Make your own free website on Tripod.com

The status of functional categories in child second language acquisition: evidence from the acquisition of CP Belma Haznedar BogI aziçi University This study examines the status of the functional categories in child second language (L2) acquisition of English. Results from longitudinally collected data are reported, presenting counterevidence for recent hypotheses on early L2 acquisition that assume (1) a structure-building approach according to which the acquisition of functional categories follows an implicational sequence of development of VP–IP–CP (Vainikka and Young-Scholten, 1994; 1996a; 1996b; 1998); and (2) a direct relationship between the acquisition of inflectional morphology and the development of functional categories (Eubank, 1993/94; 1996; Vainikka and Young-Scholten, 1994; 1996a; 1996b; 1998).The child L2 data analysed in this article show that the development of CP is not implicationally contingent on the prior acquisition of IP. The data also suggest that the lack of morphological forms in interlanguage grammars reflects a problem with the realization of surface morphology, rather than an impairment in the domain of functional projections. I Introduction A major issue in recent second language (L2) acquisition research revolves around the status of functional categories in early stages of non-native grammars (e.g.,Grondin, 1992;Eubank, 1993/94; 1996; Lakshmanan, 1993/94; Lakshmanan and Selinker, 1994; 1996; Schwartz and Sprouse, 1994; 1996; Vainikka and Young-Scholten, 1994; 1996a; 1996b; 1998; Gavruseva and Lardiere, 1996; Grondin and White, 1996; Henry and Tangney, 1996; White 1996; 1996; Haznedar, 1997; Kakazu and Lakshmanan, 2000; Haznedar, 2001). Much work on functional categories in L2 acquisition over the last several years has addressed the question of whether or not functional projections such as DP, IP and CP are present in the learner’s grammar from the start of acquisition. For some researchers, they are initially absent in interlanguage grammars and gradually develop in discrete stages on the basis of input (Vainikka and Young-Scholten 1994; 1996a; 1996b; 1998). As opposed to this © Arnold 2003 10.1191/0267658303sr212oa Address for correspondence: Bog aziçi Üniversitesi, Eg itim Fakültesi,Yabanc‡ Diller Eg itimi, Bebek-Istanbul 80815, Turkey; email: haznedab@boun.edu.tr Second Language Research 19,1 (2003); pp. 1–41 Weak Continuity hypothesis, for others – who postulate Strong Continuity in interlanguage grammars – functional categories and their projections are available in the early grammars of both adult and child L2 learners (e.g., Lakshmanan, 1993/94, Lakshmanan and Selinker, 1994; Schwartz and Sprouse, 1994, 1996; Gavruseva and Lardiere, 1996; Grondin and White, 1996; Haznedar, 1997; Prévost, 1997; Haznedar, 2001). Another group of researchers has focused on the extent to which L2 grammars diverge from the target language in terms of the representation of functional projections. Among various positions, one view holds that interlanguage grammars reveal properties found neither in the first language (L1), nor in the L2, nor in any natural language, and are permanently impaired in terms of certain properties of functional categories (e.g., Beck, 1998; Eubank and Grace, 1998).Another proposal is that L2 learners are restricted to L1 categories and features. According to Smith and Tsimpli (1995) and Hawkins and Chan (1997), for instance, L2 learners cannot acquire new functional categories or features, although they have access to L1 categories. Such claims for defective interlanguage grammars contrast with proposals that L2 grammars are not impaired, rather L2 functional categories and feature values are attainable (e.g., Schwartz and Sprouse, 1996; Duffield and White, 1999). In current research, discussion focuses particularly on the failure of L2 learners to produce verbal inflectional morphology associated with functional categories. The question is whether the frequent omission of verbal inflection means that functional categories are impaired in L2 grammars. In some of these studies – Eubank (1993/94; 1996), Vainikka and Young-Scholten (1994; 1996a; 1996b; 1998) and Eubank and Grace (1998), in particular – the absence of target-like inflectional suffixes has been taken as the primary evidence for concluding that L2 learners do not project associated functional features or categories. As opposed to this view, a number of L2 researchers postulate the Missing Surface Inflection hypothesis, according to which the lack of (or the variable use of) morphological forms in interlanguage grammars reflects a problem with the realization of surface morphology, rather than an impairment in the domain of functional projections or feature strength (Haznedar and Schwartz, 1997; Lardiere, 1998a; 1998b; 2000; Prévost and White, 2000; Haznedar, 2001). This article presents data that appear to be difficult to explain within a syntactic or feature impairment model, such as Vainikka and Young-Scholten’s successive stages theory, where the 2 Functional categories and evidence from CP acquisition Belma Haznedar 3 development is supposed to depend on the acquisition of some functional elements, as reflected in overt morphology. The L2 data from a Turkish-speaking child, Erdem, who learned English as an L2 indicate that his interlanguage grammar includes extended phrase structure despite impoverished inflectional morphology. Based on the findings, I hope to show that the gradual structurebuilding approach appears to be problematic for L2 acquisition. Rather, these results provide evidence for the position that during the course of acquisition, morphological and syntactic development can be independent from each other (Lardiere, 1998a; 1998b; 2000). The organization of the article is as follows. Section II presents a review of three current hypotheses in L2 acquisition: Minimal Trees hypothesis (Vainikka and Young-Scholten, 1994; 1996a; 1996b), Weak Transfer / Valueless Features hypothesis (Eubank, 1993/94; 1996) and Missing Surface Inflection hypothesis (Haznedar and Schwartz, 1997; Lardiere, 1998a; 1998b; 2000; Prévost and White, 2000; Haznedar, 2001). Next a brief overview of the assumptions underlying the CP system in English is discussed. Section III examines the child L2 data analysed in this article, starting with the development of yes–no questions and whquestions. Then, further data on the development of embedded clauses are discussed. Section IV presents the results and analyses of the data, focusing on Vainikka and Young-Scholten’s claims for the development of phrase structure in L2 acquisition. Finally, Section V presents a summary of our arguments. II Background 1 Minimal Trees hypothesis Vainikka and Young-Scholten (V&Y-S) (1994; 1996a; 1996b; 1998) extend the Weak Continuity hypothesis to L2 acquisition of German, arguing that only lexical categories (such as the VP) are initially available to the L2 learner. For V&Y-S, L2 learners then in sequence project the relevant functional categories gradually. Similar to the models developed for German L1 acquisition (e.g., Clahsen et al., 1994), V&Y-S argue that the clause structure is first projected only to VP, a stage termed as the bare-VP stage. The subsequent acquisition of functional projections is input-driven, resulting in a sequence of development in which, first, an early underspecified functional projection FP is acquired and then a fully specified IP or AGRP. The AGRP stage is later followed by a CP stage. As in similar L1 acquisition research, evidence for these successive stages depends heavily on overt production of functional elements such as modals, auxiliaries, verbal morphology, whquestions, I-to-C movement and complementizers. For V&Y-S, the move from the VP stage to the AGRP stage results in the productive use of auxiliaries and modals, as well as correct verbal agreement morphology. At the AGRP stage, CP is argued to be in the process of being developed, as only one or two instances of embedded CP with overt complementizers are used. What this means is that functional categories are acquired through a developmental sequence of VP–IP–CP in which the emergence of IP requires the acquisition of VP and the development of CP is dependent on the previous acquisition of IP. 2 Weak Transfer / Valueless Features hypothesis Another view on the development of functional phrase structure is Eubank’s Weak Transfer / Valueless Features (1993/94; 1996) hypothesis. Eubank’s model is similar to the Minimal Trees account in that transfer from the learner’s L1 is partial and the L2 initial state does not include all the properties of a final state grammar. Eubank proposes that both lexical and functional categories might transfer from the L1 into the L2 initial state. However, the parameter values associated with functional categories do not transfer. He specifically claims that feature values such as the strength value of agreement which determines the presence of overt verb movement do not transfer. Consequently, he stipulates that verb raising will optionally be permitted, until the relevant inflections are acquired that determine the strength of the feature. Eubank’s model predicts optionality of verb raising as an inherent property of L2 development. On the basis of data from L2 German, Eubank proposes that during the first stage, learners’ L2 grammars do not have verb movement, and hence no overt realizations of functional projections appear. At Stage 2 learners begin to acquire the target value of the verb movement parameter, but verb movement at this stage is optional and, hence, in Eubank’s account,Tense is not fully specified (for discussion of unspecified Tense in L1 acquisition, see Wexler, 1994). One important point shared by V&Y-S’s and Eubank’s accounts is the assumption that the status of abstract properties like functional features and heads in a learner’s grammar is related to the acquisition of lexical material. In other words, evidence for the presence of AGRP/ and TP/ in the grammar can be inferred from the presence of the morphological items associated with them in the learner’s interlanguage. 