WH-TOPICALIZATION IN CHINESE QUESTIONS

It is clear from the two Chinese wh-questions in (1) that the movement of shenme cai “what dish” in (1b) cannot be motivated by checking the wh-feature because the wh-phrase can stay in situ in Chinese, as in (1a). What then triggers the movement of the wh-phrase in (1b), and where does it move to? Wu (Reference Wu1999) makes an important distinction between wh-topicalization questions and wh–in situ questions. For wh-topicalization questions to be felicitous, the interlocutors in the discourse must “have a set of things in the presupposition background,” whereas in wh–in situ questions, “no particular set of things . . . is presupposed” (p. 83). The answer to the question in (1a) is not constrained by a particular set of dishes, and in fact, anything that can be called a cai “dish” can be used to answer the question in (1a). The answer to the question in (1b), however, is restricted to a set of dishes that has already been established in the previous discourse. For instance, if 10 people attended a dinner party—and everyone got food poisoning afterward except for Zhangsan—to find out what dish(es) caused the food poisoning, an investigator asked Zhangsan the question in (1b) with a preestablished knowledge of the set of dishes that was provided at the dinner party. Zhangsan was then naturally expected to answer the question by naming a dish or dishes only from the dishes provided at the dinner party. This makes it possible for the wh-phrase shenme cai “what dish” to be topicalized in (1b).

The examples in (2)–(4) demonstrate that only D-linked wh-words can undergo wh-topicalization and that non-D-linked wh-words cannot.

1. (2) a. Ni xihuan shei?

   you like who

   “Who do you like?”

2. b. *[Shei]_i ni xihuan t_i?

   who you like

   “Who do you like?”

3. (3) a. Ni xihuan na yi ge ren?

   you like which one cl person

   “Which person do you like?”

4. b. [Na yi ge ren]_i ni xihuan t_i?

   which one cl person you like

   “Which person do you like?”

5. (4) a. Ni zui xihuan shei?

   you most like who

   “Who do you like most?”

6. b. [Shei]_i ni zui xihuan t_i?

   who you most like

   “Who do you like most?”

As shei “who” in (2) is a non-D-linked wh-word, it can only stay in situ, as in (2a), and it is not felicitous if it is topicalized in Chinese wh-questions, as in (2b). In (3), the wh-phrase is D-linked, which consists of na “which” and a nominal phrase yi ge ren “one person.” In this case, both the speaker and the listener have a shared group of people in mind, and the person who answers the question is expected to choose one person from this group. This makes wh-topicalization possible in (3b). The sentences in (4) are identical to those in (2), except for the insertion of zui “most.” Here it is possible to assume that the word zui “most” creates a presupposed group of people in (4) and converts the wh-word shei “who” in (4) to a D-linked wh-word and thus enables its topicalization in (4b).

It is also possible to make a distinction between wh-elements like na “which”–noun phrase (NP) and wh-elements like shei “who”; na-NP can be considered a wh-determiner phrase (DP) and shei a wh-NP. In na-NP, the wh-word is a determiner and selects a NP as its complement, as na yi ge ren “which person” in (3),Footnote ³ whereas with shei, the wh-word is the head of the wh-NP, as in (2) and (4).Footnote ⁴ The head of the DP is typically considered to be the locus of references and is believed to host expressions related to references. Following this line of analysis, it is possible to assume that the wh-determiner has a [+/−] deictic feature attached to it, which can link the wh-DP to the relevant discourse information and cause the wh-DP to be D-linked. This makes wh-topicalization possible in Chinese. In contrast, the wh-NP shei “who” has no deictic feature attached to it and has to rely on an element like zui “most” to be related to the discourse, as shown in (2) and (4). The adverb zui “most” is an adjunct, and it adjoins to the verb phrase (VP) in (4b) and c-commands the wh-NP shei “who” in object position. Here, it is zui “most” that licenses the wh-NP as a D-linked wh-element. This creates a presupposition background in the discourse and enables the wh-NP to topicalize.

