The subjects were 84 EFL Chinese students who enrolled in the English reading and listening courses at National Sun Yat-sen Universty (NSYSU). Among them, 54 were male and 30 were female, who were around 18-22 years old. They were all non-English majors, whose English proficiency levels were similar because the enrollment was according to their proficiency levels. The reason why the adult learners were chosen as the research participants is that relative clause is a complex construction that generally appears late in learners’ language development.
Therefore, the university students were selected because they were supposed to have learned the target structures, which are included in senior high school English textbooks.
3.3 Research design
In general, there are two kinds of design researchers use in second language acquisition research: longitudinal, and cross-sectional (Gass & Selinker, 2001), in
which longitudinal approach can also be termed qualitative design, and
cross-sectional approach quantitative design. Longitudinal approach is one that usually involves a case study in a long period of time. Spontaneous speech is collected and qualitative data analysis is used. In contrast, a cross-sectional design usually involves a large number of participants and data are collected at a single point of time. In addition, qua ntitative data analysis is employed. Both approaches have its weakness and strength. In the second language acquisition research, it is
recognized that naturalistic data which collected over a long period of time can better reflect learners’ interlanguage than those which are elicited by controlled
instrumentations (Larsen-Freeman and Long, 1994). However, there are drawbacks that will occur if solely relying on learners’ spontaneous production (Larsen-Freeman and Long, 1994). First, without the imposition of constrains on the range of possible responses a subject is likely to produce, the responses may be too random and too extensive to analyze. Second, as mentioned above, the subjects may avoid
producing certain structures or vocabulary that they feel uncertain or troublesome, but which may be just those that the researcher focuses in study.
In view of the above and in accordance with the previous research (e.g., Flanigan, 1995; Gass, 1979, 1980; Izumi, 2003; Eckman, Bell, and Nelson, 1988; Hamilton, 1994), it is felt that a quantitative design is more feasible for the present study, especially due to the time pressure; however the qualitative analysis of the data is included. In this study, a questionnaire was designed, in which three kinds of controlled elicitation tasks were used to elicit the data. Table 3.1 presents the characteristics and advantages and disadvantages of cross-sectional design.
Table 3.1 Advantages and disadvantages of cross-sectional design
research design Characteristics Advantages Disadvantages Cross-sectional
3.4 The relative clause type tested
In the review of the literature (refer to Table 2.10 in Chapter two), it appears that most of the researchers focused on the six types of relative clauses as appeared on the NPAH, namely SU, DO, OPREP, IO, GEN, and OCOMP. And when considering the matrix positioning of the head NP, the four types of relative clauses are usually tested: OS, OO, SS, and SO. As we have mentioned, this obvious ly is a limitation that needs to be addressed (Izumi, 2003). Therefore, in response to this research need, 12 types of relative clauses are tested in this study, as illustrated in Table 3.2.
A particular note is made to genitive. In view of the fact that there is a separate hierarchy for genitive, (Jones, 1991, in Ellis, 2003), as presented in Table 3.3, it would be optimal to include all of the six types of genitive if a complete picture of learner interlanguage of relative clauses is to be obtained. However, the practical consideration makes it less than possible to include all the functions of genitive,
especially in view of the time limit of the present study. As a result, the type of genitive that is tested in the present research is restricted to the genitive that functions a subject, as shown in the sentence in bold in Table 3.3.
Table 3.2 12 types of relative clauses in the three tasks in this study Matrix position RC type Example
Subject SU The boy who is standing at the gate is my brother.
DO The author who he mentioned is well known.
IO The woman who Bill passed a note to is a nurse.
OPREP The candidate who I vote for didn’t win the election.
GEN The man whose wallet was stolen called the police.
OCOMP The person who John is taller than is Charles.
Object SU Jerry likes the teacher who explained the answers to the class.
DO Amy grabbed the letter whic h Jason handed to Julie.
IO The teacher looked at the girl who I explained the sentence to.
OPREP I saw the woman who I went to elementary school with.
GEN I know the man whose bicycle is new.
OCOMP I know the hotel Hilton is cheaper than.
