This chapter expounds the experimental design of the present study, which aims

Chapter Three Experimental Design

This chapter expounds the experimental design of the present study, which aims

to investigate how Taiwanese EFL learners use English RCs, in terms of its research

questions, subjects, and methodology.

3.1 Research Questions

Before formulating the research questions of the study, it is necessary to first

demarcate the scope of RCs under investigation, given a great variety of RCs in

English. The present study considers only those prototypical, fully-fledged RCs,

with the exclusion of such variant RC types as reduced RCs (i.e. prepositional,

participial, or adjectival phrases), free (headless or fused) RCs, cleft RCs, or

non-finite RCs (i.e. infinitive to phrases). More specifically, it focuses on the

following types of fully-fledged RCs: subject, direct object, indirect object,

prepositional object, and possessive determiner RCs, while excluding predicative

complement and adverb RCs. This narrowing of research focus is justified by

practical constraints of eliciting data in a classroom setting. ¹

Essentially, this study is intended as an extension from previous empirical studies

on English RC acquisition, most of which center on the impact of either L1

1

Having said this, it is also acknowledged that focusing on these five RC types only and excluding

other possible RC types is indeed a limitation of the present study, and it is hoped that this limitation

will be addressed in future RC studies.

interference or universal factors on L2 learners’ RC performance─namely, whether

their error patterns in RCs reveal the influence of their mother tongue, or whether

their learning difficulties with various RC types can be predicted by such

processing-based hypotheses as the Parallel Function Hypothesis, the Perceptual

Difficulty Hypothesis, and the Noun Phrase Accessibility Hierarchy Hypothesis.

On the one hand, given that previous RC research concentrates predominantly on

restrictive RCs, the present study extends its research scope to investigating how

Taiwanese EFL learners acquire (or use) non-restrictive RCs, with a view to gaining

more insights into what factors may interfere with their mastering this special type of

RCs. To be more precise, the study attempts to investigate (1) whether Taiwanese

EFL learners tend to overuse restrictive RCs; and (2) whether they perform differently

among various NP contexts in their use of non-restrictive RCs. When it comes to the

acquisition of non-restrictive RCs, the tendency to overuse their restrictive

counterparts warrants some attention. This is because the subtle difference between

restrictive and non-restrictive RCs can be so difficult for learners to discern or grasp

that they may fail to make a good distinction between the two in their mental

grammar of English RCs and thus have a propensity to use restrictive RCs, the

prototype of RCs, in most contexts, even where non-restrictive ones should be much

preferred. Besides, since non-restrictive RCs are commonly used with such

referentially accessible NPs as linguistically definite NPs, situationally definite NPs,

narrowly specified NPs, whole-referring NPs, generic NPs, personal pronouns, proper

NPs, and one-of-a-kind NPs, as noted in Chapter 2, it is worthwhile to examine

learners’ performance in employing non-restrictive RCs among these different NP

contexts in order to determine whether they have achieved a full understanding of

when to use non-restrictive RCs.

On the other hand, in view of an undue preoccupation with the structural aspects

of RC acquisition by the majority of previous RC research, the present study concerns

itself primarily with the functional dimensions of RC acquisition by Taiwanese EFL

learners.

The study first of all intends to investigate the extent to which Taiwanese EFL

learners are able to use RCs in various pragmatic/discourse contexts to perform such

functions as identifying, characterizing, presenting, and interpolating. Here, there

are two research foci in particular. One research focus relates to whether Taiwanese

EFL learners are inclined to use independent clauses in presentative or parenthetical

contexts, for both of which the use of RCs is normally preferred. As pointed out in

the review of relevant RC literature (Bley-Vroman & Houng, 1988, cited in

Kamimoto et al., 1992; Zhao, 1989, cited in Kamimoto et al., 1992; Li, 1996), there

are pragmatic differences between English and Chinese in their preferred structure for

presentative or parenthetical contexts. For these two contexts, English prefers RCs,

but Chinese favors independent clauses, rather than RCs, which in Chinese serve no

other functions than being a noun modifier for either identification or characterization.

If not aware of such pragmatic differences, Chinese learners are very likely to have a

tendency to underuse RCs in English by overusing independent clauses for presenting

a thematically important referent or interpolating additional information. The other

research focus relates to whether Taiwanese EFL learners are apt to misuse the

definite article the with RCs which function mainly to characterize NPs as a particular

type, rather than to identify them as a known entity. In a typical L2 classroom,

learners are often taught to always use the to introduce the information that RCs

qualify (Parrott, 2000: 355-356). Useful as it may seem, this rule of thumb is not

water-tight, inasmuch as it fails to help learners recognize and understand the function

of RCs without the—namely, those RCs with indefinite NPs, which serve to define a

particular category or class of people or things (Swales & Feak, 2003: 37-38).

