Universität Duisburg-Essen

Transcription

1 Keriman Kırkıcı The Acquisition of the Pro-Drop Parameter in Turkish as a Second Language Series A: General & Theoretical Papers ISSN Essen: LAUD 2008 Paper No. 722 Universität Duisburg-Essen

2 Keriman Kırkıcı Ankara University (Turkey) The Acquisition of the Pro-Drop Parameter in Turkish as a Second Language Copyright by the author Reproduced by LAUD 2008 Linguistic Agency Series A University of Duisburg-Essen General and Theoretical FB Geisteswissenschaften Paper No. 722 Universitätsstr. 12 D Essen Order LAUD-papers online: Or contact: laud@uni-due.de

3 Keriman Kırkıcı The Acquisition of the Pro-Drop Parameter in Turkish as a Second Language Abstract In this study, which is mainly concerned with the acquisition of the pro-drop parameter in the acquisition of L2 Turkish, subjects from Turkish-type (Afganian, Persian, Mongolian, Romanian, Russian, Ukrainian, Arabic, Spanish) and Chinese-type (Chinese, Japanese, Korean) pro-drop languages learning Turkish as an L2 at three different proficiency levels took part in a grammaticality judgment test which included 35 sentences related with prodrop, subject-verb inversion and that-trace features and another grammatical judgment test which included 18 questions referring to null subjects and referential features of pronominals. The results obtained showed that the subjects from Chinese-type languages (Group 2), were much more successful in relation to the pro-drop parameter. This does not support the claim of the indirect access hypothesis, according to which it is possible to reset parameters only before the critical period. It instead provides support for the claim that there is no L1 transfer in L2 acquisition. It was also found that there was no difference in the success of both groups choices related with pro-drop questions in accordance with their proficiency levels. These results speak for the view that UG and L2 have a direct connection in L2 acquisition and there is no unique explanation for the theories of access, at least in relation with the pro-drop parameter. A comparison between scores of the subject-verb inversion questions, that-trace questions and pro-dropping questions clearly displayed that there is no relation between subject-verb inversion and pro-dropping; just like there is no relation between that-trace and pro-dropping. These results support the claim of resetting the pro-drop parameter without any effect of syntactic features such as subject-verb inversion and that-trace and the idea of no-access in L2 acquisition. In addition, it was found that speakers, who have strong agreement in their L1 (Group 1), are much less successful than the speakers who have no agreement in their L1 s (Group 2) in acquiring an L2 which also has strong agreement. This result likewise indicates that there is no access to UG in L2 acquisition. Both L1 speakers of Group 1 and speakers of Group 2 showed a regular increase of the reference of pronouns as subjects in accordance with their L2 acquisition levels of Turkish. But it can also be seen that speakers of Group 1 were much more successful than those of Group 2, who also showed a tendency to choose another NP as an antecedent. The reason for this is thought as the possibility of a referential discourse of a pronoun in Chinese-type languages and such a binding feature can not be seen in Turkic-type languages. These 1

4 results show that people learning Turkish as an L2 make use of their L1 s in accordance with referential features and, thus, support the idea of the indirect access theory. Speakers of Group 1 also showed a regular increase of the reference of pronominals as a subject in accordance with their L2 acquisition levels of Turkish. Speakers of Group 2, however, did not show any increase in such questions related with their L2 levels and showed again a tendency to choose another NP. This result again supports the idea that L1 speakers of Group 1 are much more successful in choosing the correct antecedent than speakers of Group L2 and that speakers of Chinese-type languages transfer from their L1 s. The results of this study hence showed that L1 speakers of both Group 1 and Group 2 supported different access theories in the acquisition of Turkish as an L2. As it is possible to reach different access theories according to the kind of the parameter studied (Cook, 1995; Hawkins, 2001), it can clearly be seen that is still not possible to have a unique judgement for the pro-dropping parameter in L2 acquisition with the existing theories. 1. Introduction Human beings are born with universal principles, which are directly related with language (Chomsky, 1986). According to this approach, things that human beings know intuitively are actually principles of different sub-systems and their ways of interaction related with these principles and parameters. Thus, the things which are learned are actually the parameters of the lexicon. As is well known, the aim of Generative Grammar is to define the language of the ideal speaker who is not influenced by non-language conditions. This is the reason why Universal Grammar tries to explain the aspects which are abstract and difficult to access. According to Chomsky, acquiring a language perfectly in a short time is only possible with Universal Grammar existing at birth and valid for all languages. So, while principles are the aspects of a natural language presented in human mind, parameters are the features differentiating a language from another one with definite boundaries. According to Chomsky s assumption, grammar has principles existing in all languages and starting from birth and parameters showing differences among languages. As these principles of UG exist from the very beginning and are not formulated through learning (or acquiring), speakers apply them to the language exposed to, no matter what language is acquired. In accordance with the principles and parameters hypothesis, Chomsky (1981) also makes a distinction between core grammar and peripheral grammar. According to his approach, different principles and parameters compose the different types of different core grammars. So, the core grammar is the common accessible point whereas periphery grammar is the point falling outside of this corporation. Predictions about the language acquisition process are made on the basis of the concept of markedness. The assumption is that the different settings of a parameter are assigned relative markedness values, and children initially set a parameter in its unmarked 2

