified Ui was chosen. The antecedent from the Cf list was then identified. The next step was to apply grammatical rules, e.g., gender or number agreement, using appropriate relative pronouns, etc. By focusing on Cf lists, pronominals were resolved. If there was another relative clause, all of the steps above were followed (Miltsakaki, 2003). In example (1), it shows how the algorithm is applied to resolve the relative clauses in discourse.

(1) Today I invited David, who is my best friend, to watch the sunset.

-Have you ever been here, David?

-Yes, I have.

-Really?

-I came here with my cousin-2 who- 1 you saw last year in London.

1. The antecedent is my cousin which is the object of the preposition.

2. Grammar-driven constraints were applied. The relative pronoun who agrees with the antecedent my cousin. The Cf-list contains my cousin> who. The pronoun who-1 is identified. The Cf- list from the previous unit contains my cousin >who. Who-1 resolves to my cousin.

3. the algorithm terminates.

According to this algorithm, the score for each response ranged from 0 to 9.

The main researcher (the first scorer) evaluated the participants’ responses and scored them. After one week, the same rater analyzed the participants’ responses. In this way, the intra-rater reliability was measured. The intra- rater reliability was calculated using the Pearson product-moment correlation coefficient. The results are summarized in Table 5.

Table 5. Intra-rater reliability (writing test)

	First Scoring	Second scoring
First Scoring Pearson Correlation Sig. (2-tailed) N	1 200	707** .000 200
Second Scoring Pearson Correlation Sig. (2-tailed) N	.707** .000 200	1 200

**Correlation is significant at the 0.01 level (2-tailed)

As depicted in Table 5, there was a high consistency between the two analyses of data by the same rater (r = .707, sig. = .00), indicating high intra-correlation reliability (Cohen, 1988).

Regarding the inter-reliability of the writing test, the first rater scored the data. Then, another rater analyzed the participants’ responses once more based on the algorithm of CT, and the inter-rater reliability was calculated. The pertaining results are summarized in Table 6.

Table 6. Inter-rater reliability (writing test)

	Grade 1	Grade2
Grade 1 Pearson Correlation Sig. (2-tailed) N	1 200	.993** .000 200
Grade 2 Pearson Correlation Sig. (2-tailed) N	.993** .000 200	1 200

**Correlation is significant at the 0.01 level (2-tailed).

Based on the results concerning the writing test, as depicted in Table 6, the correlation between the two analyses performed by two raters turned out to be .993 (sig. = .00). Therefore, there was a high inter-reliability between the two ratings.

4. Data analysis

The data were analyzed quantitatively using descriptive and inferential statistics. The Chi-square tests were performed to explore relationships among variables. The lowest score and the highest score were considered in each test, and then the theoretical median (Middle value separating the greater and lesser halves of a data set) of the answers was chosen.

5. Results and discussion

5.1. Results of the first research question

The first research question investigated if the participants understood and produced English restrictive relative clauses better than non- restrictive relative clauses.

To answer this question, we employed a Chi-square test, the results of which are shown in Table 7.

As shown in Table 7, the EFL learners understood and produced RRCs better than NRRCs in English. The lowest and highest scores turned out to be 0 and 1, respectively, for RRCs and NRRCs, and the theoretical median of the answer was 0.5. Since the Sig (.001) was less than 0.05, there was a significant difference between RRCs and NRRCs. According to the results of the Chi-square test, the mean of RRCs was 0.668, which is higher than that of NRRCs (0.599), indicating that EFL learners understood and produced RRCs better than NRRCs.

Table 7. Chi-square test for RRCs and NRRCs

	N	Mean	Minimum	Maximum	Theoretical Median	Chi-Square	df	Asymp. Sig.
RRC	181	.6685	.13	1.00	0.5	69.298^a	7	.001
NRRC	181	.5990	.08	1.00	0.5	70.906^b	10	.001

The minimum expected cell frequency is 16.5.

5.2. Results of the second research question

The second research question read, “What sub-types of restrictive relative clauses i.e., Identifying or Classifying are understood or produced better?”

To examine whether Identifying relative clauses were understood and produced better than classifying relative clauses, a Chi-square test was performed. The pertaining results are presented in Table 8.