Under the influence of this false instruction, learners are very likely to get into the

habit of using the definite article with most RCs, even in a characterizing context.

The study further seeks to investigate the degree to which Taiwanese EFL

learners are capable of employing RCs in writing as a means of backgrounding

supportive materials. Learners often know very well how to construct RCs, but this

knowledge does not necessarily translate into knowledge of how to use RCs in writing

with respect to discourse grounding (Tomlin, 1994: 167). As such, they may

underuse RCs in their writing by mistakenly “foregrounding” minor information units

in independent clauses, which in fact should be “backgrounded” in such subordinate

clauses as RCs. More specifically, learners generally “have little idea about how to

accommodate different kinds of information in different sentence structures, how to

achieve textual coherence, and how to structure their thoughts in a way to which

native speakers are more accustomed” (Chen, 1999: 139). In the light of this, the

researcher sees a need to explore the issue of how well learners can utilize RCs in

written discourse to background non-essential, amplifying information.

To sum up, the main purpose of the present study is to look into the use of

English RCs by Taiwanese EFL learners with reference to the following three research

questions:

1. How is the acquisition of non-restrictive RCs by Taiwanese EFL learners?

(a) Is there a tendency for them to overuse restrictive RCs?

(b) Are there performance differences among various NP contexts in their use

of non-restrictive RCs?

2. To what extent are Taiwanese EFL learners able to use RCs appropriately in

different pragmatic/discourse contexts, including identifying, characterizing,

presentative, and parenthetical?

(a) Is there a tendency for them to use independent clauses in place of RCs for

presenting a topic or interpolating parenthetical assertions?

(b) Is there a tendency for them to misuse the definite article the with RCs in a

characterizing context?

3. To what extent are Taiwanese EFL learners able to use RCs appropriately in

writing as a useful backgrounding device?

3.2 Subjects

There were 160 subjects participating in the formal testing of the study, who

came from different senior high schools, including National Chia-Yi Girl’s Senior

High School, National Hu-Wei Senior High School, and Taipei Municipal Xisong

Senior High School According to their average years of learning English as a

foreign language, the subjects were divided into two groups: (1) 80 first-graders, who

on average had four years of English-learning experience since the first grade of

junior high school; and (2) 80 third-graders, whose mean length of English instruction

was approximately 6 years. ² In the data-analysis session, with a small number of

samples eliminated, a total of 120 subjects were included, with 60 for each group.

The rationale for selecting subjects from senior high school, as opposed to junior

2

It needs to be pointed out that some subjects in the study may have begun their study of English even

earlier at elementary or cram schools, rather than from the first grade of junior high school. Therefore,

their individual years of learning English could actually exceed an average of 4 or 6 years.

high school, is based on the consideration that for junior high school students, the

acquisition of English RCs is still at its very early stage, as it is not until the third

grade of junior high school that Taiwanese EFL learners have begun to study basic

usage of relative markers, such as who(m), which, that, and whose, in the formation of

RRCs and the difference between RRCs and NRRCs. Hence, in order to make sure

that one’s RC performance in experimental tests would reflect his/her true

competence in English RCs rather than indicate his/her incomplete learning of

English RCs, the present study selected senior high school students as the subjects for

the formal testing.

3.3 Methodology

RC data analyzed for the present study were collected from three kinds of tests:

an RC judgment test, a context translation test, and a passage-rewriting test. Most of

the test materials were taken or adapted from Azar (1999), Frodesen and Eyring

(2000), Riggenbach and Samuda (2000), and Thewlis (2000). To ensure that the

subjects’ RC performance, especially that of the first-graders, would not be affected

by their failure to understand the language of the tests, vocabulary and sentence

structure were kept at a level comprehensible to them, and difficult words or phrases

annotated with Chinese definitions in parentheses. The three tests ³ are discussed in

3

To ensure the internal validity of the present study, a pilot study was conducted with 20 senior high

school students (10 first-graders and 10 third-graders) prior to the formal testing, and based on the

results, subsequent revisions were made to the three tests.

greater detail in Sections 3.3.1 and 3.3.2, regarding their rationale, content, and

administration.

3.3.1 Instruments

The RC judgment test

The RC judgment test (see Appendix A) was used to tap into the subjects’

acquisition of English NRRCs. It required the subjects to examine English sentences

containing RCs and then make judgments as to whether these RCs should be

restrictive or non-restrictive. To preclude them from randomly guessing without

answering in reference to the degree of referential accessibility of RC heads in context,

the subjects were asked to give a brief explanation (in Chinese) in the space provided

as regards why they considered a given RC non-restrictive. This brief explanation

for a non-restrictive judgment would further help the research determine whether the

subjects were acutely aware of the close relationship between referential accessibility

and NRRCs.

