Abstract
This article presents a study investigating the effects and recoverability of L1 transfer at L2 syntax–semantics and syntax–discourse interfaces in English speakers’ L2 acquisition of Chinese daodi … wh-questions, which are considered approximate counterparts of wh-the-hell questions in English. English speakers are found to be able to acquire syntactic properties of daodi … wh-questions, but the syntactic structure of the question in their L2 Chinese is influenced by their L1 semantic and discourse constraints. Removing the L1 influences is found possible in some aspects, but recovering from the L1 transfer is persistently vulnerable at the syntax–discourse interface involving appropriate answers to Chinese daodi … wh … modal questions, where the L1 and the Chinese discourse conditions are found coexistent in their L2 Chinese. The developmental data of the L2 interfaces are discussed, and it is argued that the availability of positive evidence in the input plays an important role in L2 acquisition of interfaces, as in L2 acquisition of other linguistic aspects.
Introduction
In recent L2 acquisition research, there has been an increasing emphasis on L2 interfaces. For examples, Sorace and Filiaci (2006) formulated an Interface Hypothesis, which proposed that narrow syntax is completely acquirable in L2 acquisition, but interfaces between syntax and other cognitive domains lead to persistent problems in L2 acquisition. A distinction has been made between internal interfaces and external interfaces in L2 interface research (cf. Tsimpli & Sorace, 2006; White, 2009), with the former referring to interfaces between different modules within the computational system of grammar, such as syntax–semantics and syntax–morphology interfaces, and latter referring to interfaces between the linguistic system and grammar external components, such as syntax–discourse and syntax–pragmatic interfaces. In a study by Sorace and Serratrice (2009), it is predicted that for L2 acquisition, cross-linguistic influence takes place at internal interfaces but not at external interfaces and that structures at the syntax–discourse interface have the potential to become permanently optional in L2 grammars. As will be seen in this article, which reports on an empirical study of L2 acquisition of Chinese daodi … wh-questions by English speakers, both internal and external interfaces in L2 acquisition can be adversely influenced by learners’ L1 and the recovery from L1 transfer depends upon the availability of positive evidence in the input data that learners are exposed to.
English phrases such as what the hell, who on earth, what the dickens and so on are widely considered to belong to the same category and are generically called wh-the-hell phrases (cf. Chou, 2006, 2007; Dikken & Giannakidou, 2002; Huang & Ochi, 2004). These English phrases are generally translated into Chinese as daodi … wh-word … or jiujing … wh-word … (cf. A Chinese-English Dictionary, 1997; Concise English-Chinese Chinese-English Dictionary, 1992; Oxford Chinese Beginner’s Dictionary, 2001). As these two phrases behave the same in Chinese wh-questions, I will use daodi … wh … as a representative of the two in this article. 1 The English wh-the-hell and Chinese daodi … wh share some syntactic properties, but they are also different from each other semantically, pragmatically as well as syntactically.
Syntactic differences
Following Pesetsky’s (1987) significant work on wh-movement, Huang and Ochi (2004) 2 argue that the English wh-the-hell is continuous and synthetic, as shown in the contrast between (1a) and (1b), and it occurs as a constituent that is required to move to the specifier of complementizer phrase (CP), as in (1a). Even in multiple wh-questions, wh-the-hell is not allowed to stay in situ, as shown in (1c) and (1d). The hell cannot occur anywhere in a question without a wh-word, as in (1e).
(1) a. b. * c. Who the hell would buy what? d. * e. * Would
The Chinese daodi and its wh-associates, on the other hand, do not form a constituent; they are discontinuous and analytic with daodi and a wh-phrase staying in situ, as shown in (2a) and (2b).
(2) a. Ta He daodi will buy what “What the hell will he buy?” b. * Ta yao mai He will buy daodi what
Huang and Ochi (2004) further point out that like the wh-the-hell expression, daodi in Chinese must occur in a construction with a wh-phrase in its c-commanding domain. As we can see, daodi is not allowed in (3a) because there is no wh-phrase in the sentence. 3 Although there is a wh-phrase in (3c–e), the wh-phrase is not commanded by daodi, thus the ungrammaticality of the sentences. In contrast, the sentence in (3b) is grammatical because the wh-phrase is c-commanded by daodi.
(3) a. * Zhangsan Zhangsan daodi will buy book * ‘Zhangsan will buy books the hell.’ b. Zhangsan Zhangsan daodi will for who buy this book ‘Whom the hell would Zhangsan buy the book for?’ c. * Zhangsan yao gei Zhangsan will for who daodi buy this book d. * Zhangsan yao gei Zhangsan will for who buy daodi zhe ben shu e. * Zhangsan yao gei Zhangsan will for who buy this book daodi
As we can see in (2a, b) and (3b–e), daodi must occur in a preverbal position. In addition, it is widely recognized (cf. Chou, 2006, 2007; Huang & Ochi, 2004; Kuo, 1997) that like the wh-the-hell in English, daodi … wh is required to occur in the scope of an interrogative CP. We can use the verb xiang zhidao ‘wonder, want to know’ to demonstrate that daodi must occur in a clause with a CP having the [+Question] ([+Q]) feature. As we can see, daodi can occur in the embedded clause as in (4a) because the CP of the embedded clause has a [+Q] feature. However, in (4b), daodi is not in the scope of a CP with [+Q], leading to the ungrammaticality of the sentence.
(4) a. Wo xiangzhidao Zhangsan I wonder Zhangsan daodi will buy what ‘I wonder what the hell Zhangsan would buy.’ b. * Wo I daodi wonder Zhangsan will buy what * ‘I the hell wonder what Zhangsan would buy.’
Differences in discourse and semantics
At the discourse level, the-hell in English differs from daodi in Chinese in that while the former cannot be used with the discourse (D)-linked wh-phrase like which, as shown in (5b), the latter can, as shown in (6b). Pesetsky (1987), who labels the hell as ‘aggressively non-D-linked’ and which as ‘aggressively D-linked’, argues that it is the conflict between the aggressively D-linked which and aggressively non-D-linked the hell that makes sentences like (5b) ungrammatical. Unlike the-hell in English, daodi in Chinese can be D-linked, as well as non-D-linked, and is compatible with the aggressively D-linked na “which”. This can be clearly seen in the Chinese examples in (6b), which form a striking contrast with the English the-hell in (5b), which is aggressively non-D-linked.
(5) a. What the hell book did you read that in? b. * Which the hell book did you read that in? (=(40) in Pesetsky, 1987) (6) a. Ni daodi du-le shenme shu? You daodi read-PERF what book ‘What the hell book did you read?’ b. Ni daodi du-le na ben shu? You daodi read-PERF which CL book * ‘Which the hell book did you read?’
Not only is the-hell aggressively non-D-linked, it also converts a wh-question with a modal into a negative rhetorical question. Dikken and Giannakidou (2002) use the sentences in (7) as an illustration. Unlike the question in (7a), where the regular wh-phrase who is used, the question with who the hell in (7b) cannot be interpreted as a genuine-information question. Rather, it can only be read as requiring a negative answer compatible with the negative rhetorical answer like Nobody would buy that book. 4
(7) a. Who would buy that book? b. Who the hell would buy that book?(=(2) in Dikken and Giannakidou, 2002)
However, Chinese daodi … wh-questions with modals can be naturally interpreted as genuine-information questions and can be felicitously answered by providing genuine information, as shown in (8). This provides us with evidence that the daodi … wh-question is a genuine question and can elicit an answer of genuine information. 5
(8) Q: Daodi shei hui bangzhu ta? Daodi who will help him ‘Who the hell would help him?’ A: Zhang Laoshi hui. ‘Teacher Zhang would.’
