• 沒有找到結果。

Interplay between Forward and Backward Transfer in L2 and L1 Writing: The Case of Chinese ESL Learners in the US

N/A
N/A
Protected

Academic year: 2021

Share "Interplay between Forward and Backward Transfer in L2 and L1 Writing: The Case of Chinese ESL Learners in the US"

Copied!
50
0
0

加載中.... (立即查看全文)

全文

(1)

Interplay between Forward and Backward Transfer in L2 and

L1 Writing: The Case of Chinese ESL Learners in the US

1

Fred Jyun-gwang Chen

National Taiwan Normal University

This paper focuses on the issue of language transfer in an L2 environment. Research has shown that forward transfer from L1 to L2 appears at early stages and decreases as L2 proficiency increases. Additionally, there is evidence to suggest that backward transfer from L2 to L1 may occur in an L2 environment in which subsequent contact to L2 is limited or in an L2 academic community in which confirming evidence that L1 syntax is correct is lacking. The study of this paper investigates Chinese and Spanish learners across three English proficiency levels, who wrote in L1 and L2 at U.S. graduate schools. All participants completed sentence and discourse tasks. It is found that forward transfer occurred in the Chinese learners’ L2 writing at the discourse level and that backward transfer occurred in their L1 writing at the sentence level. Moreover, both forward transfer and backward transfer are mitigated by L2 proficiency. Furthermore, among the Level 2 Chinese learners, the relationship between forward transfer and L2 proficiency in the English sentence task follows a U-shaped curve, and the relationship between backward transfer and L2 proficiency in the Chinese sentence task also follows a U-shaped curve. The results of this study point to the complexity of language transfer and its interactions with L2 proficiency and distinctive task types.

Key words: language transfer, forward transfer, backward transfer, the U-shape curve, Principal Branching Direction (PBD), Principle of Temporal Sequence (PSE), crosslinguistic learner performance comparison, contrastive research, metalinguistic and crosslinguistic awareness

1. Introduction

One of the major issues in the field of Second Language Acquisition (SLA) is the role of language learners’ first language (L1) in the acquisition of a second language (L2). There is a question as to whether an erroneous or non-target-like L2 element (e.g. ‘Chinese English’ in Taiwan) is the result of L1 transfer or of creative construction2. This element is generally taken to be evidence of transfer (i.e. interlingual effects)3 if the presence of an L2 element is patterned on analogy to L1, whereas it is argued to be indicative of creative construction (i.e. intralingual effects) if the presence of an L2 element mirrors the one observed in L1 acquisition

1 The author wishes to offer his heartfelt thank to the 56 graduate students who participated in this study. He is grateful to the statistician Wang Yichun, who assisted with data analysis. The author’s sincere appreciation also goes to the Chief Editor Lillian Huang and the two anonymous reviewers for their insightful comments and criticisms. Any errors remain the sole responsibility of the author.

2 Creative Construction is defined by Dulay, Burt and Krashen (1982:11-12) as “the subconscious process by which language learners gradually organize the language they hear, according to rules that they construct to generate sentences. The form of the rules is determined by mental mechanisms responsible for human language acquisition and use.”

3 In this paper, ‘forward transfer’, ‘interlingual effects’, and ‘native language (NL) influence’ are used interchangeably.

(2)

(Ellis 1994). ESL teachers and SLA researchers have long been aware that ESL learners often have difficulties with academic writing in English (Lieber 1980, Reid, 1989). According to Kaplan (1966, 1983), rhetoric, in the sense of how information is arranged in writing, is shaped by culture and that there is a preference for a certain rhetorical pattern in each particular culture. Differences in rhetorical preference often have meant a clash between what they have learned in their native languages and cultures and what is expected of them from native speakers of English. As a result, ESL learners tend to transfer their L1 rhetorical patterns when they write in English.

However, the term transfer implies more than simply the effects of the L1 on the L2; the L2 also influences the L1 (Brown 2000, Su 2001, Cook 2003). This can be attested by numerous research which indicates that the effects of the L2 on the L1 play a central role in L2 environments. For instance, Sharwood Smith (1983) found, based on preliminary findings from a project in Utrecht, that native speakers of English, which is a high-status foreign language in the Netherlands, nevertheless underwent changes in their L1 English. Additionally, De Bot, Gommans, and Rossing (1991) examine Dutch immigrants in France and report that the Dutch informants’ L1 proficiency decreased over time when they had few contacts with L1 Dutch. Moreover, Carson and Kuehn (1994) investigate Chinese students across three English proficiency levels who wrote in L1 Chinese and L2 English in U.S. pre-academic intensive English programs, universities, and graduate schools. They found evidence that L1 Chinese writing proficiency declined as L2 English writing proficiency increased. More recently, Su (2001) examines transfer patterns at sentence level processing among L2 learners of English and Chinese across three proficiency levels. It is found that L2 learners not only used their L1 strategies in processing the L2 (forward transfer), but also applied their new L2 strategies in processing the L1 (backward transfer).

The purpose of this paper is two-fold. First, on the theoretical level, it is intended as a contribution to the understanding of how language learners' L2 acquisition is affected by their L1 knowledge (forward transfer) and how language learners’ L1 skills are in turn affected by their L2 proficiency (backward transfer). Researchers have long recognized that forward transfer from L1 to L2 appears primarily at the early stages of the learners’ interlanguage (IL) and decreases as their L2 proficiency increases (e.g. Taylor 1975, Kellerman 1979, Major 1986, Chan 2004). Additionally, it has been suggested that backward transfer from L2 to L1 may occur in an L2 environment where L2 learners have limited contact with their L1 or in an L2 academic community in which confirming evidence that L1 syntax is correct is lacking (e.g. Seliger 1991, Seliger and Vago 1991, Sharwood Smith and

(3)

Van Buren 1991, Carson and Kuehn 1994). Specifically, the study of this paper investigates the claim that the non-target-like L2 pattern of because-initial information sequencing frequently found in the writing of Chinese ESL learners is the result of Chinese to English forward transfer, due to the fact that this L2 English pattern corresponds to L1 Chinese discourse conventions (Lay 1975, Chiang 1981, Young, 1982). This article further seeks to identify how language transfer in general may relate to L2 proficiency. Secondly, on the pedagogical level, the purpose is to provide information for ESL teachers of writing in order to enable them to design more effective methods and materials for teaching writing to Chinese ESL learners. Thus, this paper is set up to provide data for addressing the following two research questions.

1. To what extent is the non-target-like L2 pattern of because-initial information sequencing in the L2 writing of Chinese ESL learners a function of Chinese to English forward transfer and how does forward transfer, if found to occur, relate to L2 English proficiency?

2. To what extent is the L1 discourse pattern of because-initial information sequencing in the L1 writing of Chinese ESL learners a function of English to Chinese backward transfer and how does backward transfer, if found to occur, relate to L2 English proficiency?

2. Linguistic pattern under investigation

Some fundamental issues raised by information sequencing in Chinese and English are highlighted with reference to linguistic, sociolinguistic, and psycholinguistic aspects.