Table 3.3 The Accessibility Hierarchy for –Genitive and +Genitive (Adapted from Jones, 1991, in Ellis, 2003)
Function –Genitive +Genitive
SU
The man whose wife I looked at…
The man whose wife I am bigger than…
3.5 Measuring instruments
It is acknowledged that in order to adequately describe second language learners’
interlanguage, both the actual performance of the learners and their intuitions about the target language must be taken into consideration (Schachter, Tyson, and Diffley, 1976). There are two types of data that are in question here: performance data, based on the actual linguistic production by the learners; intuitional data, based on learner reaction to already produced sentences, such as judgments of grammaticality and ungrammaticality. In this study, there are three kinds of elicitation tasks that are used: sentence combination, English-Chinese translation, and grammaticality
judgment, in which the sentence combination and translation were meant to elicit learners’ performance data whereas the grammaticality judgment generated learners’
intuitional data. The test items that were used in these tasks are adapted from Azar (1999), Celce-Murcia and Larsen-Freeman (1999) and Chen, C. S. (2004).
3.5.1 Sentence combination task
Sentence combination is a typical type of elicitation researchers use in eliciting relative clauses. Quite a number of previous studies adopted this test to collect data concerning relative clauses (see Flanigan, 1995; Gass, 1979, 1980; Izumi, 2003; Wei, 1997; Eckman, Bell, and Nelson, 1988; Hamilton, 1994).
This section contains 12 items in which each item taps into each types of relative clauses that are tested in this study. For example, in (1) the subjects were told to embed sentence (B) into sentence (A) and supply an appropriate relative pronoun, such as who, which , whose, that, so as to produce the relative clause types of OS, Jerry likes the teacher who explained the answers to the class. The distribution of each type of relative clause is at random.
(1)
(A) Jenny likes the teacher.
(B) He explained the answers to the class.
?
It is expected that evidence for the research questions 1-5 is to be gained in this test. In particular, while Gass (1980) used this task to elicit data for the evidence of learner avoidance of relative clause formation, the productive data in this study will be analyzed in this respect.
3.5.2 Chinese-English translation task
Elicited translation is a good way of exploring learners’ interlanguage, where
researcher is able to see how learners transform native language to target language and how well they grasp the rules of English (Chen, 2004). It is believed that subjects’ performance approximates natural speech productio n as they are engaged in the decoding of the stimulus sentence and the encoding of the translation
(Larsen-Freeman and Long, 1994).
In this task, there are two subsections, the first one asking learners to directly translate Chinese relative clauses to English ones and then supply appropriate relative pronouns, as in (2), in which the SO type of English relative clauses is expected: “The book which I read is interesting”. This first subsection contains ten items. As with the sentence combination test, the items for each type of relative clause were
randomly distributed, with the exception of OCOMP, which was placed in the subsection two.
(2)
我讀過的這一本書很有趣。
?
The second subsection taps into object of comparative. Due to the fact that Chinese does not allow relativization on the object of comparative, the sentence combination was designed in order that subjects were to combine the Chinese
sentence before translating it to English. Item (3) was designed to elicit the English counterpart, “The student who Bill is taller than came”.
(3)
(A) 那個學生來了。
(B) Bill 比他高。
?
As with sentence combination task, evidence is anticipated to be obtained for research questions 1-5 in translation task. Since this test involves learners
translating Chinese into English, any influence from Chinese is of special interest in this section, in addition to the difficulty order of the 12 types of relative clauses.
Three sub-questions are the focus of the task. First, because Chinese does not allow relativization of object of comparative, a possibility exists that lack of use of the comparative may occurs, or that strategies of avoidance may be used. Second, Chinese relative clauses have resumptive pronouns retained in direct object, indirect object, and object of preposition, so whether there are higher possibilities that learners might supply resumptive pronouns in the English counterparts than the other two tasks is also an interesting question. Last, it is speculated this test might receive lower accuracy rate than the other two tests due to the interference of Chinese.
