Topic prominence in Chinese EFL learners’ interlanguage

The present study aims to investigate the general characteristics of topicprominent typological interlanguage development of Chinese learners of English in terms of acquiring subject-prominent English structures from a discourse perspective. Topic structures mainly appear in Chinese discourse in the form of topic chains (Wang, 2002; 2004). The research target are the topic chain, which is the main topic-prominent structure in Chinese discourse, and zero anaphora, which is the most common topic anaphora in the topic chain. Two important findings emerged from the present study. First, the characteristics of Chinese topic chains are transferrable to the interlanguage of Chinese EFL learners, thus resulting in overgeneralization of the zero anaphora. Second, the interlanguage discourse of Chinese EFL learners reflects a change of the second language acquisition process from topic-prominence to subject-prominence, thus lending support to the discourse transfer hypothesis.


Introduction
In terms of information structure in discourse, English and Chinese are typologically different languages. Chinese is a topic-prominent language in which the topic plays an important role in the formation of a sentence, whereas English belongs to the group of subject-prominent languages in which the subject is an indispensable element that determines the English sentence pattern (Li & Thompson, 1976). In second language acquisition research, quite a number of empirical studies have been conducted based on the language typological classification of subject-prominence (SP) and topic-prominence (TP) (Cai, 1998a(Cai, , 1998bGivón, 1983;Jung, 2004;Sasaki, 1990;Shi, 1989;Yang, 2008). However, previous studies usually explored the TP/SP issue from a syntactic perspective, ignoring the discourse function of the Chinese topic, as well as the influence of the key topic structure in Chinese discourse. Consequently, the topic chain in interlanguage discourse of Chinese learners has scarcely been investigated. A topic chain is "a set of clauses linked by a topic in the form of zero anaphora" (Chu, 1998, p. 324). A good example of a topic chain will be provided later in the paper.
The present study attempts to move beyond syntax to investigate TP in the interlanguage of Chinese EFL learners from a discourse perspective. The research target for this chapter is the topic chain, which is the main TP structure in Chinese discourse, and zero anaphora, which is the most common topic anaphora in the topic chain. Topic structures mainly appear in Chinese discourse in the form of topic chains (Wang, 2002(Wang, , 2004. Zero anaphora can be used in a discourse which contains at least two clauses, and this is the reason why we have chosen it as our research target. We hope that the findings of the present study will shed light on the change of Chinese EFL learners' interlanguage from TP to SP from a discourse perspective.

Topic and topic chain
The topic in a topic-prominent language is characterized by a continuity of the referent in discourse which represents the availability or identifiability of the referent for the speakers and listeners involved (Givón, 1983). The topic explicitly establishes a point of reference for the ensuing discourse for introducing new information, that is, information that is not recoverable from the preceding text.
A topic chain has always been considered as a self-contained unit, either on the syntactic level or on the discourse level. In previous studies, however, it was shown that this unit does not always correspond to the traditional notion of a sentence. In addition, the domain or scope of a topic chain can cross not only sentence boundaries but also paragraph boundaries. This seems to suggest that a topic chain can be a unit larger than a sentence or even a paragraph.
Topic chains are a common phenomenon in Chinese. Referring expressions that can be deduced contextually by the reader are frequently omitted in Chinese d i s c o u r s e ( Y e h & C h e n , 2 0 0 3 ) . T h e r e f e r r i n g e x p r e s s i o n s h a v e e n o u g h t o p i c continuity and thus enough cohesion for readers or listeners to find the particular stretch of discourse coherent. Tsao (1979) is believed to have been the first researcher to use the term topic chain. He stated that a topic chain is a stretch of actual discourse composed of one or more than one clause, headed by a topic which serves as a common link among all the clauses. It actually functions as "a discourse unit in Chinese" (p. vii; also see Tsao, 1990, p.63). Chu (1998) presented a more restricted view. He contended that, because topic is mainly a discourse notion, it can be identified only on the level of discourse where it serves as an inter-clausal link. Consequently, he defined a topic chain as "a set of clauses linked by a topic in the form of zero anaphora" (p. 324). To him, there is no point in talking about a topic within a single clause/sentence. According to Li (2005), a topic chain is a chain of clauses sharing an identical topic that occurs overtly once in one of the clauses. All the other clauses are linked t o t h e c h a i n b y z e r o N P s ( z e r o N P s a r e m a r k e d a s Ø in (1)) coreferential anaphorically or cataphorically with the topic. In the following example, the overt NP that car is the topic in the topic chain which contains six clauses: (1) Na liang che NP/ jiaqian tai gui, ØNP yanse ye buhao, wo bu that classifier car price too expensive color either not good I not xihuan ØNP, bu xiangmai ØNP. Zuotian qu kan le yixia ØNP like not want buy yesterday go look hai kai le yihuier ØNP, haishi bu xihuan ØNP.
also drive a while still not like 'That car is too expensive, and its color is not good. I don't like it and don't want to buy it. I went to see it yesterday and I drove it for a short time, but I still dislike it.' As we can see, six unspecified NPs are identified in (1). The clauses with the unspecified NPs and the one with the overt coreferential NP ('that car') are considered to form a chain with the overt NP being the topic of the chain. The topic is usually mentioned once at the beginning of a chain in the first clause, and the following chain of clauses shares one single topic.