5 setting. If we assume that all children, regardless of the language they are learning, start off with the pro-drop parameter, this would suggest that the pro-drop setting is the unmarked one. With the opposite assumption, non pro-drop setting would be the unmarked one. The problem in this issue is that there is no consensus among the researchers as to which setting is the initial, unmarked one Parameters Each parameter has a linguistic feature and parameters are set as a result of the input the speakers exposed to so that that he/she can access the core grammar of the language being acquired. There are at least three parameters in L2 acquisition (Yates, 1990) Open Parameter This type of parameter has two values, which are mutually exclusive. Yates (1990) illustrates this relationship between values like the following: A B Figure 1: Open Parameter According to Yates (1990), children have no preset value for an open parameter. It is truly open to the choice of children. Children pick a value for an open parameter when they encounter positive evidence from their L1. However, in L2 acquisition, situation is not as simple as in L1 acquisition. As L2 learners already have a parameter setting for their own L1, first they start out with their L1 setting and assume that the L2 has the same parameter setting in their L1. When they encounter positive evidence from the L2 that triggers the resetting of the parameter, they simply switch the setting from L1 value to the L2 value. 1 Commenting on the data she gathered from published examples of children s language, Hyams (1986) assumes that the initial setting is the pro-drop. Her assumption stems from the fact that English children at the earlier stages produced sentences which are not expressed. In the later stages of language development, English children learn that their language is non pro-drop and set the switch away from its initial value. Hyams explanation for how children do this is the occurrence of expletive pronouns (it, there etc.) in non pro-drop languages. She proposes that once children hear the expletive pronouns in their language, they realise that their language is non pro-drop. Hyams claim is that children set the parameter from the positive evidence only since English children acquire expletive pronouns at about the time they acquire for full lexical subjects. White (1985) argues that the pro-drop value is marked and supports her argument by pointing out the fact that non pro-drop languages require lexical pronouns, whereas pro-drop languages allow both null and lexical pronouns. Thus, if non pro-drop is the initial, unmarked value, it can be reset to pro-drop on the basis of the positive evidence coming from the environment. 3

6 1.1.2 Default Parameter This type of parameter involves a situation where the two parameter values (A and B) have a relationship in terms of markedness. The use of the term default comes from the fact that, unlike in the case of an open parameter, children have an initial preference for a certain value of the parameter of this type (Yates, 1990). They start out with a more restrictive and therefore unmarked value of parameter. When they encounter positive evidence from their L1 that requires a different type of setting, they switch the setting to a marked value of the parameter: A B Figure 2: Default Parameter The pro-drop parameter is a good example for this. In certain languages like Spanish, Italian and Turkish, there may be declarative sentences without apparent subject. However, it is not possible to drop a pronoun in languages such as English, French and German. If these languages are the case, then a surface subject must appear. Children start out with the unmarked pro-drop value assuming that their L1 has a + pro-drop value (Hyams, 1986). Therefore, at certain stages of learning, both pro-drop language learners and non pro-drop language learners may rely on the same default (unmarked) parameter. This default pro-drop setting will be switched into the marked non pro-drop setting only when children are exposed to the linguistic input comprising the positive evidence from their L1. The L2 acquisition of the pro-drop parameter has been studied by many researchers (White, 1985, 1986; Phinney, 1987; Bulut 1996; Bulut & Can, 2000). In White s and Phinney s research, they both focused on the view that L2 learners have either a direct access to UG principles (under the assumption that the pro-drop value is unmarked) or L2 learners transfer their L1 value. However, speakers of non pro-drop languages appear to support the transfer hypothesis. This is because if L2 learners have direct access to UG principles and L1 does not play any role in L2 acquisition, even speakers of a non pro-drop language are expected to produce subjectless sentences. Therefore, it seems to be that both transfer and UG principles are in action in L2 acquisition (Can, 2000) Subset Parameter The properties of this parameter stem from the Subset Principle, versions of which have been proposed by Berwick (1985) and Wexler and Manzini (1987). As defined by Wexler and Manzini (1987), for every parameter and every given two values of it, the languages defined by the values of the parameter are one a subset of another. In other words, the sets of sentences permitted by any two values of a parameter must enter a subset relation with 4

7 each other. That is, grammars generate sentences that are in a superset/subset relationship as can be seen in figure 3: B A Figure 3: Subset Parameter Here, the grammar that generates the sentences B, also generates A. A is proper subset of B. A sentences are compatible with two grammars: the grammar that generates A sentences and the grammar that generates B sentences. The subset principle predicts that the learner s first choice is to assume a smaller grammar, that is, the grammar that is a subset of the other. The subset parameter has also been studied within the context of L2 acquisition (Thomas, 1989; Cook, 1990). These studies focused on whether or not the subset principle operates in an L2 acquisition situation where subjects L1 has the superset value while their target language requires the subset value. Results show that if the subset principle operates in L2 acquisition, L2 learners start out with the unmarked subset value and therefore, have no difficulty in choosing the correct value of the parameter. On the other hand, if the subset principle does not operate in L2 acquisition, L2 learners will transfer their superset L1 value and have difficulty in picking up the correct value of the target language Initial Zero State / Steady State This term is actually an open question in terms of development whether a child starts with all the principles of UG available or whether these principles gradually unfold as part of maturation (Cook, 1985). However, when it comes to L2 acquisition, it is generally defined as the linguistics input learned intuitively and/or the grammar learned at the very beginning (White, 2003). It is stated that the steady state achieved by L2 learners is not the same with an L1 S s and it varies from one learner to another. Referring to the initial state of an L2 learner S i which includes the knowledge of L1, and his final state as terminal state (S t ) to distinguish it from the L1 S s, she summarises the differences between L1 and L2 learning as follows: L1 learning S 0. S S L2 learning S i S S Figure 4: Initial Zero State vs. Steady State 5