4 Functional categories and evidence from CP acquisition Belma Haznedar 5 Adopting a somewhat different approach, Gavruseva and Lardiere (1996) test the assumption that lack of morphological inflection constitutes evidence for the lack of underlying phrase structure. Gavruseva and Lardiere’s (G&L) (1996) analysis of data from an 8-year-old Russian child, Dasha, focuses particularly on V&Y-S’s theory of the successive stages of development: VP is acquired prior to IP, and IP prior to CP. G&L argue that from the earliest transcripts, the data provide evidence for the early emergence of CP. Dasha produces CP-related elements prior to the productive use of agreement, tense, auxiliaries and modals, which, for V&Y-S, are required for the learner to project an IP (p. 230). This result is significant because it shows that the acquisition of CP does not depend on the acquisition of IP. At the same time, the result is also important with respect to V&Y-S’s strict criterion in German L2 acquisition. It is important to question whether overt production of morphemes associated with a functional category is necessary for one to conclude that the category is acquired. This is what I specifically address in some detail in the next section. 3 Missing Surface Inflection hypothesis As we have seen, in order to account for the fact that L2 learners frequently omit verbal inflection and use nonfinite forms in finite contexts, a group of researchers attribute the lack of verbal inflection to an impairment of functional categories and/or features in L2 grammars (e.g., Eubank, 1993/94, Eubank et al., 1997; Meisel, 1997; Beck, 1998; Eubank and Grace, 1998).1 On the alternative approach, the Missing Surface Inflection hypothesis, L2 grammars contain abstract categories and features, but the problem lies in mapping from the abstract features to the morphophonological represen-tation of these features on surface morphology (Haznedar and Schwartz, 1997;Lardiere, 1998a; 1998b; 2000;Prévost and White, 2000; Haznedar, 2001). In her pioneering work, Lardiere (1998a; 1998b; 2000) argues for the position that there is a distinction between abstract morphosyntactic features, such as [+finite], and their morphophonological representations, such as third person singular-s as in She speak-s French fluently.This proposal, which is referred to as the Separation 1Note that on the basis of data from L2 learners of German, Meisel (1997) found that nonfinite forms are used in finite contexts and finite forms in nonfinite contexts. He concluded that L2 learners made no finite–nonfinite distinctions in L2 acquisition and that L2 grammars suffer from a global impairment in the domain of abstract features. On similar grounds, Eubank (1993/94), Eubank et al. (1997) and Beck (1998) argued for a local impairment according to which functional categories are available in L2 grammars, but their feature strength is impaired. hypothesis, is based on a view of morphology according to which the features associated with an affix are distinct from the phonological realization of that affix (e.g., Beard, 1987; 1988; 1993; Halle and Marantz, 1993). Lardiere (1998a; 1998b) examines the end-state of L2 acquisition of a Chinese-speaking adult learner of English who had lived in the USA for 18 years. She shows that in spite of the low level of overt morphology, the data reveal a variety of syntactic phenomena that suggest that the learner has tense and agreement at an abstract level. According to Lardiere, these findings suggest that the learner’s problems lie in the domain of morphological mapping rather than any deficit in functional projections and features. Haznedar and Schwartz (1997) and Haznedar (2001) found similar results in examining child L2 data, from a Turkish-speaking child, Erdem, who is also the subject of this article. Despite the fact that Erdem produced many nonfinite forms in his speech, he stopped omitting subjects long before the use of verbal inflection in obligatory contexts. This shows that there is not a relationship between the use of inflectional morphology and overt subjects. At a time some functional elements are missing, there is evidence for the syntactic operations that require functional features, such as auxiliary movement and subject raising in utterances with overt subjects. Similarly, Haznedar (2001) presents robust evidence for the use of only nominative pronouns, suggesting that the Case checking mechanism is at work. The uninflected forms produced by Erdem are argued to be finite forms with missing inflections. Following Lardiere (1998a; 1998b; 2000), Prévost and White (2000) argue that if functional categories or abstract features are missing or impaired in L2 grammars, one would expect to find random placement of finite and nonfinite verbs as well as incorrect use of verbal morphology. However, Prévost and White (2000) found that, for adult L2 learners of French and German, while nonfinite verbs frequently occur in finite position, finite verbs are rarely placed in nonfinite position. On the basis of syntactic and morphological behaviour of finite verbs, Prévost and White (2000) propose that L2 learners have abstract features for finiteness and agreement in L2 acquisition and that there is no syntactic impairment in L2 grammar. In order to account for the variable use of inflectional morphology, following Lardiere (1998a; 1998b; 2000), Prévost and White (2000) also argue for a mapping problem between abstract features and surface morphological forms. They attempt to formulate this mapping problem in terms of Distributed Morphology (Halle and Marantz, 1993), according to which an inflected form is assumed to be associated with grammatical 6 Functional categories and evidence from CP acquisition Belma Haznedar 7 features such as tense, person, number and gender. In lexical insertion, the features of a lexical item should be consistent with the features of the terminal node in the syntax. The crucial point is that while the features of a syntactic node are fully specified, those of a lexical item might not be so. In other words, it is possible for a lexical item to be inserted into the hosting node, even though some of its features may be missing or partially specified. For Prévost and White (2000), L2 learners acquire the relevant grammatical features of the terminal node in the syntax via L1, UG or L2 input, but they might not have fully acquired feature specifications of the associated lexical items. They propose that while in the adult grammar nonfinite forms are specified as [–finite] and finite forms are [+finite], in the interlanguage grammar nonfinite forms can be inserted into a node bearing the [+finite] feature, because of mismatching features. Finite forms, on the other hand, are fully specified as [+finite] and therefore appear in finite contexts. Under this analysis then, there is no syntactic deficit in interlanguage grammars.2 In this article, I extend the investigation of functional categories in L2 acquisition by examining the complementizer system in child L2 acquisition of English. Assuming that the data from a large corpus analysed in this article are generalizable, they do not support the claim for a successive developmental sequence of VP–IP–CP in L2 acquisition. Further, I hope to show that the existence of functional categories does not depend on the use of morphological realizations of grammatical morphemes, suggesting that the acquisition of syntax can be independent. It is argued that the 2 It should be noted that adopting Rizzi’s truncation hypothesis Prévost (1997), who studied the acquisition of French by two English-speaking children (Gregg and Kenny), argued for truncation in child L2 acquisition. He discusses data on the development of auxiliaries and modals and finds that, consistent with his prediction, none of them occur in an Optional Infinitive (OI) / Root Infinitive (RI) environment. Similarly, subject clitics are generally found in finite contexts. Another piece of evidence for Prévost’s analysis comes from the distribution of DP subjects, which in general appear with finite verbs. With regard to null subjects, on the other hand, Prévost observes that a large proportion of OI/RI clauses had null subjects. The examination of wh-questions and embedded questions also reveals that they virtually always appear in finite contexts,which, according to Prévost, follow from Rizzi’s truncation hypothesis. With respect to negative OIs/RIs, however, Prévost observes that the percentage of negative utterances was roughly the same in finite and nonfinite contexts (over 20%). Under Rizzi’s clause structure the Neg projection is higher than other inflectional material and, therefore, negatives are not expected to occur in OIs/RIs. It appears that data provide conflicting evidence that negative nonfinite utterances occur in similar proportions to negated finite utterances. Prévost’s analysis has important theoretical implications. First, if it is the case that child L2 learners go through an OI/RI stage, one might expect to find evidence for a similar stage in Erdem’s L2 English too. Secondly, on the assumption that child L1 learners produce OI/RIs (e.g.,Wexler, 1994; Schütze and Wexler, 1996;Wexler, 1998), if a similar phenomenon is observed in child L2 acquisition, whatever the explanations are for OIs/RIs, maturational accounts of language acquisition will become irrelevant (Haznedar and Schwartz, 1997). variable use of inflection is due to problems with the realization of surface morphology, rather than to syntactic impairment, in accordance with the Missing Surface Inflection hypothesis. While our discussion focuses on the development of CP in particular, some aspects of the IP are also addressed in the discussion of the status of functional projections. To clarify the details of our analyses, at this point it is necessary to discuss some basic facts about the complementizer system in English. 4 The CP system in English Under standard assumptions (e.g., Bresnan, 1970; Chomsky, 1973), the head of CP is filled by complementizers such as for, if, that and whether. 1) a. She is not sure [CP [C if ] [he knows the truth]]. b. She thinks [CP [C that ] [he will sign the contract]]. c. I am so happy [CP [C for ] [him to succeed]]. d. She wonders [CP [C whether ] [he is leaving tomorrow]]. e. She wonders [CP [C whether ] [to accept the offer]]. The types of embedded clauses in (1) are determined by the complementizer. With respect to the syntactic distribution of complementizers, while if and that select a finite clause, for selects an infinitival complement and whether selects both types of clauses. It should be noted that interrogative complementizers, whether and if also introduce interrogative clauses.3 Finite declaratives, on the other hand, are introduced by an optional complementizer that. Another function of the head of CP is concerned with the phenomenon known as Subject–Auxiliary Inversion (I-to-C movement), which plays a central role in question formation in English. While nonauxiliary verbs in English do not move before Spell-Out, auxiliary verbs and modals raise, and they appear before the subject in matrix questions, as shown in (2). 2) a. She will present her article tomorrow. b. [CP [C’Willi [IPshe [I’ ei [VP present her article tomorrow?]]]]] Example (2b) is a yes–no question, characterized by the inversion of subject and auxiliary. While the modal will occupies the head of IP in (2a), it is moved into the head of CP in (2b). 8 Functional categories and evidence from CP acquisition 3 For arguments with respect to distributional differences between the two interrogative complementizers whether and if, see Borer, 1989; Kayne, 1991. Along the lines of Chomsky’s (1993) checking theory, Nakajima (1996) presents an analysis of distributional differences among whether, if, that and null that. Belma Haznedar 9 In addition to those constituents that appear in the head position of CP, the specifier of CP may also be filled, as exemplified in the following wh-questions. 3) a. What will you buy tomorrow? b. Where is he planning to go over the weekend? In (3a–b), the auxiliaries will and is are moved via head-to-head movement. In addition, what and where end up occupying a position in front of the auxiliaries. Following standard analyses of English wh-questions (e.g., Chomsky, 1986b), we assume that the wh-phrase moves to the specifier of CP and that the modal verb moves from I to C. 4) a. [CP[C’[IP I [I’ will [VP buy a book tomorrow]]]]] b. [CP Whati [C’ willj [IP you [ I’ ej [VP buy ti tomorrow]]]]] Example (4b) shows that the wh-phrase what, based-generated in the object position of the verb, ends up in [Spec, CP].4 One final area in regard to CP in English concerns infinitival clauses. Chomsky (1980) proposes that infinitival clauses have a nonovert subject, labelled as PRO.An important property of PRO is that it is restricted to the subject position of infinitival clauses. 5) a. Jane wants to know [CP whether [IP she should buy the book]]. b. * Jane wants to know [CP whether [IP PRO should buy the book]]. c. * PRO wants to know [CP whether [IP she should buy the book]]. d. * Jane wants to know [CP whether [IP John should buy PRO]]. e. Jane is not sure [CP whether [IP PRO to buy the book]]. 4 It should be noted that in addition to object wh-questions, English also has subject whquestions, as in Who will win the presidential election? It is important to note that there is an asymmetry between root subject and object wh-questions (e.g., Chomsky, 1986b). For some researchers, both subject and object wh-questions undergo movement,while for others root subject wh-questions do not move but remain in situ. Among others, one important difference between subject and object wh-questions that has received considerable attention in the literature is concerned with the that-trace filter (Chomsky and Lasnik, 1977). i) a. Who(m)i do you think [CP that [IP Jane will meet this afternoon ti]]? b. Who(m)i do you think [CP [IP Jane will meet this afternoon ti]]? The examples in (i) show that long distance object questions are grammatical irrespective of whether or not that occurs in the head of the embedded CP. In long distance subject questions, however, the subject cannot be extracted from the embedded CP when there is an overt complementizer appearing in the head of the lower CP. This is shown in (ii). ii) a. Whoi do you think [CP [IP ti will come this afternoon]]? b. * Whoi do you think [CP that [IP ti will come this afternoon]]? For discussion of asymmetries between object and subject wh-movement, see Chomsky, 1986b. Example (5b–c) shows that the nonovert element PRO cannot appear as the subject of a finite clause. The ungrammaticality of (5d) is due to the occurrence of PRO in the object position of buy. What one can deduce from these data is that the infinitival subject PRO occurs in ungoverned positions. The hypothesis that PRO must not be governed is referred to in the literature as the PRO theorem (e.g., Chomsky, 1986a).5 What is important is that in order for PRO to be ungoverned, there must be a CP, as CP is assumed to be a barrier to outside governors. 