Zero anaphora
Anaphora in Chinese can be classified into three categories, which are zero, pronominal and nominal forms respectively (Chen, 1987). The example in (2) is quoted from Chen (1987): (2) Tang Mingde jinghuang de wang wai pao, Ø zhuang dao yi ge dahan de Tang Mingde in panic out ran bumped into a big guy's Shenshang, Ta kan qing le naren de meiyan, Ø renchu naren shi shui. body he saw clearly that guy's eyes recognized that guy was who 'Tang Mingde ran out in panic and bumped into a big guy. He saw that guy's eyes clearly and recognized who that guy was.' In this example, Ø and ta co-refer with Tang Mingde, and na ren co-refers with yi ge da han. Ø, ta and na ren represent three kinds of anaphora in Chinese: a zero, a pronoun and a full noun phrase respectively. In Chinese discourse, the anaphora is frequently in the form of a zero morpheme, due to its prominence in discourse (Li & Thompson, 1981), which is termed zero anaphora (hereafter ZA). Zero anaphora is generally understood from the context and left unspecified. Zero anaphoras occur much more frequently in Chinese than in English. In English, even when the intended referent can be easily understood from the context, the presence of pronouns is still required to complete a clause in a discourse. In fact, the structural completion of a sentence is so important in English that the language has to resort to a suppositional pronoun it to fill in the slot of the grammatical subject. The ZA is such a common linguistic device in Mandarin that it may occur in almost any syntactic position in the sentence where a noun or a pronoun could appear. Li and Thompson (1981, p. 657) asserted that "a salient feature of Mandarin grammar is the fact that noun phrases that are understood from context do not need to be specified."

Language typology: Topic-prominent languages and subject-prominent languages
When the relation between topic and subject in languages is invoked at the typological level, topic-prominent (TP) and subject-prominent (SP) relations are the two basic relations in the structures of languages (Li & Thompson, 1976). Subject-prominent languages are those languages in which "the structure of sentences favors a description in which the grammatical relation of subjectpredicate plays a major role" (Li & Thompson, 1976, p. 459). By contrast, TP languages are those in which "the basic structure of sentences favors a description in which the grammatical relation of topic-comment plays a major role" (Li & Thompson, 1976, p. 459). It was Li & Thompson (1976) who first categorized world languages into four basic types: languages that are SP, languages that are TP, languages that are both SP and TP, and languages that are neither SP nor TP . According to this typological classification, Chinese is a TP language and English belongs to SP languages. The typological analysis above accordingly provides guidelines for identifying characteristic patterns in the study of any language and for analysis of interlanguage in terms of the acquisition of these patterns.

