HERITAGE AND L2 LEARNERS ACQUISITION OF KOREAN          IN TERMS OF IMPLICIT AND EXPLICIT KNOWLEDGE By Yeon Heo             A DISSERTATION Submitted to  Michigan State University in partial fulfillment of the requirements  for the degree of Second Language Studies  Doctor of Philosophy  2016       ABSTRACT                                                    HERITAGE AND L2 LEARNERS ACQUISITION OF KOREAN           IN TERMS OF IMPLICIT AND EXPLICIT KNOWLEDGE By Yeon Heo Recent research has demonstrated that L2 learners and heritage speakers might have two different profiles in terms of their syntactic, morpho-syntactic, and lexical-semantic knowledge of a target language (Montrul, 2004, 2006, 2010; Montrul, Foote, & Perpinan, 2008). However, the most influential of the major studies have been limited to Spanish. To expand our understandings of the incompleteness of heritage and L2 language learning from implicit and explicit knowledge perspectives (Bowles, 2011), this study investigated the differences and/or similarities between two learner groups of Korean in their knowledge of 17 target structures. One hundred fourteen learners were recruited in three groups: 65 L2 learners, 38 heritage language (HL) learners, and 11 native speakers of Korean. This research utilized an elicited imitation test (EIT), a narrative test, and an untimed aural Grammaticality Judgment Test (GJT) to measure implicit knowledge, and a metalinguistic knowledge test, a written untimed GJT, and a C-test to gauge explicit knowledge of the 17 Korean target structures.  The results of a Confirmatory Factor Analysis (CFA) demonstrated that the EIT and the aural GJT significantly explained the participants implicit knowledge of the target structures, whereas the written GJT, metalinguistic test, and C-test accounted for their explicit knowledge of the grammar topics. Further analyses showed that HL learners outperformed L2 learners when the scores of the EIT and aural GJT were combined,    highlighting HL learners superior implicit knowledge of the target structures. Regarding explicit knowledge of the target structures, L2 learners showed significantly better performances in the combined scores of the written GJT, metalinguistic knowledge test, and C-test. However, the L2 and HL groups did not demonstrate significant differences for the aural and written GJTs. Overall, the results of this study showed that 1) the HL and L2 learners linguistic knowledge of Korean have significantly different knowledge profiles in terms of the implicit-explicit dichotomy of knowledge, and 2) in addition to a time constraint and grammaticality, manipulating testing modeaural or writtencould be another valid factor to measure implicit and explicit knowledge relatively separately.   Keywords: implicit knowledge, explicit knowledge, heritage language learners, Korean, aural GJT                      Copyright by YEON HEO 2016   v                  To HIM                        vi ACKNOWLEDGEMENTS  I would like to express my sincere gratitude to Shawn Loewen, my chair, for his academic guidance, generosity, encouragement, and patience, throughout the whole dissertation process. He was always there when I needed his guidance and he made himself available to me even during vacations. He even gave me an opportunity to experience the togetherness of an American holiday by opening his home to his students. Without his considerations, this dissertation would not have been possible.   My sincere gratitude also goes to my committee members. Paula Winke has been my great role model, as she is a mother of two children and has achieved what she wants in her career. She showed me that scholars can have inner warmth in addition to being outwardly professional. I hope to, someday, be a researcher like her. Aline Godfroid has also been a great inspiration for my research. She has provided invaluable insights on major topics and helped me go one-step further whenever I felt like compromising. I wonder how on earth she sharpens her academtic agility. Ok-sook Park has been a great help concerning her knowledge of Korean and tremendous help with data collection. She introduced me to her Korean classes and even gave me an opportunity to work with them as a language facilitator. This experience will be the backbone of my teaching in the U.S. I thank my family members: my husband, Won-tae Choo; my only son, Seung-beom (Steve) Choo; my only daughter, Seung-ah (Monica) Choo; my late father, Koo Heo; my mother, Young-hae Lee; my older sister, So-mee Heo and her husband, Seung-keun Lee; my younger sister, Kyeong Heo and her husband, Yong-dap Seong; and my  vii younger brother, Jun Heo and his wife, Min-jung Park. Without their prayers, patience, and love, I would not have had the inner strength I needed to finish this dissertation.  Finally, the same thanks go to those who agreed to participate in this study, whose names I hardly remember now, but whose warm hearts and various contributions have helped inform the bulk of this dissertation. There are also indirect participants in my study who helped me recruit the participants. Professor Young-hwa Hong in University of California, Riverside, and Professor Jung-hee Kim in the University of Virginia helped me get access to their students and obtain data for this study. Their generosity will always remain in my heart. Before I finish, there is one more person I want to thank: Ms. Kui-ja Park, who has been instrumental in helping me recruit participants in her Korean community. She also prepared dinner for me several times, and encouraged me to focus on my dissertation while I was staying in California for data collection. I consider this is my turn to encourage her to continue the good fight against cancer and maintain focus on our Lord.  viii TABLE OF CONTENTS   LIST OF TABLES...x LIST OF FIGURES....xii CHAPTER 1. LITERATURE REVIEW.....1 1.1. Introduction..1 1.2. Implicit and explicit learning and knowledge...2 1.3. Measurement of implicit and explicit knowledge...11 1.4. Language learners and two types of learning and knowledge........25 1.5. Studies on heritage and L2 learners linguistic knowledge28 1.6. Research questions33  CHAPTER 2. METHODOLOGY34 2.1. Target structures.34 2.2. Pilot study I42  2.2.1. Methodology .. 42   2.2.1.1. Participants .. 42   2.2.1.2. Tasks ...42   2.2.1.3. Bio-data questionnaire . 43   2.2.1.4. Procedure ..43  2.2.2. Results .. 44  2.2.3. Suggestions .. 47 2.3. Pilot study II...48  2.3.1. Methodology ... 49   2.3.1.1. Participants ...49   2.3.1.2. Tasks ...50   2.3.1.3. Bio-data questionnaire ..51   2.3.1.4. Proficiency test . 51   2.3.1.5. Procedure .. 52  2.3.2. Results ...52  2.3.3. Suggestions .. 53 2.4. Main study..54  2.4.1. Methodology .. 54   2.4.1.1. Participants .. 54   2.4.1.2. Tasks .....55   2.4.1.3. Bio-data questionnaire ... 56   2.4.1.4. Procedure .... 57   2.4.1.5. Analyses .. 57  CHAPTER 3. RESULTS...........60 3.1. Descriptive statistics..60 3.2. Construct validity of the tests61  3.2.1. Two-factor model with four observed variables.62   3.2.1.1. Defaul model  62   3.2.1.2. Rival model 1: written ungrammatical GJT 64  ix  3.2.2. One-factor model with four observed variables .65   3.2.2.1. Default model .65  3.2.3. Two-factor model with five observed variables .68  3.2.4. Two-factor model with six observed variables71 3.3. Implicit knowledge80 3.4. Explicit knowledge81  CHAPTER 4. DISCUSSIONS.86 4.1. Research question 1: Measurements...........86  4.1.1. Valid measurements ...86   4.1.1.1. Aural GJT ..88   4.1.1.2. Written GJT ..92   4.1.1.3. C-test ..94  4.1.2. Models .....96 4.2. Research questions 2 and 3: Similarities and differences of the two groups...........99  4.2.1. Various learners ....99   4.2.1.1. Input and age .101   4.2.1.2. Orality and literacy .103   4.2.1.3. Learner variances ..106   4.2.1.4. C-test for various types of learners .108 4.3. Pedagogical implications  CHAPTER 5. LIMITATIONS.113  CHAPTER 6. FUTURE STUDIES.115  APPENDICES.................118 APPENDIX A. Stimuli (Pilot study I)....119 APPENDIX B. 129 APPENDIX C. Item total statistics (Pilot study I).151 APPENDIX D. .153 APPENDIX E. Summary o155  REFERENCES....158            x LIST OF TABLES  Table 1. Target structures in the tests...37 Table 2. Reliability coefficients for the four tests...44 Table 3. Accuracy scores for the four tests by group...45 Table 4. Results of one-way ANOVAs45 Table 5. Correlational matrix for the tests...46 Table 6. Principal component factor analysis..46 Table 7. Loadings for principal component factor analysis.47 Table 8. Reliability coefficients for the four tests...52 Table 9. Accuracy scores for the five tests by 10 participants.53 Table 10. Comparisons between Pilot studies I and II....53 Table 11. Reliability coefficients for the five tests..58 Table 12. Descriptive statistics for the five tests by group..60 Table 13. Results of one-way ANOVA on the proficiency test...61 Table 14. Correlational matrix for the six tests...61 Table 15. Modification indices from the one-factor model.65 Table 16. Summary of the important indices from the major CFAs74 Table 17. Chi-squared difference test results for models with overall good model fit indices..................75 Table 18. Factor loadings of models with good model fit indices (4 observed variables)...........................................................................................................76 Table 19. Factor loadings of models with good model fit indices (6 observed variables)..78 Table 20. Factor loadings of models with good model fit indices (5 observed variables with C-test)......78  xi Table 21. Results of one-way ANOVAs on test scores for implicit knowledge..81 Table 22. Results of one-way ANOVAs on test scores for explicit knowledge .82 Table 23. Summary of discriminant functions ........83 Table 24. Standardized canonical discriminant function coefficients ....83 Table 25. Functions at group centroids .......84 Table 26. Five tests and criteria to separate implicit and explicit knowledge ....96 Table 27. The input amounts by groups (%)......100 Table 28. Implicit and explicit knowledge and relevant measurements....117 Table 29. Item total statistics for each test focusing on cronbachs alphas if item deleted (Pilot study I) ...151 Table 30. Item total statistics for each test focusing on cronbachs alphas if item deleted (Pilot study I1) ..153 Table 31. Summary of the important indices from all CFAs .   xii LIST OF FIGURES   Figure 1. 2 Factor 4 observed variables (OVs), default model: Implicit-explicit model with written GJT......63 Figure 2. 2 Factor 4 OVs Rival model 1: Implicit-explicit model with ungrammatical items in written GJT....64 Figure 3. 1 Factor 4 OVs Rival 1: One-factor model of explicit knowledge with one set of covaried error terms between imitation test and aural GJT....66 Figure 4. 1 Factor 4 OVs Rival 2: One-factor model of explicit knowledge with one set of covaried error terms between written GJT and metalinguistic test.....67 Figure 5. 2 Factor 5 OVs C-test Rival model 1: Implicit vs. explicit model with two sets of covaried error terms....68 Figure 6. 2 Factor 5 OVs C-test Rival model 2: Implicit vs. explicit model with ungrammatical items in written GJT ..69 Figure 7. 2 Factor 5 OVs C-test Rival model 3: Implicit vs. explicit model with ungrammatical items in written GJT with two covaried error terms between written GJT ungrammatical items and metalinguistic test, metalinguistic test and C-test.70 Figure 8. 2 Factor 6 OVs Rival model 2: Implicit vs. explicit model with ungrammatical items in written GJT, three sets of covaried items ......71 Figure 9. 2 Factor 6 OVs Rival model 6: Implicit (orality+grammatical) vs. explicit (literacy+ungrammatical) model ........72 Figure 10. 2 Factor 6 OVs Rival model 6: Implicit (orality+grammatical) vs. explicit (literacy+ungrammatical) model with ungrammatical items in writtenGJT, one set of covaried error terms .....73 Figure 11. Canonical discriminant functions ..........85   1 CHAPTER 1. LITERATURE REVIEW  1.1. Introduction The topics of implict and explicit knowledge have received attention from researchers in Second Language Acquisition (SLA) in various areas. How to measure them has been an important topic, and how the two types of knowledge representations have been varied according to learner groupsi.e., HL learners and L2 learnershas been another important topic. In addition, for understanding the construct of L2 learners linguistic knowledge, a two-factor model for implicit and explicit knowledge, and a one-factor model for automatized explicit knowledge have been studied by researchers. However, the jury is still out there concerning which model is more accurate. For the first issue of various measures, this study will discuss one additional elementmodalityas another potential criteria, and one additional test, C-test, as a potential measure of explicit knowledge. For the second topic of varied learner groups and implicit and explicit knowledge, this study covers a HL learner group of Korean to present a comprehensive picture about implicit and explicit learning and knowledge of various learners. Based on the data set of the current study, a two-factor model seems to obtain supports than a one-factor model concerning the issue of the valid model of L2 learners knowledge. For pedagogical implications, this study will direct attention to similarities and differences between HL and L2 learner groups.   2 1.2. Implicit and explicit learning and knowledge Implicit and explicit learning and knowledge have been investigated in many studies (DeKeyser, 2003; DÖrnyei, 2009; N. Ellis, 1994; R. Ellis, 2004, 2005; R. Ellis, Loewen, Elder, Erlam, Philp, & Reinders, 2009; Hulstijn, 2005; Godfroid, Loewen, Jung, Park, & Gass, 2015; Reber, 1993; Rebuschat & Williams, 2009) to investigate the nature and processing of language learning, the relationship between implicit/explicit distinction, and valid measures of implicit and explicit knowledge. Reber defined implicit learning as the acquisition of knowledge that takes place largely independently of conscious attempts to learn and largely in the absence of explicit knowledge about what was acquired (Reber, 1993, p. 5). N. Ellis also emphasized that implicit learning involves unconsciousness as acquisition of knowledge about the underlying structure of a complex stimulus environment by a process which takes place naturally, simply and without conscious operations (N. Ellis, 1994, p. 1). He also suggested that implicit knowledge attainment can thus take place implicitly (a nonconscious and automatic abstraction of the structural nature of the material arrived at from experience of instances) (N. Ellis, 1994, p. 1). Dörnyei (2009) and R. Ellis (2009) have also supported this idea.  In comparison, explicit learning is defined as a more conscious operation where the individual makes and tests hypotheses in a search for structure (N. Ellis, 1994, p.1), and explicit knowledge is attained explicitly through selective learning (the  3 learner searching for information and testing hypotheses), or because we can communicate using language, explicitly via given rules (assimilation of a rule following explicit instruction) (N. Ellis, 1994, pp. 1-2). R. Ellis (2004) also suggested a working definition of explicit knowledge as The conscious awareness of what a language or language in general consists of and/or of the roles that it plays in human life (p. 229). In the two definitions, it is commonly suggested that explicit learning involves consciousness and awareness. N. Ellis (1994), Dorynei (2009), and R. Ellis (2009) agreed that explicit knowledge is obtained from explicit learning. In terms of implicit and explicit learning and knowledge, several characteristics have been investigated in previous studies. In terms of primacy and a default mechanism, implicit knowledge appears first and explicit knowledge later. Initially, L1 speakers acquire their language intuitively and utilize implicit knowledge to tell what is possible and impossible linguistically. At the age of around five, they are able to analyze L1 and develop explicit knowledge, especially metalinguistic knowledge, using conscious awareness about grammaticality of language structures through making everything clearly and explicitly (Karmiloff-Smith, 1979, p. 115). However, L1 speakers implicit knowledge from implicit learning is primary and the default system (Reber, 1993) and L1 speakers tap into implicit knowledge for spontaneous production (R. Ellis, 2009). Like L1 speakers, L2 learners default L2 production depends on implicit knowledge, but difficulty in performing an on-line task may cause them to  4 utilize explicit knowledge when they lack enough implicit knowledge (R. Ellis, Loewen et al., 2009).  Concerning awareness, implicit learning entails acquisition that occurs at an unconscious, lack-of-awareness level (Reber, 1993) and in unintentional and meaning-focused conditions (Rebuschat & Williams, 2009). Explicit learning, on the other hand, involves conscious awareness with an intention to find regularities and memorization (Hulstijn, 2005). Verbalizabilityi.e.self-reporting, is another criterion. Implicit knowledge, which learners acquire without awareness, is used without being known, therefore, it cannot be verbalized. In comparison, explicit knowledge, which results from a conscious and analyzing process, can consciously be applied and verbalized entailing at least some degree of metalinguistic knowledge. Therfore, it is suggested that learners are able to verbalize a subset of their explicit knowledge, even if they are not likely to verbalize the entire contents of it with proper metalanguage (R. Ellis, 2009).  In terms of processing and accessibility, explicit learning requires effort and strategic expertise, while implicit learning is an automatic process (Dorynei, 2009; R. Ellis et al., 2009). In other words, explicit knowledge is accessible through controlled processing, whereas implicit knowledge is accessible through procedural processing (DÖrnyei, 2009). Therefore, implicit knowledge is available for fluent, spontaneous use of language, whereas explicit knowledge is accessible through monitored control in planned usages of language. However, the controlled vs. automatic types of processing  5 and the accessibilites of the two types of knowledge can be measured from their speeds. The boundaries of implicit knowledge, automatized explicit knowledge, and explicit knowledge have been blurry, due to: 1) proficient L2 learners practice and speeded-up processing of information, and 2) lack of proper measures for the implicit and explicit knowledge. For example, DeKeyser (2003) has suggested that explicit knowledge can be fully automatized and functionally equivalent to implicit knowledge. This raises an important question of how to differentiate automatized explicit knowledge from implicit knowledge in terms of accessibility, processing, and even learning. This question makes researchers keep speculating on theories of interface between the two types of knowledge (DeKeyser, 2003; Hulstijn, 2002; N. Ellis, 2005).  