L2 acquisition of Wh-interrogatives at the syntax-discourse interface: interface hypothesis again

1. Introduction

Recent second language acquisition (SLA) research has witnessed a remarkable shift in scope and orientation, focusing more on the acquisition of interfaces of syntax with other modules of the grammar rather than on the syntax proper. Studies conducted in this line of thought focused on examining second language (L2) acquisition of two or more typologically different languages. The interest in examining two or more typologically different languages in such bilingual SLA studies is to see to what extent these (typological) differences cause difficulty for L2 learners. The differences may include a “partial structural overlap across the two languages where language A uses construction X in context X and construction Y in context Y, while language B uses construction X in both context X and context Y” (Sorace and Serratrice 2009: 196). In other words, such studies aim to identify the influence of this bilingual overlap and whether, and to what extent it causes non-convergence from native or near-native L2 endstate grammar. However, the focus of these studies may be different in the sense that each study may focus on one particular aspect to investigate. For instance, some studies focus on the ability of L2 learners to re-set the parameters according to L2 they are acquiring (see e.g. Belletti et al. 2007). Some others investigate L2 learners’ residual optionality in interpreting uninterpretable features in their endstate L2 grammars (see e.g.Tsimpli and Dimitrakopoulou 2007). And some others aim to find out where these difficulties lie and which module of the grammar they belong to (see.e.g., Sorace 2005, 2006; Sorace and Serratrice 2009): do they belong to syntax proper, or syntax interfaces with other cognitive domains?

This study examines the availability of the Interface Hypothesis (IH), first proposed by Sorace and colleagues (Sorace and Filiaci 2006) that L2 learners’ endstate grammars are vulnerable at the syntax-discourse interface by first language (L1) Yemeni Arabic (YA) learners, who acquire English as L2, through examining the participants’ rejection of resumptive pronouns in English ungrammatical wh-interrogatives and acceptability of gaps/empty categories in grammatical ones. The study concerns primarily two sets of wh-interrogatives: grammatical and ungrammatical, at two main levels, namely (D-)iscourse linking and non-d-linking conditions, and in terms of subject and object contexts. It involves three groups of participants (n=41): intermediate learner group (IMG, n=17), advanced learner group (AVG, n=14) and control group (NSG, n=10). The intermediate learners are BA English students, at their fourth year, and the advanced learners are MA English students (after completing three semesters of courses, and before commencing writing their MA Theses). Both learner groups are native speakers of YA, and the criterion followed to determine their proficiency in English is the Online University of Michigan grammar (placement) Test (OUMG) parts 1-4. The control group are native speakers of (British) English, MA students at the University of Essex.

The reason why I study this phenomenon is threefold: first, YA is a null/pro subject language, while English is not.¹ Therefore, wh-question formation mechanism is expected to be different in both languages. For example, while YA uses base-generation strategy, English uses wh-movement strategy.² These two different mechanisms lead to two different resumption options: YA accepts resumptive pronouns in wh-interrogatives, English does not. Second, since YA typologically differs from English, YA (and its speakers) provides a good model for an experimental study. Thus, it is expected that L1 (i.e.YA) interference may have a role to play in the difficulty encountered by YA learners of L2 English. And third, the paucity of studies in Yemeni context that tackle this important aspect of L2 acquisition research. To the best my knowledge, there has been no study in the literature investigating L2 acquisition at the syntax-discourse interface by L1 YA learners of English as L2. Thus, this study serves to fill this gap, by giving empirical support to the IH in SLA and contributing to the ongoing controversy on this unprecedented hypothesis.

The study not only provides strong support to the IH’s findings, but also for its further refinements that, while it is relatively easy to acquire linguistic phenomena related to the core syntax and those related to syntax and internal interfaces such as syntax-semantics, acquiring linguistic phenomena related to syntax and external interfaces like syntax-discourse is vulnerable even at advanced stages of L2 acquisition. The results show that whereas both groups, specifically advanced learners, performed near-native like at the syntax proper (non-d-linking condition), they were far short of native or near-native competence at the syntax-discourse interface (d-linking condition). Their performance was also better in object condition than in subject condition.