Typological transfer and discourse transfer
Topic-prominent Chinese and SP English, as two typologically salient languages, share some similarities but retain more differences. Consequently, when a Chinese learner of English makes an attempt to acquire SP English, both his previous knowledge of L1 and his present knowledge of L2 are adopted to decide whether TP features are available in English. Research has shown two contradictory claims about the role of topic/subject prominence typology in L2 acquisition. One claim is that irrespective of the learners' L1, the process of L2 acquisition is characterized by an early universal TP stage and that the typology of topic/subject prominence is not transferable. The opposing view on the role of topic/subject prominence typology in L2 acquisition is that the learners' L1 plays a role in their L2 learning and that as their L2 proficiency increases, native speakers of TP languages gradually increase the use of SP features in their L2 production.
Discourse transfer refers to the use of some of the discourse patterns of the learner's L2 in the same way in which they are employed in the learner's L1 (see Kasper & Schmidt, 1996;Kellerman, 1995;Wu, 2001). Discourse transfer is operationally defined in the present study as the transfer that happens when the language learner transfers L1-based discourse patterns to the L2 context. Discourse transfer studies focus on an L1 discourse strategy that is negatively transferred to L2 contexts, and demonstrates how learners transfer the L1 criteria of discourse processing. Along this line, discourse transfer studies look at the structure of L2 learners' output to see how it is organized. Some previous studies (Kasper, 1992;Kellerman, 1983) view discourse transfer as a cognitive activity in that it reflects the selection of some discourse patterns in the learner's L2 to be used in the same way in which they are employed in the learner's L1, and focus on the learner's cognitive contribution in selecting and producing L1-based discourse patterns in L2 contexts. The cognitive view emphasizes the relationship between L2 input, learner internal processing, and learner output in order to discover how the existing knowledge of the L1 influences the acquisition of L2. Following this research tradition, Bartelt (1992) states that "discourse transfer is a rule-governed cognitive process" (p. 113), in which the known rules of the native language are used as hypotheses in mastering the L2. As a TP language, Chinese is characterized basically as a highly context-dependent language, which stresses semantic coherence rather than formal cohesion, whereas English is an SP language (Wang, Hsu, & Chen, 1998). Chinese is called a discourse-oriented language (Huang, 1984(Huang, , 1989Shi, 1989) with a rule of topic NP deletion, which operates across discourse to delete the topic of a sentence under identity with a topic in a preceding sentence. The present study will focus on the discourse patterns of the learners' L2 to see how they are organized and try to explore whether Chinese EFL learners will actively select and produce L1-based discourse patterns in L2 contexts due to the difference between Chinese and English discourse mentioned above.