8 2. Universal Grammar in Second Language Acquisition 2.1 Initial State / Final State L1 and L2 acquisition might be considered to be similar since in both processes, learners are faced with the problem of making sense of the input. However, they are different in that in L1 acquisition, the child s task is to progress from the initial state of knowledge (S 0 ) to the steady state (S S ) of the adult s complex and subtle competence, whereas in L2 acquisition, the adult learner s initial state does not equal that of the child s. That is to say, L2 learner starts the acquisition process with the knowledge of his L1 present in his mind, and this knowledge might be interfering with the acquisition process. Although UG s Principles and Parameters theory has been able to set basic models for L1 acquisition, there hasn t still been reached absolute ones for L2 acquisition. However, according to many researchers studying L2 acquisition, UG s principles and parameters are still current for L2 learners (White, 1989; Flynn, 1987; Phinney, 1987; Bulut,1996). 2.2 Accessibility to UG: Four Hypotheses The description of L2 development within the principles and parameters framework of UG is based on underlying linguistic principles and parameters values and so, two languages may be compared through ways in which they utilize the same linguistic principles while fixing certain parameters differently. According to many researchers, UG is theory of the initial state of the language faculty, prior to any linguistic experience. In other words, the principles and unfixed parameters of UG constitute the initial state of an L1 learner. L2 acquisition represents the situation in which learners have already activated UG and have therefore gone through the process of parameter setting. Direct Access Hypothesis Direct UG accessibility assumes that L2 learners have direct access to UG. According to this view, L2 parameter values are fixed on the basis of L2 input alone, without any influence from L1. Thus, L2 acquisition is accepted to be the result of a direct interaction between UG and the L2 data (White, 1989). Indirect Access Hypothesis According to this view, access to UG is possible only through L1. However, many researchers also think that indirect UG accessibility claims that although the L! parameter values may initially be transferred to L2, they will eventually be changed into the values appropriate for L2 (White, 1996; Cook, 1993). One view, which is called the window-ofopportunity hypothesis, supporting the indirect access hypothesis claims that it is not entirely possible to access UG while acquiring an L2; it is only possible to access the UG principles but it is not possible to reset the parameters (Schachter, 1989). Resetting the parameters can not be performed after the period called critical period. 6

9 Weak Indirect Access Hypothesis In this hypothesis, UG principles and parameter values exemplified in L1 are available to an L2 learner (Schachter, 1996). However, this hypothesis differs from the previous one in that it considers L1 core grammar provide an L2 learner s only access to UG, for parameters not activated in L1 will be inaccessible and parameter resetting to the L2 values of parameters whose L1 values differ from the ones appropriate for the L2 will not be possible (Flynn, 1996; Schachter, 1996). No-Access Hypothesis This hypothesis assumes that an L2 learner s grammar cannot be defined in terms of UG principles and parameter at all, not even in terms of L1 parameter values at the initial state (Felix, 1991). This implies that L1 transfer cannot be described in terms of UG, that there is no evidence of UG transfer and that UG is thus no longer available to L2 learners. According to some researchers the L2 grammar, compared with the L1 grammar, is much more limited acquired (Schachter, 1989; Bley-Vroman, 1989). 3. Null Subjects and the Pro-Drop Parameter Empty categories are not only analysed with the topic of pronouns related with pro-drop, but also with the PRO, NP-trace and WH-trace. But in order to distinguish pro from the others, it would be useful to mention at the difference between the PRO and pro. 3.1 Null Subjects and PRO Control Theory, which is a sub-theory of the Government and Binding Theory, deals with the NP that show [+anaphoric, +pronominal] features. This NP, which is called PRO, can be proved by means of the theta role. According to this approach, when analysing the following sentences, it can at first seen that because of not having an argument, the sentences do not obey to the Extended Projection Principle which claims that a subject is required. But it can also be seen that even adding an NP fitting the argument structure will not be sufficient: (1) a. [Adam öldürmek] kötüdür. b. Ali [adamı öldürmek] zorunda kaldı. c. *[Kadın adam öldürmek] kötüdür. According to the Theta Criterion, there is a subject which does phonologically not exist: (2) a. [PRO adam öldürmek] kötüdür. b. Ali i [PRO i adamı öldürmek] zorunda kaldı. PRO can also be proved in accordance with the binding principles. When putting an anaphor instead of the NP adam in (1a), the anaphor will be bound to the PRO situated 7