3.5.3 Grammaticality judgment task
Learners’ intuitional data were long to be stressed as important when an adequate account of learners’ interlanguage is to be elicited, and grammaticality judgment is a valid means of obtaining the necessary intuitional data (Schachter, Tyson, and Diffley, 1976). The idea that acquisition is evident in what learners know intuitively is also supported by Ellis (2005). Many studies incorporates grammaticality judgment task, which signifies that grammaticality judgment is a standard tool in second language acquisition research (Gass, 1979, 1980; Izumi, 2003; Ioup and Kruse, 1977).
Grammaticality judgment in this study also compensates what sentence combination and English-Chinese translation cannot do to reveal what is lacking in learners’
interlanguage (Gass & Selinker, 2001). For example, learners may correctly produce object of preposition type of relative clause, such as, I looked for the book whcih Tom was talking about. But they may not be able to correctly identify the error in the sentence, *I looked for the book who Tom was talking about, where the relative pronoun who should be which instead. In addition to the need of identifying what is not correct, the present study also required the subjects to correct the errors.
There are 29 items in this task in which 18 are grammatically correct and 11 are grammatically incorrect, which are randomly distributed. Three categories of errors were used in formulating the incorrect items, namely relative marker omission, pronoun retention, and incorrect use of relative marker. Table 4 illustrates the grammatically incorrect sentences.
Table 3.4 Three categories of errors in the grammaticality judgment task
Categories Example sentences
1. Relative marker omission * Almost all of the people appear on television wear makeup.
2. Pronoun retention * The car that the man drove it was very fast.
3. Incorrect use of relative marker
*Bob admires the professor which John lives next to.
3.6 Procedures
As a pilot study, the questionnaires incorporating the three tests were first given to 45 English- major student s who were at their second year in National Sun Yat-sen University (NSYSU). There were three sections in the questionnaire: (1)
background information, (2) test questions which were categorized into the three test types, and (3) a difficulty ordering task which required subjects to make subjective judgment on 12 types of relative clauses and rearrange them in descent order.
According to pilot study, amendments were made to eliminate the inappropriate items.
Furthermore, the test items were made to better tap into the problems that were meant to be explored in this study. The revised questionnaires were then given to 120 non-English major students in NSYSU. Those that left too many items unanswered were excluded from the study. Finally, data from 84 questionnaires were collected and analyzed.
This study is intended to inspect the learners’ underlying knowledge of relative clauses. For some researchers (e.g., R. Ellis, 2003, 2005), it is termed implicit knowledge (intuitive and unconscious) as opposed to explicit knowledge (metalingual and conscious). In his 2005 study, Ellis confirmed that while time-pressured tests will require learners to rely on their implicit knowledge, tests without time constraints will permit learners to draw on their explicit knowledge. Therefore, 30-minute time limit was imposed on the completion of the three tests to ensure that subjects were not to have spare time to look back and made use of explicit rather than implicit
knowledge that is the focal point of the study.
As in Daughty’s (1991) and Izumi’s (2003) studies, participants were told that all words were spelled correctly and no errors were involved regarding to tense or aspect.
In addition, lexically, the sentences in the test were kept as simple as possible so as to preclude the possibility of judgments being made on the basis of something other than that which was specifically being tested. What is more, they were told that Chinese was allowed if they happened to forget the spelling.
3.7 Scoring
The scoring of the data in the present study largely follows the method used by Izumi (2003). For the Sentence combination and translation testes, one point was given for correct production of the form and zero point for incorrect production of the form, errors involving articles, tense, and spelling were ignored, and omission of the relative pronoun in the DO and OPREP is permissible, so it was deemed correct. In the grammaticality judgment test, any items not judged or judged correctly but
without corrections were given zero point. For the resumptive pronoun analysis, one point will be given to whoever retained the pronoun and zero point for those who did not.
Recall that avoidance strategies are behaviors that the subjects use when they deliberately and consciously choose not to use the target form, possibly due to the difficulty or the partial mastery of the structure (Kleinmann, 1977). The subjects of this study are university students who have learned the target structures after
graduation from senior high school. Therefore, the deviant form they produced should be the result of the structure complexity or lack of full knowledge, rather than entire lack of knowledge of the target structure. In the present study, other than the deviant form, the total lack of use of the target form will be included in the data for examination (see Schachter, 1974). Points will be assigned according to the type of strategy that was used by the subjects. That is, one point for each type of the strategy.