Previous studies on topic prominence in interlanguage
The introduction of the notion of TP by Schachter and Rutherford (1979) and Rutherford (1983) into the research field of SLA has triggered quite a number of studies concerned with this issue. Fuller and Gundel's (1987) study suggests that TP/SP is not a transferable typology and that L2 learners of different language backgrounds may go through a similar stage of universal TP. The reason why Fuller and Gundel obtain such results might be the students recruited for their study. Perhaps they were so advanced that any L1 effects would wash out. In contrast, our cross-sectional study may offer counterevidence against such a washout, since the low proficiency students do show what we would expect if transfer is at work.
However, in recent years it is commonly accepted that learners' L1 plays a role in their L2 learning. In his longitudinal study of a Hmong (a TP language) adult learner of English, Heubner (1983) found that the learner's interlanguage was found to progress from the initial TP to the SP stage through morphological syntactization. Rutherford (1983) detected evidence of overproduction of TP structures by TP speakers, especially Chinese speakers whose language has typical TP features. Jin (1994) found that when learning Chinese, the English learners go through a process of systematically transferring English SP features to Chinese until they reach a requisite proficiency when the concept of topic emerges, thus supporting Rutherford's (1983) claim on typological transfer. One of the present authors (Yang, 2008) showed that Chinese students at the preliminary level are strongly influenced by their native language and transfer their Chinese form/function to English directly. However, there is a general tendency that the frequency of using TP features in the interlanguage decreases and that the use of SP structures increases with higher English proficiency levels.
In sum, two conflicting claims have emerged from the studies reviewed above. One is that the process of L2 acquisition is characterized by an early universal topic-comment stage, independent of a learner's native language; the other is that the early TP stage is evidence of typological transfer from L1 to L2. Such conflicting conclusions may arise from the fact that the studies above were based only on the learning of English, which is an SP language.
Although the previous findings lent support to typological transfer by examining interlanguage, the role of topic/subject prominence typology in L2 acquisition is still not clear, as these studies have explored this issue in a purely syntactic way. The exact status of typology in L2 acquisition cannot be fully accounted for until more studies are carried out to investigate SP interlanguages of TP L1 learners. It remains to be seen whether previous findings can be substantiated from a discourse perspective. It is not clear whether some L2 sentence-level discourse features will show similarity to the learner's L1, that is, whether language learners will transfer L1-based discourse patterns to the L2 context. Accordingly, the present study will look at the structure of L2 learners' output to see how it is organized. Consequently, the present study attempts to provide additional cross-linguistic examination by looking at how learners perform on a translation task, with the belief that the findings obtained from English learners of Chinese may represent a useful contribution, enabling us to better understand the role of typology in L2 learning.

Research questions and hypotheses
The study reported here attempts to determine whether the TP features in Chinese discourse exist in the interlanguage of Chinese learners of English. In order to deepen our understanding of TP features of Chinese learners of English at different proficiency levels, we pose the following research questions: 1. Do learners use fewer TP structures and more SP structures at higher proficiency levels than at lower ones? 2. Are zero anaphoras transferable in connected written discourse?
Based on the evidence of some previous studies (Cai, 1998a(Cai, , 1998bGivón, 1983;Jin, 1994;Sasaki, 1990;Shi, 1989;Yang, 2008;Yip, 1995), we put forward the following two hypotheses: 1. The characteristics of the Chinese topic chain will be transferred to the interlanguage of Chinese EFL learners, thus resulting in the overgeneralization of ZAs. 2. The development of interlanguage discourse of Chinese EFL learners undergoes the process of discourse transfer.

Participants
The participants of this study were 90 Chinese students of English in China. Group 1 consisted of 30 high school first year students (15-16 years old), Group 2 consisted of 30 first-year undergraduate English majors (19-20 years old) and Group 3 consisted of 30 first-year postgraduate English majors (24-26 years old). Groups 2 and 3 included students educated at the same university. Since there was a 3-or 4-year difference in the length of formal schooling between each of the groups, it was assumed that they represented three proficiency levels: beginning (A2 of CEFR 1 ), intermediate (B1) and beginning advanced (B2). Of course, the dividing lines are by no means absolute.

Instrument and procedure
A task was designed to compare the English of Chinese university students with English passages from textbooks, with the passages being translated into Chinese as the source texts for the task given to the students. The task was a translation of two pieces of Chinese discourse into English, each of which consisted of two paragraphs that were characterized by typical Chinese TP features, namely, topic chains and ZAs. These two pieces of Chinese discourse were translated from two pieces of English discourse, which were excerpted from primary school textbooks Go for It (Book 3; Nunan, 2002) and Essential English for Foreign Students (Book 1; Eckersley, 1955). This task was designed to elicit the learners' production of anaphoras. There were 6 topic chains in which 26 anaphoras including zero, pronominal and nominal anaphoras were identified in the two pieces of Chinese discourse. We can refer to Example (1) in 2.2, which is the second piece of Chinese discourse, for more information.
That piece of discourse is made up of two topic chains. In the first topic chain, which contains 6 ZAs, the overt NP na liang che ('that car') is the topic. The difficult English words were provided to the participants in order to 1 CEFR = Common European Framework of Reference for Languages ensure that these words would not prevent them from completing the task successfully. After reading the instructions and making sure they understood them, the participants proceeded with the task. A maximum of 20 min was allowed for this task. It is possible that some students had used the books as course texts, but it is assumed they did not remember all the details. And even if they did remember some, the results suggest that less proficient students were much more influenced by ZAs than the more proficient ones. Data for this investigation were collected in the participants' regular classes. All the participants were asked to perform the task independently.