The article is set up as follows. Section 2 lays out the theoretical foundations of SLA at the interfaces, reviewing some important related studies, specifically those supporting the IH. Section 3 articulates the typological syntactic differences between wh-interrogative formation in YA and English, placing emphasis specifically on the wh-mechanism and resumption strategy used in both languages. Section 4 spells out the design and methods of the study, focusing among other things on the participants of the study, the nature of the data utilized, instrument used and analysis of the data collected from the questionnaires administered to the three groups. Section 5 presents the results, tabulating these results in terms of d-linked and non-d-linked, and subject and object wh-interrogatives, and presenting the One-way ANOVA and Post hoc Scheffe analyses of the findings at the level (p = .05). Section 6 discusses the results reached in terms of three aspects, namely learners’ performance in d-linking versus non-d-linking conditions, their performance in object versus subject conditions, and Access to UG and L1 interference, and Section 7 concludes the article.

2. L2 acquisition at the interfaces

Within earlier generative approaches to syntax, Principles and Parameters (P&P) framework,³ for example, SLA studies focused on the role of Universal Grammar (UG), and whether L2 leaners still have access to it in the acquisition of L2 (or even Ln). There are actually three positions: i) Full Access (see e.g. White 1986, Cook 1988; Flynn 1996; Shormani and AlSohbani 2015), ii) Partial Access (see e.g. Hawkins and Chan 1997; Smith and Tsimpli, 1995), and iii) No Access (see e.g. Bley-Vroman 1989).⁴ While it was somehow difficult in P&P Framework to account for the difficulties encountered by (adult) L2 learners, and more importantly where (exactly) they lie within L2 grammars (i.e. either within the core syntax or within its interfaces with other high cognitive domains that are not part of the core grammar), recent conceptions and developments in generative approaches to syntax such as the Minimalist Program (Chomsky 2000, 2008), cartography approaches (see e.g. Cinque 2002; Rizzi 2004; Belletti et al. 2007),⁵ discourse interface studies (see e.g. Reinhart 2006; Erteschik-Shir 2007; Frascarelli 2007; Shormani 2017) explicitly assist SLA scholars to account for the non-ultimate (or ultimate) attainment, the failure of native-like competence.

From a minimalist perspective, consider feature (un)interpretability; for instance, it is held that interpretable features at the syntax–discourse interface involve considerable difficulty for L2 learners (Sorace 2005). For others, (see e.g. Tsimpli and Dimitrakopoulou 2007), however, it is uninterpretable features that constitute considerable difficulty for L2 leaners at the syntax-discourse interface, which results in variability in L2 endstate grammar. It was reported that these higher interface domains can also be extended to interpret “delays” experienced in L1 acquisition by children (e.g. Platzack 2001) and L1 attrition (e.g. Chamorro and Sorace 2018). Given this, it was not surprising to find evidence of variability or residual optionality in adult L2 leaners’ endstate grammar (see e.g. Sorace 2000, 2005, 2006, 2011, 2016; Tsimpli and Dimitrakopoulou 2007; Tsimpli and Mastropavlou 2007; Sorace and Serratrice 2009, Sorace et al., 2009).

Cartography approaches and discourse studies also help scholars identify where the difficulty encountered by L2 learners lies, in that they make explicit the structural representations beyond the syntax proper, or relating it to other cognitive domains. For example, cartography approaches make available linking the syntax proper with discourse/pragmatics interface (see e.g. Shormani 2017). Discourse has been identified in the left-periphery of constituent structure, specifically with the informational/coda structure, while the syntax proper with the propositional structure, the C-domain and T-domain, respectively (see e.g. Rizzi 1997, 2004, 2006; Cinque 2006; Shormani 2017; Shormani and Qarabesh 2018). To represent the discourse, the CP layer has been proposed to project to three (and sometimes four including FiniteP) projections, namely ForceP, TopicP, and FocusP (Rizzi 1997, 2004, 2006; Cinque 2006), and sometimes even SpeakerP/AddresseeP (see e.g. Espinal 2013). In Rizzi’s (1997: 283) own words, the C-domain is “the interface between a propositional content (expressed by IP) and the superordinate structure (a higher clause or, possibly, the articulation of discourse, if we consider a root clause).” Thus, TopicP, for instance, has been referred to as a projection structurally representing left-dislocation phenomena such as clitic-left dislocation, or topicalization, which is a discourse property. Several studies have tackled this aspect to account for the difficulty encountered by L2 learners at the syntax-discourse interface (see e.g. Tsimpli and Sorace 2006; Dugarova 2014; Chamorro and Sorace 2018; Smeets 2018).