Table 8. Chi-square test for identifying and classifying relative clauses

	N	Mean	Minimum	Maximum	Theoretical Median	Chi-Square	df	Asymp. Sig.
Identifying	191	.6767	.00	1.00	0.5	72.586^a	4	.001
Classifying	191	.6545	.00	1.00	0.5	44.00	4	.001

a. 0 cells (0.0%) have expected frequencies less than 5. The minimum expected cell frequency is 38.2.

Here again, the lowest and highest scores were calculated to be 0 and 1, respectively, for identifying and classifying relative clauses, and 0.5 was chosen as the theoretical median of the answers. Based on the results of the Chi-square test, the significance level (.001) was less than 0.05; therefore, there was a significant difference between identifying relative clauses and classifying relative clauses. According to the findings in Table 8, the mean of identifying relative clauses (0. 676) was higher than that of classifying relative clauses (0. 654), suggesting that identifying relative clauses were understood and produced better than classifying relative clauses.

5.3. Results of the third research question

The third research question examined the sub-types of non-restrictive relative clauses i.e., Subjectivity, Continuative, or Relevance, which were comprehended or produced better.

A Chi-square test was conducted to investigate if subjectivity relative clauses were understood and produced better than continuative and relevance relative clauses.

As shown in Table 9, the lowest score and highest scores turned out to be 0 and 1, respectively, for relevance, continuative, and subjectivity relative clauses. Moreover, 0.5 was chosen as the theoretical median of the answers. The significance level of .001 (p < 0.05) indicated that there were significant differences between continuative relative clauses, relevance relative clauses, and subjectivity relative clauses. Furthermore, the mean of continuative relative clauses (0.664) was higher than those of subjectivity relative clauses (0.639) and the relevance relative clauses (0.520), suggesting that continuative relative clauses were understood and produced better than subjectivity and relevance relative clauses.

Table 9. Chi-square test for relevance, continuative, and subjectivity

	N	Mean	Minimum	Maximum	Theoretical Median	Chi-Square	df	Asymp. Sig.
Relevance	183	.5202	.00	1.00	0.5	74.022^a	4	.001
Continuative	183	.6648	.00	1.00	0.5	51.645^b	3
Subjectivity	183	.6393	.00	1.00	0.5	40.197^a	4	.001

a. 0 cells (0.0%) have expected frequencies less than 5. The minimum expected cell frequency is 36.6.

b. 0 cells (0.0%) have expected frequencies less than 5. The minimum expected cell frequency is 45.8.

5.4. Discussion

Given the emphasis on understanding the role of RCs in discourse cohesion, the current study sought to explore the function of English relative clauses in discourse. Different sub-types of restrictive relative clauses (Identifying, Classifying) and non-restrictive relative clauses (Subjectivity, Relevance, and Continuative) were identified. The study also investigated the sub-types of restrictive and non-restrictive relative clauses, which were understood and produced more appropriately. The results obtained from RRCS and their subtypes were compared to those of NRRCS and their subtypes to examine whether there existed significant differences among them.

5.5. Discussion of the first question

The first question addressed the issue of whether students understand and produce restrictive relative clauses (RRCs) better than non-restrictive relative clauses (NRRCs) in English. The results of the first research question manifested that the participants understood and produced RRCs better than NRRCs in English. In this regard, Miltsakaki (2003) examined the differences between restricting and non-restricting relative clauses. She concluded that the type of relative clauses had a significant effect on subsequent reference to the head noun. However, the type of relative clause did not affect the topical status of the head noun referent. Consistent with our Miltsakaki and Kukich (2004), Miltsakaki (2005)) highlighted the role of relative clauses within the framework of Centering Theory in textual cohesion. In a similar vein, Poesio et al. (2000), Poesio et al. (2004a), and (2004b)) pointed to the influential role of English relative clauses in discourse processing.

Investigating the role of NRRCs in relational coherence, Ferrari (2005) found that NRRCs were minimal units of a text and that there were pragmatic relations between NRRCs and their main clauses and other units of co-text.

The results of the current question are congruent with those of Grodner et al. (2005), who used a self-paced reading methodology to investigate the processing of restrictive and non-restrictive relative clauses in a supportive and null context. Their findings showed that the processing of restrictive relative clauses was harder than the processing of non-restrictive relative clauses in a null context. However, the processing of non-restrictive RCs was more difficult than restrictive RCs in a supportive context.

Our findings are in sharp contrast with those of Wiesemann (2009) in that RC type did not have any effect on the dependent variables “Pre-RC Degree of Cohesion, post-RC Degree of Cohesion, and Subsequent Mention.” Thus, the RC type did not have any effect on the cohesion of the text.

