Transfer and access to universal grammar in adult second language acquisition

Summary This dissertation focuses on the roles of first language transfer and Universal Grammar in adult second (or foreign) language acquisition. It contributes to the ongoing debate whether second language acquisition is constrained by Universal Grammar. According to generative linguists, Universal Grammar (UG) is innately given knowledge which constrains first language acquisition. Assuming the correctness of this approach, generative second language researchers are interested in the question whether UG continues to constrain non-primary language acquisition by adult learners. Positions on the accessibility of UG in second language acquisition range from those who assume that UG operates in second language acquisition as it does in first language acquisition, to those who assume that UG plays no part at all in second language acquisition, either directly or indirectly via the first language grammar. There are six partially overlapping positions, which differ from each other with respect to the roles they attribute to first language transfer and (direct) access to UG.1. No Transfer/No Access: UG is not involved at any stage of second language acquisition, either indirectly via the instantiations of the first language or directly.2. No Transfer/Full Access: the instantiations of the first language do not constitute the learner’s initial theory about the second language (no transfer). Both used and unused properties of UG fully constrain second language acquisition (full access).3. Partial Transfer/No Access: some, but not all properties of the first language transfer, and second language is only constrained by those UG properties that do transfer.4. Partial Transfer/Full Access: parts rather than all of the first language transfer, and UG fully constrains second language development.5. Full Transfer/No Access: all instantiations of the first language make up the initial theory about the second language; UG constrains subsequent development only via the first language instantiations.6. Full Transfer/Full Access: all instantiations of the first language make up the initial theory about the second language and UG fully constrains subsequent second language development.The aim of this dissertation is to arbitrate between these positions. For clarity’s sake, the questions of transfer and UG access are teased apart.On the one hand the study tries to define to what extent the native grammar determines the non-native grammar of adult learners of a second language. Which position best captures the role of the first language: No Transfer, Partial Transfer or Full Transfer? On the other hand, this study investigates whether adult second language learners acquire previously unactivated options of Universal Grammar. In other words, which is the correct hypothesis: No Access or Full Access?Research questions: do learners transfer and acquire parameters of UG?Throughout this study, the focus is on parameterised properties of UG. Generative linguists account for some differences between languages by incorporating a limited number of options into UG, called parameters. A parameter links a range of seemingly unrelated properties. This greatly reduces the learning task, since parameterised properties do not have to be learnt one by one; instead, UG parameters provide the child with innate, advance knowledge of a limited range of options. The input data helps to decide between the parameter values and triggers setting of the appropriate parameter value. The cluster of properties related to the chosen parameter value then follows automatically.Parameters provide a suitable testing ground for the questions of transfer and UG access that this study addresses. Firstly, parameters indicate the operation of UG. Parameters are assumed to involve formal properties too subtle and abstract to be learnt on the basis of input alone. Evidence of parameter setting is thus evidence for UG. Secondly, the idea of various parameter values can capture variation between languages. By focusing on a learning situation with contradictory parameter values, it is possible to determine whether second language learners start off with one parameter value and switch to another value later. Thirdly, the assumption that parameters involve clusters of superficially unrelated properties allows for strong, falsifiable research hypotheses. If a complete cluster of properties associated with the L1 parameter value carries over into the L2, this suggests that the L1 parameter value transfers; in other words, clustered transfer is evidence for the position of Full Transfer. Likewise, clustered acquisition of the properties associated with the target language (or non-first language) parameter value is evidence for the position of Full Access. However, if there is evidence for transfer and/or acquisition of some properties, but not of the complete cluster of properties, the hypotheses of Full Transfer and Full Access are refuted. The idea of clustering is thus central to the four research questions of this study :1. Is the L2 initial state defined by clustered transfer of properties that relate to L1 parameter