Wu (Reference Wu1999) makes a clear distinction between wh-movement in English and wh-topicalization in Chinese. In English, the wh-word raises to the specifier (Spec) of the complementizer phrase (CP) to check the wh-feature,Footnote ⁵ whereas in Chinese, the wh-word raises to the Spec of the topic phrase (TopP) to check the topic feature, as shown in (5) for (4b). According to Wu, the topic feature in Chinese is strong and must be checked before spell-out.

1. (5) [_TopPShei_i [_IPni [_VPzui [_VPxihuan t_i]]]]?

   who you most like

   “Who do you like most?”

The distinction between D-linked wh-words and non-D-linked wh-words has been recognized since Pesetsky (Reference Pesetsky, Reuland and ter Meulen1987). In English, for example, the D-linked wh-word which is not subject to the superiority condition, as in (6).Footnote ⁶ The behavior of which is in striking contrast to that of what and who in (7), both of which Pesetsky considers to be non-D-linked wh-words. The superiority effect is also absent in Chinese D-linked wh-words, as shown in (8b), which is perfectly acceptable in Chinese in spite of the fact that the lower wh-phrase shenme shu “what book” raises over the higher wh-word shei “who” to the left periphery of the multiple wh-question. In (8), there is a presupposed set of books, and the answer must be from this set. The absence of the superiority effect should not be surprising if the movement in sentences like (8b) is considered an instance of wh-topicalization rather than wh-movement.

1. (6) a. [Which singer]_i t_i sang which song?

2. b. [Which song]_i did which singer sing t_i?

3. (7) a. [Who]_i t_i sang what song?

4. b. *[What song]_i did who sing t_i ?

5. (8) a. Shei zui xihuan shenme shu?

   who most like what book

   “Who likes what book most?”

6. b. [Shenme shu]_i shei zui xihuan t_i?

   what book who most like

   *“What book does who like most?”

Following Wu’s (Reference Wu1999) assumption that the economy condition (Chomsky, Reference Chomsky1995) only applies to wh-feature checking in wh-movement but not to wh-topicalization, economy considerations will rule out any derivation like (7b) in which the lower wh-phrase what song raises over the higher one who, given that the alternative derivation in (7a) in which the higher wh-word who raises is more economical. Unlike wh-movement, wh-topicalization involves checking the topic feature, which is assumed not to be inherently attached to a wh-word. As a result, it is legitimate to raise either the higher or the lower wh-word to serve as a topic; thus the absence of the superiority effect in wh-topicalization.

It is clear from the examples in (3), (4), and (8) that D-linking is a necessary condition but not a sufficient condition for wh-topicalization in Chinese because D-linked wh-words are allowed to stay in situ, as in (3a), (4a), and (8a). What then triggers wh-topicalization in Chinese? Pan (Reference Pan2006) argues that D-linked wh-words bear an underspecified [+/− topic] feature and that two requirements are involved in wh-topicalization: (a) a [+topic] feature and (b) a D-linked wh-word that is linked to a set of things or a group of people in the presupposition background. It seems feasible to assume that when the head of TopP has a [+topic] feature, it can trigger wh-topicalization, attracting a D-linked wh-word to Spec TopP to get its value and specification as a wh-topic with a [+topic] feature. However, the wh-topicalization triggered by the [+topic] feature at the head of TopP is subject to a condition under which the wh-word must have a D-linked reading to be topicalized, and the wh-topicalization must be licensed by a presupposed set of things or a presupposed group of people. In other words, both the [+topic] feature and the relevant discourse condition are indispensable for wh-topicalization in Chinese, and neither of them can trigger the wh-topicalization by itself. In English, the wh-word which is inherently D-linked lexically, whereas the other wh-words are not. However, Chinese wh-words are underspecified with regard to the D-linked reading and become D-linked when they are licensed by a relevant discourse—for example, by a set of things or a group of people in the presupposed background, as in (1b) and (3); by words like zui “most,” as in (4), (5), and (8); or by both. This analysis is in conformity with the proposal that Chinese is a discourse-oriented language (see Huang, Reference Huang1984, Reference Huang, Jaegglie and Safir1989; Shi, Reference Shi1989; Tsao, Reference Tsao1977). A D-linked wh-word with a [−topic] feature stays in situ in Chinese wh-questions, as in (3a) and (4a). As discourse lies outside the linguistic computation, the syntax-discourse interface involved in Chinese wh-topicalization is considered an external interface in the sense of Belletti and colleagues (2007), Sorace (Reference Sorace, Cornips and Corrigan2005, Reference Sorace2009), Sorace and Serratrice (Reference Sorace and Serratrice2009), Tsimpli and Sorace (Reference Tsimpli, Sorace, Bamman, Magnitskaia and Zaller2006), and White (Reference White, Ritchie and Bhatia2009). This interpretation of topicalized wh-words is also in conformity with the widely accepted definition of topics proposed by Li and Thompson (Reference Li, Thompson and Li1976, Reference Li and Thompson1981); that is, topics are generally definite and provide old, shared, and preestablished information. A wh-word becomes definite when it is D-linked because it is what both the speaker and the listener have in their presupposition.