The implicit learning mechanism is also different from the explicit learning mechanism in terms of variability, consistency, and certainty (R. Ellis, 2005, 2009). As previously suggested, the implicit learning system is primary and default; the majority of L1 competence depends on implicit learning and implicit knowledge. In addition to this, when the remarkable similarities of L1 speakers in their end-state of L1 acqusition are considered, it is reasonable to suggest that their implicit knowledge plays an important role in L1 speakers homogeneity in L1. In fact, implicit learning is less subject to learner-to-learner variation or to period-to-period variation across a life span of learners than explicit learning. Therefore, individual differences such as IQ and aptitude do not influence the implicit learning system of learners as well as they affect  6 the explicit learning system (DÖrnyei, 2009; Robinson, 2005). This is based on the postulation that explicit learning involves standard smartness because it requires elaboration and deep processing of the task (DÖrnyei, 2009; N. Ellis, 2005; Robinson, 2005). Thus, explicit learning ability might be correlated with learners general cognitive capacity, resulting in much variability between individuals. However, for less proficient L2 learners in instructional contexts, their implicit knowledge should be limited compared to their explicit knowledge. In this case, less variability in implicit knowledge between learners and more certainty in implicit knowledge might not be relevant. Instead, their explicit knowledge that depends on their invididual differences (IDs) counts in instructional settings.        Researchers in SLA have been interested in how both types of representations of knowledge are processed, interacted, and related to pedagogical implications. It has been suggested that even though learners cannot use implicit and explicit knowledge concurrently due to their distinctive natures, these two types of knowledge and cognitive processing may interact in a reasonable degree at the level of performance (DÖrnyei, 2009; R. Ellis, 2004; Ellis et al., 2009). In addition to their interaction at the level of language use, researchers in SLA and neurocognition are especially interested in the possibility of conversions from one type of knowledge into the otheri.e.interface between the two types of knowledge (DeKeyser, 2003; N. Ellis, 2005; R. Ellis, 2005, 2009; Krashen, 1981; Paradis, 2009; Ullman, 2001, 2004).    7 There are three major views about the interface of two types of knowledge and their views entail different pedagogical implications. The noninterface position assumes an absolute distinction between implicit and explicit knowledge (Hulstijn, 2005; Krashen, 1981; Paradis, 2009). There is no converting explicit knowledge directly into implicit knowledge, and vice versa; the two types of knowledge are not influenced by each other. According to this position, the implicit learning situation should be encouraged for the attainment of more implicit knowledge. The reason is that implicit knowledge is a primary factor in explaining language competence and implicit knowledge is obtained from implicit learning. In this perspective, communicative tasks should be encouraged for automatic and spontaneous learning, whereas tasks to enhance metalinguistic knowledge in teacher-centered learning should be discouraged (Krashen, 1981).  The strong interface position claims that even though implicit knowledge and explicit knowledge are two distinct systems, learned explicit knowledge, such as grammatical regularities, can be converted to implicit representation. This transformation is possible through communicative practice and use (DeKeyser, 1998). Likewise, conversion of implicit knowledge to explicit knowledge is possible through the process of conscious reflection and analysis. At the neural level, this conversion was explained using strengthening and weakening of declarative memory (DM) and procedural memory (PM) networks. Crowell (2004) suggested (as cited in Ellis, 2009, p.  8 22) that the conversion of explicit knowledge to implicit knowledge in the strong-interface position might be a strengthening of connections in the procedural network. This occurs with a weakening of connections in the declarative network, which leads to the proceduralization of declarative knowledge. However, it is not clear what conversion of two separate knowledge systems exactly means (Dörnyei, 2009). Pedagogically, in this position, instruction receives a strong support, and form-focused instruction as well as meaning-focused activities finds essential roles in SLA.        Last, the weak interface position puts restrictions on improving learners implicit knowledge through explicit learning. For the transformation, learners should be ready developmentally and psychologically. Moreover, explicit learning only facilitates implicit learning by promoting controlled practice, but does not change the natural learning sequence of implicit learning processes, nor does it lead to implicit knowledge (Ellis, 2009). This suggests that explicit knowledge contributes indirectly to implicit knowledge by promoting certain processes where implicit learning can occur (Dörnyei, 2009). This guarantees that implicit and explicit learning processes work in concert. Pedagogically, the magnitude of restrictions and degrees of indirect contributions should be investigated to assess the exact role of instruction in enhancing implicit knowledge.      The issue of interface requires a great deal of further research at the levels of psychology, neurocognition, and behavioral studies. In psychology and neurocognition, there is another strong and related distinction between declarative memory (DM) and  9 procedural memory (PM) (Anderson, 1983; Paradis, 2009; Ullman, 2001). In nature, DM is factual and PM is rule-based. Concerning their access and processing, DM is accessed slowly, whereas PM is highly automated, and therefore, quickly accessed. The DM and PM networks should be regarded as dissociated, but functionally cooperative. Regarding the relationship of DM and PM with implicit and explicit knowledge, as it manifests in the intertwined relationships between SLA, cognitive psychology, and neuroscience, DM and explicit knowledge have been viewed as parallel, as have PM and implicit knowledge (DeKeyser, 1997; Dörnyei, 2009; R. Ellis, 2005; Hulstijn, 2002; Paradis, 2004, 2009; Ullman, 2001, 2004, 2005). Eretin and Alptekin (2013) also suggested that declarative/procedural (DP) memories have a complementary relationship with the implicit and explicit knowledge systems that are embedded in DM and PM systems. Therefore, in the language learning of late L2 learners, it is natural for conscious awareness and controlled processing to govern the ways in which explicit learning and DM processing take place.  Ullman (2001) presented the declarative/procedural (DP) model and suggested that the interface between the two systems might be impossible due to a disparity between memory systems and types of processing for adult L2 learners. L1 speakers learning procedure activates both declarative and procedural components of the memory system. However, late L2 learners, whose age of onset (AoO) is after puberty, experience a continuing weakness of PM and a strengthening of DM (Ullman, 2001).  10 This change differentiates late L2 learners processing mechanisms from L1 acquisition. In other words, L1 speakers process morphosyntactic and phonological information of L1 implicitly through the procedural system, whereas lexical and semantic information through the declarative system. However, L2 learners strengthened declarative system and weakened procedural system seem to find a different optimum processing mode from that of L1 speakers. Since the declarative mechanism is associated with conscious awareness and controlled processing in an instructional setting, adult L2 learners seem to process all L2 linguistic informationsuch as morphosyntactic, phonological, lexical, and semantic informationexplicitly. Therefore, it is more plausible that the learning of adult L2 learners results in more declarative and explicit knowledge due to their highly controlled, conscious, and monitored learning context in a grammar-oriented class.  However, Ullman (2001, 2004, 2005) also suggests that due to L2 learners practice and use in an L2, the DM-based processing of explicit knowledge is expected to be gradually replaced by the PM-based, automatic use of implicit knowledge. In other words, from extensive use, L2 learners mismatched, but strengthened DM for morphosyntactic and phonological associations slowly prepares for the PM-based procedure (Eretin & Alptekin. 2013). This means that L2 learners PM system becomes more available to acquire grammatical knowledge, resulting in better learning in PM, and leading to higher proficiencies in L2 learners. Based on the L2 learners similar neurocognitive patterns to those of L1 speakers, Ullman (2005) suggests that this  11 advanced level of L2 knowledge was almost like that of L1 speakers in terms of grammatical dependence on the PM system, and depended on the nature of the L2 exposure and instrinsic procedural learning abilities. This suggests that at the processing level, highly proficient L2 learners experience qualitative change in the type of information processing. However, based on a lack of enough evidence concerning their direct interface, this change of L2 proficient learners type of processing does not necessarily mean a direct conversion from explicit knowledge into implicit knowledge, especially due to the distinct nature of DP memory networks that underpin implicit and explicit knowledge systems (Ullman, 2001, 2004). More studies at the behavioral level should be conducted as a preparatory step for investigating the interface issue. In this context, SLA research has seen quite a few important studies on valid measures of implicit and explicit knowledge.  1.3. Measurement of implicit and explicit knowledge To gain insight in the nature and operationalization of the two distinct types of linguistic knowledge, a handful of recent studies have investigated possible measurements with high construct validity (Bowles, 2011; R. Ellis, 2005; R. Ellis, et al., 2009; Kim & Nam, 2016; Loewen, 2003; Spada, Shiu, & Tomita, 2015; Suzuki & DeKeyser, 2015; Vafaee, Kachinske, & Suzuki, 2016; Zhang, 2015). The following criteria have been proposed as characteristics or qualities of implicit knowledge:  12 absence of awareness, time pressure, focus on meaning, consistent responses, certainty about correctness/incorrectness of the responses, no use of metalinguistic knowledge, y for children (R. Ellis, 2005, 2006). As criteria for explicit knowledge, the following have been proposed: higher degree of correctness/incorrectness of the responses, use of metalinguistic knowledge, and learnability for learners with form-focused instruction (R. Ellis, 2005, 2006). Based on these criteria, elicited oral imitation tests (EIT), oral narrative tests, and timed GJTs are interpreted as better measures of implicit grammar knowledge,while untimed GJTs and metalinguistic knowledge tests are regarded as better measures of explicit grammar knowledge with the highest validity for learners of English as L2 (Loewen, 2003; R. Ellis, 2005, R. Ellis, et al., 2009), as well as for learners of Spanish as a HL and L2 (Bowles, 2011).   Time pressure has been found to be an important factor in distinguishing measurements of implicit and explicit knowledge in various tasks (Bowles, 2011; R. Ellis, 2005; Godfroid et al., 2015; Kim & Nam, 2016; Spada et al., 2015; Zhang, 2015). In their studies, time availability meant whether or not learners were under pressure to conduct a task spontaneously, or whether or not they had an opportunity to plan their response carefully before responding. Operationally, to have or not to have time available involved distinguishing tasks that were demanding on learners short-term  13 memories from those that allowed L2 learners to utilize their L2 processing capacity fully and comfortably (R. Ellis, 2009, p. 38). Accordingly, three testsan EIT, oral narrative, and timed GJTincorporated time constraints, and all of them loaded on the implicit knowledge factor, while the two unpressured testsan untimed GJT and metalinguistic testloaded on the explicit knowledge factor (R. Ellis, 2009, p. 60)    An EIT requires learners to focus on meaning rather than form and time pressure, which is inherent to an online and spontaneous natural situation of language use. An EIT that is also reconstructive in nature would have certain features that distinguish it from a test that might allow learners to rely on simple rote repetition of target stimuli. To avoid simple rote repetition and to enhance meaning-focused and spontaneous production, the test needs to include some delay between the presentation and repetition of the stimulus. The EIT is inherently time pressured in that learners listen to each statement only once in real time, and the participants produce their ideas and repeat the statement in their own time, i.e., self-paced (Ellis, 2009, p.79). Spada, Shiu, and, Tomita (2015) investigated the construct validity of an EIT as a measure of implicit knowledge. 73 EFL learners took part in the study which utilized an EIT, timed aural and written GJTs, a written error correction test (ECT) with three sectionsidentification, correction, and explanationand an OPT. The passive construction was the target structure. The results from a factor analysis were that the EIT, timed aural, and timed written GJTs loaded highly onto a factor identified as  14 implicit knowledge, whereas all the sections in the ECT loaded on the other factor labeled explicit knowledge. The timed written GJT also loaded on explicit knowledge relatively highly, with a .51 loading. They suggested that the EIT was a valid measurement of implicit knowledge, and more studies would be necessary for better understanding of the construct validity of EITs as a measurement of implicit knowledge.  For the purpose of investigating if an EIT is a valid measure of implicit knowledge in a refined way, additional time pressure was added for further study. Kim and Nam (2016) added time pressure to the EIT to see if time pressure makes a difference to the validity of the EIT as a measure of implicit knowledge. 66 adult advanced EFL learners participated in the study. The authors replicated the tests and test content, and the same 17 English structures utilized by Ellis (2005). They manipulated the EIT with or without time constraint. The time limit for each stentence was calculated based on nine native speakers time for production of each sentence without time pressure. The researchers produced a three-factor modelstronger implicit, weaker implicit, and explicit knowledge. The EIT, regardless of time pressure and grammaticality, measured stronger implicit knowledge. Their study suggests that the EIT is a valid measure of implicit knowledge regardless of time pressure embedded, confirming the previous studies of R. Ellis (2005) and Spada et al. (2015). Gutiérrez study (2012) suggested that the grammaticality of test items in written GJTs is a potential variable over time pressure for high construct validity in  15 separately measuring implicit and explicit knowledge of Spanish. In his study with 53 L2 learners of Spanish at low and high proficiencies, Gutiérrez (2012) utilized a self-paced PowerPoint-slide-show format in the timed GJT. The amount of time that the stimuli remained on the screen varied between 6 and 9 seconds. In the discussion section, he suggested that the grammatical sections in timed and untimed GJTs constitute good measures of implicit knowledge regarding 16 Spanish target structures. In comparison, the ungrammatical sections in timed and untimed GJTs and the metalinguistic knowledge test measured explicit knowledge with high validity. He maintained that grammatical and ungrammatical sentences might lead learners to tap two different knowledge representations. R. Ellis (2005) found that, depending on the tasks, grammaticality is important as well as time constraint. In his study, L2 learners scores on grammatical items in the untimed GJT loaded on both the implicit and explicit knowledge factors. Thus, Ellis (2015) and Godfroid et al., (2015) excluded the grammatical items only in the untimed GJT as a measure of explicit knowledge; but in the timed GJT, no exclusion was required for the grammatical items. This seems to mean that time constraints play a role in measuring two types of knowledge separately, regardless of the grammaticality of items in the GJT. Since learners can tap either type of knowledge representation when they are given unlimited time, learners can use either type of grammar knowledge. In Ellis study (2005), the L2 learners seemed to tap both implicit and explicit knowledge for judging the grammaticality of grammatical items  16 when there was no time pressure, but not for ungrammatical items. However, it is still unclear what variables influence learners to tap which knowledge for grammaticality judgment.  The effects of time constraint have been highlighted in comparisons between timed and untimed GJTs. In studies using L2 learners, significant differences have been found between the two tests, demonstrating the importance of time constraint for separate measures of implicit and explicit knowledge (Ellis, 2009). Bowles (2011) identified time pressure as the sole important factor for written GJTsboth grammatical and ungrammaticalfor measuring implicit and explicit grammar knowledge separately. This supports Ellis results in terms of timed GJTs, but not grammatical items in untimed GJTs. Her study deserves attention in that: 1) R. Ellis study (2005) required grammaticality for the untimed GJT as a measure of explicit knowledge, and 2) Gutiérrez (2012) found that grammaticality is the important factor in distinguishing measures of two types of knowledge. Later, Zhang (2015) extended R. Ellis and Bowles results to Chinese learners of English as a foreign language. The study with 100 university-level students replicated the results of R. Ellis study (2005) in that the EIT and timed GJT had high and significant loadings onto implicit knowledge, whereas ungrammatical items in the untimed GJT and metalinguistic test loaded onto explicit knowledge. Zhangs study also replicated Gutiérrezs study (2012) in that the EIT, and grammatical items in the timed and untimed GJTs loaded highly onto implicit  17 knowledge, whereas the metalinguistic knowledge test and ungrammatical items in the timed and untimed GJTs were loaded to explicit knowledge. This result emphasizes the importance of the grammaticality of GJT items for tapping implicit and explicit knowlege separately, but item grammaticality does not seem to override the validity of the EIT as a whole. Therefore, the total items of the EIT tap into implicit knowledge. Zhangs study (2015) also highlights and confirms R. Ellis study (2005) that when learners answered GJT items under no time pressure, they tapped into either explicit or implicit knowledge freely, except when the items were ungrammatical. Therefore, ungrammatical items led the participants to tapping explicit knowledge.  The relationship between time pressure, grammaticality and accuracy should be clarified for better understanding the proper models for the two types of grammar knowledge. It has been supported that the learners recorded higher scores in grammatical sections than ungrammatical sections, regardless of the existence of time constraints (Gutiérrez, 2012; Loewen, 2009). When judging sentences in terms of grammaticality, learners engage in three processing operations: 1) understanding the meaning, 2) deciding whether or not there is something ungrammatical, and 3) identifying what is incorrect and why it is incorrect (Loewen, 2009; R. Ellis, 2005). According to this explanation, judging ungrammatical sentences requires at least one more operation, which results in lower accuracy (Gutiérrez, 2012; Loewen, 2009). However, this notion does not necessarily mean that judging ungrammatical sentences  18 requires more time. This was confirmed in Loewens study (2009), where he found that grammatical sentences require more time than ungrammatical sentences. However, it is not clear how fast the three operations can occur. In other words, it should be clarified whether or not these three operations can indeed occur almost simultaneously and, if they can, in what context this is possible. Moreover, since Gutiérrez study did not present response time for GJTs, it is hard to conclude at this point whether either the time restriction or the grammaticality of items are, in isolation, important factors for GJTs as measurements of implicit or explicit grammar knowledge.     