Wm. . H _,,.. n..~o~4wt P , 2, ,. g. a _. . _ ,V a . , . . r. ‘ a m . h‘ : ~ ‘ . . . V w .1. M i . L ‘ V. L . v . . o. .. n, .. ,, 4 .. . ":3 ljmalnlnlrl- / L” {’3 ”1'7 ' LIBRARY Michigan State Unlverslty This is to certify that the dissertation entitled THE ACCESSIBILITY 0F UNIVERSAL GRAMMAR IN LANGUAGE ACQUISITION: A CROSS-LINGUISTIC PERSPECTIVE presented by Ahmed Abdullah M. AlaBanyan has been accepted towards fulfillment of the requirements for Ph.D. degree in LingUiStiCS Major professor Date Sept. 23. 1996 MS U is an Affirmative Action/Equal Opportunity Institution 0-12771 CE IN RETURN 30X to move this checkout «om your record. or before data duo. PLA To AVOID FINES Mum on DATE DUE DATE DUE ‘ DATE DUE THE ACCESSIBILITY OF UNIVERSAL GRAMMAR IN LANGUAGE ACQUISITION: A CROSS-LINGUISTIC PERSPECTIVE By Ahmed Abdullah M. Al-Banyan A DISSERTATION Submitted to Michigan State University in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY Department of Linguistics and Languages 1996 D: No 8t] Stl‘ ABSTRACT THE ACCESSIBILITY OF UNIVERSAL GRAMMAR IN LANGUAGE ACQUISITION: A CROSS LINGUISTIC PERSPECTIVE By Ahmed Abdullah M. Al-Banyan I:l'his research endeavor seeks to examine the acquisition of English by Arabic native speakers and the acquisition of Modern Standard Arabic (Arabic) by English native speakers. It explores the theory of Universal Grammar (DO) and its relationship to the acquisition processes of first (1.1) and second language (L2). Particularly, it emloys a principles and parameters approach to UG, as realized in Chomsky's (1981) Government and Binding theory. It investigates whether or not advanced adult L2 learners as a foreign language have access to D6 principles and parameters; specifically, the Subjacency and the Empty Category Principles, and the Null Subject Parameter are tested. In addition, the study researches whether or not there is a difference between perception and production tasks in measuring UG principles/parameters. Moreover, it discusses the nature of DG-based acquisition studies and comments on DG properties that considered in such studies . 1777.. _h .4.. .___- l- . ..- Por data collection, two testing instruments are utilized, a perception (grammaticality judgment) task, and a production (question formation) task. These tasks are constructed in.English for the English.Experiment (EE), and in Arabic for the Arabic Experiment (AE). For each experiment, two groups (controls and subjects) are given the two tests. The subjects are 60 male adults who are advanced learners of English in Riyadh, Saudi Arabia (for the EE), and 34 male/female adult advanced learners of Arabic in the USA (for the AE). The collected data are tabulated, and descriptivo statistics, Regression, and Chi-square are computed for data analysis. The study reports that DC is still accessible to adult L2 learners but its accessibility is partly hindered by late acquired linguistic, cognitive, and socio-psyohological components. In comparison, child L1 learners can directly access Core grammar (06) while adult L2 learners access core grammar through late acquired peripheral components. Among other things, the study also reports some factors that may condition functional computation of components in the bilingual mind/brain. Copyright by Ahmed Abdullah E. Al-Banyan 1996 Dedicated To my beloved wife and children ACKNOWLEDGMENTS I would like first and for most to thank Allah, the almighty, for the health, protection, and help He bestowed on me throughout my life. I also would like to extend my deepest thanks and gratitude to the chairperson of my doctoral committee. I am greatly thankful for the guidance, encouragement, and support he provided throughout my graduate study. Eis constant advice and suggestions have made a difference in my academic life. He is not only a linguist and teacher but also a big brother and mentor to all his students. This is Dr. Dennis R. Preston. I am greatly indebted to the members of my doctoral committee: Drs. Grover Hudson, Yen-Ewei Lin, and Susan Gass for their valuable cements and suggestions on an early draft of this dissertation. Their influence on this work is incalculable. I also would like to thank Dr. Charlene Polio, the outside reader of this dissertation, for her praise and CMtI . vi I am also thankful to Imam Mohammad Ibn Saud Islamic university of Riyadh, Saudi Arabia for granting me a scholarship to pursue my higher studies in the USA. My gratitude is also extended to Michigan State University, represented by the Department of Linguistics and Languages, for giving me a Teaching.Assistantship and a College Research Fellowship. I also would like to thank five people who helped.me‘with the data collection of this study, namely-Dr. Mahdi.Alosh (the Director of the Arabic Language Program at The Ohio State University of Columbus, Ohio), Dr. Roger Allan (the Director of the Arabic and Islamic Studies Program.at the Univarsity of Pennsylvania of Philadelphia, Pennsylvania) and his assistant Dr. Sabah Gandour, Dr. Taj-Alsir Eamza Idris (the Director of the program.of Teaching Arabic to Speakers of Other Languages of Fairfax,'virginia), and.Dr. Omar.AleAfaleg (the Director of the English. Language Center at the Institute of Public Adminstration of Riyadh, Saudi Arabia). moreover, I would like to thank all the students for participating in this study. Last but not least, my appreciation and gratitude go to my wife, ana, and my children, Leena and Mbtaz, for their moral support and endless love that kept me going. I also vii want to thank my mother, Tarfa, my mother-in-law, Hessa, my father, Abdullah, brothers and sisters for their continual praise and encouragement. May Allah bless them all. viii TABLE OF CONTENTS CW1 ImomCTIONOO..OOOOOOOOOOOOOOOOOOOOO00.0.1 1 1 Background................................1 1 2 Plan Of The Study.........................7 1.3 Rationale For The Study..................10 1 4 Research Questions.......................15 1 5 Testable Hypotheses......................15 cmz LITERATURE “VIEWeeeeeeeeeeeeeeeeeeeeeeee19 Background...............................19 UG As Realized In GB Theory..............21 The Subjacency Principle.................27 The Empty Category Principle.............29 The null Subject Parameter...............31 UG-based Studies In L1 Acquisition.......35 UG-based Studies In L2 Acquisition.......42 Child L2 Acquisition 8tudies.............43 Adult L2 Acquisition Studies.............47 The an-Accessibility Of UG Studies......48 The Pull-Accessibility of DC Studies.....54 The Partial-Accessibility of U0 Studies..63 Summation................................74 IO. 00.... O. 0.. (UMP e e NNNNH eee UMP NNNNNNNNNNNNN O mhhhhbbWNNNNP 5; lmDmLMYeeeeeeeeeeeeeeeeeeeeeeeeeeeeee77 Subjects.................................77 Group One................................78 Group Two................................79 Group Three..............................81 Group Pour...............................82 Research Instruments.....................86 The Grammaticality Judgment Test.........88 The Question Formation Test..............91 Procedures...............................93 methods Of Data Analysis.................95 co. co .UNH O O O O NIH ucscutnuouiuiwcuiu O O Ob93538383F|FAF|H*H 0 ix Coding And Scoring.......................95 statistical TestBOOOOOOOOOOOO0.00.00.00.09? CHAPTER 4 DATA ANALYSIS AND RESULTS...............100 bhbhbbhhhhbbhhbhhh g UIMUI O O O O O O O O O I O 0 O O O O O ht»IOAJNDNIMIOIJhiH|HI‘E‘F‘P‘H‘H UNH e e N P 00 UNNNHPH e O 0 MIA 0 O O N H e e N H WNNNHHH The English Experiment..................100 The Control Group.......................100 Results Of The Two Tests................102 Across Test Comparison..................107 The Experimental Group..................108 Results Of The Two Tests................111 Across Test Comparison..................116 Across Group Comparison.................117 The Arabic Experiment...................121 The Control Group.......................121 Results Of The Two Tests................123 Across Test Comparison..................126 The Experimental Group..................127 Results Of The Two Tests................130 Across Test Comparison..................135 Across Group Comparison.................136 Across Experiment Comparison............141 Foreign Language vs. Second Language....143 DISCUSSIONS AND CONCLUSIONS.............147 Discu‘sion‘...’00......0.00.00.00.0000001‘7 conclusions.0.0...0......0.00.00.00.0000156 Recommendation For Further Research.....160 ”PHICESeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeee161 BIBLImWMOOIOO0.0..0...OO0....0......0.0.0.000000000184 CHAPTER 1: INTRODUCTION 1.1 Background Do the cases of child first (L1) and adult second (L2) language learning involve the same or different processes? Do adult L2 learners depend on the same Language Acquisition Device (LAD) as child L1 learners? In the last thirty years or so, these questions, among others, have received a lot of attention in language acquisition research. Some of the earliest answers to these questions are provided by Lenneberg (1967) who claims that adult L2 learners proceed in ways different from child L1 acquisition. lenneberg argues for his Critical Period Hypothesis (CPR) , “according to which there is a fixed span of years during which language learning can take place naturally and effortlessly, and after which it is not possible to be completely successful" (Ellis, 1994:4844) . This critical period is due to certain maturational (plastic) changes in the human brain where the optimum period for language acquisition falls within the first ten years of life (see Singleton (1989) for details about the controversy over the age factor). Further support for CPR comes from researchers (e.g. Schachter, 1988; Bley-Vroman, 1989; and others) who show that there are major differences between the two language acquisition cases. For example, L1 knowledge significantly influences the acquisition process of L2 learners in many areas such as pronunciation. There is, however, some evidence to argue against Lenneberg' s hypothesis . Several studies have shown that L2 learners frequently proceed in ways similar to child L1 acquisition. For instance, 8. P. Corder (1967), Bailey et a1. (1974), Dulay and Burt (1974a,b) and others investigate the developmental trends of both L1 and L2 learning processes. Their research is based on the theoretical assumption that there are underlying basic similarities of all languages in structure and organization. These similarities are referred to as linguistic universals which are innate rather than the results of learning. They investigate the previous question by comparing the utterances used by first and second language learners, focusing on structural differences and similarities between the languages. They claim that first and second language acquisition processes are the same, in spite of 3 differences such as motivation, previous knowledge, and so on. For example, the order in which particular features of the target language are acquired is usually the same in both L1 and L2 acquisition. More recently, the relationship between first and second language acquisition has been explored by linguists working within the domain of a principles and parameters approach to universal Grammar (UG) , as realized in Government and Binding (GB) theory (Chomsky, 1981 a 1986). According to GB, UG is “the system of principles, conditions, and rules that are elements or properties of all human languages ... the essence of human language" (Chomsky, 1975329). Chomsky argues that human beings are innately endowed with universal language- specific knowledge (i.e. UG) which constrains first language acquisition. In other words, human beings utilize certain principles that are biologically determined for language learning. This notion, according to Jensen (1990:4): had been developed in the seventeenth century by Descartes and his followers, and it forms an important part of the 'Grammaire generale et raisonnee' of Lancelot and Arnauld (1660) and Bumboldt's (1836) 'Uber die Verschiedenheit des menschlichen Sprachbaues. ' Modern Linguistics has done much to make specific claims about the content of universal grammar, especially under the aegis of Chomsky's Extended Standard Theory. 4 Being' a. theory' of knowledge (not of behavior), UG concerns the internal structure of the human mind. It attempts to integrate human mind and gramatical principles in the processes of language acquisition. “It givos the child advance knowledge of many abstract and complex properties of language so that these do not have to»be learned solely'on the basis of linguistic input or by means of general learning strategies" (White, 1989:5). UG consists of a certain.number of abstract principles and.parameters‘which.constrain the fonm and functioning of grammars. In Cook's words, I'the speaker knows a set of principles that apply to all languages, and parameters that vary within clearly defined limits from one language to another" (1988:1-2). Two examples of UG principles that are relevant to my study, namely the Subjacency Principle and the Empty Category Principle (ECP) are given here. First, Subjacency is a principle within the Bounding Theory, a sub-component of DC which defines the boundaries for movement. This principle says that movement may not cross more than one bounding node. For example, in English, like modern Standard Arabic (henceforth, Arabic) (Aoun, 1981), the bounding nodes are the Sentence (SIIP) and the noun Phrase (NP). One of the most cited categories constrained by the Subjacency principle is 5 wh-movement in questions (see 5 2.2.1 for further details about Subjacency) . Regarding the ECP, it is a constraint on the distribution of empty categories (such as wh-trace, NP-trace, and so forth), which states that empty elements have to be properly governed. Proper government can be achieved either by a lexical element (such as V(erb), N(oun), P(reposition) etc.) or by a coindexed antecedent. (See 5 2.2.2 for information about the EC?) While UG principles which guide acquisition are universal, there is considerable variation between different languages. The grammar of Arabic, for example, differs in important respects from that of English. In Arabic, sentences exhibit V(erb) 8(ubject) O(bject) word-order as the basic one while basic English word order is SVO. The two languages are similar, then, in ‘principles' in that they both contain 8, V, and 0 but differ ‘parametrically' in their ordering of these elements. A child experiences one order or the other in his/her acquisition and ‘fixes’ the parameter in that direction. An exanple of a parameter that concerns this study is the Hull Subject parameter. This parameter is also called the ‘pro(noun) -drop(ping)" parameter. It is intended to explain, 6 among other things, the property exhibited by pro-drop languages such as Spanish and Italian (as opposed to non pro- drop languages like French and English) of allowing subject pronouns in tensed sentences to be dropped ( Chomsky, 1989; Jaeggli, 1982; Rizzi, 1982) (for more details about this parameter, see 5 2.2.3). universal Grammar has been supported by the I'poverty of the stimulus" argument, which states that the input children receive is insufficient to bring about the linguistic knowledge that they finally attain. Consider the following two sentences: (1)a. Which car did April think that John bought? b.*Which car did April accept Sue's news that John bought? Native speakers who are exposed to such sentences know intuitively that sentences like (6a) are grammatical while sentences like (6b) are ungrammatical. The question is how one can account for this knowledge. It has been assumed that the linguistic data to which language learners are exposed will not provide clear evidence for distinguishing between grammatical and ungranmlatical cases. Consequently, it is supposed that there is built-in knowledge of DC which prevents learners from making UG violations. 7 Assuming that children have innate linguistic properties (UG) that aid them in acquiring a first language, one can ask if adult second language learners still have access to the cognitive system called UG. In other words, is UG still available to L2 learners? And if so, what could be the role of UG in L2 learning? Three different positions on these questions are represented in the literature: A. UG is not available to adult L2 learners (e.g. Clahsen a uuysken, 1986; Clahsen, 1988). L2 learning proceeds only by means of problem-solving and hypothesis-testing. B. UG is fully available to adult L2 learners ( White, 1985b; 1988; Bleybvrommn.et al., 1988; Flynn, 1988); C. UG is not available in its entirety to adult L2 learners, but only those portions of U0 which are instantiated in the learner's first language are available in second language acquisition (Schachter, 1989). 1.2 Plan Of The Study This is an experimental cross-linguistic cross- directional study which aims to examine the acquisition of 8 English by native speakers of Arabic and the acquisition of Arabic by native speakers of English. It will explore the theory of UG and its relationship to the processes of first and second language acquisition cases. In particular, it will be concerned with a principles and parameters approach to UG, as realized in Government and Binding Theory (Chomsky, 1981 and 1986) . It will focus on the question of the accessibility of UG in L2 acquisition. That is, it will investigate whether or not advanced L2 learners of Arabic and advanced L2 learners of English as a foreign language rely on the principles and parameters of UG. In addition, the study will investigate if there is a difference between perception and production tasks in measuring UG Principles and Parameters. moreover, the study will discuss the nature of UG-based acquisition studies and cement on the U0 properties that are considered in such studies. The particular properties of DO which I plan to test are: (1) The Subjacency Principle; (2) The Empty category principle; and, (3) The Null Subject Parameter. These properties of UG are selected for two reasons. First, they are appropriate for testing because they are accessible to surface level analysis (Schachter, 1989) . Second, they have appeared in a number of L2 acquisition studies, making it possible to compare the findings of this study to others. The 9 purpose of comparison is to find out if the obtained results are consistent with the No Access -- i.e. UG is dead -- Hypothesis, or with the Access to DC hypothesis -- i.e., UG is available to L2 learners whether completely or partially. The subjects will be asked first to complete a personal and academic questionnaire. This questionnaire will help reveal possibly significant variables among subjects such as age, age of first exposure to English or Arabic, years of formal instruction in English or Arabic, length of study of English or Arabic in English speaking countries or/and in Arabic speaking countries, and finally, educational level and area of specialization. Then they will be tested using two different types of tasks, namely perception and production tasks. The perception task contains a gramticality judgment test (test one) which includes a syntax test. For this task, subjects have to respond grammatical, ungrammatical or not sure to the sentences, and to try to correct the sentences» they label ungranmiatical. Such corrections help the examiner in deciding whether subjects are rejecting the ungrammatical sentences for the right reasons (i.e. because these sentences violate one of the principles/parameters of UG) . The production task contains written question formation (test two) which also includes a syntax test. For this task, subjects QC' ta 10 will make wh-questions out of sentences, questioning only the underlined phrase. The UG test items test for the presence or absence of the Subjacency Principle, the ECP and the Null Subject Parameter. The syntax test items, however, test for whether the learner has comand of the relevant structures that are being examined by the DO test. If subjects fail to recognize the grammatical sentences (syntax test), this will show that they have not yet learned the structures in question. The tests will be constructed in Arabic and English. They will be applied to four groups of subjects:(1) Native speakers of English (Control Group); (2) Native speakers of Arabic who are advanced learners of English in Saudi Arabia (Target Group); (3) Native speakers of Arabic (Control Group) 3 and (4) Native speakers of English who are advanced learners of Arabic in the USA (Target Group). 1.3 Rationale For The Study The present literature on the UG-accessibility topic (in adult L2 acquisition) provides three different positions ranging from the no access to UG to complete access to DC (see 5 2.4.2 for details). In fact, we can find researchers who ll disagree with their own earlier findings - for example, Clahsen and Nuysken (1986) and Clahsen and Nuysken (1989) . In the 1986 paper, the authors argued that UG is totally unavailable; in the 1989 paper, however, they claimed that only fixed principles but not parameters are available. Although this difference obviously reflects an open-mindedness on the part of the authors, it clearly shows that the UG- accessibility question is still in need of further research before one can reach a satisfactory result (White, 1989) . Such different positions indicate that current research has not provided an unequivocal answer about the role of UG in L2 acquisition. The available literature regarding this question is not yet completely developed. Thus, there is still great necessity for many studies that can contribute to the determination of the role of UG in L2 acquisition. By reaching a decision about the role of DC in L2 acquisition, we will, first, come to grips with the impact of UG on the formation of second language grammars. Second, we can significantly contribute to the development of a principled theory of L2 acquisition. Finally, we can ultimately contribute to the development of a generative theory of language (Flynn, 1991) . 12 In my view, most of the UG-studies (e.g. Clahsen & Nuysken, 1986) conducted so far seem to be inconclusive for several reasons. First, to date only small fragment of the central core principles and parameters have been investigated. Second, researchers (e.g. Schachter, 1989) tend to concentrate mostly on European languages which has left many languages out of the picture, including Arabic. If we examine the current literature, we will find that there are very few, if any, studies that have native speakers of Arabic learning English as subjects for the UG test. Also, there is not any study, at least to my knowledge, that takes native English speakers learning Arabic as subjects for the UG test. It is extremely crucial to focus on many different languages so that we can have data from the whole spectrum which, in turn, will help in deciding the answer of the UG question and in finding common linguistic universals. Third, most of the previous experiments look at the operation of principles and parameters of UG in only one direction. In other words, in their studies, researchers (e.g. Clahsen, 1988) tend to focus only on the acquisition of 1 (say English) by native speakers of I (say French) or 2 (say German). They rarely try to investigate (within the same study) the reverse acquisition; i.e. the acquisition of Y or 13 2 by native speakers of x. It is useful to look at the operation of principles and parameters in both directions (White, 1989) . It is useful in the sense that one can compare learners whose L1, say English, has value A (of the Null Subject Parameter) learning an L2, say Arabic, with value B with learners whose L1 (Arabic) has value B learning an L2 (English) with value A. Different effects may appear, depending which setting is found in the L1 and which in the L2. Raving cross-directional studies is crucial because it is not always easy to find pairs of previous studies each covering one direction. Further, it is not easy to find two one-directional studies that examine the same principles and parameters. Saving the same principles/parameters is important for comparative reasons. Further, one can hardly find two one-directional studies that employ the same method as far as the choice of UG-tests and subjects is concerned (Liceras, 1989) . UG-accessibility studies, when conducted on a cross-linguistic cross-directional basis, will certainly help us find answers to the many remaining questions concerning, for example, the various ways in which UG might intervene in the acquisition process, similarities and differences across languages which can be revealed by the l4 concept of parameters within UG, and relationships between the mother tongue and the target (i.e. second) language. Fourth, methodologically speaking, most studies try to focus on a single test for tapping UG principles and parameters in L2 acquisition. For example, White (1985b), Liceras (1989) and Bley-Vroman (1988) used only a grammaticality judgement task to elicit perceptions of L2 learners. In contrast, Flynn (1988) used an elicited production task. Only few researchers (such as Billes, 1986) used multiple tasks. It will be more reliable and revealing if we adopt a multimethod experiment in order to scrutinize the operation of UG in L2 acquisition. A multimethod experiment will enable us to see if L2 learners act differently from one type of task (perception) to another (production) and aid in our understanding of the process of acquisition. Finally, there are very few, if any, studies that have been conducted in countries where English or Arabic is taught as a foreign language (EFL/AFL) . When conducted in such environments, UG studies will certainly help us find answers to the many remaining questions concerning, for example, the role of instruction when it is the main source of the linguistic input. Moreover, they may shed some light on the 15 differences and/or similarities between the two learning environments (e.g. ESL versus EFL). 1.4 Research Questions This study will attempt to find answers to the following questions: 1. Do adult L2 learners have access to the principles and parameters of UG? 2. If so, do they have full access to UG as do child L1 learners? 3. Or, do they have partial access to UG? 4. Given the results of the two tests, do subjects perform similarly/differently across the DO principles and parameters being tested? 