Thus, such recent conceptions and developments give SLA scholars enough room to interpret the difficulties encountered by L2 learners, as well as L1 acquirers alike, like the residual optionality, deficit parsing or misanalyzing of L2 structures, variability, permanent fossilization, etc. in adult L2 acquisition (cf. also Long 2003; Shormani 2013). For example, Sorace and colleagues (Sorace and Filiaci 2006) propose the Interface Hypothesis in an attempt “to explain the non-convergence and optionality revealed in very advanced adult second language learners in the comprehension and production of certain structures” (Chamorro and Sorace 2018: 2). The interface Hypothesis states that learners’ endstate knowledge of L2 is vulnerable and this vulnerability is due to interface effects (Sorace and Filiaci 2006; Sorace and Serratrice 2009). That L2 advanced learners encounter considerable difficulty acquiring linguistic phenomena at the syntax-discourse interface, according to Sorace and colleagues, is due to the fact that this interface is a higher level of cognitive domains which requires high processing to be acquired.

The term Interface has been used to signify a level at which a module of the grammar is linked with other cognitive domains such as syntax-semantics, syntax-morphology, syntax-discourse (see e.g. Sorace 2005, 2006, 2011, 2016; Sorace and Serratrice, 2009). There are two types of interface: internal and external. The former refers to “those between narrow syntax and the other linguistic modules (phonology, morphology, semantics) and external interfaces, those between syntax and other cognitive modules” (Slabakova and Ivanov 2011: 638). Put differently, internal interfaces are said to be between internal modules of the grammar such as syntax-semantics, and syntax-morphology, while external interfaces take place between syntax and external cognitive domains such as syntax-discourse, syntax-pragmatics.

As for the difference between internal and external interface, Tsimpli and Sorace (2006: 653) postulate that “[t]he distinction between the two interfaces is based on the assumption that the syntax-discourse interface is a ‘higher’ level of language use, integrating properties of language and pragmatic processing, whereas syntax-semantics involve formal properties of the language system alone” (emphasis mine). Furthermore, Sorace and Serratrice (2009) argue that the difference between internal and external interfaces lies in that the former involves operations internal to the ‘core computational system’, while the latter involves operations external to this computational system.⁶ In their own words, “the syntax-semantic interface involves formal features and operations within syntax and Logical Form, whereas the syntax-discourse interface involves pragmatic conditions that determine appropriateness in context” (p. 197). The overall view held regarding these two types of interfaces, and which causes difficulty for L2 learners is that external interfaces are more difficult to acquire than internal ones. Sorace (2011: 1) writes: “language structures involving an interface between syntax and other cognitive domains are less likely to be acquired completely than structures that do not involve this interface” (see also Chamorro and Sorace 2018).⁷ The claim is that external interfaces require high processing because they lie out of the core computational system, as has been alluded to above. In what follows, I will discuss some example studies whose results provide support to the IH.

The IH effects, as well as the difference between internal and external interfaces, were supported by empirical studies cross-linguistically with a wide range of L1s and L2s (see e.g. Sorace 2006, et seq; Tsimpli and Sorace 2006; Sorace and Serratrice 2009; Belletti et al. 2007; Antonova-Ünlü 2015; Chamorro and Sorace 2018; Smeets 2018). For example, Tsimpli and Sorace (2006) investigate the interface hypothesis in a study involving 27 adult Russian learners of Greek. They investigated learners’ ability to distinguish between focus (a syntax-semantics interface property) and clitic-left dislocation, or topicalization (a left-periphery, or more specifically a syntax-discourse interface property). The results indicate that the leaners involved were able to analyze focus data, but not left-dislocated clitics. They conclude that while the syntax-semantics interface is relatively (completely) acquirable, the syntax-discourse interface is vulnerable.