In addition to time pressure and grammaticality, modality has also been investigated as a variable to explain learners linguistic knowledge (Bialystok, 1979, 1982; Granena, 2012; Kim & Nam, 2016). The rationale has been that aurality and orality predispose learners use of implicit knowledge. Granena (2012) utilized GJTs in different modalities to investigate implicit and explicit knowledge of learners of Spanish using three groups: L1 speakers, early L2 learnersage of onset (AoO) was from 3 to 6 years oldand late L2 learners, whose AoO was 16 years old or later. She utilized time pressure and modality to measure implicit and explicit knowledge of Spanish. To measure implicit knowledge, timed visual and timed auditory GJTs were utilized, whereas for explicit knowledge, untimed visual and untimed auditory GJTs were used. As in R. Ellis study (2005), for the timed visual GJT, 120% of the time that L1 speakers spent in responding to each item was allocated. For the timed auditory GJT, 120% of the  19 time that L2 speakers spent in responding to each item was also allocated. This means that for the auditory GJT, not only modality but also time pressure was involved. For the untimed auditory GJT, each item was presented twice under no time pressure to respond. It was found out that the effect of modality was qualified by an interaction with groupL1 speakers and early L2 learners performed better on auditory tests than on visual tests, whereas late L2 learners performed better on visual tests than on auditory tests. The researcher suggested that instead of using the timed auditory GJT, a natural auditory GJTpresenting the sentence only once without time pressuremight be a better measure of implicit knowledge of Spanish (Granena, 2012).  Kim and Nam (2016) conducted another study about the effects of test modality. They employed 17 English structures, which were selected based on difficulty, acquisition timing in the developmental process, pedagogical timing, and linguistic aspects of the structures (R. Ellis, 2005). 66 adult advanced EFL learners participated in the study. The researchers manipulated modality for the timed GJT, aural, and written GJTs. As previously mentioned, a three-factor model resultedstronger implicit, weaker implicit, and explicit knowledge. The aural GJT with time pressure measured both stronger and weaker implicit knowledge. In comparison, the timed written GJT and timed grammatical aural GJT measured weaker implicit knowledge. Finally, the metalinguistic test measured explicit knowledge. They concluded that the timed GJT in aural and written mode measured weaker implicit knowledge. This study suggests that a  20 continuum should be utilized to explain the two types of knowledge and depending on the measurements, implicit knowledge can be more specified. However, it is difficult to assess the effectiveness of modality as a factor in determining valid measures of implicit and explicit knowledge due to the fact that time constraint was also embedded in the aural and written GJTs. Therefore, it is difficult to tease apart the influence of time constraint from that of modality in the model. Aural GJTs with a natural time constraint and written GJTs without time constraint might measure implicit and explicit knowledge effectively as measures with high validities.  Two studies are noteworthy in that they challenge the findings of previous research by introducing new measures of implicit knowledge. First, Suzuki and DeKeyser (2015) recruited 63 advanced Chinese learners of Japanese as L2. Five particle-related structures in Japanese were selected. The researchers developed an EIT with a built-in word monitoring task and they labeled it an EIM task. The EIT scores were positively correlated with metalinguistic knowledge, not with the probabilistic serial reaction time (SRT) task scores, which was supposed to measure learners implicit sequence learning ability. In contrast, the word monitoring (WM) test scores were not correlated to metalinguistic knowledge, but with SRT scores among L2 learners with longer residence in Japan. Based on these results, the authors suggest that the EIT measure automatized explicit knowledge, not implicit knowledge. A couple of issues should be addressed concerning the EIM task. First, the participants were instructed to  21 convert ungrammatical sentences into grammatical sentences. This could have predisposed learners toward tapping explicit knowledge, making them focus on grammar in an analytical way. Second, their EIM task was different from the EITs utilized in previous research in that the WM test was embedded in the EIM task. Therefore, it is difficult to compare DeKeyser and Suzukis result directly with those in previous research. In this case, a claim is possible that even if the EIM was highly correlated with the metalinguistic test, this does not necessarily mean that the EITs utilized in previous research would have correlated with the metalinguistic test. In addition, the time pressure learners had before they produced imitation sentences was a maximum of eight seconds for every sentence, which may have arguably enabled them to tap explicit knowledge.  Second, Vafaee, Kachinske, and Suzuki (2016) utilized written GJTs, a metalinguistic test, a self-paced reading (SPR) task, and a word-monitoring task (WMT) to measure 79 advanced L2 learners implicit and explicit knowledge of English. Four target structurespresent hypothetical conditional, third-person s, simple past/present perfect, and mass/count nounswere selected due to the difficulty of mastering them and their easiness to incorporate in the tests. The researchers claimed that written GJTs, even with time constraint, might be too coarse to measure implicit knowledge. Instead, they presented an SPR task and a WMT. A series of confirmatory factor analyses demonstrated that the timed, ungrammatical GJT, the untimed, ungrammatical written  22 GJT, and the metalinguistic test measured explicit knowledge, whereas the SPRT and WMT measured implicit knowledge. They thus concluded that GJTs with/without time pressure and/or grammaticality of the sentences were not fine-grained measures of implicit knowledge. Instead, the GJTs measured two different levels of explicit knowledge.  It seems that there has not been an agreement on valid measures of implicit knowledge in SLA. However, one measure of explicit knowledge has been confirmed in all previous test validity research. Regardless of the number of factors in the final model, metalinguistic tests highly loaded on explicit knowledge in previous research (Bowles, 2011; Ellis, 2005; Kim & Nam, 2016; Suzuki & DeKeyser, 2015; Vafaee, Kachinske, & Suzuki, 2016; Zhang, 2015). This confirms that, on the continuum of implicit and explicit knowledge, the metalinguistic test is positioned in the far end of explicit knowledge as a valid measurement.   More studies are required to investigate the relationship between grammaticality of test items, time pressure, and modality in order to obtain a more comprehensive picture of the construct validity of various measurements. However, the results from previous studies help draw a temporary summary as follows:  First, rigorous time pressure based on native speakers processing time as in various timed GJTs (Bowles, 2011; Ellis, 2005; Godfroid et al., 2015; Kim & Nam, 2016; Zhang, 2015), seems to predispose learners toward the use of implicit knowledge  23 and tends to override the grammaticality of test items. In other words, when there is a rigorous time restriction for each stimulus based on L1 speakers reaction times, implicit knowledge can be measured through timed GJTs regardless of the grammaticality of the items. Rigorous time pressure seems to override modality, too. In Kim and Nams study (2016), the timed written GJT and timed aural GJT both loaded on the same construct labeled implicit knowledge. This means that as long as time pressure was manipulated in a rigorous manner, different modes did not affect the loadings.  Second, modality is also an important factor. In Kim and Nams study (2016), the timed written GJT and timed aural GJT loaded differently on weaker implicit knowledge. The timed written GJT loading on implicit knowledge was .61, whereas the timed aural GJT loading was .85. Therefore, the timed aural GJT can arguably explain weaker implicit knowledge more strongly than its written counterpart (Kim & Nam, 2016; Spada et al., 2015).    Third, grammaticality alone does not seem to play a primary role in distinguishing between measurements of implicit from explicit knowledge. When there is no time restrictionas in the untimed written GJTs (R. Ellis, 2005)or no rigorous time restriction in a timed written GJT (Gutiérrez, 2012), it is possible that ungrammatical items in a written mode can be used to measure explicit grammar knowledge effectively. Moreover, grammaticality seems to play an important role when time pressure conditions are the same. Grammaticality turned out to be important to  24 distinguish between two constructs of knowledge, labeled as stronger implicit and weaker implicit knowledge. Kim and Nam (2016) found that the ungrammatical items in the timed aural GJT loaded heavily on stronger implicit knowledge, whereas the grammatical counterparts in the same timed aural GJT loaded heavily on weaker implicit knowledge. This demonstrates that, when conditions of modality and time pressure are the same, grammaticality is an important factor in measuring measure weaker and stronger implicit knowledge relatively separately.  Fourth, tasks that address both oral and aural skills seem to predispose learners toward tapping into their implicit knowledge. When EITs and aural GJTs are compared, EITs are confirmed to be more valid measurements of implicit knowledge than aural GJTs (Kim & Nam, 2016). This might possibily be due to EITs elements of both decision on plausibility and reconstruction/production, whereas GJTs have only a decision/comprehension element. When a task requires a higher cognitive load, it provides a better measure of that knowledge. In support of this argument, Kim and Nam (2016) found that the EIT loaded on stronger implicit knowledge, with or without time pressure, and regardless of the grammaticality of the test items.  Fifth, each test in the battery of tests utilized by R. Ellis (2005) had its own characteristics and that could become more valid measurements by manipulating time pressure, modality, and grammaticality. More studies using different groups of learners, languages, and tests are required in order to investigate validity claims regarding  25 measurements of implicit and explicit knowledge, and to obtain a battery of tests with high validity for each type of knowledge. For example, direct relationships between modes, time pressure, and grammaticality for the battery of the tests, including GJTs, and their construct validities for measurements of implicit and explicit knowledge are worth investigating further.   1.4. Language learners and two types of learning and knowledge There is a wide consensus that L1 learners acquire their L1s primarily through implicit learning (e.g., DÖrnyei, 2009). When children learn and use language, they use implicit knowledge from implicit learning, which is stored in procedural memory and executed automatically (Paradis, 2004, 2009). Generally, they do not make a conscious attempt to learn the target material until they receive explicit instruction in classroom settings. During automatic implicit learning, L1 learners are not aware of either their learning process or the outcome of it (DÖrnyei, 2009). Learning occurs in a natural setting where children are exposed to a considerable amount of natural input.    On the other hand, adult L2 learners utilize a different cognitive system for L2 acquisition, explicitdeclarative memory (DM) system, arguably because maturational constraints affect the implicit learning system used in childhood (Bley-Vroman, 1989, 2009; DeKeyser, 2000, 2003; Ullman, 2001). This loss of implicit cognitive mechanisms for language learning forces adult L2 learners to rely on explicit learning  26 and knowledge. Furthermore, because adult L2 learners start to learn L2s later than L1 speakers do, they usually start to learn L2s in an instructional setting. In class, they are presented with taught language rules, encouraged to derive hypotheses and put them to the test. In terms of input amount, L2 learners total amount of input is much less than that of L1 speakers.    Heritage language (HL) learners are defined as those who have been raised in a home where a HL is spoken, who speak and understand the HL, and who are bilingual in their heritage language and the dominant language (Valdés, 2001). They have some characteristics of native speakers and some characteristics of L2 learners. Concerning the characteristics shared with native speakers, HL learners start to learn L1s through implicit learning in a natural learning setting first. Their major input mode is aural and stems from family members. In this stage, their knowledge of the L1 is implicit. After the age of 5, when metalinguistic skills develop (Oh, Jun, Knightly, & Au, 2003; Montrul, 2008), they come to learn L1s both implicitly and explicitly. However, HL learners metalinguistic knowledge might be limited due to their limited exposure to instructional settings relative to L1 learners. Since general cognitive skill acquisition theory is based on the premise of age-related maturational constraints on language learning (DeKeyser, 2007, 2009; DÖrnyei, 2009), heritage learners should go through a transition period by the age of six from more implicit learning in terms of syntactic structuressuch as relative clauses (Polinsky, 2011)to more explicit learning as they  27 mature. In addition to the change in their L1 learning, the amount and frequency of L1 input is reduced or varied as heritage learners acquire and use their dominant language. Resulting from this lack of instruction and input, their L1 knowledgeparticularly explicit knowledgeexperiences attrition (Polinsky, 2011), which leads to incomplete knowledge representations (Montrul, 2008; Paradis 2004). Therefore, heritage learners L1 knowledge is similar to native speakers in that they have implicit knowledge for some aspects of their L1. At the same time, HL learners knowledge is similar to that of L2 learners in that their linguistic knowledge is not fully developed. Finally, both HL and L2 learner groups explicit knowledge might be fragile, but not HL learners implicit knowledge might not be that as fragile as their explicit knowledge (Allen & Reber, 1980; Montrul, 2008).  However, Bowles (2011) included HL learners as another group for the research on language learners implicit and explicit grammar knowledge. Before her study, the majority of studies about the two types of grammar knowledge were about L2 learners (R. Ellis, 2004, 2005, 2006; Ellis, Loewen et al., 2009). However, in this study (Bowles, 2011), the participants were HL learners whose instruction length was no more than 2 years in addition to L2 learners whose instruction length in secondary and post-secondary settings was 6 years on average. The results demonstrated that along with L2 learners, HL learners scores on the battery of five tests (i.e., an EIT, an oral narrative test, a timed GJT, an untimed GJT and a metalinguistic knowledge test), loaded highly  28 onto a two-factor model with high construct validity, where time pressure and mode are crucial. Therefore, it can be claimed that HL learners can provide further support for the distinction between implicit and explicit knowledge measures.  1.5. Studies on heritage and L2 learners linguistic knowledge Previous studies about HL learners have mainly focused on similarities and differences in the linguistic performances of HL learners and L2 learners. In terms of what I propose to interpret as implicit knowledge, a handful of studies can be categorized as demonstrating that HL learners have an advantage over non-heritage learners in implicit knowledge in terms of phonological, morphological, syntactic, and lexical aspects. In perception tasks of voice onset time (VOT) measurement, Spanish heritage learners were shown to perceive six Spanish consonants VOT better than L2 learners (Au, Knightly, Jun, & Oh, 2002). Korean heritage learners also demonstrated native-like abilities in the perception of VOT of six Korean consonants in a phoneme perception test (Oh, Jun, Knightly, & Au, 2003), meaning that, in terms of perception, the heritage learners implicit knowledge of the phonological feature is as solid as that of native speakers of Korean. Spanish heritage learners also performed more accurately than L2 learners on a computer-based, untimed oral production task involving Spanish gender agreement (Montrul et al., 2008). Even though the task was untimed, it could be a measure of implicit knowledge since it was a meaning-based, oral assignment (R. Ellis,  29 2005, 2006). In a word recognition task aimed at measuring knowledge about gender, French heritage learners were asked to orally imitate the last word in phrases that contained matches and mismatches of gender markers and nouns. When the phrases were grammatically correct, their reaction time was the same as native speakers and significantly shorter than L2 learners, which might mean heritage learners have: 1) the same implicit knowledge of gender markers as native speakers and 2) better implicit knowledge than adult L2 learners (Guillelmon & Grosjean, 2001). Finally, one study suggests that HL learners might have better implicit and explicit knowledge on various aspects of language than do L2 learners. In a guided narrative test (i.e., measure of implicit knowledge), Korean heritage learners outperformed L2 learners in mimetics, particles, complex predicates, connectives, and honorifics, but not in idiomatic expressions (Lee, Kim, Kong, Hong, & Long, 2005). In this study, the most advanced learners among the heritage learners found honorifics the easiest, which suggests that honorifics might be a complex and potentially proficiency-dependent, but not age-of-onset (AoO) dependent structure.  In contrast, the Korean HL learners in OGrady, Lee, and Choo (2003) did not show better knowledge about subject and object relative clauses than L2 learners. In a forced-choice picture task, participants heard aural cues only once and were asked to choose the one picture that matched the cue. The interval between each item was eight seconds. Considering the task was in an aural mode and had an element of time  30 constraint, the participants implicit knowledge of Korean relative clauses was able to be measured. However, since the Korean heritage learners in the study had not been exposed to Korean input at home, the results might indicate the importance of input and output for young heritage learners rather than informing us about the exact degree of implicit knowledge of the target rule of general heritage learners and L2 learners, respectively. Regarding explicit knowledge, two studies found that heritage learners do not have an advantage over non-heritage learners. Spanish HL learners did not outperform L2 learners in untimed GJTs in terms of verbal agreement, gender agreement, tense, aspect, mood, or clitics (Au et al., 2002), all of which could have measured participants explicit knowledge (R. Ellis, 2005). Another study reported that L2 learners had better explicit knowledge about Spanish gender marking. Montrul et al. (2008) found that the L2 learners outperformed Spanish heritage learners in written gender recognition and written picture interpretation in terms of gender agreement accuracy, but not in an oral description task, which might have measured implicit knowledge. Based on these findings, Montrul et al. (2008) suggest that heritage learners and L2 learners have an advantage over each other in different types of modes and knowledge.  