10 before itself due to the Binding Principle A, which states that an anaphor must be bound to an antecedent inside the Governing Category: (3) [PRO i kendini i öldürmek] kötüdür. While the PRO in (3a) can be almost everyone, the PRO in (3b) must be Ali. According to this approach, the PRO in (1a) obeys the Binding Principle A by referring to a structure not included in the sentence and the PRO in (1b) obeys the Binding Principle B by referring to a structure included in the sentence. So, as stated before, PRO s have [+anaphoric, +pronominal] features. As it is not possible to value the PRO in accordance with the government concept, another way of interpreting it related with the Control Theory is the concept control. According to this approach, Ali in sentence (1b) is the controlling and the PRO is the controlled structure. Controlling which is compulsory like in these sentences is called obligatory control. However, the PRO in (4) can both be bound with the NP in the previous sentence and with another item outside the sentence. These kind of controls are called optional control. (4) Ali i [PRO i/j kapıyı açarken] onu gördü. It can also be seen that the PRO in sentence (1a) can not be bound to a specific NP. These types of arbitrary PROs control are called arbitrary control. 3.2 Null Subjects and pro Beside the empty category PRO, it is also possible to mention another one which is called pro. This kind of categories can be seen in pronoun-dropping languages such as Turkish. They are seen in sentences which have AGREEMENT and on the contrary of the PRO, they can have an apparent NP instead: (5) a. [pro adamı öldürmesi] kötü oldu. b. [onun adamı öldürmesi] kötü oldu. Whereas PRO is the only structure to be seen in subject position of an inflection with (-) AGREEMENT, pro can exist in subject position of an inflection with (+) AGR. Researchers like Taraldsen (1981), Rizzi (1982) and Chomsky (1982) claim that null subjects generally exist in specifier positions of an inflectional phrase which has (4) AGR. For example Turkish and Spanish are known as languages that have strong subject-verb AGR. On the contrary of PRO, pro are not only be seen in clauses, but they can also be seen in sentences: (6) pro adamı öldürdü. pro, which is an empty subject, does not exist in all languages. For instance in languages such as Turkish (7), Chinese (8), Spanish (9) and Italian (10), it is possible to drop the pronoun which is in subject position: 8

11 (7) a. O konuşuyor b. Konuşuyor (8) a. Ta shuo b. Shuo (9) a. Él habla b. Habla (10) a. Lui parla b. Parla However, it is not possible to drop the pronoun in languages such as French (11), German (12) and English (13): (11) a. Il parle b. *Parle (12) a. Er spricht b. *Spricht (13) a. He speaks b. *Speaks When it comes to comparing Turkish-type pro-drop languages with Chinese-type pro-drop languages according to AGR, it can be seen that although Chinese-type languages do not have a morphological AGR, it has been pointed that pronouns both in subject and object position can be dropped (Huang, 1984; Yuan, 1997; Wakabayashi, 1997). Most Turkishtype pro-drop languages, however, make it possible to drop the pronoun and to put an empty category which has features o + pronoun and anaphora. It has been mainly claimed that this category s anaphor is supplied with person AGR morphemes (Kornfilt, 1984; Özsoy, 1987). Another feature of most pro-drop languages is the subject-verb inversion feature. If the language is a (+) pro-drop one, then it is possible to move the subject to verb final position. But this feature is not possible if there does not a (+) pro-drop parameter exist: (14) a. O şarkı söyledi b. Şarkı söyledi o (15) a. He sang a song b. *Sang he a song Pro-drop languages are also said to have a that-trace filter. In Spanish, for example, the complementiser que (that), can be seen after the wh-movement (16), a feature which cannot be seen in non pro-drop languages surface structure (17): (16) Quien dijiste que vino? 9

12 (17) *Who did you say that came? According to Rizzi (1982, 1990), however, there are two important syntactic features related with null subjects. The first one is the subject-verb inversion such as in Greek (18), Italian (19) and Spanish(20) : (18) Petros pandreftike tin Ilektra the Peter married the Ilektra Peter married Ilektra Pandreftike o Petros tin Ilektra (19) Gianni ha telefonato John has telephoned Ha telefonato Gianni (20) El medico viene! The doctor is coming! Viene el medico! The second related syntactic feature according to Rizzi is the probability of extending an inserted subject to a complementizer in a question, which is not possible in English: (21) Chi credi [adıl telefoneral ] Who you think [that [adıl will telephone ]]? *Who do you think [that [adıl will tlephone]]? When it comes to the relation between the that-trace and pro in Turkish, researchers like Kornfilt (1984) and Sezer (1991) claim that there are empty operators in the C of a clause. For example, the empty component in relativisation, is related to the A through the empty operator. This can be illustrated as follows: (22) [[Polislerin e i tutukladığı] İş i ] adamları i yankesiciydi. CP Spec C' IP C polislerin e i tutukladığı İş i Another important feature related wit the pro-drop parameter are expletive pronouns, which fill the subject position of non pro-drop languages (23). However, pro-drop languages such as Turkish, do not show such expletive pronouns (24): 10

