Study on Chinese College EFL Learners' Comprehension and Production of English Relative Clauses: Testing Three Hypotheses

2.1 Predictor hypotheses of RC Requisition

In previous research, several important hypotheses have been proposed to aecount for the relative ease and difficulty of processing and acquiring different types of RC sentences. They are Parallel Function Hypothesis (PFH), Perceptual Difficulty Hypothesis (PDH), Noun Phrase Accessibility Hierarchy Hypothesis (NPAHH), and so on.

For better understanding of these hypotheses, Table provides some example sentences of different types of relative clauses that will be investigated in this study.

2.2 Literature review

A number of studies have been conducted to test the predictions of different hypotheses of RC acquisition both in L1 and L2. Among them, those in L2 are of our Special interest. Table 2 presents a summary of important empirical studies of RC processing and acquisition in English as L2. The table is enriched on the basis of Izumi's table (Izumi, 2003, pp. 293-294).

Although care needs to be taken in interpreting the table (because of differences in the focus of investigation, data and participants used in different studies), several things can be noted in examining the table. First, it is clear that many studies found support for NPAHH, indicating it is most likely a valid hierarchy for understanding linguistic facts not just about RC, but about other grammatical facts. PDH also finds some support but has received less attention in L2 studies than NPAHH has to date.

None of the studies in Table 2 provide support for PFH. This fact was also attested by Doughty (1991, p. 437). After discussing the relevant studies in both L1 and L2 acquisition, she concluded that the PFH has not been supported by the studies of L2 relativization since it was suggested. In a study of the comprehension of relative clauses by children EFL learners, Sheldon (1974) found that children could better understand sentences in which the HN and the RN served the same function in their respective clauses (e. g., SS and OO relativization). However, these findings have not been replicated by other EFL researchers (Bowerman, 1979). They were not supported by an early study of relativization (Gass & Ard, 1980), and have been directly countered by evidence from studies of adult native Speakers' comprehension of relative clauses (Prideaux & Baker, 1986). Despite Doughty's negative Statements, this study still takes the PFH into consideration in order not to miss a chance to test it further together with other hypotheses.

Second, with regard to the data elicitation measures, Table 2 indicates that most of the studies rely on data from production measures. According to Izumi (2003, p. 294), only one study in SLA used a receptive task to examine the difficulty order of different RC sentence types. Ioup and Kruse (1977) used a grammaticality judgment task and found support for the PDH. Gass's (1979, 1980) studies included multiple testing instruments, but only the information from the sentence combination task was used to examine directly the difficulty order among different RC types. Data from other measures (a grammaticality judgment task and free composition) were used to address the issues of language transfer in SLA. Likewise, Doughty (1988, 1991) used multiple testing measures, but the analyses were performed on the composite scores of all the testing measures combined.

It is important to analyze both the production data and the comprehension data, because tasks in different modalities may present different degrees and/or types of difficulty in learners' language processing. For example, Ioup (1983) suggested that different elicitation measures may yield different results with regard to support for different hypotheses of RC acquisition. Previous SLA research in other areas also suggests that task type exerts great influence on L2 learners' linguistic Performance (e. g., Bygate, 1999). Therefore, the investigation of the task effect is worthwhile.

In sum, the review of previous literature suggests that more research is needed to test different hypotheses of RC acquisition in SLA by using different types of tasks. Ideally, research should examine these issues with the same participants within the same study so that direct comparisons of the results from different elicitation measures will be possible. The present study aims to do this and examines the effects of different elicitation tasks on the production and comprehension of the English RCs in the context of Chinese-speaking environment. The specific research question is: To what extent are the predictions of different hypotheses of RC acquisition supported by the results of different elicitation tasks performed by Chinese College EFL learners?

3.1 Participants

The participants in this study were randomly selected from the first-year classes of a university in Shanxi Province. They had learned English mainly in the classroom from teachers' instruction, and had just been enrolled as College students after the national College entrance examination.

3.2 Instruments

In order to test the predictions of the three hypotheses of RC acquisition from multiple angles, the participants' knowledge of RC was examined using a sentence combination test and a grammaticality judgment test. The first test was used to examine the participants' productive abilities, and the second was to examine their comprehension. The study focused on eight RC types that are formed by relating the HN (subject or object) to the RP (subject, direct object, indirect object, object of a preposition). The inclusion of indirect object contrasted with Izumi (2003), who only considered the other three RPs. For each type of RC included, half of the testing items were embedded in the subject position and half in the object position in the matrix sentence. The following are details of each of these tests:

Sentence combination test. This test was adapted from Eckman, Bell and Nelson (1988) and Doughty (1991). In this test, the participants were directed to embed sentence B into sentence A, as in (5):

1. A. The film was directed by David Lean.

2. B. Mary said that she had seen it in Venice.

3. Expected response: The film which Mary said that she had seen in Venice was directed by David Lean.

The participants were prohibited from using any coordinating conjoiners such as and, because, since, so, and when. The test included 16 items, with 2 items for each of the eight RC types.

Grammaticality judgment test. This test was constructed for this study on the theoretical basis of Doughty (1991) and Izumi (2003). Some of the testing items were borrowed from grammar books and relevant studies. It consisted of 32 testing sentences, with 4 items for each RC type, 2 of them correct and 2 incorrect. In Gass (1982), four error types were identified, including pronoun retention, non-adjacency, incorrect relative marker morphology, and inappropriate relative marker omission. Besides, a newly found error type, inappropriate omission of the HN, was also included. An example of each type of error is given in (6a-6e):

1. The man who you met him went to the hospital. [pronoun retention]

2. The woman is young who likes John, [non-adjacency]

3. They forgot many words what they had studied. [incorrect RC marker morphology]

4. The man saw the girl is my brother. [inappropriate omission of RC marker]

5. People have enough money to buy which will be fine for their health. (inappropriate omission of the head noun).

For each sentence that the participants deemed to be incorrect, they were asked to provide relevant corrections.