On the other hand, in the aforementioned study, Lee et al. (2005) also used a written test (i.e., a potential measure of explicit knowledge) that consisted of multiple- 31 choice questions and fill-in-the-gap questions. Korean HL learners outperformed L2 learners in all the target features except idiomatic expressions. These results contradict those of the previous two studies that demonstrated L2 learners outperforming HL learners in the untimed GJTs and the written tests that arguably measured participants explicit knowledge. However, a large gap between the numbers of the two groups23 HL learners versus only four L2 learnersmake the results less conclusive and calls for more research on this issue (Lee et al., 2005).    In sum, HL learners in Spanish and Korean have been shown to outperform L2 learners in oral tasks and in tasks that involve fast and less consciously analyzable linguistic knowledge. This can be indirect evidence supporting the claim that HL learners have stored more implicit linguistic knowledge about their HL than L2 learners have about their L2, presumably because the HL learners have learned their HL implicitly (Montrul, 2008). However, the measurements utilized in the previous studies were not originally designed to measure implicit and explicit knowledge, per se. Therefore, a study with robust construct validity is required to understand the nature of the two groups knowledge representations more clearly.  To the best of my knowledge, only one study (Kim & Nam, 2016) has recently tried to investigate the direct relationship between different test modalities (especially in GJTs) and test construct validity with regard to measuring implicit and explicit knowledge. At this juncture, this investigation that is the subject of this paper will build  32 upon HL learning studies by identifying linguistic differences or similarities between HL learners and L2 learners of Korean in terms of implicit and explicit knowledge. In addition, this study will extend the scope of existing research on measurements of implicit and explicit knowledge by considering another novel population, i.e., HL learners of Korean and another variable, i.e., modality.  33 1.6. Research questions The following research questions guided this study: 1. How well does a battery of tests (elicited imitation test, aural GJT, written GJT, and metalinguistic test) measure implicit and explicit linguistic knowledge?  2. Do HL and L2 learners differ on their implicit knowledge measured through the elicited imitation test and aural GJT? 3. Do HL and L2 learners differ on their explicit knowledge measured through the written GJT and metalinguistic test?       34 CHAPTER 2. METHODOLOGY   This study is a cross-sectional quantitative investigation of the implicit and explicit knowledge of native speakers, HL learners, and L2 learners of Korean. This study is a conceptual replication of: 1) previous studies that used a battery of tests in an attempt to measure implicit and explicit knowledge separately (Ellis, 2005; Gutiérrez, 2012), and 2) Bowles study (2011) in terms of using both HL and L2 learners performances to validate measurements of implicit and explicit knowledge, and to confirm the similarities/differences of the two groups concerning that knowledge.   2.1. Target structures In some of the aforementioned studies, 17 language structures were selected and used as target. As a replication study, I tried to select 17 target structures for the current study, too. A handful of researchers have investigated the order of acquisition either by L1 learners of Korean (Cho, 1982; Jeong et al., 2006; Kim & Kang, 2010; Kwon et al., 1979; Lee et al., 2008; Lee & Lee, 2010; Seo & Lee, 1999) or by L2 learners of Korean (Jeong, 2008; Kim, 2006; Seol, 2010; Um, 1989). However, those studies are limited to a couple of target structures such as phonemes (Jeong, 2008; Jeong et al., 2006; Kim, 2006; Kwon et al., 1979), particles (Cho, 1982; Lee et al., 2008; Seol,  35 2010; Um, 1989), or conjuntions (Kim & Kang, 2010; Lee & Lee, 2010; Seo & Lee, 1999). To my knowledge, few of these studies have considered Korean acquisition order, which covers all aspects of Korean structures comprehensively as in the studies that were based on processibility theory (Pienemann, 1998). Therefore, it is hard to put forward well-established criteria for selecting target structures based on previous studies.  As an alternative, I used two criteria to select target structures for this study; 1) target structures should be problematic for learners and result in frequent and persistent errors, and/or 2) structures should cover a full range of proficiency levels in their pedagogical grading (i.e., the order in which the structures are presented in the textbooks or teachers pedagogical grading). To satisfy the two criteria, literature on Korean error analysis (Lee, 2003), a couple of textbooks that have widely been used in the United States and in Korea (Cho, Lee, Schulz, Sohn, & Sohn, 2010; Sogang Korean Institute, 2008), and grammar books of Korean were consulted (Kim, Park, Lee, Lee, Jeong, Choi, & Heo, 2005; Kim-Renaud, 2009). Based on the criteria, 17 grammar structures of Korean were selected (Bowles, 2011; Godfroid et al., 2015; Gutiérrez, 2012; R. Ellis, 2005; Zhang, 2015): word order, particle, conjugation, number and counter, possessive, noun as adverb, adverbial suffix, conjunction, tense, negative polarity adverb, nominalization, relativization (Ko & Ku, 2008), complementation, passive, causative, honorifics, and collocation. Table 1 presents the list of the target  36 structures and a summary of their characteristics in terms of learner errors, teachers pedagogical grading (Sheffler, 2011), and grammatical types, such as lexical, morphological, and/or syntactical categories.  37 Table 1.  Target structures in the tests  Structure Example of the ungrammatical items  Pedagogical grading Knowledge type  1 Word order *Minswu-ka cal kulim-ul kulin-ta. Minswu-Nom well picture-Acc draw-Dec *.R4n˚V 9î  . c Minswu-ka kulim-ul cal kulin-ta.  Minswu-Nom picture-Acc well draw-Dec .R4n˚V  9î . Minsu draws pictures well. Beginning Syntactical 2 Particle (case marker)  *Namtongsayng-i sakwa-lul-man meknun-ta.  Male younger brother-Nom apple-Acc-only eat-Dec *"þ&/339Ê  ,‘$ê%:. c Namtongsayng-i sakwa-lul meknun-ta. Male younger brother-Nom apple-Acc eat-Dec "þ&/339Ê  ,‘$ê%: c Namtongsayng-i sakwa-man meknun-ta. Male younger brother-Nom apple-only eat-Dec "þ&/339Ê  ,‘$ê%: My younger brother eats only apples. Beginning Morphological 3 Conjugation   *Elin ai-ka acwu yeyppu-yo.  Small child-Nom very pretty-Dec. * 7ı9Ê˚V 7ı;R 8^2®8ê. cElin ai-ka acwu yeyppe-yo.  Small child-Nom very pretty-Dec.  7ı9Ê˚V 7ı;R 8^1&8ê The little child is very pretty. Beginning Morphological  38 Table 1.  4 Number and counter  *kyosil-ey haksayng-i tases-kay iss-ta.  Classroom-in students-Nom five-Clf  exist-Dec *˜¦5:8& G¯339Ê %:3– ˚r 9Þ%:. ckyosil-ey haksayng-i tases-myeng iss-ta.  Classroom-in students-Nom five-Clf  exist-Dec ˜¦5:8& G¯339Ê %:3– ,Û 9Þ%:.  There are five students in the classroom. Beginning Morphological 5 Possessive  *Ikestul-un na khemphyuthe-wa phulinthe-ita.  These-Top I computer-and printer-Cop/Dec *9Ê˚Ù':9Œ "î AJG>Cƒ8Œ . cIkestul-un nay khemphyuthe-wa phulinthe-ita.  These-Top I-Poss computer-and printer-Cop/Dec 9Ê˚Ù':9Œ "î9® AJG>Cƒ8Œ .  These are my computer and printer.  Beginning Morphological 6 Nouns as adverbs *Swumi-nun nayil-ey mwehay-yo? Swumi-Top tomorrow-at what-do-Q *4n.N$ê  -¦GÊ8ê? cSwumi-nun nayil mwehay-yo? Swumi-Top tomorrow what-do-Q 4n.N$ê  -¦GÊ8ê?  What does Sumi do tomorrow Intermediate Morphological 7 Adverbial suffix *Na-nun mikwuk-eyse chenchenhakey wuncenhan-ta. I-Top the UN-in slowly drive-Dec *"î$ê .N˜Ã8&3r >ò>òG®˚â . cNa-nun mikwuk-eyse chenchenhi wuncenhan-ta.  I-Top the UN-in slowly drive-Dec "î$ê .N˜Ã8&3r >ò>òIÞ . I drive slowly in the US. Intermediate Morphological  39 Table 1.  8 Conjunction   *Hankwuke-lul paywuko siph-ciman sewul-ey ka-yo.  Korean-Acc  learn-Conj want to-but-Conj Seoul-to go-Dec *    ˚V8ê.  cHankwuke-lul paywuko siph-ese sewul-ey ka-yo.  Korean-Acc  learn-Conj want to-thus-Conj Seoul-to go-Dec     ˚V8ê.  I want to learn Korean, so I go to Seoul. Intermediate Syntactical 9 Tense  *Nayil sophwung-un cham caymi-iss-ess-ta.  Tomorrow picnic-Toc very fun-be-Pst-Dec * 3âFã9Œ >” 9Þ8˚%:. cNayil sophwung-un cham caymi-iss-ul-kes-ita. Tomorrow picnic-Toc very fun-be-Fut-Dec  3âFã9Œ >” 9Þ9ı ˚Ù9Ê%:. Tomorrow the picnic will be very fun. Intermediate  Syntactical 10 Negative polarity adverb *Hankwukmal-un kyelkho swip-ta. Korean-Toc never easy-Dec *G²˜Ã,&9Œ  4Ó%:. cHankwukmal-un kyelkho swipci-anh-ta. Korean-Toc never easy-Neg-Dec G²˜Ã,&9Œ   7€%:.  Korean is never easy. Advanced Morphological 11 Nominalization  *Ceyleymi-nun sihem-eyse silswu-hayss-ki-lul al-ass-ta.  Ceyleymi-Toc exam-in mistake-do-Pst-Nominal-Acc know-Pst-Dec *:r*^.N$ê 52H.8&3r 5:4nGÞ ƒ+Ò 7¢7®%:. cCeyleymi-nun sihem-eyse silswu-han-ke-sul al-ass-ta. Ceyleymi-Toc exam-in  mistake-do-Pst-Mod it-Acc know-Pst-Dec :r*^.N$ê 52H.8&3r 5:4nG² ˚Ù9ı 7¢7®%:. Jeremy knew that he had made a mistake in the exam. Advanced Syntactical  40 Table 1.   12 Relativization (Ko & Ku, 2008) *Na-nun yeca-lul chayksang yephey issnun an-ta. I-Toc woman-Acc desk side-Loc exist-Rel know-Dec *"î$ê 8B9æ+Ò  8\8& 9Þ$ê 7ž%:. cNa-nun chayksang yephey issnun yeca-lul an-ta. I-Toc  desk side-Loc exist-Rel woman-Acc know-Dec "î$ê  8\8& 9Þ$ê 8B9æ+Ò 7ž%:.  I know the woman who is by the desk. Advanced  Syntactical 13 Complementation  *Swucengssi-ka khephi-lul cohahanta an-ta.  Swucenngssi-Hon-Nom coffee-Acc like-Dec know-Dec *4n:k7~˚V A:G™+Ò :á7ıG²%: 7ž%:. cSwucengssi-ka khephi-lul cohahanun kesul an-ta.  Swucenngssi-Hon-Nom coffee-Acc like-Pres-Mod it-Acc know-Dec 4n:k7~˚V A:G™+Ò :á7ıG®$ê ˚Ù9ı 7ž%:. I know that Sujeong likes coffee. Advanced  Syntactical 14 Passives  *Khunos-i cal phan-ta.  Big clothes-Nom well sell-Dec  *BÆ 8“9Ê 9î Ef%:. cKhunos-i cal phal-lin-ta.  Big clothes-Nom well sell-Pass-Dec  BÆ 8“9Ê 9î .  Big clothes sell well. Advanced  Syntactical 15 Causative  *Emeni-ka aki-lul can-ta.  Mother-Nom baby-Acc sleep-Dec * 7ı ƒ+Ò 9ê%:. cEmeni-ka aki-lul cay-wun-ta. Mother-Nom baby-Acc sleep-Caus-Dec   7ı ƒ+Ò . The mother puts the baby to sleep. Advanced  Syntactical  41  Table 1.  16 Honorifics  *Halapeci-kkeyse ppang-ul masisskey meknun-ta.  Grandfather-Nom-Hon bread-Acc deliciously eat-Dec-Hon * 0Ë9ı ,19Þ˚â ,‘$ê%:. c Halapeci-kkeyse ppang-ul masisskey tusin-ta.  Grandfather-Nom-Hon bread-Acc deliciously eat-Hon-Dec-Hon  0Ë9ı ,19Þ˚â '256%:. The grandfather eats bread delightedly. Intermediate Lexico, morpho-syntactical 17 Collocations  *Chwuw-ese  moca-lul sinnun-ta.  Cold-because hat-Acc wear-Dec *?ê9"3r ,þ9æ+Ò 56$ê%:. c Chwuw-ese  moca-lul ssun-ta.  Cold-thus hat-Acc wear-Dec ?ê9"3r ,þ9æ+Ò 7J%:. I put on a hat because it is cold. Advanced Lexical Note: C means the correct sentence with the problematic part(s) corrected. Nom means nominative. Acc means accusative. Dec means declarative. Clf means numeral classifier. Top means topic. Cop means copula. Conj means conjunctive. Pst means past. Fut means future tense. Neg means negative. Nominal means nominalizer. Mod means modifier. Rel means relative. Loc means locative. Pass means passive. Caus means causative. Hon means honorific. 42 2.2. Pilot study I This is the first of two pilot studies conducted before the main study. The purpose of this study, as a preparatory step before the main study, was to investigate the following: 1) the reliabilities of the test items, 2) the descriptive statistics of three groups with different language backgrounds, and 3) the results of a One-way ANOVA.     2.2.1. Methodology 2.2.1.1. Participants Thirty-six participants (nine L1 speakers, three HL learners, and 24 L2 learners) participated in the first pilot study; seven of them were males and 29 were females. Their average age was 23.48 years old, with ages that varied from18 to 37. HL learners were in third or fourth-year Korean classes, whereas L2 learners were either in first-year or second year Korean classes. They were studying Korean in the same U.S. institution. The average instruction period of L2 learners was 11.5 months.  2.2.1.2. Tasks An EIT was developed (Appendix A, 1). It contained 34 sentence stimuli (i.e., two statements for each of the 17 grammar structures). Half of them were grammatical, and half were ungrammatical. The participants listened to the stimuli. First, they were asked to think about the meaning and decide if they agreed with the statements. Then, they repeated the stimuli orally in correct Korean. Participants imitations were recorded. When the obligatory occasions were created and supplied correctly, the response scored one point. When the obligatory occasions were created, but not supplied or no obligatory occasions were created, the response scored zero points (Erlam, 2009).    Computer-delivered, written, timed and untimed GJTs were designed using E- 43 Prime 2 (Appendix A, 2). In the timed GJT, participants judged the grammaticality of each stimulus after reading each sentence, whereas in the untimed GJT, they read each sentence and judged grammaticality without time limits. Both GJTs contained the same 68 sentences out of which four sentences were judged for each of the 17 target structures. Half of the sentences were grammatical and half of them were ungrammatical. Participants were asked to choose among correct/incorrect/I dont know for each stimulus. The items were randomly distributed for each participant. Total reaction times were also recorded to identify potential differences between the three groups. For the oral EIT, and written GJT, the number of the words in each stimulus was over five Korean words. It was decided based on the pilot study in which several L2 learners imitated the sentences successfully when the word count was less than five. A computer-delivered metalinguistic knowledge test was designed (Appendix A, 3). For this test, participants read 17 ungrammatical sentences, one sentence for each structure, and selected the best explanation of the error out of three choices. The participants were encouraged to use their explicit knowledge about the target structures. A total percentage accuracy score was calculated out of the total number of points possible.  2.2.1.3. Bio-data questionnaire The following items of information were obtained through a bio-data questionnaire (Appendix A, 4): participants age of onset (AoO) of L2 learning, parents use of Korean at home, participants use of Korean at home, at school, and in social contexts, input/output amount and mode before and after five and after puberty, past and present learning environment, and Korean proficiency levels. 2.2.1.4. Procedure In this study, participants proceeded in order of: EIT, written timed GJT, written  44 untimed GJT, metalinguistic test, and, finally, bio-data questionnaire. After that, participants were asked to participate in a follow-up interview, in which they were asked if they had enough time to read and judge the items for the written timed GJT and their comments about the tests in general. Data collection was done individually.   Data collection for HL learners was conducted in the US. However, data for L2 learners was collected in Korea due to the limited pool of L2 learners in the US.   2.2.2. Results Table 2 demonstrates the reliability coefficient for each of the four tests. All of Cronbachs alpha coefficients were above .80, indicating that the tests were internally consistent. The item total statistics for each test are in Appendix C, Pilot study I. The focus is the Cronbachs Alpha values if each item is deleted.  Table 2. Reliability coefficients for the four tests Test Items Participants Reliability coefficient Oral imitation 34 36 .987 Written Timed GJT 68 36 .964 Written Untimed GJT 68 36 .952 Metalinguistic knowledge 17 36 .843   Table 3 shows the means and standard deviations of scores on the four measures by native speakers (NS), HL and L2 learners. Native speakers scored near ceiling (91% or higher), with very little variance on all tests. Their scores were higher than those of both the HL and L2 learners. The HL learner group outperformed the L2 learners on all measures.   45 Table 3. Accuracy scores for the four tests by group Tests (100%) NS (n=9) HL (n=3) L2 (n=24) M SD M SD M SD Oral imitation 95.8 3.3 69.6 13.3 11.8 12.5 Written Timed GJT  91.0 2.3 51.5 7.4 30.8 10.0 Written Untimed GJT  93.5 2.7 70.6 11.1 44.7 11.5 Metalinguistic knowledge  87.6 6.9 70.6 5.9 40.4 15.6  Participants scores on the four tests were subjected to separate one-way ANOVAs, with Group as the between-subject factor. There were statistically significant between-group differences on all four tests. Table 4 shows the results of the ANOVAs and post hoc analyses to determine the sources of the between-group differences. In comparison with the results in Bowles study (2011), two points are noteworthy. Firstly, the HL groups mean accuracy score on the untimed GJT was significantly higher than that of L2 learners. Secondly, L1 and HL groups showed significantly higher mean accuracy scores on the metalinguistic knowledge test than L2 learners. The majority of the L1 learners major, which was second language acquisition, and HL learners longer period of instruction than that of L2 learners can explain these results.    Table 4. Results of one-way ANOVAs Test F Results of post hoc analysis (Hochbergs GT2) Bowles study (2011) Oral imitation 201.49** L1> HL, HL > L2, L1 > L2 The same  Timed GJT 160.40** L1> HL, HL > L2, L1 > L2 The same Untimed GJT  78.47** L1> HL, HL > L2, L1 > L2 The same except HL = L2 Metalinguistic knowledge T  41.75**         HL > L2, L1 > L2 L1> HL, HL = L2, L1 = L2 Note. ** p < .01  46 The Kaiser-Meyer-Olkin value of the data was .79, exceeding the recommended value of .6, and Bartletts Test of Sphericity reached statistical significance, supporting the factorability of the data. Table 5 presents the correlational matrix for the four tests. All coefficients were .3 and above.   Table 5. Correlational matrix for the tests   Test Timed GJT Untimed GJT Metalinguistic T Oral imitation .71** .94** .88** Timed GJT --- .73** .64** Untimed GJT  --- .81** Note . **p < .01  Table 6 shows the eignenvalues of the two factors and Table 7 shows the pattern and structure matrix from a factor analysis to investigate the underlying structures of the four tests (Ellis, 2005). The results in Table 5 revealed the presence of two components with eigenvalues exceeding .4, explaining 84.0% and 10.0% of the variance in Korean knowledge. This two-component solution explained a total of 94.0 % of the variance. Even though the eigenvalue for the second factor was below 1.0 (.4), it accounted for a meaningful increase in the shared variance: 10%.   Table 6. Principal component factor analysis  Component Eigenvalue Variance Cumulative 1 3.36 84.05 84.05 2  .40  9.99 94.04  Table 7 demonstrates the pattern and structure matrix of the factor analysis with oblimin rotation. The metalinguistic test, EIT, and untimed GJT loaded heavily at .8 or  47 higher on Factor 1. The timed GJT loaded heavily at .97 on Factor 2. Factor 1 can be named explicit knowledge. Since the Eigenvalue of Factor 2 is less than 1, it is difficult to name the component implicit knowledge of Korean. Moreover, due to the small sample size of the data and low proficiency levels of the majority of L2 learners in this pilot study, it should be acknowledged that these results neither guarantee nor deny the construct validity of the four measurements for running a factor analysis in the main study.    