Five types of avoidance strategies are included in this analysis, in which type 1~4 are those that were generalized by Gass (1980) and type 5 is added for the present study, as in Table 3.5. As we can see in Table 3.5, type 1 mainly involves avoidance
in the OCOMP relativization, where the object of comparative, the syntactically more complex structure, is to be circumvented by a lexical change, and finally the subject function, the syntactically less difficult one is produced. Type 3 is similar to type 1 in that it results in changing the more complex construction (OCOMP relatives) to the less complex one (SU relatives), though by different means. Type 4 involves
avoidance from DO (or IO, OPREP, and GEN) relatives to SU relatives. Type 2 and 5 are alike, both related to avoidance of center embeddedness, an issue that involves processing of matrix sentence. Note that when the head NP functions as object, type 2 is related to the avoidance from the more difficult relativization positions to the less difficult ones, as the example sentence below shows. In fact, as will be shown in Chapter Four, quite a number of learners use this strategy in object matrix positioning, though it did not outnumber the case in which the learners used type 2 to avoid
center-embedding. Besides, type 5 occurred only in the case of subject matrix positioning, when there was a need for learners to avoid center embedding. In fact, there was no such case that the subjects in this study used type 5 in object matrix positioning sentences.
【Type 2】
I know the woman.
Bill passed a note to her.
The correct combination: I know the woman who Bill passed a note to.
Avoidance: Bill passed a note to the woman who I know.
Table 3.5 The categorized avoidance strategies (adapted from Gass, 1980) Type of avoidance Example sentence
1. Substitution of one lexical item for another.
(Avoid OCOMP relatives)
I know the man.
Joseph is thinner than him.
→I know the man who is fatter than Joseph.
2. Switching the order of the two sentences so as to embed the sentence which
→Bill passed a note to the woman who is a nurse.
3. Changing the identical NP.
(Avoid OCOMP relatives)
He saw the wo man.
The man is older than the woman.
→He saw the man who is older than the woman.
The man kissed the woman.
? He saw the woman who was kissed by the man.
5. Changing the head NP (Avoid center embedding)
The man called the police.
His wallet was stolen.
? The man called the police whose wallet was stolen.
Note that only the data in the sentence combination test and translation test are submitted for analysis. Specifically, two questions are expected to be answered in this test, as follows:
1. To report the avoidance strategies that were used.
2. To see the tendency of the avoidance: that is whether the learners tend to avo id relativization on positions that are low on the NAPH?
CHAPTER FOUR RESULTS
4.1 The accuracy order of the relative clauses
4.1.1 Sentence Combination test
Table 4.1 presents the total correct responses and the percentage (in brackets) thereof for the six types of relative clauses, which are placed in the two different matrix positions. The rightmost column shows the total correct responses for matrix subject position versus matrix object position, and the bottom row of the Table
demonstrates the total results of each type of relative clauses without distinguishing between the two types of matrix positions. As can be seen from the bottom row of Table 4.1, the highest scores were obtained for SU, followed by GEN, DO, OPREP, IO, and OCOMP. Except GEN, OPREP and IO, the order generally matches the universal markedness as predicted by NPAH. This is in conformity with the research prediction made earlier in this study. In terms of the matrix positions, the score obtained for the matrix object position is higher than the subject position: 60.2 vs. 49.2. This matches the prediction by PDH, and also the research hypothesis.
Figure 4.1 graphically shows the result. As for the difficulty order obtained for each different type of relative clauses, it is OGEN> OS> OO> SS> SGEN>
OOPREP>SO> OIO>SOPREP> SIO> OOCOMP> SOCOMP (> means gets more accurate responses than). This order is generally as predicted by SOHH, except OGEN, which gets a slightly higher score than OS. To restate, the order for
>SOCOMP. The total mean accuracy rate for this test is 65.1%.
Table 4.1
Total correct responses on sentence combination test by relative clause type and
Total correct responses on sentence combination test by relative clause type and