1.5 Testable Hypotheses The testable hypotheses for this study are the following: B1: The background (demographic) independent variables will have no statistically significant effect on the Syntax and UG scores obtained by group one in the two tests . H2: H3: 84: BS: E6: 87: 16 Following Schachter (1989), G1 subjects who pass the syntax test will also pass the UG test, and subjects who fail the syntax test will also fail the UG test. There will be no statistically significant difference in the Syntax and UG scores obtained by group one across the two tests. The background (demographic) independent variables will have no statistically significant effect on the Syntax and UG scores obtained by group two in the tWO teltl. G2 subjects who pass the syntax test will also pass the UG test, and subjects who fail the syntax test will also fail the UG test. There will be no statistically significant difference in the Syntax and UG scores obtained by group two across the two tests. There will be no statistically significant difference in the Syntax and UG scores obtained by groups one and two across the two tests. The background (demographic) independent variable will have no statistically significant effect on the Syntax and UG scores obtained by group three in the two tests. 89: 810: 811: 812: 813: 814: 815: 17 G3 subjects who pass the syntax test will also pass the DO test, and subjects who fail the syntax:test'will also fail the DO test. There will be no statistically significant difference in the Syntax and UG scores obtained by group three cross the two tests. The background (demographic) independent variables will have no statistically significant effect on the Syntax and UG scores obtained by group four in the tWO telts. G4 subjects who pass the syntax test will also pass the UG test, and subjects who fail the syntax test'will also fail the DU test. There will be no statistically significant difference in the Syntax and UG scores obtained by group four across the two tests. There will be no statistically significant difference in the Syntax and UG scores obtained by groups three and four across the two tests. There will be no statistically significant difference in the Syntax and UG scores obtained by emerimental groups two and four across the two tests. 18 In sumary, (1) Hypotheses 1, 4, 8, and 11 assume that the background demographic variables of the subjects will have no biasing effect on test scores. (2) Hypotheses 2, 5, 9, and 12 state that subjects who pass the Syntax test (i.e. show knowledge of syntactic constructions such as ‘that-trace' effects of the ECP test) will also pass the UG test (i.e. demonstrate evidence of UG properties such as the ECP) , and vice versa (i.e. subjects who fail the Syntax test will also fail the UG test. (3) Hypotheses 3, 6-7, 10, 13-15 deal with the comparison of subjects' scores within/between groups on both tests. They assume that subjects will behave similarly across the two tasks. In other words, there will be no significant difference within/between subjects' perceptive and productive performances. In conclusion, as my hypotheses suggest and following Flynn (1987) and White (1989), I believe that there is at least a partial role for UG in the process of L2 acquisition. Further, UG is one component of an acquisition theory, whether of L1 or L2. This component will interact with various others (such as personality, aptitude, etc.), and the failure of L2 learners may be attributable to these other components and not necessarily to the non-accessibility of UG. CHAPTER 2: LITERATURE REVIEW 2.1 Background Chomsky (1986a) says that there are two broad approaches to the study of language, namely the Externalized (E-) approach and the Internalized (I-) approach. The former aims to collect samples of a particular language or languages and then to describe their properties. The latter, however, is concerned with a native speaker' s knowledge about language which is treated as an internal property of the human mind/brain rather than something external. Chomsky (1988) indicates that the recent history of linguistics demonstrates a move from an E-approach to an I- approach, as shown in the switch from the structural to the generative tradition in American linguistics. However, both approaches are relevant to the study of language universals . According to Eckman (1988) , there are two prominent approaches to the study of language universals. The first is the study of typological universals as presented by Greenberg (1966) and has been continued by Comrie (1981) , among others. The second 19 20 is the study of Universal Grammar (UG) as presented by Chomsky (1981a,b). The Greenbergian approach is data-driven where data from a wide range of languages are analyzed in order to discover universal patternings (i.e. features that languages have in common) . In comparison, the Chomskyan approach is theory- driven. It seeks to discover universals by the in-depth analysis of the properties of individual languages so as to identify the abstract principles of UG that constrain the form of any language gramar. Comparing the UG and typological approaches, Eckman (1988) argues that they are similar in the sense that the formulation of parameters in UG is associated with the work of the typological school; such parameters predict the types of variation that can and cannot exist among languages. Moreover, he goes on by saying that “both approaches can be called ‘typological' in this respect, since they both make predictions regarding the types of languages that can and cannot exist with respect to some grammatical property or construct" (p.418) . According to Eckman, however, a basic difference between the two approaches is one of degree; that is, a distinction between them can be based on the degree of theory-dependence of the constructs used to make 21 generalizations . The constructs invoked by the UG approach are considered more theory-dependent than those used by the typological approach. The universal statements of the UG approach are formulated using theory-dependent, abstract concepts such as x-bar theory, parameters, etc. , and they are stated with regard to abstract levels of representation. On the other hand, the typological approach makes universal statements that are indicative generalizations, using relatively theory-independent constructs such as linear order, lexical category, etc., and formulates these principles with respect to surface representation. Since the focus of this research is on the accessibility question of DO in language acquisition, I will now turn to the Chomskyan view of linguistic universals and examine its contributions to the field of language acquisition. 2.2 UG As Realized In GB Theory UG, species-specific, is a theory of the biological endowment of language faculty that characterizes the essential properties of gramars. In other words, UG is assumed to be an innate language faculty that sets the limits within which human languages can vary. It “consists of a highly structured 22 and restrictive system of principles with certain open parameters to be fixed by experience..." (Chomsky, 1981b:38). Such abstract, innate principles apply to all languages. They are principles of the initial state of knowledge which provides the basis for the language learner to develop his/her grammar. Two examples of UG principles that are relevant to this study, namely the Subjacency principle and the Empty Category Principle (ECP) will be discussed in detail in sections (2.2.1) and (2.2.2) respectively. As far as parameters are concerned, they vary in certain restricted ways from one language to another. In Jaeggli and Safir's words, “parameters [are] a set of language specific options expressed as postulates that interact with universal principles to form the grammars of particular languages" (1989:2) . The parameters within a theory of UG provide the mechanism with which we can account for the role of the mother tongue in L2 acquisition. Thus, they account for the differences between groups of various L1 speakers learning a common L2 (Flynn, 1988). An example of a parameter that is relevant to this study, namely the pro-drop parameter will be discussed in detail in section (2.2.3) . According to Chomsky's (-1981a) theory of markedness, those rules that are determined by fixing the parameters of UG 23 on the basis of input from the language being learned constitute the Core grammar. This Core grammar is believed to be unmarked because it is acquired with minimal triggering data. In contrast to a Core granular is the Peripheral grammar which is the set of marked rules that are outside of Core grammar and have to be learned without the help of UG. Chomsky (1981a) says that both theories of UG and markedness work together in the processes of language acquisition. He explains that In our idealized theory of language acquisition, we assume that the child approaches the task equipped with UG and an associated theory of markedness that serves two functions: it imposes a preference structure on the parameters of UG, and it permits the extension of core grammar to a marked periphery. Experience is necessary to fix the parameters of core gramar. In the absence of evidence to the contrary, unmarked options are selected. (1981a:8) The goal of UG is generally twofold. First, it attempts to explain the complexity and richness of grammar. Second, it tries to explain how knowledge of a language is acquired despite limited and often degenerate linguistic experience] input. White (1990:132) has specifically summarized the functions of the theory of UG for L2 acquisition by saying that What the UG perspective offers is the means to identify abstract properties of language, to study certain aspects of L2 competence in depth, and to reach a greater 24 understanding of precisely what formal properties L2 learners unconsciously internalize (or fail to internalize) . Chomsky (1981a) argues that it is impossible for a child to learn his/her first language without having a set of universal linguistic principles that comprise UG. The input a child receives is insufficient to bring about the linguistic knowledge that he/ she finally attains. For instance, speakers of English intuitively know that (2a) is gramtical sentence while (2b) is ungramatical: (2) a. Who did Mary believe would be late? b.*Who did Mary believe that would be late? The conjunction (comlementizer) that must not be present in (2) . Children are not explicitly taught that (2b) is ungrammatical. Thus, this knowledge is not based on the learner's experience of the world, but it comes from the built-in knowledge of UG which prevents him/her from making UG violations. To give another example, Ellis (1994) reported a dative alternation as an illustration of insufficient input a child receives. He said that English allows only two constructions with dative verbs such as “give" as illustrated by the following sentences : 25 (3) a. She gave the pen to John. (Noun Phrase [NP]+ Propositional Phrase [PP]) b. She gave John the pen. (NP 4- NP) c.*She gave to John the pen. (*PP + NP) However, other dative verbs such as “explain" do not permit the NP+NP pattern as shown below: (4) a. She explained the answer to John. (NP 4- PP) b.*She explained John the answer. (*NP 4- NP) c. She explained to John the answer. (PP + NP) Given these examples, how do children find out that (3c) and (4b) are unacceptable? One way to attempt answering this question is by examining two types of evidence, namely positive evidence and negative evidence. Positive evidence, which comes from osmosure to other speakers' speech, is not an adequate answer because it underdetermines the linguistic competence of children; that is to say, children acquire gramar that goes far beyond the actual sentences (input) that they have been exposed to. In contrast, negative evidence, which comes from adult/parent feedback, rarely occurs. Research on L1 acquisition indicates that children do not usually get negative feedback that tells them what is ungrammatical in their utterances (Chomsky, 1965). Even if 26 children are corrected, they usually ignore corrections (White, 1989a). Having said that there is an insufficient positive evidence and no negative evidence, the answer to the previous question is that children must rely on innate knowledge which “in some sense compensate for the lack of availability of negative input" (White, 1989a:14). Arbib and Hill (1988:96) say that “many linguists have assumed that the lack of negative feedback to children creates a logical problem for language acquisition that can be solved only by reference to innate constraints that prevent children from ever formulating overly general grammars." In other words, L1 learners achieve competence which goes beyond the linguistic input received. So UG is motivated due to the mismatch between the limited linguistic input to which a child is exposed and the complexity of then grammar which he/she achieves. The nature of this gramr is inseparable from the problem of how it is acquired, and this problem is approached by postulating that UG mediates language acquisition. 27 2.2.1 The Subjacency Principle The subjacency principle is a principle within the Bounding theory, a sub-component of UG which defines the boundaries for movement. Thus, it defines the restrictions that govern how far an element can be moved from deep to surface structures. This principle says that movement may not cross more than one bounding node. For example, in English, like Arabic (Aoun, 1981), the bounding nodes are the Sentence (in GB terminology, SnIP, i.e. Inflectional Phrase ) and the Noun Phrase (NP). One of the most cited categories constrained by the Subjacency principle is wh-movement in questions. Consider the following English (5) and Arabic (6) examples where the type of extraction is a wh-word and the domain of extraction is a noun phrase complement: (5) a. *Whom does the president have evidence that the Senator trusted? b. *[CPl Whom: does [IPl the president have [NP evidence [CP2 t': that [IP2 the Senator trusted trill”? (6) a. *man rafadha al-qadhi al-dali:la ?anna al- walada dharaba? 28 b. *[CPl man: [IP1 rafadha al-qadhi [NPal-dali:1 Whom refused the-judge the-evidence [CP2 t'I bi-?anna [al-walada dharaba tr]]]]]? that the boy hit “*Whom did the judge refuse the evidence that the boy hit?" The sentences in (5) and (6) violate the Subjacency condition. The wh-phrases 'whom' in (5) and ‘man' ‘in (6) have crossed two bounding nodes, namely NP and IP1. They have moved stepwise: first they move to the lowest [Spec (ifier) , C (omplement)P(hrase)2]l as step one, then they move to the matrix (Spec, CP1] as step two. Such movement is called successive cyclic: it applies in successive cycles, from bottom to top. Since each clause (CP) defines a syntactic domain in which wh-movement can apply, step one is legitimate (because only one bounding node (IP2) is crossed) and step two is illegitimate (because two bounding nodes (NP and 1P1) are crossed). With regard to Arabic, I assume that the Subject is generated VIP-internally. Thus, VSO word order is derived by Verb movement to I, and SVO order is further derived by the Subject movement to Spec, IP (Aoun et al., 1994). 29 It is noteworthy that there are other types of extraction, namely relative pronoun, and topic, and other domains of extraction, namely sentential subject, relative clause, and embedded question. Regarding parameters, the principle of Subj acency allows some parametric variation across languages with respect to the bounding nodes that they adopt. White (1990: 125) reports that “languages differ as to the bounding status of 8; Italian and French, for example, [unlike English] have NP and S' as bounding nodes, but not S (Chomsky, 1981b; Rizzi, 1982; Sportiche, 1981).“ Therefore, the variation is a choice between IP and CP(-S'). (For further examples about Subjacency, see Cook, 1988.) 2.2.2 The Empty Category Principle (ECP) The ECP is a constraint on the distribution of empty categories such as wh-trace, NP-trace, and so forth. It states that empty elements have to be properly governed. Proper government can be achieved either by a lexical element such as V(erb), N(oun), P(reposition), etc., or by a coindexed antecedent. Consider the following sentences: 30 (7) a. *Who: do you think [CP t'I that [IP tr leftll? b. Who: do you think [CP t’I [IP t1 left”? These two sentences show that a subject can only be extracted from clauses without an overt complementizer (that). This phenomenon was described in terms of the that-trace filter. This filter has been reinterpreted in terms of the government requirement for traces. In (7a), the complementizer intervenes between the trace (tr) in the subject position and the intermediate trace (t'r) in [Spec, CP] . Thus, the intermediate trace cannot antecedent-govern the subject trace. Hence, the subject-trace violates the ECP and (7a) is ungrammatical. In contrast to a Subjacency violation which is considered a movement violation, this ECP violation is said to be a government violation. In (7b), however, there is no overt complementizer to block government from the intermediate trace. The subject trace is properly governed. (I will talk about the ECP in Arabic when I discuss the properties of the Null Subject Parameter later.) Regarding parameters, the ECP is parameterized with regard to the elements which constitute proper governors in different languages rather than with regard to whether it holds in some languages but not in others. White (1989) 31 reports that in English but not in French, preposition stranding is possible because in English the preposition constitutes a proper-governor but not in French. (For more examples on the ECP, see Cook, 1988.) 2.2.3 The Null Subject Parameter (NSP) The Null Subject parameter (also called the pro(noun)- drop(ping) parameter) is intended to eamlain, among other things, the property exhibited by pro-drop languages such as Spanish, Italian, and Arabic (as opposed to non-pro-drop languages like French and English) of allowing subject pronouns in tensed sentences to be dropped (Chomsky, 1989; Jaeggli, 1982; Rizzi, 1982). Other properties that cluster together with the property of phonologically null subject pronouns and that have been associated with this parameter are the free inversion of subject and verb in declarative sentences and the extraction of a subject out of a that - clause under wh-movement (i.e. that-trace) . Liceras (1989) argues that these three properties are related in that acquiring the null subject is a condition for acquiring the other two properties. Further, she says that there is an implicational hierarchy of difficulty relating the three properties of the pro-drop parameter (i.e. pro- drop>inversion>that-trace) so that if that-trace has been acquired, free inversion and null subject pronoun have been acquired too. For example, the gramar of English is set negatively for this parameter (i.e. [— pro-drop]) while the grammar of Italian is set positively for this parameter (i.e. [4- pro-drop] ) : (8) English Italian a.*Smokes. Fuma. b.*Smokes Mary. Fuma Mario. c.*Who did you say Chi hai detto that smokes? che fuma? Unlike English, Arabic -like Italian - (Lakshmanan, 1986; Renstowicz, 1989) is a pro-drop language. However, it differs from Italian with regard to the properties associated with the Null Subject parameter. Arabic allows subject pronouns to be omitted as (9a) shows: (9) a. Sa-yadhabu. Will go “He will go." As for the property of free subject inversion, Arabic (Owens, 1988) has the order of VSO as the most frequent and basic word order. An example (9b) is in order: 33 (9) b. Daraba zayd-un 9aliyy-an. hit Zayd-Nom Ali-Acc “Zayd hit Ali." It should be pointed out that Arabic has two different types of sentence structure, namely the verbal sentence and the nominal sentence. The verbal sentence is one that begins with a verb and has the V80 word order. In contrast, the nominal sentence is one that starts with a noun and has, in certain constructions such as a nominal sentence that contains a verbal sentence, the SVO word order. In other constructions, the nominal sentence has no overt verb in its surface structure. Regarding the extraction of an embedded subject (leaving a trace) out of a clause containing a complementizer (that- trace effects), Arabic allows ‘that-trace' effects only in certain contexts but not in others. This asymetry is due to the fact that Arabic (Aoun, 1981) is more complex (than Italian, for example) in that extraction of a subject is dependent on the main verb. In Arabic, there are two types of complementizers, namely Nominal ?anna that and Verbal ?an that. Nominal ?anna can only cooccur with “believe-type" verbs like yadunnu think while Verbal ?an can only cooccur with “want-type" verbs such as yuri:du want. Also, Nminal 34 ?anna must be followed by a noun or a cliticized pronoun while Verbal ?an must be followed by a verb. In Aoun's (1981:639) own words, “?anna is a case- assigning element and ?an a mood-assigning element: ?anna and ?an assign accusative and subjunctive, respectively. ... The subjunctive feature will be paired with the verb [while the accusative feature] will be paired with a following lexical NP if there is one or spelled out as a clitic if there is none." Noun phrases (NPs) that occur, for examle, in sentential complements of “believe-type" verbs cannot be extracted if they occur in the subject position (i.e., immediately following the complementizer ?anna ). However, if NPs occur in the object position, extraction is possible. An example of impossible extraction is (10c): (10) c.*man1 dhanna muhamad-un[CP2t'I Who think Mohammad-Nam ?annaIIP2 t1 ishtara al-siyarataljl? that . bought the-car “*Who did Muhammad think that _ bought the car?" The presence of the complementizer ?anna prevents the intermediate trace (t'i) from counting as the antecedent- governor of the lower trace (ti) in subject position. Hence, the ECP is violated. 35 In comparison, NPs that occur, for example, in sentential complements of “want-type" verbs can be freely extracted whether they occur in the subject or object positions. (For further examples about the NSP, see White, 1985b; 1989a.) 2.3 UG-based Studies In L1 Acquisition The theory of UG is closely tied to a theory of L1 acquisition. Chomsky argues that children come to the acquisition task with innate knowledge which aids them to recognize and figure out linguistic structures of their language on their own. In White's words, “UG provides a kind of blueprint as to what the grammar will be like, but details can only be filled in by input from the language being learned" (1989:16) . Thus, UG help explain the speed and ease with which language is learned. Goodluck (1991) argues that the child is biologically equipped with knowledge of UG. This fact can be supported by examining language development which shows many of the properties of biologically given behaviors. Two examples of such properties are in order. First, there is an “orderly progression of stages"; that is, children develop their linguistic abilities by following distinct stages. For 36 instance, the ‘cooing' stage (when the child begins producing vocalizations with a vowel-like quality) emerges before the ‘babbling' stage (the process of combining consonant-like and vowel-like sounds); the ‘one word' stage precedes the ‘two word' stage, and certain grammatical morphemes exhibit similar order of acquisition across different children learning the same language (Brown, 1973). The second.property'concerns the fact that linguistic development is to a degree “independent of external stimuli". In other words, a speech stimulus is either unnecessary or only minimally needed because children possess a biologically timed system.that is not dependent on ewesure to speech. For example, some researchers (e.g. Goodluck, 1991) support this fact by reporting that even deaf children go through the ‘babbling' stage. Having said that, I should point out that a child has to be exposed to language in order to achieve normal linguistic development (White, 1989a). In the literature of language acquisition, there are two types of studies, namely observational studies (based on collecting speech data in natural settings) and experimental studies (based on eliciting linguistic structures from language learners). 37 In an observational research, Brown (1973) longitudinally studied first language development on three unacquainted children (Adam.and Sarah, both 27 months old; Eve, 18 months old). The aim of Brown's study was to investigate the development of L1 (English) grammar acquisition after the emergence of syntax in the child's two-word (or more) utterances. Brown discovered that the three children generally appear to develop English grammatical morphemes in a roughly similar order of acquisition. Such.morphemes were present progressive [-ing], plural [-(e)s], irregular past [went], possessive [-'s], uncontracted copula [is], articles [the], regular past [-ed], regular third person singular [-s], and contracted auxiliary ['s]. Brown found that the same results when be compared his study with.other studies dealing with the acquisition of morphemes by children in the same age range. As a support of Brown's (1973) results, De‘Villiers and De Villiers (1973) reported a similar order of morpheme acquisition.across all subjects. They reached this conclusion by conducting a cross-sectional study which involved twenty- one children (ages 16-40 months old) learning English as their first language. 