13 (23) a. There are children in the school. b. It is raining. (24) yağıyor yağmur yağıyor 3.3 Null Subjects and Reference of Pronominals The reference of null subjects and objects in pro-drop languages also differ very much from the reference of pronominals in non pro-drop languages. Huang (1984), for example, studied these types of empty objects in Chinese, Japanese and Korean. The researcher thinks that the empty object in Chinese (25-26), Japanese (27) and Korean (28) can not be bound as an antecedent to the NP in the clause or main clause. In Chinese (25-26), for example, the empty object can refer to a discursive context but the empty subject, can only be bound to an NP as an antecedent coming before it. (25) Lisi hen xihuan e. L. very like Lisi i likes [him] D,*i very much. (26) Zhangsan shuo [Lisi bu renshi e]. Z. say L. not know Zhangsani said that Lisi did not know [him]d,*i. (27) John-wa i [ e i siken-ni otita] no omada siranai. J. exam failed that yet not-know John i still doesn t know that [he] i failed the exam. (28) John-i [ e e po-ess-ta-ko] malha-ess-ta. J.-Subjsee-Past-Decl-Comp say-past-decl John1 said that [he] i,d saw [him] D,*i. As the morphological approach was sufficient for many researchers, it has been started to look for other explanations recently. Hawkins (2001), for example, claimed that Chinesetype languages, differ from Greek-type languages and English-type languages by means of two parameters. The first of these is the topic feature in C, which needs a constituent that must be seen in CP s specifier. According to the second parameter, C in Chinese, is licensing a null constituent, which is identified by the context of the utterance. Hawkins (2001) thinks that if this approach is true, then English learners of Greek-type languages will have to reset the parametric differences related with the inflection while speakers Chinese-type languages will have to reset the parametric differences related with C. When it comes to the roles of possible antecedents Turkish, it is seen that the subject position of the antecedent has a constraining feature related binding. The following 11

14 sentences for example are ungrammatical, except that the pronominal o can be bound to an antecedent outside the sentence: (29) a. *Ali i [onun i kitabın]-ı verdi. b. *Ali i [onun i konuştuğun]-u söyledi. The reason why these kind of sentences are ungrammatical in Turkish, is the impossibility of an overt pronoun; only a pro can exist: (30) a. Ali i [pro i kitabın]-ı verdi. b. Ali i [pro i konuştuğun]-u söyledi. Although sentences (29b) and (30b) show that it is not possible to bind the pronominal o to an antecedent in subject position, it can be seen that the antecedent o in sentence (31) can be bound to an NP in subject position. The pronominal in subject position in sentence (32), also, can be bound to an antecedent in object position: (31) Ali i [Ayşe nin j onu i/*j gördüğün]-ü söyledi. (32) Ben Ayşe ye i [onun i yaptıkların]-ı anlattım. Sezer (1991), explains the reference relations of the pronominal o with a domain which he calls Obviation Domain and makes the following definition: The pronoun o in subject poisition, refers to another subject with the smallest T, which includes a host. The host mentions here, means the NP/T including the pronoun in subject position with the narrowest domain. 3.4 Null Subjects and Second Language Acquisition As L2 researchers try to explain the grammar with UG theory, most syntactic features of L2 s, such as the pro-drop parameter, are handled with UG s Principles and Parameters Hypothesis. For example White (1985, 1986) and Phinney (1987) studied the resetting of the pro-drop parameter of Spanish, Italian and French speakers who learned English as a second language. While Spanish and Italian are pro-drop languages, French is a non prodrop one. It was seen that at the beginning of the acquiring, Italian and Spanish speakers tend to produce sentences without subjects. It can be stated that these results resemble the L1 acquisition process as the speakers started with an unmarked feature of the (+) pro-drop parameter which is a feature supporting the direct access hypothesis. Another possibility is that the speakers transfer these features from their L1. The only way of proving this idea is looking at the results of the French speaking group. This is because French, is a non prodrop language and according to the direct access hypothesis, these speakers are expected to start with the unmarked value which means composing sentences without subjects at the very beginning. But the results of the French speaking group showed that they composed sentences which have subject. This means that there is transfer in resetting the pro-drop parameter and it is possible to say that these results support the indirect access hypothesis. 12

15 Liceras (1989) states that the peripheral aspects of grammar, the marked ones, should appear while acquiring L2 and their acquisition should be much more difficult. At this point, White (1985) stated that if the UG parameters which adult L2 learners acquired while acquiring their L1 do not operate in L2, then there will be big difficulties. She thinks that the learner will transfer the parameter from L1 to L2 and transfer errors will occur. Studies related with the pro-drop parameter are relatively new. There have been some studies searching for results related with the pro-drop parameter while being a pro-drop language speaker such as Italian and Spanish and acquiring a non pro-drop language such as French and/or English as an L2 since the 1980 s (White, 1985; Hilles, 1986; Phinney, 1987). There have also been some studies stating the processes of being a non pro-drop L1 speaker such as English and/or French and acquiring a pro-drop language such as Italian and/or Spanish as L2 (Liceras, 1989; Phinney, 1987). Studies which look for the results of being a non pro-drop L1 speaker such as English while acquiring a pro-drop language such as Turkish as L2 are also rather new (Bulut and Can, 2000; Gürel, 2002). One of the important studies in this field is White s (1986) study which searched for the results of being both pro-drop L1 speakers such as Spanish and Italian and non pro-drop L1 speakers such as French, acquiring a non pro-drop L2 language like English. 32 Spanish L1 speakers and 2 Italian L1 speakers composed the first group whereas 37 French L1 speakers constituted the second group. All the participant were attending to English L2 classes of elementary and advanced levels. As both French and English are non pro-drop languages, one might expect at the beginning that the two groups were going to get quite different results. But this was not the case. White asked the participant to take two tests, a grammatical judgement test consisting of 28 sentences and another test consisting of 12 sentences asking for specific phrases. It was found that there was an L1 effect in L2 acquisition and that Spanish and Italian L1 speakers transferred the null subject and thattrace features to English as an L1 much more than L1 speakers of French. White (1985), similarly applied a grammatical judgement test to 54 L1 speakers of Spanish and 19 L1 speakers of French learning English as L2. But this time she tried to choose the participants from three different levels: Elementary, Intermediate and Advanced. It was found that L1 speakers of Spanish who were in the elementary level of English as L2, had a greater tendency of transferring an overt subject to the target language than L1 speakers of French. However, it was also seen that on the contrary of L1 speakers of French, L1 speakers of Spanish transferred the null subject feature to the target language English much more. Phinney (1987) found that Spanish participants were gradually able to state a difference between types of pronouns in accordance with the L2 learning level. Expletive pronouns such as it and there, for example, were more rarely used than referential pronouns such as I, you, he, she etc in advanced English L2 classes. After studying on two 12 year long English L2 students written data, he found that referential pronouns were 13