Furthermore, Sorace and Filiaci (2006) conducted a study examining the Anaphora resolution in near-native speakers of Italian. They recruited a control group of monolingual Italian native speakers and a group of native English speakers who had attained near-native proficiency in Italian. Their study supported the IH, pointing out that the residual optionality they found is still persistent at the syntax–discourse interface, which may be attributed to “indeterminacy”, failing to link pronouns to their antecedents. Similar findings were also reported by Belletti et al. (2007) who conducted a study to examine the null subject parameter (which is a discourse related phenomenon) pro versus overt subject pronoun in L2 Italian by 17 learners whose native language is (American) English. They concluded that there is significant divergence between L2 learners and native speakers, attributing this divergence to two sources: i) the visible outcome is the overproduction of overt subject pronouns, and the misinterpretation of overt subject, and ii) the unsystematic use of the low focus position in the VP-periphery of the clause dedicated to the new-information postverbal subjects (Belletti et al. 2007: 682). Further, Antonova-Ünlü (2015) conducted a study whose focus was L2 acquisition of Case markers in Turkish by highly proficient learners of L1 Russian. According to Antonova-Ünlü, Case markers in Turkish involve syntax-semantics/morphology interface and syntax-discourse interface, the latter of which requires pragmatic understanding for accurate interpretation. She concluded that while L2 learners easily acquire syntax-semantics/morphology interface materials, syntax-discourse interface linguistic phenomena, however, were vulnerable.

In addition, Tsimpli and Dimitrakopoulou (2007) conducted a study on the acquisition of English wh-interrogatives by L1 Greek learners. In this study, the authors examined the residual optionality in L2 acquisition, proposing the Interpretability Hypothesis, which states that uninterpretable features, but not interpretable features, constitute considerable difficulty for L2 learners. The study recruited two groups of learners, 21 intermediate and 27 advanced, and 26 native speakers of English, as a control group. The data utilized consisted of 30 wh-interrogatives (and 21 distractors). They examined resumption, d-linking and animacy effects on the responses of their participants. They found that while non-d-linked wh-questions were relatively easy to learn by both learner groups, d-linked ones were difficult. They concluded that L2 acquisition is vulnerable at the syntax-discourse interface, thus adding strong support to the IH. In addition, Smeets (2018) conducted a study examining near-native grammars at the syntax–discourse interface of two different domains of object movement in Dutch exhibiting two properties lied within syntax-semantics and syntax-discourse interfaces. In this study, the author recruited 56 participants, 15 of which were native speakers of Dutch, 16 native speakers of German and 25 native speakers of English. Her instrument was felicity judgment tasks and a truth value judgment task. The results showed that while syntax-semantics interface properties were easy to acquire, syntax-discourse were not so. However, she opines that for L2 convergence, discourse cues should be sufficiently salient in the input. She added that due to L1-L2 typological differences, “L2 learners may not fully acquire L2 discourse conditions on syntax when their L1 allows the relevant syntactic construction as well but does so in different discourse settings” (p. 21).

3. Wh-interrogatives in English and Arabic

Recall that YA is a null subject, or pro-drop language, while English is not. Therefore, it is expected that both languages differ in the way each forms wh-interrogatives. I will discuss the differences between English and YA in terms of d-linked [+DL] and non-d-linked [-DL] wh-interrogatives.⁸ Section 3.1 addresses the former, and Section 3.2 tackles the latter.

4. The present study

5. Results

The results collected from the acceptability judgment from the three groups are presented in Tables 5 and 6. The results are presented in two formats according to the conditions, i.e. d-linking and non-d-linking, and subject and object. I first count the participants’ responses for the d-linked wh-interrogatives in both subject and object contexts. Then I count the responses for the non-d-linked wh-interrogatives also in both subject and object contexts. For differences between groups, I used One-way ANOVA analysis for the differences between groups, and Post hoc Scheffe test was used for differences within groups.