5 Summary This section has focused on the properties of CP in adult English, such as subject–auxiliary inversion, wh-movement and the distribution of PRO. It is shown that the derivation of nonroot subject wh-questions involves moving a wh-phrase to [Spec, CP] and the verb to the head of CP, the latter also being the case in yes–no questions.We have also illustrated that in English the head of the functional category CP is filled by wh-complementizers or lexical complementizers such as if, for and that. In what follows, I first present the child L2 data in terms of CP-related elements and then discuss my findings in light of V&Y-S’s claims on the development of phrase structure in L2 acquisition. First, a brief note about the subject of the study. III The current study The subject of this study, Erdem, is a native speaker of Turkish who arrived in the UK in November 1993. He was four years and three months old at the time the data collection began. He had no previous knowledge of English prior to his arrival in the UK and received virtually no exposure to English during the first two months of his stay, as he was always at home with his Turkishspeaking parents. Erdem started nursery school in January 1994, where he spent two and a half hours a day.When the data collection started at the beginning of March 1994, Erdem had been attending the nursery school for one and a half months. At school, he received no instruction in English but had contact with both native and 10 Functional categories and evidence from CP acquisition 5 In addition to licensing PRO in the subject position of infinitival complements, certain verbs such as want and expect, treat subjects of infinitival clauses as their objects. These verbs – known as Exceptional Case-Marking verbs – can govern into an IP and assign case to the subject of the infinitival clause. i) a. I expect [IP them to leave early]. b. I want [IP her to take a decision on this matter]. Belma Haznedar 11 nonnative speakers of English. By the time of the first sample (9 Mar 94), he had been producing some English words, mainly nouns, which were basically the names of animals such as dog, cat, lion, pig, bird, duck, etc. In September 1994 Erdem started infant school. Unlike at the nursery, all of the other children were native speakers of English. At school, he spent six hours a day, playing and learning with other children. Starting in September 1994, drastic changes were observed in Erdem’s English in terms of the use of a wide variety of phenomena, such as modals, yes–no questions, wh-questions and complement clauses. The data analysed in this article consist of 46 recordings, covering a period of approximately 18 months. Apart from the first three sessions, all of the data, collected approximately three times a month, were tape-recorded. Each session was conducted in English and continued for about 40 to 90 minutes. 1 CP-related elements in Erdem’s L2 English This section presents data on the development of questions in Erdem’s interlanguage, first focusing on yes–no questions and then wh-questions. We discuss yes–no questions with intonation, copula be, auxiliaries be and do and modal verbs. a Yes–no questions: The first yes–no questions are found in Sample 6 (22 Apr 94). Of the 3 questions, 2 of them are formed by intonation alone. 6) a. It’s a ball?6 (S 6) b. It’s a television? (S 6) In Sample 7 (6 May 94), there are two more instances of intonation questions. 7) a. It’s a triangle? (S 7) b. This is a zero zero? (S 7) Of the 11 yes–no questions produced up until Sample 15 (16 Sep 97), 5 (45.6%) are intonation questions. It should be noted at this point, however, that in the rest of the data, few occurrences of intonation questions are found (see Appendix 1). What is more, given the context, all these intonation questions seem to be appropriate, as shown in (8). 6One might argue that these early utterances beginning with It’s can be formulaic in that they are not analysed.However, while It’s may be unanalysed, the fact is that Erdem is still using intonation in these examples and, therefore, they are included in the analysis of questions (see Haznedar, 2001). 8) a. Investigator: What? What can you see here? Erdem: I don’t know #7 you know this? (S 18) b. Erdem: We do it this # What you think? (S 21) c. Erdem: You think ok? You say something. (S 21) In addition to intonation questions until Sample 15 (16 Sep 94), there are some occurrences of yes–no questions, mainly with copula be.8 9) a. Are you ready? (S 8) b. Is it very very big? (S 15) As can be seen in Appendix 1, there are more yes–no questions with copula be in the following samples. 10) a. Is it one boy holding ball in his hand? (S 26) b. Are you tired? (S 27) c. Is it funny? (S 27) In addition to the use of copula be, auxiliaries such as be and do and modal verbs are also used in yes–no questions. As mentioned previously, the number of questions increases after Sample 15 (16 Sep 94), when Erdem started at infant school in the autumn of 1994. Let us now examine the development of questions with the auxiliaries be and do and modals. The earliest instances of yes–no questions with the auxiliary be are found in Sample 15. 11) a. And I say you # # are you not listen me? (S 15) b. Are you coming? (S 15) We find more yes–no questions with the auxiliary be starting at Sample 30. 12) a. Are you going to take your books in your holiday? (S 30) b. Is the boy running in the playground? (S 33) c. Are you going to stay in my house? (S 33) 12 Functional categories and evidence from CP acquisition 7 # indicates a pause. 8There are some examples of yes–no questions with the modal verb would at this early stage. i) a. Would you like to playing this # play # this is very good playing. (S 9) b. Would you like to eggs? (S 11) c. Would you like to go my new house? (S 14) However, I do not think that would has been analysed as a modal verb at this stage. It only occurs in one context – namely, would you like to – while making an offer. There are no cases in which it is used in positive or negative declaratives in the same recordings. Due to its unanalysed nature, it is not included in my counts. Belma Haznedar 13 As shown in Appendix 1, in comparison with the relatively infrequent use of yes–no questions with auxiliary be (27 instances in the whole corpus), there are large numbers of yes–no questions that contain do (216) and modal verbs (189). The first use of auxiliary do in Erdem’s questions occurs in Sample 16 (4 Oct 94). 13) a. Do you know what I got? (S 16) b. Do you know what this say? (S 16) In subsequent recordings, we observe that auxiliary do is produced in different contexts: (1) the use of do-support with verbs such as want and (2) the correct use of do-support in past tense contexts, as shown in (14). 14) a. Do you want to look at # look at that? (S 17) b. Do you want to see it? (S 18) c. Did you colour your picture? (S 22) Yes–no questions also occur with modal verbs. It should be noted that the early uses of modals are restricted to can, which first appears in Sample 15 (16 Sep 94). Starting in Sample 16 (4 Oct 94), Erdem produces can in questions as well. 15) a. Can you say xxx? (S 16) b Can you push me? not now # push my car. (S 17) c. Can you help me in that one? (S 17) Similar to declarative contexts, modals other than can appear gradually in yes–no questions. 16) a. Would you get this? (S 24) b. Will you take this? (S 28) c. Could you get this? (S 35) d. Shall I read this? (S 42) Erdem’s data on yes–no questions indicate that it is only auxiliaries, modals and the copula be that undergo subject–auxiliary inversion. Erdem never inverts main verbs, i.e. there are no errors of the type in (17), where the thematic verb and the subject are inverted. 17) a. * Makes daddy a cake? b. * Reads mummy a book? We find that yes–no questions are formed correctly from their earliest occurrences. As shown in Appendix 1, of the 536 yes–no questions in Samples 6–46 (22 Apr 94–24 Aug 94), there are only 22 using intonation alone, 5 of which, as mentioned earlier, are produced in the first 15 recordings. In Table 1, we present the rate of intonation questions vs. subject–auxiliary inversion at different time periods in Erdem’s data. What we find is that the rate of intonation questions in Samples 6 through 14 appears to be high, occurring 45.5% of the time, but it subsequently drops to 10.3% in Samples 15 and 21. As mentioned previously, most are pragmatically appropriate. Overall, however, the rate of intonation questions is low, which might suggest that Erdem quickly acquired subject–auxiliary inversion in English yes–no questions. Overall, we find that the number of yes–no questions in the earliest samples is low. In addition to some occurrences of early intonation questions, there are only 11 yes–no questions until Sample 15 (16 Sep 94).However, evidence from the use of be (both as a copula and an auxiliary), do and modals (primarily can at the beginning) in subsequent samples shows that Erdem produces questions in which auxiliaries/modals are used in correct sentenceinitial position. b Wh-questions: This section first discusses the earliest and the most frequently used forms of wh-questions, which include what and where. The following examples in (18) show the first instances of Erdem’s wh-questions. 18) a. What’s this? (S 7) b. What’s four? (S 7) c. What’s this name? (S 7) We cannot rule out the possibility that the questions in (18) are formulaic, as all of them start with what’s. Erdem’s utterances with This is/It’s . . . in Samples 6 and 7 appear to be formulaic where the copula be is not yet analysed. These earliest wh-questions with what’s in Sample 7 (6 May 94) are therefore not included in the counts. From Sample 8 (20 May 94) on, however, what’s varies with what, which might suggest that it is not morphologically unanalysed. While some of them are still not target-like, examples of the nonformulaic what questions with missing auxiliaries are shown in (19). 