16 only 6% and 13% NOT used where necessary. Expletive pronouns, however, were 56% and 76% NOT used and this supported the approach stating that these features are lightening the UG theory as expletive pronouns are null in Spanish and overt in English. If they were not analysed by means of UG, it would rather difficult to explain the delay of the acquisition of expletive pronouns. The results of Tsimpli and Roussou s (1991) study, analysing the data of L1 speakers of Greek, which is a pro-drop language, acquiring English as an L2 have similarities with the ones before. 6 intermediate and 7 advanced learners of L2 English were given a grammatical judgement test consisting of 30 sentences related with the pro-drop parameter and 10 grammatical Greek sentences which included the null subject feature, the subjectverb inversion feature and overt complementisers and which were asked to translate them into English. It was seen that all the participants refused the null subjects in the grammatical judgement test and translated them also overtly to English. They also did not accept any subject-verb inversion and translated them according to the English word order characteristic. More than 95% of the participants, also, do not approve the that-trace feature both in the grammatical judgement test and the translations. Although there is no morphological AGR in Chinese, it is possible to drop the pronoun both in subject and in object position (Huang, 1984; Yuan, 1997 and Wakabayashi, 1997). Studies of Chinese Korean and Japanese speakers and Greek speakers acquiring a non prodrop L2 such as English showed that Chinese L1 speakers are much more successful in learning the overt subject feature of English than Greek L1 speakers (Hawkins, 2001). As it can be seen from the above stated studies, there is no parallelism in acquiring all features of the pro-drop parameter and the L2 level is gains much importance. When it comes to Turkish, it can be seen that subjects can be dropped optionally in Turkish because the inflection at the end of the sentence already indicates the subject itself. As Chomsky (from Braidi, 1999) stated, the pro-drop parameter is set at the beginning and children (D 0 ) start with the (+) pro-drop parameter feature. Learners start to reset the correct parameter which is already available in their minds while getting the correct input. According to this approach, a marked language which does not allow null subjects, is a sublanguage of an unmarked language that allows both null subjects and overt. One of the important studies related with the pro-drop parameter in Turkish as an L2 was performed by Bulut (1996), which analysed the data of Turkish L1 speakers acquiring English as L2. The findings supported the indirect access hypothesis because they showed that the learners access UG by means of their L1. Similar results were achieved in Bulut s and Can s study (1999), which indicated that participants have an indirect access to UG and reset the parameters. According to the researchers, these results also supported the windowof-opportunity approach. In Bulut and Can s (2000) another study, however, English L1 speakers s acquiring Turkish as L2 were analysed and it was shown that, on the contrary of Bulut (1996) and 14

17 Bulut and Can (1999), participants could not transfer the English pro-drop parameter value to Turkish and so, that they did not use their L1 settings. The difference of the adults s valuing the pro-drop parameter in English L1 and Turkish L1 can be explained with the subset principle. The participants started acquiring the pro-drop parameter in L2 with the unmarked feature and on the contrary situations, as stated in Bulut (1996) and Bulut and Can (1999), transfers existed. Another study performed by Öztürk (2000) related with the acquisition process of the pro-drop parameter in L2 Turkish distinguishes the ones stated above. According to the writer, Turkish is NOT a pro-drop language and overt pronouns are actually topic pronouns completely based on discursive aims, which do not have subject function sentences. Öztürk also claimed that subject agreement morphemes have a pronoun feature and are the real subjects of the sentence. These agreement morphemes are generated in VP and turn into a morphological state by means of fusion and are adjuncts of the verb. In Enç (1986), it was claimed that overt pronouns in Turkish have pragmatic functions to be learned by the L2 learner in certain contexts and that these pronouns distributions are NOT optional or obligatory. However, according to Erguvanlı-Taylan (1986), when subjects have a topic changing feature in discourse, the usage of overt pronouns is obligatory and empty categories lead to ungrammatical structures. 4. The Present Study 4.1 Research Questions and Predictions This study is mainly trying to analyse the effect of resetting the parameters when both L1 and L2 have the same parameter but have different AGR features. In other words, it will be analysed if there is any difference in resetting the pro-drop parameter between L1 speakers with strong AGR and L1 speakers without AGR while acquiring Turkish as an L2, which has strong AGR. If L2 learners setting the pro-drop parameter happens without depending on AGR, then it will be expected to related with another stimulus. This study aims to establish a relation with the pro-drop parameter and (i) the that-trace filter and (ii) subject-verb inversion. It is expected to achieve a parallelism in the success of L2 learners pro-drop features and the that-trace filter in accordance with subject-verb inversion. If L2 learners resetting the pro-drop parameter without depending on AGR does exist, then L1 speakers of Chinese-type languages are be expected to be more successful than the other ones. And this will mean that L2 learners reset the pro-drop parameter related to other stimulus. The reference relations between pro-dropping and pronominals of Turkish L2 learners will also be analysed. As L1 speakers of Chinese-type languages have empty subjects with a discursive reference, empty subjects can both be bound to the subject and the object and even another NP, which is not possible in Turkish. The pronoun in subject position in 15