Data analysis
Raw data in the task were classified, calculated and tabulated into various tables for data analysis. The frequencies of various types of anaphora employed in the task by the participants were counted. The following provides an example of data analysis for the second piece of Chinese discourse: Data analysis was conducted to examine the number and proportion of ZAs used by the three participant groups when performing the translation task. It focused on the inappropriate use of ZAs in discourse. For the translation production, every piece of discourse produced by the participants was compared with its original version in the textbook to identify the ZAs produced by the participants, and the total number of wrong ZAs was computed. For example, in the sample above a participant at the beginning level (Group 1) employed 5 ZAs in the translation of the second piece of Chinese discourse where pronominal anaphoras were necessary. However, a participant at the intermediate level (Group 2) employed 2 ZAs wrongly and 3 pronominal anaphoras properly, and a participant at the advanced level (Group 3) made no mistake in employing anaphoras. In the analysis, the results of the three groups were compared in order to reveal whether there were comparable tendencies among beginning, intermediate and advanced learners in the production of TP constructions. With respect to data analysis, descriptive statistics such as percentages were employed to describe relative proportions and tendencies, and at times the data were assessed with the help of the SPSS package. A one-way ANOVA was employed to measure significant differences between each level.

Distributions of different types of anaphora in the test among the three groups
The distribution of types of anaphora used by the three participant groups is graphically represented in Figure 1. The figure shows that Groups 1 and 2 used the ZA more extensively than the Group 3, who in turn used the pronominal and nominal anaphora more frequently.  participants at lower levels demonstrated the tendency to produce more ZAs in the test. As the learners' L2 proficiency increased, such a tendency gradually became weaker. Although all the participants had learned English for 3 up to 10 years, they still displayed a tendency to rely on Chinese TP structures.

Multiple comparisons between proficiency levels
A one way ANOVA was conducted to measure differences between levels and their level of significance. The results are given in Table 1, which demonstrates significant difference in the use of ZAs across the three proficiency groups (F = 57.466, p = .000). Since the ANOVA only shows that at least one mean is different from all the other means, to be specific, a further LSD post hoc analysis was undertaken and the result of multiple comparisons between proficiency levels is shown in Table 2. It shows that the difference between every two proficiency levels was significant (p < .05), with the level of significance being especially meaningful with respect to the difference between Group 1 and Group 3 (p < .01). At the beginning level, many ZAs were produced by all participants. At the intermediate level, relatively fewer ZAs were produced by the participants than at the beginning level, but the percentage is also very high: up to about 50%. Generally speaking, the ZAs produced by advanced participants were obviously less numerous than those of the other two groups. The proportion of such structures decreased by almost one-third in comparison to the intermediate level, demonstrating that the participants were less influenced by their native language when using the L2.