14 Functional categories and evidence from CP acquisition Table 1 Percentage of inversion vs. intonation in yes–no questions Sample: 6–14 Sample: 15–21 Sample: 22–46 Intonation 5/11 45.5% 6/58 10.3% 11/467 2.4% Inversion 6/11 54.6% 52/58 89.7% 456/467 97.6% 19) a. OK # hello # what you doing here? (S 8) b. Hello dad # what are you doing? (S 9) c. What do you want? (S 11) As can be seen in Appendix 2, the first wh-questions with where are also found in Sample 8 (20 May 94). Of the 2 instances, one occurs with missing copula be, the other with the uncontracted form. 20) a. Where carrots? (S 8) b. Where is ‘karpuz’ [= watermelon in Turkish] (S 8) We do not find large numbers of where questions, although there are more of them as time passes. 21) a. Where’s chocolate biscuits? (S 11) b. OK # where is your friend? (S 11) c. Where is article? (S 18) Among the other types of wh-phrases in questions are why, who, which and how, although there are only a few in the early transcripts. 22) a. Why don’t you go home? (S 12) b. This is me # this is me this is # # who is it that? (S 16) c. Which one do you like in this? (S 17) d. How can you make another control? (S 24) In the whole corpus, while there are only a few instances of whquestions with when and whose, when is used extensively in complement clauses. 23) a. When this lives? (S 36) b. When is the bus coming down? (S 38) c. When do you want to make a ginger bread? (S 42) d. Whose birthday is it? (S 45) Belma Haznedar 15 Table 2 Distribution of wh-phrases in wh-questions what 187 (+5 wh-in situ)/388 49.5% where 64 (+1 wh-in situ)/388 16.8% how 38/388 9.8% which 32/388 8.3% why 32/388 8.3% who 18/388 4.6% when 8/388 2.1% whose 3/388 0.8% The total number of wh-questions broken down by type of whphrase is given in Appendix 2.9 As can be seen in Table 2, we do not find the same number of questions for each type of wh-phrase. Overall, Erdem asks fewer who, when and whose . . . questions than what, where, how, which and why.10 There is a gradual increase in the use of wh-phrases, what and where being the most common types used. The overall distribution of different types of whphrases in all wh-questions until Sample 46 (24 Aug 95) is given in Figure 1. We now discuss the types of errors found in these wh-questions. It has been reported that L1 English-speaking children make errors of subject–auxiliary inversion and of missing auxiliary (e.g., Klima and Bellugi, 1966; Stromswold, 1990). In Erdem’s L2 English, we also find similar errors. One type of error consists of cases in which the auxiliary be or do is missing. 24) a. OK # hello # what you doing here? (S 8) b. What you eating? (S 11) c. What I get here? (S 20) The second kind of error is concerned with the lack of subject–auxiliary inversion. 25) a. What you’re saying? (S 19) b. What we are doing again? Again those pictures. (S 26) c. What we can play? Do you know? (S 29) 16 Functional categories and evidence from CP acquisition 9One should also note that there are only 6 instances of wh-in situ questions in the entire corpus, as shown in (i) (see the last column in Appendix 2). i) a. Investigator: There might be little animals in it # you should be careful. Erdem: It’s not animals # look. Investigator: OK # take it. Erdem: Put where? (S 11) b. Investigator: Let’s play something else. Erdem: Play what? (S 12) c. Investigator: Of course I do listen to what you say. Erdem: You say what? (S 15) d. Erdem: This is # # Investigator: What? Erdem: This is what? (S 21) e. Investigator: Why doesn’t Elifcan talk to you? Erdem: You tell me what? (S 36) f. Investigator: End of the week. Erdem: What do you mean? End of what? (S 41) 10 In a study of questions in child L1 English, Stromswold (1990) also reports that children did not ask the same number of questions for each type of wh-phrase. Overall, she finds fewer questions with when and which than other types of wh-questions. Belma Haznedar 17 In order to examine both types of errors, we divide all wh-questions into two groups: nonsubject wh-questions vs. (root) subject whquestions, as the latter does not involve subject–auxiliary inversion.11 First, consider some of the subject questions given in (26): 26) a. Which number suit the pictures? (S 21) b. Who want bread? (S 27) c. Which plane goes slowly? (S 29) Appendix 3 presents the total number of subject vs. nonsubject wh-questions. As we can see, of the 382 wh-questions,12 357, being nonsubject wh-questions, require subject–auxiliary inversion; only 25 are subject wh-questions (6.7%). Appendix 4 shows the distribution of the 357 instances of nonsubject wh-questions in each sample by wh-phrase. These 357 questions form the basis of Appendix 5, which looks at questions with missing auxiliary errors. As can be seen from Appendixes 4 and 5, of the 15 wh-questions in Samples 8 through 11 (20 May 94–17 June 94), 6 (40%) have a missing auxiliary. Appendix 5 shows that the proportion of these errors gradually decreases, in Samples 24 through 36 (8 Dec 94–24 Mar 95), for what where how which why who when whose wh-questions Percentage Figure 1 Distribution of wh-phrases in wh-questions 11Our discussion here is mainly restricted to root clauses, since Erdem produces few longdistance wh-questions. 12The figure of 382 does not include the six instances of wh-in situ questions; 5 with what, 1 with where, discussed in footnote 9. instance, there are 154 nonsubjects wh-questions (Appendix 4) and 29 have missing auxiliary errors (18.83%) (Appendix 5). After Sample 36, we do not find missing auxiliary errors. Overall, however,we find that in earlier samples the rate of missing auxiliary errors is higher than that of subsequent samples. We now turn to the rate of inversion errors. Inversion errors in nonsubject wh-questions are detailed in Appendix 6. In order to find inversion errors, one needs to subtract missing auxiliary errors from the total number of nonsubject wh-questions (Appendix 4). In other words, in cases where the auxiliary is missing, one cannot look at inversion errors, precisely because the auxiliary is not provided. Therefore, for all instances of inversion errors in Appendix 6, we first look at each of the cells in Appendix 4 which shows the number of nonsubject wh-questions; then we find out whether the question has a missing auxiliary; if it does, for each cell, we subtract the number of missing auxiliary errors, and then calculate the number of inversion errors. In Sample 27, for example, there are 6 instances of wh-questions with what but 3 have missing auxiliary, so 3 is subtracted from 6. Of the remaining 3, 1 has an inversion error, as shown in Appendix 6. The first instances of inversion errors are found in Samples 19 and 20 (1 Nov 94–8 Nov 94). Of the 19 nonsubject wh-questions in these two samples (see Appendix 4), 2 have missing auxiliary errors (Appendix 5). In order to find the number of inversion errors, we first subtract 2 from 19 (the total number of nonsubject whquestions in these two samples), of the remaining 17, 2 have inversion errors (11.8%) (Appendix 6). No inversion errors occur between Samples 21 and 25 (15 Nov 94–29 Dec 94). In Samples 26 through 30 (5 Jan 95–4 Feb 95), however, there are 8 (19.5%) inversion errors out of 41 (50 – 9 = 41) nonsubject wh-questions (50 being the total number of nonsubject wh-questions, and 9 being the number of missing auxiliary errors; see Appendix 4 and Appendix 5). The percentage of inversion errors between Samples 33 and 35 (1 Mar 95–16 Mar 95) is somewhat higher, occurring 27.3% (9/33) of the time (42 being the total number of nonsubject wh-questions (Appendix 4), and 9 being the total number of missing auxiliary errors (Appendix 5) ). However, in subsequent samples, we find very few instances of inversion errors. The number of inversion errors with various types of wh-phrases differs due to the fact that not all wh-phrases are produced at the same rate. As what is the most prevalent wh-word in the corpus, it is predictable that it would predominate.As predicted, Erdem fails to invert what more than any other wh-phrases. Of the 23 inversion 18 Functional categories and evidence from CP acquisition Belma Haznedar 19 errors in wh-questions, 11 (47.8%) of them involve what. Similarly, missing auxiliary errors with what predominate. Out of the 46 instances of missing auxiliary (Appendix 5), 31 have what (67.4%). Overall, however, despite a certain number of inversion errors and missing auxiliaries discussed in this section, it is not the case that Erdem systematically fails to invert subject and auxiliary or consistently fails to produce auxiliaries in questions. In most cases, these errors occurred alongside utterances where there was correct subject–auxiliary inversion and auxiliary use, as shown in (27). 