18 Turkish, can only be bound to a NP in subject position or another NP. So, it will be expected that L1 speakers of Chinese-type languages will be less successful than L1 speakers of Turkish-type languages because they will be expected to bind the pro to a NP in object position. The following research questions related with the pro-drop parameter were formulated for the present study: 1. Is there a significant difference in resetting the pro-drop parameter between L1 speakers of Chinese-type languages and L1 speakers of Turkish-type languages while acquiring Turkish as L2? 2. Is there a specific relation between resetting the pro-drop parameter and acquiring the AGR, subject-verb inversion and that-trace filter features? This study also searched for results related with pro-dropping or not pro-dropping: the strategies that L2 learners follow while using pronominals. In accordance with this question, a question related with reference relation was also formulated as the following: 3. What are L2 learners performances related with the referential differences between pro-drop and pronominals? 4.2 Subjects In this study, which is mainly concerned with the pro-drop parameter in L2 acquisition, subjects were chosen exclusively from pro-drop languages like Turkish (Afganian, Persian, Mongolian, Romanian, Russian, Ukrainian, Arabic, Spanish) and like Chinese (Chinese, Japanese, Korean). All of the subjects were attending TÖMER (Centre for the Teaching of Turkish) and were learning Turkish as an L2 at 3 different proficiency levels (elementary, intermediate and upper-intermediate). There were 17 L1 Turkish-type language speakers (Group 1) and 3 L1 Chinese-type language speakers (Group 2) in elementary L2 classes (total 20); 15 L1 Turkish-type language speakers (Group 1) and 8 L1 Chinese-type language speakers (Group 2) in intermediate L2 classes (total 23) and 19 L1 Turkish-type language speakers (Group 1) and 5 L1 Chinese-type language speakers (Group 2) in upperintermediate L2 classes (total 24). The subjects were given a grammatical judgment test which included 35 sentences related with pro-drop, subject-verb inversion and that-trace features and another grammatical judgment test which included 18 questions referring to null subjects and referential features of pronominals. All the participant were asked to fill in an inquiry form concerning their age, education, occupation, the period of staying in Turkey, the age of learning Turkish and being able to speak another foreign language etc. Special attention was paid to the adults ages as all of them were chosen according to their starting learning L2 Turkish after the critical period. In order to approve the tests reliability, both of them were also applied to a controlling group consisting of 10 L1 Turkish speakers. 16

19 4.3 Materials and Procedures Each subject was given two test. The first one was a grammatical judgement test including 35 sentences related with pro-drop, subject-verb inversion and that-trace filter in Turkish. Here the aim is finding results for resetting the parameter in case the participants both L1 s and L2 s share the same parameter but have different AGR features. In other words, it will be analysed if there is any difference in success in resetting the pro-drop parameter while having strong AGR in their L1 or not. The participants got 30 minutes to judge the test and were asked to put tick next to the sentence that sounded correct to them, put a cross if they thought it is not expressed in good Turkish and put a question mark if they could not decide. In order not to draw attention to a single structure, sentences which were not directly related with the topic were also added to test. The participants took the test in groups in a classroom environment. While evaluating the test, each correct answer of the learners was graded with one (1) point and both the incorrect answers and even question mark put (?) answers were graded with zero (0) points. After rating the total points, the percentage rate of the correct answers were clarified. The second test consisted of 18 sentences including pronouns or overt pronominals in subject position. The participants were asked questions aiming to find the antecedent which the components were bound to and were given 15 minutes to answer it in groups in a classroom environment. The test included questions based on both pro and pronominals related with sentences and also clauses and the subjects were asked to find if the possible antecedent in the sentence or clause was related with the subject, object or another NP by means of choosing the correct answer. Evaluating the second test was also performed by grading each subject s correct answer with one (1) point and incorrect answer with zero (0) points so that the percentage rate could be set. The percentage rate of the correct answers were rated by means of dividing them into number of participants. 5. Results 5.1 Findings of Test 1 When comparing the Turkish L2 data of L1 speakers of Turkish-type pro-drop languages (Group 1) with the data of L1 speakers of Chinese-type languages (Group 2), it can be seen that there is no increase or decrease in acquiring the pro-drop in L2, rein accordance with the learners proficiency level. Both groups were much more successful at answering the questions related with prodrop in sentences, rather than clauses. Although not much difference was stated in correct answers related with pro-drop in sentences, there was a notable difference in the ones related with pro-drop in clauses. While Group 2 s score reached to (60, 68%), Group 2 s score remained at (48,39%). 17