2.1 Linguistic aspects of information sequencing

It has been well documented that, in general language use, the preferred or unmarked sequence in Chinese complex sentences is for the subordinate clause (SC) to precede the main clause (MC). Wang (1984:96) observes that “in Chinese, the main component comes at the end, and the subordinate component comes at the beginning”. Li and Zhang (1986) also indicate that the natural clause sequence in Chinese complex sentences is subordinate-to-main clauses (SC—MC), although the salient and less common main-to-subordinate clause (MC—SC) sequence is possible (Osgood 1980). Kirkpatrick provides an example which illustrates the most common SC—MC sequence in Chinese in (1) (Kirkpatrick 1993:31).

(4)

(1) yinwei feng tai da suoyi bisai gaiqi-le Because wind too big, so competition change time-A. ‘Because the wind was too strong, the competition was postponed.’

Example (1) is a causality sentence in which the conjunction yinwei (because) appears in the sentence-initial position. Additionally, Scollon (1993) indicates that there are two main structures in which because is used in English. The unmarked structure is as follows.

X because Y

where Y is taken to be the cause of X or explanation of X. The marked structure is

Because Y, X

Other researchers also point out that in the sequence of English main and subordinate clauses, the unmarked sequence has the main clause first (Clark and Clark 1977, Quirk, Greenbaum, Leech and Svartvik 1985, Prideaux and Hogan 1993). For example, in a study of 40 Chinese letters of request, Kirkpatrick (1991) demonstrates that the unmarked SC—MC sequence in complex sentences is also a fundamental principle for sequencing the information in discourse. He provides the English translation of a Chinese letter which reveals the Chinese writing preference for prefacing a request with reasons as noted in (2) (Kirkpatrick 1991:195).

(2) (Because) I was listening to a program last night (date) and heard the news that a lady colleague who had returned from Singapore was offering New Year's gifts, this excited my interest. I hope that she can send me some gifts and a photograph of herself, but because I have rudely forgotten her name, I hope you can take the trouble….

Kirkpatrick notes that the request—for some gifts and a photograph—in the body of the letter is preceded by the two reasons for the request, both following the

because-initial sequencing.

The contrastive information sequencing in Chinese and English is also consistent with the notion of Principal Branching Direction (PBD)4. Chinese and English are

4 Principal Branching Direction is defined by Lust (1983:138) as “the branching direction which holds consistently in unmarked form over major recursive structures of a language, where ‘major recursive

(5)

categorized as typologically distinctive languages with respect to the PBD: Chinese is a left-branching language, while English is largely a right-branching language5. This typological distinction accounts for the Chinese preference for an SC—MC sequence and the English preference for an MC—SC sequence.

2.2 Sociolinguistic aspects of information sequencing

Young (1982), in her analysis of Chinese subjects’ spoken English, argues that different ways of structuring information receive different social evaluations or rankings. She demonstrates that Chinese speakers prefer to delay their topic or thesis statement until supporting statements have been given as in the following example (Young 1982:77).

(3) Theta: One thing that I would like to ask. BECAUSE MOST OF OUR RAW MATERIALS ARE COMING FROM JAPAN AND THIS YEAR IS GOING UP AND UP AND IT’S NOT REALLY I THINK AN INCREASE IN PRICE BUT WE LOSE A LOT IN EXCHANGE

RATE AND SECONDLY I UNDERSTAND WE’VE SPEND A LOT OF MONEY IN TV AD

LAST YEAR. So, in that case I would like to suggest here: chop half of the budget in TV ads and spend a little money on Mad magazine.

As indicated by Young, the subordinate marker because initiates the listing of reasons in the supporting statements (capitalized), which establish the situational framework for evaluating the significant information to follow in the main clause (italicized). Young argues that Chinese speakers tend to minimize confrontation in formal social relationships, and this can be traced to culture-specific notions of acceptable discourse strategies. Chinese speakers find it uncomfortable to introduce their request at the outset, and thus by sequencing information differently from English native speakers, Chinese speakers are actually displaying a culturally appropriate discourse strategy: i.e. they are minimizing the imposition by exerting ‘negative politeness’ (Brown and Levinson 1987). Young concludes that there are correspondences between linguistic behavior and social evaluation and that difficulties in cross-cultural interactions will tend to occur when speakers are faced with an unfamiliar sociolinguistic tradition.

In the same vein, Scollon and Scollon (1991) claim that confusion in intercultural communication often arises as a result of differing discursive strategies in the placement of the topic statement. The Chinese discourse convention is an

complementation.”

5 That is, English allows for left-branching at the word level (e.g. a rich man), but is right-branching at the phrase (e.g. a man of wealth) and the sentence (e.g. This is the man whom I met yesterday) levels.

(6)

inductive one: i.e. the topic, such as a request, is generally deferred until after a considerable amount of discourse (and even “small talk”) which encodes reasons is provided. This reason-request sequence is the inverse of the deductive Western discourse pattern, in which the topic statement is given first and then followed by the supporting arguments, as exemplified by Schiffrin (1987:207) in (4).

(4) Can you work any of this with just the two of us, or you'll have to wait for Irene?

Cause I don't know how long she'll be.

In (4), the speaker made a request (accomplished by an indirect question)—Can we do without Irene?—and then gave a reason for it. However, to place a request at the beginning is deemed presumptuous in Chinese conversation, where the small talk is valued as a kind of extended ‘facework’ (Goffman 1967) that can mitigate the imposition later to appear in the topic statement. According to Scollon and Scollon, the Chinese inductive pattern of topic introduction is a natural outcome of interpersonal relationships, and the linguistic structure of SC—MC information sequencing which facilitates this interpersonal position is thus preferred.

2.3 Psycholinguistic aspects of information sequencing

The psycholinguistic aspects of information sequencing are pertinent to the concept of iconicity. For example, Ungerer and Schmid (1996:251) describe the Principle of Iconic Sequencing as “the sequence of two clauses corresponds to the natural temporal order of events”. They provide the example in (5).

(5) He opened the bottle and poured himself a glass of wine. *He poured himself a glass of wine and opened the bottle.

Here the first sentence clearly corresponds to the natural temporal order of events; the second sentence is unacceptable because the order in which the clauses are arranged violates the principle of iconic sequencing, although not ungrammatical according to the rules of syntax.

Sequential iconicity is best illustrated by Clark (1977) in the context of language development in children. Clark proposes that children utilize the Order of Mention Strategy in processing sentences. That is, children tend to interpret the order of mention of events as the linear (temporal) order of the events used as the reference point. Several research findings have corroborated that both Chinese and American

(7)

children encounter comprehension difficulty when the order of mention of events conflicts with the temporal order of events specified by the sentence (e.g. Kuo 1985, Kwoh 1997). Thus, it appears that children’s comprehension difficulty is largely one of organizing or representing information in the mind, rather than a purely ‘linguistic comprehension’ problem (Chang 1991).

The notion of sequential iconicity in Chinese is substantively explicated by Tai (1985, 1993), who has posited the Principle of Temporal Sequence (PTS): “the relative word order between two syntactic units is determined by the temporal order of the states which they represent in the conceptual world” (1985:50). According to the PTS, when two Chinese sentences are linked by temporal connectives, the first event always precedes the second one and the reverse is not possible as illustrated in (6).

(6) ni gei wo qian cai neng zou You gave me money then can leave ‘You can’t leave until you give me the money.’

The Chinese sentence in (6) would be ill-formed if the second event (“then can leave”) were ordered before the first event (“You give me the money”). Thus, the iconic nature of information sequencing in Chinese suggests a close parallel between surface linguistic behavior and underlying cognitive activities.