5.5.1. Discussion of the second question

The second question asked whether the identifying relative clauses are understood and produced better than classifying relative clauses. The results of the second research question illustrated that identifying relative clauses were understood and produced better than classifying relative clauses. As put forth by Prince (1990), resumptive pronouns can appear in RRCs that modify definite heads, while RRCs which modify indefinite heads cannot have resumptive pronouns. If RRCs describe definite heads, it constitutes known information. Therefore, Prince (1990) concluded that there was a significant distinction between identifying and classifying RRCs.

However, our findings are inconsistent with those of Wiesemann (2009), who concluded that there was not any significant difference between Identifying and Classifying RRCs. Based on her study, RRCs (Identifying and Classifying) followed by a referring expression could help identify antecedents in RCs. She also found that the number of RCs in the Hierarchical—Hypotactic approach was less than RCs in the sequential approach.

5.5.2. Discussion of the third question

The third research question investigated the sub-types of non-restrictive relative clauses (Subjectivity, Continuative, or Relevance), which were comprehended or produced better. The analysis of quantitative data indicates that relevance, continuative, and subjectivity had a significant role in textual cohesion. The findings show that continuative relative clauses were understood and produced better than subjectivity and relevance relative clauses. This is in contrast with Loock’s (2006, 2007, 2010a, 2010b)) and Loock and O’Connor (2013) findings. Loock (2006) found that continuative ARCs were too difficult for his subjects to understand. Only four subjects out of 30 in his study labeled continuative ARCs as perfectly clear. Therefore, the continuative ARCs in a text are essential to their comprehension.

Loock (2006) confirmed that subjectivity ARCs were more suppressible than the continuative ARCs and the relevance ARCs. He held that one can use subjectivity ARCs to add her/his point of view, so if he/she deletes this kind of ARC from the text, the addressee could understand the discourse easily.

Although Loock’s (2006, 2007)) findings showed that continuative NRRCs were the most difficult NRRCs type to understand, the results of this study represent that the participants understood and produced this kind of NRRCs better than subjectivity and relevance NRRCs.

6. Conclusion and implications

Coherence, a defining feature of discourse, functions as a link between the units of a text to maintain connected discourse. Iranian EFL learners encounter difficulties in producing coherent texts, or they may not understand the logical connection between the units of the texts. CT is a practical approach by which researchers can evaluate and model discourse coherence within a text. According to its pronoun rule and transitions, the reader and writer can comprehend and produce more coherent texts. Since the clause can be the unit of a discourse in CT at the local level, we chose relative clauses as the central unit of discourse analysis and examined whether restrictive or non-restrictive relative clauses were appropriately understood or produced by EFL learners. Likewise, the sub-types of RRCs (Identifying and Classifying) and NRRCs (Continuative, Subjectivity, and Relevance) were identified.

An important finding of this study is that Iranian EFL learners could understand and produce RRCs better than NRRCs in English. Thus, according to the results, RRCs substantially modifying their previous nouns, are easier to understand and produce than NRRCs. Besides, using RRCs can promote textual cohesion. The present research also confirms Wiesemann’s (2009) study in which Centering Theory was found to be the best model depicting how RCs contribute to cohesion.

Another significant outcome of this study was that identifying RRCs and continuative NRRCs were the most accessible type of RCs to understand and produce. Thereupon, EFL learners are suggested to be more meticulous about using various subtypes of RCs properly. Informed of subtle differences of diverse forms of NRRCs and RRCs, EFL learners can appropriately use them in their essays, making their essays more coherent.

Centering rules and transitions were employed to construct comprehension and written tests and to evaluate the EFL learners’ responses to written tests. Consequently, given the prominence and saliency of CT rules and transitions as well as the awareness of RCs and their subtypes, EFL teachers are proposed to integrate centering rules into their instruction. They are also suggested to use RCs in test construction and evaluation of EFL learners’ essays. Results of such an evaluation may assist teachers in measuring the coherence of EFL learners’ essays. Besides, centering rules and transitions provide EFL learners with a tangible way to analyze their texts.