As Table 5 shows, the IMG participants performed less significantly than the AVG participants in both subject and object d-linked wh-interrogatives [IMG: (subject) 42.4%; and (object) 54.1%; AVG: (subject) 58.6%, and (object) 65.7%, accuracy]. It is also clear that both learner groups performed better in object d-linked wh-interrogatives than in subject d-linked ones. However, both groups differ significantly from the control group.

Table 6 shows that both learner groups do not differ significantly; they both fare somehow better than in d-linking condition, particularly in object contexts [IMG: (subject) 68.2%, and (object) 78.9%; AVG: (subject) 82.8%, and (object) 87.2%, accuracy].

The overall results in both d-linking and non-d-linking conditions, presented in Tables 5 and 6, show that the control group performed as expected. They rejected the resumptive pronouns in ungrammatical sentences and accepted gaps in grammatical ones. Both learner groups fare in non-d-linked better than in d-linked wh-interrogatives. This shows that non-d-linked wh-interrogatives are easier to acquire than d-linked ones, which in turn implies that they have encountered less difficulty in the syntax proper data than that at the syntax-discourse interface.

As for the d-linking condition, the two experimental groups performed significantly less than the control group (cf. Table 5). A One-way ANOVA performed on the d-linked wh-interrogatives revealed statistically significant differences between the control group and each of the learner groups on both subject and object conditions F(2, 38) = 91.248, p <.001], and F(2, 38) = 19.961, p < .001], respectively.

For comparisons within groups, Post hoc Scheffe test indicated high statistically significant differences between the control group and each of the learner groups [IMG: (subject) MD = 4.364, p <.001, p <.05; (object) MD = 5.723, p <.001, (p<.05); AVG: (subject) MD = 2.417, p =.01, (p <.05), (object) MD =3.351, p =.01)].¹²

As for the learner groups, there were significant differences between the intermediate and advanced learners in both subject and object conditions [IMG: (subject), MD =3.907, p <.001, (p<.05), (object) MD = 3.037, p <.001), (p<.05); AVG: (subject) MD =3.015, p <.001), (p <.05), (object) MD = 1.907, p = .01, (p <.05)].

The above findings in terms of the d-linking condition clearly indicate that both learner groups performed far short of the control group. Put differently, both learner groups seem to accept resumptive pronouns in d-linking condition in ungrammatical wh-interrogatives, and reject gaps in grammatical ones, which results in bad performance. They also show clearly that the advanced group performed better in the object condition than the intermediate group.

Turning now to the non-d-linking condition, the intermediate learners and the advanced learners fare better in the object condition. However, they both differ significantly from the control group, though they performed better than they did in the d-linked wh-interrogatives (cf. Table 6). A one-way ANOVA conducted on the non-d-linked wh-interrogatives indicated that there were statistically significant differences between the control group and the learner groups each [F(2, 38) = 88.985, p <.001] for subject, and [F(2, 38) = 4.119, p <.024] for object.

As for comparison within groups, Post hoc Scheffe test revealed statistically significant differences between the control group and the intermediate group [IMG: (subject) MD= 9.623, p <.001, (p<.05); (object) MD = 1.764, p =.01, (p < .05)]. However, there were no significant differences between the control group and advanced group [AVG: (subject) MD= .481, p = 1, (p >.05), (object) MD = .341, p = 1, (p >.05)].

This is an important finding of this study, because it highlights the better performance of the advanced group at the syntax proper. That is to say, they performed near-native like in judging the non-d-linked wh-interrogatives.

However, Scheffe test also revealed that there were no statistically significant differences between the intermediate group and advanced group [IMG: (subject) MD=1.714, p =.058, (p >.05), (object) MD =.514, p =1, (p >.05); AVG: (subject) MD=.462, p =1, (p>.05), (object) MD = .317, p = 1, (p >.05)].

This is also a significant finding due to the fact that both groups, intermediate and advanced learners, seem to perform near-native like in wh-interrogatives. This implies that the linguistic phenomena related to the syntax proper is easier to acquire.