38 Generally speaking, the major argument of the above two studies is that the order of the acquisition of morphemes was not influenced by the frequency of morpheme appearance in the children's environment. Rather, child language acquisition is guided by Universal Grammar. Moreover, child language acquisition cannot be adequately emlained by imitation and the like. In comparison to the above observational studies, Carol Chomsky (1969) conducted an experimental study that was based on eliciting certain gramatical structures (easy vs. eager to see, tell vs. promise) from English-speaking children. She tested her subjects' ability to distinguish between sentences such as: (11) a. John is eager to please. b. John is easy to please. Although these two sentences (that have missing object constructions) have similar surface structures, they have different underlying (deep) structures. In (11a) ‘John' is said to please other people. In (11b), however, ‘John' is said to be pleased by other people. It is claimed that children -can discriminate between such structures that they could not have acquired from their surrounding environment; children rely on UG in figuring out such structures. Further, 39 C. Chomsky explained that children learn to use linguistic processes in the simpler cases first (e.g. 11a), and then proceeds to employ them in the more complex cases (e.g. 11b) . More recently, Hyams (1986) conducted an analytic study focusing on the pro-drop parameter. She analyzed published examples of children's language. Some examples are in order: (12) a. Want look a man. b. Read bear book. c. He ride bike. d. I want hold it. Hyams discovered that English-speaking children produced subjectless sentences (12a,b) at the earlier stages. At the same time, they could also produce sentences with subjects (12c,d). Hyams argued that children start with a positive value of the pro-drop parameter that allows null-subject sentences regardless of their L1 languages (e.g. English or Arabic). Then English-speaking children (like German-speaking children) go on to realize that their language does not allow subjectless sentences (i.e. a non-pro-drop language), setting the parameter to a negative value. Hyams explained that English-speaking children, being exposed to positive evidence, reach such realization when they first acquire expletive 40 subjects (i.e. the use of ‘it' and ‘there' as subjects) such as “It's time to go." Clahsen (1988) argued that child L1 acquisition primarily follows from principles of UG. To this end, the author studied the emergence of agreement markings (e.g. subject-verb agreement) in child L1 learners of German. He provided three kinds of evidence in favor of the parameter view of L1 acquisition. They are the following: 1. Possible grammar: In all stages of the L1 acquisition process of German syntax, the gramar of the child is fully definable within the theory of gramar. 2. Developmental interactions: given the INFLlection] /V parameter, it is expected that in child L1 development, the acquisition of correct verb placement and the attainment of the agreement system takes place in the same developmental phase. 3. Triggering experience: The availability of complementizers [such as dab ‘that'] and certain items of the agreement paradigm allow the child to reset the values chosen for the INFL/V and the CCMP/INFL parameter. (Clahsen,1988:65) It is noteworthy that Clahsen's claim that children have access to principles of UG was originally made by Clahsen and Muysken (1986) who studied the acquisition of German word order in L1 acquisition. Thornton and Crain (1994) studied 21 three-and-four-year- old childrens' emerging knowledge of successive cyclic movement (see section 2.2.1 for an explanation of such 41 movement). This study contained four experiments which investigated children's productions and comprehension of wh- questions. It showed that young children have the mechanisms of long-distance wh-movement available to them. Most children produced well-formed questions such as “What do you think Cookie Monster eats?" In conclusion, pioneer L1 acquisition studies suggest that children are biologically equipped with knowledge of Universal Grammar. Such contention, however, does not entail that all principles of UG are available to the child from the outset. In fact, there are three broad approaches relating to the question whether some UG principles emerge only after a period in which they are not present in the child's grammar. These three positions are the following: 1. The Strong Continuity approach. All principles of UG are present in the child's grammar at the outset of syntactic development. 2. The Weak Continuity approach. All principles of UG such as the X'-theory are available to the child from the beginning of language development, but the actual syntactic tree has to be posited based on the input. 3 . The Maturational approach. The development of gramar is taken to result from maturation processes. That is, 42 principles of grammar are biologically programed to emerge only after a certain period of development. Different researchers have supported one or the other of the above three positions. However, according to Goodluck (1991), the Weak Continuity position is the most popular one in L1 literature since the 1970s. Next, literature on child and adult L2 acquisition will be reviewed in detail. 2.4 UG-based Studies In L2 acquisition Does the theory of UG have relevance for the field of L2 acquisition as it does for the field of L1 acquisition? L1 children progress from an initial state of knowing nothing except UG to a final state of knowing everything about their L1 language. In contrast, L2 learners, whether children or adults, already know a first language. Children learning an L2 are not in full possession of competence in their L1 while adults possess complete competence in their L1. In the literature of language acquisition (e.g. Cook, 1988), there are five different methods of learning, other than UG. PrOPOsed to explain language acquisition. These are: (1) Imitation (e.g. repeating sentences after a competent speaker); (2) Grammatical explanation (e.g. explaining 43 ‘reflexives' such as ‘himself ' in pedagogical grammar books); (3) Correction and approval (e.g. correcting grammatical mistakes in the classroom); (4) Social interaction; and (5) Dependence on other faculties. These methods have been rejected for L2 acquisition as well as L1 acquisition (Bley- Vroman et al., 1988; Cook, 1988) . Many studies have revealed that L2 learners know things that they could not have acquired from the environment around them. This leads us to the question of UG-accessibility in L2 acquisition. Before trying to answer this question in terms of the present research, I will review some important previous literature dealing with child L2 acquisition (5 2.4.1) and adult L2 acquisition (5 2.4.2) . 2.4.1 Child L2 Acquisition Studies O'Grady et al. (1989) reported on both cross-sectional and longitudinal morpheme acquisition studies of children learning English as a second language. Those studies found that L2 children, regardless of their first language, showed a similar order of grammatical morpheme acquisition. Dulay and Burt (1974a) conducted a morpheme-acquisitional study on 115 children (6-8 years old) learning English as an 44 L2. These children came from different (Chinese and Spanish) linguistic backgrounds. They were tested for the sequence of acquisition of eleven grammatical morpheme. Then, the results of the L2 sequence were compared with the corresponding order in L1 acquisition of English. The authors found that the sequence of acquisition, among the subjects, was the same regardless of their L1. Further, they said that their findings revealed a different sequence from that of L1 acquisition. This difference can be explained by the fact that L2 learners “need not struggle with the same kinds of semantic notions already acquired in earlier childhood" (Dulay and Burt, l974a:52). Furthermore, Dulay and Burt (1974a,b) spelled out their above position by advancing a major theory for L2 acquisition, namely the theory of Creative Construction (CC). According to CC, L1 and L2 acquisition cases follow from the same set of innate principles. L2 acquisition patterns are determined by the language structure of an L2 and the creative construction powers of the L2 learner. The fact that the L2 learner already has knowledge of a language is considered irrelevant. Heckler (1975) conducted a study that dealt with the acquisition of English verb morphology by Arabic, Japanese, and Spanish speakers. He based his research on Berko's (1958) 45 experimental L1 study on English morphology which tested production rather than comprehension. He found that L2 learners tend to learn English morphemes in a non-randomized order. For example, the past allomorphs /—t/ and /-d/ were mastered before the allomorph /-id/. Heckler's results revealed similar order of acquisition in first and second language learning. Rrashen (1982) discussed the results of 21 English- morpheme-acquisition studies that had L1 and L2 learners as subjects. He found that there was a similar order of acquisition for L2 learners, regardless of differences in age and L1 background; such an order was referred to as the “natural order" of acquisition. These results agreed with Dulay and Burt's (1974a,b) findings. Further, he said that the order of acquisition for a second language was not identical to that for a first language. Sabra (1987) conducted an observational study that dealt with the process of L2 acquisition of English syntax by four Arabic children (ages 6-7) in kindergarten and first-grade classrooms. They were observed for five weeks in the two classroom; then, they were observed for eight weeks out of school. Sabra considered the subjects' developmental stages (e.g. negative and interrogative structures), learning process 46 in comparison to that of other L1 and L2 learners of English, L1 interference, and communicative strategies used to compensate for language deficiencies. In general, the author found that children demonstrated the following: (1) significant development in the acquisition of English syntax; (2) similar stages of acquisition to those found in L1 and L2 studies; (3) no L1 interference; and finally, (4) different communicative strategies such as repetition, gestures, noises and simplification in order to compensate for their language deficiencies. To sum up, the above studies and others have revealed that (1) there is a similar order of acquisition for L2 learners; (2) In spite of differences such as motivation, previous knowledge, and so on, first and second language acquisition processes are the same in terms of studies of negation and interrogation development. For example, the developmental stages in which particular features of the target language are acquired are usually the same in both L1 and L2 acquisition. This gives evidence that child L2 learners, like child L1 learners, have access to UG which facilitates language acquisition; (3) According to morpheme studies, however, the order of acquisition for an L2 is not identical to that for an L1 (e.g. Dulay and Burt, 1974a). 47 Larsen-Freeman (1976) explains this difference by saying that L1 order of acquisition is conditioned by cognitive development, while L2 acquisition order correlated with frequency in input . 2.4.2 Adult L2 Acquisition Studies Assuming that child L1 and L2 learners have innate linguistic principles (UG) that help them in acquiring a first or second language, one can ask if adult L2 learners still have access to UG. Put differently, is UG still accessible to adult L2 learners? And if so, what could be the role of UG in L2 learning? Three different positions on these questions are represented in the literature: ( 1) UG is not accessible to adult L2 learners (e.g. Clahsen and Muysken, 1986; Clahsen, 1988, Meisel, 1991) (see 3 2.4.2.1 for literature review); (2) UG is fully accessible to adult L2 learners (e.g. Bley-Vroman et al., 1988; Flynn, 1988; White, 1988) (see 5 2.4.2.2 for review of literature); and, (3) UG is not accessible in its entirety to adult L2 learners, but only those portions of UG which are instantiated in the learner's L1 are accessible in L2 acquisition (e.g. Clahsen and Muysken, 1989; Schachter, 1989; Clahsen, 1990) (see 5 2.4.2.3 for review of studies). 48 2.4.2.1 The Non-Accessibility Of UG Studies Lenneberg (1967) was among the first to tackle the question of the accessibility of UG in L2 acquisition. He argued for his Critical Period Hypothesis (CPH) which states that L2 learners are cut off from UG after a certain age. This critical period is due to certain maturational (plastic) changes in the human brain where the optimum period for language acquisition falls within the first ten years of life (or around the age of puberty). In other words, L2 learners have no access to the principles and the parameters of UG after they reach a specific age. In support of Lenneberg's CPH, Johnson and Newport (1989) argued that a critical period for language acquisition extends its effects to second language acquisition. To test their argument, they conducted a study on 46 Chinese and Korean learners of English as a second language. These subjects ranged in age of arrival in the United States from ages 3 to 39. They were tested on their knowledge of English syntax and morphology, using a grammaticality judgment task of spoken English sentences of varying types. Johnson and Newport found that. there is a gradual decline in language acquisition skills over the period of on-going maturational growth. Subjects' 49 performance gradually declined from about age seven on until adulthood. Arguing against the CPH, Cook (1985) says that Lenneberg's hypothesis is not concerned with acquisition processes but with physical or cognitive maturational processes. Further, she wonders how one can retrieve the L1 parameters to fix their values differently in the L2 if he/ she has no access to UG. Moreover, Flynn and Manuel (1991) questioned the validity of Lenneberg's CPH by reviewing a large number of relevant studies . They argued that the logical problem of L1 and L2 acquisition is the same. They indicated that there are both post-puberty L2 learners who achieve full command of a second language as well as pre- puberty learners who fail to attain native-like competence. Clahsen and Muysken (1986) compared the acquisitional sequences of German word order in L1 and L2 acquisition, especially the question of whether learners of German treat the verb phrase as head-initial or head-final. They found major differences between the two (L1 and L2) cases of iacquisition. For example, unlike L1 learners, adult L2 learners have consistent difficulties in discovering the underlying SOV word order of German. They argued that adult L2 learners, unlike L1 learners, resort to general problem- 50 solving (learning) strategies to acquire the second language. Thus, UG is not available to adult L2 learners. In a reply to the above study, DuPlessis et al. (1987) reanalyzed Clahsen and Muysken's data in terms of three parameters: head position, adjunction (of ADV/PP to IP), and proper government (e.g. COMP as a proper governor of INFL) . They found that the rules that adult L2 learners employ, the stages that they go through, and the errors that they make are completely consistent with a UG that incorporates such parameters. ’To further support their position, DuPlessis et al. discussed their own data on the acquisition of German and Afrikaans which revealed that adult L2 learners still have access to UG. For example, adult L2 learners are said to restructure their grammar, when they find out that German is not an SVO language, but an SOV language. In addition, White (1989a) argued that Clahsen and Muysken failed to distinguish between L2 performance and the acquisition of L2 competence. In White's words, “German is a particularly problematic case, allowing a number of surface word orders in declaratives, including: SVO, SV"OV" , AVSO, and AV"SOV", in main clauses, and SOV and ”SOV V in subordinate clauses" (p. 104) . She concluded that since both SVO and SOV are possible word orders, one cannot claim that DO 51 is not accessible in L2 acquisition simply because L2 learners adopt an incorrect word order for German. In support of his earlier position, Clahsen (1988) argued that adult L2 acquisition follows from general learning strategies not observed in child L1 acquisition in the construction of a target language. He reached his conclusion by comparing the emergence of agreement markings in adult L2 and child L1 learners of German. He argued that adult L2 learners have great difficulty in acquiring the intricacies of German word order. He concluded by saying that there were essential differences between children and adults with regard to the process of acquisition. Flynn and O'Neil (1988) argued against Clahsen's (1988) conclusion by saying that UG might in fact be involved in the process of adult L2 acquisition. They explained that the essential language faculty does not change significantly over time. Moreover, the argument of pattern matching, advanced by Clahsen, (i.e. patterns of acquisition for the L2 should exactly match those for the L1 acquisition in order for one to argue for UG) by itself does not necessarily and completely rule out UG-accessibility in L2 acquisition. For example, Flynn and O'Neil said that 52 in order to develop an explanatory theory of L1 acquisition, we must account for both the learner's Ll knowledge and principles independent of this knowledge. Clahsen ... assumes a “clean slate” approach to L2 learning. .. . The role of L1 knowledge is not accounted for in Clahsen's framework. Perhaps, if we assumed that L2 learners do not start with a clean slate ., then by reformulating questions in terms of interactions of the L1 and the L2, we might be able to explain the Clahsen data without invoking general indicative learning principles at the level he investigated. (p. 10) Finally, Flynn and O’Neil say one can support the Access- to-UG hypothesis if we try to reformulate assumptions about the structure of German as argued by DuPlessis et al. More recently, Meisel (1991) argued that UG is not available any more in L2 acquisition. In support of this position, he provided three types of evidence: 1. L2 typically involves “learning” in the traditional sense (trial and error, gradual approximation to the norm, etc.) Whereas L1 development can be described adequately as triggering of implicitly available knowledge (UG), resulting in “instantaneous acquisition". 2. The observed developmental patterns in L2 can be accounted for without referring to US principles. 3. There are crucial differences between Ll development and L2 acquisition. which cannot. be explained. as resulting from parameter resetting (P. 272-3). 53 In a response to Meisel (1991), Schwartz (1991) explained that Meisel's argument (that 06 is not available to L2 learners) is weak on three grounds. First, UG-based explanation of L2 acquisition is preferred over any other approach (e.g. general learning mechanisms) because of both the higher predictive power and the explicitness of HG principles. Second, it is not sufficient to prefer a different approach other than UG simply because it accounts for the same data as UG-based approach. Third, it is theoretically simpler to rely only on 00 to explain certain facts about L2 acquisition processes than one that must resort to a combination of mechanisms. In conclusion, proponents of the no-access to no (in adult L2 acquisition) argue that L1 acquisition requires a linguistic theory (i.e. UG) whereas L2 acquisition requires a cognitive theory (e.g. the Multidimensional Model (mm). The ma, proposed by Meisel, Clahsen, and Pienemann (1981) as a predictive framework, focuses on the relationship between implicit knowledge and output by indicating the strategies which have to be mastered in order to produce different structures. It attempts tp provide an mlanation of observed development in learner-language. (See Ellis (1994:382-88) for full account of the m.) 54 Arguing against it, Ellis (1994) explained that the am only gives an explanation of acquisition with regard to learner production. There is nothing in this model that shows how learners come to comprehend grammatical structures, and how comprehension and production interact. 2.4.2.2 The Full-Accessibility Of UG Studies Dulay and Burt (1974), Bailey, Madden, and Krashen (1974), and others (e.g. Bley-Vroman et al., 1988; Flynn, 1988; White, 1989) found that L2 acquisition is, in essential respects, similar to L1 acquisition. That is, regardless of their age, language learners proceed in similar ways. These studies demonstrated that, first, developmental L2 errors are similar to those committed by L1 learners; and, second, the acquisition order of certain morphemes is the same in both L1 and L2 acquisition cases. Third, both first and second language acquisition involves the ‘logical problem' (also called ‘Plato's problem') which states that the input language learners are exposed to is, as Cook (1988), White (1989), and rlynn and Manuel (1991) argued, (a) underdetermined (i.e. linguistic competence learners attain goes far beyond the actual sentences they are exposed to); (b) degenerate (i.e. 55 learners are exposed to an input that is not always perfect); and (c) finite (i.e. lack of negative evidence). In other words, the poverty of the stimulus argument applies equally to L2 acquisition. Thus, there is no significant difference between children and adults as far as language acquisition is concerned. Eanania and Gradman (1977) conducted a case study on a Saudi adult female in Bloomington, Indiana. They investigated the subject's acquisition of mlish sentence structure, mean length of utterance, and order of acquisition over a period of eighteen months. They found that the subject progressed in a manner similar to the process of L1 acquisition. Further, the study demonstrated that the subject did not seem to be influenced by her first language (Arabic) in the production of English utterances and that she dealt with English creatively. Flynn (1984; 1987; 1988) argues that adult L2 learners have access to the same language faculty as L1 learners by showing, in an experimental study, that the Principal Branching Direction (PBD) is operative in adult L2 acquisition. The PD!) is also called the Head-position parameter which has two values, namely head-initial (i.e. the head such as a noun or a verb which occurs before a complement as in English and Arabic, which are described as right- 56 branching languages) and head-final (i.e. the complement occurs before the head as in Japanese, a left-branching language). Flynn's subjects, consisting of 51 Spanish native speakers and 53 Japanese native speakers coming from different levels of proficiency, were tested with a standardized elicited imitation task and an act-out comprehension task that dealt with English complex sentences. Flynn wanted to see if L2 acquisition of anaphora is constrained by branching direction. The results indicated that Spanish subjects were more successful than Japanese in performing the two tasks. This is due to the fact that Spanish, unlike Japanese, is head-initial like English which gave Spanish speakers an advantage over Japanese speakers . Generally speaking, Flynn concluded that US is still available to adult L2 learners and learning is facilitated if the L1 and L2 parameter settings are the same. However, if the L1 and L2 have different parameter settings, the pattern of acquisition will resemble the early stages of L1 acquisition. In other words, there is no need to reassign a value in the Native Language (NL) to match the Target Language (TL) if a parameter-value in the NL and the TL match. However, a process of parameter-value reassignment is necessary if the NL and the TL do not match. 57 Felix (1988) conducted an experiment designed to find out whether or not adult L2 learners have access to principles of UG. He tested 48 German college students who had learned English as an L2 for their intuitions about grammatical contrasts in constructions involving different principles hypothesized to be part of UG. The test sentences reflected contrasts which did not have any parallels in German, for example sentences with ‘that-t' effects which are explained in terms of the ECP violations: (13) a. Who does John believe that he saw? b.*Who does John believe that saw him? Such sentences and others representing different kinds of structures formed the questionnaire that was given to the subject. The subjects were asked to mark each sentence as either ‘grammatical' or ‘ungrammatical' . The same questionnaire was also given to a control group of 23 native speakers of English. Results of Felix's study indicated that L2 learners were able to make correct gramatical judgments on these sentences . Felix argued that these results strongly indicated that adult L2 learners are able to achieve grammatical knowledge which can neither be learned from available speech data (i.e. positive evidence) nor from what 58 is explicitly taught in the foreign language classroom. Thus, adult L2 learners do have access to principles of UG. Further support of Felix' results comes from White (1988) who examined the acquisition of Subjacency and the ECP by native speakers of French learning English. White's subjects consisted of 43 adults and 23 adolescents. A number of different tests (cloze test, a multiple-choice judgment task, a paced judgment task, and a comprehension task) were devised to test for the two principles of UG. White hypothesized that 0G is available to the L2 learner. She argued that the errors that the L2 learners will make in the TL will not constitute violations of UG-principles; In other words, they will not constitute iwossible errors. White's hypothesis was borne out by the results of her study. Clahsen and Muysken (1989) criticized the above two studies (Felix, 1988 and White, 1988) by stating that the results of these two studies ‘can only demonstrate that L2 learners can use their 0G knowledge to some extent in the evaluation of target sentences, but most of the evidence is consistent with the notion that is only UG knowledge which is instantiated in the L1 of the learners" (p. 26). They concluded by saying that the main difference between child L1 acquisition and adult L2 acquisition is that children have 59 direct access to UG, while adults only have access to those 0G principles and parameters which are present in their first language. Bley-Vroman, Felix and Ioup (1988) addressed the debate about whether adult L2 learners have access to the principles/parameters of UG. They investigated two principles of UG, namely Subjacency and ECP, by testing 92 advanced Korean speakers learning English and a control group of 34 native speakers of English. The test was a grammaticality judgment task, with 32 randomized wh-movement sentences, approximately half of which were ungrammatical sentences which violated Subjacency or the ECP. The subjects were required to demonstrate whether a sentence was ‘possible' or ‘impossible' in English, or whether they were ‘unsure' . The results revealed that the subjects achieved scores (at 759‘) considerably higher than chance suggesting that 0G is still available to adult L2 learners. The study of Bley-Vroman et al. is not without problems. There were not enough sentences for each syntactic phenomena (such as relative clauses, noun complements, etc. which form part of Subjacency) . In fact, we can find only one test sentence for some of the structural categories. More that one 60 test sentence is needed in each category (e.g. relative clauses) so that we can have reliable results. Liceras (1989) investigated the setting of the pro-drop parameter by 30 English and 32 French speakers learning Spanish as a foreign language. English and French are non- pro-drop languages while Spanish is a pro-drop language. Thus, a “resetting" (i.e. from -pro-drop to +pro-drop) of L1 parameter was forced, which necessarily influenced a cluster of linguistic properties (i.e. null subjects, verb-subject inversion, and that-trace violations). Liceras suggested that there may be an implicational hierarchy among these properties such that if ‘that-trace' has been acquired, subject-verb inversion and null subject will have been acquired at an earlier stage. She also hypothesized that Spanish represents the unmarked option of the parameter and that L2 learners of Spanish may begin with the unmarked setting rather than their (marked) L1 setting. Liceras' subjects were asked to respond to a written gramaticality judgment task consisting of 17 items written in Spanish. Each item contained one or two instances of the properties of the pro-drop parameter. The results indicated that most L2 learners do not start with the L1 setting as far as the null subject is concerned. In other words, there is no 61 L1 transfer into the L2. In addition, the results suggest that acquiring the null subject (and verb inflection) comes first in order before acquiring inversion and ‘that-t' effects. The study showed that adult L2 learners have access to all possible parameter settings, not just the ones instantiated in their native languages. Thomas (1991) conducted an experiment as an attempt to answer the question of whether 0G is available to adult L2 learners. She investigated the interpretation of English reflexive pronouns by 132 native speakers of Japanese and of Spanish, and the interpretation of the Japanese reflexive zibun by 41 native speakers of English and of Chinese. The subjects were given two experimental tasks: One contained elicited imitation of sentences in L2 with pronouns and anaphors in varied syntactic structures, whereas the second task was a multiple-choice comprehension test of the interpretation of pronouns and anaphors in L2 . Consider the following example: (14) Mary heard that Sue told the doctor about herself. The subjects were asked whether the reflexive ‘herself ' refers to ‘Mary', ‘Sue', some other person not mentioned in the sentence, or any combination of ‘Mary', ‘Sue', or some other person . 