Another property of wh-the-hell that Dikken and Giannakidou (2002) find is that regular wh-phrases are always presuppositional and are linked to discourse-familiar values, but the wh-the-hell differs in this respect; it is not presuppositional. This can be seen in (9b) (= (35) in Dikken and Giannakidou, 2002), where who the hell cannot be linked to Someone in spite of the fact that a piece of discourse is created that forces an interpretation of who the hell to be anaphoric to a previously introduced discourse referent, that is, Someone. This is because the domain of who-the-hell is an open set, or the entire domain of the universe, including all persons in the universe and cannot be a presupposed subset of it and cannot be bound by a default existential referent in the discourse. On the other hand, who in (9a) can be linked to the discourse-familiar Someone. Here, the use of who presupposes that a specific person exists and is yet to be identified.
(9) a. Someonei has stolen that book. John knows whoi b. Someonei has stolen that book. *John knows whoi the hell.
In contrast with the unacceptable English sentence in (9b), the Chinese counterpart is completely acceptable as in (10), where daodi and its associate shei ‘who’ are linked to the discourse-familiar referent, that is, youren ‘someone’, which demonstrates that the Chinese wh-word associated with daodi can be presuppositional and anaphoric to a referent previously mentioned in the discourse.
(10) Youreni tou-le naben shu. Zhangsan zhidao daodi shi sheii. Someone steal-PERF that book Zhangsan know daodi is who. ‘Someone has stolen that book. *Zhangsan knows who the hell that person is.’
Furthermore, the English phrase wh-the-hell is incompatible with veridical verbs. Factive verbs such as know, tell, confirm and realize are veridical, which carry a presupposition of truth in the proposition of their complement clauses. As wh-the-hell is not presuppositional, it is not licensed in the complements of these verbs, as shown in (11b). However, negation of these veridical verbs can license wh-the-hell in the embedded question of the veridical verbs, as in (12). 6 Dikken and Giannakidou (2002) argue that the insertion of negation provides the required non-veridical c-commanding element which wh-the-hell depends upon for its legitimate occurrence in the complement clauses.
(11) a. John knew (told me/confirmed/realised) who would buy that book. b. * John knew (told me/confirmed/realised) who the hell would buy that book. (12) a. John didn’t know (tell me/confirm/realise) who would buy that book. b. John didn’t know (tell me/confirm/realise) who the hell would buy that book.
In contrast to the English wh-the-hell, the Chinese daodi … wh possesses the properties of presupposition, and it can occur in complement clauses of veridical verbs such know, tell, confirm and realize in Chinese, as in (13), where daodi … wh can occur in the complement clause of affirmative forms of veridical verbs, as in (13a), as well as in the complement clause of their negative forms, as in (13b). As we can see, the veridical verbs’ presupposition of truth in the proposition of their complement clauses does not prevent daodi … wh from occurring in their complement clauses. This is believed to be due to the fact that daodi … wh can be presuppositional, unlike wh-the-hell in English.
(13) a. Zhangsan zhidao (gaosu-le wo/zhengshi-le/yishidao) daodi shei hui mai Zhangsan know (tell-Perf me/confirm-Perf/realise) daodi who will buy na ben shu. that CL book * ‘John knew (told me/confirmed/realised) who the hell would buy that book.’ b. Zhangsan bu zhidao (mei gaosu wo/zhengshi/yishidao) daodi shei hui mai Zhangsan not know (not tell me/confirm/realise) daod who will buy na ben shu. that CL book ‘John didn’t know (tell me/confirm/realise) who would buy that book.’
The structure of the daodi … wh-question in Chinese
A number of researchers (Cinque, 1999; Rizzi, 1997) have proposed that information relevant to the interface between syntax and pragmatics can be encoded syntactically. Speas (2004), Tenny (2006) and Tenny and Speas (2003) further propose that pragmatic features can not only be encoded syntactically but can also affect syntactic computation. In this article, I follow Huang and Ochi (2004) in treating daodi as an adverb. In addition, I assume that daodi left-adjoins to an inflectional phrase (IP) in which there must be a wh-word. 7 The adverb daodi and its wh-associate form a chain with uninterpretable [+Q] and [+wh] features, which are to be checked by some elements with interpretable [+Q] and [+wh] features in CP. This requires that daodi must have a wh-word in its c-commanding domain and daodi … wh must be c-commanded by a CP with a [+Q] feature.
In English, we can assume that the entire phrase wh-the-hell is first merged in its base-generated position, then raises to the Specifier of CP to check the [+wh/+Q] feature. 8 Table 1 provides a summary of the contrasts and comparisons between the-hell and daodi in English and Chinese questions.
Features and properties of the-hell and daodi in English and Chinese questions.
CP: complementizer phrase; D: discourse.
Cross-linguistic influence at L2 interfaces
It is well documented in the L2 literature that cross-linguistic influence plays an important role in L2 acquisition and that L2 learners are found to transfer language phenomena from their L1 to the target language. Earlier generative studies of L1 transfer mainly focus on syntactic structures in L2 acquisition; some researchers argue for the transfer of the entire L1 grammar in L2 acquisition (e.g. the Full Transfer/Full Access Hypothesis proposed by Schwartz & Sprouse, 1994, 1996), while others believe that only parts of L1 syntactic structures are transferred into L2 grammars (e.g. the Minimal Trees Hypothesis by Vainikka and Young-Scholten, 1996a, 1996b, and the Valueless Features Hypothesis by Eubank, 1994a, 1994b). In recent years, an increasing number of researchers have shifted their attention to interfaces in L2 acquisition, and the Interface Hypothesis formulated by Sorace and Filiaci (2006) proposes that narrow syntactic properties in L2 acquisition are completely acquirable, but grammatical aspects that involve an interface between the syntactic domain and other cognitive domains may not be fully acquirable in L2 acquisition. 9 As L2 research becomes more nuanced and sophisticated, a distinction is made between internal interfaces and external interfaces, with the former including syntax–morphology, syntax–lexicon and syntax–semantics interfaces, and the latter including syntax–discourse and syntax–pragmatics interfaces (see White, 2009), and it is recognized that not all interfaces are equally problematic. For example, it is proposed in the study by Tsimpli and Sorace (2006) that the syntax–semantics interface, an internal interface, causes fewer problems in L2 than the syntax–discourse interface, an external interface.
In a study by Sorace and Serratrice (2009), it is predicted that for L2 acquisition, cross-linguistic influence takes place at internal interfaces but not at external interfaces and that structures at the syntax–discourse interface have the potential to become permanently optional in L2 grammars. They propose that ‘in the case of syntactic choices that are governed by grammar-internal semantic features, the role played by the language-specific setting of the parameter played a determining role’ (p. 206), and for L2 acquisition, they ‘predict potentially permanent optionality with respect to structures at the syntax–discourse interface, but not with respect to those at the syntax–semantics interface, regardless of language combination’ (p. 207). However, evidence has been reported in recent L2 literature that cross-linguistic influence takes place at both internal and external interfaces in L2 acquisition.
Belletti, Bennati, and Sorace (2007) conducted a study focusing on the production and interpretation of post-verbal subjects, and null and overt pronominal subjects, by near-native speakers of Italian whose native language was English. The results of the study show that while the near-native speakers were similar to the native Italian controls in their production of null subjects, their use of overt pronominal subjects was significantly higher than the controls. It was also found that overt subject pronouns in embedded clauses were interpreted as coreferential with the matrix subject at a significantly higher rate by the near-native speakers than by the native controls. Moreover, post-verbal subjects were underused across verb classes by the near-native speakers, in contrast with the native controls. Based on these findings, Belletti et al. argue that the near-native speakers had acquired the crucial properties related to the null subject in Italian, but they optionally referred to grammatical and discourse-related features of their L1 English. The implications of these findings are that L1 transfer of both syntactic and discourse-related features take place in the English speakers’ L2 Italian and that the cross-linguistic influence is persistent even at the near-native level.