27) a. What do you do this car now? (S 28) Hey don’t do. Now what you do my this special car? You got to be careful. b. What do you mean? (S 31) What cushion mean? (S 31) c. What he say here? (S 33) What do you want to do? (S 33) Perhaps even more importantly, similar to the observation made by Klima and Bellugi (1966) for L1 English (see also Hyams, 1994), at a time that Erdem sometimes fails to invert auxiliary in whquestions, he does however invert auxiliaries consistently in yes–no questions. As mentioned previously, the first instance of an inversion error in wh-questions is found in Sample 19 (1 Nov 94). The following yes–no questions in (28) come from earlier samples, exemplifying subject–auxiliary inversion even before inversion errors in wh-questions appear. 28) a. Are you not listen me? (S 15) b. Are you coming? (S 15) c. Is it very very big? (S 15) d. Do you know what I got? (S 16) Let us consider the example in (28a), which shows that auxiliary be and the subject you both precede the negative not. On the assumption that there is a NegP between the VP and the IP, the subject and the verb are both positioned in a higher position. If this is so, the auxiliaries and modals in (28) must be in a higher position. This finding might suggest that Erdem’s L2 grammar already has a CP projection, and syntactic factors do not appear to be responsible for the errors found in Erdem’s wh-questions. 2 Embedded clauses This section examines the development of embedded clauses in Erdem’s L2 English.We first discuss clauses with because and if as well as with wh-complements. Secondly, we examine the use of the complementizer that and clauses with verbs such as want. Appendix 7 presents the number of embedded clauses in the corpus. Erdem’s first utterance with because occurs in Sample 13 (23 Aug 94). From Sample 15 (16 Sep 94) onwards, there are numerous instances of clauses with because and if consistently used in subsequent samples, as shown in (29) and (30). 29) a. Because it’s bedtime. (S 13) b. I just eat my hands because I not cut. (S 15) c. You can get it because it is good. (S 15) 30) a. If we go there and if we go there # we coming there. (S 15) b. I don’t know if I pass it. (S 20) c. If you want to jump # you press this or press this. (S 20) d. I don’t know if my mummy wash it. Is it dirty? (S 39) e. I don’t know if I can get it out. (S 43) Note that Erdem’s if-clauses in (30) include both conditionals and the cases where if introduces complements of verbs such as know, as in (30b) and (30d–e). Similar to the other acquisition facts I have discussed up to now, complement clauses with wh-phrases also start after Sample 15 (16 Sep 94). From the earliest occurrences, Erdem’s wh-complement clauses occur both in questions and declarative sentences. Some examples are given in (31). 31) a. Do you know what this say? (S 16) b. Yes # that’s what I say. (S 16) c. I don’t know where is it. (S 16) d. I don’t know who is it. (S 16) I should note at this point that the production of the complementizer that is never frequent in Erdem’s data.13 This is shown in (32) by the 5 occurrences of that in the whole corpus.14 20 Functional categories and evidence from CP acquisition 13 A similar finding is reported in Bloom et al. (1989). In a study on the acquisition of complementation in child L1 English, Bloom et al. examine longitudinal data from four 2–3- year-old children. They specifically discuss the acquisition of four complement-taking verbs: two epistemic verbs (think and know) and two perception verbs (see and look). What they find is that the production of particular complement types mainly depend on the matrix verb. Think, for example, appeared with sentential complements, whereas know occurred almost exclusively with wh-complements. The use of that as a complementizer, however, was rare. Of the 179 complement clauses with think, only 3 had that. For Bloom et al. the rare occurrences of that might be related to the fact that as well as being a plurifunctional lexical item, that may not be frequent in the input children hear. 14There is only one instance of that in a relative clause construction, as shown in (i). i) I was just doing something that you don’t know. (S35, 16 Mar 95) Belma Haznedar 21 32) a. Do you remember that I show you a one red man? (S 37) b. She tells the other boys that she tell the other she tell the other boys that that she say that let’s play now outside. (S 37) c. I don’t think that I could get xxx. (S 39) d. I just bring him a bear book but I didn’t know that he like bear. (S 44) Apart from the examples given in (32), Erdem does not produce the complementizer that, as shown in (33). 33) a. You say me go # I’m go home. (S 12) b. I know it’s a big bird. (S 16) c. I think it’s my dad. (S 18) d. But you say I can’t catch it. (S 19) A final area we consider is infinitival clauses with verbs such as want. Infinitival constructions with want appear quite early, around Sample 10, and are consistently present throughout the full time under consideration. Some examples are given in (34). 34) a. I want to go new playground. (S 10) b. No # please please # I don’t want to get off the bike. (S 11) c. Do you want to go to Ankara? (S 16) d. I don’t want to play that game. (S 17) The last column in Appendix 7 presents the number of utterances with the verb want in Samples 10–46.15 What we have shown up to this point is that utterances with because, if and the verb want, appear relatively early. Furthermore, yes–no questions, wh-questions and wh-complement clauses are in general used correctly in their earliest occurrences. IV Discussion 1 Implications for the Minimal Trees hypothesis Up to this point the following properties of Erdem’s early L2 grammar have been discussed: . Yes–no questions, where correct subject–auxiliary inversion predominates; . wh-questions, where what and where appear first; . the productive use of embedded clauses on the basis of data from 15The number for want given in Appendix 7 does not include the use of want with a nominal complement, as in the following examples. i) a. I want coca cola. (S 20) b. I want these chocolate biscuit. (S 23) indirect wh-questions, clauses with if and because, and infinitival clauses embedded under want. I would now like to discuss these findings with respect to the Minimal Trees hypothesis and the Missing inflection hypothesis. I argue that a number of properties in Erdem’s L2 grammar provide evidence against the successive development of phrase structure proposed by V&Y-S. One of the criteria for the existence of CP in the Minimal Trees model is the production of questions. Under this account, L2 learners should produce wh-questions with a fronted wh-phrase or yes–no questions with a fronted auxiliary/modal verb in order to be able to conclude that they have developed a CP. Furthermore, V&Y-S also require the use of embedded clauses with overt complementizers for the CP system and claim that at the VP or the FP/AgrP stage, there is no evidence for embedded clauses. On the basis of the absence of overt complementizers and productive questions, they conclude that learners at the VP and FP/AgrP stage have neither acquired the CP projection nor transferred it from their L1. In Erdem’s very early recordings, we do not find regular and productive use of questions or overt complementizers. As we have seen, the relatively productive use of these constructions appears mainly after Sample 15 (16 Sep 94).The question we are faced with is whether or not we should conclude that Erdem has no knowledge of a CP system up to this point. What has been repeatedly emphasized in V&Y-S’s work is the lack of functional categories in early stages. However, in my opinion, it is not clear what is meant by ‘early stages’ in V&Y-S’s account. It is not known how long a VP stage lasts or how long it takes for an L2 learner to pass on to the IP/AgrP stage. Their analysis appears to require the production of CP elements from the start of data collection. In an attempt to argue against Lakshmanan and Selinker, V&Y-S (1996a) note that: If we look at evidence for the development of the CP by the two children, as discussed in Lakshmanan and Selinker (1994), we see that they do not produce any embedded clauses during the first two sessions . . . CP is not present from the start of data collection. (V&Y-S, 1996a: 27–28; emphasis added) At this point, we are left with a puzzle: do we expect Erdem to produce CP-related phenomena such as subject–auxiliary inversion, wh-questions and wh-complementizers in the first recorded session? This was the period during which Erdem was basically 22 Functional categories and evidence from CP acquisition producing his first multi-word utterances in English. It is obvious that language acquisition proceeds through time. Nobody would expect an L2 learner to start producing, for example, relative clauses or embedded clauses in the first two or three sessions. The learner must necessarily be exposed to input before language acquisition can take place.16 Certain predictions follow from V&Y-S’s Minimal Trees hypothesis: one prediction is concerned with the acquisition of subject vs. nonsubject wh-questions. If one assumes that root subject wh-questions do not require movement, but other types of wh-questions do, V&Y-S should predict that root subject whquestions would be acquired first, as movement is not possible due to the lack of CP. In regard to the distribution of subject vs. nonsubject wh-questions in Erdem’s L2 English, we find that of the 48 wh-questions until Sample 20 (8 Nov 94), none of them are subject wh-questions. Overall, of the 382 wh-questions until Sample 46 (24 Aug 95), only 25 (6.7%) are subject wh-questions. Our results do not appear to support the (potential) prediction that subject wh-questions should occur earlier than nonsubject whquestions. We have also found early uses of infinitival clauses as the complement of the verb want. Starting in S 10 (5 June 94), Erdem consistently produced want clauses, as in (35). 