62 Thomas' results supported the hypothesis that UG is available to adult L2 learners and disconfirmed the claim that learners can activate in L2 only that form of HG instantiated in L1. Having said that, I can see at least one problem with Thomas' study; namely some of her subjects are not true adult learners. That is, some of the subjects began to study the L2 at ages as young as 4 years old. White, Travis, and Maclachlan (1992) investigated the question of whether adult L2 learners whose L1 is Malagasy (a VSO language in the Western Malayo-Polynesian branch of the Austronesian family) observe constraints on wh-movement in English. They pointed out that Malagasy and English differ in terms of the domains from which wh-extractiom is possible, as well as in what may be extracted. They hypothesized that UG is available to post-puberty learners. White et al's subjects were 38 adult Malagasy learners of English as a foreign language in Madagascar. They were given two tasks: A gramaticality judgment task and a question formation task. These tasks were not timed. The results indicated that Malagasy learners of English at different proficiency levels (i.e. high and low intermediate) knew what domains permit extraction and what domains prohibit it in 63 English. Such results are consistent with the claim that UG is available in non-primary (L2) acquisition. All these experimental studies by Hanania and Gradman (1977), Flynn (1984; 1987; 1988), Felix (1988), White (1988), Bley-Vroman et al. (1988), Liceras (1989), Thomas (1991), and White et al. (1992) argue, whether directly or indirectly, that UG is fully accessible in adult L2 acquisition just as in child L1 acquisition. These studies investigated several principles and parameters of HG such as the PHD, Subjacency, the ECP, pro-drop, and other syntactic structures. They found that adult learners are able to reset UG parameters which are different from their first language suggesting the availability of HG in L2 acquisition. Moreover, the results of these studies came as a counterevidence against the proponents of the CPH. 2.4.2.3 The Partial-Accessibility Of UG Studies As an argument against the full-accessibility of CG (in L2) position, Schachter (1988) and Slay-Vroman (1989) discussed some differences between first and second language learning situations. These differences reflect that it is not possible to suggest that the two learning cases are entirely 64 similar. First, Schachter (1988) explored four basic areas of differences between L1 and L2 acquisition cases. These are: completeness, equipotentiality, previous knowledge, and fossilization. These basic differences show that while children attain complete mastery of their L1, adults cannot achieve complete mastery of their L2. Adults are also not equipotential for any natural language. Schachter (1988:225) said that “equipotential is a cover term to indicate the ability to learn any arbitrary natural language 1: in the same amount of time and with the same ease as the time and effort required to learn a completely unrelated language Y." Adults' previous knowledge (i.e. their L1s) has a substantial effect on the second language learning situation. Finally, adults demonstrate instances of fossilization. That is, there is a “regular reappearance or re-emergence in IL [InterLanguage] productive performance of linguistic structures which were thought to be eradicated" (p. 228) . For examle, English native speakers may use the indicative form of the verb in Spanish in cases where it is inappropriate. Having discussed the major differences between L1 and L2 learning cases, Schachter argued that underlying processes for the two types of language acquisition are not the same. Further, she defended her position by cementing on the 65 morpheme and negation studies which show similarities between first and second language learning situations. She said that “such goes on in language use that is not attributable to the language faculty (or module) per se, but rather involves other systems independent of but interacting with it" (p. 230). The similarities so far discovered between child L1 acquisition and adult L2 acquisition are not due to the language module, but rather are due to other systems (or modules). Schachter raised the interesting possibility that the native language mediates access to UG. In other words, 0G is available to adult L2 learners only in the form in which it is instantiated in L1. Further support for Schachter's proposal came from Hley- Vroman (1989) who made several arguments about the role of 0G in L2 acquisition, in light of the fundamental differences between first and second language acquisition cases. He discussed nine basic characteristics in which he found adult language learning significantly different from child language learning, as a support for what he called the “Fundamental Difference Hypothesis". These are: Lack of success; general failure; variation in success, course, and strategy; variation in goals; fossilization; indeterminate intuitions; the 66 importance of instruction; negative evidence; and role of affective factors. These nine characteristics demonstrate that adults lack general, guaranteed success. Complete success is rare especiallwaith.respect to accent (pronunciation). Bley-Vroman explained that adults differ from one another with regard to success, course of learning, and strategies of learning. They also vary‘with regard to the type of attainment -- some seem concerned about grammatical corrections, while others seem concerned about having a good pronunciation. Adults tend to reach only a particular stage of learning, a stage short of native-like ability. They seem to lack clear grammaticality judgments. Formal instruction matters to them in second language learning; Negative evidence is also sometimes useful and necessary in. the L2 learning situation. Finally, affective factors such as motivation, personality and the like are crucial to L2 acquisition. Hley-Vroman (1989) noted that the adult second language learner can reconstruct principles of universal Guammar by observing his/her native language. He cited Eellerman (1977) who ‘showed that adult learners had ideas of what, in their native languages, was universal ... and.what was specific to the native language" (p. 53). 67 White (1985) conducted a study about the effects of the pro-drop parameter in adult L2 learning. She hypothesized that adult L2 learners have certain problems when their Lls have different parametric value (e.g -pro-drop) than the L2 (e.g. +pro-drop) . She suggested that the L2 learner would transfer the L1 value of the pro-drop parameter to the L2. She tested these hypotheses on 19 French (controls) and 54 Spanish (experimental group) learners of English. These subjects were given a gramaticality judgment task that contained 31 randomized English sentences, some of which were incorrect. They were asked to read the sentences at their own pace and to distinguish the correct from the incorrect. The results showed that French learners did significantly better than Spanish learners with regard to the property of the null subject. This is due to the fact that French, like English, is a non-pro-drop language -- learning was facilitated -- while Spanish is a pro-drop language -- learning was difficult. This finding supported White's claim that only those 0G principles and parameters which are instantiated in the learner's first language are available to adult L2 learners. However, the differences between the responses of the two groups with regard to the properties of inversion and “that-t" 68 effects were not significant. While they correctly rejected the sentences that contained Verb-Subj ect word order in declarative sentences, the subjects had problems in correctly identifying the ungrammaticality of subject extraction from clauses with the complementizer present ' (i.e.‘that-t" effects). In response to White's (1985) study, Lakshmanan (1986) argued that White's results are inconclusive because adult L2 learners from different pro-drop L1 backgrounds were not considered. Taking this issue into consideration, Lakshmanan conducted a study to investigate whether adult L2 learners from different pro-drop L1 backgrounds (Arabic, Spanish, and Japanese) would carry any property of the pro-drop parameter over from L1 to L2 (English) causing transfer errors. The subjects were 53 adults in total: 16 Spanish speakers, 21 Arabic speakers, and 16 Japanese speakers. As for the control group, White's (1985) data on the control group of 19 French speakers was adopted. The subjects were tested on the three pro-drop properties, using a grammaticality judgment task. The test used was identical to the one used by White in her study. Lakshmanan's results indicated that the differences in the responses between all the experimental groups and the 69 control group were not significant as far as the null subject is concerned. Furthermore, for all of the sentences that contained null subjects, the performance of the three experimental groups was at the chance level. These results differ from White's findings. However, similar results as those obtained by White were observed with respect to subjects' responses to Subject-Verb inversion and “that-t" effects. Lakshmanan concluded that it is not certain that Arabic, Spanish, and Japanese speakers are consulting their Lls in the processing of the null-subject property. An exception was noted with regard to “that-t" effects: The Arabic subjects appeared to transfer the asymmetry of their L1 into English (see 5 2.2.3 for an explanation of this asymmetry) . Schachter (1989) carried out a study to find out whether adult second language learners have access to the principles of UG. In particular, Subjacency was tested. The subjects included: 19 native speakers of English (the control group) ; 21 advanced Korean speakers of English; and 40 advanced Chinese and Indonesian speakers of English. Korean does not show evidence of Subjacency, while Chinese and Indonesian demonstrate some evidence of Subjacency. Subjects were given a grammaticality judgment task that contained four syntactic 70 construction tests (i.e. Sentential Subjects (SS), Relative Clauses (RC), Noun phrase Complements (NC), and Embedded Questions (EQ) as the syntax test and four Subjacency (violation) tests (i.e. *SS, *RC, *NC, and *EQ) as the UG test. Some examples of these constructions, which were used in Schachter's (1989:86-7) study, are in order (The use of asterisks indicates ungrammaticality, and italicized t indicates the original extraction site.) : (15) a. That oil prices will rise again this year is nearly certain. (SS) b. *Which party did for Sam to join t shock his parents? (*88) c. The professor that gave the most interesting lecture just left for Harvard. (RC) d. *Which problem did Bill find a principle which solves t? (*RC) e. There is a good possibility that we can obtain the information elsewhere. (NC) f. *Who did the police have evidence that the mayor murdered t? (*NC) g. The police didn't discover who the murderer was? (EQ) h. *Who did the police wonder who saw t? (*EQ) 71 The results showed that the control group passed the tests (i.e. the tests were appropriate), while many of the other subjects did not perform well in determining Subjacency violations. In other words, many subjects failed the Subjacency test, although they were proficient in English. These results constitute a major challenge for those who believe that UG is fully accessible to adult L2 learners. Furthermore, Schachter' s results provided some support for the position that only those UG principles which are present in the learner's first language will be accessible in the L2 learning situation. This position is well-supported by the Korean, Indonesian, and Chinese data. Since the Korean language demonstrates no evidence of Subjacency, many Korean subjects failed the test. On the other hand, both the Indonesian and Chinese languages show evidence of Subjacency; as a result, native speakers of those languages performed better as a whole than did the Korean subjects. Clahsen and Muysken (1989) argued that full knowledge of the properties of the L1 gramar is available in L2 acquisition. In other words, adult L2 learners only have access to UG as it is mediated through their L1 grammar. To support their argument, they discussed studies that dealt with the fundamental differences between L1 and adult L2 72 acquisition; such differences are said to be attributed to different kinds of learning mechanisms. In particular, they discussed their previous studies of L1 and L2 acquisition of three different areas of German grammar, namely verb placement, verb inflection, and negation (e.g. Clahsen and Muysken, 1986; Clahsen, 1988) . They concluded that adult L2 learners (when asked to give a grammaticality judgment in any language) fall back on UG principles that have instantiations in their first language. Clahsen (1990) supported Clahsen and Muysken's (1989) claim that adult L2 learners have access only to those HG principles which are instantiated in their L1. He stated that “the crucial differences . . . between L1 and L2 acquisition are (a) that adult L2 learners no longer have open parameters and (b) that they only have access to stable UG principles in so far as these principles have instantiations in the speakers' native language" (p. 143). White (1990) discussed the motivation for universal Grammar, and reviewed the three different positions regarding the availability of 0G to adult L2 learners. She reported on a UG study (presented at the 13th Annual Boston University Conference on Language Development, 1988) by Schachter who studied 18 adult native speakers of Dutch and 21 native 73 speakers of Korean. All subjects were advanced English as a second language (ESL) learners. Dutch, like English, shows evidence of Subjacency, while Korean.does not show evidence of it. The subjects were tested on their knowledge of Subjacency. The test was a grammaticality judgment task that contained Subjacency violations and grammatical sentences of equivalent complexity. Schachter's results demonstrated that while Korean speakers were accurate in their judgment of the grammatical sentences, many of them failed to reject Subjacency violations. In comparison, the Dutch speakers were considerably accurate in accepting the grammatical sentences as well as rejecting the Subjacency violations. The Dutch speakers, unlike the Korean speakers, behaved just like the control group. These results suggested that only 0G principles that are present in the learner's L1 are available for acquisition of the target language. Further, White (1990) discussed some studies (e.g. Bley- vroman.et al., 1988) which argue for the full-accessibility of DC in the L2 acquisition situation. She said that we cannot rule out at least indirect effects from the L1 despite the researchers' original hypotheses. She concluded her article by saying that “the question of UG accessibility in L2 74 acquisition is still unresolved. However, at the very least, the studies discussed in [this] article indicate that there is accessibility via the L1." 2.5 Summation It is generally accepted that child L1 and child L2 learners resort to innate linguistic properties (i.e. UG) during language acquisition. In other words, children have direct access to 0G (e.g. Chomsky, 1969; Brown, 1973; Dulay and Burt, 1974; Heckler, 1975; Sabra, 1987; Clahsen, 1988; O'Grady et al., 1989; Goodluck, 1991; and others). In contrast, it is still controversial whether the adult L2 learner, like the L1 learner, has access to the principles of UG. Three different positions on this issue can be detected. One position (e.g. Clahsen and Muysken, 1986), which focuses on the difficulties experienced by adult L2 learners, is that 0G is not accessible; a second position (e.g. Hley-Vroman et al., 1988) is that UG is fully accessible to adults; and a third position (e.g. Schachter, 1989) is that 0G mediates the acquisition of the L2 through the L1. Some researchers try to explain the status quo. Some examples are given now. First, White (1990:125) states that: 75 the differences in these positions can be traced in part to whether emphasis is on the relative lack of success of L2 learners (their knowledge, fluency, and ultimate attainment rarely approach that of native speakers) or their success (they do acquire many complex properties of language which are not transparent in the input). Second, Cook (1988:183) claims that “the problem in choosing between the three [positions] is that they might be true for different learners, or for different aspects of language for the same learner; L2 learning depends on an interaction between learner and situation.“ But they are intended to be true of all, for UG. Third, in his attempt to make a distinction between the learning experience of adults and children with regard to UG, Felix (1985) argues that the crucial difference is due to the separate faculties that adults and children bring to bear upon the learning experience. For adults, UG is in competition with late- acquired general principles of a problem-solving nature; children rely on UG alone. Finally, Hudson (personal communication) says that one explanation is certainly the flexibility of the theory. As I have said elsewhere, there is some evidence to suggest that adult L2 learners proceed in ways similar to child L1 learners (e.g. Bailey et al., 1974). White (1989) argues that adults, like children, achieve 76 competence (knowledge of language) which goes beyond the linguistic input received. That is, the competence of adults, like children, includes properties which are not immediately obvious and which are not explicitly taught. This fact gives support to the notion that there is no principled difference between child and adult language acquisition. Furthermore, Flynn (1987) argues that the biological endowment for language does not change substantially over time. Thus, UG can be argued to hold in adult L2 acquisition as well as child L1 acquisition. In support of this position, Cook (1985) wonders how the L2 learner knows that the question ‘*Isr the book that ti on sale is good?’ is ungrammatical unless there is something (UG) which guides him/her to understand that fact. Generally speaking, adults seem to do as well as children in acquiring many aspects of language and they manifest a similar development route in L2 acquisition (CooK, 1973; Taylor, 1974) . Given such considerations based on these studies and others (such as White, 1985; Bley-Vroman et al., 1988), there is some reason to believe that UG may continue to operate in adult L2 acquisition process. In other words, findings of these studies suggest that UG is still available to adults, if not totally, at least partially. CHAPTER 3: METHODOLOGY This research utilized a combination of perception and production methods for the collection of English and Arabic data and a combination of qualitative and quantitative techniques for analyzing the data. Presented in this chapter are descriptions of the research sample population (subj ects) , research instrumentation (the questionnaire), and procedures followed for data collection and statistical analysis. Prior to data collection, approval of the research methods of this project was granted from the (Michigan State) university Committee on Research Involving Human Subjects (UCRIHS) . 3.1 Subjects Data sought for this investigation was solicited from four groups: (1) Native speakers of American English as a control group; (2) Native speakers of Arabic who are advanced 77 78 learners of English as a foreign language (EFL) as a test- group; (3) Native speakers of Arabic as a control group; (4) Native speakers of American English who are advanced learners of Arabic as a foreign language (AFL) as a test-group. Groups one and two participated in the English experiment whereas groups three and four took part in the Arabic experiment. 3.1.1 Group One Group one (henceforth, G1) consisted of native speakers of American English; of those, 9 people did not return their questionnaires, and 7 people were dropped from the sample because their questionnaires were incomplete. The remaining 25 subjects were 9 males and 16 females, ranging in age from 19 to 50 (the majority of the subjects - 18 - are in their early twenties). They were all Michigan state residents, and had no background in Linguistics. Most of them are either undergraduate or graduate students at Michigan State university (MSU), East Lansing. Table 1 shows the distribution of subjects according to age, major of study, and educational level . 79 Table 1 Subjects' Profile I CATEGORY 1 2 3 4 TOTAL Age 18 2 3 2 25 Major 5 10 9 1 25 E. Level 20 5 0 0 25 A. Age: (1)19-29; (2)30-39; (3)40-49;-(-4+) 50-0ver. H. Major:(1)Science 5 Engineering; (2)Social Science; (3) Humanities; (4)fine Arts. C. Level : (1)0ndergraduate; (2)graduate The selection of this group was made in two ways. First, some friends and neighbors were approached for participation in this study. Second, some MSU instructors were asked to allow their students to take part in this research. 3.1.2 group Two Group two (hereafter, G2) consisted of 77 advanced learners of English enrolled in the English Language Center (ELC) at the Institute of Public Administration (IPA, a governmental agency which offers training programs related to civil service such as accounting, banking, sales, hospital administration, etc.) in Riyadh, Saudi Arabia, during the Sumner semester of 1995. Of them, 11 subjects were dropped from the sample population because they did not fill out their 80 questionnaires. The remaining 60 subjects were adult male Saudi Arabic speakers, ranging in age from 19 to 30 years (most of them - 51 - were under 25 years old). Except for 4 of them who are university graduates, all of the subjects are high school graduates. They have received basic instruction in English during their intermediate (3 years) and high school (3 years) studies. Specifically, they had been taught in a traditional, grammar-oriented fashion for four hours per week. In the ELC, however, they studied different language skills (grammar, writing, reading, conversation, and listening) for 24 hours per week. Obviously, the subjects had been exposed to English as a foreign language (i.e. in a non-English speaking environment), rather than English as a second language (i.e. in an English speaking environment). In comparison, there may well be considerable differences in the type and amount of input available in the two environments. The subjects were classified as advanced learners on the basis of their level in the ELC program. This level was determined by using a placement examination which was developed and administered by the ELC teaching staff. Table 2 indicates the subjects' profile according to age of first wosure to English (Exposure), years of formal instruction in 81 English (Instruction), length of study of English in English- speaking countries (E. in ESC), and major of study (Major). Table 2 Subjects' profile CATEGORY Exposure Instruction E . in ESC 1 4 54 54 2 53 6 5 11 3 3 0 1 16 4 O 0 O 11 TOTAL 6 O 60 60 6 O I A. Exposure: (137-11; (2)12-15; (3)16-20. I H. Instruction: (1)6-9; (2)10-13. C. E.in ESC:(1)None; (2)1-6 months; (3)6-1 year. D. Major: (1)Accounting; (2)8anking; (3)Sales; (4)Hospital Administration. This group was selected by personally contacting the Director of the ELC at the IPA to seek his permission to allow his students to be involved in this project. 3.1.3 Group Three Group three (henceforth, G3) consisted of 22 native speakers of Saudi Arabic. Of those, 10 people did not returned their questionnaires despite repeated attempts to get them back. The remaining 12 subjects were adult male Saudi 82 Arabians who were (MA or Ph.D.) Graduate students at MSU during the fall of 1995. The age of participants ranged from 30 to 39 years. They had no background in Linguistics and their majors of study were varied: 6 were majoring in Science and Engineering, 2 in Social Sciences, and 4 in Humanities. This group was selected by approaching my local friends and seeking their participation in this study. Since Modern Standard Arabic is not spoken natively by anybody, a clarification word is in order. I claim, following Lewkowicz (1978) and many other modern Arabists, that any speaker of a variety of Arabic who has at least completed a college education in an Arab country (conducted primarily in Arabic) will be a valid source of information about Arabic. 