20 It was also found that Chinese-type language L1 speakers were much more successful (39,17%) in choosing the correct answers related with subject-verb inversion than Turkishtype language L1 speakers (5,94%) while acquiring Turkish as L2. When comparing the scores of the subject-verb inversion questions with the ones of pro-drop questions, it was seen that there was no relation between subject-verb inversion and pro-drop. However, although it was found that subjects of Chinese-type languages were more successful in questions related with subject-verb inversion, there was no considerable success in questions related with pro-drop. When comparing both the scores of the that-trace filter questions with the ones of subject-verb inversion of both groups, it was seen that there was a higher relation between these two types of questions than the relation between the ones between pro-drop and subject-verb inversions. Speakers of Chinese-type languages were more successful (47,18%) in all questions related with the that-trace filter than subjects of Turkish-type languages (27,78%), both in subject and object position. Answers given by both groups related with pro-drop questions were rather similar to the ones related with subject-verb inversion. But it can be thought that most subjects did not make a conscious choice and made an overgeneralization in wrongly inverted questions. This might be due to the fact that most pro-drop languages do have the inversion feature as a syntactic property. Subjects of Turkish-type L1 speakers performed a score of (33,37%) at wh-phrases being in subject position and a score of (22,20%) in object position. Subjects of Chinesetype L1 speaker, also, reached a score of (43,61%) at wh-phrases being in subject position and a score of (50,77%) in object position. 5.2 Findings of Test 2 L1 Speakers of both Turkish-type languages and Chinese-type languages performed a regular increase in success of questions related with pro in subject position in accordance with their proficiency level. This regular increase, supports the hypothesis of resetting parameters in L2 acquisition. However, despite the regular increase in success of both groups, the subjects of Chinese-type languages performed a lower success rate in quantity than the subjects of Turkish-type languages. When comparing the scores of L1 speakers of Turkish-type languages (Group 1) with the ones of L1 speakers of Chinese-type languages (Group 2) related with the questions of pronominals in subject position, it can be seen that subjects of Group1 performed a regular increase in success, in accordance with their L2 proficiency level. Participants of Group 2, however, performed a regular decrease in success at answering the questions related with binding it to another NP, in accordance with their L2 proficiency level, although they performed again a regular increase in success at binding it to a subject. Subjects of Turkish-type L1 languages performed a similar success in questions related with the reference of the NP s pronominal subjects in clauses, both at the elementary 18

21 level and the intermediate level. But participants of these languages (Group 1) attending to the upper-intermediate level classes, performed a much better performance. Participants of Chinese-type L1 speakers, however, performed here again a regular increase of success in questions related with the reference of the NP s pronominal subjects in clauses. 6. Conclusion In conclusion, the present study looked at how UG exists in adult second language acquisition starting off the pro-drop feature in subject position. It has been tried to state the relationship between the pro-drop parameter and syntactic features such as subject-verb inversion and the that-trace filter while acquiring a second language. This study also searched for the results of the pronoun as an empty category and the reference features of pronominals in subject position when it is not possible to drop a pronoun in L2 acquisition. The study analysed the syntactic approaches of acquiring Turkish as L2 by means of the pro-drop parameter, which is a developmental problem in L2 acquisition. In this study, which is mainly concerned with the pro-drop parameter in L2 acquisition, subjects were chosen exclusively from pro-drop languages like Turkish (Afganian, Persian, Mongolian, Romanian, Russian, Ukrainian, Arabic, Spanish) and like Chinese (Chinese, Japanese, Korean). All of the subjects were attending TÖMER (Centre for the Teaching of Turkish) and were learning Turkish as an L2 at three different proficiency levels (elementary, intermediate and upper-intermediate). The first research question was asking if there was a significant difference in resetting the pro-drop parameter between L1 speakers of Chinese-type languages and L1 speakers of Turkishtype languages while acquiring Turkish as L2. According the data collected from the subjects from Turkish-type languages (Group 1) and the data of the subjects from Chinese-type languages (Group 2), it was found that Group 2 s choices related with resetting pro-drop parameter were much more successful. This does not support the widely supported claim of the indirect access hypothesis, according to which it is possible to reset parameters only before the critical period It provides support for the claim that there is no L1 transfer in L2 acquisition. It might also be the case that L2 learners have access to UG, just like L1 acquisition, which means that they have direct access to UG without applying to L1 parameters. However, it was found that there was no difference in the success of both groups choices related with pro-drop questions in accordance with their levels. According to the direct access hypothesis, however, access to UG is expected to regular, just like in L1 acquisition and L2 parameters are expected to be set without any effect of L1 (White, 1996). The results here show that UG and L2 have a direct connection in L2 acquisition and there is no unique explanation for the theories of access, at least related with the pro-drop parameter. The second research question was searching for the answer if there does exist a specific relation between resetting the pro-drop parameter and acquiring the AGR, subjectverb inversion and that-trace filter features or not. The data collected from both groups 19