To compare the effects of L1 transfer on pure syntactic versus interface phenomena, Montrul (2010) examined whether the influence of L2 learners’ L1 is similar to that of heritage speakers’ dominant language on their acquisition of Spanish. Both the L2 learners’ L1 and the heritage speakers’ dominant language in Montrul’s study were English, and she focused on three linguistic phenomena in Spanish: the placement of clitics, clitic doubling (= clitic left dislocation) and differential object marking. All these three phenomena are absent in English. The results of Montrul’s study indicate that L2 learners and heritage speakers were as accurate as native Spanish speakers on the purely syntactic phenomenon of the clitic placement, but in their judgment of object topicalization with and without clitic doubling, both groups exhibited transfer from English at the syntax–pragmatics interface. Furthermore, both groups were inaccurate with differential object marking, which involves the interface between syntax and semantics. Montrul’s findings suggest that both the internal interface (i.e. the syntax–semantics interface involved in the differential object marking) and the external interface (i.e. the syntax–pragmatics interface involved in the clitic doubling) are vulnerable to L1 transfer.
Similar findings are also reported in Sánchez, Camacho, and Ulloa (2010), who investigated Shipibo speakers’ acquisition of the first-person subject pronoun in their L2 Spanish. As mentioned above, both null and overt subjects are allowed in Spanish, but their distributions are governed by discourse constraints. The results of the study show that Shipibo speakers were able to use null subjects in tensed clauses in their L2 Spanish and that they rarely produced illegal overt pronominal subjects with non-tensed clauses. They were also found to be able to acquire the morphological subject-verb agreement paradigm in the target language. However, they had persistent problems with the discourse constraints on the use of null and overt subjects and, in particular, with the distribution of first-person pronouns. This suggests that grammatically encapsulated properties are readily acquirable in L2, but discourse constraints on the distributions of null and overt pronouns are vulnerable to L1 transfer.
Not only transfer of L1 discourse/pragmatic constraints takes place at L2 interfaces, transfer of L1 syntactic structures is also found at L2 interfaces. Yuan (2010) carried out an investigation of the role of the semantics-syntax interface in the representation of wh-words as existential polarity words (EPWs) in L2 Chinese grammars of English and Japanese speakers. It was found that Japanese speakers and English speakers were significantly divergent in that the wh-word is licensed by the yes–no particle ma in L2 Chinese grammars of the former but not the latter. Unlike English, Japanese uses question-marking particles to mark its yes–no questions, and Yuan argues that the L1 transfer of this Japanese syntactic structure can help Japanese speakers identify the Chinese yes–no particle as a licenser for the EPW in their L2 Chinese grammars. Although English speakers have little difficulty in accepting simple Chinese yes–no questions with the particle ma but without any wh-words, they have persistent problems in syntactically wiring between the yes–no particle ma and the wh-word as an EPW in their L2 Chinese yes–no questions.
Problems with the underlying syntactic structure at L2 interfaces were also found in a study by Belletti and Leonini (2004), who examined the L2 Italian syntax–pragmatics interface involving the Italian verb–subject (VS) word order. Most of the L2 Italian learners involved were German speakers. Recall that in Italian, the subject in post-verbal position constitutes the focus of new information, and Belletti and Leonini report that although there was no association between the VS word order and focus in the learners’ L2 Italian grammars, the learners were found to be sensitive to the pragmatic content of the VS word order in Italian because they relied on their L1 transfer and adopted equivalent L1 strategies, such as stress on the focused element to express focus in their L2 Italian. Belletti and Leonini maintain that the L2 Italian learners are able to appreciate the discourse context involved in the VS order but have difficulties in consistently projecting the Focus Projection, required in the L2 representation but absent in the L1. In other words, Belletti and Leonini argue that the problem the L2 Italian learners have with the interface involving the VS word order is grammatical in nature.
The studies reviewed above suggest that both internal and external interfaces are vulnerable to L1 transfer and that the distinction between the internal and external interfaces (see Tsimpli & Sorace, 2006; White, 2009) does not help explain cross-linguistic influence in L2 interfaces. However, it is not clear from the studies above whether L2 learners are able to recover from L1-related grammars and what can help expunge the L1 influence from their L2 grammars. These questions will be taken into consideration when we discuss the findings of our study.
Empirical study
To examine English speakers’ L2 Chinese daodi … wh-questions, an empirical study was conducted with a focus on the syntax–semantics and syntax–discourse interfaces involved in L2 Chinese questions of this type. We asked the following research questions:
Syntactically, is it possible for English speakers to establish the requirements in their L2 Chinese daodi … wh-questions that daodi … wh is c-commanded by a CP[+Q], the wh-word is required in the c-commanding domain of daodi and the discontinuous form daodi … wh is used?
At the syntax–semantics interface, is daodi … wh allowed in the complement clause of the affirmative form of veridical verbs in English speakers’ L2 Chinese, in contrast with wh-the-hell in their L1 English?
At the syntax–discourse interface, (a) do English speakers transfer the aggressively non-D-linked property of the-hell from their L1 to daodi, which is not aggressively non-D-linked in the target language Chinese? (b) can daodi … wh be linked to a discourse-familiar referent in English speakers’ L2 Chinese, unlike its counterpart wh-the-hell in their L1 English? and (c) will English speakers be able to provide genuine information to the Chinese daodi … wh … modal question, instead of providing a negative rhetoric answer, as in the case in the wh-the-hell … modal question in their L1 English?
If any L1 transfer occurs, will English speakers’ L2 Chinese grammars be able to recover from the cross-linguistic influence and acquire the native-like competence in the aspects concerned?
Instruments and procedures
A cloze test was administered to all subjects involved to establish their Chinese proficiency levels. The subjects also had to do three other tasks, an acceptability judgment task, a sentence-combination task and a discourse-completion task.
Acceptability judgment task
The acceptability judgment task included, among other things, 12 sentence types related to the research questions above, and each type had four tokens. So in total, there were 48 test sentences concerning the focus of the article. 10 All sentences were presented in Chinese characters, but instructions were given to the English speakers in English and to native speakers of Chinese in Chinese. The subject was asked to judge each test sentence by circling a number on a Likert scale, as shown in (14), which was presented below each test sentence.
In order to minimize any possible effect of vocabulary on the subjects’ judgment, efforts were made to include only those basic words of daily life. In addition, English translation and the Chinese pinyin (a Chinese phonetic system) were provided for some words potentially unfamiliar to the subjects. The 12 sentence types used in the test and their examples are listed in (15).
(15) Sentence types and their examples in the acceptability judgment test Type 1: The discontinuous form of daodi … wh-object Ta daodi xihuan shei? he daodi like who ‘Who the hell does he like?’ Type 2: * The continuous form of daodi wh-object * Ta xihuan daodi shei? he like daodi who ‘Who the hell does he like?’ Type 3: * The continuous form with the wh-object preceding daodi * Ta xihuan shei daodi? ta like who daodi ‘Who the hell does he like?’ Type 4: daodi…wh-subject Daodi shei xihuan ta? daodi who like her ‘Who the hell likes her?’ Type 5: * wh-subject daodi * Shei daodi xihuan ta? who daodi like her ‘Who the hell likes her?’ Type 6: daodi c-commanded by CP[+Q] Xiao Wang xiang zhidao ni daodi xihuan shei. Xiao Wang want know you daodi like who ‘Xiao Wang wonders who the hell you like.’ Type 7: * daodi c-commanded by CP[-Q] * Xiao Wang daodi xiang zhidao ni xihuan shei. Xiao Wang daodi want know you like who * ‘Xiao Wang the hell wonders who you like.’ Type 8: * daodi in yes-no questions (which have no wh-word) * Ni daodi xihuan Li Ying ma? you daodi like Li Ying y/n-Q * ‘Do you the hell like Li Ying?’ Type 9: Wh-questions with na ‘which’ Ni xihuan na ben shu? you like which CL book ‘Which book do you like?’ Type 10: Wh-questions with daodi and na “which” Ni daodi xihua na ben shu? you daodi like which CL book * ‘Which the hell book do you like?’ Type 11: Veridical verbs with daodi … wh Ta gaosu-le wo ta daodi xihuan shei. she tell-FTV me she daodi like who * ‘She told me who the hell she liked.’ Type 12: Negated veridical verbs with daodi … wh Ta mei gaosu wo ta daodi xihuan shei. she didn’t tell me she daodi like who ‘She didn’t tell me who the hell she liked.’