In sum, wh-words can be topicalized to Spec TopP in Chinese, but the wh-topicalization takes place at the syntax-discourse interface. This syntactic movement has to be licensed by reference to a set of things or a group of people in the presupposition background; that is, only D-linked wh-words can be topicalized, and non-D-linked wh-words cannot. Unlike the wh-word which in English, which is inherently D-linked lexically, Chinese wh-words become D-linked when they get licensed by a set of things or a group of people in the presupposition background, by words like zui “most,” or by both.

L2 INTERFACE STUDIES

Why are adult L2 learners rarely able to acquire native competence of the target language? What are the possible causes of the divergence between L2 grammars and the native grammar? To find answers to these questions, an increasing number of L2 researchers have investigated interfaces in L2 grammars as a possible cause of the divergence. Sorace and Filiaci (Reference Sorace and Filiaci2006) studied the interpretation of Italian intrasentential anaphora by native Italian speakers and English near-native speakers of Italian. Results showed that the English near-native speakers of Italian had nativelike interpretations of Italian null subject pronouns in both forward and backward anaphora conditions. However, with respect to overt subject pronouns, the near-native speakers were found to be divergent from the native Italian speakers, particularly in the interpretation of overt subject pronouns in the backward anaphora condition. On the basis of this finding, Sorace and Filiaci propose the interface hypothesis, according to which narrow syntactic properties in L2 acquisition are completely acquirable, but aspects of grammar that involve an interface between the syntactic domain and other cognitive domains may not be. This is believed to be due to the fact that L2 grammars have insufficient processing resources at their disposal to integrate the multiple types of information involved at the interfaces between syntax and other cognitive systems, and that these insufficient processing resources make it difficult for L2 grammars to coordinate the L2 syntax domain with other cognitive domains.

Evidence in support of the interface hypothesis has been found in a number of studies in the L2 literature. These include Hopp’s (Reference Hopp, Foster-Cohen, Sharwood-Smith, Sorace and Ota2004) study of the syntax-discourse interface in the acquisition of German scrambling by English and Japanese speakers and Sorace’s (Reference Sorace2004, Reference Sorace, Cornips and Corrigan2005, Reference Sorace, Belletti, Bennati, Chesi, Di Domenico and Ferrari2006) investigations of pragmatic-discourse factors in near-native speakers’ L2 Italian use of null and overt subjects. However, there is also evidence in the L2 literature showing that targetlike grammars are attainable in L2 interfaces and that vulnerability in L2 interfaces is not inevitable. For example, Dekydtspotter, Sprouse, and Thyre’s (1999/Reference Dekydtspotter, Sprouse and Thyre2000) study showed that English-speaking learners of French and native French speakers exhibited similar knowledge of the syntax-semantics interface involving the interpretations of French quantification at a distance. Similar results were also obtained in studies by Borgonovo, Bruhn de Garavito, and Prévost (Reference Borgonovo, Bruhn de Garavito, Prévost, Burgos, Clark-Cotton and Ha2005) and by Borgonovo, Bruhn de Garavito, Guijarro-Fuentes, Prévost, and Valenzuela (Reference Borgonovo, Bruhn de Garavito, Guijarro-Fuentes, Prévost, Valenzuela, Foster-Cohen and Krajnovic2006). In Borgonovo et al. (Reference Borgonovo, Bruhn de Garavito, Prévost, Burgos, Clark-Cotton and Ha2005), English-speaking learners were found to be able to acquire the syntax-semantics interface involved in the choice of mood in relation to the specificity of the DP head in Spanish relative clauses, whereas Borgonovo et al. (Reference Borgonovo, Bruhn de Garavito, Guijarro-Fuentes, Prévost, Valenzuela, Foster-Cohen and Krajnovic2006) showed that Brazilian-Portuguese speakers were able to acquire the syntax-semantics interface regarding the specificity restrictions on null objects in their L2 Spanish.