The findings of the current study may contribute to both EFL teachers and language learners. For example, Iranian M.A. students majoring in English-related courses are required to do their theses in English in which they should be able to produce well-organized and comprehensible discourse. Equipped with the knowledge of RCs and their usage, they can produce coherent and comprehensible discourse. EFL teachers are suggested to raise their language learners’ awareness of the differences between RRCs and NRRCs and their subtypes to facilitate discourse understanding and production. Besides, EFL teachers can benefit from using the centering algorithm to analyze and score language learners’ writing productions and inform EFL learners of their strengths and weaknesses. Consequently, language learners will be able to comprehend reading texts more easily and to produce well-organized texts. Overall, the results of the present study hold some significant pedagogical implications for syllabus designers and material developers. Syllabus designers and program developers can incorporate different relative clauses and their subtypes into EFL textbooks where EFL learners can find sufficient exercises on relative clauses, different kinds of relative clauses, and their functions in the discourse. These textbooks can also contain interactive and task-based activities requiring language learners to use the correct form of relative clauses to communicate. Such activities can facilitate learners’ use of different relative clauses by creating contexts in which these clauses are required. Further, books on writing and reading skills should provide language learners with the necessary tasks and activities concerning centering rules and transitions.

Overall, the study offers new insights into the discourse structure of English relative clauses within the framework of CT. Even though much remains to be done, the results of the present study may be of high relevance to language pedagogy. However, this study suffers from some limitations. The first one stems from the sample size. Only 200 Iranian EFL learners selected from Islamic Azad University, Shiraz, and Marvdasht Branches were involved in the study. Undoubtedly, larger sample sizes may yield more reliable and generalizable results. Another limitation concerns the lack of a specific e-rater essay scoring device that can be used for scoring language learners’ answers to writing tests and analyze the coherence of their text. Moreover, the findings of this study concern advanced EFL learners. Therefore, caution should be exercised when interpreting the results concerning EFL learners of other proficiency levels. More studies are then called for concerning the function of English restrictive and non-restrictive relative clauses in discourse cohesion of EFL learners of other proficiency levels.

Additional information

Funding

The authors received no direct funding for this research.

Notes on contributors

Masoomeh Benshams

Masoomeh Benshams is a Ph.D. candidate of TEFL at Islamic Azad University Shiraz branch, Iran. Her research areas include second language acquisition, teacher education, and syntax studies.

Firooz Sadighi

Firooz Sadighi is a professor of applied linguistics. He received his Ph.D. from the University of Illinois. His research areas include first/second language acquisition, second language education, and syntax studies.

Mohammad Reza Falahati Qadimi Fumani

Mohammad Reza Falahati Qadimi Fumani is an associate professor of computational linguistics at the Regional Information Center for Science and Technology, Shiraz, Iran. His areas of interest are linguistics, contrastive linguistics, and computational linguistics.

Naser Rashidi

Nasser Rashidi is a professor of applied linguistics in the English department of Shiraz University, Iran. He has presented and published many papers in different (inter)national conferences and reputable journals. His areas of interest include critical pedagogy, critical discourse analysis, and teacher education.

Appendix A.

Reading Comprehension: Read the following sentences, then choose the correct choice. You have 40 minutes to choose the correct answers

1. A: Madam, I’d like to know some information about the Eastern Island.

B: It’s one of the most beautiful places on the earth and is famous for its large statues. These interesting statues, which were carved by ancient people to resemble human heads, line its coast.

The large statues are the most beautiful place.

1. Eastern Island is made by ancient people.

2. The large statues located near the island’s coast.

3. The large statues are the most beautiful place.

1. Brockenhurst School, which opened as an educational experiment in 1974, is to close this summer. All the teachers who are currently employed by school will be relocated to the other school.

   1. Brockenhurst School will be moved very soon.

   2. The teachers will not be unemployed.

   3. Teachers have been employed since 1974.

2. -How may I help you?

-I want to buy a plane ticket.

-I can help you with that. Where are you traveling to?

- London.

- Sure. I have booked a seat on a flight to London in the morning. Your ticket will arrive in the mail within three days. What you’ve got is E and G which is aisle and next to it on the window.

a. E and G are the gates in the airport.

b. The travel agent will book some tickets for the next three days.

c. E and G are their seats.

1. Rose, who crashed her car into a valley, was faced down in the water and airlifted to the hospital where died hours later.

   1. Rose died in a plane crash.

   2. Rose died in the hospital.

   3. Rose died because she had drowned.

2. The first book John took from the library was Darwin’s Origin of Species, which is inspired by the dream of becoming a geologist.