The SC—MC information sequencing in Chinese is also consistent with the PTS, since the cause (reason) always precedes the effect (consequence) in real time. However, the English translation in (6) shows that conjoined sentences in English need not incorporate the cognitive structure of the PTS, because the normal clause order in English is not constrained by the sequence of events. Hence, in terms of word order, Chinese is more iconic than English. Although psycholinguistic studies have revealed that human cognition and perception do not seem to vary considerably (e.g. Kwoh 1997), the linguistic patterns used to encode the conceptual principles in each culture may differ to a great extent, as exemplified by the disparate preferences for sequencing information in Chinese and English.

2.4. Summary

This section illustrates that the preferred or unmarked information sequence in Chinese follows an SC—MC order from the linguistic, sociolinguistic, and psycholinguistic perspectives. At the linguistic level, it has been shown that Chinese prefers to preface a main part with a subordinate one at both the sentence

(8)

and discourse levels, which reflects its typological feature as a left-branching language. At the sociolinguistic level, it has been pointed that a common Chinese discourse strategy is to introduce reasons at the outset and provide the intended request at the end. This inductive discourse pattern is to minimize imposition and maintain social relationship. At the psycholinguistic level, the unmarked SC—MC information sequencing (e.g. reason/cause precedes request/consequence) in Chinese reflects the iconic nature of word order in Chinese, which follows the conceptual principle of the PTS in actual time. By contrast, it is indicated that English generally follows an MS—SC information sequencing, prefers a deductive discourse pattern, and is less iconic in word order than Chinese. Thus, the study of this paper is to examine if Chinese ESL learners would use their preferred L1 SC—MC information sequencing in their L2 English writing and if they would use the unmarked L2 MS—SC information sequencing in their L1 Chinese writing. The following research design is set and the data are collected and analyzed statistically to address these two research questions raised earlier in this study.

3. Research design

3.1 Participants

While a contrastive approach is a necessary condition for determining the possibility of interlingual effects, it is not a sufficient one (Odlin 1989, Ellis 1994). The fact that an L1 pattern occurs in both L1 and L2 does not constitute sufficient evidence that transfer from L1 to L2 has taken place, since there can be other causes such as IL developmental factors (Rutherford 1987) or L2 learners’ lack of familiarity with the TL norms (Mohan and Lo 1985, Fakhri 1994). Thus, it is essential to incorporate into research design a crosslinguistic learner performance comparison that includes at least two language groups in which one group has the relevant pattern in the L1 and the other does not. If the data analysis further indicates a statistically significant difference between the two language groups with respect to the linguistic pattern under investigation, a claim that NL influence has occurred in the ILs of the learners would be then justified. Otherwise, interlingual or intralingual (i.e. developmental) effects as the sole explanation for the occurrence of a non-target-like L2 pattern in learners’ ILs would be equally questioned.

Therefore, the present study has included two ESL groups which constitute an experimental group and a control group. As mentioned earlier, Chinese is a left-branching language and English is a right-branching language. This typological distinction is reflected from the unmarked SC—MC (because-initial)

(9)

information sequencing in Chinese and the unmarked MC—SC (because-medial) information sequencing in English. Spanish, like English, is a right-branching language, and therefore Spanish native speakers should prefer the unmarked MC—SC information sequencing in the L1. This unmarked preference is further supported by a survey of eight native speakers of Spanish6. Thus, the Chinese ESL group, as the purpose of this investigation, is assigned to the experimental group. The Spanish ESL group, as a methodological requirement, is assigned to the control group.

Additionally, according to Demaris (1992), statistical inference in logistic regression relies largely on the asymptotic behavior of sample statistics, that is, the sampling distribution of deviance becomes closer and closer, or “converges” to the

Chi-squared distribution as the sample size increases toward infinity. In order for

these asymptotic properties to be approximately valid, the average cell or level size (that is, the sample size divided by the number of cells or levels) should be at least five. The same minimum requirement for sample size in inferential statistics is also indicated in Hatch and Lazaraton (1991). Based on the statistical considerations, this study has recruited a sample size of 24 for each ESL group, with eight participants on each proficiency level. The participants in the experimental group are 24 Chinese native speakers from Taiwan, and those in the control group are 24 Spanish native speakers from Spain. Eight native speakers of American English are further recruited as a baseline group for data comparison with the ESL groups. Thus, the database contains a total of 56 graduate students drawn from various departments at three universities in the Greater Philadelphia area.

Moreover, all ESL participants hold a bachelor’s degree in their native countries to ensure that the acquisition of the relevant linguistic pattern in their L1s has been completed through both formal and informal input. This extra criterion is based on research findings that suggest writing skills which have never been acquired in the L1 cannot be transferred to the L2 (Cummins 1981, John and Tetroe 1987).

6 Eight native speakers of Spanish from three academic communities were surveyed by the present author for their use of [a] Because A, B and [b] B because A in their L1 (Spanish). Their responses are indicated as follows. (1) Three ESL students from the Dominican Republic indicate that [a] is ungrammatical; (2) A community college student from Ecuador indicates that [a] is not commonly used; (3) A community college teacher from Ecuador indicates that both [a] and [b] are okay, but that [b] is better; (4) An ESL teacher from Puerto Rico indicates that [b] is better; (5) a Ph.D. student from Bolivia, who majors in philosophy, indicates that [b] is the common form, while A sounds “poetic”; and (6) a Ph.D. student from Spain, who majors in linguistics, indicates that both forms are correct, but that [a] is unusual. In summary, the because-medial structure (i.e. Because A, B) is either considered to be marked or ungrammatical by the Spanish native speakers surveyed.

(10)

3.2 English proficiency

It has been noted that a major methodological flaw of many earlier contrastive research is the lack control for L2 English proficiency in making claims about transfer of L1 rhetorical patterns or discourse structure in learners’ L2 production (Hinds 1982, 1983, Burtoff 1983). For instance, it has been claimed by a few previous contrastive studies conducted without reference to L2 proficiency that Arabic and Hebrew L1 writers are apt to transfer L1 rhetorical patterns, such as repetition and parallelism in their L2 English writing (e.g. Berman 1980, Ostler 1987). However, other researchers have found that low English proficiency writers, no matter if English is their first or second language, tend to use repetition and parallelism for cohesive purposes (e.g. Jacobs 1982, Scarcella 1984).

Thus, in assessing possible effects of L2 English proficiency on language transfer in this study, each of the two ESL groups is further divided into three proficiency levels on the basis of their TOEFL score. Specifically, Level 1 participants obtain TOEFL scores in the 500-547 range, Level 2 participants obtain TOEFL scores in the 550-597 range, and Level 3 participants have TOEFL scores in the 600-650 range. Since all the ESL participants are graduate students at American universities, their English proficiency levels for this study are based on their graduate school admission TOEFL scores. For most universities, international students with a TOEFL score in the 500-547 range may be granted probational admission into graduate studies, but they are required to take courses to upgrade their English in a university-affiliated English language program in addition to regular university courses. To be considered for full admission into graduate studies at an American university, international students must score at least 550 on the TOEFL. To further apply for a teaching assistantship, the minimal admission score is usually 600.