A more fine-tuned approach to L2 interfaces has been proposed and argues that not all interfaces between syntax and other cognitive domains are problematic (Belletti et al., Reference Belletti, Bennati and Sorace2007; Sorace, Reference Sorace, Cornips and Corrigan2005, Reference Sorace, Belletti, Bennati, Chesi, Di Domenico and Ferrari2006, Reference Sorace2009; Sorace & Serratrice, Reference Sorace and Serratrice2009; Tsimpli & Sorace, Reference Tsimpli, Sorace, Bamman, Magnitskaia and Zaller2006; White, Reference White, Ritchie and Bhatia2009). A distinction should therefore be made between internal interfaces and external interfaces, in which internal interfaces include syntax-morphology, syntax-lexicon, and syntax-semantics, and external interfaces include syntax-discourse and syntax-pragmatics. According to these researchers, internal interfaces involve different modules internal to the grammar. Because the operations are within the linguistic computation, internal interfaces are relatively easier for L2 learners to acquire. However, external interfaces (e.g., syntax-discourse and syntax-pragmatics interfaces) involve external conditions of contextual appropriateness, which are outside the grammar proper and therefore impose extra burdens on the processing of information from different resources.

The distinction between internal interfaces and external interfaces is useful in accounting for the findings in the L2 interface studies reviewed previously. Syntax-semantics interfaces, being internal in the linguistic computation and involving formal properties of the language system alone, are acquirable, whereas syntax-discourse or syntax-pragmatics interfaces are particularly problematic in L2 grammars because they involve external conditions and are likely to lead to interface vulnerability. This distinction is also supported by the results in Tsimpli and Sorace’s (Reference Tsimpli, Sorace, Bamman, Magnitskaia and Zaller2006) study of Russian learners of Greek. In the study, Russian learners were found to be able to acquire the syntax-semantics interface involved in focusing and verb raising in their L2 Greek. In contrast, the vulnerability of the syntax-discourse interface was found in Russian learners’ overuse of Greek overt subject pronouns, whose distribution was regulated by discourse factors.

However, the contrast between the relatively less problematic acquisition of internal interfaces and the vulnerability of external interfaces in L2 is not consistently supported by data of all relevant studies. Lozano and Mendikoetxea (Reference Lozano, Mendikoetxea, Chandlee, Franchini, Lord and Gudrun-Marion2009) examined verb-subject (VS) structures using experimental data from English and Greek speakers’ L2 Spanish and corpus data from Spanish and Italian speakers’ L2 English. Their results indicated that both groups of learners were sensitive to discourse status (i.e., focus) in their L2s, but they had problems encoding the discourse information syntactically. This raises the possibility that vulnerability of L2 syntax-discourse interfaces can be related to a deficit in underlying representations of L2 grammars. Evidence in support of this argument is also provided in a study of Belletti and Leonini (Reference Belletti, Leonini, Foster-Cohen, Sharwood-Smith, Sorace and Ota2004), in which L2 learners of Italian were shown to have problems with the VS word order related to focus. There was, however, clear evidence that they were sensitive to the pragmatic focus of the construction because they were found to use stress—transferred from their first language (L1)—on the focused element in their L2 Italian. Belletti and Leonini attributed this problem to defective representations of a functional category (i.e., the focus projection) in L2 Italian grammars.