Research question 1
To summarize the results of the current study, the beginning learners showed predominant use of TP features, and the use of such features gradually decreased, in accordance with the learners' L2 proficiency, across the group of learners as a whole. The results of the test have demonstrated a general shift from the use of zero to pronominal to nominal anaphoras in the interlanguage of Chinese learners at the discourse level as their proficiency level increased. The results show that the interlanguage discourse of the Chinese students of the three proficiency levels is characterized by TP constructions. Even the advanced learners are still subject to the influence of the Chinese TP feature. It clearly suggests that L1 transfer effect in the form of the TP feature of interlanguage production prevails over a relatively long period of time for learners with a TP L1 background such as the Chinese students in this study.
The present findings do not support the findings of Fuller and Gundel's (1987) study regarding an early universal stage of TP in L2 acquisition, but rather support the argument put forward by others that learners transfer their L1 features to L2 learning and that, with growing L2 proficiency, learners gradually become sensitive to the characteristics of the target language, approximating its norms (Cai, 1998a(Cai, , 1998bGivón, 1983;Jin, 1994;Sasaki, 1990;Shi, 1989;Yang, 2008;Yip, 1995). Schachter (1983) argues that "it is not lik ely to be the case that the process of L2 acquisition will prove to be the same process as in the first language acquisition case" (p. 256). She points out that one obvious source of difference is the learner's L1 and the facilitating and interfering effects it produces: "It is clear to anyone who takes the trouble to look at the production of a non-native speaker of a language that there is a first language, or transfer effect" (Schachter, 1983, p. 105). Actually the reliance on the knowledge of the native language, especially the native-language-related typological differences, is a universal learning strategy, which is obvious in the process of L2 acquisition.

Research question 2
In discussing discourse transfer, Odlin (1989) dealt with it in terms of structural factors and nonstructural factors. This is because discourse involves a wide array of nonstructural as well as structural characteristics. If many TP structures are present in the text, the impression conveyed to the native-speaker reader of English will be of disjointed and fragmented development of the messages embodied in the discourse.
The non-native-like English discourse produced by Chinese learners could be related to the absence of adequate cohesive ties. This is not surprising if, in the deep structure, the learner conceives his TL (English) surface coding in a TP manner. In a TP native language, a topic stands at the head of the discourse and the ensuing discourse is taken by the producer and the receiver to be related to it until a new topical referent is introduced. Subject-prominent languages such as English, however, undergo a complicated process of textualization to achieve a tightly-spun surface structure.
As discussed earlier, discourse transfer is the use of some of the discourse patterns in the learner's L2 in the same way in which they are employed in the learner's L1. It then follows that the learners are likely to negatively transfer the norms from their L1 to their L2 production largely due to their knowledge gaps about the L2 discourse norms. First, if learners are unaware of the L2 norms in particular contexts, they will be unable to produce the equivalent grammatical forms; in other words, a mismatch arises between their grammatical knowledge and the demands of particular discourse contexts. Therefore, they may resort to the s i m p l e r g r a m m a t i c a l f o r m s t h e y a r e m o r e c o n f i d e n t a b o u t b u t m a y n o t b e grammatically appropriate. Second, learners may have acquired the grammatical structures of certain utterances in the discourse. However, they may not understand or be aware of their discursive meanings in different contexts. Even in the data of advanced learners, discourse transfer still occurs, as they have not yet acquired the equivalent L2 forms of discourse in some contexts. In order to play safe, they tend to fall back on L1 literal translation when their "linguistic resilience" falls short. Transferring their L1 discourse norms in the form of literal translation and direct form-meaning mapping appears to be a safe strategy to help them finish the task.

Conclusion
The present study has provided some preliminary evidence that in the process of Chinese learners' acquisittion of English, discourse structures may be influenced by L1-L2 typological differences. Some conclusions can be drawn from this study. First of all, TP properties in the interlanguage of Chinese students are evident. Second, ZAs occur pervasively in the interlanguage of Chinese students of English at different proficiency levels. These ZAs are likely to be motivated by native-language-related typological transfer and discourse transfer. Third, the interlanguage discourse of Chinese EFL learners also reflects a change of the L2 acquisition process from topic-prominence to subjectprominence. Han (2010) has conducted longitudinal studies on fossilization and a similar longitudinal study should be carried out to examine the developmental sequences of SP features in English L2 acquisition by speakers of TP languages. Furthermore, research needs to investigate the acquisition of a SP language as an L2 by comparing L2 learners from both TP and SP L1 backgrounds, which will enable us to better understand typological influence on L2 learning.