3.1.4 Group Four Group four (hereafter, G4) consisted of 58 native speakers of American English who were advanced learners of Arabic in the USA. Of them, 17 did not return their questionnaires and 7 were dropped from the sample because either they did not complete the two tests devised for data collection or they were not English native speakers. The remaining 34 subjects were adult (24) males and (10) females. 83 They came from different educational institutions, namely the Arabic Language Program (ALP) at The Ohio State university (080), The Program of teaching Arabic to Speakers of Other Languages (TASOL) at the Institute of Islamic and Arabic Sciences in America (IIASA), and Arabic and Islamic Studies Program.(AISP) at the university of Pennsylvania (0 of Penn). Table 3 shows the distribution of the subjects according to school, age, age of first exposure to Arabic, years of formal instruction in Arabic, length of study of Arabic in Arabic- speaking countries (A in ASC), major of study, and educational level. Table 3 Subjects' Profile L _ I CATEGORY 1 2 I 3 4 TOTAL _ School 11 12 11 0 34 Age 19 5 5 5 34 I Exposure 10 19 3 2 34 Instruction. 20 14 0 O 34 A. in ASC 19 6 6 3 34 Major 10 6 18 O 34 E. Level 1 10 23 0 34 A. School:(1)OSU,(2)IIASA;(3)U of Penn. B. Age:18-29;(2)30-39;(3)40-49;(4)50-over. C. Exposure:(1)0-8;(2)16-27;(3)33-37;(4)42-50 D. Instruction:(1)1-3;(2)3-6. E. A.in ASC:(1)None;(2)1-6 months;(3)6-1 year;(4)1-2 years F. Major:(1)Science & Eng.;(2)Social Science;(3)Humanities G. Level:(1)High School;(2)Dndergrad;(3)Grad. 84 As the table 3 illustrates, there were some subjects who have been exposed to Arabic at an early age. But, this was really not a true ‘exposure’. In fact, the majority of them had been only exposed to Quranic (i.e. the holy book of Islam which contains the revelations of Allah to Muhammad peace be upon him) Arabic since they are Muslims. Usually, they read transliterated text along with English translation. Thus, this is only a reading process done mainly in English. The rest of these subjects have been exposed only to Colloquial Arabic (not Modern Standard Arabic). Further, English has been used predominately because they have been raised in the USA and schooled in American Public schools. The selection of this group started in early fall of 1995 and was made in several steps. First, a request for student participation was sent to the LINGUIST List via e-mail. Upon this request, several responses from different individuals were received; of them, only the Coordinator of the ALP at OSU agreed to assist in the endeavor. Second, the Internet was searched for possible web sites dealing with Arabic studies. Two people were located and contacted via e-mail and telephone. Of these, only one agreed to help with the research, namely the Coordinator of Arabic language instruction at the University of Pennsylvania. Third, the 85 researcher knew of the IIASA which is funded by the government of Saudi Arabia. Thus, the Director of the TASOL program at this institute was contacted by phone and agreed to assist in this study. Now, a brief description of the above three Arabic programs is in order. First, the ALP at the OSU is located in Columbus, Ohio. It offers a proficiency-oriented and functionally-based approach that emphasizes student-student and student-teacher oral interaction. It focuses on the pragmatic meaning of language without neglecting the structural aspect, which is taught implicitly at the beginning stage. This approach combines oral interaction with audio, video, and interactive CALL (Computer-Assisted Language Learning) components. The basic learning skills of writing, reading, listening, and speaking are emphasized throughout the courses of instruction. Second, the AISP at the University of Pennsylvania is located in Philadelphia, Pennsylvania. It offers proficiency- based courses, implying that all activities are aimed at placing learners in the context of the native-speaking environment. Emphasis is on all four language skills: speaking, listening, reading and writing. 86 Third, the TASOL program is located in Fairfax, Virginia. It is part of an educational institution (IIASA) linked to the Islamic University of Imam Muhammad Ibn Saud of Riyadh, Saudi Arabia. This program emphasizes all four language skills and gramar. On the basis of their teachers' assessments and their level in their respective programs, the subjects were advanced learners of Arabic. Such assessments were adopted in this study. Admittedly, since these assessments came from three different programs, it is not clear whether ‘Advanced' is really equivalent among the students in this group. But, in agreement with White (1989a), it is hard to come by a similar measure of proficiency that can be applied to all subjects, given different circumstances. Moreover, given the small number of students who study Arabic in the USA, in general, and the small number of Advanced learners of Arabic in particular, it was necessary to find an adequate number of Advanced learners from different programs. 3.2 Research Instruments In order to test the question of the availability of UG in adult L2 (English or Arabic) acquisition, two tasks were 87 devised in English and Arabic, a grammaticality judgment tasks (henceforth, test one) and a question formation task (hereafter, test two). These two tasks were preceded by an introductory sheet eliciting demographic information which are: age, age of first exposure to English or Arabic, years of formal instruction in English or Arabic, length of study of English or Arabic in English-speaking countries or/and in Arabic-speaking countries, and finally, educational level and area of specialization (Appendix A). Such biographical data would be helpful in revealing possibly significant variation among subjects. The use of test one has both a practical and a theoretical rationale. Practically speaking, it is better than oral production tests in the sense that some phenomena are not available to examination in production data because of their rare or zero occurrence. This is due to the observation that learners try to avoid grammatical structures that they find difficult. Theoretically speaking, the use of grammaticality judgments enables the researcher to investigate the learner's linguistic competence (“knowledge") (Kellerman, 1986). Therefore, “learner judgments of acceptability are a reflection of that learner's competence in the target language" (Arther, 1980:182). Grammaticality judgments 88 provide “something about mental structures and processes that make learning possible, and about their interaction with the learner's input and environment" (Slay-Vroman et al., 1988:2) . Further, such tasks have been extensively utilized by generative linguists who have depended on their own intuitions about grammaticality as a basis for testing hypotheses about underlying principles of gramar (Ellis, 1994). Recently, however, criticisms have been advanced towards the use of grammaticality judgment task as the only source of information about the learner's linguistic competence (e.g. White, 1989a). These criticisms have been based on the observation that there can be response biases associated with such tasks. Thus, it is necessary to provide supporting evidence from different sources. For this reason, test two was used as an alternative means of assessing whether UG constraints adult L2 acquisition. 3.2.1 The Grammaticality Judgment Task (Test One) This perception test consisted of 36 randomized items requesting intuitional responses of grammaticality; 15 constitute the syntax test and 21 constitute the UG test. The syntax test was designed to ensure that subjects have reached 89 a level at which Subjacency, Empty Category Principle (ECP) , and the Null Subject Parameter (NSP) should have appeared in the L2. As for the UG test, it was devised to test for the presence or absence of the relevant structures that were being examined by the Syntax test. This test was constructed in English (Appendix C) and Arabic (Appendix D). The English version of the test includes grammatical sentences (the syntax test) and ungrammatical sentences (the UG test). In contrast, the Arabic version of the test contains 21 grammatical sentences, 15 of which constituted the Syntax test and the rest (items #s 5,6,14,16,26, and 28) were part of the UG test, and 15 ungrammatical sentences (the remaining part of the UG test). The ungrammatical sentences in this test violated the Subjacency principle, the ECP, or the NSF. For this test, subjects have to respond ‘grammatical', ‘ungrammatical' or ‘not sure' to the test-sentences, and to try to correct the sentences they think ungramatical. Such corrections provide the researcher with more evidence whether or not subjects' rejections of ungrammatical sentences come as a result of their awareness that these sentences violate UG principles [parameters . Each of the three UG properties being tested was represented by certain number of constructions. First, 90 Subjacency was represented by three syntactic construction tests, each consisting of three grammatical sentences, and three Subjacency tests, each consisting of three ungrammatical sentences. The constructions were as follows: Relative Clauses (RC), Noun-phrase Complements (NC), and Embedded Questions (EQ1) . Second the ECP was represented by two syntactic construction tests, each consisting of three grammatical sentences, and two ECP tests, each consisting of three ungrammatical sentences. These constructions were the following: That-trace Effects (Th-T), and Embedded Questions (EQ2) . Third, the NSP was represented by two syntactic construction tests, each consisting of three grammatical sentences, and two NSP tests, each consisting of 3 ungrammatical sentences for the English test or three grammatical sentences for the Arabic test. These constructions were: Subject Omission (SO), and Subject-Verb Inversion (SVI) . Finally, there was an overlap between some of the constructions; that is, EQ2 involved not only the ECP but also Subjacency, and Th-T involved the NSP as well as the ECP. Table 4 shows the categorization of English and Arabic sentences according to sentence type, item number of the Syntax test-sentences, and item number of the UG test- BODtODCOB . 91 Table 4 Categorization of Sentences 8 Type English STI Arabic STI English 19:13- Arabic ua'rr RC 12128 1825 152232 101929 INC 81831 22235 62629 111531 E01 52335 32030 91124 72336 Th-T 21316 41233 31734 172427 EQ2 41225 91832 101936 132134 ISO 21316 41233 71430 51426 SVI 21316 41233 202733 61628 S(entence) ;STI=Syntax Test Items ;UGTI-UG Test Items. Note that sentence types 4,6, and 7 have the same item is for the syntax test since they constitute the properties of the same parameter, the NSP. 3.2.2 The Question Formation Task (Test Two) This production test was constructed in English (Appendix E) and in Arabic (Appendix F). It consisted of 11 randomized declarative sentences, each containing and underlined phrase. Subjects were asked to write a grammatical English or Arabic question which questioned the underlined phrase. There were three simple sentences that required the formation of simple The questions; these served as the Syntax test. types of extraction out of these simple sentences are: object of a preposition, direct object, and subject. The extraction of 92 any one of these types will result in a grammatical English or Arabic question. The Syntax test was used to make sure that subjects were capable of making basic wh-questions. The other eight sentences served as the UG test. As far as domains of extraction are concerned, two of the sentences could have resulted in (Subjacency) ungrammatical extraction from RC, two in extraction from NC, and two in extraction from EQ1. AS for the ECP and the NSP, two of the sentences could have resulted in Th-T violations. The UG test was devised to examine for the presence or absence for Subjacency, the ECP, and the NSP. Table 5 indicates the categorization of English and Arabic sentences according to sentence type, item number of the Syntax-test sentences, and item number of the UG-test sentences . Table 5 Categorization of Sentences Sentence Type Syntax Test-item UG Test-item RC - 2’ 6 NC - 3’ 7 4. 93 Following Liceras (1989), if Th-T has been acquired, SVI and 80 would have been acquired too. For this test, subjects would find some way of phrasing their questions to avoid violating a principle/parameter of UG, if they know constraints on wh-movement in English/Arabic (White et al., 1992) . 3.3 Procedures The administration of the questionnaire to the four groups was done either directly (for G1 and G3) or indirectly (for G2 and G4) by the researcher. As for G1 and G3 (the control groups), subjects were contacted (in/outside classroom) by the researcher who handed in the questionnaire himself. As for G2 and G4 (the emerimental groups), the Directors of the different English/Arabic programs were instructed (either face-to-face for G2, or by phone and e-mail for G4) on how to administer the questionnaire to their students. For G4 only, mail was used to send questionnaires to the three Arabic program Directors. Subjects were told that they were needed to help with a doctoral research into the acquisition of English and Arabic. They were assured that they would not be assessed in any way 94 by the two tests and that the results of the tests would be kept confidential. Then, they were instructed to fill out a personal and academic information sheet. After that, they were asked to be concerned with certain points while answering the tests. First, they should concentrate on the grammar rather than meaning, spelling, vocabulary, and the like. Second, they should not change their first answer. Third, they should take a break whenever they need one. Fourth, they should do the tests by themselves and not consult a gramar book or work with anybody. To ensure optimum understanding, test instructions were translated into Arabic/English and typed on the instruction sheets (Appendix H) . As for test one, subjects were given three choices (“grammatical'l'ungramatical'l'not sure") . They were told to read the sentences and to judge them according to these three choices. They were required to correct any sentence that they found grammatically incorrect. With regard to test two, subjects were asked to make wh-questions out of the sentences, questioning only the underlined phrase. No time limit for completion of the two tests was reposedm Thus, subjects were free to spenduwhatever thee they needed. Imposing a time limit would result in failing to complete the tests by some learners. 95 After test completion, questionnaires were collected and classified according to groups (G1, G2, G3, G4) . Each questionnaire was assigned a sequential number (1, 2, etc.); then the subject's personal information along with his/her test-responses were coded, transferred into op-scan forms, and then entered into (computer) spreadsheet so that scoring and computation can be done more easily. 3.4 Methods of Data Analysis 3.4.1 Coding and Scoring The subjects' personal and academic information (i.e. independent variables such as age, major of study, etc.) were coded by grouping them into units in order to index group membership. For instance, the variable ‘age' was divided into units (1- 18-29; 2-30-39; etc.)and subjects were grouped into these units according to their own age (see I 3.1 for full account of such variables). As for subjects' responses, they were coded differently depending on the type of the test. Responses to test one were coded as follows: (1) Correct; (2) Incorrect with the right correction; (3) Incorrect without correction; (4) Incorrect with the wrong correction; and ( 5) Not sure. The status of these responses were the following: 96 (1) will be the right answer for only a grammatical sentence; (2) and (3) will be the right answers for only an ungrammatical sentence; (4) and (5) will be regarded as wrong responses. As for test two, responses were coded as follows: (1) Correct; (2) Simmle violation; (3) RC violation; (4) NC violation; (5) EQl violation; (6) Th-T violation; and (7) Other. For scoring the two tests, one point was assigned for a correct response and zero for an incorrect response. As for test one, the way for passing each of the syntax construction tests was for the subject to judge two of the three sentences as grammatical. Thus, a subject will pass the RC test if he/ she has judged as grammatical two of the three grammatical sentences containing RC, and so on. The criterion for passing each of the UG tests was for the subject to judge two of the three sentences as ungrammatical. Thus, a subject will pass the RC Subjacency test if he/she has judged as ungrammatical two of the three ungrammatical sentences containing a wrong extraction from relative clauses, and so on. There was an exception to the last criterion which concerned the Arabic test: The criterion for passing the SO and SVI tests was for the subjects to judge two of three sentences as grammatical. 97 Regarding test two, the criterion for passing the syntax test was for the subject to correctly supply two gramatical questions out of the three simple sentences. With regard to the UG tests, the criterion for passing the RC Subjacency test was for the subject to correctly supply one grammatical question out of the two sentences containing RC, and so on. 3.4.2 Statistical Tests Descriptive statistics, in the form of percentages, simple frequency tables, means, and standard deviations, was utilized. The aim of such procedure was to draw generalized descriptions regarding subjects' performance in the two tests. Further, a simple logistic Regression analysis was performed to investigate the effects of each background independent variable (e.g. age) on the two dependent variables, the Syntax test and the UG test. Regression analysis estimates a model in which the dependent variable (y) is approximated by a linear combination of the independent variables (x1,...,x2). It shows the relationship between a set of independent variables and one (or more) dependent variable(s). It answers the question: to what extent can a dependent variable be explained and predicted by one or more 98 independent variables (Hatch a Farhady, 1982)? For example, I estimate how the number of years of formal instruction in English/Arabic affects subjects' scores on the UG test. Further, a Chi-square test ( or simply x3 ) for a 2x2 contingency table was performed. This test was used to determine whether statistically significant difference exists among the groups (G1,G2,etc.) and/or between the two tasks used. Schachter's (1989:79) design of a contingency table was adopted in this study. This design is presented in Figure 1. Figure 1 Research Design for Testing for UG This research design helps reveal certain important predictions about the accessibility of UG in L2 acquisition. First, subjects who pass the Syntax test are predicted to pass the UG test also; thus, subjects who fall into cell A will be said to observe UG principles/parameters. On the other hand, subjects who fall into cell D have failed the Syntax test and, as a result, have failed the UG test as well. Finally, subjects who fall into cells B and C would constitute a 99 serious challenge to those who hold the position of the full- accessibility of UG in adult L2 acquisition. The data were further analyzed by converting the P(robability)-value into a percentage form.for easy reading. To do this, the following formula was utilized: “(1- P-value) x 100%". The probability level of .05 (95%) was selected as a threshold of statistical significance. Thus, a 95% probability of nonchance results was accepted as significant. CHAPTER 4: DATA ANALYSIS AND RESULTS This chapter presents the analysis and results of the two eameriments in this research, the English Experiment (EE) and Arabic Experiment (AE) . The data was obtained from the two tests given to G1 and G2 for the and from the two tests given to G3 and G4 for the AE. 4.1 The English Experiment This section provides the findings of the control (G1) and experimental (G2) groups in the two tests. Also, it presents across task and across group comparisons. 4.1.1 The control Group Hypothesis 1 assumes that the background independent variables (age, major, educational level) will have no statistically significant effect on the Syntax and UG scores obtained by group one in the two tests. To test this 100 a T fie meme at“: 5 s: we 101 hypothesis, three separate regression procedures were performed, two for test one --one with scores on the Syntax test (Table 6.1) and one with scores on the UG test (Table 6.2); and the third procedure was for test two (Table 6.3). The results were nonsignificant in the two tests. Thus, Hypothesis 1 was supported. Tables 6.1, 6.2, and 6.3 show’the p-values along with their percentages computed on the data which clearly indicate the nonsignificance of the attribute variables on test scores. Notice that an overall statistically significant regression value is set to be significant at p<.05 (9596). Table 6.1 Regression Test On The Syntax Scores (T.1) MO 0.7015922 0.7015922 29.85 0.7015922 29.85 0.246739 75.33 0.246739 75.33 0.5690341 43.1 0.5873026 41.27 0.3904715 60.96 0.4707726 52.93 0.09948679 90.06 0.8346364 16.54 0.8297221 17.03 0.8297221 17.03 0.8297221 17.03 0.1464118 85.36 0.09948679 90.06 0.8346364 16.54 0.8297221 17.03 °- ””379 _ °- ”“3“ 0.8297221 Table 6.2 Regression Test On The UG Scores (T.1) 102 . .... ... * ...... * . ...... . RC 0.05258698 94.75 0.5609153 43.91 0.2092923 79.08 NC 0.3792179 62.08 0.3792179 62.08 0.3792179 62.08 301 0.4120182 58.8 0.7249469 27.51 0.4046165 59.54 lTH-T 0.8261385 17.39 0.4463903 55.37 0.3825315 61.75 892 0.4780725 52.2 0.3248069 67.52 0.5690341 43.1 80 0.6300769 37.0 0.2301086 76.99 0.1753483 82.47 8V1 1 0.0 1 0.0 1 0.0 6.3 Regression Test On Test Two Scores SIM 0.4214111 57.86 0.04335453 95.67 0.4131625 58.69 RC 0.6504092 34.96 0.8020747 19.8 0.7197408 28.03 INC 0.7433741. 25.67 0.5362194 ‘46.38 0.4046165» 59.54 391 0.5534034 44.66 0.7823668 21.77 0.1278598 87.22 TH-T 0.3792179 62.08 0.3456719 0.863423 Note. SIM-Simple constituted the Syntax Test. 4.1.1.1 Results Of The Two tests Hypothesis 2 assumes that G1 groups subjects who pass the Syntax test will also pass the UG test and that subjects who fail the Syntax test will also fail the UG test. This group of native speakers overwhelmingly perform as expected in the two tasks. Thus, Hypothesis 2 is substantiated. The majority of subjects pass both the Syntax and UG tests for the 103 constructions tested; thus, they tend to fall in the A cell. Specifically, the mean score of test one is 21.44 and of test two is 17.67. Tables 7.1, 7.2, and 7.3 indicate the Subjects' results in test one and Tables 7.4, 7.5, and 7.6 show their results in test two. The method of presenting these Tables is that the first one gives Subjects' results in each construction separately, followed by Subjects' results in each property of UG as whole in terms of Means (M) and Standard Deviations (SD), and then Subjects' results in the whole test presented by Ms and SDs too. The frequency distribution of Subjects' responses in the two tests is listed in detail in Appendix G. Table 7.1 Test One Results “mmm— so svr 20 0 19 1 19 3 21 1 23 0 22 1 24 1 4 1» 5 0 3 0 3 0' 2 0 ‘2 0 0 0 104 Table 7.2 Means and Standard Deviations of Results M Sy. is short for Syntax. Note . Sub (j acency) -ac, we, :91; scrum-'1', r92 ; nap-so, svr . Table 7.3 Overall Means and Standard deviations Table 7.4 Test Two Results RC NC 301 18 3 17 1 17 1 18 1 I1 3 2 5 SD M 8D 8D Sub. Pass 23.33 1.16 20.67 1.16 19.33 0.58 Fail 1.67 1.16 4.33 1.16 0.33 0.58 ECP Pass 24.5 0.71 22.5 0.71 22.0 1.41 Fail 0.5 0.71 2.5 0.71 0.0 0.0 NSP Pass 24.0 0.0 24.0 1.41 23.0 1.41 Fail 1.0 0.0 1.0 1.41 0.0 0.0 105 Table 7.5 Means And Standard Deviations of Results L_ Sy. I no I syaoc u so a so u so Sub. Pass 19 o 19 1.73 17.33 0.58 t. Fail 6 o 6 1.73 4.33 1.16 IECP Pass 19 0 19 0 18 0 I Fail 6 01r 6 o 5 o Table 7.6 Overall Means and Standard deviations Sy. 00 l ihflfifii_ I M SD M SD M SD IPaSs 19.0 0.0 19.0 0.0 17.67 0.48 Fail 6.0 0.0 6.0 0.0 4.67 0.48 These results demonstrate that the two tasks effective and appropriate ones for UG testing. native speakers know both the syntactic constructions of the principles/parameters Subjacency'and.ECP constraints on.domains of extraction. are The English (grammatical) tested, and the They also recognize the ungrammaticality of subjectless sentences and verb-subject sentences (NSP). UG. Thus, they have access to 106 It is noteworthy, however, that the Subjects' relatively poor performance on test two was not due to violation of a property of UG; rather, some of the subjects did not follow the directions of the test properly. For example, instead of questioning the underlined word, they questioned some other phrase. Having said that, one may argue also that not following the test directionwwas a strategy to avoid'violating a UG principle. Finally, as presented in the above Tables, the results show that the correctness of answers varied across different construction types. In other words, this group performed generally better in some constructions than in others. Table 7.7 gives the rank.order of the principles/parameter tested.as measured by the percentage of those who pass both the Syntax and UG tests. Table 7.7 Rank Order of UG Types in percentages no me C NSP 92% - ECP 88% 72% Subjacency 77.32% 69.32% 107 One possible explanation for such an order is that a certain principle/parameter is easier/more difficult to access than others. It is interesting to note that the order is consistent across the two tests. 