The overall findings reached regarding the d-linking condition suggest that: i) both learner groups, though with a varying degree in favor of the advanced group, performed better in the non-d-linked than they did in the d-linked wh-interrogatives, indicating that (a) they have acquired the syntax proper linguistic phenomena, and (b) the fact that the advanced learners performed like native speakers indicates that they face no difficulty in processing the non-d-linked wh-structures, ii) both learner groups, particularly the advanced learners, performed much better in both ungrammatical and grammatical wh-questions, which suggests that they rejected resumptive pronouns and accepted gaps, respectively, relatively close to native speakers’ performance in the non-d-linking condition, and iii) the fact that the advanced group performed like native speakers in the non-d-linked, but far less than them in the d-linked wh-questions shows clearly that they have acquired the proper way of using the [-DL], but not [+DL] features, and iv) the fact that both groups performed far better in the non-d-linked than in the d-linked wh-questions indicates that they encountered considerable difficulty in processing or analyzing linguistic phenomena in the d-linking condition, or more specifically, at the syntax-discourse interface.

6. Discussion

The overall results reached revealed that the intermediate and advanced learners performed significantly differently from the native speakers in the d-linking condition, and significantly far less than they did in the non-d-linking condition, in both subject and object contexts. This indicates that both groups accepted the resumptive pronouns in the ungrammatical, but did not accept gaps in the grammatical wh-interrogatives. This otherwise clearly indicates that they were successful in acquiring the syntax proper wh-questions, but were not so in acquiring the syntax-discourse ones, thus supporting the IH. Thus, I will discuss the results of the study in terms of:

As for (i), given the statistics in section 5 above, briefly repeated below for convenience, both learners performed in the d-linking condition statistically significantly less than the control group [IMG: (subject) 42.4%; and (object) 54.1%; AVG: (subject) 58.6%, and (object) 65.7%, accuracy] (cf. Table 5). It is also clear that advanced learners performed better than intermediate learners. A One-way ANOVA performed on the d-linked wh-interrogatives revealed statistically significant differences between the control group and each of the learner groups both in subject and object conditions [F(2, 38) = 91.248, p <.001], and F(2, 38) = 19.961, p < .001], respectively.

As for the non-d-linking condition, both learner groups performed much better than they did in the d-linked wh-questions [IMG: (subject) 68.2%, and (object) 78.9%; AVG: (subject) 82.8%, and (object) 87.2%, accuracy] (cf. Table 6). This means that both learner groups have acquired the syntax proper in which these non-d-linked wh-interrogatives are processed and analyzed, more specifically in object contexts. A one-way ANOVA conducted on the non-d-linked wh-interrogatives indicated that there were statistically significant differences between the control group and the learner groups each [F(2, 38) = 88.985, p <.001] for subject, and [F(2, 38) = 4.119, p <.024] for object.

Post hoc Scheffe test, however, shows that there were no significant differences between the control group and advanced group [AVG: (subject) MD= .481, p = 1, (p >.05), (object) MD = .341, p= 1, (p >.05)]. As noted so far, this is a very significant result which clearly highlights the fact that both learner groups performed near-native like in non-d-linked constructions, which in turn indicates that they have acquired the syntax proper relatively well.

Regarding (ii), the fact that both learner groups’ performance in object condition is far better than in subject condition is not surprising. Given the syntactic facts of YA: i) YA is a null subject language, ii) YA does not use wh-movement strategy, but rather base-generation one, and iii) YA does not exhibit that-t effects, this finding is again not surprising. Given the base-generation strategy utilized by YA, and the assumption that resumptives are the spell out of (uninterpretable) agreement features and Case (cf. Alexiadou and Anagnostopoulou 1998), resumptive pronouns are likely to persist in L1 YA learners’ endsate grammars, or their Interlanguage (cf. Lardiere 2009). Thus, when the wh-word/phrase is base-generated in the topmost Spec, CP, these agreement features and Case are likely to spell out as resumptives, which is not the case in English, which in turn indicates that learners will have a “tendency” to accept these resumptives in their English production.

The fact that YA does not exhibit the that-t effects results in YA learners being unsensitive to such effects in English. If this is on the right track, the learners will accept resumptives in ungrammatical wh-questions, and will not accept gaps in grammatical ones. Given that that-t effects are related to subject wh-questions, but not to object ones, their bad performance in subject wh-questions could be a direct result of that.