Table 7. Loadings for principal component factor analysis  Test Pattern matrix Structure matrix  Component 1 Component 2 Component 1 Component 2 Meta knowledge  1.03 -.11 .95 .59 Oral imitation  .93  .07 .98 .71 Untimed GJT   .82  .19 .95 .75 Timed GJT  .03  .97 .70 .99  However, in the main study, I decided to use a confirmatory factor analysis (Bowles, 2011; Isemonger, 2007). This is to confirm that the data should fit the two-factor modeli.e., the oral EIT, and aural GJT are valid tests of implicit knowledgewhile the untimed GJT and metalinguistic test are good tests of explicit knowledge based on the previous studies (Bowles, 2011; R. Ellis, 2005).    2.2.3. Suggestions  It was expected that implicit knowledge measured through the oral EIT and timed GJT would be significantly greater than that of L2 learners (Bowles, 2011). However, explicit knowledge measured through the untimed GJT  48 would not be significantly different from that of L2 learners. In particular, scores on the meta-linguistic test would be marginally lower than that of L2 learners due to L2 learners instruction-oriented learning setting. AoO and input/output would be a significant factor for implicit knowledge. On the other hand, instruction would be a significant factor for explicit knowledge, especially on the meta-linguistic test (Roehr & Gutiérrez, 2009).   The expected discussions would revolve around two major arguments. Firstly, is similar in terms of being incomplete and unstable compared to that of L1 speakers, which would be borne out through comparisons between L1 speakers and HL/L2 learners knowledge representations. Secondly, HL is different in terms of implicit and explicit knowledge due to their different learning environments, which would point to the significance of HL learners natural setting in comparison to L2 learners formal instructional setting. In addition, the contribution of the two group different amount of input/output to their two types of Korean knowledge would also be discussed.                            2.3. Pilot study II y dissertation proposal defense, and as well as, further email exchanges and in-depth, face-to-face meetings, the following changes were made to the methodology of Pilot study I.  1.  A proficiency test was  proficiency in order to investigate RQ 3 thoroughly. Originally, proficiency levels were supposed to be measured through the level of instruction most recently received. However, in the process of recruiting HL learners for Pilot study II, the researcher realized that the majority of the candidates had not received  49 any formal instruction at the college level. As such, there was no way to measure their proficiency in Korean, so a general proficiency test was required.   2.  A narrative test was developed as an additional measurement of learners implicit knowledge. 3.  The timed GJT was modified into an Aural GJT as a more valid measure of implicit knowledge. 4.  The word count for each stimulus in the oral EIT, the Aural GJT, and the Written GJT was controlled rigorously. 5.  The problematic parts in the stimuli in the metalinguistic test were underlined for higher validity.  6.  To enhance reliability, an additional section was added to the previous metalinguistic test, in which participants were asked to find an example of each of the 17 target structures from a passage.  7.  The bio-data questionnaire was modified, highlighting age as an important  8.  Instead of running an Explanatory Factor Analysis, a Confirmatory Factory Analysis would be employed to investigate the validities of the tests in terms of implicit and explicit knowledge of Korean.  2.3.1. Methodology 2.3.1.1. Participants Ten participants (one male L1 speaker and nine HL learners) participated in the second pilot study; out of the nine HL learners, two were males and seven were females. Their average age was 21.7 years old, with ages that varied from18 to 28average instruction period was 0.9 months. Two of the nine HL learners received four  50 months of instruction at a college, and none of the others had formal instruction. 2.3.1.2. Tasks The oral EIT was modified (Appendix B, 1). The word count was controlled. To prevent participants from imitating without having proper knowledge, the word count for the grammatical sentences became five. This decision was made based on the I. However, the word count for ungrammatical sentences remained four.  An oral narrative test (Appendix B, 4) was designed to elicit the use of a number of the target structures such as particles, tenses, Sino numbers, noun classifiers, conjunctions, relativization, complementation, and honorifics. The participants read a story that was modified from the oral narrative test from the Marsden Project (Ellis, Loewen, et al., 2009). Participants were asked to read the text twice and then retell the story orally in three minutes. Their narratives were audio-recorded and transcribed. An obligatory occasion analysis was carried out to obtain the percentage of correct suppliance of each target structure (Ellis & Barkhuizen, 2005). Computer-delivered aural and written untimed GJTs were designed using E-Prime 2 (Appendix B, 2). The minimum word count for the oral EIT was four for ungrammatical sentences and five for grammatical sentences. Since the cognitive load that the participants would experience in the aural and written GJT might be less than that in the oral EIT, the word count for the stimuli in these GJTs was five or more. Non-obligatory adverbs were inserted in the proper positions of the sentences. In the aural GJT, participants judged the grammaticality of each stimulus after listening to each sentence (Bylund, Abrahamsson, & Hyltenstam, 2012; Granena, 2012) without time constraint, whereas in the written GJT, they read each sentence and judged the  51 grammaticality without time limits.  A two-section metalinguistic knowledge test was designed (Appendix B, 3). The first section is a modified version of the computer-delivered test from Pilot study I (Appendix B, 3, Section 1). Another fourth choice was added to each sentence-level stimulus and the problematic parts were underlined. The second section is a paper-based test. In this test, participants were asked to find examples of 17 specific target structures from the discourse-level text (Appendix B, 3, Section 2). A total percentage accuracy score was calculated out of the total number of points available.  2.3.1.3. Bio-data questionnaire A more age-oriented questionnaire was developed to investigate input and output levels for each age bracket (Appendix B, 5).   2.3.1.4. Proficiency test A C-test (Appendix B, 6) was adopted and modified for the proficiency test (Lee-Ellis, 200 means both 2009). Therefore, a desirable test should have two components of measurement: knowledge and the production of knowledge. It is judged that this modified C-test has these two components of measurement. The original C-test was rigorously investigated and found valid and reliable (Lee-Ellis, 2009). In the current study, for a logistical reason, the original five passages were decreased to four passages with comparably high reliability and validity.  Before this decision on the C-test was made, Test of Proficiency in Korean (TOPIK) was considered as a possible option. However, TOPIK has two critical disadvantages: 1) TOPIK takes much longer than the C-test, which makes it difficult for TOPIK to be utilized efficiently, and 2).the reading and listening sections in TOPIK are  52 formatted such that they consist of multiple-choice questions, which might not measure ficiency holistically. However the C-test has a component of production, which would measure participants receptive and productive proficiency levels more appropriately. Due to the C-tests more holistic aspect and logistic advantage, it was decided that the test would be utilized in this study.        2.3.1.5. Procedure The participants proceeded in order of: proficiency test, EIT, narrative test, aural GJT, written GJT, metalinguistic test and, finally, bio-data questionnaire. Data collection was done individually.    2.3.2. Results Table 8 alpha coefficients were above .80 except the first section of the metalinguistic knowledge test.   Table 8. Reliability coefficients for the four tests Test Items Participants Reliability coefficient Oral imitation 34 10 .884 Aural GJT 68 10 .872 Written (untimed) GJT 68 10 .848 Metalinguistic knowledge 1 17 10 .708 Metalinguistic knowledge 2 17 10 .887   Table 9 shows the descriptive statistics on the five measures by NS, and HL learners.   53 Table 9. Accuracy scores for the five tests by 10 participants  NS (n=1) HL (n=9) Test (100%) M SD M SD Proficiency test  95.0 - 34.9 12.0 Oral imitation  91.2 - 61.8 17.0 Narrative test  100 - 92.5 4.8 Aural GJT  94.1 - 69.3 11.3 Written (untimed) GJT 94.1 - 69.8 10.1 Metalinguistic knowledge 1   70.6 - 40.5 16.2 Metalinguistic knowledge 2  17.6 - 19.0 16.0  2.3.3. Suggestions  Table 10 highlights the comparisons between the results of Pilot studies I and II. Generally speaking, the Cronbach alpha coefficients became lower after Pilot study II. For the oral imitation as well as aural and written GJTs, it was decided that additional and non-obligatory adverbs should be readjusted for better reliability, making the minimum word count for the oral imitation, Aural and Written GJTs commonly four. Concerning the metalinguistic test 1, two items8 and 13were problematic. The ungrammatical sentences and the four choices were modified for higher reliability (Appendix B, 3, Section 1).  Table 10. Comparisons between Pilot studies I and II  Pilot study I  Pilot study II   Mean of HL learners  Mean of HL learners Oral imitation .987 69.6 .884 61.8 Timed /Aural GJT .964 51.5 .872 70.6 Untimed GJT .952 70.6 .848 69.8 Metalingustic test .843 70.6 .708(Section 1) .887(Section 2)  40.5  54 2.4. Main study  Based on the previous two pilot studies, the main study was conducted with a total of 114 participants.   2.4.1. Methodology 2.4.1.1. Participants Three groups of native speakers (n=11), HL learners (n=38), and L2 learners (n=65) participated in this study (N=114). 43 of them were males and 71 of them were females; their average age was 22.98, varying from 18 to 33. HL learners were categorized under two conditions: 1) HL learners born in the US and 2) both parents were Korean and provided opportunities for considerable amounts of input and output of Korean for implicit learning at home. The input/output amounts were measured through the questionnaire that asked about amount of input and output of Korean using a 100% scale. The instruction periods at college level for the HL learners vary from none to 22 months. The data collection was conducted in several college areas in Maryland, Virginia and California. Data collection for the HL learners lasted for approximately eight months in the United States, from Fall 2013 through Summer 2014.    Concerning L2 learners, participants were learners of Korean as a second language. Their minimum period of learning Korean as a second language was 10 months and maximum was 64 months in an instructional setting. All of the L2 learners L1 was Chinese to control for any discrepancies caused by their L1s and/or L3s in terms of word order, case markers including topic markers, frequent omission of words including plural marker, simple tense system, lack of relative clauses, and honorifics (Sohn, 1999). The data collection was conducted in the International Language Education Center (ILEC) affiliated to a national university in a metropolitan city in  55 Korea. The L2 learners either used to learn Korean or were learning Korean in the ILEC. The data collection lasted for approximately three months in Fall and Winter, 2014. The 11 native speakers of Korean were undergraduate students at three four-year universities in Busan, whose ages were comparable to the HL and L2 learners of Korean. Their majors varied, but none of the participants majors or minors were related to linguistics or languages.  2.4.1.2. Tasks The modified oral EIT from Pilot study II was used, which controlled the word count. In the test, the minimum word count for the majority of the grammatical sentences was five to keep participants from mimicking the sentences without processing the meanings and getting one point for the mere act of mimicking. However, the word count for the ungrammatical sentences was four because participants had to modify the incorrect sentences to get a score. The oral narrative test in Pilot study II was modified by altering difficult Korean expressions. However, the main ideas remained the same. Participants were asked to read the text twice carefully. Their recalling performances were audio-recorded and transcribed. An obligatory occasion analysis was carried out to obtain the percentage of correct suppliance of each target structure (Ellis & Barkhuizen, 2005). The computer-delivered aural and written untimed GJTs from Pilot study II were used. Since the cognitive load that the participants would experience in the aural and written GJT might be less than that in the oral EIT, the word count for the stimuli in these GJTs was five or more. To regulate the meanings and word counts of the sentences in both GJTs, if necessary, non-obligatory adverbs were inserted in the proper positions. In the aural GJT, participants judged the grammaticality of each stimulus after listening to each sentence, whereas in the written GJT, they read each sentence and judged the  56 grammaticality. Participants took both GJTs without time limits.  The two-section metalinguistic knowledge test from Pilot study II was translated in Chinese for Chinese L2 learners of Korean by a Chinese PhD student. Her major was Korean language and linguistics in the national university where the data collection was conducted. The original English version was used for the HL learners and the Chinese version was used for the L2 learners of Korean. For the native speakers of Korean, the researcher developed a Korean version of the metalinguistic knowledge test. The first section is a modified version of the computer-delivered test from Pilot study II with the fourth choice added to each stimulus and with problematic parts underlined. The second section is a paper-based test in which the three groups of participants were asked to find examples of 17 specific target structures from the text in Korean. The instructions were in English, Chinese, or Korean. For each section, an accuracy score was calculated out of the total number of points available and the scores for both sections were summed up. Finally, raw scores were converted into a 100% scale.   The proficiency test was analyzed based on the Korean morphemes in the text, i.e., the smallest unit that has meaning (Ko & Ku, 2008). The total number of morphemes was decided to be 172 after the researcher consulted with the co-rater in an exchange of emails and several face-to-face meetings. Each correctly answered morpheme was given one point and the total point for each participant was converted into a 100% scale.   2.4.1.3. Bio-data questionnaire The modified version of questionnaire from Pilot study II was used. For Chinese L2 learners, a Chinese graduate student whose major was Korean literature and linguistics developed a Chinese version.    57 2.4.1.4. Procedure Participants proceeded individually in order of: the proficiency test (C-test), EIT, narrative test, aural GJT, untimed written GJT, metalinguistic test, and finally bio-data questionnaire. This order is the same as the instruments in previous studies (Ellis, 2005; Bowles, 2011) except the proficiency test. Like any study in which participants take a couple of tests in a fixed sequence, potential test order effects were unavoidable in the current study. Considering the fact that this study is about implicit and explicit knowledge and learners different levels of awareness of target structures, the current orderi.e., from the least awareness to the most awarenessmight be better than the the other way around. However, ideally speaking, counterbalancing the tests would have dealt with this test order effect (S. Loewen, personal communication, June 6, 2016).             The data collection time for each individual varied, but it lasted as little as one and a half hours for native speakers of Korean and as much as two and a half hours for either HL learners or L2 learners. At the end of the data collection, all participants were paid 25.00 in U.S. dollars or 25,000 in Korean won.    2.4.1.5. Analyses  The reliabilities of the five tests and more detailed tests were calculated using Cronbach alpha. For examining the interrelationships between the five tests, Pearson product moment coefficients were run. For research question 1, which asked about implicit and explicit knowledge of the participants and about the model fitting of the data, four major CFAs were carried out as follows: 1) implicit/explicit model with the written GJT as an observed variable for explicit knowledge (Bowles, 2011), 2) implicit/explicit model with the written ungrammatical GJT as an observed variable for explicit knowledge (Ellis, 2005; Philp, 2009; Zhang, 2015), 3) grammatical/ ungrammatical model (Gutiérrez, 2012), and 4) one-factor model with explicit  58 knowledge as the only latent variable (DeKeyser, 2003). For research question 2, which asked about the three groups knowledge assessed using the EIT and the aural GJT, One-way ANOVAs were employed with the aural/oral elicited imitation and aural GJT both separately and combined. For research question 3, which asked about the three groups knowledge assessed using the written GJT and the metalinguistic knowledge test, One-way ANOVAs were run with the ungrammatical written GJT and metalinguistic test both separately and combined.       Table 11 shows the Cronbachs alpha coefficients for the battery of tests.   Table 11. Reliability coefficients for the five tests Test Items Participants  Reliability coefficient Aural/oral EIT 34 114 .94 Narrative test - 114 r=.98 Aural GJT 68 114 .85 Written GJT  68 114 .87 Ungrammatical  34 114 .87 Metalinguistic knowledge test   1 17 114 .75   2 17 114 .85 Total 34 114 .78 Proficiency (C-test) 172 114 (38) r=.99 Note. Ungrammatical means the ungrammatical items in the written GJT.  All of the reliability coefficients were above .80 except of the metalinguistic knowledge test. Concerning the narrative test, the researcher rated the test twice. The interval between the two ratings was one and a half years. The intra-rater reliability was r=.98. In terms of the proficiency test, due to logistical reasons, only part of the total  59 participants C-test scores were used for the inter-rater reliability. 38 C-test scores (33.3%) out of the total 114 obtained from Rater 1 were randomly selected and compared with the selected 38 scores from Rater 2 using the intraclass correlation coefficient. The inter-rater reliability was r=.99.   60 CHAPTER 3. RESULTS  3.1. Descriptive statistics Table 12 presents the results of descriptive statistics of the battery of tests from all three groups.   Table 12. Descriptive statistics for the five tests by group  NS (n=11) HL (n=38) L2 (n=65) Test (100%) M SD M SD M SD Aural/oral EIT  94.65 3.68 58.20 19.95 46.02 15.96 Narrative test 95.11 1.66 76.80 11.61 56.05 12.37 Aural GJT  93.85 5.50 66.60 11.60 64.66 9.78 Written GJT  95.86 1.59 66.56 11.65 70.61 10.19 Written Ungra GJT  96.49 3.20 55.96 18.33 63.62 16.81 Metalinguistic knowledge test  1  84.49 7.57 48.40 18.12 66.79 27.87 2  36.36 14.60 26.12 14.74 47.91 19.02 Total 60.43  9.43  37.15  14.24  57.35  20.37 Proficiency (C-test)  96.59 1.31 46.35 12.83 60.11 17.02  To compare the three groups general proficiency of Korean, a one-way ANOVA was run using the scores of the C-test (Lee-Ellis, 2009). This written-mode test measures the participants general proficiency and involves a controlled production aspect. A post-hoc test revealed that the native speakers proficiency was significantly different from those of HL and L2 learners. Moreover, L2 learners proficiency was significantly higher than that of HL learners, as demonstrated in Table 13.    