4.1.1.2 Across Test Comparison Hypothesis 3 states that there will be no statistically significant difference in the Syntax and UG scores obtained by group one across the two tests. The chi-square test was used to test this hypothesis. The results showed that there was no significant difference in the subjects' scores across the two tests; thus, the hypothesis was supported. The p-values and their percentages are supplied in Table 8. Table 8 Chi-square For Test Comparison m———' 5.905263 0.1163115 88.3796 6.396825 0.0938215 90.6296 6.311111 0.0974170 90.26% 6.230769 0.1009070 89.91% 108 4.1.2 The Experimental Group Hypothesis 4 says that the background independent variables (age of first exposure to English, years of formal instruction in English, English in English-speaking countries, and major of study) will have no statistically significant effect on the Syntax and UG scores obtained by group two in the two tests. To test this hypothesis, three separate regression procedures were performed, one for the Syntax test of task one (Table 9.1), one for the UG test of task one (Table 9.2), and one for task two in general (Table 9.3). The regression test results show some random significant effect of some of the independent variables on the dependent variables (i.e. RC, NC, etc.). Thus, Hypothesis 4 was not fully supported. In fact, one can trace only one consistent effect of “Exposure" on “80" in the Syntax test and UG test of task one. Tables 9.1, 9.2, and 9.3 indicate the p-values and their percentages commuted on the scores of the two tasks. 109 Table 9.1 Regression Test On The Syntax Scores (T.1) -------T- J---- Type Exposure % Instruction % RC 0.9270966 7.3 0.9379218 6.21 NC 0.3322787 66.78 0.1222775 87.78 EQl 0.3074646 69.26 0.2709474 72.91 TE-T 0.001634416 *99.84 0.1335307 86.65 EQ2 0.09018724 90.99 0.3350756 66.5 80 0.001634416 *99.84 0.1335307 86.65 SVI 0.001634416 “99.84 0.1335307 86.65 Table 9.1 (cont'd). I. In ESC % _ , 0.1277144 87.23 0.816326 18.37 0.001188633 '99.89 0.5636998 43.64 I 0.3606661 63.94 0.01791276 *98.21 0.3263482 67.37 0.3072984 69.28 0.1472112 85.28 0.721474 27.86 0.3263482 67.37 0.3072984 69.28 l0.3263482 67.37 0.3072984 Note. *P<.05 Table 9.2 Regression Test On The UG Scores (T.1) IEEEEII Exposure % Instruction % ,_ .i_l 0.2910545 70.9 0.9356642 6.44 0.2384059 76.16 0.7154724 28.46 0.0165465 *98.35 0.001674686 “99.84 0.6158443 38.42 0.2382089 76.18 0.2677423 73.23 0.001036542 *99.9 0.02897469 *97.11 0.1326322 86.74 0.2209572 0.3002859 .a T = =o._o._o._o._o_o_fl_m T =u=sinrfaiwe Foroiohloihem Table 9.2 (cont'd). ——-:::=--: 0.3720305 0.7507975 110 0.4417537 55.83 0.4642319 53.58 #I 0.704513 29.55 0.1693198 83.07 0.6478563 35.22 0.01073157 *98.93 0.7756574 22.44 0.7537754 24.63 I 0.5782007 0.03650714 0.704513 0.3929697 Nete. *P<.05 Table 9.3 Regression Test On Test Two Scores Type Exposure % Instruction % son 0.3502923 ‘64.98 0.1125675 66.75 ac 0.506615 49.32 0.06236171 91.77 ac 0.267148 73.29 0.3609504 63.91 :01 0.3625261 63.75 0.1879762 61.21 rs-r 0.1797332 62.03 0.757628 24.24 Table 9.3 (cont'd). In ESC 0.3357125 0.8821459 0.8072692 0.04930279 Note. *P<.05 0.3957236 60.43 0.2802787 71.98 0.4235362 57.65 0.3108651 68.92 I 0.7942093 20.58 0.3346315 66.54 7 ofw pro ‘SV 6ft tht fo 111 Table 9.1 demonstrates the following significant effects of variables: (1) “Exposure" had an effect on the scores of the properties of the Null Subject parameter, ‘TH-T', ‘SO', and ‘SVI'; (2) “English in English-speaking countries" had an effect on the ‘NC' scores; and (3) “Major" had an effect on the ‘EQ1' scores. As for Table 9.2, we can trace the following effects: (1) “mosure" had an effect on the ‘EQl' and ‘80' scores; (2) “Instruction" had an effect on the ‘EQl' and ‘EQ2' scores; and (3) “Major" had an effect on the ‘TH-T' and ‘SO' scores. Finally, Table 9.3 shows that only “major" had an effect on the scores of ‘RC'. 4.1.2.1 Results Of The Two Tests Hypothesis 5 assumes that G2 subjects who pass the Syntax test will also pass the UG test and that subjects who fail the Syntax test will also fail the UG test. The results of this group of non-natives were varied; that is, some subjects behaved as the hypothesis predicted, others did not. Thus, Hypothesis 5 was partly supported/rejected. Tables 10.1, 10.2, and 10.3 summarize the results of test one whereas Tables 10.4, 10.5, and 10.6 summarize the findings of test two. The way of presenting these Tables is that the first one gi' fo th SD th Ta = 1_2L_ = = ~E~ m Id 112 gives Subjects' results in each construction separately, followed by Subjects' results in each property of UG as a whole in terms of Means (M) and Standard Deviations (SD), and then Subjects' results in the whole test presented by Ms and SDs too. The frequency distribution of Subjects' responses in the two tests is listed in detail in Appendix H. Table 10.1 Test One results EQ2 80 8V]: 13 10 10 4 25 5 32 5 23 14 '39 5 '32 14 27 3 I20 3 I Table 10.2 Means and Standard Deviations of Results Fail 8.0 0.0 26.5 4.94 3.0 Note. Sub(jacency)-RC,NC,EQ1;ECP-TH-T,EQ2;NSP-SO,SVI. Sy. is short for Syntax. 113 Table 10.3 Overall Means and Standard Deviations 24.67 21.33 35.33 8.96 6.67 0.58 000 33 30 7 Table 10.6 Overall Means and Standard Deviations -I:'!- _ 0° __ _— sy. - M SD M SD M SD _PasS 50.0 0.0 28.83 5.89 25.67 6.13 r511 ”10.04_ 0.0 _31.16 5.69 6.634 0.24 l O N O ‘1 O fa D6 tk fa 0t] 114 As the above Tables demonstrate, about three quarters of the subjects show knowledge of the constructions in question by falling into cells A or C and, thus, passing the Syntax tests in the two tasks. Put differently, subjects correctly judge grammatical sentences of different constructions as grammatical (task one), and supply simple grammatical questions (task two). However, less than one sixth of the subjects exhibit that they fail both the Syntax and the DO tests in the two tasks; thus, falling into cell D. This performance supports Hypothesis 5. One may say that these subjects are not as advanced in their English as the other subjects are. On the other hand, only about one third of the subjects pass both the Syntax and.UG tests in the two tasks. So, they fall into cell A. Such performance supports Hypothesis 5. Penultimately, about 50 percent of the subjects either pass the Syntax test and fail the DO test or'vice versa; thus, they fall into cells C or B. The behaviors of these subjects fail to support Hypothesis 5. Ultimately, the results indicate that the correctness of answers may vary across different construction types. Subjects perform better in certain constructions than in others. Table 10.7 provides the rank order of the 115 principles/parameter tested as measured by the percentage of those who pass both the Syntax and UG tests in both tests(tasks). Table 10.7 Rank Order Of UG Types in Percentages Test 1 47.5% Subj acency 2 3 . 8 5% ECP 5 0% I 19.17% lSubjacency 35.55% I As the above Table shows, subjects' intuitions differ with different UG properties. But, the order is not consistent across the two tests. Thus, it is difficult to say which property is easier/more difficult to access than others. In comparison, the rank order of UG types is similar across the subjects (G2) and the controls (G1) on test two. In test one, however, the rank order is the same in terms of only one UG property, the NSP. It must be concluded that the NSP is the easiest to access. 116 4.1.2.2 Across Test Comparison Hypothesis 6 states that there will be no statistically significant difference in the Syntax and UG scores obtained by group two across the two tests. To test this hypothesis, the chi—square test was employed. Except for one construction (NC), the results indicate that there was a significant difference in the results across the two tests. Thus, the hypothesis was weakly rejected. The chi-values and p-values along with their percentages are presented in Table 11. Table 11 Chi-Square For Test Comparison m Chi-value P-value Percent RC 18.414067 0.0003613 *99.97% NC 1.623925 0.6539772 34.61% I EQl 8.397044 0.0384806 *96.16 TH-T 13.172907 0.0042772 *99.58 J Note . *P< . 05 As the previous section shows, the non-natives did generally better in test two than in test one. In other words, they performed much better in the production task (test two) than in the perception task (test one). One possible h) DJ: di 117 explanation is that non-natives are more experienced in production tasks than in perception tasks. 4.1.3 Across Group Comparison Hypothesis 7 assumes that there will be no statistically significant difference in the Syntax and UG scores obtained by groups one and two across the two tasks. The chi-square test was used to verify this hypothesis. The results indicate that there is a significant difference in the results of test one across the two groups; thus, the hypothesis is rejected. However, test two results show that there is a partially significant difference across the two groups; thus, the hypothesis is partly rejected/accepted. Tables 12.1 and 12.2 provide the chi-values, p-values, and percentages of differences in test one and test two respectively. Table 12.1 Chi-value For Group Comparison (Test 1) 118 P-value f Percent 25.170719 0.0000142 *99.99% NC 11.101709 0.0111884 *98.89% 1801 19.742708 0.0001919 *99.98% TH-T 24.160320 0.0000231 *99.99% EQ2 43.397980 0.0000000 *100% S0 11.360283 0.0099290 *99.01% SVI 7.969642 0.0466433 *95.34% the. *P<.05 Table 12.2 Chi-value For Group Comparison (Test 2) 3.220666 0.3588379 64.12% 16.512381 0.0008901 *99.92% 18.200901 0.0003998 *99.97% difference, the constructions , 4.4215278 0.2193975 ac (Subi acency) . 78.07% Table 12.2 demonstrates that there is no between the scores of the two groups, and significant in two of THJT (ECP). difference is due to the fact that the non-natives performed significantly better in the second task, in comparison to the first task. 119 However, the overall results of the two groups indicate that the native speakers (G1) did significantly better than the non-natives (G2) across the two tasks. Tables 12.3 and 12.4 present the overall results of both tests of group one and group two respectively. These results are presented in terms of means and standard deviations of subjects who pass and/or fail the Syntax test, the UG test, and both the Syntax and DO tests. Table 12.3 Overall Subjects' Results of Both Tests (G1) IPass 21.47 3.50 lFail 3.53 3.50 _ Note. N-25 Table 12.4 Overall Subjects' Results of Both Tests (G2) L Sy. - as ‘ sysoc - M SD M SD M SD % Pass 48.5 2.12 26.22 3.69 21.89 5.34 36.49 I Fail 11.5 2.12 33.78 3.69 7.17 0.47 11.95 I Nete. N-60 120 In agreement with White et al. (1992), natives and non- natives who passed the UG tests in the second task resorted to different strategies of paraphrasing sentences in order to avoid violating a principle of UG. Such paraphrasing appeared in different forms such as changing a complex sentence into a simple one, using wh-in situ (i.e. a wh-phrase which has not moved to [Spec, CP]), and so on. Some examples of paraphrases are in order. (16) a . The manager bought the dog that had brought the ring. (RC) Target Subjacency violation:*What did the manager buy the dog that had brought? Paraphrase: What did the dog bring? b. The father has proof that his son bought th_e red house. (NC) Target Subjacency violation:*What does the father have proof that his son bought? Paraphrase: Of what does the father have proof? c. John wondered who would sell the diamond ring. (EQl) Target Subjacency violation:*What did John wonder who would sell? Paraphrase: Who would sell what? Va Si 9: th ta: tht th 121 d. Tom suspected that Liza liked John. (Th-T) T_ar;get ECP violation: *Who did Tom suspect that liked John? Paraphrase: Who liked John? Or, Who did Tom suspect liked John? 4.2 The Arabic Experiment This section presents the analyses/results of controls (G3) and subjects (G4) in the two tasks of the Arabic experiment. Further, it provides across task and across group comparisons . 4.2.1 The Control Group Hypothesis 8 says that the background independent variable (major of study) will have no statistically significant effect on the Syntax and UG scores obtained by group three in the two tasks. To examine this hypothesis, three separate regression procedures were performed, two on task one (one on the set of the Syntax scores and the other on the set of the UG scores) , and the last procedure on task two. The results indicate that the background attribute had 1.1-1.... .. Eflfimifld 122 no significant effect on the subjects' scores of both tasks. Thus, the hypothesis is accepted. Tables 13.1 and 13.2 demonstrate the p-values and percentages of the effect of “major" on test one and test two respectively. Table 13.1 Regression Of Test One Scores % . - RC 0.3376009 66.24 0.1589887 84.11 NC 0.3096396 69.04 0.1037541 89.63 801 0.2618949 73.82 1 0.0 TH-T 0.3575508 64.25 0.7194269 28.06 EQ2 0.7881004 21.19 0.6231033 37.67 I SO 0.3575508 64.25 1 0.0 J SVI 0.3575508 64.25 0.3575508 64.25 I I —__ Note. Major 1-Regression of Syntax Scores; 2-UG Table 13.2 Regression of Test Two Scores Sin 1 0.0 RC 1 0.0 NC 0.3575508 64.25 EQl 0.1293913 87.07 C: 1i 123 4.2.1.1 Results Of The Two Tests Hypothesis 9 states that G3 subjects who pass the Syntax test will also pass the UG test and that those who fail the Syntax test will also fail the UG test. The results indicate that G3 (Arabic native speakers) performed as expected in all constructions in the two tests, as demonstrated by Tables 14.1, 14.2, 14.3 (for test one), and 14.4 (for test two). They pass both the Syntax and UG tests for the constructions tested; thus, they tend to fall in the A cell. The mean score for those who pass the first test is 10.28 and the second test is 12 (recall that the number of subjects is 12). The way of presenting these Tables is that the first one gives Subjects' results in each construction separately, followed by Subjects' results in each property of UG as a whole in terms of Means (M) and Standard Deviations (SD), and then Subjects' results in the whole test presented by Ms and SDs too. (Since the subjects achieved perfect score on the second task, I only provide one Table which shows their scores on each construction separately.) More details about the frequency distribution of subjects' responses in the two tests are listed in Appendix I. Table 14.1 Test One Results 124 Table 14.2 Means And Standard Deviations of Results [ZZZ—... .. _——‘ M SD M SD M SD Sub. Pass 12.0 0.0 11.33 0.58 11.33 0.58 Fail 0.0 0.0 0.67 0.58 0.0 0.0 ECP Pass 11.0 1.41 10.0 0.0 9.0 1.41 Fail 1.0 1.41 2.0 0.0 0.0 0.0 I NSP Pass 12.0 0.0 10.5 0.71 10.5 0.71 Fail 0.0 0.0 1.5 0.71 0.0 0.0 Table 14.3 Overall Means And Standard Deviations I Sy. no syaoo M SD M SD M SD Pass 11.67 0.58 10.61 0.68 10.28 1.18 Fail 0.33 0.58 1.39 0.68 0.0 0.0 Table 14.4 Test Two Results RC NC EQ1 TH-T 12 0 12 0 12 0 12 0 o I I0 0 0 0 0 0 0 125 These results show that the two tasks are effective and appropriate ones for testing of UG. The Arabic native speakers know' both the syntactic constructions for the principles/parameters being tested and the UG constraints on domains of extraction. They know what is grammatical and ungrammatical in Arabic. Thus, they have access to UG knowledge. Finally, the results show that subjects performed differently across different constructions. In other words, subjects tend to pass both the Syntax and UG tests in a certain property of UG more than others. Table 14.5 gives the rank order of DO types as measured.by those who pass both the Syntax and UG tests in the two tasks. Table 14.5 Rank Order of UG Types in Percentages Test 1 j Subj acency 94 . 41% 100% _.5" 126 In comparison, the above rank order is totally different from the rank order reported by English native speakers (G1) . This difference is not easy to explain; however, I agree with Felix (1988:290) who says that “in the absence of more detailed studies about UG-generated knowledge in adult L2 learners, the rank order figures must be viewed with much caution. " 4.2.1.2 Across Test Comparison Hypothesis 10 assumes that there will be no statistically significant difference in the Syntax and UG scores obtained by group three across the two tasks. This hypothesis was supported by the chi-square test that showed that there was no significant difference between the scores of the two tests. Table 15 supplies the results of the chi-square test. Table 15 Chi-square For Test Comparison —— Type Chi-value P-value Percent RC 0.086956 0.9933553 0.67% INC 0 1 0.0% IEQl 0.086956 0.9933553 0.67% lTH-T 0.363636 0.9476476 5.24% J IA 127 4.2.2 The Experimental Group Hypothesis 11 states that the background independent variables will have no statistically significant effect on the Syntax and UG scores obtained by group four in the two tests. To test this hypothesis, three separate regression procedures were carried out, one with scores on the Syntax tests (RC, NC, etc.) as the dependent variables (test one, Table 16.1), one with scores on the UG tests as the dependent variables (test one, Table 16.2), and one with scores on the Syntax and UG tests as the dependent variables (test two, Table 16.3) . Generally speaking, the results demonstrate that the attribute variables (i.e. school, age, age of first exposure to Arabic, years of formal instruction in Arabic, Arabic in Arabic- speaking countries, major, and educational level) did not have a biasing effect on test scores. Thus, the hypothesis was supported. Tables 16.1, 16.2, 16.3 provide the breakdown of the effect of the independent variables on the dependent variables in terms of p-values and percentages. 128 Table 16.1 Regression Test On The syntax Scores (T.1) 0.527355 0.495106 0.737984 NC 0.851055 14.9 0.312259 68.78 0.482892 51.72 EQ1 0.554112 44.59 0.907846 9.22 0.590879 40.92 TH-T 0.244273 75.58 0.674658 32.54 0.119060 88.1 EQ2 0.676534 32.35 0.259114 74.09 0.124238 87.58 SO 0.244273 75.58 0.674658 32.54 0.119060 88.1 I SVI 0.244273 75.58 0.674658 32.54 0.119060 88.1 Table 16.1 (cont'd). Instruct % AsASC % Major % E.Level % 0.664534 33.55 0.560670 43.94 0.703766 29.63 0.133043 86.7 0.468046 53.2 0.066850 93.32 0.475526 52.45 0.500923 49.91 0.937227 6.28 0.267435 73.26 0.541570 45.85 0.570589 42.95 0.733948 26.61 0.361444 63.86 0.693046 30.7 0.157584 84.25 [0.413617 58.61 0.728715 27.13 0.901648 9.84 0.818432 18.16 l0.733948 26.61 0.361444 63.86 0.693046 30.7 0.157584 84.25 l0.733947 26.61 0.361444 63.86 0.693046 30.7 0.157584 84.25 129 Table 16.2 Regression Test On The UG Scores (T.1) RC 0.746912 25.31 0.708371 29.17 0.737556 i NC 0.341925 65.81 0.250184 74.99 0.852860 14.72 I E91 0.978993 2.11 0.555204 44.48 0.563652 43.64 lTH-T 0.532265 46.78 0.689152 31.09 0.730130 26.99 EQ2 0.270921 72.91 0.500430 49.96 0.346623 65.34 80 0.432438 56.79 0.183289 81.68 0.275173 72.49 0.338240 0.572034 0.454528 A.ASC 0.656132 34.39 0.525664 47.44 0.451142 54.89 0.147816 85.22 0.647810 35.22 0.155701 84.43 0.358502 64.15 0.311836 68.82 0.613636 38.64' 0.562494 43.76 0.416634 58.34 0.177970 82.21 0.357858 64.22 0.365466 63.46 0.567298 43.28 0.696119 30.39 0.812990 18.71 0.620668 37.94 I 0.951327 4.87 0.098960 90.11 0.027964 *97.21 0.703061 0.951327 0.409887 0.008500 0.728799 *P<.05 IHJCGH Table 16.3 Regression Test On Test Two Scores LType School % Age %j Exposure % Sim 0.542507 45.75 0.291356 70.87 0.403529 59.65 RC 0.126840 87.32 0.956680 4.34 0.106092 89.4 NC 0.351210 64.88 0.616224 38.38 0.928462 7.16 “ EQ1 0.053969 94.61 0.884494 11.56 0.077946 92.21 H 1 _.__J"-48 130 Table 16.3 (cont'd). JIZEEEEEI x prose: is I may»: 1 as runwel IIIIII V0.312375 66.7 0.609941 39.01 0.04932 *95.07 0.747995 25.21 0.358501 64.15 0.06523 93.46 0.910063 9 0.645197 35.49 I 0.613904 16.61 0.177969 62.21 0.357656 64.22 0.572934 42.71] 0.690465 10.96 0.106002 69.2 0.172124 62.79 0.423364 57.67 I 0.69671 30.13 0.06355 93.65 0.419722 56.03 0.566145 41.19 the. *P<.05 Tables 16.2 and 16.3 show some random significant effect of two variables (“Instruction", and “Major") on some of the constructions. Table 16.2 shows that :(1) “Instruction" had an effect on the ‘NC' scores; (2) “Major" had an effect on the ‘SO’ and ‘SVI' scores. Table 16.3 indicates that “major" had an effect on ‘SIM' (SIMple . the syntax test). These variables' effects were not consistent throughout the data; so, one cannot claim.that these attributes had any biasing effect on test scores. 4.2.2.1 Results Of The Two Tests Hypothesis 12 predicts that G4 subjects who pass the Syntax test‘will also pass the DC test and that those who fail the Syntax test will also fail the UG test. This hypothesis 131 was partly supported by the results, as indicated by Tables 17.1, 17.2, and 17.3 for test one; 17.4, 17.5, and 17.6 for test two. The way of presenting these Tables is that the first one gives Subjects' results in each. construction separately, followed by subjects' results in each property of UG as a whole in terms of Means (M) and Standard Deviations (SD), and then Subjects' results in the whole test presented by'Ms and SDs too. Full details of the frequency distribution of subjects' responses in the two tests are provided in Appendix J. Table 17.1 Test One Results “*— =01 m-w * 602 so sv: '102 53135 124 161 209 227 18 4 15 11 11 5 11 7 14 3 3 2 1 4 132 Table 17.2 Means And Standard Deviations Of Results — Sy. UG M SD M SD Sub. Pass 24.0 4.0 12.67 5.03 Fail 10.0 4.0 21.33 5.03 ECP Pass 26.5 4.95 16.5 0.71 Fail 7.5 4.95 17.5 0.71 5.0 2.82 INSP Pass 23.0 0.0 29.0 0.0 21.0 1.41 I Fail 11.0 0.0 5.0 0.0 3.0 1.41 Table 17.3 Overall Means And Standard Deviations [Fail 9.5 1.80 14.61 8.54 4.89 1.83 __ Table 17.4 Test Two Results RC NC 301 TH-T 26 0 24 0 24 0 23 0 8 2 I9 2 133 Table 17.5 Means and Standard Deviations of Results t Sy. UG SyaUG M SD M SD M SD ISub. Pass 32.0 0.0 24.67 1.16 24.67 1.16 I I Fail 2.0 0.0 9.33 1.16 2.0 0.0 I ECP Pass 32.0 0.0 23.0 0.0 23.0 0.0 I Fail 2.0 0.0 11.0 0.0 2.0 0.0 Table 17.6 Overall Means and Standard Deviations M 1 \ Pass 32.0 0.0 23.83 1.19 23.83 1.19 I Fail 2.0 0.0 10.17 1.19 2.0 0.0 I __ These Tables indicate that the adult Arabic learners can recognize the constructions tested, as shown by the fact that about three quarters of them pass the Syntax tests; thus, falling into cells A or C. In other words, most subjects judge grammatical sentence items as grammatical. However, about one half of the subjects demonstrated knowledge of both Syntax tests and their equivalent UG construction tests. These subjects fall into cell A, thus, substantiating hypothesis 12. 134 A smaller number of subjects (about 10 percent) failed both the Syntax and‘UG tests. The behaviors of these subjects were predicted by Hypothesis 12. These subjects must have lower competence of Arabic than the other subjects. On the other hand, about 30 percent of the subjects either passed the Syntax test and failed the UG test or vice versa; thus, falling into cells C or H. The behaviors of these subjects failed to support Hypothesis 12. Finally, the findings of this group show' that the correctness of answers varied across different construction types. In other words, subjects tend to pass both the Syntax and DC tests in a certain property of UG more than others. Table 17.7 presents the rank order of UG types as measured by those who pass the UG principles/parameter. It shows that the order is not consistent across the two tests. Thus, it is difficult to say which UG property is easier/more difficult to access than others . 17.7 Rank Order of UG Types in Percentages NSP ECP Subj acency 135 In comparison, the rank order of UG types is similar across G3 (the controls) and G4 (the subjects) on test two. In test one, however, the rank order is completely different across the two groups. It is interesting to note, however, that the natives (G1 and G3) were consistent in their performance across the two tests (though they demonstrated different order from each other). On the other hand, the non-natives (G2 and G4) did not show consistent order across the two tests (and the order was different across these two groups too). They also demonstrated similar order with their respective natives on only test two. These results may indicate that the natives and non-natives are productively similar but perceptively different . 4.2.2.2 Across Test Comparison Hypothesis 13 states that there will be no statistically significant difference in the Syntax and UG scores obtained by group four across the two tasks. To test this hypothesis, a chi-square analysis was performed. This hypothesis was partly accepted/rejected, as the results in Table 18 indicates. 136 Table 18 Chi-square Of Test Comparison 'm——_l RC 11.333333 0.0100534 *99% INC 17.241379 0.0006304 *99.94% 891 4.432432 0.2183969 78.17% 4.742857 0.1916242 80.84% the. *P<.05 This partially significant difference in the two tests' scores was due to the fact that the non-natives performed generally better in test two than in test one. 4.2.3 Across Group Comparison hypothesis 14 assumes that there will be no statistically significant difference in the Syntax and UG tests' scores obtained by groups three and four across the two tasks. The chi-square test was utilized to test this hypothesis. The results of test two supported this hypothesis. In test one, however, the results indicated that there was a significant difference between the two groups in only two Subjacency constructions, RC and NC. Thus, Hypothesis 14 was partly accepted/rejected. Tables 19.1 and 19.2 present the chi- values, p-values, 137 and test two respectively. Table 19.1 Chi-value for Group Comparison (T.1) and percentages of difference in test one 11.732026 0.0083598 *99.17% NC 24.830450 0.0000167 *99.99%? lEQl 6.569444 0.0869636 91.31% ITH-T 5.778966 0.1228734 87.72% I802 2.127451 0.5463792 45.37% 80 1.900653 0.5932802 40.68% I SVI 1.629827 0.6526457 34.74% Note. *P<.05 Table 19.2 Chi-value for Group Comparison (T.2) mar—"516' 0.965944 0.8094916 19.06% INC 1.647058 0.6487672 35.13% 891 1.647058 0.6487672 35.13% TH-T 1.647058 0.6487672 35.13% The overall results of the two groups demonstrate that the native speakers (G3) performed generally better than the non-natives (G4) across the two tasks. Tables 19.3 and 19.4 provide the overall results of both.tests for groups three and 138 four respectively. These results are given in terms of means and standard deviations of subjects who pass and/or fail the Syntax test, the UG test, and both the Syntax and UG tests. Table 19.3 Overall Subjects' Results of Both Tests (G3) Note. N-12 Table 19.4 Overall Subjects' Results of Both Tests (G4) 06 , --_- M SD % 21.61 3.13 19.30 6.40 56.76 12.39 3.13 3.44 2.04 10.11 I _ Note . N-34 As noted in the analysis of the English experiment, natives and non-natives who passed the UG tests in the second task resorted to different ways of paraphrasing sentences in order to avoid violating a principle of UG. The data revealed different forms of paraphrasing such as changing a comlex sentence into a simple one, using wh-in situ, and so forth. Some examples of paraphrasing are provided below. 