3.3 Materials

Early contrastive analyses which focus on sentence-level analysis are under attack for describing errors in isolation. Contrastive rhetoric studies are criticized for concentrating on L1-L2 surface differences and similarities. More recent research calls for an awareness that there is often an interaction between different levels of analysis—e.g. syntax and discourse—such that they need to be examined simultaneously. This study thus consists of two instruments to investigate the participants’ linguistic performance at both the sentence and discourse levels. The first instrument is a sentence-combining task (SCT) designed to elicit the pattern of

(11)

because-initial information sequencing (or alternatively termed because-initial

structure) in a complex sentence. The second instrument is a discourse task (DT) designed to elicit this pattern in discourse.

3.3.1 The sentence-combining task (SCT)

The SCT contains 20 test items in five semantic relations, encompassing causality, concession, contrast, condition, and temporality. The semantic relation of causality is the purpose of this task and other semantic relations merely function as distracters. Therefore, out of the 20 test items, 12 target items focus on causality, as signaled by because, to elicit the linguistic pattern at the sentence level. The remaining 8 test items contain the other four semantic relations, which were conveyed by conjunctions such as whereas, before, though, and unless as illustrated inthe English SCT in Table 1 below.

Table 1. Semantic relations in the English SCT7

(1) a. He arrived late at the station. b. He missed the morning train.

(causality: because)

(2) a. They want to live in a fancy big house. b. We would rather live in an apartment.

(contrast: whereas)

(3) a. Prices are going up. b. We’d better something.

(temporality: before)

(4) a. He is always late for work. b. He gets up early every morning.

(concession: though)

(5) a. I am forced to stay inside because of the rain. b. I usually go out for a walk early every day.

(condition: unless)

Additionally, the combining order in the SCT is flexible to allow the present author to see if because-initial structure occurs in a complex sentence. Moreover, the sentence pairs to be combined are all independent sentences, and the sentence order of potential because-clauses is counter-balanced lest the participants would combine each pair of sentences in the order in which they occur. The examples below illustrate the counter-balanced order and independent nature of the sentences to be combined which signal the semantic relation of causality as indicated in Table 2.

7 The semantic relations provided in parentheses do not appear in actual tasks. For a complete set of the English, Spanish, and Chinese tasks, see Appendices 2.1, 2.2, and 2.3.

(12)

Table 2. The counter-balanced sentence-combining order for the test items on causality

(1) a. He arrived late at the station b. He missed the train.

(reason) (consequence) (2) a. He has to work hard in his old age.

b. He was idle in his youth.

(consequence) (reason)

3.3.2 The discourse task (DT)

In the DT, the ESL learners are asked to write a narrative of 150-200 words from six sequential pictures. For each picture, the ESL learners are instructed to explain why an activity is taking place. It is hoped that by providing a reason for the activity in each picture, the ESL learners will have least six opportunities to produce the linguistic pattern under investigation at the discourse level. In other words, there is a potential number of 144 tokens for the 24 participants in each ESL group (6 x 24) to produce either the because-initial structure (native-like L1 Chinese pattern) or because-medial structure (target-like L2 English pattern) at the discourse level. Thus, dividing the actual tokens of occurrence by the 144 potential tokens of occurrence will reveal the two ESL groups’ preference for sequencing information at the discourse level. An example of the DT is given in Figure 1.

(13)

Figure 1. The Discourse Task8

You will see a series of six pictures. In each of these pictures, AN ACTIVITY is going on. Please write a narrative of about 150-200 words in English that describes the sequence of the activities. For example, in picture one, the worker is putting a sign on the bench. Go through each picture and explain WHY that activity is happening. Make sure that you do this for EACH PICTURE, as you construct your narrative. You will have 20 minutes to complete this task.

3.4. Procedures

There are two different sets of tasks (i.e. two SCTs and two DTs), one in the participants’ L1s and the other in their L2 English. The L1 tasks are in English, Chinese, and Spanish, and are completed by the American, Chinese, and Spanish participants. The American participants complete L1 tasks to establish the English baseline; this ends their participation in this study. The Chinese and Spanish ESL participants have a 2-part procedure. They first complete one set of tasks in one language and then complete the other in another language. The L1 tasks they complete are syntactically identical to but semantically different from the L2 tasks

8 Only the self-drawn picture series in the L1 DT are included in this article. The L1 Chinese/Spanish DT is identical in format to the L2 English DT but different in the picture contents. The picture series on the English SCT is an excerpt from a TSE (Test of Spoken English) materials issued by Educational Testing Service (ETS). Permission for using the picture series in this study is granted by ETS, but the copyrighted picture series are not to be distributed or published by the author.

(14)

they complete. This is necessary to ensure that they would not be completing the exact same sets of tasks, only in a different language. In addition, the presentation of L1/L2 task order to the Chinese and Spanish ESL participants is counter-balanced to control for possible practice effects: half of the Chinese and Spanish ESL participants receive the L1/L2 task order, and the other half receive the L2/L1 task order.

4. Results

4.1 The first research question

The first research question concerns issues related to forward transfer and L2 proficiency. In order to address the first part of this question, it is hypothesized that

(1a) Chinese ESL learners, when writing in English, will supply significantly more of the non-target-like L2 pattern of because-initial information sequencing at both the sentence and discourse levels, which resembles that of their L1, than will Spanish ESL learners who do not prefer this pattern in their L1.

The overall performance of because-initial and because-medial information sequencing produced by the Chinese and Spanish ESL groups in the English SCT and English DT is first provided in relative frequencies and followed by significance tests as indicated in Table 3 and Table 4, respectively.

Table 3. The relative frequencies of using information sequencing by the Chinese and Spanish ESL groups in the English SCT and English DT

ESL Groups The English SCT The English DT

Information Sequencing because- initial because- medial because- initial because- medial Chinese 99 38.22% 160 61.78% 32 52.46% 29 47.54% Spanish 19 7.88% 222 92.12% 3 4.48% 64 95.52%

As indicated by the descriptive statistics in Table 3, the Chinese ESL group appears to use more of because-initial structure than the Spanish ESL group in both the English SCT (38.22% vs. 7.88) and the English DT (52.46% vs. 4.48%). However, the number of because-initial structure used by the Chinese ESL group in the

(15)

English tasks must significantly exceed the number used by the Spanish ESL groups in order to argue for the existence of interlingual effects (i.e forward transfer) from L1 Chinese to L2 English. Thus, to determine if such interlingual effects exist, significant tests are conducted to assess the evidence provided by the sample data as given in Table 4.

Table 4. Difference between Chinese and Spanish ESL groups in their use of

because-initial structure in the English SCT and English DT

Tasks Null Residual Dispersion df χ2

Deviance Deviance

English SCT 252.5044 183.6215 3.4847 1 19.76** English DT 95.4097 54.1442 1.5333 1 26.91**

**p < 0.01 df = degrees of freedom

As seen in Table 4, the results of inferential statistics using the Chi-square test show that there is a significant difference between the performance of these two ESL groups in both English tasks (p < 0.01). The statistical results indicate that the two ESL groups have significantly different tendencies in sequencing information at both the English sentence and discourse levels. The Chinese ESL group exhibits a distinctive preference for using because-initial information sequencing, whereas the Spanish ESL group shows a greater tendency to use because-medial information sequencing. Hypothesis (1a) is therefore supported.

The second part of the first research question addresses the extent to which forward transfer relates to L2 English proficiency. Based on previous research findings that forward transfer decreases as L2 proficiency increases, it is hypothesized that

(1b) There is a negative relationship between L2 English proficiency and the non-target-like use of because-initial information sequencing in the L2 writing of Chinese ESL learners at both the sentence and discourse levels.