61 Table 13. Results of one-way ANOVA on the proficiency test Test F (p value) Results of post hoc analysis Proficiency (C-test) 48.86 (p=.000) **NS > L2, **NS > HL, **L2 > HL    3.2. Construct validity of the tests Research question 1 was; how well does the battery of tests (EIT, narrative test, aural GJT, written GJT, metalinguistic test, and C-test) measure implicit and explicit linguistic knowledge? Table 14 presents the correlation matrix for HL and L2 learners scores on the battery of tests.   Table 14. Correlational matrix for the six tests  Narrative  Aural GJT  Written GJT  Written Ungra GJT  Metalinguistic knowledge test  Proficiency test (C-test) Aural/oral EIT .723** .696** .616* . 495** .326** .496** Narrative  - .506** .333** .301** .089 .216* Aural GJT   - .811** .639** .475** .642** Written GJT    - .804** .656** .741** Written Ungra GJT     - .577** .698** Metalinguistic knowledge test      - .711** Note. ** means Correlation is significant at the 0.01 level (2-tailed). * means Correlation is significant at the 0.05 level (2-tailed).  All of the tests were significantly correlated with each other at the 0.01 level. The written GJT was highly correlated with the aural GJT, r=.811, whereas the  62 ungrammatical items in the written GJT were moderately correlated with the aural GJT, r=.639. The metalinguistic knowledge test was weakly correlated with the aural/oral EIT, r=.326. Likewise, the proficiency test was moderately correlated with the aural/oral EIT, r=.496. However, the narrative test was not related to the metalinguistic test. It is noteworthy that the written GJT was highly correlated with the proficiency test, r=.741, and the metalinguistic test was also highly correlated with the proficiency test, r=.711.  Dozens of confirmatory factor analyses (CFA) were run using AMOS version 23.0.0, either with a two-factor model or with a one-factor model. Depending on the numbers of the latent factors and observed variables, the results were divided into four categories: four observed variables (the EIT, aural GJT, written GJT, and metalinguistic test), five observed variables with the narrative test in addition to the four observed variables, five observed variables with the C-test in addition to the four observed variables, and six observed variables with the narrative test and C-test in addition to the four observed variables.  3.2.1. Two-factor model with four observed variables (OVs): Implicit-explicit knowledge model using the EIT, aural GJT, written GJT and metalinguistic test  3.2.1.1. Default model In this model, two different modesaural and writtenwere highlighted. The two latent variables were labeled implicit and explicit knowledge (Bowles, 2011; Ellis, 2005, 2009; Kim & Nam, 2016; Spada et al., 2015). For implicit knowledge, the observed variables were the aural/oral EIT and aural GJT. For explicit knowledge, the observed variables were the written GJT and metalinguistic test. The results of the analysis are presented in Figure 1, which is titled implicit-explicit model with written GJT. The summary statistics for the model fit are = .410, df=1, p= .522, normed fit  63 index (NFI) = .998, root mean square error of approximation (RMSEA)= .000.      Figure 1. 2 Factor 4 observed variables (OVs), default model: Implicit-explicit model with written GJT  At a glance, the indices seem good enough as evidence for a good model fitting. However, as Figure 1 demonstrates, the estimate of the standardized regression weight of the whole written GJT for explicit knowledge is 1.064, which is above 1. This makes the squared multiple correlation of written GJT above 1, at 1.132. Moreover, e3, the error variance of the whole written GJT is negative at -15.464. Based on these problematic Heywood caseswhich could potentially be caused by a combination of a small sample size and only two indicators per factor, nonidentification of the model, or the presence of outliers (Kline, 2015, p. 158)it was decided that this default model be modified based on previous studies (Brown, 2015; Byrne, 2010; R. Ellis, 2005). Rival model 1 is the modified model of the default model.      64 3.2.1.2. Rival model 1: written ungrammatical GJT  Figure 2 illustrates the results of the CFA of implicit-explicit model with written ungrammatical GJT as an observed variable of explicit knowledge. The summary statistics for the model fit are = .48, df=1, p= .49, NFI = .997, RMSEA = .000. The low nonsignificant  value points to a good fit for this model. The NFI, which is greater than .95, indicates a good model fit. The RMSEA, an important parsimonious index, finally indicates that this model fits the current set of data very well. Even though the standardized regression weight for the metalinguistic test was .65, which was slightly less than the desirable level of .7 (Byrne, 2010), this model fits the data very well (Brown, 2015; Byrne, 2010)  Figure 2. 2 Factor 4 OVs Rival model 1: Implicit-explicit model with ungrammatical items in written GJT     65 3.2.2. One-factor model with four observed variables: Explicit knowledge model using the written GJT, metalinguistic test, EIT, and aural GJT  3.2.2.1. Default model The results from a CFA using a one-factor model with the four observed variables are presented in Appendix E (See 1F-4-Default. 4 OVs Default Explicit). All of the tests were grouped into the only latent variable, explicit knowledge. The summary statistics for this model fit are = 20.573, df=2, p= .000, NFI = .914, RMSEA = .302, which evidence the poor fit of this model.  This explicit-only model produced modification indices that recommended covarying two pairs of error terms: 1) EIT and aural GJT, and 2) written GJT and metalinguistic test. Table 15 demonstrates the details of the two pairs of error terms whose modification indices are above 9, and that have theoretical relevancy in terms of aural and written modes.    Table 15. Modification indices from the one-factor model Pairs of error terms M.I. Par Change e3    e4 (imitation test  aural GJT) 11.430 27.138 e1    e2 (written GJT  metalinguistic test) 9.228 22.155  Based on the results in Table 15, two more rival CFAs were run. First, the one-factor model was employed with the error terms of imitation and aural GJT covaried. Second, another one-factor model was run with the error terms of written GJT and metalinguistic test covaried. Figure 3 demonstrates the covaried model between the error terms of the imitation test and aural GJT.   66  Figure 3. 1 Factor 4 OVs Rival 1: One-factor model of explicit knowledge with one set of covaried error terms between imitation test and aural GJT  The summary statistics for the model fit are = .410, df=1, p= .522, NFI = .998, RMSEA = .000. Figure 3 demonstrates that the estimate of standardized regression weight of the written GJT for explicit knowledge is abnormal at 1.064, which is above 1. Moreover, e1, the error variance of the whole written GJT is negative at -15.464. Thus, it was decided that this rival model is not admissible (Brown, 2015; Byrne, 2010). Figure 4 demonstrates the covaried model between the error terms of written GJT and metalinguistic test.   67  Figure 4. 1 Factor 4 OVs Rival 2: One-factor model of explicit knowledge with one set of covaried error terms between written GJT and metalinguistic test  The summary statistics for the model fit are = .410, df = 1, p = .522, NFI = .998, RMSEA= .000. However, in this model, the standardized regression weight of the metalinguistic test was .49, which is much lower than the desirable level of .7, and even lower than that of .65, which was the standardized regression weight of the metalinguistic test for the Rival model 1 (See Figure 2).       68 3.2.3. Two-factor model with five observed variables: Implicit and explicit knowledge model using the imitation test, aural GJT, written GJT, metalinguistic test, and C-test Since the C-test has literacy and form-focused elements like the written GJT and the metalinguistic test, the researcher decided to use the C-test as another observed variable, which was expected to load on explicit knowledge. Figure 5 demonstrates the the five observed variables of the expanded version of the previous model illustrated in Figure 1.  Figure 5. 2 Factor 5 OVs C-test Rival model 1: Implicit vs. explicit model with two sets of covaried error terms   However, in Figure 5, based on the recommendation of modification indices of the two-factor and five-observed-variable model with the C-test (See Appendix E. 2F-5-C-test 5 CVs Default), two sets of error terms were correlated between 1) the written GJT and metalinguistic test, and 2) the metalinguistic test and C-test. The summary statistics for the model fit are = .66, df = 1, p = .42, NFI = .998, RMSEA = .000.  69 These indices support that this is a good model for the data set.  Figure 6. 2 Factor 5 OVs C-test Rival model 2: Implicit vs. explicit model with ungrammatical items in written GJT  Figure 6 demonstrates the results of the CFA with the same five observed variables as the model in Figure 5. This time, however, instead of whole items in the written GJT, the ungrammatical items in the written GJT were utilized. The summary statistics for the model fit were = 7.749, df=4, p= .101, NFI = .972, RMSEA = .096. Since the RMSEA index was not statistically significant, it was decided that two sets of error terms would be correlated, following the recommendation indices. The results are presented in Figure 7.   70  Figure 7. 2 Factor 5 OVs C-test Rival model 3: Implicit vs. explicit model with ungrammatical items in written GJT with two covaried error terms between written GJT ungrammatical items and metalinguistic test, metalinguistic test and C-test   Following the recommendation of modification indices from Figure 6, two sets of error terms were correlated between: 1) the ungrammatical items in written GJT and metalinguistic test, and 2) the metalinguistic test and C-test. The fit indices became much better. Figure 7 shows that the results of the summary statistics for the model fit were = .677, df = 1, p = .411, NFI = .998, in comparison with .972 in Figure 6, and RMSEA = .000, in comparison with .096 in Figure 6. These indices confirm that this model fits the data set very well.    71 3.2.4. Two-factor model with six observed variables: Implicit and explicit knowledge model using the imitation test, aural GJT, narrative test, written GJT, metalinguistic test, and C-test  Figure 8. 2 Factor 6 OVs Rival model 2: Implicit vs. explicit model with ungrammatical items in written GJT, three sets of covaried items  The researcher decided to add the narrative test as another observed variable to the two-factor and five-observed-variable model with the C-test, because the narrative test has orality and meaning-focused elements. It was expected that the narrative test would load on implicit knowledge. Figure 5 demonstrates the results of the CFA adding the narrative test to the previous five variables in Figure 6. Based on the recommendation of modification indices, three sets of error terms were correlated  72 between: 1) the imitation test and narrative test, 2) the Aural GJT and narrative test, and 3) the metalinguistic test and C-test. In Figure 8, the summary statistics for the model fit are = 8.323, df = 7, p= .305, NFI = .986, RMSEA = .019.  Figure 9. 2 Factor 6 OVs Rival model 6: Implicit (orality+grammatical) vs. explicit (literacy+ungrammatical) model   Finally, the researcher decided to combine two elements in a CFA: grammaticality and modality. It was expected that the narrative test and the grammatical items in the imitation test and Aural GJT would load on implicit knowledge, whereas the ungrammatical items in the Written GJT, metalinguistic test and C-test would load on explicit knowledge. In Figure 9, the summary statistics for the model fit are = 11.339, df = 8, p = .183, NFI = .953, RMSEA = .064. The RMSEA index was above the desirable level of .05 (Brown, 2015).   73  Figure 10. 2 Factor 6 OVs Rival model 6: Implicit (orality+grammatical) vs. explicit (literacy+ungrammatical) model with ungrammatical items in writtenGJT, one set of covaried error terms    Following the recommendation of modification indices from Figure 9, two error terms were correlated between the metalinguistic test and C-test. The fit indices became better. In Figure 10, the statistics for the model fit are = 8.323, df = 7, p = .305, NFI = .966, RMSEA = .043. In this model, the standardized regression weight of the grammatical items of the aural GJT was .42, which is much lower than the desirable level of .7. Table 16 summarizes the results of all the CFAs that were employed; the summary of the important indices from all CFAs is in Appendix E.  74 Table 16. Summary of the important indices from the major CFAs   Observable  variable (O.V.)  Factor        Models INDEX   NFI ( .90) RMSEA (.00RMSEA  .05)  (p  .05) df CMIN/DF (0CMIN/DF2) AIC (smaller than comparison model) 4 OVs  EIT,  Aural GJT, Written GJT, Metalinguistic 2 Figure 1. 2F-4-Default. 4 OVs Default Implicit vs. explicit .998 .000 0.410 (p=.522) 1 .410           Heywood case 26.410 Figure 2. 2F-4-Rival 1. 4 OVs Ungrammatical items in written GJT .998 .000 0.483 (p=.487) 1 .483 26.483 1 Figure 3. 1F-4-Rival-1. 1 Covaried error terms btw imitation and aural GJT .998 .000 0.410 (p=.522) 1 .410 Heywood case 26.410 Figure 4. 1F-4-Rival-2. 1 Covaried error terms btw written GJT and metalinguistic test .998 .000 0.410 (p=.522) 1 .410 26.410 5 OVs   EIT Aural GJT Written GJT Metalinguistic C-test 2 2F-5-C-test-Default 5 CVs Default: Implicit vs. explicit .927 .269 25.201 (p=.000) 3 8.400 59.201 Figure 5. 2F-5-C-test-Rival-1. 2 Covaried error terms btw written GJT and metalinguistic test, metalinguistic test and C-test .998 .000 .656 (p=.418) 1 .656 38.656 Figure 6. 2F-5-C-test-Rival-2. Ungrammatical items in written GJT .972 .096 7.749 (p=.101) 4 1.937  41.749 Figure 7. 2F-5-C-test-Rival-3. 2 Covaried error terms btw written GJT_ungra and metalinguistic test, metalinguistic test and C-test .998 .000 0.677 (p=.411) 1 .677 38.677 6 OVs  EIT Narrative test Aural GJT Written GJT Metalinguistic C-test 2 2F-6-Rival-2. Ungrammatical items in written GJT .873 .217 46.283 (p=.000) 8 5.783 84.283 Figure 8. 2F-6-Rival-3. Ungrammatical items in written GJT. 3 Covaried error terms btw aura GJT and narrative, narrative and imitation, metalinguistic and C-test .986 .019 5.191 (p=.393) 5 1.038 49.191 Figure 9. 2F-6-Rival-5. Orality+grammatical vs. Literacy+ungrammatical  .953 .064 11.339 (p=.183) 8 1.417 49.339 Figure 10. 2F-6-Rival-6. Orality+grammatical vs. Literacy+ungrammatical. 1 covaried error terms btw metalinguistic test and C-test .966 .043 8.323 (p=.305) 7 1.189 48.323   75 Table 17. Chi-squared difference test results for models with overall good model fit indices Observed variables Model  df  Difference btw dfs p 4  Imitation Aural GJT Written GJT Metalinguistic test 1F-4OVs-Rival 1 0.410 1 - - - 1F-4OVs-Rival 2 0.410 1 0 0 P>.05 2Fs-4OVs-Default 0.410 1 0 0 P>.05 5  (4 OVs + C-test)    Whole items in written GJT 2Fs-5OVs-Default 25.201 3 - - - 2Fs-5OVs-Rival 1 0.656 1 24.545     2 P<.05 Ungra. items in written GJT 2Fs-5OVs-Rival 2 7.749 4 - - - 2Fs-5OVs-Rival 3 0.677 1 7.072 3 P>.05 6   (5 OVs+ Narra-tive test)  Ungra. items in written GJT 2Fs-6OVs-Rival 2 46.283 8 - - - 2Fs-6OVs-Rival 3 5.191 5 41.092 3 P<.05 Oral/ Aural+Gra.  Vs. Lit.+ Ungra. 2Fs-6OVs- Rival 5 11.339 8 - - - 2Fs-6OVs-Rival 6 8.323 7 3.016 1 P>.05 Note. OV means observed variable, and F means factor.  Models 1 Factor (F) -4-Rival 1, 2 F-5-Rival 2, and 2 F-6-Rival 5 were compared to the other nested models with acceptable fit indices in each category of the observed variables. Formal chi-squared difference tests were conducted by 2 (df1-df2) = 2df1-df2  and were distributed as a chi-squared distribution with df=df1-df2 (Brown, 2015). Table 17 shows the results produced by the chi-squared difference tests. The tests demonstrate that two models were statistically better than their comparison models. First, in the category of the 2-factor model with five observed variables, (i.e., EIT, aural  76 GJT, written GJT, metalinguistic test, and C-test), the Rival 1 model was statistically better than the default model, (2)=24.545, P<.05. The Rival 1 model has two sets of error terms between: 1) written GJT and metalinguistic test, and 2) metalinguistic test and C-test; whereas the default model does not have covaried error terms. Second, in the category of 2-factor models with six observed variables (i.e., EIT, narrative test, aural GJT, ungrammatical items in the written GJT, metalinguistic test, and C-test), the Rival 3 model was statistically better than the Rival 2 model, (3)=41.092, P<.05. The Rival 2 model is the default model that has the ungrammatical items in the written GJT, whereas the Rival 3 has three sets of error terms between: 1) aural GJT and narrative test, 2) narrative test and imitation, and 3) metalinguistic test and C-test. However, the chi-squared difference test results should be interpreted only within a category of models with the same observed variables, since this test is based on nested models (Brown, 2015). Therefore, the formal chi-squared difference test results do not provide all of the necessary information to select the best model across models with different numbers of observed variables. The next step was to examine the model fit indices via factor loadings of the observed variables in order to decide the best model. The results are in Tables 18 through 20.    Table 18. Factor loadings of models with good model fit indices (4 observed variables) Model Imitation Aural GJT Written GJT Written GJT Ungra Metalinguistic  2F-4-Rival 1 .73 .96 - .89 .65 1F-4-Rival1 .58 .76 1.06 - .62 1F-4-Rival2 .73 .95 .85 - .49 Note. All of the factor loadings are significant at the .001 level.  77  When 2 F-4-Rival 1 and 1 F-4-Rival 2 are compared, 2 F-4-Rival 1 is judged to have a better fit. First, the factor loading of the metalinguistic test in 1 F-4-Rival 2, .49, is below the desirable level, whereas that of 2 F-4-Rival 1 is good enough at .65. This difference means that the metalinguistic test in 2 F-4-Rival 1 explains 42% of the variance in the explicit knowledge construct, whereas in 1 F-4-Rival 2, the test explains only 24%. This undesirably low factor loading of the metalinguistc test, which is less than .5, could result in removing the factor from the path diagram, which worsens the results of the CFA. Thus, 2 F-4-Rival 1 has a better set of factor loadings on explicit knowledge than 1 F-4-Rival 2. Second, discriminant validity of 2F-4-Rival 1 is better than that of 1 F-4-Rival 2. Concerning factors, 2F-4-Rival 1 presents the two latent variables using two factors for each. By comparsion, in 1F-4-Rival 2 model, the same four variables are grouped into one factor. When other things are equal, a model with fewer observed variables per factor has a higher fit than a model with more indicators per factor (MacCallum et al., 1996). The reason is that more observed variables per factor provide a more powerful, precise test than a rival model with fewer observed variables (MacCallum et al., 1996). 1F-4-Rival 2 has four variables for a factor, whereas 2F-4-Rival 1 has two variables for each factor of the two. Third, theoretically a one-factor model cannot explain the current data of HL learners. If the one-factor is labeled explicit knowledge, there is not much explanation left for the 38 HL learners, who must have relatively little explicit knowledge, but also considerable implicit knowledge due to their early AoO, and also a considerable amount of input/output of Korean at home and in their social communities. Moreover, considering the fact that the 15 HL learners out of the total 38 had not received any Korean instruction at an institute, the one-factor  78 model labeled explicit knowledge looks problematic to explain this set of data.   