It should be pointed out that the studies of L2 interfaces reviewed previously have mostly dealt with L2 interface issues in holistic terms. In other words, researchers tend to treat the L2 syntax-semantics interface, the L2 syntax-discourse interface, or L2 interfaces in general in a domain-wide fashion. This is questioned by Yuan (Reference Yuan2010), who proposed a variable-dependent approach to L2 interfaces, arguing that success or failure in acquiring a particular interface in L2 may not necessarily be across the whole interface domain and is likely to be affected by different variables involved. Yuan’s argument was based on a study of the syntax-semantics interface involved in English speakers’ acquisition of Chinese wh-words used as existential polarity words (EPWs). Chinese wh-words are ambiguous and can be interpreted as EPWs as well as interrogative words.Footnote ⁷ When used as EPWs, Chinese wh-words are subject to strict syntactic and semantic conditions. Semantically, an EPW must fall within a context in which the proposition containing the EPW is a nonfact or in which the truth value of the proposition is not positively fixed in a definite manner. Syntactically, an EPW must occur in the c-commanding scope of its licenser. The licensers for wh-words used as EPWs in Chinese include negators, the yes-no particle ma, words such as ruguo “if” in conditional clauses, uncertainty adverbs, the sentence final particle le indicating inference, nonfactive verbs, and so forth. The results of Yuan’s study indicated that the syntax-semantics interface could be established between EPWs and some of their licensers but not the others, which provides evidence of a nondomain-wide success or failure in L2 acquisition of interfaces. On the basis of a variable-dependent approach, Yuan argued that English speakers’ L2 acquisition of the syntax-semantics interface involved in Chinese wh-words used as EPWs is affected by a number of variables, including the categorical natures of individual elements involved in the interface relationship, the status of these elements in the target language speakers’ grammar, the input that L2 learners are exposed to, crosslinguistic influences, and so forth.

However, Yuan’s (Reference Yuan2010) study only involved an internal interface (i.e., the syntax-semantics interface), which is considered to be relatively less problematic in the L2 acquisition of interfaces. Can Yuan’s variable-dependent approach apply to L2 acquisition of external interfaces like the syntax-discourse interface, which is deemed to be vulnerable in L2 grammars (see Belletti et al., Reference Belletti, Bennati and Sorace2007; Sorace, Reference Sorace, Cornips and Corrigan2005, Reference Sorace2009; Sorace & Serratrice, Reference Sorace and Serratrice2009; Tsimpli & Sorace, Reference Tsimpli, Sorace, Bamman, Magnitskaia and Zaller2006; White, Reference White, Ritchie and Bhatia2009)? This study is expected to provide an answer as it focuses on the syntax-discourse interface involved in English speakers’ L2 acquisition of Chinese wh-topicalization.

THE EMPIRICAL STUDY

Participants

To answer these questions, an empirical study was conducted involving 71 English-speaking learners of Chinese and a control group of 20 native speakers of Chinese. The English-speaking learners included students studying Chinese at universities in Beijing as well as students and lecturers of Chinese studies at some British universities. The lecturers were teaching Chinese or Chinese-related courses. A purpose-designed cloze test, which consisted of three passages written in both Chinese characters and pinyin (i.e., a Chinese phonetic writing system), was used as a proficiency test. The total score of the proficiency test was 40.

To examine wh-topicalization in L2 Chinese, it is a prerequisite to ensure that general topicalization has been established in the L2 Chinese grammars of the participants involved in the experiment. To do this, a stringent criterion was adopted to select participants who were considered to have acquired topicalization in Chinese. There were four types of NP-topicalization control sentences in categories I–IV, each of which contained three tokens (see Appendix). In total, there were 12 NP-topicalization sentences in the experiment. Only those learner participants who accepted at least 11 out of the 12 NP-topicalization sentences were included in this study; that is, 92% of their judgments of the sentences with the NP-topicalization had to be +1 or higher on the Likert scale provided.

On the basis of this criterion, 19 English speakers were selected for the study who had mastered topicalization in Chinese. Those who did not meet the criterion were eliminated from the analyses of L2 Chinese wh-topicalization. Detailed information on the English group (English) and the native Chinese group (NS) is provided in Table 1.