The overall performance of because-initial information sequencing produced by the Chinese ESL groups in the English SCT and DT across proficiency levels is first provided in frequency distributions and followed by significance tests as indicated in Table 5 and Table 6, respectively.

(16)

Table 5. The relative frequencies of using information sequencing by the Chinese ESL groups across three proficiency levels in the English SCT and English DT

Participants The English SCT The English DT

Information Sequencing because- initial because- medial because- initial because- medial Level 1 58 69.05% 26 30.95% 16 69.57% 7 30.43% Level 2 11 13.75% 69 86.25% 11 55% 19 45% Level 3 30 31.58% 65 68.42% 5 27.78% 13 72.13% Chinese Total 99 38.22% 160 61.78% 32 52.46% 29 47.54%

As indicated by the descriptive statistics in Table 5, the Chinese ESL group exhibits different use of the because-initial structure at different proficiency levels in both English tasks. To determine if the differences among the three Chinese subgroups achieve statistical differences, inferential statistics using the Chi-square are further conducted as indicated in Table 6.

Table 6. The overall relationship between L2 English proficiency and the use of

because-initial structure among the Chinese ESL group in the English

SCT and English DT

Tasks Null Residual Dispersion df χ2

Deviance Deviance

English SCT 137.2334 79.1889 3.3983 2 17.81**

English DT 35.1894 27.8360 1.0617 2 6.93*

**p < 0.01 *p < 0.05 df = degrees of freedom

The results of inferential statistics reveal a significant difference among the three English proficiency levels in both the English SCT (p < 0.01) and the English DT (p < 0.05). Given that statistical results indicate a significant effect of L2 proficiency on the production of because-initial structure in both English tasks, a one-tailed

t-test9 is further conducted to locate the exact differences among the three L2

9 Whether a one-tailed or two-tailed t-test is used depends on if the direction (either positive or negative) of the alternative hypothesis is specified. On the basis of previous research, if a researcher finds a strong relationship between a dependent variable and the dependent variable(s), then the direction of that relationship can be specified; that is, whether there will be higher (positive) or lower (negative) than more typical scores. But since no direction is specified for the null hypothesis, it is necessary to consider both tails of the distribution. Therefore, a null hypothesis is called a two-tailed or nondirectional hypothesis. In contrast, an alternative hypothesis can be either a two-tailed (nondirectional) or one-tailed (directional) hypothesis, depending on if the direction has been specified (Hatch and Lazaraton 1991). In this study, a one-tailed t-test

(17)

proficiency levels.

4.1.1 The English SCT

A significant difference is found in all three pairs of English proficiency levels (p < 0.01), between Level 1 and Level 2, Level 1 and Level 3, and Level 2 and Level 3, in the English SCT as noted in Table 7.

Table 7. Differences in the use of because-initial structure among the pairs of English proficiency levels of the Chinese ESL group in the English SCT

Level Level 1 Level 2 Level 3

Log odds 0.8266 0.2205 -0.9555

Level 1 -6.5752** -4.8756**

Level 2 -2.7084**

** p < 0.01

The frequency rate of using because-initial structure by the Chinese ESL group across the three proficiency levels in the English SCT, as indicated in Table 5 earlier, is 69.05% at Level one, 13.75% at Level 2, and 31.58% at Level three. The numerical differences can be further illustrated in Figure 2.

0 0.2 0.4 0.6 0.8

Level 1 Level 2 Level 3

L2 English Proficiency

Figure 2. The frequency rate of using because-initial structure among the three English proficiency levels of the Chinese ESL group in the English SCT

Figure 2 exhibits a pattern in which the frequency rate across the three proficiencies is initially high (Level 1), then declines (Level 2), and rises again (Level 3). Hypothesis (1b) is not supported at the sentence level, since the relationship between forward transfer and L2 proficiency is not a strictly negative one.

(18)

4.1.2 The English DT

A significant difference is found between Level 1and Level 3 (p < 0.01), and between Level 2 and Level 3 (p < 0.05) in the English DT. Moreover, there is a statistical trend in the difference between Level 1 and Level 2 (p < 0.17), although not significant at the 0.05 level. Statistical results are displayed in Table 8.

Table 8. The frequency rate of using because-initial structure among the three English proficiency levels of the Chinese ESL group in the English DT

Level Level 1 Level 2 Level 3

Log odds 0.8266 0.2205 -0.9555

Level 1 -0.9808 -2.5663**

Level 2 1.7707*

** p < 0.01, *p < 0.05

The frequency rate of using because-initial structure by the Chinese ESL group across the three proficiency levels in the English DT, as indicated in Table 5 earlier, is 69.57% at Level one, 55% at Level 2, and 27.78% at Level three. The numerical differences can be further illustrated in Figure 3.

0 0.2 0.4 0.6 0.8

Level 1 Level 2 Level 3

L2 English Proficiency

Figure 3. The frequency rate of using because-initial structure among the three English proficiency levels of the Chinese ESL group in the English DT

Figure 3 shows a downward linear curve. Thus, it is clear that the relationship between L2 proficiency and L1 transfer is in a negative direction. That is, the more advanced the Chinese participants are in L2 English proficiency, the fewer this non-target-like pattern they tend to produce in their L2 writing, and vice versa. Hypothesis (1b) is therefore supported at the discourse level.

(19)

4.2 The second research question

The second research question concerns issuesof howbackward transfer, if found to occur, relates to L2 proficiency. Based on previous research findings that L1 skills may decline in an L2 environment, it is hypothesized that

(2) There is a relationship between L2 English proficiency and the use of the L1 rhetorical pattern of because-initial information sequencing in the L1 writing of Chinese ESL learners at both the sentence and discourse levels.

The overall performance of because-initial information sequencing produced by the Chinese ESL groups in the Chinese SCT and Chinese DT across proficiency levels is first provided in relative frequencies distributions and followed by significance tests as indicated in Table 9 and Table 10, respectively.

Table 9. The relative frequencies in the performance of the Chinese ESL groups on the Chinese SCT and Chinese DT across three English proficiency levels

Participants The Chinese SCT The Chinese DT

Information Sequencing because- initial because- medial because- initial because- medial Level 1 60 68.97% 27 31.03% 27 87.1% 4 12.9% Level 2 26 29.21% 63 70.79% 28 93.33% 2 6.7% Level 3 59 64.84% 32 35.16% 30 88.24% 4 11.76% Chinese Total 99 38.22% 160 61.78% 32 52.46% 29 47.54%

As indicated by the descriptive statistics in Table 9, the Chinese ESL group exhibits different use of the because-initial structure at different proficiency levels in both the Chinese SCT and Chinese DT. To determine if the differences among the three Chinese subgroups achieve statistically significant differences, inferential statistical tests using the Chi-square are further conducted as indicated in Table 10.

(20)

Table 10. The overall relationship between L2 English proficiency and the use of because-initial structure among the Chinese ESL group in the Chinese SCT and Chinese DT

Tasks Null Residual Dispersion df χ2

Deviance Deviance

Chinese SCT 126.9455 92.1013 3.6999 2 9.42**

Chinese DT 28.0038 27.3490 1..4407 2 0.46

**p < 0.01 df = degrees of freedom

As displayed in Table 10, inferential statistics reveals a significant L2 proficiency effect in the Chinese SCT (p < 0.01), but not in the Chinese DT (p < 0.80). In what follows, inferential statistical tests are further conducted to identify the possible differences among the three L2 proficiency levels.