   1. John was a geologist.

   2. John studied the origin of Species.

   3. Darwin’s studies give the idea of becoming a geologist.

3. -How many hours a day do you work?

   • We are all working too long hours, which is making us ill, less productive, and is not supporting life.

     1. They are not happy to work long hours.

     2. They believe that working too hard is beneficial.

     3. To live their lives, they don’t have to work hard.

4. My friend goes into debt for the kids. He spends five hundred dollars on them, and it’s enough which I think is silly but that’s the way of things today.

   1. His friend borrowed five hundred dollars.

   2. The speaker believes that his friend must not spend all his money.

   3. The kids borrowed five hundred dollars from his father.

5. A plane with 155 passengers on board had a crash landing. Only a few were rescued which I think was a miracle.

   1. All of the passengers were rescued.

   2. All of the passengers died.

   3. It’s incredible that some passengers were not killed in the crash.

6. He had been to see a film with his friend, Home Alone 2, which wasn’t as good as Home Alone 1, and his friend had said that modern films were too commercial.

   1. Both of them had seen Home Alone 1.

   2. Both of the films were good.

   3. His friend believed that modern films are better than classic films.

7. Robinson College asks candidates to complete a questionnaire on their educational background, which is then used to help decide whether they are offered a place.

   1. The candidates have to complete a questionnaire to be accepted at the college.

   2. All of the candidates will be accepted after completing the questionnaire.

   3. Candidates’ educational backgrounds are not essential for accepting at the college.

8. - Helen volunteered to help with the program. She had to go to San Francisco, which is in northern California.

   1. Helen had to go to another state to work.

   2. San Francisco is near California.

   3. Helen went on vacation to San Francisco.

9. We visited a seafood restaurant, which has been opened since 1888, in the afternoon, and had an early dinner. The waiters, who were all dressed in traditional fishermen’s clothes, were very friendly and told us about the history of the restaurant whose name was The Joly Whaler.

   1. The fishermen worked there as waiters.

   2. The restaurant dates back to the 17^th century.

   3. The fishermen told them the history of the restaurant.

10. We went to Vancouver where I have some old friends. They have some dinosaur skeletons that local people have found in the area.

    1. We have some old friends in Vancouver.

    2. Dinosaur skeletons have been discovered by archaeologists.

    3. We stayed with our friends in Vancouver.

11. The scientists who have strong ideas about how the future might be based on his knowledge of science is Ray Kurzweil. He was born in 1948 and grew up in New York City. He is an inventor and is well known for his work in technology.

    1. Ray Kurzweil studies how the future should be.

    2. Ray Kurzweil predicts the future.

    3. Ray Kurzweil studies about the future invention.

12. The new restaurant that Mr. Jones bought was losing money, was dirty inside, and the food was terrible, greasy and undercooked.

    1. The owner of the restaurant is Mr. Jones.

    2. The new restaurant was making a profit.

    3. The food tastes very well.

13. Last year my best friend, Sophie, invited me to stay with her during the summer holiday. Sophie lives in Nancy, which is a beautiful town in the eastern part of France. She has a big apartment that overlooks the river.

    1. Sophie lives in Nancy during the summer holiday.

    2. Nancy is Sophie’s friend.

    3. We could see the river from her apartment.

14. The skies were blue on Thursday, and we spent some time out on the sea in a large boat that we hired. I caught a big fish that the captain said was the biggest he’d seen this year. I felt very proud. We left on Thursday evening after a mini-vacation that helped me to relax a lot, and now I have returned to work.

    1. They hired the biggest boat.

    2. They had an interesting and long vacation.

    3. They went to a city by sea for their vacation.

15. A volunteer is someone who helps other people but doesn’t get paid. According to American publication, %26 of Americans did volunteer work in 2008.

    1. A volunteer gets some money for his/her work.

    2. Americans volunteer to help people in need.

    3. A volunteer is a person who needs help.

16. Chess boxing is a very unusual sport that was invented by a Dutch artist whose name is a Lepe Rubingh. If you’ve never heard of it, the name chess boxing can help you to imagine. It’s a mixture of chess and boxing.

    1. Lepe Rubingh was a sportsman who invented chess boxing.

    2. Chess boxing is a combination of two sports: chess and boxing

    3. The inventor of chess boxing was from Russia.

17. Lying is a part of everyday life. A psychologist asked people to keep a diary of the lies they told over a week. Here is what their diaries revealed:

-There wasn’t a day when the participants didn’t tell at least one lie

- Over the week they decided about 30 percent of people with whom they interacted.