Table 19. Factor loadings of models with good model fit indices (6 observed variables)  Model Imita-tion Aural GJT Narrative test Written GJT C-test Meta-linguistic  Imitation_ Gra AuralGJT_Gra Written_ GJT Ungra 2F-6-Rival 6 - - .66 - .80 .65 .96 .42 .88 2F-6-Rival 3 .73 .95 .37  .68 .86 - - .82 Note. All of the factor loadings are significant at the .001 level.   Both models demonstrate a small factor loading of less than .6. The aural GJT_Grammatical items factor loading in 2F-6-Rival 6 on implicit knowledge is .42. The narrative tests factor loading in 2F-6-Rival 3 on implicit knowledge is .37, which is lower than .6.  Table 20. Factor loadings of models with good model fit indices (5 observed variables with C-test) Model Imitation Aural GJT Written GJT Written GJT Ungra Metalinguistic C-test 2F-5-Rival3 .74 .95 - .83 .61 .84 2F-5-Rival1 .73 .95 .97 - .56 .77 Note. All of the factor loadings are significant at the .001 level.  2F-5-Rival 3 shows a better set of factor loadings on implicit and explicit knowledge than 2F-5-Rival 1, especially concerning the factor loadings of the metalinguistic test on explicit knowledge. Therefore, 2F-5-Rival 3 is a better model of the data than 2F-5-Rival 1. When 2F-4-Rival 1 with the four observed variables is compared with 2F-5-Rival 3 with the additional C-test, 2F-5-Rival 3 is a more comprehensive model. 2F-4- 79 Rival 1 has four observed variables, whereas 2F-5-Rival 3 has five in total, including the C-test. When the data set is explained with a more complex five-variable model, choosing a simpler rival model with one less observed variable could lead to ignoring the significance of discriminating power, which in turn may lead to explaining the data set in an underidentified way. In other words, 2F-5-Rival 3 provides more information than 2F-4-Rival 1, which gives a more comprehensive explanation of the two-factor model by adding the C-test as another valid observed variable.  Another reason for selecting 2F-5-Rival 3 as the best model is that the two sets of covaried error terms in the model can be explained substantiallytheoretically or conceptually. Usually, methodological effects cause covaried error terms (Brown, 2015). In CFA construct validation studies, covaried error terms are necessary to explain method covariance, such as in the analysis of observed variables obtained from different assessment modalities (Brown, 2015). In a questionnaire of multiple items, the effects are associated with similarly-worded, or reverse-worded items. From this perspective, two questions arise in the current study: 1) why and how are the errors not associated with the latent variable, explicit knowledge? and 2) why are they correlated? For the first question, the two sets of error terms of the writtenGJT_Ungrammatical items, metalinguistic test, and C-test are not associated with explicit linguistic knowledge due to their additional meaning-oriented reading process as well as form-focused processing. Since the participants had unlimited time for the tasks, they could have focused on meaning for comprehension as well as forms for grammatical accuracy. The latent variableexplicit knowledgeunderlying the observed variable might not have explained their meaning-focused process and knowledge (R. Ellis, 2005; 2009). For the second question, concerning their correlations, covarying error terms means that the  80 involved indicators are similar in some aspects (Brown, 2015). The written GJT_Ungrammatical items and metalinguistic test are correlated arguably due to their sentence-level reading, whereas the metalinguistic test and C-test are correlated due to their paragraph-level reading. In fact, the metalinguistic test has two sections: sentence-level processing and discourse-level processing. In the sentence-level processing, the participants were instructed to choose the best explanation of the ungrammatical part in the sentence. In the discourse-level processing, they were asked to find examples of 17 target structures from the discourse-level text. Sentence-level meaning-oriented reading involves lexical access and from-clause-to-sentence integration using syntactic information, whereas discouse-level meaning-oriented reading, in addition to the previous processing for sentence-level processing, involves genre familiarity, discourse style, internal and external coherence, and building a macrostructure (Danks & End, 1987). The difference could have resulted in two separate sets of correlations: 1) one between error terms of the written GJT and metalinguistic test from sentence-level processing, and 2) the other between error variances of the C-test and metalinguistic test from discourse-level processing. Therefore, compared to the other models, 2F-5-Rival 3 model (Figure 10) was determined to be the best model based on the good model-fit indices and factor loadings.   3.3. Implicit knowledge  Research question 2 was; do the groups differ on the EIT and the aural GJT?  To answer this research question, the values of two observed variables (i.e., aural/oral imitation and aural GJT for implicit knowledge), were summed up and a one-way ANOVA was employed for the three groups.   81   Table 21 shows the results of the one way ANOVAs for the three groups concerning aural/oral EIT and aural GJT, both separately and combined.  Table 21. Results of one-way ANOVAs on test scores for implicit knowledge Test F (p value) Results of post-hoc analyses Aural/oral imitation 40.82 (p=.000) **NS > HL, **NS > L2, **HL > L2 Aural GJT 39.65 (p=.000) **NS > HL, **NS > L2   Aural/oral imitation and Aural GJT 46.02 (p=.000) **NS > HL, **NS > L2, *HL > L2 Note. The symbol > indicates that the first group scores are significantly higher than those of the second group. One asterisk means that p < .05, and two asterisks mean that p < . 001. Depending on the results from the tests of homogeneity of variances for each test, appropriate post-hoc tests were employed (Field, 2009).   For the aural/oral EIT and aural GJT separately and combined, native speakers showed significantly higher scores than HL or L2 learners. Concerning the aural/oral EIT, HL learners knowledge was significantly larger than that of L2 learners. However, in the aural GJT, HL learners mean score was not significantly different from that of L2 learners. When the two tests scores were combined, HL learners demonstrated a significantly greater implicit knowledge than L2 learners did.     3.4. Explicit knowledge Research question 3 was; do the groups differ on the Written GJT, the Metalinguistic test, and the C-test?  The three variables for explicit knowledge, which are written ungrammatical GJT, metalingustic test, and C-test, were summed up and a one-way ANOVA was run for the three groups. Table 22 demonstrates the results.   82 Table 22. Results of one-way ANOVAs on test scores for explicit knowledge Test F (p value) Results of post-hoc analyses Written Ungrammatical GJT  25.51 (p=.000)  *NS > HL, **NS > L2,  Metalinguistic test  21.36 (p=.000) **NS > HL, **L2 > HL,   C-test 48.86 (p=.000) *NS > L2, **NS > HL, **L2 > HL Written Ungrammatical GJT and Metalinguistic test 25.47 (p=.000)  *NS > L2, **NS > HL, **L2 > HL Written Ungrammatical GJT, Metalinguistic test, C-test 41.47 (p=.000)  *NS > L2, **NS > HL, **L2 > HL Note. The symbol > indicates that the first groups scores are significantly higher than those of the second group. One asterisk means that the significance is p < .05, and two asterisks means that p < . 001. Depending on the results from the tests of homogeneity of variances for each test, appropriate post-hoc tests were employed (Field, 2009).    For the written ungrammatical GJT and metalinguistic test both separately and combined, native speakers demonstrated significantly higher results than the other two groups. Concerning the written ungrammatical GJT, HL  knowledge was not significantly different from that of L2 learners. However, on one hand, concerning the metalinguistic test, L2 learners knowledge was significantly greater than that of HL learners. On the other hand, native speakers metalinguistic knowledge was not significantly different from that of L2 learners. When both tests were combined, L2 learners explicit knowledge was significantly greater than that of HL learners. When the C-test was added, L2 learners explicit knowledge was significantly greater than that of the comparison group.  To fully understand the relationships between the tests and the learner groups in terms of the two types of linguistic knowledge, a discriminant function analysis (DFA)  83 was employed. The purpose was to see how the tests, in context of two functions, could discriminate between the native speakers, HL learners, and L2 learners. Table 23 demonstrates the summary of the results.   Table 23. Summary of discriminant functions Function Eigenvalue % of Variance Cumulative %          Canonical correlation  Wilks Lambda  Chi-Square p 1 1.727 64.7 64.7 .796  .189  181.69 .000 2 .942 35.3 100 .696  .515  72.35 .000  The DFA identified two discriminant functions. Function 1 accounted for 64.7% of the total variance, canonical R2 =.63, and Function 2 accounted for 35.3% of the total variance, canonical R2 =.48. In combination, Functions 1 and 2 significantly differentiated HL and L2 groups,  = 0.189, 2(10) = 181.69, p =.000. Removing Function 1 also indicated that Function 2 significantly differentiated HL and L2 learner groups,  = 0.52, 2(4) = 72.35, p =.000.  Table 24 presents the standardized canonical discriminant function coefficients.    Table 24. Standardized canonical discriminant function coefficients Factor Function 1 Function 2 Imitation 1.075 -.274 Aural GJT .595 .201 Written ungrammatical GJT -.099 .101 Metalinguistic test -.846 -.229 C-test -.625 1.066  The EIT, along with the aural GJT, had the opposite effect in comparison to the written ungrammatical GJT, metalinguistic test, and C-test concerning Function 1,  84 which differentiated the EIT/aural GJT from the written ungrammatical GJT/metalinguistic test/C-test. Comparatively, the EIT had a negative relationship with Function 2, whereas the written ungrammatical GJT and C-test showed positive relationships. Even though the aural GJT loaded on both Functions positively, the loading for Function 1 is stronger than that for Function 2, which was expected. In contrast, the metalinguistic test loaded on both Functions negatively. The researcher expected that there would be a negative relation of the test to Function 1 and a positive relation to Function 2. However, the magnitude of the negative loading to Function 1 is greater than that to Function 2. The small and negative loading to Function 2 may be related to the tests smallest loading to the explicit knowledge in the CFA model (See Table 20). Thus, these arguably evidenced the validity of this DFA. Based on the results, with due caution, the researcher suggests that Function 1 relates more to implicit knowledge, whereas Function 2 relates more to explicit knowledge.   Table 25 provides two functions at the group centroids.  Table 25.  Functions at group centroids Group Function 1 Function 2 L1 1.686 2.635 HL 1.435 -.843 L2 -1.124 .044  In terms of the mean function scores for each group, Function 1, consisting of the EIT and the aural GJT, differentiated L1 and HL groups from L2 learners. HL learners demonstrated a larger mean score than L2 learners. However, Function 2, consisting of the written ungrammatical GJT, metalinguistic test, and C-test, differentiated L1 and L2 groups from HL learners. HL learners showed a lower mean score than L2 learners. In sum, HL and L2 learners were two different groups in terms  85 of implicit and explicit linguistic knowledge, and L1 speakers had the largest amounts of both types of linguistic knowledge. Figure 11 presents the visual representation of the group centroids. Function 1 separates the L1 and HL groups from L2 groups clearly. This supports that Function 1 is more related to implicit knowledge. In comparison, Function 2 does not differentiate the HL group from L2 group. Instead, Function 2 separates the L1 group from the other two groups. Considering the fact that this data set encompasses the whole groups from native speakers of Korean, HL learners, to L2 learners, the amount of explicit knowledge of native speakers seems to be much larger than the other two groups due to the L1 speakers education at schools. That might be why this DFA failed to differentiate the HL and L2 leaner groups explicit knowledge effectively.  Figure 11. Canonical discriminant functions 86 CHAPTER 4. DISCUSSIONS   With this study, I investigated HL and L2 learners implicit and explicit linguistic knowledge of Korean, focusing on test validity and comparisons between the two types of knowledge for each group. The results summarize that as the best model fit, a two-factor model with five observed variablesi.e., the imitation test, aural GJT, written GJT, metalinguistic test, and C-testdemonstrated the best results. Concerning implicit knowledge, the HL learners showed significantly better results in the EIT and aural GJT. In terms of explicit knowledge, the L2 learner group performed significantly better in the written GJT, metalinguistic test, and C-test.         4.1. Research question 1: Measurements The first research question asked whether scores on the five tests would load on two separate factorsone representing implicit knowledge, and the other representing explicit knowledge.   4.1.1. Valid measurements Concerning valid tests for measuring implicit and explicit linguistic knowledge relatively separately, several types of tests have been investigated and found valid. For implicit knowledge, aural/oral EITs (Erlam, 2009; Kim & Nam, 2016; Spada, Shiu, & Tomita, 2015), oral narrative tests (Bowles, 2011; Ellis, 2005), and timed GJT (Bowles, 2011; Ellis, 2005, Kim & Nam, 2016; Zhang, 2015) have been studied and found valid. In comparison, for explicit knowledge, the untimed written GJT (Bowles, 2011), ungrammatical items in untimed written GJTs (Ellis, 2005; Zhang, 2015), and metalinguistic tests (Bowles, 2011; Elder, 2009; Ellis, 2005; Kim & Nam, 2016) have  87 been investigated and found valid. In addition, more tests have been studied with the intention of measuring implicit knowledge even if the focus of some studies was not exclusively the validity of the measurements: word monitoring tasks (Godfroid, 2015; Suzuki & DeKeyser, 2016) and aural GJTs (Abrahamsson, 2012; Abrahamsson & Hyltenstam, 2008; Bialystok, 1979, 1982; Granena, 2012; Kim & Nam, 2016).  The results of the EIT in the current study replicated the results of previous studies, which supported the claim that the EITs tap implicit knowledge (Bowles, 2011; R. Ellis, 2005; Spada et al., 2015; Zhang, 2015). Spcifically, Kim and Nam (2016) found that the EIT loaded on strong implicit knowledge, whereas timed aural and written GJT loaded on weaker implicit knowledge. The differences between the two tests were indirectly confirmed in the current study. In this study, both of the EIT and aural GJT loaded on implicit knowledge. However, the EIT demonstrated significantly lower mean scores of HL and L2 learners compared to those of the aural GJT. This discrepancy is potentially due to the EITs production element, which the aural GJT does not have and, therefore, represents an additional task component to the aural mode of the EIT. For native speakers, the difference in the scores of the aural GJT and the EIT was not significant, M = .802, p = .640 (See Table 12). However, the difference between the HL learners scores for the two tests was significant, M = -8.39, p < .001; and so were the L2 learners results, M = -18.64, p < .001. Moreover, the HL learners EIT scores, M=58.20, were significantly higher than those of L2 learners, M= 46.02, p=.001. However, concerning the aural GJT scores, the HL group, M=66.60, was not different from the L2 group, M=64.66, p=.367. These results suggest that different tasks require different cognitive loads, leading HL and L2 learners to tap into knowledge differently to varying degrees. In the current study, the EIT and aural GJT were similar in: 1) using natural time constraints, and 2) using aural mode except the EIT required production  88 process in addition to decision on grammaticality, just like the aural GJT. According to Laufer and Goldstein (2004), strength of knowledge can be explained using two sets of concepts: active/passive and recall/recognition. They suggested that a hierarchy of four degrees of strength exist from the strongest to the weakest in the order of active recall, passive recall, active recognition, and passive recognition. They claim that stronger knowledge requires deeper processing (Laufer & Goldstein, 2004). Considering that L2 learners have a limited attentional capacity (McLaughlin, Rossman, & McLeod, 1983), the cognitive load that tasks require should be appropriate compared to learners individual capacity and proficiency levels for successful performance. When cognitive loads for EIT and aural GJT are compared in the current study, the EIT most likely requires learners to go through deeper processing of stronger knowledge. This might have resulted in the discrepancy of their mean scores. In addition, the EITs deeper processing might have taken a larger toll on the L2 learners, reflecting their smaller amount of implicit knowledge than that of HL learners. However, this claim should be supported by further research on grammar knowledge because Laufer and Goldsteins study (2004) was about vocabulary. 4.1.1.1. Aural GJT Considering L2 learners relatively high mean score on the aural GJT compared to their EIT score, how learners process linguistic information when they do not meet the limits of their cognitive capacity might provide useful hints. When the cognitive capacity of proficienct L2 learners is more than the cognitive load that a test requires for measuring implicit knowledge, their remaining cognitive capacity would raise their monitoring and awareness level. This could keep learners from depending on only spontaneous and effortless implicit knowledge, and sequentially enabling them to tap into explicit knowledge. This seems to suggest that the short sentences in the aural GJT  89 did not use up L2 learners cognitive capacity. Therefore, either artificial time constraints or more complex sentences could have enabled learners to access implicit knowledge relatively exclusively, as planned.    