4.2.1 The Chinese SCT

Further statistical results using the two-tailed t-test10 indicates a significant difference between Level 1 and Level 2 (p < 0.05), and between Level 2 and Level 3 (p < 0.05) in the Chinese SCT. However, no significant difference is found between Level 1 and Level 3 (p < 0.62). That is, the negative L2 effect on

because-initial information sequencing in Chinese SCT (resulting in low frequency

of use in this pattern) is most conspicuous in the Level 2 Chinese learners as shown in Table 11.

Table 11. The frequency rate of using because-initial structure among the three English proficiency levels of the Chinese ESL group in the Chinese SCT

Level Level 1 Level 2 Level 3

Log odds 0.7985 -0.8850 0.6118

Level 1 -5.1220* 0.5849

Level 2 -4.6751*

*p < 0.05

The frequency rate of using because-initial structure by the Chinese ESL group across the three proficiency levels in the Chinese SCT, as indicated in Table 9 earlier,

10 Given that hypothesis (2) is stated in a non-directional form due to the relative lack of strong evidence between backward transfer and L2 proficiency in previous studies, the two-tailed, rather than one-tailed,

(21)

is 68.97% at Level one, 29.41% at Level 2, and 64.84% at Level three. The frequency rate at Level 2 is only 29.21%. This is rather low, compared to the frequency rates at Level 1 and Level 3, which are 68.97% and 64.84%. The minimal difference between Level 1 and Level 3 (68.97% vs. 64.84%) clearly indicates why inferential statistics in Table 11 yields no significant difference between these two levels (p < 0.62). In contrast, the vast differences between Level 2 and Level 1 (29.21% vs. 68.97%) and between Level 2 and Level 3 (29.21% vs. 64.84%) indicate why inferential statistics reveals a significant difference between Level 2 and Level 1, and between Level 2 and Level 3 (p < 0.05). The numerical differences among the three proficiency levels at the Chinese SCT can be further illustrated in Figure 4.

0 0.2 0.4 0.6 0.8

Level 1 Level 2 Level 3

L2 English Proficiency

Figure 4. The frequency rate of using because-initial structure among the three English proficiency levels of the Chinese ESL group in the Chinese SCT

Figure 4 exhibits a pattern in which the frequency rate for the three proficiency levels is initially high (Level 1), then declines (Level 2), and rises again (Level 3). Thus, hypothesis (2) is partly supported at the sentence level, since there is a negative L2 effect on the use of because-initial structure among the Level 2 Chinese learners in Chinese SCT but not on the Level 1 and Level 3 Chinese learners. In other words, the Level 2 Chinese participants are likely to experience backward transfer from L2 English to L1 Chinese, a point that is returned to in the Discussion Section.

4.2.2 The Chinese DT

As shown in Table 10, there is no significant effect of L2 English proficiency on the use of because-initial structure by the Chinese ESL students in the Chinese DT (p < 0.80). The frequency rate of using because-initial structure by the Chinese

(22)

ESL group across the three proficiency levels in the Chinese DT, as indicated in Table 9 earlier, is 87.1% at Level 1, 99.33% at Level 2, and 88.24% at Level 3. The numerical differences can be further illustrated in Figure 5.

Figure 5. The frequency rate of using because-initial structure among the three English proficiency levels of the Chinese ESL group in the Chinese DT

Figure 5 exhibits a virtually horizontal linearity, indicating that the Chinese ESL learners at all proficiency levels show a strong preference for the L1 discourse pattern of because-initial structure in the Chinese DT. Thus, L2 proficiency does not appear to condition the use of this L1 structure in the Chinese DT. Hence, hypothesis (2) is not supported at the discourse level.

5. Discussion

5.1 The first research question

To address the first research question on forward transfer, hypothesis (1a) predicts that Chinese ESL learners, when writing in L2 English, will supply more of

because-initial structure, as influenced by their L1, than will Spanish ESL learners

who do not prefer this pattern in their L1. The results, as shown in Table 4, indicate a statistically significant difference between the two ESL groups in both the English SCT and the English DT. This finding lends further support to the previous L2 research claim that Chinese ESL learners tend to transfer their L1 discourse strategy of SC—MC information sequencing into the their L2 English at the both sentence and discourse levels (Young 1982, Scollon and Scollon 1991 Kirkpatrick 1993).

Regarding the second part of the first research question on the relationship 0

0.2 0.4 0.6 0.8

Level 1 Level 2 Level 3

(23)

between forward transfer and L2 proficiency, hypothesis (1b) predicts that there is a negative relationship in these two variables at both the English sentence and discourse levels. However, statistical results, as respectively shown in Figure 2 and Figure 3, reveal that hypothesis (1b) is only supported at only the discourse level but not the sentence level. This is because the Level 2 Chinese ESL learners appear to produce high frequency occurrences of target-like MS—SC information sequencing (i.e. because-medial structure) in the English SCT (86.25%, see Table 5). In an attempt to unravel the mystery of the Level 2 Chinese learners’ unusual performance on the English SCT, the present author further conducts within-task and between-task data comparisons for the English SCT and English DT.

5.1.1 The within-task data comparison: Overuse

The section first discusses the Chinese learners’ performance on the English DT, because it is consistent with previous research findings. As shown in Table 5 and Figure 3, the Chinese learners’ performance data in the English DT exhibit a downward linear curve: 69.57% at Level 1, 55% at Level 2, and 27.78% at Level 3. These are the percentages in which the Chinese learners use a marked L2 pattern (i.e.

because-initial structure), although it is considered unmarked in their L1. These

performance data fit in well with previous research claims that there is a negative relationship between L2 proficiency and forward transfer (e.g. Takahashi and Dufon 1989, Maeshiba, Yoshinaga, Kasper and Ross 1996). That is, the least proficient Level 1 Chinese learners are found to aptly transfer their L1 discourse pattern of

because-initial information sequencing into their L2 (69.57%), whereas the most

proficient Level 3 Chinese learners resort less to this L1-based discourse strategy in their L2 writing (27.78%) with the Level 2 intermediate Chinese learners in between (55%).

In contrast, the Chinese learners’ performance data in the English SCT, as shown in Table 5 and Figure 2, exhibit a U-shaped behavior pattern11 (Kellerman 1979, 1983, 1985, Gatbonton 1983, Lightbown 1985, Gershokoff-Stow and Thelen 2004). That is, the frequency rate of using L1-based discourse strategy in their L2 writing is initially high (69.05% at Level 1), then declines (13.57% at Level 2), and rises again (31.58% at Level 3). These results do not support the negative relationship between L2 proficiency and forward transfer. That is, the Level 2 (intermediate)

11 The notion of U-shaped behavior in SLA is first proposed by Kellerman. That is, L2 learners may sometimes pass through an early stage of development where they exhibit correct use of a target-like L2 form if this form corresponds to an L1 form and then replace it with a developmental non-target-like L2 form before finally returning to the correct target-like L2 form again. Thus, error frequency in the three stages is initially low, then rises, and finally declines again; conversely, accuracy level is initially high, then declines, and finally rises again, yielding the U-shape in a graphic representation of the performance data.