The RC judgment test was composed of 18 questions with 19 RCs, of which 7

were restrictive and 12, non-restrictive. The antecedents of the 12 NRRCs included

six common types of referentially accessible NPs ⁴ , as categorized in Chapter 2,

namely, NPs whose domain of reference rarely needs any further restriction for either

4

The test focused on NRRCs whose antecedents were NPs; thus, those NRRCs with sentential heads

were not considered here.

identification or characterization: (1) NPs which are definite by virtue of the prior

discourse; (2) NPs which are definite by virtue of the speech situation; (3) NPs which

are narrowly specified with extra modifications; (4) NPs which refer to the entire

class, rather than part of it; (5) NPs which refer to a particular class or category of

people or things; (6) personal pronouns, proper NPs, and one-of-a-kind NPs, all of

which are referentially identifiable due to their lexical connotations of “uniqueness”

or “being the only one.” The following are the 12 NRRCs in the RC judgment test,

introduced in terms of the type of referentially accessible NPs to which their

antecedents pertain:

(65) Linguistically definite NPs

Question 2: According to the news, the police arrested the man who had

robbed the First National Bank. The man, who was wearing a

plaid shirt and blue jeans, was caught shortly after he had left

the bank

(Cf. the police arrested the man who had robbed the First National Bank.)

(66) Situationally definite NPs

Question 15: This book, which I finished reading yesterday, has many

interesting stories.

(Cf. Question 3: The magazine which you just talked about has photos of

famous stars.)

(67) Narrowly specified NPs

Question 13: She just bought a beautifully-tailored, pink silk skirt, which I

have always wanted to have.

(Cf. Question 7: Susan is wearing a dress which I have never seen before.)

Question 18: Their expensive, new car, which broke down last night, is

being fixed in a garage today.

(68) Whole-referring NPs

Question 5: I taught English to some children today. To my surprise, all the

children, who were 9 years old, were really quiet and attentive

during the class.

(Cf. Question 16: We took some children to a playground for different

games. As soon as we arrived, children who wanted to

play soccer ran to an open field and the others just

stayed nearby to play basketball.)

(69) Generic NPs

Question 9: Humans, who need water to survive, may someday run out of

pure water.

(Cf. Question 12: People who drink should not drive, because it is very

dangerous to do so.)

Question 11: An elephant, which is the world’s largest land animal, has few

natural enemies other than human beings.

(70) Personal pronouns

Question 4: I, who am a student at this school, come from a country in Asia.

(Cf. Question 14: Anybody whom you give a ticket to will be welcome in my

music concert.)

Proper NPs

Question 8: Jean, whose English is very good, applied for a job as a

secretary.

(Cf. Question 10: Women who are able to type 50 words in a minute can

apply for this job.)

Question 1: PS2, which stands for Play Station 2, has become quite popular

with both adults and youngsters in Taiwan.

One-of-a-kind NPs

Question 6: Mary looked up at the moon, which was very bright that

evening.

Question 17: Jos’s father, who exercises every day, is still healthy and

energetic.

The context translation test

The context translation test (see Appendix B) was employed to scrutinize the

subjects’ use of English RCs in different pragmatic/discourse contexts. In the test,

the subjects were to translate Chinese sentences into appropriate English according to

contexts given. They were told that instead of strictly adhering to exact Chinese

wording or sentence patterns, they were free to write down whatever they thought

should be appropriate English counterparts for these Chinese sentences in particular

contexts. To make this test less demanding, difficult English words and phrases

were initially given for the subjects to readily incorporate into their English

translation.

The context translation test consisted of 10 questions. Among them, questions

6 and 8 were distracters, namely, items designed to lower the subjects’ awareness of

the target structure in the test by eliciting non-relative constructions, as shown in (71):

(71) Question 6: He is an English teacher of junior high school.

Question 8: She caught a cold today.

The rest of the questions involved Chinese sentences constructed in such a context

that their appropriate English equivalents would preferably involve the use of RCs to

perform such functions as identifying, characterizing, presenting, and interpolating.

These RC-targeting questions are introduced as follows:

(72) RCs serving to identify NPs as a known entity

Question 4: The one who is talking with Mary is my father.

Question 7: I went to see the movie which you introduced to me yesterday.

(73) RCs serving to characterize NPs as a particular type

Question 5: He is looking for a girl who shares the same interests.

Question 9: It is a device which can save computer files.

Question 10: People who speak loudly in public places are very annoying.

(74) RCs serving to present an important, new topic into the discourse

Question 1a: There was a king who loved his daughter very much.

Question 1b: A beggar came who wore very strange clothes.

(75) RCs serving to interpolate parenthetical assertions

Question 2: Today I ran into Mr. Smith, who used to live next door to me.

Question 3: My favorite animal is the elephant, which is considered the

biggest land animal in the world.