Test sentences in Types 1–3 were mainly used to examine whether daodi and the wh-word formed a discontinuous form or were incorrectly synthesized in the subjects’ L2 Chinese as in the case of wh-the-hell in their L1 English. As we can see in the examples of Types 1–3, the sentence in each type is exactly the same as its corresponding sentences in the other two types, except for one difference, that is, the formation of daodi and the wh-word. Test sentences in Types 2 and 3 were also expected to help examine whether English speakers’ L2 Chinese adopts the L1-based linear word order of wh-daodi if the continuous form is incorrectly transferred from their L1.
The possible discontinuous or continuous form of daodi and the wh-word can only be tested with the form in object position but not in subject position. This is because at the surface level in Chinese, daodi is always adjacent to the wh-subject even though in the underlying structure, daodi is separated from the wh-subject, with the former left-adjoining to IP and the latter merging into the Spec of IP. The purpose of including test sentences in Types 4 and 5 was to find out whether daodi … wh is acceptable in subject position (Type 4) in English speakers’ L2 Chinese and whether they would be able to reject Type 5 sentences, where the wh-word is not c-commanded by daodi.
Recall that both wh-the-hell in English and daodi … wh in Chinese have to be c-commanded by CP[+Q] and that they are used in wh-questions in both languages. Test sentences in Types 6 and 7 were used to examine this aspect of English speakers’ L2 Chinese grammars, and Type 8 was used to test whether daodi would be incorrectly allowed in a c-commanding domain of CP[+Q] without any wh-word, that is, in a yes–no question, in L2 Chinese.
In English, the-hell is aggressively non-D-linked, while which is aggressively D-linked. The example sentences in Types 9 and 10 form a minimal pair, which was designed to collect data concerning the possible discourse compatibility of daodi with na “which” in English speakers’ L2 Chinese. Similarly, wh-the-hell is not licensed in the complement clause of veridical verbs in English although this is possible in the complement of negated veridical verbs. However, no such semantic constraint exists in Chinese, and we used the paired sentences in Types 11 and 12 to collect data of possible L1 transfer of this semantic constraint in English speakers’ L2 Chinese.
Sentence-combination task
The sentence-combination task was used to investigate whether English speakers would be influenced by their L1 English and incorrectly reject Chinese sentences with daodi-wh linked to a referent in the discourse. In each question in the sentence-combination task, subjects were presented with two individual clauses (A) and (B) and were asked to indicate to what extent it is appropriate for Clause B to follow Clause A. They were told that both (A) and (B) are GRAMMATICALLY CORRECT and were asked to indicate the degree of appropriateness of the combination by circling one of the choices under each question, as shown in (16). There were four control questions and four experimental questions, and there were also four distracters. The control question and its corresponding experimental question were exactly the same except for one difference, that is, the latter used both daodi and a wh-word, but the former used only the wh-word.
(16) (A) Suiran wo baba zhi shuo ta gei wo mai-le yi ge liwu, although my father only say he for me buy-PFV one CL present ‘Although my father only said that he had bought a present for me,’ (B) keshi wo yijing caichulai ta (daodi) gei wo mai-le shenme. but I already figure out he (daodi) for me buy-PFV what ‘but I had already figured out what (the hell) he had bought for me.’ If Clause B follows Clause A here, it is (a) completely appropriate. (b) ok. (c) not so appropriate. (d) completely inappropriate. (e) I don’t understand the clauses.
Discourse-completion task
The aim of the discourse-completion task was to examine whether English speakers can overcome the L1 English transfer, if any, and correctly interpret the Chinese daodi … wh … modal question as a genuine-information question, rather than a negative rhetorical question. The subject was presented with daodi … wh … modal questions, and after each question, multiple answers were provided for the subjects to complete the discourse. Subjects were told that ‘all the questions and answers are GRAMMATICALLY CORRECT, but some answers are APPROPRIATE to the questions given and some ARE NOT APPROPRIATE’, and they were asked to indicate ‘which answer(s) is/are APPROPRIATE and which answer(s) is/are NOT APPROPRIATE to the question given’ by using ‘√’ for the former and ‘X’ for the latter. In (17), an example is provided of a question and the multiple answers used in the discourse-completion task. As we can see, the set of answers included an answer of genuine information, as in (17c); a negative reinforcement, as in (17a); as well as a response irrelevant to the question, as in (17b). 11 Subjects could choose ‘I don’t understand the question/answers’, as in (17d), if this was the case. There were four tokens of the type in the task, and there were also four distracters. Before subjects started the discourse-completion task, they were given three examples for practice. The purpose of this practice was to make sure that they mark ALL the answers in terms of appropriateness and that they feel free to mark more than one answer as appropriate or inappropriate.
(17) Daodi shei hui bangzhu ta? daodi who would help her ‘Who the hell would help her?’ a. Shide, meiyou ren hui bangzhu ta. right no person would help her ‘That’s right, no one would help her.’ b. Laoshimen ye changchang huxiang bangzhu. teachers also often each other help ‘The teachers also often help each other.’ c. Li Laoshi yiding hui bangzhu ta. Li Teacher definitely will help her ‘Teacher Li will definitely help her.’ d. I don’t understand the question/answers.
Selection of subjects suitable for the study
The empirical study involved 95 English speakers and 18 native speakers of Chinese as controls. The English-speaking subjects were undergraduate students, post-graduate students, lecturers and professors of Chinese from universities in the United Kingdom as well as students studying at universities in Beijing at the time of the data collection. The native Chinese speakers were university students or office workers in China. Conversations with some Chinese language teachers revealed that they were aware of the word-order difference between Chinese daodi … wh-questions and English wh-the-hell questions but not of the differences in the underlying syntax or at the syntax–semantics or syntax–discourse interface. None of the teachers interviewed said they had given any specific instruction on daodi … wh-questions in their classroom teaching. Among over a dozen of Chinese textbooks or grammar books the author checked, none of them gives any specific explanation on Chinese daodi … wh-questions.
In this study, we treat daodi … wh in Chinese as an approximate counterpart of wh-the-hell in English. 12 This is not only based on the proposal by Huang and Ochi (2004) and Huang (2010), and on a number of well-established dictionaries, (e.g. A Chinese-English Dictionary, 1997; Concise English-Chinese Chinese-English Dictionary, 1992; Oxford Chinese Beginner’s Dictionary, 2001), but it is also based on the author’s personal experience of teaching Chinese as a foreign/second language to English speakers. This treatment is supported by the translations of Chinese daodi … wh-questions into English by English-speaking learners of Chinese, which was made available to the author by Z. Mai (personal communication). 13 Some of the sentences in the translation task were daodi … wh-questions, and it was found that learners at high-intermediate and advanced levels translated Chinese daodi … wh-questions into English as wh-on-earth questions, which are considered to belong to the same category as the wh-the-hell question. Here are two examples of the English wh-on-earth questions translated from Chinese daodi … wh-questions: ‘Where on earth does Xiao Li need to go?’ and ‘What on earth is Xiao Li going to eat?’. However, it was also found in the translation task that beginner and low-intermediate learners were either unable to handle the Chinese daodi … wh-questions or translated daodi in the questions as ‘indeed’, ‘really’, ‘after all’, ‘once and for all’, ‘so’, ‘finally’, ‘in fact’ or ‘actually’. 14 This shows that daodi in Chinese daodi … wh-questions is not always interpreted as an approximate counterpart of the-hell by L2 Chinese learners. It is also reasonable to believe that if these interpretations occur, daodi in these learners’ L2 Chinese daodi … wh-questions will not be sensitive to the syntactic restrictions, let alone to the syntax–semantic and syntax–discourse conditions involved. In the absence of any interpretation data in our study, we used a syntactic criterion to identify subjects suitable for the present study. Recall that both daodi and the-hell are required to be c-commanded by CP[+Q]. However, none of the other interpretations found in the students’ translation task above is subject to this syntactic restriction. If a subject can consistently (i.e. four out of the four tokens) accept daodi c-commanded by CP[+Q] in Chinese daodi … wh-questions and at the same time also consistently (i.e. four out of the four tokens) reject daodi c-commanded by CP[−Q], 15 this will give us confidence in believing that daodi is treated as an approximate counterpart of the-hell, at least syntactically, in the subject’s L2 Chinese daodi … wh-questions.