139 (17) a. Ra'a zaid-un al-fatat allati ishtarat al-sa9a. Saw Zaid-nom the-girl who bought the-watch “Zaid saw the girl who bought the watch." Target (RC) Subjacency violation: *matha ra'a zaid al-fatat allati ishtarat? What saw Zaid the-girl who bought “*What did Zaid see the girl who bought?" Paraphrase: ayyu fatatin ra'a zaid? Which girl saw Zaid “Which girl did Zaid see?" b. saddaqa ali al-khabar ?anna muhammad saraq accepted Ali the-news that Muhammad stole al-tufaha. the-apple “Ali accepted the news that Muhammad stole the apple." Target (NC) Subjacency violation: *matha saddaq ali al-khabar ?anna.muhammad saraq? What accepted Ali the-news that Muhammad stole “*What did Ali accept the news that M. Stole?" 140 Paraphrase: matha saraq muhammad? What stole Muhammad “What did Muhammad steal?" c. sa'al al-muzar9 al-rajul man sa-uhthitu asked the-farmer the-man who would-bring al-baqara. the-cow “The farmer asked the man who would bring the cow." Target (EQ1) Subjacency violation: *matha sa'al al-muzar9 al-rajul man sa-uhthiru? what asked the-farmer the-man who would-bring “*What did the farmer ask the man who would bring?" Paraphrase: matha sa'al al-muzar9 al-rajul? What asked the-farmer the-man “What did the farmer ask the man?" d. i9taqad zaid ?anna at-tifl sharab al-halib. thought Zaid that the-child drank the-milk “Zaid thought that the child drank the milk." 141 Target (TH-T) ECP violation: *matha i9taqad zaid ?anna sharab al-halib? what thought Zaid that drank the-milk “*What did Zaid think that drank the milk?" Paraphrase:i9taqad zaid ?anna man sharab alhalib? thought Zaid that who drank the-milk “Zaid thought that who drank the milk." 4.3 Across Experiment Comparison Hypothesis 15 states that there will be no statistically significant difference in the Syntax and UG test scores obtained by experimental groups two and four across the two tasks. To test this hypothesis, a chi—square analysis was employed. This hypothesis was generally accepted, as the results of test one indicate in Table 20.1. There was a significant difference in only one construction, namely 80. In test two, however, Hypothesis 15 was partly accepted/rejected, as the findings show in Table 20.2. There was a significant difference in the results of both groups in two constructions, NC and EQ1. 142 Table 20.1 Chi-square for Group Comparison (T.1) g g RC 1.452399 0.6932994 30.68% NC 4.718585 0.1936022 80.64% EQ1 2.331372 0.5065374 49.35% TH-T 5.905725 0.1162881 88.38% EQ2 5.746973 0.1245908 87.55% 1 SO 8.762178 0.0326258 *96.74% I SVI 4.718300 0.1936255 80.64% I Note. *P<.05 Table 20.2 Chi-square for Group Comparison (T.2) Type Chi-value Percent I 2 .625219 0.4530853 54.7% NC 14.102534 0.0027688 *99.73% EQ1 15.897765 0.0011900 *99.89% I TH-T 1.604957 0.6582663 34.18% I Note. *P<.05 Generally speaking, group four (Arabic learners) performed better than group two (English learners) in both tests. The percentage of those who passed both tests was 56.76% in G4, and 36.49% in G2. One possible explanation for this difference is that G4 were more advanced in their knowledge of Arabic than G2 in their knowledge of English. 143 Other factors may have contributed to this difference such as motivation, and the like. In general, the two experimental groups performed better on the second (production) task than on the first (perception) task. This performance may be due to the fact that the subjects have greater difficulty in perceiving UG principles] parameters. This difficulty level can be attributed to lack of experience in such tasks. 4.4 Foreign Language Vs. Second Language In this section, I compare the overall results of test one Subjacency scores of the English learners (G2) to those of Schachter's (1989) subjects. The purpose of this comparison is to find out if there are significant differences between learning English as a foreign language (EFL) and learning English as second language (ESL). Schachter's subjects were 20 Chinese, 21 Korean, and 20 Indonesian people. They were advanced learners of ESL. They were given a grammaticality judgment task that contained four constructions (88,RC,NC,EQ) which aimed to test for the accessibility of Subjacency (UG) in L2 acquisition. For the purpose of comparison, I focused on only three of these four 144 constructions since this study did not include the construction of Sentential Subject (SS). The method of comparison was to collapse the three sets of scores (i.e. the scores of RC, the scores of NC, and the scores of E01) into one 232 table for each group; and then performed a chi-square analysis on the number of those who pass/fail the Syntax tests, pass/fail the CG tests, and pass/fail both the Syntax and HG tests. Also, percentages were calculated. Each one of Schachter's three groups was compared to group two of this study. Table 21 presents the results of the chi—square test along‘with percentages. Table 21 Chi-square for Group Comparison - .... I... —_ NO % NO % Chi-value Sy.?ass 126 70 48 80 *34.97 Fail 54 30 12 20 *26.73 IUG Pass 67 37.22 24 40 *20.32 Pail 113 62.78 36 60 *39.80 Bo.Pass 43 23.89 22 36.67 *6.78 I rail 30 16.67 10 16.67 *10.0 145 Table 21 (cont'd). I Chinese 7 7 NO 96 Chi-value NO 95 Chi-value 50 83.33 *32.82 46 73 *37.20 10 16.67 *30.25 17 27 *19.28 27 45 *17.02 10 15.87 *42.19 33 55 *43.83 53 83.13 *21.69 24 40 *5.39 10 15.87 *20.54 l7 11.67 *l4.30 17 26.98 3.6 ‘ Note. Bo(th) -Sy(ntax)&UG;Indo is short for Indonesian. *P<.05 (df-1). The percentages are for the ms, not for the Chi-values. As indicated in the above Table, the results of each of Schachter's three groups differed significantly from those of the Arabic-speaking learners of English in this study. The Chinese and Indonesian subjects performed generally better than the Arabs and the Koreans did. In contrast, the Arab subjects performed better as a whole than the Korean group did. According to Schachter (1989) , Korean shows no evidence of Subjacency; as a result, the Koreans performed relatively poorly on the test. The other groups (Chinese, Indonesian, and Arab groups) performed generally better than the Korean because their languages show evidence of Subjacency; and , thus, accessing 0G was somewhat facilitated. 146 Now, the question is why did the Chinese and Indonesian subjects perform generally better than the Arabs, even though Subjacency obtains in Arabic as it does in English? One possible explanation is that the learning context may have some effect on the outcomes of those subjects. In other words, ESL context may well be a positive factor that contributed to the better performance demonstrated by the Chinese and Indonesian subjects. Of course, there may be other contributing factors, one of which is different teaching materials and/or methods. Finally, I must point out that for lack of literature, no attempt was made to compare the performance of G4 (learners of Arabic) with that of other groups in a different study. CHAPTER 5: DISCUSSIONS AND CONCLUSIONS 5.1 Discussions Given the results of the English and Arabic experiments, do adult L2 learners have access to the principles and parameters of UG? The answer should be affirmative because a good number of subjects passed both the Syntax and HG tests of both tasks in the two experiments. According to White (1989a), chance performance would be 25% because there are four possible cells into which subjects could fall into. The results indicate that the non-natives scored above chance in both experiments. The percentage of those who passed both tests is 36.49% for the English learners and 56.76% for the Arabic learners. The difference between the two experimental groups might be attributable to the fact that they are not equally advanced learners of the target language. Thus, the results of the two experiments constitute a challenge to advocates of the no-UG-access hypothesis. If adult L2 learners have no access to UG (and only have access 147 148 to problem-solving and hypothesis-testing mechanisms), they should be totally unable to work out the appropriate restrictions on wh-movement in English/Arabic which cannot be directly induced from surface structure (White et al., 1992) . Those who passed the CG tests did not violate the principles of UG. Arguing for the UG-accessibility position, do adult L2 learners have FULL access to UG as in the case with L1 children? The answer is no. The results of both experiments demonstrate that there were many subjects who failed the 0G tests, thus falling into cell C. If UG was fully available to the subjects, they would have done well on determining UG violations for the constructions they knew in English/Arabic. Moreover, if adult L2 learners have full accessibility of UG, why did they not do as well as the English/Arabic native speakers, even when the principles of Subjacency and ECP operate in both English and Arabic. The above argument leads us to the third position that 0G is only partially accessible to adult L2 learners. According to this view, only those portions of 0G which are instantiated in the leaner's first language are available in the second language. Given the findings of this research, this position is not supported. Although both languages English and Arabic 149 show evidence of Subjacency and ECP, a number of subjects in each experiment failed to recognize UG violations for the constructions they already knew. Reaching similar results, Schachter (1989) explained that L2 learners might have difficulty in accessing their linguistic knowledge reliably. Furthermore, the results of the Null Subject Parameter (NSP) in both experiments constitute a serious challenge to proponents of the UG-partial accessibility position. The NSP is set differently in English and Arabic; English is a non- pro-drop language (negative) while Arabic is a pro-drop language (positive). Yet, many subjects performed significantly better in the NSP tests than in the tests of Subjacency and ECP which operate almost similarly in both languages. The present findings do not support any of the three positions on the question of the accessibility of 0G in adult L2 learning. The subjects did not perform as bad (below chance level) to say that 0G is not available to them; and they did not do as well as child L1 learners (or native speakers) to say that UG is available to them. Further, subjects' results came as a counterargument to the prediction of the position that claims that 0G is only partially accessible to adults . 150 For the moment, the picture seems more complicated; UG is available but neither fully nor partially. The proponents of the full-access hypothesis claim that the L2 learner proceeds in the same way as the L1 learner. On the other hand, the advocates of the partial-0G access hypothesis believe that the L2 learner has only access to 0G principles/parameters that are operative in his/her L1. The findings of this paper form a major challenge to these positions (and, of course, to the no-access position). Actually, we can reach a solution to this problem by carefully examining the rationale behind the conflicting positions. First, the no-access hypothesis should be totally disregarded for the following reasons: (1) There are relatively few studies which claim that UG is entirely not available to adult L2 learners. If we combine the studies which claim full UG-accessibility with those which claim partial UG- accessibility, we can conclude that these studies demonstrate that UG is available in one way or another; thus; these studies together constitute a serious challenge to the no- access position. (2) Two prominent advocates of this position have changed their no-access position to the partial- access position, namely Clahsen and Muysken (1989) . (3) Since there is no major biological restructurings in the human brain 151 after birth, it is reasonable to assume that, in some way, UG also determines L2 acquisition (Lust, 1988). In other words, UG is available to adult L2 learners because biological endowment for language does not change substantially over time, despite CPH (i.e. Critical Period Hypothesis) (Flynn, 1987). (4) Differences between L1 and L2 acquisition have nothing to do with the presence or absence of UG. They only mean that adult L2 learners face more learning difficulties than their L1 counterparts. UG is one component of an acquisition theory, whether of L1 or L2. Such component will interact with various others (such as personality, aptitude, emotional states, etc.) and the failure of L2 learners may be attributable to these other components, and not necessarily to the non-accessibility of UG. (5) Finally, to say that 0G is not available is to go backward in time and say that adults learn a habit system, a system of disposition to behavior, acquired through training and conditioning (Chomsky, 1986a) . Such an idea has long been abandoned as a way of emlaining L1 acquisition and should be abandoned for L2 acquisition too. Second, we should do away with the position which states that UG is partially available to adult L2 learners. It is partial-access in the sense that only L1 UG- principles/parameters are available to them. This position is 152 not convincing for several reasons: ( 1) many studies (e.g. Bley-Vroman et al., 1988) have shown that adult L2 learners perform above chance in grammaticality judgment tasks even if the L1 (e.g. Korean) of these learners do not show evidence of a UG principle (e.g. Subjacency) as in the L2 (e.g. English). (2) Other studies (e.g. this paper) indicate that adult L2 learners do not do as well as L1 learners even when a VG principle is involved in both languages, the L1 and the L2. (3) Differences between L1 and L2 acquisition should not be taken to mean that the accessibility of CG is diminished. Such differences may only contribute to the difficulties adult L2 learners face while accessing UG. Thus, the accessibility issue should be separated from the difficulty issue. (4) A number of studies (e.g. Schachter, 1989) did not fully support this position. They only provided limited support. (5) Finally, the proponents of this position fail to distinguish between L2 performance and the acquisition of L2 competence (White, 1989) . Finally, the hypothesis that UG is fully available should be reformulated and reinterpreted. It should be reconstructed for several reasons: (1) Most studies (e.g. Bley-Vroman et al., 1988) did not fully support this position. Although subjects of these studies achieved scores considerably higher 153 than chance, they did not score as the native speakers did. (2) Many studies indicated that adult L2 learners rarely if ever achieve native-like competence (Ellis, 1994). (3) The advocates of this position claim that if the L1 and the L2 have similar parameter settings, learning is facilitated because there is no need for parameter-resetting; but when the L1 and L2 have different parameter-settings, learning pattern will resemble the early stages of L1 acquisition (Flynn, 1987) . This argument is not supported by the findings of this research which show that subjects performed generally better in constructions that differed parametrically from English to Arabic (e.g. the NSP) than constructions that operated similarly in both languages (e.g. Subjacency) . Given the above arguments, we can try to solve the 0G paradox by saying that 0G is still available to adult L2 learners but it interacts with other linguistic, cognitive, and socio-psychological components that have been acquired by the learner over the years. Such interactions cause learning processing difficulties (e.g. contradictions, blockage, etc.) which contribute in partly distorting the accessibility of the underlying principles/parameters of UG. The mind/brain of the adult L2 learner has more linguistic and cognitive components 154 to compute than the child L1 learner; hence, the differences that are found between the two cases of acquisition. It is very important to assume the interactions of various components of the human mind which process language, vision, memory, etc. in order to solve the CG paradox. Perhaps we can describe UG in isolation as one component, but we cannot analyze and explain its operation without understanding, describing, explaining, and analyzing other faculties of the mind. Of course, we are far from understanding what the human mind/brain is doing while processing language, but at least we can observe its results (Chomsky, 1988) . Stated somewhat differently, children have direct access to Core grammar which is a particular instantiation of built- in UG principles and parameters. In contrast, adults have access to core grammar but only through layers of peripheral components, an idiosyncratic linguistic component, a cognitive component, and a socio-psychological component. Figure 2 will help illustrate this position. 155 1 < ------ INPUT ------ > 1 2 ~ 3 4 Child L1 Acquisition Vs. Adult L2 Acquisition 1 . Core Grammar ; 2 . Peripheral Linguistic Component ; 3 . Peripheral Cognitive Component ; 4 . Peripheral Socio- psychological Component . Figure 2 UG-accessibility in Language Acquisition One may be tempted to ask why we have tremendous variations in subjects' results from one UG-study to another. One way to answer this question is to consider the factors that may condition functional computation of components in the bilingual brain as suggested above and as are represented in levels 2, 3 and 4 in adult L2 acquisition in Fugure 2. 156 5.2 Conclusions This study was carried out to examine the acquisition of English by Arabic native speakers and the acquisition of Arabic by English native speakers. It explored the theory of CG and its relationship to the processes of first and second language acquisition. Particularly, it was concerned with a principles and parameters approach to UG, as realized in Government and Binding theory (Chomsky, 1981). It focused on the question of the accessibility of UG in adult L2 acquisition by investigating whether or not advanced learners of Arabic and advanced learners of English as a foreign language rely on the principles/parameters of UG. The specific properties of HG which were investigated were the Subj acency Principle, the Emty Category Principle (ECP) , and the Null Subject Parameter. In addition, the study investigated whether or not there was a significant difference between perception and production tasks in measuring UG principles and parameters. Moreover, the study discussed the nature of (JG-based acquisition studies and commented on the 0G properties that were considered in such studies. The study contained two experiments, English and Arabic, with two groups of subjects in each experiment. In the 157 English experiment, there were 25 controls and 60 advanced Arabic-speaking learners of English, while in the Arabic experiment, there were 12 controls and 34 advanced learners of Arabic. The subjects were asked to fill out an information sheet about their age, age of first exposure to English or Arabic, years of formal instruction in English or Arabic, length of study of English or Arabic in English-speaking countries and/or in Arabic-speaking countries, and finally, educational level and area of specialization. Then they were tested using two different types of tasks, a grammaticality judgment task and a question formation task. The first task tested the receptive abilities of the subjects while the second task tested their productive abilities. Upon data collection, statistical procedures were utilized for analysis of data. Such procedures were descriptive statistics (means, standard deviation, etc.), a generalized logistic regression test, and a chi-square test. The regression test was used to investigate the effects of each background variables on the scores of the two tasks. In contrast, the chi-square test was used to determine whether or not statistically significant difference existed within/ between groups . 158 A number of conclusions have been drawn from this research: 1. Rejection of the position of no-UG-access in adult L2 acquisition and the position of partial-UG-access in adult L2 acquisition. These two positions are shown to be inconclusive for several reasons explained earlier. 2. Reformulation of the position of full-UG-access in order to solve the UG paradox. This position reconstruction reads as follows: UG may still be accessible to adult L2 learners but its accessibility is partly distorted/delayed by late acquired linguistic, cognitive, and socio-psychological components which are in constant interaction with UG. In other words, children directly access core grammar (UG) while adults access core grammar through late developed peripheral components . 3. Generally speaking, subjects (and controls) tend to perform better in production tasks than in perception tasks. One possible reason is that language learners are more emerienced in production tasks than in other tasks. 4. The results of the rank order of HG types have indicated that the natives (G1 and G3) and the non-natives (G2 and G4) are productively similar but perceptively different. 5. There are some factors that may condition functional computation of components in the bilingual brain. Some of 159 these factors are the representation of language in the learner's mind, the language acquisition context, the structural relationship between the L1 and L2, and so on. 6. Learning a language as a second language (rather than as a foreign language) may contribute positively in accessing UG more accurately. This is due to the fact that the type (and quality) of input is different from one environment to another. 7. English-speaking learners of Arabic performed generally better than Arabic-speaking learners of English in both tasks even though their respective languages show evidence of Subjacency and ECP. 8. The experimental groups (G2 and G4) performed generally better in the NSP tests than any other test even though the parameter-setting is different from one language to another. 9. The results of the second task revealed some of the subjects' processing strategies such as changing a complex sentence into a simple one, using wh-in situ, etc. in order to avoid violating a principle of UG. 160 5.3 Recommendation For Further Research In order to better understand UG and to permanently solve the controversy over the UG-accessibility question found in adult L2 language acquisition literature, we must draw on the work of cognitive scientists such as neurolinguists and psycholinguists. Pursuing this question should be tied with the question of the organization of the bilingual brain. Particularly, we should cooporate with those who pursue the question of whether the bilingual's two languages are mentally represented as a single system or as two separate systems. This call for close cooporation is legitamte since we are dealing with the theory of innateness, the biological endowment of language . Further research on the accessibility of 0G in adult L2 learning is extremely crucial. We should focus on many different languages so that we can have data from the whole spectrum which, in turn, will help in deciding the answer of the UG question and in finding common linguistic universals. Moreover, we should focus on more varied principles and parameters than those which have previously been examined. APPENDICES APPENDIX A Personal And Academic Questionnaire Native Language : .................................. Sex: Please circle one: MALE FEMALE Age of first exposure to English (for native speakers of Arabic only) or to Arabic (for native speakers of English Length of study of English (for native speakers of Arabic only) or Arabic (for native speakers of English only) : a. In English speaking countries: ---------------- b. In Arabic speaking countries: """"_' -------- Educational level: ------------------------------- Area of specialization: ------------------------------ 161 APPENDIX B Instruction Sheet There are two voluntary tests to do. The aim of these tests is to help with Doctoral Research into the acquisition of English and Arabic. You will 3192 be assessed in any way by these tests, and the results of the tests will be kept confidential . Please concentrate on grammar rather than meaning, spelling, and the like. When answering the tests, do not try to change your first answer. Take a break whenever you feel like it. PLEASEDONOT.CKAGRAMMARBOOK. DONOTWORKWITH ANYBODY . DO THE TESTS BY YOURSELF. TEST ONE Please read the following sentences. Put a J" in the parentheses next to any sentence that you think is GRAMMATICAL. Put an X in the parentheses next to any sentence that you think is UNGRAMMATICAL. Put a ? if you are NOT SURE. 162 163 If you think that a sentence is UNGRAMMATICAL, please correct it by writing the complete correct sentence in the space provided . TEST TWO Thank you for completing the information sheet and doing the first test! For this test, please make wh-questions (i.e. questions that start with what, w_here, w_hy, who, etc.) out of the following sentences, questioning only the underlined phrase. Write the questions in the provided spaces. EXAMPLE: She saw the car. ------- > What did she see? APPENDIX C English Test One 1. The house I rented two days ago will be sold next week. 2. I think that he saw Mary. ( ) 3. Who did you think that got married? ( ) ........................................................ 7 4 . I wonder when John bought these books. ( ) 5. The department chairman asked me whom I wanted to have as my main adviser. ( ) 6. Whom did the man have proof that John liked? ( ) ........................................................ 7 7. Wanted to run for re-election. ( ) 8. There is a good chance that you can pass the test next month. ( ) 165 9. What does Mary wonder who would steal? ( ) .................................................... 7 10. Who do you wonder when sold the house? ( ) 11. 12. 16. 18. .................................................... 7 Whom did the emloyee ask the manager where he would send?( ) .................................................... 7 She wondered where he would read the letter. ( ) John believes that his brother will be late. ( ) Spent $10 on his shirt. ( ) What did Susan reject the letter that had in it? ( ) .................................................... 7 We believe that John hit Sue. ( ) Who did Mary believe that would be late? ( ) .................................................... 7 There is a good opportunity that he can buy a new house today. ( ) 166 19. Who did you wonder what would sell? ( ) 20. .................................................... 