3.3 Data elicitation procedures

In September 2012, the testing papers including the three elicitation measures were handed out to the students in the classroom. They were asked to finish the papers as required. Altogether, 114 papers were collected after 100 minutes. Subsequently, 30 papers were randomly selected forming the database for the follow-up analysis.

3.4 Data analysis

The first step in data analysis is scoring. The tests were scored separately for different types of RC and also for the two different positions of the RC in the matrix sentence. In some cases, elimination of a small number of test items was necessary because of the participants' failure to complete all the test items in time or to follow the procedure specified in the test (e. g., not providing corrections in the grammaticality judgment test), which might result in an unequal number of total items included for each participant. The scores for each participant were calculated by dividing the total correct scores by the total number of applicable items. Therefore, each participant received a percentage score for each test.

The production test was scored by assigning 1 point for each correct use of the form and 0 point for incorrect use. When the participant produced, for example, the S type relative while the DO type was required in the item, the sentence was scored as incorrect. Errors involving articles, tense, and spelling were ignored, as long as they did not pertain to the formation of the RC. Because omission of the relative pronoun is permissible in the DO and OPREP types, it was considered correct as long as all else relevant to RC formation was correct.

Because of the potential problems the scoring of a grammaticality judgment test can pose, such as uncertainty as to why a participant judged a sentence to be ungrammatical, care was taken in scoring the test to check first whether the best item was indeed judged on the basis of the targeted form. This was done by examining the participants' corrections of errors. If the correction for a particular item indicated that the participant was judging the sentence as ungrammatical for reasons other than the RC-related problems, that item was regarded as non-applicable and was eliminated from further analysis. RC-related corrections typically involved movement of the RC to an adjacent position of the HN if it was not already there, change of the word order inside the RC, change of the RP, insertion of RP when there was none (this was applicable only for the S clause type), and insertion or deletion of pronoun copies. Common non-RC-related corrections were change of tense or aspect in the matrix clause, change of articles, and change, insertion, or deletion of prepositions in the matrix clause. Any items not judged, or judged incorrect but without corrections, were excluded from the analysis. After determining the applicable items, these items were scored by assigning 1 point for correct judgment and 0 point for incorrect judgment.

4.1 Results

Table 3 presents the mean accuracy scores of different tests for the four types of RCs placed in the two different matrix positions.

4.1.2 Grammaticality judgment test

The results of the accuracy scores for each RC sentence type on the grammaticality judgment test are displayed in Figure 2.

Figure 2.

Mean accuracy scores of grammaticality judgment test

As PFH predicts, the acquisition of RC is easier when the grammatical function of the HN is identical with that of the RP, while it becomes more difficult when the function of the HN is not parallel with its RP. Figure 2 shows that the scores of S-S is higher than O-S, and the acquisition of RCs like O-DO, O-IO, and O-OPREP is easier than that of RCs as S-DO, S-IO, and S-OPREP. Therefore, generally speaking the result is consistent with the PFH, with the sole exception of S-OPREP > O-IO. Second, O-S has a lower score than S-S and S-OPREP, which contradicts the prediction of the PDH, that is, O-S & O-DO & O-IO & O-OPREP > S-S & S-DO & S-O & S-OPREP. Besides, in either case of the HN as the subject or the object, the highest score of RC type is not S-S, indicating that the NPAHH also fails to be proved by the grammaticality judgment test.

Table 4 shows that in the context of a Chinese-speaking environment, the results of the production test of English RC like sentence combination contradict PFH, but partly support PDH and NPAHH with the exception of S-S type; the comprehension test such as grammaticality judgment Supports PFH generally but contradicts PDH and NPAHH.

Table 4.

The summary of the results

Test	PFH	PDH	NPAHH
Sentence combination	contradict	partly true	generally true
Grammaticality judgment	generally true	contradict	contradict

4.2 Discussion

The findings presented in the last section can be explained in relation to the participants. All participants in the study had just been enrolled in the university and learned English mainly in high school classroom settings, where grammatical rules were especially emphasized. Whenever the learners comprehended and produced a linguistic item, they would naturally resort to certain grammatical rules as guidelines. Their Performance on certain linguistic items largely depends on their understanding and practice. With regard to the acquisition of RCs, it seems that these learners' comprehension and production are more related to the relevant instructions they received than to certain universals. Another possible reason is that the participants had already mastered grammatical rules of RCs on the whole, though unsatisfactorily on some types due to the significant differences between Chinese and English relative clauses, and the negative transfer from their mother tongue.

As for the data elicitation measures, it is noted from Tables 3 and Figures 1 and 2 that the participants performed better in the sentence combination test than in the grammaticality judgment test. This Variation across different elicitation measures suggests that the students who participated in the study had already learned to understand all types of RCs, but still could not identify some subtle errors especially when they interfered with some distracting correct usages. Therefore, teachers and learners at university level should not be satisfied in understanding the general meanings of the RCs encountered in reading, but should strive to be precise in using the RCs.

Figure 1.

Mean accuracy scores of sentence combination tes