Table 1.

Information on participants who mastered topicalization in Chinese

Note.

Values enclosed in parentheses represent standard deviations.

The result of a two-sample Kolmogorov-Smirnov test indicates that there was no significant difference between the two groups in their performance in the cloze test (Z = 1.314, p = .063). Given (a) the lack of a significant difference between the English group and the NS group in the cloze test, (b) the fact that the English group’s average length of learning Chinese was 178 months (i.e., nearly 15 years), and (c) the fact that their average stay in China or Taiwan was 66 months (i.e., 5.5 years), the English group is considered a possible representative of English speakers’ end states of L2 Chinese grammars.

Results

The results presented in this section are based on the judgment data of wh-questions by the English and NS groups. The judgment data of sentences with a topicalized NP (i.e., the NP-topicalization control sentences) are not included because participants in the English group have all met the stringent criterion for Chinese topicalization and are therefore considered to have established Chinese topicalization in their L2 Chinese grammars. In Figure 1, it is possible to see that—like the NS group—the English group accepted wh-questions with a topicalized D-linked na-NP “which-NP.” The same was found in wh-questions with a topicalized D-linked na-NP when it was reinforced by zui “most.” The results of Kolmogorov-Smirnov tests show that there was no significant difference between the two groups in their judgments of either type of sentences. For wh-questions with a D-linked topicalized na-NP, Z = 1.002, p = .268; for wh-questions with a D-linked topicalized na-NP + zui, Z = 1.183, p = .122. In regard to judgments of wh-questions when na-NP was left in situ, the result of a Kolmogorov-Smirnov test indicates that no significant difference existed between the two groups, Z = 0.304, p = 1.000.

Figure 1.

Mean scores of groups’ judgments of Chinese wh-questions with D-linked which-NP topicalized, with most and which-NP topicalized, and with which-NP in situ (see examples in Appendix, section I).

These results suggest that a D-linked which-NP is allowed to be topicalized in English speakers’ L2 Chinese grammars as long as the Chinese topicalization is established in their L2 Chinese grammars. However, as the wh-word which is an inherently D-linked word in English, it could be argued that the English group’s nativelike behaviors in judging Chinese wh-questions with na-NP topicalized may be influenced by properties of the wh-word which in their native language. For this reason, it is necessary to look at the behaviors of other wh-words in the English group’s L2 Chinese when they are D-linked and topicalized.

Figure 2 presents the results of the judgments of wh-questions with a topicalized D-linked shenme-NP “what-NP,” wh-questions with a topicalized D-linked shenme-NP reinforced by zui “most,” and wh-questions with a D-linked shenme-NP left in situ. The results of Kolmogorov-Smirnov tests indicate that there was no significant difference between the two groups in their judgments of any of the sentence types. For the wh-questions with a topicalized D-linked shenme-NP, Z = 0.748, p = .631; for the wh-questions with a topicalized D-linked shenme-NP reinforced by zui, Z = 0.863, p = .447; and for the wh-questions with a D-linked shenme-NP in situ, Z = 0.682, p = .741. These results clearly demonstrate that it is possible to topicalize D-linked shenme-NP in English speakers’ L2 Chinese wh-questions in spite of the fact that the English equivalent of shenme-NP is not D-linked in their native language.

Figure 2.

Mean scores of the groups’ judgments of Chinese wh-questions with D-linked what-NP topicalized, with most and D-linked what-NP topicalized, and with D-linked what-NP in situ (see examples in Appendix, section II).

Figure 3 presents results concerning another D-linked wh-argument sheide-NP “whose-NP” in Chinese wh-questions. The English group’s judgment was similar to that of the NS group, and the participants in the English group accepted (a) Chinese wh-questions with the D-linked wh-argument topicalized, (b) wh-questions with a D-linked topicalized wh-argument reinforced by zui “most,” and (c) wh-questions with the D-linked wh-argument in situ. The results of Kolmogorov-Smirnov tests show that there was no significant difference between the two groups in their judgments of any of these sentence types. For the wh-questions with a topicalized D-linked sheide-NP, Z = 0.394, p = .998; for the wh-questions with a topicalized D-linked sheide-NP reinforced by zui “most,” Z = 1.060, p = .211; and for the wh-questions with a D-linked sheide-NP in situ, Z = 0.723, p = .673.