35) a. I want to play. (S 10) b. I want to fly one. (S 13) c. Do you want to look at # look at that? (S 17) Under the assumption that the subject of an infinitival clause is PRO, the analysis of (35a), for example, is as follows. Belma Haznedar 23 16Note that the data from the earliest recordings indicate that Erdem correctly interprets wh-questions and provides appropriate answers. i) a. Investigator: Which colour do you want? Erdem: Red. (S 4, 4 Apr 94) b. Investigator: Where is your dad now? Erdem: My dad school. (S 5, 11 Apr 94) These examples show that Erdem’s answers to wh-questions with wh-phrases such as which and where are appropriate,even in the earliest recordings, and he had no problems processing utterances that contain a CP. Crucially, the data given in (i) come from the earliest recorded sessions, presumably fulfilling V&Y-S’s criterion of early stage. In earlier work (Haznedar, 1997), following Grondin and White (1996), I argue that the correct interpretation of whquestions might suggest that Erdem understands and processes sentences with a CP, and therefore his interlanguage grammar had a CP projection. As has been pointed out by an anonymous reviewer, however, Erdem’s correct interpretation of wh-questions could be argued to be pragmatic. 36) a. I want [CP [IP PRO to play]]. These arguments give further support to the view that Erdem’s L2 English has a CP system.17 Another source of evidence for CP is concerned with the acquisition of complement clauses, which are assumed to require CP. In addition to evidence for syntactic operations such as subject–auxiliary inversion in questions, which roughly starts by Sample 15 (16 Sep 94),we have also shown that complement clauses with wh-constituents and because and if were widely produced at a time Erdem was not yet producing certain lexical complementizers, in particular, that. 37) a. Do you know what this say? (S 16) b. No I don’t know where I go. (S 17) c. Look # what I can do. (S 22) The presence of overt complementizers certainly provides evidence for a complementizer system. However, the lack of overt complementizers does not mean that Erdem’s grammar has no CP system, especially when one considers the fact that the overt complementizer, that, is optional in English. It may well be that the examples in (33) show that Erdem does not have that in the initial stages of his interlanguage grammar.18 Rather, he only has a null complementizer, in accordance with the Missing Surface Inflection hypothesis. What this suggests is that the dearth of the complementizer that in Erdem’s English cannot be attributed simply to the absence of the CP projection, since other CP-elements and complementizers are readily available, as well as the productions of clauses with Ø complementizers, as in (38). 38) a. I think this boy is crying. (S 21) b. It say it’s very good drink. (S 22) c. You say I don’t ask you some questions but you are asking questions. (S 26) 24 Functional categories and evidence from CP acquisition 17An anonymous reviewer argues that the use of want in these examples do not seem to provide strong evidence for evidence of CP in that as an exceptional case marking verb want would take an IP rather than a CP. First, one should note that Erdem used want as an ECM verb, as in (i), but not until much later. i) a. I want my mummy to hold me. (S 28) b. Because my mummy don’t want me to find where is the biscuit. (S 42) The data discussed in this article show that a CP projection develops around Sample 15, while the examples with want as an ECM (Exceptional Case Marking) verb come from Sample 28 onwards. Secondly, the analysis of sentences with want within an IP projection does not account for the use of complementizers at the same period under discussion. 18This point has suggested to me by an anonymous reviewer. V&Y-S might argue that Erdem did not produce embedded clauses in the first several recordings. As we have discussed previously, the types of clauses depend on the properties of complementizers. The delay in the use of embedded clauses might be accounted for by the fact that Erdem first had to acquire individual lexical words for each complementizer. In other words, the absence of complementizers might be a lexical problem in the sense that he may have simply not yet learned the appropriate lexical items. 2 V&Y-S: Successive developmental stages: VP–IP–CP In addition, one of the major proposals in the Minimal Trees hypothesis is that L2 learners acquire functional categories through successive developmental stages, namely, VP–IP–CP. Under this account, the acquisition of VP precedes the acquisition of IP; likewise, the acquisition of IP precedes the acquisition of CP. The early emergence of CP prior to IP would provide counter evidence for V&Y-S’s theory of the successive development of phrase structure. Under standard assumptions, the copula be, auxiliaries be, do and have, modal verbs, tense and agreement markings are all assumed to be associated with inflectional categories. It should be noted that while auxiliary and copula be appeared fairly early in Erdem’s L2 English, the development of tense and agreement morphology is rather gradual.19 For the purposes of this article, the next section examines the development of verbal morphology20 in Erdem’s L2 English. 3 Verbal morphology a Subject–verb agreement (third person singular -s): Of the 46 samples examined in this study, the first obligatory context for third Belma Haznedar 25 19While the use of copula be seems to be productive by Sample 12 (9 Aug 94) over 90%, the third person singular -s, let alone being productive, first appears at Sample 15 (16 Sep 94). It is only by Sample 41 (26 May 95), ten months later that, the percentage of verbs inflected with the third person singular -s reaches 70%. As discussed in the next section, similarly regular or irregular past tense morphology on verbs is absent for a long period of time. 20One should note that in regard to the acquisition of IP, this article examines only the development of verbal morphology.This has been done for a particular reason: At a time at which Erdem’s L2 data provide evidence for the early development of copula be, auxiliary be and overt subjects, there is no evidence for the use of verbal morphology such as third person singular -s and tense marking. If one adopted V&Y-S’s model of structure building approach, on the basis of data from be and overt subjects, one would conclude that an INFL category is present, but because Erdem does not produce tense or agreement markings in his early L2, INFL must be missing. As discussed in great detail in Haznedar (2001), this would lead to a contradiction: that is, it does not seem to be plausible to argue that IP is both present and absent. person singular (3sg) -s occurs in Sample 9, there are 4 such cases but none of the verbs are inflected. One should note that the first occurrence of the inflected verb is not found until Sample 15, and the use of inflection is not productive for a considerable length of time. Until Sample 23, only a few utterances with 3sg -s are produced. Table 3 presents the gradual development of 3sg -s in different time periods.21 In Samples 15 through 28, we do not find much evidence for the productive use of the inflected verb. As can be seen in Table 3, only 11.8% of the verbs are inflected. It is only after S 28 that Erdem produces 3sg -s relatively frequently. In S 29, for instance, there are 13 utterances which have 3sg -s, as opposed to 11 utterances where the agreement marker is not provided. This does not mean that the number or percentage of verbs inflected with 3sg -s rapidly increases in the following recordings. On the contrary, the suppliance of the 3sg -s is gradual. In Samples 29 through 35, we find 111 (32.2%) utterances with 3sg -s, while the number of instances in which the subject–verb agreement is missing is 234 (67.8%). Between Samples 36 through 40, on the other hand, the use of the 3sg -s appears to change noticeably, inflection occurring 62.0% (119 verbs with 3sg -s), in comparison to 38.0% (73) uninflected forms. It should be noted, however, that it is only after Sample 40 that the number of instances with 3sg -s is always higher than that of instances where the 3sg -s is not provided. The use of uninflected verbs finally occurs somewhere around 28% in Samples 41–46. In sum, these findings suggest that the development of 3sg -s is gradual. 