values?2. What are the implications of the empirical results of the present research for theories of Partial Transfer and Full Transfer?3. Is L2 development characterised by clustered acquisition of properties that relate to TL (or non-L1 parameter values)?4. What are the implications of the empirical results of the present research for theories of No Access and Full Access?The Null Subject ParameterThe parameter investigated in this study is known as the Null Subject Parameter (or the Pro-drop Parameter). This parameter captures the phenomenon that some languages may omit subject pronouns, while other languages require subject pronouns to be expressed. For instance, Italian allows the subject pronoun to be dropped in main and subordinate clauses, whereas their English counterparts require an overt subject pronoun. Moreover, Italian may leave the expletive subject unexpressed, while English must use overt expletives.The Null Subject Parameter is amenable to investigation because the constraints on empty elements such as null subjects are taken to be too abstract to be learnt consciously and are therefore related to UG. Moreover, there is a rich body of second language acquisition research on this parameter which can serve as a point of departure for this study. Finally, and crucially, the Null Subject Parameter relates a cluster of properties. Chapter two presents an overview of other second language acquisition studies on the Null Subject Parameter.Chapter 3 introduces the version of the Null Subject Parameter adopted in this study. Although there are different versions of the Null Subject Parameter, all accounts agree on the following: languages like Italian and Spanish, which allow null subjects, represent the [+ null subject] value of the parameter, whereas languages like English, which require lexical subject pronouns, represent the [- null subject] value of the parameter. Most versions of the parameter also share the assumption that its [+ null subject] value clusters with the property of rich verbal agreement. This study further includes subject-verb inversion and (SVAO) word orders that indicate verb raising in the cluster of [+ null subject] properties. Throughout this study, I adopt a version of Speas’ (1995) null subject theory which accounts for this cluster of typical null subject phenomena in terms of two syntactic properties: empty SpecAgrP and V-to-Agr raising.Multiple case studies: Romance learners of Germanic languagesChapter 4 describes the longitudinal data of nine adult Romance learners of Germanic languages: Spanish learners of Swedish and Italian learners of German and English.122 The Romance languages Spanish and Italian have[+null subject] properties: missing subjects, subject-verb inversion and word orders with the finite verb preceding the adverb (SVAO); the Germanic languages Swedish, German and English generally lack these properties and are considered to exhibit the [- null subject] parameter value. The multiple case studies investigate to what extent the properties of the first language [+ null subject] parameter value carry over into the [- null subject] second language. Moreover, the data are also searched for evidence of acquisition of the cluster of properties related to the [- null subject] parameter value of the second language. The key word in this quest is clustering. Only clustered transfer of the omission of subjects from main and subordinate clauses, subject-verb inversion and placement of the finite verb before the adverb counts as evidence for Full Transfer. Likewise, only clustered loss of these interrelated properties counts as evidence for Full Access.Conclusion: Full Transfer, No AccessChapter 5, discusses the empirical results and the implications of these results for the hypothesis of Full Transfer. The results show that the frequency of occurrence of the parameterised properties varies among the learners, but regardless of the language pairs (Italian-English, Italian-German, Spanish-Swedish). Despite this quantitative variation, all nine learners produce utterances with missing subjects in main and subordinate clauses, post-verbal subjects and (X)SVAO adverb placement orders. Moreover, these properties are found throughout the two-year-period during which the data were collected. This strongly suggests that the properties of the learners’ Spanish/Italian [+ null subject] parameter value transfer as a cluster. Disconfirming No Transfer and Partial Transfer, the results support the notion of Full Transfer.A secondary aim of this study was to look into the question of UG access. There is no evidence suggesting that the [+ null subject] parameter value is unlearnt by any of the learners involved in this study. Despite the fact that the data cover approximately two years, there are no innovations in the relevant properties of any individual learner which characterise a new stage in his or her development. In fact, the term development may not be appropriate for the data that were studied: the absence of new phenomena suggests a lack of development in the properties under investigation. Since the results fail to provide evidence for Full Access, they are best accounted for by the notion of No Access.