(24)

Chinese learners (13.57%) are found to produce less of non-target-like L2

because-initial structure than both the Level 1 (beginning) Chinese learners (69.05%)

and the Level 3 (advanced) Chinese learners (31.58%). In other words, the Level 2 Chinese learners produce significantly more of target-like L2 because-medial structures (86.25%) than both the Level 1 (30.95%) and the Level 3 (68.42%) Chinese learners. The Level 2 Chinese speakers even outperform the native English baseline group (80.23%) in producing this target-like L2 unmarked pattern in the English SCT (see Appendix 1.1). It is unlikely that the Level 2 intermediate Chinese ESL learners are actually ‘better’ than the Level 3 advanced Chinese ESL learners and the English native speakers in acquiring the target-like L2 English pattern of because-medial information sequencing. Thus, the U-shaped pattern may have resulted from a so-called ‘overuse’ behavior by the Level 2 Chinese learners in overindulging in their use of this target-like L2 pattern, thus leading to their underrepresented use of the L1 discourse pattern (Levinston 1971).

Overuse of because-medial structure as a result of transfer of training

How can the overuse behavior of the Level 2 Chinese learners be explained? The reason may have been due to transfer of training (Selinker 1972). In other words, in a context of learning L2 writing in the U.S. academic environment, the Level 2 Chinese learners might have been instructionally directed toward the unmarked L2 use of because-medial information sequencing, and thus have made a conscious effort to approximate this usage (Corder 1978). Because of this likely transfer of training effect, they appear to overuse the target-like unmarked L2 pattern in the English SCT. This explanation via a transfer of training effect seems plausible since six of the eight Level 2 Chinese students, whose TOEFL scores are lower than 575, come from a university which has required them to take additional ESL courses.

L2 availability and restructuring process

However, why did overuse behavior, as a possible result of the transfer of training effect, occur only to the Level Chinese 2 learners, but not to the Level 1 and the Level 3 learners? Two major theories in SLA may help to interpret this result. Firstly, it may be related to the Chinese learners’ L2 processing level, an issue of learnability (Pienemann 1984). Moreover, it has been pointed out that the L2 resources and access level available to learners might act as constraints on development (Freeman and Long 1991). Secondly, the research view of SLA as a

(25)

restructuring process offers a further explanation. According to Lightbown (1985),

SLA is not simply linear and cumulative, but is characterized by a processing of restructuring according to a U-shape, first declining as more complex internal representations replace less complex ones, and increasing again as skills become proficient. Similarly, Corder (1978) views the process of SLA as a restructuring

continuum in which learners first utilize their L1, gradually restructuring it as they

discover how it differs from the L2. In other words, learners’ IL is viewed as a series of stages consisting of the gradual replacement of L1 rules by L2 rules. Moreover, Anderson (1983) maintains that the early stages of SLA involve the creation of independent and autonomous linguistic systems, at least partly distinct from the L2, which he termed nativization; this is distinguished from denativization, referring to the later stages of SLA, in which the circumstances cause learners to restructure their linguistic system to conform more closely to that of the L2.

Thus, at the earler stages of the IL continuum, it is only natural for the Level 1 Chinese learners to utilize their L1 for hypotheses about the L2, as evidenced by their high frequency of use in the L1 discourse pattern of because-initial information sequencing (69.05%) in the English SCT. This result suggests that forward transfer from L1 to L2 is primarily a processing and acquisition strategy in the autonomous, pre-restructuring stage in L2 learners’ IL system. On the other hand, the Level 3 Chinese learners, who are more fluent in L2, may have been approaching the later stages of restructuring and therefore have a better command of the target-like unmarked L2 pattern. This could explain their relatively fewer use of

because-initial information sequencing (31.58%) in the English SCT, as compared

to that of the Level 1 Chinese learners (69.05%), although still considerably more than that of the English baseline group (19.77%). In contrast, the level 2 Chinese learners, at the intermediate course of IL development, may have begun to separate the L1 and L2 systems, as a result of their developing an independent set of L2 rules distinct from those of L1. But since they are still at the transitional stages of restructuring, the Level 2 Chinese learners may not yet have a proper sense of the natural (unmarked) use of this target-like L2 item but tend to overuse it.

The theories of L2 processing level and restructuring process can also account for why the transfer of training effect may not affect the Level 1 and Level 3 Chinese learners. Transfer of training does not appear to be an issue for the Level 1 Chinese learners, due to their heavy reliance on L1 acquisition strategy (i.e.

because-initial information sequencing) in an autonomous, pre-restructuring stage in

their IL (Anderson 1983). Moreover, even if the Level 1 Chinese learners were exposed to the linguistic pattern (i.e. because-medial information sequencing) under investigation in an ESL classroom, transfer of training would not take effect so

(26)

easily (Pienemann 1984, Freeman and Long 1991) before their processing level is elevated (say, to the intermediate level). Additionally, transfer of training does not affect the Level 3 Chinese learners, because they are advanced learners who might have developed crosslinguistic awareness of the markedness difference between information sequencing in Chinese and English, and would not regard

because-medial structure as the sole information sequencing strategy in English.

By contrast, the Level 2 Chinese learners are presumably at the unstable, transitional stages of restructuring in their IL and are therefore more susceptible to the transfer of training effect.

5.1.2 The between-task data comparison: Task effect

The suggestion that the Level 2 Chinese learners may not have internalized the target-like unmarked L2 pattern can be further attested by a comparison of their performance data from the two different tasks. As can be seen from Table 5, the frequency with which the Level 2 Chinese learners produce the non-target-like L2 pattern (i.e. because-initial structure) increases drastically when they turn from the English SCT (13.75%) to the English DT (55%). This difference in performance at the two different linguistic levels reflects a task effect (i.e. the learners behave linguistically different between different task types). In contrast, for the Level 1 and the Level 3 Chinese learners, the difference in their performance data for the two linguistic levels is only minimal and does not show a task effect. For the Level 1 Chinese learners, their consistently high frequency of use in the non-target-like L2 pattern between both the English SCT and DT (69.05% vs. 69.57%) reveals a strong forward transfer from L1 to L2. For the Level 3 Chinese learners, their consistently lower frequency of use in this non-target-like L2 pattern in both English tasks (31.58% vs. 27.78%) suggests a weaker forward transfer and reflects a more target-like L2 use, pointing more to the direction of the native-like use by the English baseline group in both English tasks (19.77% vs. 16.67%).

Distribution of attention due to the task types

What requires further explanation, however, is why the overuse of the L2 target-like pattern by the Level 2 Chinese learners is only manifest in the English SCT but not in the English DT. There is a possible explanation for it. That is, it may have to do with the attention given to different task types. As pointed out by L2 researchers (e.g. Ney and Fillerup 1980, Zamel 1980, Azabdaftari 1986, Suhor 1987, Sajjadi and Tahririan 1992, Ito 1997), sentence-combining is a grammar task

(27)

used to help students understand sentence structure. It requires that they pay attention to individual grammatical forms and not to communication of ideas. In contrast, the picture description task is a communicative task, one which has been found to produce more interesting ideas and be very helpful to students unfamiliar with the subject matter or the cultural context of a literary text (Sarkar 1978).

Thus, the Level 2 Chinese learners may have seen the English SCT as a grammar task which requires that they attend to grammatical forms. This may have facilitated their use of the learned, but not yet internalized, unmarked L2 pattern, thereby leading to their overuse behavior. On the other hand, English DT is a writing task which requires that they focus on communicating ideas, a task which does not allow them to focus all their attention on the target-like L2 form. Hence, concentrating on generating ideas, they revert to their L1 discourse strategy and carry it over into the English DT. This can explain why the overuse of target-like L2 because-medial information sequencing is not manifest in the English DT.