It needs to be pointed out that this experimental test essentially differs from the

traditional Chinese-to-English translation due to its provision of contextual

information. For example, question 1 supplies one with a presentative context,

wherein the Chinese sentence to be translated opens up a discourse unit to introduce a

new topic; questions 2 and 3, a parenthetical context, wherein the Chinese sentence to

be translated acts as a thrown-in remark or afterthought. With reference to

contextual information, the subjects could concentrate more on finding the

appropriate structural counterpart in English for a specific context, and less on giving

a literal translation from Chinese into English. Moreover, as such contextual

information could serve as clues that motivated the subjects to use the target structure,

the need was obviated to employ “Europeanized” Chinese items, namely, those

awkward Chinese sentences whose syntactic patterns are “assimilated” into English

for ease of eliciting the target structure in the translation (Wu, 2005) (e.g. deliberately

changing Chinese sentences from independent clauses to RCs in order to prompt the

subjects to use RCs accordingly in their English translation).

The passage-rewriting test

The passage-rewriting test (see Appendix C) was to examine the subjects’ ability

to make good use of RCs in writing in relation to discourse grounding. In the test,

the subjects were directed to study a short passage containing several pairs of

sentences and then rewrite it in a better way without leaving out any of the

information; basically, they were to combine each pair of sentences into one sentence

by using appropriate conjunctions. Different from the traditional mechanical

sentence-combining, this rewriting task allowed the subjects more freedom as to how

they would like to combine sentences, and foremost, provided context for them to

choose proper conjunctions by taking into account such discourse factors as

propositional arrangement and textual coherence (Kameen, 1978). ⁵

The passage to be rewritten in the test was adapted from text (76) below by

changing all its RCs into independent clauses:

(76) It’s interesting to talk with women who have had similar experiences.

Balance, for example, is a topic to talk about. Achieving balance may be a challenge for those women who have jobs and family responsibilities.

Some women don’t have jobs outside their homes. They sometimes feel criticized by other people who think that all women should have careers.

This is an attitude that more and more people share toward women today.

Some women work at jobs and have young children. They also feel criticized by other people who think that all women should stay at home with their children. This is a belief that makes some women feel a lack of balance in their lives. Some women never have children. They may feel pressure from their own parents, who worry that they will never have grandchildren. The above are three examples that show how it can be difficult for women to achieve balance between jobs and family responsibilities.

The idea of designing such a rewriting task to test grounding ability with RCs was

actually inspired by Bardovi-Harlig (1990), who suggests that a text can slightly be

modified, for instance, by changing all subordinate clauses into independent clauses,

and used for learners to identify what he calls “pragmatic mismatches” (pp. 58-59)

and rewrite these mismatches for a better fit in terms of discourse grounding. By

analyzing the subjects’ clause-combining strategies observed in their revised text, the

researcher could probe into their communicative competence in utilizing appropriate

5

Kameen (1978) suggests three types of sentence-combining tasks according to the degree to which

these tasks impose control on and give context to L2 learners in combining sentences: mechanical,

meaningful, and communicative.

sentence structures to package different idea units with reference to their contribution

to the development of the discourse—that is, independent clauses for marking

foreground information, and subordinate clauses (here RCs, to be exact), background

information.

3.3.2 Test Administration

The present study is by its nature a blindfolded experiment inasmuch as its

subjects were not informed that the tests they took would be analyzed for research

purposes. Prior to the formal testing, they were only told that in the next two periods,

there would be three tests to help them review what they had learned before (but they

did not know that the target structure in these tests was English RCs); the testing was

conducted during their regular English classes by their own school teachers (rather

than by the researcher himself), who were beforehand given a copy of test

administration guidelines (see Appendix D). The purpose of blindfolding the

subjects was to increase the reliability and validity of the study, considering that if

they had been conscious of the research nature of the tests, they might have taken the

tests in a perfunctory and reluctant manner, instead of doing their best to answer them

(since how well they performed in these tests had no bearing at all on their school

“grades”), and their true competence in English RCs would not have been observed.

The three experimental tests were administered in two separate sessions. In the

first session, the subjects were initially given 30 minutes to complete the

passage-rewriting test. After their passage-rewriting test sheets were collected, they

were given another 20 minutes to take the context translation test independently in

order to avoid the possible influence of the previous test.

In the second session, the subjects were given 30 minutes to complete the RC

judgment test. The purpose of conducting this RC judgment test last was to prevent

the subjects from being aware that it was English RCs that were being targeted in the

passage-rewriting and context translation tests. If the order of test administration

had been reversed, i.e. administering the RC judgment test at the very beginning, then

by the time for the passage-rewriting and context translation tests, the subjects might

have become cognizant of the target structure, thus deliberately using more RCs in

these two tests, and such conscious over-production of RCs would have detracted

from the internal validity of the study.