We originally used subjects’ performance in the cloze test to divide the 95 subjects into five Chinese proficiency groups: beginner, post-beginner, intermediate, advanced and very advanced groups. However, based on our stringent selection criterion, we discarded the data of all 19 beginners and all 18 post-beginners as none of them met the criterion. Four subjects (out of 25) in the intermediate group, 16 (out of 22) in the advanced group and 10 (out of 11) in the very advanced group met the criterion. 16 For the sake of simplicity, we put the 4 subjects from the intermediate group in the advanced group in our analyses. Information about each of the groups thus identified is given in Table 2.
Information about each group.
: significantly different from the group of native speakers of Chinese (NS group) at p < 0.001.
A one-way analysis of variance (ANOVA) result shows that there is a significant difference between groups in their performance in the cloze test (F(2, 313) = 155.084, p < 0.001), and the post hoc Scheffé tests indicate that the advanced group, but not the very advanced group, is significantly different from the native Chinese group.
As we can see in Table 2, subjects in the very advanced group, in average, had a history of more than 14 years of studying Chinese, and their average stay in China or Taiwan was nearly 4 years. In fact, many of them were lecturers and professors teaching either Chinese or Chinese-related courses in their universities. Given this information, we consider subjects in this group very advanced learners of Chinese.
Results
Results of the acceptability judgment task
Recall that in the acceptability judgment task, the subject had to judge each test sentence by circling a number on a Likert scale from ‘−2’ to ‘+2’, as shown in (14). In our analyses of the data, we treat any score of ‘+1’ or above that the subject assigned to a particular sentence as a sign of accepting that sentence. Conversely any score of ‘−1’ or below as a sign of rejecting that sentence, and mean scores between ‘−1’ and ‘+1’ are treated as signs that subjects were indeterminate about the sentence. 17
From Table 3, we can see that the advanced and very advanced groups behaved native-like in accepting daodi c-commanded by CP[+Q] and rejecting daodi c-commanded by CP[−Q]. Native-like behaviours can also be observed in their rejection of daodi in the Chinese yes–no question, which has a CP[+Q] but does not have a wh-word (see the last column of Table 3). These findings suggest that daodi was syntactically treated as an approximate counterpart of the-hell in their L2 Chinese grammars because like the-hell in English, daodi must be c-commanded by CP[+Q] in Chinese, and neither of them is allowed to occur in a question without a wh-word (e.g. in a simple yes–no question).
Group mean scores in the judgment of ‘daodi’ c-commanded by CP[+/-Q] and ‘daodi’ in the yes–no question (which does not have a wh-word).
↔: significantly different between the two at p < 0.001.
Table 4 provides further evidence that subjects in the two L2 groups had acquired syntactic behaviours of daodi in Chinese wh-questions. Like native speakers, they correctly accepted the discontinuous form of daodi … wh and correctly rejected the continuous form, and they also made clear distinctions between the two. Significant differences were found in one-way ANOVA tests between the three types of sentences in every group’s judgment, 18 and post hoc Scheffé tests indicate that all the groups, without any exception, were able to make significant distinctions between the grammatical discontinuous form daodi … wh, on the one hand, and the ungrammatical continuous forms of daodi … wh and the incorrect order of wh-daodi, on the other, as shown by the arrows between sentence types in learner groups’ mean scores in Table 3. This suggests that in these English speakers’ L2 Chinese daodi … wh-questions, like the L1 Chinese grammar, daodi was not synthesized with the wh-word and it occurred in a position c-commanding the wh-word.
Group mean scores in the judgment of ‘daodi … wh’ questions with the wh-word in object or subject position.
: significantly different from the NS group at p < 0.001; “↔” = significantly different between the two.
Table 4 also shows that the native Chinese speakers did not reject ungrammatical Chinese wh-questions with daodi appearing at the end of the wh-question and preceded by the wh-object (mean = −0.21). This is rather unexpected. An explanation for this is that in colloquial Chinese, it is acceptable to use daodi at the end of the wh-question (with a pause or a comma before it). It is possible that some native Chinese speakers had this pattern in mind when they judged the incorrect wh-questions with daodi at the end of the wh-question. 19
Recall that there are minimal pairs in the acceptability judgment test and that the sentence in one type is exactly the same as its corresponding sentence in the other type except for one difference, that is, whether daodi precedes or follows the wh-word. As we can see in the last two columns of Table 4, the two L2 groups, like the native Chinese group, accepted the grammatical Chinese sentences with daodi … wh in subject position and rejected the ungrammatical sentences with wh-daodi in subject position. Results of paired-samples t-tests indicate that like the native Chinese group, the two L2 groups were able to make significant distinctions between the two types of sentences. 20
The data in Tables 3 and 4 provide us with evidence that the syntactic properties of daodi were established in these learners’ L2 Chinese daodi … wh-questions and that these learners were suitable as subjects for examining the syntax–semantics and syntax–discourse interfaces in English speakers’ L2 Chinese daodi…wh-questions.
Recall that in English, the-hell is aggressively non-D-Linked and cannot be used with the aggressively D-Linked wh-word which (see Pesetsky, 1987). However, such a conflict does not exist between their counterparts in Chinese. As we can see in Column 3 of Table 5, the native Chinese group readily accepted Chinese sentences with daodi and na ‘which’ coexisting in a question, and their mean score in judging this type of question is 1.99. Similar behaviours were also found in the judgment of the very advanced group, whose mean score is 1.98. Although the advanced group’s mean score (= 1.65) is well above the acceptance threshold of +1, results of paired-samples t-tests show that subjects in this group made a significant distinction between the which-control questions and the questions with daodi coexisting with na ‘which’. It is possible that daodi in these subjects’ L2 Chinese grammars was influenced by some L1-based non-D-linking property, which made the co-occurrence of daodi with the Chinese na ‘which’ in the wh-question less acceptable to the L2 learners. In spite of the evidence of possible L1 transfer, native-like behaviours of daodi can be acquired by English speakers as subjects in the very advanced group clearly accepted both the which-control questions and the questions with daodi coexisting with na ‘which’, and like the native Chinese, they did not make a distinction between the two.
Group mean scores in the judgment of presupposition and D-Linking of ‘daodi’.
D: discourse; *: significantly different from the NS group at p < 0.05; **: significantly different from the NS group at p < 0.01; ***: significantly different from the NS group at p < 0.001; ↔: significantly different between the two.