Another possible interpretation of this subject-object asymmetry is to look at the nature of the object wh-questions in English, i.e. English wh-questions allow the complex ‘that-he/it’, which is not different from YA. Examples in (3a) and (3b) illustrate the point in question.

(3) a. ʔayyan dars qaal ʕali *(ʔinn-uh) katab*(-uh)?      D-linked

    which lesson said  Ali that-he    wrote-it

    ‘Which lesson did Ali say that he wrote (*it)?

  b. mu qaal ʕali *(ʔinn-uh) ?aštri*(-uh) ?           Non-D-linked

    what said Ali that-he bought-it

    ‘What did Ali say (that) he bought (*it)?

Thus, the only difference between YA and English in object -wh-questions turns to be the obligatory-optional use of the resumptive pronoun it, which suggests that YA learners find object wh-questions easier to process/acquire than subject ones. This may be thought of as a “facilitator” aiding the learners to perform better in object wh-questions than in subject ones. In fact, the subject-object asymmetry found in our study provides support to previous studies (see e.g. Tsimpli and Dimitrakopoulou 2007, and the references there).

As for (iii), considering the nature of the data utilized, they are complex wh-interrogatives, involving long-distance wh-movements, because they each consist of a matrix and complement clause. These complex wh-interrogatives involve resumptive pronouns “as a saving device to redeem derivations and structures that would otherwise violate a fundamental principle of the grammar” (Rouveret 2011: 2, see also Leal Méndez and Slabakova 2012). These structures are expected to be more difficult to process or analyze by the learners, specifically if the L1 does not use wh-movement strategy and allows resumptive pronouns. Put differently, the wh-movement in these complex structures will undergo several cyclic wh-movements from the base position to the topmost Spec, CP. Thus, learners will face considerable difficulty in processing these A’-chain wh-questions much more than that faced in processing simple wh-questions (i.e. one-clause wh-questions).¹³

At the level of the syntax proper, i.e. non-d-linked wh-questions, L2 learners in this study performed approximately well. The fact that they performed well, at least at the syntax proper indicates that the learners, specifically the advanced learners, have acquired how resumptive pronouns are used in English (non-d-linked) wh-interrogatives, by rejecting these pronouns in ungrammatical sentences and accepting gaps/empty categories in grammatical ones. This is evidence that the UG activated by the learners’ L1 (before puberty, cf. Pinker 1995; Shormani 2014c) has worked its path in L2 acquisition, and is still accessible, allowing them to re-set the resumptive parameter according to L2 grammar in the syntax proper. Form a minimalist perspective, L2 learners were able to use Merge and Agree operations properly. As for Merge, one could argue that learners, specifically the advanced ones, were able to apply the internal Merge (wh-movement) accurately, and the wh-movement was applied cyclically, involving several wh-movement stances. Assuming with Chomsky (2000, 2001), for instance, that the head C has an EPP (Extended Projection Principle) (or [uwh]) feature, EPP triggers the wh-word/phrase to remerge in Spec, CP, and the learners seem to have acquired it. Regarding Agree, it is possible to argue that L2 learners have acquired where to or not to apply it in YA and English wh-questions. That is to say, while Agree applies in YA resulting in resumption, learners did not apply it in English (non-d-linked) wh-questions.