The most common lies that people tell are those in which they pretend to like something to avoid hurting or those in which they make up an excuse.

1. People tell lies to hurt other’s feelings.

2. According to the study carried out, most people don’t tell any lies.

3. Some people tell lies because they were punished after telling the truth.

Appendix B.

Complete the relative clause in each sentence with your own ideas. You have 60 minutes to complete the following sentences

1. –Today I invited David, who is my best friend, to watch the sunset.

• Have you ever been here, David?

• Yes, I have.

• Really?

• I came here with my cousin … … … … … … … … … … … … … … … … … ….

  • (2) Amazingly, water was something of which Michael was afraid when he was a little boy. He couldn’t swim. His mother … … … … … … … … … … … … … … … ….has always motivated him.

  • (3) –What do you think?

• It looks great. You must purchase it.

• But I don’t like it. I’d like to buy a dress … … … … … … … … … … … … … … … … … ….

  • (4) Joe wanted to apply for a job. Could he find his favorite job? Could he find his favorite job?

-Yes, he finally decided to go and work in London, … … … … … … … … … … … … … … …

• (5) I met Mr. Smith, … … … … … … … … … … … … … … … … … … … … … … … … … ….He offered me a job in his school as soon as he found that I was in the teaching business too. He struck me as being a very nice man.

• (6) Vanessa Mae, … … … … … … … … … … … … … … … … … … … … … … … … … … … …

had to make a big decision. She was equally gifted at both the violin and the piano, but she had to concentrate on just one instrument. Although she had just won a prize at a famous piano competition, Vanessa Mae chose the violin.

• (7) Albert Einstein, … … … … … … … … … … … … … … … … … … … … … … … … … … …., had trouble in elementary school. He did poorly in certain subjects such as history and languages. Einstein is best known for his general theory of relativity.

• (8) Mary, my close friend, was unemployed for about one month. Now, she is hired by Mr. Brown, … … … … … … … … … … … … … … … … … … … … … … … … … … … … …

• (9) What is Nick like?

- He has a high self-confidence. He imagined himself to be an artist, … … … … … … … ….

• (10) Clerk: Can I help you?

Customer: I’m looking for a pair of shoes.

Clerk: You will find them on the second floor.

Customer: Oh! Unfortunately, most of them were made of plastic … … … … … … … ….

• (11) Michael has been employed as a computer software engineer for a couple of months. He works for a large company. He hates his boss … … … … … … … … … … … … … ….

• (12) On the plane, I worried about being all alone in London. But when I met Mr. Flannery, … … … … … … … … … … … … … … … … … … … … … … … … …., his warm welcome made me feel at ease. He was very kind and treated me like his own daughter.

• (13) The Olympic Games are an important international event featuring summer and winter sports. Olympic Games are held every four years. Tennis is possibly the least popular sport on the Olympics, … … … … … … … … … … … … … … … … … … … … … … … … …

• (14) In Istanbul, Yesim is getting ready for class. Her mother brings breakfast, … … … … … … … … … … … … … … … … … … … … … … … … … … … … … …, and is looking at her e-mail on her phone. She downloads a chapter from her online textbook to read on the train.

• (15) After university, Kyle Berner tried many different jobs. He tried producing music, selling computers, and even selling hot dogs on the street. Kyle traveled to a small village, and work as an English teacher and teaches the local people … … … … … … …

• (16) The school where John works is close to his apartment. John studied French at University. He teaches English to the students … … … … … … … … … … … … … … … … … … … … … … … … ….

• (17) Brian went to Paris where he has some old friends that he has not seen for about three years. The first thing he did was to visit the associate of the businessman … … … … … … … … … … … … … … … … … … … … … … … … … … … … … … … ….

• (18) There was a murder at a hotel recently. A man … … … … … … … … … … … … ….asked the receptionist to phone the authorities.

• (19) Most people feel at least a little bit uncomfortable when they lie. Work is one place where people sometimes lie to avoid getting in trouble. A liar is someone … … … … … … … … … … … … … … … … … … … … … … … … … … … … … …

• (20) I’m fortunate that I had the opportunity to receive a good education. I know that there are lots of kids … … … … … … … … … … … … … … … … … … … … … … … … … … I’d like to give them a chance to learn.