Clearer evidence of L1 transfer is observed in the advanced group’s judgment of Chinese sentences with daodi embedded in a complement clause of (negated) veridical verbs. Recall that in English, the-hell is incompatible with veridical verbs and cannot occur in a complement clause of affirmative forms of these verbs although it is compatible with negative forms of veridical verbs. The (in)compatibility seems to transfer from English speakers’ L1 into their L2 Chinese, and this can be clearly seen in the advanced group’s judgment of daodi in a complement clause of a veridical verb and in a complement clause of a negated veridical verb because this group accepted the latter but not the former. Results of paired-samples t-tests indicate that there is a significant difference between the group’s judgment of the former and the latter (t(79) = −8.267, p < 0.001). Such a difference disappears in the very advanced group’s judgment (t(39) = −1.098, p = 0.126), which suggests that with more exposure to Chinese input and with the improvement of their L2 Chinese proficiency, English speakers are able to expunge the semantic constraint transferred from their L1 English and acquire the native-like competence in this aspect of grammar. A significant difference was found in the native Chinese group’s judgment between the two types of sentences (t(71) = −3.377, p < 0.01). This is rather unexpected. However, given that the native Chinese group’s mean score (= 1.61) for daodi in a complement of a veridical verb is well above the acceptance threshold of +1, the difference is considered to indicate only a preference of sentences with daodi in a negated veridical verb rather than a rejection of sentences with daodi in a complement clause of a veridical verb. 21
Results of the sentence-combination task
Recall that in English, the domain of the wh-the-hell question is an open set or is the entire domain of the universe and cannot be a presupposed subset of it and cannot be bound by a default existential referent in the discourse. However, the Chinese daodi … wh-question behaves differently, and the wh-word in the question can be linked to a discourse-familiar referent. Table 6 presents data concerning the groups’ performance in the sentence-combination task involving linking daodi … wh to discourse referents. In our analysis of the data from the sentence-combination task, if subjects considered a sentence combination ‘completely appropriate’, we converted it to ‘+2’; if the combination was considered just ‘ok’, it was converted to ‘+1’; if ‘not so appropriate’, it was converted to ‘−1’ and if ‘completely inappropriate’, it was converted to ‘−2’. The statement of ‘I don’t understand the clauses’ was given the value of ‘0’. 22
Group mean scores of linking ‘daodi … wh … ’ to discourse-familiar referents in the sentence-combination task.
: significantly different from the NS group at p < 0.01; ↔: significantly different between the two.
As we can see in Table 6, the advanced and very advanced groups were native-like in accepting the control combinations. However, a different picture appears when daodi is used in the experimental combinations, in which daodi … wh refers to a familiar referent in a previous sentence in the discourse. In paired-samples t-tests comparing each group’s judgment of the control combinations (without daodi) with its judgment of the experimental combinations (with daodi), a significant difference was found between the control combinations and the experimental combinations in the advanced group. 23 A one-way ANOVA result also indicates that there is a significant difference between the groups in doing the experimental combinations (F(2, 7.622) = 5.442, p < 0.01), and the advanced group was found significantly different from the native Chinese group in post hoc Scheffé tests, in spite of the fact that the group behaved native like in doing the control combinations. The implication of this finding is that subjects in this group transferred the discourse constraint on wh-the-hell from their L1 English into their L2 Chinese and did not allow the daodi … wh to be linked to an existential referent in the discourse. However, no significant difference was found in the very advanced group between the control combinations and the experimental combinations. This suggests that in spite of the earlier occurrence of L1 transfer of the discourse constraint, English speakers are able to acquire the native-like competence, eventually allowing the daodi … wh to be bound by a referent in the discourse.
Results of the discourse-completion task
Recall that in English, the wh-question with wh-the-hell cannot be interpreted as a genuine-information question when used with a modal verb and that it can only be felicitously interpreted as a negative rhetorical question. However, Chinese daodi … wh-questions with modals cannot be interpreted as negative rhetorical questions and can be naturally answered by providing genuine information. What was used for statistic analyses in the study here was the frequency rates at which individual subjects chose a particular answer to answer the daodi … wh … modal question. The purpose was to find out whether any L1 transfer took place in this aspect and whether any difference existed between the native Chinese group and any learner group in interpreting Chinese daodi … wh … modal questions as genuine-information questions or as negative rhetoric questions. That is, the frequency rates of genuine-information answers and negative-reinforcement answers in subjects’ responses to the daodi … wh … modal questions were used as numerical data in the analyses. Recall that in the discourse-completion task, there were four tokens of the daodi … wh … modal question and that the subject could choose the genuine-information answer or the negative-reinforcement answer or both in response to each question. As a result, the maximum score a subject could get in providing each of the two types of responses was 4 and the minimum was 0. As we can see in Table 7, the two L2 groups readily interpreted daodi … wh … modal questions as genuine questions as they were as frequent as the native Chinese group in providing answers of genuine information to this type of question (4 out of 4). This suggests that Chinese daodi … wh … modal questions can be acquired as genuine questions by English speakers. However, the acquisition does not seem to imply the removal from English speakers’ L2 Chinese grammars of L1 transfer of interpreting this type of question as a negative rhetorical question. As shown in the last column of Table 7, the English speakers were rather optional in incorrectly interpreting the Chinese questions as negative rhetorical questions, and the average frequency rates of the advanced and very advanced groups’ incorrect interpretations were as high as 2.15 and 2.20 (out of 4). Results of a one-way ANOVA indicate that a significant difference exists in the groups’ interpretation of the questions as negative rhetorical questions (F(2, 22) = 19.998, p < 0.001), and significant differences were found in post hoc Scheffé tests between the native Chinese group and each of the two L2 groups. The implication of these findings is that the daodi … wh … modal question can be acquired by English speakers as a genuine question, but this is more likely to be in addition to, rather than as a replacement of, the question incorrectly interpreted as a negative rhetorical question in their L2 Chinese grammars. English speakers optionally interpret the questions as negative rhetoric questions, and their L2 Chinese grammars seem to have difficulty expunging this interpretation from their L2 Chinese grammars. This L1-related constraint is so persistent that it exists even at very advanced levels of L2 Chinese.
Mean frequency rates of the subjects in each group in interpreting the ‘daodi … wh … modal’ question as a genuine question or a negative rhetoric question in the discourse-completion task (the highest possible frequency rate = 4, the lowest = 0).
: significantly different from the NS group at p < 0.001.
Discussion
Table 8 provides a summary of the findings in our study of Chinese daodi … wh-questions. Recall that all the subjects in the advanced and very advanced groups were selected on a very stringent syntactic criterion that they had to consistently accept daodi c-commanded by CP[+Q] in Chinese daodi…wh-questions and at the same time also consistently reject daodi c-commanded by CP[−Q]. This criterion helped to identify subjects who syntactically treated daodi as an approximate counterpart of the-hell in their L2 Chinese daodi … wh-questions.
Summary of the findings concerning ‘daodi … wh’ questions.
√: accept; √¢: accept but not completely; X: reject; ?: indeterminate.
Subjects thus identified were found to have acquired other syntactic properties of daodi in Chinese daodi … wh-questions, that is, daodi is not permitted in questions without a wh-word (e.g. in a simple yes–no question); only the discontinuous form of daodi … wh is possible in Chinese daodi … wh-questions. In addition, their acceptance of daodi … wh-subject and rejection of wh-subject-daodi suggests that in their L2 Chinese daodi…wh-questions, daodi is likely to left-adjoin to IP, where the wh-subject is base-generated in the specifier position, just like the case in the native Chinese grammar. 24 It has been reported in the L2 literature that wh-in-situ in Chinese wh-questions does not cause problems to English-speaking learners of Chinese in spite of the fact that wh-movement is required in their L1 English (see Yuan, 2007a, 2007b). Given that English-speaking learners are exposed to the discontinuous form of daodi … wh and the position of daodi preceding the wh-subject in the input data, they should be able to learn that unlike the synthetic wh-the-hell in their L1 English, daodi and the wh-word in Chinese occupy different positions, with the former left-adjoining to IP and the latter in its base-generated position.
In addition, the requirement in learners’ L1 English that the wh-the-hell must be c-commanded by CP[+Q] seems to have also played a facilitating role in their acquisition of the syntactic structure of the Chinese daodi … wh-question. As we have seen, all the subjects in the advanced and very advanced groups consistently accepted daodi c-commanded by CP[+Q] in a Chinese daodi … wh-questions and at the same time also consistently rejected daodi c-commanded by CP[−Q]. This suggests that daodi in their L2 Chinese grammars had a [+Q] feature attached to it and had to be checked by a [+Q] feature in CP in their L2 Chinese daodi … wh-questions. It is obvious that there is no positive evidence in the input data that could inform English speakers that daodi CANNOT be c-commanded by CP[−Q] in Chinese. Given this fact, it seems reasonable to assume that the requirement in their L2 Chinese grammars that daodi must be c-commanded by CP[+Q] is transferred from the same requirement in their L1 English concerning the-hell. This L1 transfer enables them to meet the requirement concerning the c-commanding by CP[+Q] in their L2 Chinese daodi … wh-questions.