However, at the syntax-discourse interface, L2 learners involved in this study fail to apply the parameter re-setting, or Merge and Agree operations successfully. In other words, L2 learners were not able to re-set the parameters of L1 according to L2, in such a way as to reject resumptive pronouns in their L2 grammars at the syntax-discourse interface, i.e. when discourse comes to play. The fact that [+DL] features are imposed by discourse such as discourse (co)referentiality (see e.g. Shormani 2017), context, situation, etc. in resumption, for instance, to disambiguate a grammatical structure, or what has been referred to by Tsimpli and Dimitrakopoulou 2007) as “garden-path effects”, manifests itself clearly in their L2 vulnerability. This suggests that L2 learners’ grammar cannot reach endstate or native speakers’ competence, which also suggests that access to UG is partially restricted, the reason of which might be attributed to L1 interference, among other reasons.¹⁴ One could assume that though these L2 learners have reached high level of learnability in I-language of L2 (cf. Chomsky 2013), they were just unable to re-set the parameters concerning the use of resumptive pronouns in L2 English. From a minimalist perspective, it seems that L2 learners even at advanced levels are not able to distinguish where to apply the operations Merge and Agree appropriately in their performance. Put simply, they were unable to figure out that a resumptive pronoun cannot be merged with a moved wh-word/phrase in L2 English, and while Agree is spelled out as a resumptive pronoun in YA (cf. Alexiadou and Anagnostopoulou 1998, see also Tsimpli and Dimitrakopoulou 2007, for Greek), it is not the case in English wh-questions at the syntax-discourse interface. Having this deficit characterization of wh-interrogatives at the syntax-discourse interface, L2 learners fail to perform like native speakers, as predicted by the IH.

This makes it clear, though, that L1 interference may play a role in the learners’ performance. That is, the difficulty encountered by L2 learners could be attributed to L1 interference in acquiring linguistic phenomena relating to the syntax-discourse interface. Or, L2 learners find it difficult to process and analyze a particular structure involving discourse-domain related phenomena, thus, resulting in resorting to their L1, as a “subconscious” strategy to overcome such difficulties (Shormani 2014a &b). The resumption strategy of YA could be said to have been transferred to L2 counterpart structures, hence resulting in ungrammatical performance. Assuming with Alexiadou and Anagnostopoulou (1998) and Tsimpli and Dimitrakopoulou (2007) that features like subject-verb agreement and Case are uninterpretable features and spell out as resumptive pronouns, it could be argued that these features are transferred from L1 YA to L2 English at the syntax-discourse interface. This L1 interference could also be said to make L1-L2 counterpart wh-structures “difficult to identify and analyse in the L2 input due to persistent, maturationally-based, L1 effects on adult L2 grammars” (Tsimpli and Dimitrakopoulou 2007: 217).

7. Conclusions and further implications

This study examined L2 acquisition at the syntax-discourse interface by L1 YA learners acquiring English wh-interrogatives, at both d-linking and non-d-linking conditions, or syntax proper and syntax-discourse interface, respectively. The aim was to answer one major question: do linguistic phenomena relating to syntax-discourse interface constitute difficulty for Yemeni learners of English? According to the results reached, it is clear that the answer is yes; L1 YA learners’ competence of English at the syntax-discourse interface is still vulnerable.

In our corpus, I found strong support to IH predictions, i.e. L2 acquisition is vulnerable at the syntax-discourse interface. Both groups of learners, intermediate and advanced, involved in this study found considerable difficulty in acquiring the wh-interrogatives at the syntax-discourse interface. However, they performed significantly better in the syntax proper, i.e. the non-d-linked wh-questions. Given this, our study, thus, provides strong support to previous studies on this phenomenon (see e.g. Tsimpli and Sorace 2006; Sorace and Filiaci 2006; Tsimpli and Dimitrakopoulou 2007; Tsimpli and Mastropavlou 2007; Sorace and Serratrice 2009; Sorace 2005, 2006, 2016).

In addition, our study provides important findings that could be employed by university teachers while prescribing textbooks for their students, or even while their teaching of syntax courses. In the latter aspect, university teachers are advised to involve comprehensive comparisons of YA with English wh-question formation in both languages, concerning specifically the different wh-mechanism and resumption strategies utilized by each language. In this way, teachers could provide students with linguistic input sufficient for acquiring such differences, and students would be made aware of these differences between their L1 and L2 they are acquiring from early stages of their acquisition, and more specifically before these structures get fossilized.

To conclude, it is true that the results of this study support numerous previous studies, but these studies were conducted in different contexts, different L1s, and L2s, different from those of this study. Thus, it goes without saying that further research be conducted to see whether the results concluded with in this study are supported, not only concerning wh-interrogative acquisition, but also including acquisition of other linguistic phenomena such as articles, focus, topicalization, subject-verb agreement, etc. in English, and I leave these for future research.