Figure 3.

Mean scores of the groups’ judgments of Chinese wh-questions with D-linked whose-NP topicalized, with most and D-linked whose-NP topicalized, and with D-linked whose-NP in situ (see examples in Appendix, section III).

The results, however, concerning shei . . . zui “who . . . most” present a different picture, as shown in Figure 4. Although the English group, like the NS group, allowed the D-linked shei “who” to remain in situ in their L2 Chinese wh-questions, unlike the NS group, they did not accept Chinese wh-questions with the D-linked shei topicalized—despite the presence of the word zui “most” in the sentence. The results show that participants in the English group were indeterminate about the acceptability of Chinese wh-questions with the D-linked shei topicalized because their mean score in judging this type of wh-questions was −0.12. The result of an independent-samples t test shows that a significant difference existed between the two groups in their judgments of the wh-questions with topicalized shei . . . zui, t(37) = 5.029, p < .001. However, no significant difference was found in a Kolmogorov-Smirnov test between the two groups in their judgments of wh-questions zui . . . shei “most . . . who” in situ, Z = 0.674, p = .755. The English group’s nonacceptance of the topicalization of the D-linked shei with the presence of zui in the wh-question is rather unexpected, and this provides some evidence that not all D-linked wh-arguments can be topicalized in L2 Chinese wh-questions.

Figure 4.

Mean scores of the groups’ judgments of Chinese wh-questions with most and D-linked who topicalized, and with most and who in situ (see examples in Appendix, section IV).

The results presented have all been concerned with D-linked wh-arguments. To examine wh-topicalization in L2 Chinese, it is also necessary to find out whether L2 grammars, at the same time, are able to reject topicalization of non-D-linked wh-arguments in L2 Chinese wh-questions. Figure 5 presents data of this type. The English group, like the NS group, rejected incorrect wh-questions with non-D-linked shenme-NP “what-NP,” shenme “what,” or shei “who” that were topicalized. At the same time, they were nativelike in accepting wh-questions with these non-D-linked wh-arguments in situ. No significant difference was found in independent-samples t tests between the two groups in rejecting wh-questions with a topicalized non-D-linked shenme-NP, t(37) = −1.799, p = .078, and in rejecting incorrect wh-questions with topicalized non-D-linked shenme and shei—for shenme, t(37) = 0.116, p = .909; for shei, t(37) = −0.939, p = .354. Like the NS group, the English group readily accepted wh-questions with non-D-linked wh-arguments in situ, and there was no significant difference in Kolmogorov-Smirnov tests between the two groups in accepting the non-D-linked shenme-NP in situ, Z = 0.164, p = 1.000, and in accepting wh-questions with the non-D-linked shenme and shei in situ—for shenme, Z = 0.164, p = 1.000; for shei, Z = 0.329, p = 1.000.

Figure 5.

Mean scores of the groups’ judgments of ungrammatical Chinese wh-questions with non-D-linked what-NP, what, and who topicalized, and grammatical wh-questions with non-D-linked what-NP, what, and who in situ (see examples in Appendix, section V).

Figure 4 shows that the English group did not accept Chinese wh-questions with the D-linked shei “who” topicalized in spite of the presence of the word zui “most” in the sentence. An examination of individual data reveals that of the 19 participants in the English group, only 3 participants were found to be able to consistently accept the correct topicalization of the D-linked shei while—at the same time—also consistently rejecting the incorrect topicalization of the non-D-linked shei in their L2 Chinese wh-questions.Footnote ⁸ Although this suggests that Chinese wh-questions with the D-linked shei topicalized are not completely unacquirable by English speakers, the fact that only a very small number of very advanced learners managed to do this indicates that this is a vulnerable area in English speakers’ L2 Chinese grammars.