7 Thought she that her boss would give her a payraise. The plan we discussed last week will be tested for efficiency. ( ) Whom did John make the claim that he hit last month? I wonder whom Sue will invite. ( ) 24. Whom did Sue tell you where she had seen? ( ) 25. .................................................... 7 He wondered where his wife spent the weekend. ( ) Whom does the president have evidence that the Senator trusted? ( ) .................................................... 7 Brought he too many toys for his kids. ( ) 167 28. Mary doesn't like dresses that have flowers on them. 29. What did the fact that you didn't buy surprise your wife? ( ) ........................................................ 7 30. Has built many houses in the past 10 years. ( ) 32. Whom did the Principal refuse the student that had criticized? ( ) ........................................................ 7 33. Knows Mary that her husband doesn't love her anymore. 34. Who does he believe that died? ( ) ........................................................ 7 35. He doesn't remember what he wrote. ( ) 36. Who did Alice wonder what would buy? ( ) ........................................................ 7 APPENDIX D Arabic Test One 1. Jaa'a al-walad allathi kataba al-qisata. came the-boy who wrote the-story “The boy who wrote the story came." 2. Sharah al-rajul al-hadath bi-?anna al-lisa saraq explained the-man the-incident that the burglar stole al -muj awharat . The-jewelry “The man explained that the burglar stole the jewelry." 3. Sa'ala al-mudir al-ustatha an matha kasar al-taleb. Asked the-principal the-teacher what broke the-student “The principal asked the teacher what the student broke . " 4. I9taqad muhammad ?anna khalid tazawaj al-fatat al- thought Muhamd that Khalid married the-girl the- sagira. Young “Muhamad thought that Khalid married the young girl ." 5. Sa-yahthuru. Will-come “Se will come." 6. Qatal zaid-un nafsah. Killed Zaid-Nom himself “Zaid killed himself." 7. *matha sa'ala omar salem man katab? What asked Omar Salem who wrote “*What did Omar ask Salem who wrote?" 8. Al-waled allathi harab min al-madina mat abu-h. The-boy who escaped from the-city died father-his “The father of the boy who fled the city died." 168 169 9. Sa'ala ali fatimah.mata jaa'at. Asked Ali Fatimah.when arrived-she “Ali asked Fatimah.when she arrived." 10. *matha ahaba al-rajul al-fatah allati qara'at? What liked the-man the-girl who read “*What did the man like the girl who read?" 11. *man saddaq salem al-hadath ?anna lyla saraqat? Who believed Salem.the-claim.that Lyla robbed “*Whom did Salem.believe the claim.that Lyla robbed?" 12. Thanna zaid ?anna muhammad baa9 al-siyyarah. Thought Zaid that Muhammad sold the-car “Zaid thought that Muhammad sold the car." 13. *man thanna al-rajul matha ?anna saraq? Who thought the-man.what that stole “*Who did the man think what that stole?" 14. Thahaba ila al-mataar. Went to the-airport “He went to the airport." 15. *man rafatha al-qathi al-iddi9a bi-?anna al-walad tharab? Whom rejected the-judge the-claim.that the-boy hit “*Whom did the judge reject the claim.that the boy hit?" 16 . Nama al-awlaad. Slept the-boys “The boys slept." 17. *man thanna muhammad ?anna ishtra al-siyyarah? Who thought Muhammad that bought the-car “*Who did Muhammad think that bought the car?" 18. Iktashaf rijaal al-shurtah.man saraq al-bank. Discovered men the-police who robbed the-bank “The policemen discovered who robbed the bank." 170 19.*man ra'aytu al-siyyarah allati thanan-tu ?ann-hu Who saw-I the-car Which thought-I that-he hatamaha? Destroyed-it “*Who did I see the car that I thought that-he destroyed?" 20. Sa'ala ali al-fatah.matha ahthara-t. Asked Ali the-girl what brought-she “Ali asked the girl what she brought." 21. *matha i9taqada ali man ?anna ishtara? What thought (Ali who that bought “*What did.Ali think who that bought?" 22. Ashaa9 muhammad al-khabar bi-?anna al-taleb kasar al- Spread.Muhammad the-news that the-student broke the- zujajeh. glass “Muhammad spread the word that the student broke the glass." 23. *man sa'al al-mudir al-taleb man tharab? Whom asked the-principal the-student who hit “*Whom did the principal ask the student who hit? 24. *man yathun ali ?anna ra'a al-bait? Who think Ali that saw the-house “*Who does Ali think that saw the house?" 25. Ahthar-tu al-kutub alti ista9ar-tu-ha.min-k. Drought-I the-books that borrowed-I-them.fromryou “I brought the books that I borrowed from.you.' 26. Jaa'a min al-safar. Came from the-travel “He came back from the trip." 27. *matha i9taqad al-walad ?anna inkasar? What thought the-boy that broke “*What did the boy think that broke?" 28. Al-mara'a naamat fi al-siyyara. The-woman slept in the-car “The woman slept in the car." 171 29. *matha tharabat al-mara'a al-walad allathi akal? What hit The-woman the-boy who ate “*What did the woman hit the boy who ate?" 30. Araf muhammad ayy kitaab qara'at. Knew Muhammad which book read-I “Muhammad knew which book I read." 31. *matha rafatha al-walad al-iddi9a ?anna al-fatah What rejected the-boy the-claim that the-girl hatamat? Destroyed “*What did the boy reject the claim that the girl destroyed?" 32. Sa'alat al-mara'a ayna thahabat al-qittah. Asked The-woman where went the-cat . “The woman asked where the cat went." 33. 9alima zaid ?anna al-walad mujtahid. Knew Zaidl that the-boy good “Zaid knew that the boy is good." 34.*matha haddatha zaid ali man ?anna qara'a? What told Zaid Ali who that read “*What did Zaid tell Ali who that read?" 35. Thamma omar al-iddi9a ?anna al-zawaj utiil al-umer. Criticized mar the-claim that the-marriage prolong the-life “Omar criticized the claim that marriage prolongs one’s life." 36. *matha sa'al muhammad fatimah ayna akala-t? What asked Muhammad Fatimah where ate-she “*What did Muhammad ask Fatimah where she ate?" APPENDIX E English Test Two 1. He went to Paris. ........................................................ 7 2 . The manager bought the dog that had brought the ring. ............_................1 .......................... 7 3. The fact that you didn't tell the truth proves your dishonesty. ........................................................ 7 4 . My brother wondered who saw E . ........................................................ 7 5. He gave me a present. ........................................................ 7 6. John has sold the cat that ate the rat. ........................................................ 7 7. The father has proof that his son bought the red house. ........................................................ 7 8. John wondered who would sell the diamond ring. ........................................................ 7 172 173 9. Paul thought that a virus entered the computer system yesterday. ........................................................ 7 10. £2 bought a new house. ........................................................ 7 11. Tom suspected that Lisa liked John. ........................................................ 7 APPENDIX F Arabic Test Two 1. Thahab mihamad ila al-suq. Went Muhammad to the-mall “Muhamad went to the mall." 2. Ra'a zaid al-fatah alti ishtarat al-saa9a. Saw Zaid the-girl who Bought the-watch “Zaid saw the girl who bought the watch." 3. Qaddam al-a'ab al-ddalil ?anna al-rajul kasar Presented the-father the-evidence that the-man broke al-bab. the-door “The father presented the evidence that the man broke the door." 4. Sa'al al-ustaath al-taleb man akhath al-qalam. Asked the-teacher the-student who took the-pen “The teacher asked the student who took the pen." 5. A9daa-ni al-ustaath hadiyyah. Gave-me the-teacher present “The teacher gave me a present." 6. Baa9 zaid al-qittah allati akalat al-samakah. Sold Zaid the cat that ate the fish “Zaid sold the cat that ate the fish." 7. Saddaq ali al-khabar ?anna muhammad saraq al-tufaha. Accepted Ali the-news that Muhammad stole the-apple “Ali accepted the news that Muhammad stole the apple." 8. Sa'al al-muzar9 al-rajul man sa-uhthiru al-baqara. Asked the-farmer the-man who would-bring the-cow “The farmer asked the man who would bring the cow." 174 175 9. Thannat hind ?anna al-rajul ishtara al-khatem al- Thought Hind that the—man bought the-ring the- thahabi. Golden “Hind thought that the man bought the golden ring." 10. Ishtara1muhammad siyyarah jadidah. Bought Muhammad car new “Muhammad bought a new car." 11. I9taqad zaid ?anna at-tifla sharab al-haleeb. Thought Zaid that the-child drank the-milk “Zaid thought that the child drank the milk." APPENDIX C Frequency Distribution Of Responses (G1) TEST ONE 176 177 TEST TWO Item Grammar R#1 R#2 R#3 R#4 R#S R#6 R#7 Total Rs 1 9 4 9 9 9 4 6 7 6 U'IU'IU'IU1U1U'IUTU'IU'IUl APPENDIX H Frequency Distribution Of Responses (G2) GHHTT can: . #5 Total 84 . 83 R. 1. 0 Item Grammar Status 00000000000 00000000000 37629114942 60915256468 01234567890123456789012345 178 179 TEST TWO It Grammar R#1 R#2 R#3 R#4 R#5 R#6 R#7 Total Rs 1 l 54 5 O O O O l 60 2 23 O 5 O O 0 32 6O 3 9 0 0 O O O 51 60 4 8 O O 7 0 5 60 l 43 6 O O O 0 11 6O 6 2 24 l 0 O D S 60 3 l6 0 0 7 O D 37 60 8 4 15 0 O 0 5 4O 60 5 22 O O 0 O 5 33 60 10 1 42 B O 0 O 0 16 6O APPENDIX I Frequency Distribution Of Responses (G3) TEST ONE Item Grammar Status . #1 R. #2 R. #3 . #4 . .5 Total Rs NNNNNNNMNNNNNNNNNNNNNNNNNN 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 O 1 2 3 4 5 180 181 TEST TWO Grammar Rtl R112 R#3 R#4 R115 R#6 R#7 Total It RS 0.2 12 12 12 E12 12 12 12 12 £12 12 12 11 12 12 12 11 12 11 10 APPENDIX J Frequency Distribution Of Responses (G4) TEST ONE Item Grammar Status R. #1 R. #2 R. #3 . 44 R. 85 Total Rs 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 182 183 TEST TWO Item Grammar R41 R42 R83 R84 R85 R46 R87 Total Rs .KEY fer Test One Tables Grammar: 1.RC;2.NC;3.EQl;4.TR-T;5.EQ2;6.SO;7.SVI. Status: 1.Correct;2.Incorrect. R (esponse) : 1 . Correct ; 2 . Incorrect with right correction; 3.Incorrect without correction; 4.Incorrect with wrong correction; 5.Not sure. KEY'for Test TWo Tables Grammar:1.Simple;2.RC;3.NC;4.EQl;5.TR-T. R(esponse):1.Correct;2.Incorrect Simple; 3.Incorrect RC; 4.Incorrect NC;5.Incorrect E01; 6.Incorrect TH-T;7.0ther. BIBLIOGRAPHY BIBLIOGRAPHY Al-Ani, Salman B (ed.) 1978. Readings in Arabic Linguistics. Bloomington, IN: Indiana University Linguistics Club. Al-Rajahi, Abdu. 1988. Al-Tatbeeq Al-Nnahawi. Beirut, Lebanon: Dar AlNahthah AlArabeeah Publishing Co. Aoun, Youssef. 1981. “ECP, move a, and Subjacency." In Linguistic Inquiry, 12, 4:637-645. Aoun, Youssef. 1982. “Expletive pros." In Papers in Syntax, Alec Marantz and Tim Stowell (eds.). Cambridge, Mass: “To Ppe 1-‘e Aoun, Joseph. 1994. “Agreement, word order, and conjunction in some varieties of Arabic." In Linguistic Inquifl, 25, 2:195-220. Arbib and J. C. Bill. 1988. “Language acquisition: schemas replace universal granular." In Elaining Language Universals, John A. Hawkins (ed.), New York: Basil Blackwell. Pp. 61-80. Arther, B. 1980. “Gauging the boundaries of second language competence: A study of learner judgements." In Language Learning, 30:177-94. Bailey, K., M. Long, and S. Peck. 1983. Second Language Acquisition Studies. Mass: Newbury Rouse. Bailey, N., C Madden, and S. Krashen. 1974. “Is there a 'Natural Sequence' in adult second language learning?" In Language Learning, 24: 235-243. 184 185 Bakir, Murthadha. 1990. “Notes on Subjacency as a syntactic constraint in Arabic and English." In Further Insights into Contrastive Analysis, Jacek Fisiak (ed.). Amsterdam; John Benjamins Publishing Co. Pp. 329-355. Berke, J. 1958. “The child's learning English.morphology;" In Birdsong, David. 1990. “universal grammar and second language acquisition theory." In. Studies in Second Language Acquisition, 12: 331-340. Bl.y-vrm' Re We 3 8e We P.118, and Ge Le Ioupe 1988a “The accessibility' of universal grammar in. adult language learning . " In Second Language Research, 4 : 1-32 e Bleybvroman, R. 1989. “What is the logical problem of foreign language learning." In Linguistic Perspectives in Second Language Acquisition, S. Gass and J. Schachter, (eds.), New York: Cambridge university Press, pp. 41-68. Brown, R. 1973. A First Language: The Early Stages. Cambridge, Mass: Barvard university Press. Butterworth, Brian, Bernard Comrie, and O. Dahl. 1984. Explanations for Language universals. Berlin: walter de Gruyter & Co. Carroll, 8. and Meisel, J. M. 1990. “universals and second language acquisition: some comments on the state of current theory." In Studies in Second Language Acquisition, 12: 201-208. Chomsky, Carol. 1969. The Acquisition ofSyntax in Children from 5 to 10. Cambridge, Mass: MIT Press. Chomsky, N; 1965..A£pects of the Theory of Syntax. Cambridge, Mass: MIT Press. Chomsky, N; 1975. Reflections on language. New York: Pantheon Press. Chomsky; N. 1981a. Lectures on. Government and. Binding. Dordrecht: Foris. 186 Chomsky, N. 1981b. “Principles and parameters in syntactic theory." In Elastion in Linguistics:The Logical Problem Of Language Acquisition, N. Hornstein and D. Lightfood, (eds.). London: Longman, pp. 32-75. Chomsky, N. 1986a. Knowledge of Language: Its Nature, Origin and Use. New York: Praeger. Chomsky, N. 1986b. Barriers. Cambridge, Mass: MIT Press. Chomsky, N. 1988. Language and problems of knowledge. Cambridge, Mass: MIT Press. Clahsen, B. 1988. “parameterized grammatical theory and language acquisition: a study of the acquisition of verb placement and inflection by children and adults." In L_i_n_guistic TheoLy in Second Language Acquisition, S. Flynn and W. O'Neil (eds.). Boston: Kluwer Academic Publishers, pp. 47-75. Clahsen, B. 1990. “The Comparative study of first and second language acquisition." In Studies in Second Language Acquisition, 12: 135-153. Clahsen, B. and P. Muysken. 1986. “The availability of universal grammar to adult and child learners: a study of the acquisition of German word order." Second Language Research, 2: 93-119. Clahsen, B. and P. Muysken. 1989. “The UG paradox in L2 acquisition." In Second Language Research, 5: 1-29. Comrie, B. 1981. Language universals and Linguistic Typology. Oxford: Basil Blackwell. Comrie, B. 1990. “Second language acquisition and language universals research." In Studies in Second Language Acquisition, 12:209-218. Cook, Vivian J. 1973. “The comparison of language development in native children and foreign adults." In International Review of Ap_plied Linguistics, x1:1. Cook, v. J. 1985. “Chomsky's Universal grammar in second language learning.” In A_p_plied Linguistics, 6 :2-18. 187 Cook, V. J. 1988. Chomsky's Universal Grammar: An Introduction. Oxford: Basic Blackwell. Corder, S. P. 1967. “The significance of learners' errors." In Applied Linguistics, 5:161-170. De Villiers, J. and P. De Villiers. 1973. “A cross-sectional study of the acquisition of grammatical morphemes in child speech." In Journal of Psycholinguistic Research, 2:267-278. Dulay, 3., and M. Burt. 1974a. “Natural sequences in child second language acquisition." In Language Learning, 24: 37-53. Dulay, B., and M. Burt. 1974b. “A new perspective on the creative construction processes in child second language acquisition." In Language Learning, 24:253- 258. DuPlessis, Jean, D. Solin, L. Travis, and L. White. 1987. “UG or not UG, that is the question: A reply to Clahsen and Muysken." In Second Language Research, 3:56-75. Eckman, F. 1988. “Typological and Parametric views of universals in second language acquisition." In Linguistic Theory in Second Language acquisition, S. Flynn and W. O'Neil (eds.). Boston: Kluwer Academic Publishers. Pp. 417-29. Eckman, F. and A. Hastings, (eds.). 1979. Studies in First and Second Language Acquisition. Mass: Newbury Rouse. Eckman, F., L. Bell, and D. Nelson (eds.). 1984. Universals of Second Language Acquisition. Rowley, Mass: Newbury House. Eid, Mushira and Clive Boles, (eds.). 1993. Persflctives on Arabic Linguistics V. Amsterdam: John Benjamins Publishing Co. Ellis, Rod. 1994. The Study of Second Language Acquisition. Oxford: Oxford University Press. 188 Ervin, S. And C. Osgood. 1965. “Second language learning and bilingualism." In C. Osgood and T. Sebeok (eds.), Psycholinguistics: A Survey Of Theory And Research Problems. Bloomington, IN: Indiana University Press. Eubank, E. (ed.) . 1991. Point Counterpoint: Universal Gramar in the Second Language. Amsterdam: John Benjamins. Felix, S. 1985. “More evidence on competing cognitive systems." In Second Language Research, 1: 47-72. Felix, S. 1988. “UG-generated knowledge in adult second language acquisition." In Linguistic Theory in Second Language Acquisition, S. Flynn and W. O'Neil (eds.). Boston: Kluwer Academic Publishers. Pp. 277- 94. Flynn, S. 1984. “A universal in L2 acquisition based on a PBD typology." In Universals of Second Language Acquisition. Rowley, Mass: Newbury Bouse. Pp. 75-87. Flynn, S. 1987. A Parameter-setting Model of L2 Achuisition: Experimental Studies in Anaphora. Dordrecht: D. Reidel Publishing Co. Flynn, S. 1988. “Nature of development in L2 acquisition and implications for theories of language acquisition in general." In Linguistic Theory in Second Langua_g_e Acquisition, S. Flynn and W. O'Neil (eds.). Boston: Kluwer Academic Publishers. Pp. 76-89. Flynn, S. 1991. “Government-binding: parameter setting in second language acquisition." In Crosscurrents in Second Language Acquisition and Linguistic Theories, Thom Buebner and Charles A. Ferguson (eds.) . Amsterdam: John Benjamins. Pp. 143-167. Flynn, S. and W. O'Neil, (eds.) 1988. Linguistic Theory in Second Language Acquisition. Boston: Kluwer Academic Publishers. Flynn, S. and Sharon Manuel. 1991. “Age-dependent effects in language acquisition: an evaluation of “critical period" hypotheses." In Point Counterpoint: Universal Grammar in the Second Language. Amsterim: John Benjamins. Pp. 117-145. 189 Gass, S. 1989. “Language universals and second-language acquisition." In Language Learning, 29(4): 497-534. Gass, S. and J. Schachter (eds.). 1989. Linguistic Perspectives on Second Language Acquisition. New York: Cambridge University Press. Gass, S. and L. Selinker. 1994. Second Language Acquisition. Hillsdale, New Jersey: Lawrence Erlbaum Associates Pub. Goodluck, Helen. 1991. Language Acquisition: A Linguistic Introduction. Cambridge, Mass: Blackwell. Greenberg, H. 1966. Universals of Language (2nd ed.). Cambridge, Mass: MIT Press. Hanania, E. and H. Gradman. 1977. “Acquisition of English structures: a case study of an adult-native speaker of Arabic in an English-speaking environment." In Language Learning, 27:75-91. Hatch, Evelyn and H. Farhady. 1982. Research Design and Statistics for Aguied Linguistics. Rowley: Newbury House Publishers, INC. Heckler, E. 1975. The Acquisition of English Verb Morphology fl Non-native speakers. Ph.D. dissertation, Michigan State University. Hilles, S. 1986. “Interlanguage and the pro-drop parameter." Second Language Research, 2(1) : 33-52. Hoekstra, Teun and Bonnie D. Schwartz. 1994. Language Acquisition Studies in Generative Grammar. Philadelphia, PA: John Benj amins Publishing Company. Hudson, Grover. 1995. Personal Communication. MSU. Huebner, T. And C. Ferguson (eds.). 1991. Crosscurrents in Second Language Acquisition and Linflistic Theories. Amsterdam: John Benjamins Publishing Co. Hyams, Nina. 1986. Language Acquisition and the Theory of Parameters . Dordrecht : Reidel . 190 Hyams, Nina. 1989. “The null subject parameter in language acquisition." in The Null Subject Parameter, O. Jaeggli and K. Safir (eds.). Dordrecht: Kluwer. Pp. 215-238. Jaeggli, O. 1982. Topics in Romance Syntax. Dordrecht: Foris. Jaeggli, O. and k. Safir. 1989. “The null subject parameter and parametric theory." In The Null Subiject Parameter, O. Jaeggli and K. Safir (eds.). Dordrecht: Kluwer. Pp. 1- 44. Jensen, John. 1990. Morphology: Word Structure in Generative Grammar. Amsterdam: John Benjamins Publishing Co. Johnson, J. and E. Newport. 1989. “Critical period effects in second language learning: the influence of maturational state on the acquisition of English as a second language." In Cignitive Psychology, 21:60-99. Keenan, E. and B. Comrie. 1977. “Noun phrase accessibility and universal gramar." In Linguistic InqL, 8:63-100. Kellerman, E. 1986. “An eye for an eye: crosslinguistic constraints on the development of the L2 lexicon." In Crosslinguistic influence in Second Langu__g__ge Acquisition, E. Kellerman and M. Sharwood Smith (eds. ). Oxford: Pergamon. Pp. 35-48. Kenstowicz, Michael. 1989. “The null subject parameter in modern Arabic dialects." In The Null Subject Parameter, O. Jaeggli and K. Safir (eds.) . Dordrecht: Kluwer. Pp. 263-276. Klein, Wolfgang. 1991. “SLA theory: prolegomena to a theory of language acquisition and implications for theoretical linguistics." In Crosscurrents in Second Language A_cauisition and Limistic Theories, T. Huebner and C. Ferguson (eds.) . Amsterdam: John Benjamins Publishing Co. Pp. 169-194. Krashen, S. 1982. Principles and Practice in Second Language Acquisition. University of Southern California: Pergamon Press. 191 Lakshmanan, U. 1986. “The role of parametric variation in adult second language acquisition: A study of the “pro-drop" parameter." P_apers in Applied Linguistics- Michigan, Ann Arbor. Pp. 97-118. Larsen-Freeman, D. 1976. “An explanation for the morpheme acquisition order of second language learners." In Language Learning, 26:125-134. Larsen-Freeman, D. 1991. “Second language acquisition research: staking out the territory." In TESOL Quarterly, 25: 315-350. Larsen-Freeman, D. and M. Long. 1991. An Introduction to Second Language Research . London: Longman. Lenneberg, E. H. 1967. Biological Formations of Language. New York: Wiley. Lewkowicz, Nancy K. 1978. “Topic-Conant and relative clause in Arabic." In Readings in Arabic Linguistics, Salman Al- Ani (ed.). Bloomington, IN: University Linguistics Club. Pp. 553-575. Liceras, Juana M. 1989. “On some properties of the “pro- drop" parameter: looking for missing subjects in non- native Spanish." In Linguistic Perspectives in Second Language Acquisition, S. Gass and J. Schachter (eds.). New York: Cambridge University Press. Pp. 109-133. Lust, Barbara. 1988. “Universal grammar in second language acquisition: promises and problmns in critically relating theory and empirical studies." In Linguistic Theory in Second Language Acquisition, S. Flynn and W. O'Neil (eds.) Boston: Kluwer Academic Publishers. Pp. 309-328. McLaughlin, B. 1987. Theories of Second Language Learning. London: Edward Arnold. Meisel, J. 1991. “Principles of universal gramr and strategies of language learning: some similarities and differences between first and second language acquisition." In Point Counterpoint: Universal Grammar in the Second Language, L. Eubank (ed.). Amsterdam: John Benjamins. Pp. 231-276. 192 Meisel, J., H. Clahsen, and M. Pienemann. 1981. “On determining developmental stages in natural second language acquisition." In Studies in Second Language Acquisition, 3:109-35. O'Grady, W. , M. Dobrovolsky and M. Aronoff. 1989. Contemporary Linguistics. New York: St. Martin's press. Owens, J. 1988. The Foundation of Grammar: An Introduction to Medieval Arabic Grammatical Theory. Amsterdam: John Benjamins Publishing Company. Preston, Dennis. 1989. Sociolinguistics and Second Language acquisition. Oxford: Blackwell. Ritchie, W.C. 1991. “Linguistic theory: contributions to second language acquisition." In Studies in Second Language Acquisition, 13: 77-85. Rizzi, L. 1982. Issues in Italian Syntax. Dordrecht: Foris. Rutherford, W. (ed.) 1984. Language Universals and Second Language Acquisition. Amsterdam: John Benjamins Publishing Co. Rutherford, W. 1987. Second Language Grammar: Learning and Teaching. London: Longman. Rutherford, W. 1988. “Grammatical theory and L2 acquisition: a brief overview." In Linguistic Theogy in Second Language Acquisition, S. Flynn and W. O'Neil (eds.). Boston: Kluwer Academic Publishers. Pp. 404-416. Sabra, S. M. 1987. The Acquisition of English Syntax by Four Arabic-speaking Children. Ph.D dissertation, Michigan State University. Schachter, J. 1988. “Second language acquisition and its relationship to universal grammar." In Applied Linguistics, 9: 219-235. Schachter, J. 1989. “Testing a proposed universal." In Linguistic Perspectives in Second Language Acquisition, S. Gass and J. Schachter, (eds.) . New York: Cambridge University Press. Pp. 73-88. 193 Schwartz, Bonnie. 1991. “Conceptual and empirical evidence: a response to Meisel." In Point Counterpoint: Universal Grammar in the Second Language, L. Eubank (ed. ) . Amsterdam: John Benjamins Pub. Co. Pp. 277-304. Singleton, David. 1989. Language Acquisition: The Ag} Factor. Philadelphia: Multilingual Matters Ltd. Taylor, B. 1974. “Toward a theory of language acquisition." In Language Learning, 24: 23-35. Thomas, Margaret. 1991. “Universal grammar and the interpretation of reflexives in a second language." In Language, 67(2):211-239. Thornton, Rosalind and Stephen Crain. 1994. “Successful cyclic movement." In Language Acquisition Studies in Generative Grammar, T. Hoekstra and B. Schwartz, (eds.). Philadelphia, PA: John Benjamins Publishing m. Pp. 215-252. Vainikka, Anne and Martha Young-Schotten. 1994. “Direct access to K'-theory: evidence from Korean and Turkish Adults Learning German." In Language Acquisition Studies in Generative Grammar, T. Hoekstra and B. Schwartz, (eds.). Philadelphia, PA: John Benjamins Publishing Company. Pp. 265-316 . Wahba, Wafaa A. B. 1991. “LF movement in Iraqi Arabic." In Leggical Structure and Linguistic Structure, C.-T. James Huang and Robert May, (eds) . Boston: Kluwer Academic Pub. Pp. 253-276. White, L. 1985a. “The acquisition-of parameterized grammars: Subjacency in second language acquisition. In Second Language Research, 1 : 1-17 . White, L. 1985b. “The Pro-drop parameter in adult second language acquisition." Language Learning, 35: 47-62. White, L. 1988. “Island effects in second language acquisition." In L_inguistic Theory in Second Lang_u__age Acquisition, S. Flynn and W. O'Neil (eds.). Boston: Kluwer Academic Publishers. Pp. 144-172. 194 White, L. 1989a. Universal Grammar and Second Language Acquisition. Amsterdam: J. Benjamins Publishing Co. White, L. 1989b. “Linguistic universals, markedness and learnability: Comparing two different approaches." In Second Language Research, 5: 127-141. White, L. 1990. “Second language acquisition and universal grammar." In Studies in Second Language Acquisition, 12:121-133. White, L., Lisa Travis, and Anna Maclachlan. 1992. “The acquisition of wh-question formation by Malagasy learners of English: evidence for universal grammar." In Canadian JOurnal of Linguistics, 37(3):341-368. Woolford, Ellen. 1982. “Island constraints, Subjacency, and ECP." In Pepers in Syntax, Alec Marantz and Tim Stowell (eds.). Cambridge, Mass: MIT. Pp. 271-291.