However, successful acquisition of the syntactic properties of the daodi … wh-question does not seem to imply successful interfaces with other cognitive domains in English speakers’ L2 Chinese. As we can see in Tables 5 and 8, the advanced group, who had no problem with the syntactic properties of the Chinese daodi … wh-question, did not readily accept the Chinese wh-question with daodi and na ‘which’ coexisting in it. This, again, can be assumed to be influenced by their L1 English. In English, the-hell is aggressively non-D-linked and which aggressively D-linked, and as a result, the two cannot coexist in a sentence because this would lead to a conflict in discourse (see Pesetsky, 1987). Such a conflict does not exist in Chinese. If English speakers cannot successfully expunge the L1-based discourse constraint from their L2 Chinese, they will have problems with the Chinese wh-question with daodi and na ‘which’ coexisting in it. This can explain why the advanced group did not have native-like behaviours in judging Chinese wh-questions of this type. However, learners are exposed to positive evidence of this type of wh-questions in their Chinese input, which can inform their L2 Chinese grammars that unlike the-hell in their L1 English, daodi in Chinese is not aggressively non-D-linked and can coexist with na ‘which’. With the exposure to this type of positive evidence in the input, English speakers can eventually recover from the L1 influence and acquire the native-like competence in this aspect of the Chinese grammar, and this is supported by the native-like behaviours of the very advanced group in their judgment of Chinese daodi … na ‘ daodi …which’ questions.
Successful recovery from L1 transfer was also found in English speakers’ judgment of daodi in the complement clause of veridical verbs. In English, veridical verbs, such as know, tell, confirm or realize, carry a presupposition of truth in the proposition of their complement clauses and this is incompatible with the property of the non-presupposition of the hell, and as a result, wh-the-hell is not licensed in the complements of these verbs. Dikken and Giannakidou (2002) point out that the incompatibility disappears when the veridical verbs are negated because the negation of these verbs creates a required non-veridical proposition in which wh-the-hell can occur legitimately. Evidence of the L1-based (in)compatibility of the veridical and negated veridical verbs with wh-the-hell was found in the advanced group’s judgment data as they accepted sentences with daodi … wh in the complement clauses of negated veridical verbs but not those with daodi … wh in the complement clauses of affirmative forms of veridical verbs. This shows that although this group of learners had good commands of the syntactic properties of the Chinese daodi … wh-question, their L2 Chinese syntactic structure was very much governed by the semantic constraint in their L1 English, which led to a problem at the syntax–semantics interface in their L2 Chinese grammars. However, the L1-based semantic constraint seems to disappear at the very advanced stage because our data show (see Table 5) that the very advanced group’s mean score for daodi in complements of affirmative forms of veridical verbs is +1.53, and there is no significant difference in their judgment between daodi with affirmative veridical verbs and daodi with negated veridical verbs. It is believed that this can be achieved with the help of the positive evidence in the input data that daodi can be used in the complement clause of affirmative forms of veridical verb in Chinese daodi … wh-questions, which can send a clear signal to English speakers’ L2 Chinese grammars that there is no semantic conflict between daodi and affirmative forms of veridical verbs in the target language Chinese.
Recall that daodi … wh in Chinese and wh-the-hell in English also behave differently at the discourse level; while the wh-word in the daodi … wh-question can be linked to an existential referent in the discourse, wh-the-hell in English, due to its non-presuppositional property, cannot. Evidence of L1-based discourse constraint was found in the advanced group’s sentence-combination task. As can be seen in Table 6, this group of learners behaved differently from the native Chinese group and made a significant distinction in the sentence-combination task between daodi … wh linked to a discourse-familiar referent and the linking of a wh-word to a discourse-familiar referent without daodi; they allow the latter but not the former. The implication of this finding is that in spite of their successful acquisition of the syntactic properties of the daodi … wh-question in Chinese, their L2 Chinese syntactic structure is regulated by the discourse constraint transferred from their L1 English, which results in a breakdown at the syntax–discourse interface in their L2 Chinese. The transfer of L1-based discourse constraint can make the discourse domain of daodi … wh in the learner’s L2 Chinese more like an open set, similar to that of wh-the-hell in their L1 English and, therefore, cannot be linked to a referent already existing in the discourse. While there is evidence of L1 transfer of the discourse constraint governing the syntactic structure of daodi … wh in English speakers’ L2 Chinese, our data suggest that the L1 transfer of the discourse constraint can be removed from English speakers’ L2 Chinese because the data of the very advanced group show that they had native-like behaviours in the sentence-combination task, allowing the wh-word in daodi … wh to be linked to a discourse-familiar referent. The removal of the L1 transfer of the discourse constraint can be triggered by the positive evidence that learners are exposed to in the Chinese input. The reference linking like that between yi jian dongxi ‘something’ and shenme ‘what’ in the sentence in (18) can provide robust evidence to the learner that daodi … wh can indeed be bound by a discourse-familiar referent in Chinese.
(18) Wo baba zhi shuo ta gei wo mai-le yi jian dongxii, my father only say he for me buy-PFV one CL thing
keshi wo yijing caichulai ta daodi gei wo mai-le shenmei.
but I already figure out he daodi for me buy-PFV what * ‘My father only said that he had bought something for me, I have already figured out what the hell he has bought for me.’
However, removal of L1 transfer does not seem to be always possible, and this can be seen in the L2 groups’ performance in the discourse-completion task, which was designed to examine the syntax–discourse interface in English speakers’ L2 Chinese involving the Chinese daodi … wh … modal question as a genuine-information question or as a negative rhetorical question. While the Chinese daodi … wh-questions with modals can be naturally answered with genuine information and cannot be interpreted as expecting a negative rhetorical answer, the English wh-question with wh-the-hell … modal can only be felicitously answered with a negative rhetorical answer. As shown in Table 7, the two L2 Chinese groups did not have difficulty in interpreting the daodi … wh-question as a genuine-information question, and they were as frequent as the native Chinese speakers in answering the questions with genuine information. This provides us with robust evidence that English speakers are able to acquire Chinese daodi … wh … modal questions as genuine questions. However, this native-like behaviour in English speakers’ L2 Chinese should not be taken as evidence of the overall success of the syntax–discourse interface involved in Chinese daodi … wh … modal questions because our data suggest that the native-like interpretation of the Chinese daodi … wh … modal question as a genuine question is only in addition to, rather than a replacement of, the L1-based interpretation of the question as a negative rhetorical question in English speakers’ L2 Chinese. As we can see in the last column of Table 7, English speakers, in both the advanced and very advanced groups, optionally responded to the Chinese daodi … wh … modal question with a negative rhetorical answer, and more than 50% of the questions were responded to by learners in these groups with negative rhetorical answers.
Recall that subjects in the very advanced group are considered to be at very advanced levels of L2 Chinese because they had a history of over 14 years of studying Chinese and their average stay in Chinese-speaking environments was nearly 4 years. Given this group of learners’ background and given that they had a native-like command of the syntactic properties of the Chinese question but optionally made the incorrect interpretation of the question, it is reasonable to argue that there is a persistent problem at the syntax–discourse interface in this aspect of English speakers’ L2 Chinese because their native-like syntactic structure of the question is optionally regulated by the discourse constraint transferred from their L1 English, where wh-the-hell … modal questions are read as expecting negative rhetorical answers in the discourse.