The task types, however, do not seem to have an effect on the Level 3 and Level 1 Chinese learners. This is presumably because the Level 3 Chinese learners, at their current advanced stage, have generally move beyond certain syntactic processing constraints (in this case, because-medial structure) and are capable of directly processing the semantic content, an automaticized process that distinguishes proficient readers from non-proficient ones no matter in their L1 or L2 (Tzeng and Hung 1981). Therefore, the Level 3 Chinese learners should not be constrained by this specific grammatical form no matter in the English SCT or English DT, since they may have internalized/automaticized that form. The task effect will not affect the Level 1 Chinese learners, either since they may have not been aware of the L1-L2 difference between because-initial and because-medial structures in terms of markedness (i.e. frequency of use).

The tendency of target-like L2 patterns to appear and disappear at various points of development is labeled by Selinker (1972) as backsliding. Selinker indicates that L2 learners may backslide and lose target-like L2 forms and structures, ones they have seemingly mastered previously in isolation, when their attention is focused on new and difficult intellectual subjects. This is likely to be the case of the Level 2 Chinese learners in the present study, since the English DT is more cognitively demanding than the English SCT12. The suggestion that the contrasting performance data of the Level 2 Chinese learners may have been due to the task

12 The task effect between the SCT and DT that conditions the Level 2 Chinese learners (the experimental group) in terms of conscious attention to form does not apply to the Spanish control group, since Spanish, like English, uses the unmarked MC—SC (because-medial) information sequencing. Therefore, the issue of language transfer (either forward or backward) between L1 Spanish and L2 English due to the task effect will simply not occur.

(28)

effect is supported by previous research. For instance, it has been shown that more communicative tasks, which require conscious attention to content, tend to yield lower accuracy in grammatical forms, whereas sentence-combining, which is more formal and demands more conscious attention to form, will generally result in greater accuracy in the use of grammatical forms (Hulstijn and Hulstijn 1984, Sajjadi and Tahririan 1992).

5.2. The second research question

To address the second research question on possible backward transfer, hypothesis (2) predicts that there is a relationship between L2 proficiency and the use of the L1 discourse pattern of because-initial information sequencing in the L1 writing of Chinese ESL learners. As shown in Table 9 and Figure 5, the Chinese learners at all proficiency levels are found to predominantly use the L1 discourse pattern in the Chinese DT, confirming the research claim that SC—MC information sequencing is a fundamental principle in Chinese discourse. However, the Level 2 Chinese learners behave rather differently in the Chinese SCT. As shown in Table 9 and Figure 4, L2 proficiency has a negative effect on the performance data in the Chinese SCT, due to the meager use of the L1 discourse pattern by the Level 2 Chinese learners at the sentence level. Explanations pertaining to the negative L2 effect are suggested below.

5.2.1 The possible source of the negative L2 effect: Backward transfer (L1 attrition and L1 change)

There is some research evidence to indicate that proficient L1 writers are shown to gradually decline in their L1 writing skills (Carson and Kuhn 1994) or change in their L1 processing strategy (Su 2001) in an L2 environment. From another perspective, L1 attrition or L1 change, together termed more generally as backward transfer, is considered to be a rule-governed process, one that is affected by the kinds of data or sources of knowledge available to the learner to be used as evidence in testing the fitness of his construction of L1 grammar. In a similar vein, it has been argued that L1 speakers need evidence not only to develop an L1 system but also to maintain it (Seliger 1991). In other words, backward transfer occurs not because of lack of use but because of lack of confirming evidence that L1 syntax is correct in an L2 community (Sharwood Smith and Van Buren 1991). Thus, for the Chinese ESL learners, it is possible for an L2 pattern to begin to influence on the linguistic domain of L1, since their linguistic context is that of a U.S academic

(29)

community where the L1 discourse pattern is either not preferred (lack of confirming evidence) or not much used (lack of exposure). However, it remains to be explained as to why backward transfer, as the result of a negative L2 English effect, has occurred only with the Chinese SCT but not the Chinese DT.

5. 2. 2 Distribution of attention due to different task types

The inconsistent results for the Chinese SCT and the Chinese DT may have been due to attention distribution between different task types, as has been discussed in section 5.1.2 with respect to the English SCT and English DT. To recapitulate, the Chinese learners might have equally seen the Chinese SCT as a grammar task, which facilitates their overuse of the unmarked L2 pattern (i.e. because-medial structure) leading to their underrepresented use of the L1 discourse pattern (because-initial structure), even in their L1 Chinese writing. On the other hand, the cognitively demanding nature of the Chinese DT prevents the Chinese learners from focusing solely on the L2 form, so they revert to their preferred L1 discourse pattern in the Chinese DT. This may explain why the backward transfer or reverse transfer of an L2 form into the L1 (Cook 2003), is not manifest in the Chinese DT. Hence, it appears that whether backward transfer will occur has something to do with the Level 2 Chinese learners’ ability or inability to attend to the target-like L2 pattern, which may in turn be contingent upon the nature of the task types.

Also worthy of special mention is the significant issue of L1 attrition in terms of the distinction between competence and performance. The former refers to the tacit linguistic knowledge of the learner, whereas the latter the control of that knowledge, which has to do with the nature of the “online accessing and processing” (Seliger and Vago 1991). In general, L1 attrition literature appears to accept a competence dimension as well as a control dimension (Sharwood Smith and Van Buren 1991). However, it is clear that the backward transfer (a more neural and general term than language attrition), which occurs to the Level 2 Chinese learners in the Chinese SCT rather than the Chinese DT, has more to do with the online access mechanisms (performance and control) than with the available tacit knowledge itself (competence).

The restructuring process of the Level 2 Chinese ESL learners

In their L1 Chinese task, the Level 2 Chinese learners produce their L1 native discourse pattern significantly less than the Level 1 and the Level 3 Chinese learners. Similar to their unusual linguistic behavior in the English SCT (i.e. overuse of the

數據

Table 1. Semantic relations in the English SCT 7
Table 2. The counter-balanced sentence-combining order for the test items on          causality
Figure 1. The Discourse Task 8
Table 6. The overall relationship between L2 English proficiency and the use of
+7

參考文獻

相關文件

6 《中論·觀因緣品》,《佛藏要籍選刊》第 9 冊,上海古籍出版社 1994 年版,第 1

• One technique for determining empirical formulas in the laboratory is combustion analysis, commonly used for compounds containing principally carbon and

 Promote project learning, mathematical modeling, and problem-based learning to strengthen the ability to integrate and apply knowledge and skills, and make. calculated

Teachers may consider the school’s aims and conditions or even the language environment to select the most appropriate approach according to students’ need and ability; or develop

Wang, Solving pseudomonotone variational inequalities and pseudocon- vex optimization problems using the projection neural network, IEEE Transactions on Neural Networks 17

Hope theory: A member of the positive psychology family. Lopez (Eds.), Handbook of positive

Define instead the imaginary.. potential, magnetic field, lattice…) Dirac-BdG Hamiltonian:. with small, and matrix

According to a team at Baycrest’s Rotman Research Institute in Canada, there is a clear link between bilingualism and a delayed onset of the symptoms of Alzheimer ’s and other