4 Tense marking The development of past tense forms also follows a gradual pattern. This section examines the development of regular and irregular past tense morphology in Erdem’s L2 English. 26 Functional categories and evidence from CP acquisition a Irregular past tense: The first obligatory context for irregular forms appears in Sample 10, and in 5 out of 5 cases Erdem fails to produce the irregular form of the verb. The earliest occurrence of irregular past forms (4/16) is found in Sample 13. Until Sample 20, very few utterances with the irregular past appear in Erdem’s English. Of the 81 contexts, only 7 (8.6%) instances of the irregular past form are found. In the remaining 74 utterances (91.4%), the base form of the verb is produced. Starting in Sample 22 (22 Nov 94), although the irregular past forms are still not frequently used, their incidence increases gradually; As can be seen in Table 4, until Sample 28 (20 Jan 95), there are 43 (20.8%) cases of irregular past in comparison to 164 (79.2%) instances where an irregular form is absent. In Samples 29 through 35 (26 Jan 95–16 Mar 95), of the 159 obligatory cases, only 67 (42.1%) instances of irregular past forms are attested. In the following samples, the percentage of irregular past tense forms continues to gradually increase. A total of 80 instances of irregular past forms (41.0%) are attested in Samples 36 through 40 (24 Mar 95–19 May 94), as opposed to 115 occurrences of missing irregular past (59.0%). Uninflected verbs still account for 48.6% of all irregular past contexts used in Samples 41–46. b Regular past tense -ed: Similar to the distribution of irregular past forms, verbs inflected with regular past tense -ed fail to appear for a long time. Although the first obligatory context for -ed occurs in Sample 8; no verbs inflected with the regular past are produced until Sample 15. As can be seen in Table 5, even after Sample 15, the number of verbs inflected with past tense -ed is rather low. In Samples 15 through 28, only 10.5% (6/57) of the verbs are inflected, whereas 89.5% (51/57) of the obligatory contexts lack the inflected form of the verb. Up to Sample 35, a total of 10 (10.0%) regular past forms are produced out of 100 obligatory contexts. Similarly, in Samples 36–40 while the percentage of uninflected forms was 75.4%, only 24.6% of the verbs had the inflected form. Unlike the production of 3sg -s, the suppliance of past -ed is still low even after Sample 40. In Samples 41–46, 57% of the verbs still lack regular past tense morphology. In striking contrast, Tables 6, 7, 8 and 9 show the virtually errorfree acquisition of I-to-C movement and the suppliance of dosupport in yes–no questions. When we compare the numbers of tense and agreement marking with those of yes–no questions, the percentages in Tables 6, 7, 8, and 9 illustrate the claim that I-to-C movement with modal verbs, auxiliary and copula be, and do are acquired prior to overt suppliance of agreement and tense morphology in obligatory contexts. In order to show this with a particular example, let us examine the numbers and percentages of agreement marking in Samples 29–35.While Erdem’s suppliance of agreement marking is around 33%, as shown in Table 3, a year after 28 Functional categories and evidence from CP acquisition data collection started, Erdem inverts auxiliary and copula be, auxiliary do and modal verbs almost 100% of the time during the same period (copula be occurring at 100%, modal verbs at 100%, do-support at 97.6%, auxiliary be at 92.9%). Similarly, the suppliance of past tense marking in Samples 29–35 remains low, irregular past occurring at 42%, regular past at 15% and never reaches the 80 or 90% that at least some studies use as a criterion for acquisition. To sum up, the child L2 data analysed in this article have shown that verbal inflectional elements 3sg -s and past tense forms are not produced early on, but rather the development of inflection is gradual (see also Haznedar and Schwartz, 1997; Haznedar, 2001). We observe that 3sg -s, for instance, is not attested until Sample 15 (16 Sep 94). Even after Sample 15, it is not produced consistently. In contrast, yes–no questions and wh-questions are productively used, in particular after Sample 15. Similarly, there are numerous utterances with complementizers such as because, if and various kinds of wh-constituents in embedded contexts before the acquisition of verbal inflection. Moreover, some utterances which unambiguously invoke the CP level occur with an uninflected form of the verb. Some examples are illustrated in (39). The examples in (39) show that although 3sg -s is not provided, Erdem has knowledge of syntactic operations that involve CP. For example, (39a) is a yes–no question that requires the projection of a CP. Likewise, (39d) shows that Erdem is able to produce a complex CP sentence while the main verb want lacks 3sg -s. According to V&Y-S’s developmental sequence of VP–IP–CP, one would not expect CP sentences either to occur prior to the use of the 3sg -s and past tense or to occur without inflection on the verb. In Erdem’s data, however, CP-related elements have been acquired prior to the use of agreement and tense morphemes. V Conclusions By analysing longitudinal data obtained from a Turkish-speaking child L2 learner of English, I have hoped to show that the findings in this study do not support Vainikka and Young-Scholten’s successive stages of development.One cannot overlook the fact that yes–no questions, inverted wh-questions and complement clauses are used consistently at a point in which certain inflectional verbal morphology – i.e., 3sg -s and past tense – does not appear to be productive. In contrast to V&Y-S’s claim, these data do not show that Erdem has acquired IP-related elements and, hence, their projections before the acquisition of CP-related elements.What this means is that the development of CP in Erdem’s English does not appear to be implicationally dependent on the prior acquisition of IP, if one assumes that evidence for IP requires overt production of morphological items associated with IP. This result is compatible with previous work on child L2 acquisition reported by Gavruseva and Lardiere (1996). The findings reported in this study provide further support for the Missing Surface Inflection hypothesis in that the absence of inflectional morphology (3sg -s or past tense) cannot be attributed to the lack of functional projections. Rather, the learner appears to have problems with surface morphological realizations of the verbs. What this means is that morphological inaccuracy appears to reflect difficulties with the mapping rather than any underlying syntactic deficit. The results presented here show that Erdem’s L2 data provide us with robust evidence for a CP projection in the syntax 30 Functional categories and evidence from CP acquisition long before he was reliably productive in using the inflectional system. In conclusion, these results are consistent with Lardiere (1998a; 1998b; 2000) who argues for a discrepancy between the development of the syntactic representation of functional categories and inflectional morphology. In this sense, this study presents another example of mapping problems through which the L2 learner faces difficulties in associating surface forms with some underlying abstract feature. Much work still needs to be done in order to understand the nature of mapping from feature to form in L2 acquisition. Acknowledgements The research reported in this article was conducted as part of a doctoral dissertation submitted to the University of Durham, UK. I would like to thank my thesis advisor Bonnie D. Schwartz for her discussion of issues addressed in this article. My sincere thanks also go to Bernadette Plunkett and Martha Young-Scholten who shared their insights with me.Earlier versions of this article were presented at SLRF 2000 (Second Language Research Forum), University of Wisconsin, Madison, September 2000 and Boston University Conference on Language Development, Boston, November 2000, and I thank those in attendance for their suggestions. I am also grateful to three anonymous Second Language Research reviewers for their invaluable comments and suggestions. Many thanks to Judy Monthie-Doyum for proofreading the article. All errors are, of course, mine. This research was conducted with support from the Turkish Ministry of Education, for which I am grateful.