Why is it possible for English speakers to acquire the native-like interpretation of the Chinese daodi … wh … modal question as a genuine-information question but difficult for them to get rid of the L1-based discourse condition on this type of question? Again, we would argue that this is due to the availability of positive evidence that English speakers are exposed to in the Chinese input. Learners are exposed to the Chinese discourse where daodi … wh … modal questions are answered with genuine information, which can provide them with clear evidence that this type of questions can be genuine-information questions in the target language Chinese. However, there is no positive evidence in the input data that this type of questions CANNOT be answered with a negative rhetorical answer. In other words, nothing in the input can help remove the L1-based discourse condition in English speakers’ L2 Chinese in spite of the successful acquisition of syntactic properties of the question and an alternative but native-like interpretation of the question as a genuine-information question. Because there is neither positive evidence in the input data for the removal of the L1-based discourse condition nor positive evidence to confirm the legitimate use of the L1-based discourse condition in the target language, negative rhetorical answers remain optional answers to Chinese daodi … wh … modal questions at the discourse level, even at very advanced levels of their L2 Chinese. In the study by Sorace (2011), cross-linguistic influence is considered a result that L2 speakers fail to inhibit the unwanted language, and she argues that exercising executive control to constantly inhibit the unwanted language may take attentional resources from other tasks, which may lead to residual optionality in the L2 end state. This analysis could be used to account for English speakers’ optional interpretations of the daodi … wh … modal question as a negative rhetorical question. However, it will be unable to explain why they are able to make determinate and correct interpretation of the question as a genuine-information question, and it cannot provide a satisfactory answer with regard to very advanced learners’ native-like performance in D-linking and linking daodi … wh- to a discourse-familiar referent.
The findings in our study only partially confirm the Interface Hypothesis by Sorace and Filiaci (2006), according to which, narrow syntactic properties in L2 acquisition are completely acquirable but grammatical aspects that involve an interface between the syntactic domain and other cognitive domains may not be fully acquirable in L2 acquisition. Our data suggest that English speakers are able to successfully acquire syntactic properties of Chinese daodi … wh-questions, but interfaces are not completely unacquirable in L2 acquisition. Our results have shown that in spite of problems caused by L1 transfer, English speakers are able to eventually acquire native-like competence in most, although not all, interfaces involved in the study.
As discussed earlier, a more fine-tuned approach to L2 interfaces in recent L2 studies has argued that not all interfaces in L2 acquisition are problematic and that a distinction should be made between internal interfaces and external interfaces (see Belletti et al., 2007; Sorace, 2005, 2006, 2009; Sorace & Serratrice, 2009; Tsimpli & Sorace, 2006; White, 2009). However, the findings in our study cannot be easily ascribed to the distinction between the relatively less problematic acquisition of internal interfaces and the vulnerability of external interfaces in L2. As predicted by the distinction, the syntax–semantics interface, an internal interface, in the compatibility between the daodi … wh-question and veridical verbs can be successfully acquired by very advanced English–speaking learners of Chinese, but the same also applies to some external interfaces in our study. As we have seen, the conflict of the aggressively D-linked wh-word na ‘which’ and the L1-based property of aggressively non-D-linked the-hell attached to the Chinese word daodi in earlier stages can be removed and daodi can coexist with the aggressively D-linked wh-word na ‘which’ in very advanced learners’ L2 Chinese wh-questions. While one could argue that the aggressively D-linking property of both the English wh-word which and its Chinese counterpart na is lexically inherent and does not necessarily require the presence of discourse, and therefore may not be a truly valid linguistic entity for the study of L2 syntax–discourse interfaces, the result from the sentence-combination task demonstrates very clearly that in spite of L1 transfer of the discourse constraint in earlier stages, daodi … wh can be linked to a discourse-familiar referent in very advanced learners’ L2 Chinese. This has obviously put into question the reliability of the distinction between internal and external interfaces in L2 acquisition studies.
It has been proposed that problems at L2 interfaces can be due to various limitations in processing the target language by L2 learners, which include insufficient processing resources at their disposal to integrate the multiple types of information involved at the interfaces between syntax and other cognitive systems and inefficient coordination that makes it difficult for L2 grammars to coordinate the L2 syntax domain with other cognitive domains (see Sorace, 2005; Sorace & Filiaci, 2006; Sorace & Serratrice, 2009). However, it seems unlikely that the interface problems that English speakers have at various stages in their L2 acquisition of Chinese can be accounted for in terms of processing factors because our results have shown that some interfaces (including both internal and external interfaces) are acquirable in English speakers’ L2 Chinese, but the syntax–discourse interface involving the appropriateness of answers to the Chinese daodi … wh … modal questions is persistently vulnerable even at very advanced levels of their L2 Chinese. If processing factors are to blame here, it would be difficult to explain why they are not in operation at the syntax–discourse interface involving daodi … wh linked to a discourse-familiar referent, which has been shown to be acquirable in English speakers’ L2 Chinese.
An L1-based feature of negative rhetorical questions improperly attached to Chinese daodi … wh … modal questions will trigger negative rhetorical answers in learners’ L2 Chinese discourse, and as there is no positive evidence in the input for the removal of the improper discourse feature, negative rhetorical answers remain possible answers to this type of questions, even in very advanced learners’ L2 Chinese. More specifically, because of the L1 influence, the L2 speakers have a wider range of options available to them in answering Chinese daodi … wh … modal questions, which is in contrast with the native Chinese speakers who do not have the additional option available to them. Our findings show that the information-seeking feature can be properly attached to the daodi … wh … modal question in English speakers’ L2 Chinese, allowing genuine information to be used as answers to the questions. However, the successful specification of information-seeking feature is only an addition to, rather than a replacement of, the negative rhetorical feature attached to English speakers’ L2 Chinese daodi … wh … modal questions.
Conclusion
The latest reformulation of the Interface Hypothesis (Sorace, 2011) claims that this hypothesis should be tested via online measures, and we acknowledge that the tests we used in our empirical study were mostly offline. 25 However, the varying successes/failures at different L2 Chinese interfaces found in our study seem to render it legitimate to argue that the vulnerability in English speakers’ L2 Chinese daodi … wh-questions is unlikely to be due to processing problems at interface levels. Recall that Sorace and Serratrice (2009) predict that while the syntax–semantics interface is influenced by L1 transfer, the syntax–discourse interface is not and that structures at the syntax–discourse interface have the potential to become permanently optional in L2 grammars. Our results have demonstrated that cross-linguistic influence can occur at both internal and external interfaces in L2 acquisition and that the vulnerability at the syntax–discourse interface may not necessarily be permanent.
Is ‘daodi, ‘the-hell’ in English speakers’ L2 Chinese? Syntactically, the requirement in their L2 Chinese grammars that daodi … wh must be c-commanded by CP[+Q] is believed to be due to the requirement attached to wh-the-hell in their L1 English because there is no positive evidence in the Chinese input that could inform English speakers that daodi … wh CANNOT be c-commanded by CP[−Q] in Chinese. Semantically and in view of D-linking property, there are traces of the-hell at earlier stages of L2 Chinese. At the discourse level, there is evident influence of the-hell regarding possible linking of daodi … wh to a discourse-familiar referent. Nevertheless, English-speaking learners can eventually recover from the-hell and become native-like in these aspects. However, the daodi … wh … modal question is very much affected by the discourse condition of the wh-the-hell as a negative rhetorical question in learners’ L1 English, and this L1 influence is so persistent that this infelicitous syntax–discourse interface continues to be optional in learners’ L2 Chinese, even at very advanced levels. This provides us with evidence that recovery from L1 transfer may not always be possible in L2 interfaces, and this is likely to depend on the availability of relevant positive evidence in the input, as in L2 acquisition of other language aspects.
Footnotes
Acknowledgements
I am very grateful to those students and teaching staff in China and in the United Kingdom who participated as subjects in the empirical study. Without their kind help and support, this project would have been very difficult, if not impossible. I would also like to acknowledge the assistance that the National Research Centre for Foreign Language Education of Beijing Foreign Studies University gave to me as its Overseas Research Fellow. An earlier version of the article was presented at the GASLA-11 Conference at the University of Washington, Seattle, and I thank the audience for their comments and suggestions.