The relatively strong loading of the Aural GJT on implicit knowledge demonstrates that modality can be another important factor for measuring two types of linguistic knowledge relatively separately. The results can be explained and supported from two perspectives. First, aurality/oralityaural/oral language talk or conversational interactionis natural, therefore spontaneous, whereas literacy is not. L1 speakers do not learn how to segment streams of sound consciously or intentionally. They just listen and pick up the phonological information during a narrow period, which ranges from the ages of 1 to 5 (Hulstijn, 2015; Paradis, 2009) or 6 (Abrahamsson, 2012) and, in a timely manner, produce the L1 after they are exposed to continous input; primarily they acquire the language implicitly. Therefore, speechoralityis considered as the primary, fleeting, and unconscious form of language in natural communication (Danks & End, 1987). Comparatively, in literacy acquisition, learners learn consciously and intentionally how to decode the written language in words, phrases, sentences, and discourse (Horowitz & Samules, 1987, p. 9). All living languages are primarily used through the aural/oral medium and languages remain alive even without their writing systems, making literacy secondary. Thus, literacywritten language textis a secondary, artifactual, permanent, and conscious form of communication, which restructures consciousness (Horowitz & Samules, 1987). In addition to the ideas of primary orality and secondary literacy, implicit linguistic knowledge is generally said to be primary, stable, and pervasive, whereas explicit knowledge is generally said to be secondary, unstable, and peripheral for language learners (Dörnyei, 2009; N. Ellis, 1994; R. Ellis, 2004; R. Ellis et al., 2009; Reber, 1993). Such arguments can  90 generallystill arguablybe summed up as claiming that implicit knowledge acquired through aurality/orality should be primary, whereas explicit knowledge obtained through literacy should be secondary, especially for L1 speakers and even highly proficient L2 learners. Therefore, it follows that implicit and explicit knowledge could be measured through different modalities because different modes prevail in different types of learning, resulting in different types of knowledge (Bialystok, 1979; Montrul, 2008; Granena, 2012). Accordingly, basic linguistic knowledge of adult heritage speakersearly bilingualsseems to be acquired in childhood like an L1. HL learners processing and storing of what is heard should be automatic and accurate like L1 speakers due to the two groups similar age of onsets (AoOs). This means that HL learners knowledge should be similar to L1 speakers to some degrees, but different from L2 learners (Montrul, 2004, 2008). This prediction was confirmed from the results that implicit knowledge measured from the aural/oral elicited imitation (EI) test and aural GJT of HL learners was significantly larger than that of L2 learners.  Second, the aural GJTas a means for measuring auralityhas an inherent time constraint, which is similar to timed written GJTs. Previous studies used written GJTs with time pressure to measure the implicit knowledge of learners (Bowles, 2011; Ellis, 2005; Godfroid et al., 2015; Gutiérrez, 2012; Zhang, 2015). In the current study, the aural GJT was utilized for the same purpose. The timed GJTs in previous studies and the aural GJT in the current study have one thing in common: time constraints. For the timed written GJTs, 120 % of the time that native speakers had spent in judging the grammaticality of the items in the GJTs was allocated artificially. On the other hand, the aural GJT in the current study is inherently time-constrained. It is maintained when humans hear the sound stream of L1 or L2, their attention-limited phonological storage holds the information for about two seconds (Ardila, 2003; Cole & Pickering, 2015).  91 Afterward the information decays when it is not rehearsed sub-vocally using a sub-vocal rehearsal system in the phonological loopa slave component of the central executive in their working memory (Baddeley, 2003). It is hard to tell that no participant engaged in subvocalization, but the researcher found that no participant engaged in audible subvocalization. Therefore, it seems that the learners time for sub-vocal rehearsal could have been somewhat limited, if not eliminated. The aural GJT had natural time constraints and this factor arguably facilitated the learners access to implicit knowledge like timed written GJTs. Nonetheless, the natural aural GJT might not completely have prevented the participants from sub-vocally repeating what they heard and from focusing on the problematic parts in the GJT items. If they repeated, these attention-drawing processes would hinder them from accessing implicit knowledge totally separately from explicit knowledge. To suppress the learners potential sub-vocal rehearsal, making the items in aural GJTs longer and more complex might be a solution to the issue of potential sub-vocal rehearsal. Moreover, combining time pressure with the aural element might make aural GJTs (Granena, 2012; Kim & Nam, 2016) and EIT (Kim & Nam, 2016) more fine-grained as valid measures of implicit knowledge, distinguishing differences in representations of linguistic knowledge between HL and L2 learners. Actually, in previous research, aural GJTs have been utilized for the purpose of eliciting implicit knowledge (Abrahamsson, 2012; Abrahamsson & Hyltenstam, 2008; Bialystok, 1979, 1982; Granena, 2012) and this received empirical support (Kim & Nam, 2016). In the current study, the natural aural GJT was arguably supported as an authentic way for gauging implicit knowledge similar to timed written GJTs, highlighting orality as a valid factor like time constraints in GJTs. This addition would enrich the thread of research on valid measurements for implicit and explicit linguistic knowledge.   92 4.1.1.2. Written GJT Concerning the written GJT, two major factorstime pressure and grammaticalityhave been highlighted in previous research. Time pressure has been an important factor in studies for L2 learners to distinguish tests for implicit knowledge from those for explicit knowledge in written GJTs (Bowles, 2011; Ellis, 2005, Godfroid et al., 2015; Kim & Nam, 2016; Zhang, 2015). The assumption is that, under time pressure, learners resort to automatic and proceduralized processing, leading them to access implicit knowledge (Ellis, 2005). However, without time pressure, learners can tap into either explicit or implicit knowledge. In further explaining the relationship between time constraints and the linguistic knowledge that learners access, grammaticality seems to play an important role, depending on the item similarities. When items in written GJTs are repeated in timed and untimed contexts sequentially, the grammaticality of the items could be an important factor in measuring implicit and explicit knowledge separately. In studies using L2 learners, R. Ellis (2005) and Zhang (2015) found that only the ungrammatical items in the untimed written GJTs loaded highly onto the factor of explicit knowledge, but not grammatical items. This demonstrates that the grammaticality plays an important role for learners in accessing explicit knowledge when there is no time constraint. This means that if the items of a GJT have a rigorous restriction, implicit knowledge can be measured through timed GJT regardless of the grammaticality. This was also confirmed in the current study. It is noteworthy that R. Ellis (2005) and Zhang (2015) utilized the same items for the timed and untimed written GJTs. In comparison, Bowles (2011) utilized, in the untimed GJT, different tokens of the same type of items from the timed GJT. The results were that the entire untimed GJT, regardless of their grammaticality, loaded highly onto the factor of explicit knowledge, which attenuated the importance of grammaticality in the untimed  93 written GJT. These different results could demonstrate the confounding learning effects of repetition on learners when the same sentences were administered consecutively in the timed and untimed written GJTs. In other words, learners might tap into linguistic knowledge differently, depending on the item similarities or differences in two sequential GJTs. When the items in the two GJTs are the same, their learning effects might automatize grammaticality judging process. In this case, the different types of knowledge, which learners are supposed to access due to time pressure in timed and untimed GJTs, might be blurred, enabling learners to process items in the untimed GJT more quickly as a whole. Thus, processing grammatical items in the untimed written GJT becomes more proceduralized and requires a less analytic approach. This means that in the untimed GJT, to a lesser degree, learners need to analyze the items consciously or take advantage of time as planned by researchers. When the items are different tokens in the GJTs, however, grammatical items in the untimed written GJT still require some degree of conscious analysis (Bowles, 2011; Loewen, 2009), making absence of time constraints an important factor for measuring explicit knowledge. This is also true for a sequence of aural and written untimed GJTs as substantiated in the current study where the same items were utilized in the two consecutive GJTs. Regardless of the different modes, only ungrammatical items in the untimed GJT loaded highly onto explicit knowledge. Therefore, different modalities do not affect the learning effect generated from the same items in two consecutive GJTs. In summary, grammaticality can be an important factor in measuring implicit and explicit knowledge separately when the presented items are the same in two consecutive GJTs, and the second GJT is not under time constraints. However, if the items are different tokens for the same target structures, time pressure, in itself, is a significant factor for measuring the two types of knowledge relatively separately. In addition, these results hold true  94 regardless of modality. 4.1.1.3. C-test In previous studies, a C-test has been designed for the purpose of gauging the readability of L1 speakers related to comprehension and aptitude (Taylor, 1957), and later used for L2 learners (Lee-Ellis, 2009; Oller, 1972; Tremblay, 2011; Tremblay & Garrison, 2010). In this study, the C-test was first utilized to gauge the participants proficiency, where the participants engaged in decoding while reading the passages with blanks, and encoding while filling up the gaps through writing in visual mode. Unexpectedly, the C-test turned out to be a valid measure of explicit knowledge, which can be justified based on Ellis criteria (2009) to operatinalize the constructs of L2 implicit and explicit knowledge. First, in terms of the degree of awareness, a C-test involves heavy literacy dependence and a written mode that is associated with a controlled production aspect, which enhances learners monitoring of linguistic forms. Therefore, a C-test naturally encourages learners to use rules to respond correctly. Second, the time limit for the C-test in the current study was 20 minutes. Considering the fact that almost all participants finished the test in timewith the exception of a couple of low-level participantstime pressure for the test was quite limited. This ample time could restrict learners from utilizing automated and spontaneous knowledge for the C-test. Third, the C-test requires focus on form. It primarily asked for correct forms of Korean structures such as particles, conjugations of the verbs and adjectives, conjunctives, and collocational words. The learners were supposed to pay careful attention to all the candidates they could think of and determine the correct forms for the blanks. Fourth, regarding systematicity of Ellis (2009), the C-test resulted in variable responses, which was confirmed by the highest standard deviations (SDs) among the tests for both groups in the current study (See Table 12). Finally, in terms of  95 learnability, the C-test favored L2 learners who have received form-focused instruction. This was confirmed by the significantly better score of L2 learners compared to that of HL learners in the results (See Table 13). In addition, C-tests could be a good candidate as a written counterpart of EITs. Its reading/written mode and production elements would make the test a valid measurement of explicit knowledge compared to EITs aural/oral mode, and its production elements as a valid measure of implicit knowledge. Therefore, in addition to the aforementioned criteria that Ellis suggested (2009), modalityorality and literacycould be a valid candidate as a criterion to operationalize the contructs of L2 implicit and explicit knowledge relatively separately.    In sum, the topic of fine-grained and valid measurements of implicit and explicit knowledge requires more rigorous study in the future, but such studies should be conducted with a comprehensive battery of tests, which contains a whole gamut of tests concerning dichotomies such as orality/literacy, with and without time pressure, decision/production, and meaning/form. Therefore, EITs, timed/untimed written GJTs, aural GJTs, metalinguistic tests, and C-testsalong with other new measures such as a word monitoring testshould be investigated in one study with proper rival models to find if the tests measure two types of knowledge with construct validity.  Table 26 summarizes the five tests based on the criteria to separate implicit and explicit linguistic knowledge of HL and L2 learners of Korean. This table demonstrates the tests according to the categories of grammaticality, time pressure, modality, and production. Time pressure, aural/oral modes seem to be important factors for measures of implicit knowledge, whereas limited or no time pressure and visual mode seem to be influential on measures of explciit knowledge. Concerning production, the EIT and C-test, which are at the ends of the continuum between implicit and explicit knowledge, have a production element. When cognitive load is enhanced using a production test,  96 learners might have a strong tendency to utilize either implicit or explicit knowledge relatively exclusively. That might mean that learners cannot tap into both types of knowledge with flexibility when they engage in a more cognitively taxing task. This postulation is supported by the results of the DFA (See Table 24). The EIT and C-tests standardized canonical discriminant function coefficients for Fuctions 1 and 2 are highest among the factors. Therefore, deeper processings of tasks with a production element seem to tap into stronger and pure knowledge.    Table 26. Five tests and criteria to separate implicit and explicit knowledge  Test Category EIT Aural GJT Written GJT Meta- linguistic test 1 Meta- linguistic test 2 C-test Grammaticslity Ungrammatical Yes Yes Yes Yes n/a Yes Grammatical Yes Yes Yes No n/a Yes Time pressure Yes Yes No No No Limited Modality Aurality Yes Yes No No No No Orality Yes No No No No No Production Yes No No No No Yes Note. * means that the time pressure and production elements of the C-test were limited because 1) the time limit was 20 minutes, and 2) they were supposed to produce relevant syllables in context, not whole words or sentences.  4.1.2. Models Regarding the proper factor model of implicit and explicit knowledge, results of the current study confirmed that a two-factor model was valid, which aligns with the findings of previous studies (Bowles, 2011; R. Ellis, 2005; Zhang, 2015). The three previous studies scopes have been extended from learners of English as L2 (R. Ellis, 2005), to Spanish learners of Spanish as FL and HL (Bowles, 2011), and to learners of  97 English as FL (Zhang, 2015). These studies, as well as the current study, found the two-factor model of implicit and explicit knowledge valid regardless of language, target structures, or types of learners. In the currenty study, the two-factor model also seems to be favored over a one-factor model. Among the total 32 CFAs using the data set, six models showed good fit indices (See Appendix E or Table 16 for a summary version). For models with four observed variables, one two-factor model (2 Factor 4 Observed variables Rival 1) and one one-factor model (1 Factor 4 Observed variables Rival 2) demonstrated good model fit indices without Heywood cases. Even if the two-factor models factor loadings were stronger than those of the one-factor model, the model fit indices were comparable with each other. However, when the number of observed variables increased from four through five to six, none of the one-factor models presented good model fit indices, whereas the two-factor models demonstrated four models with good fit indices. These results support that two-factor models explain the data set in a more valid way than one-factor models regardless of various contexts such as numbers or combination types of observed variables. In this case, the two-factor model might have better discriminant validity than the one-factor model by explaining the data set with more discriminating power. In comparison, the one-factor model explains the data in an underidentified way, increasing the risk of empirical underidentification (Brown, 2015).  Moreover, the results of the discriminant function analysis (DFA) also supported the two-factor model (See Tables 19-20). In the statistical analyses, the data of native speakers were also included and this confirmed that one of the factors in the two-factor model is highly likely to be implicit knowledge rather than an automatized type of explicit knowledge. Moreover, the first factor explains a total 64.7 % of the variance. This is the majority of the variance, which is also in line with the claims that  98 implicit knowledge is primary whereas explicit knowledge is secondary and peripheral (Dörnyei, 2009; N. Ellis, 1994; R. Ellis et al., 2009; Reber, 1993; Rebuschat, 2013). (See Results section for detailed information):  The results of the current study suggest that adding HL learners linguistic data to L2 learners is a great support for investigating the nature of L2 learners linguistic knowledge. Previous research suggests that the concept of L2 learners implicit and explicit knowledge can be explained using a dissociated dichotomy (Bowles, 2011; Ellis, 2005; Spada et al., 2015; Zhang, 2015). However, DeKeyser (2003) explains L2 learners knowledge using the concepts of proceduralized or automatized explicit knowledge and analyzed explicit knowledge. When a data set comes from only L2 learners, it can be posited that L2 learners knowledge measured through an EIT, timed GJT, or aural GJT might be either implicit knowledge or proceduralized explicit knowledge (DeKeyser, 2003; Suzuki & DeKeyser, 2015). However, when data from L2 and HL learners are combined and the sample size is large enough, the results should be interpreted from a new perspective. Thus, it is logical to presume that HL learners, due to their special learning environment since birth, should possess a considerable amount of implicit knowledge and varied degrees of explicit knowledge, depending on their periods of instruction. If the two groups do not violate test invariance and the equivalence of the measurement model is determined from multiple-group solutions using CFAs, the L2 learners presumed implicit knowledge should arguably be as implicit as the HL learners, not proceduralized explicit knowledge. However, if the two groups data violates any significant measurement invariance, this means that the tests measure different skills/constructs in the groups. In such cases, the discrepancy suggests that the nature of HL learners implicit knowledge is different from that of L2 learners, which might be proceduralized explicit knowledge. Thus combining the HL groups  99 data with L2 learners could help interpret CFA results correctly, qualifying Similarities and differences of HL and L2 learners two types of linguistic knowledge. However, it is too early for a final verdict. For more fine-grained and valid measurements, EITs, timed written GJTs, and aural GJTs along with other measures should be investigated in one study with HL and L2 learners using proper rival models.    4.2. Research questions 2 and 3: Similarities and differences of the two groups Research questions 2 and 3 concern similarities or differences between HL and L2 learner groups two types of knowledge. HL and L2 learners learning trajectory is different from L1 speakers in terms of AoO, input, and ultimate attainment. In addition, HL learners language learning profile is also different from L2 learners regarding AoO, input, and types of linguistic knowledge. Therefore, it is necessary to investigate the roles of AoO and input for explaining those differences between L1, HL, and L2 learner groups.   4.2.1. Various learners HL and L2 learners linguistic knowledge of a target language is similar in that it consists of incomplete implicit and explicit linguistic knowledge compared to the knowledge and the ultimate attainment of native speakers (Montrul, 2008). Previous research demonstrates that, in varying degrees, L1 monolinguals possess larger implicit and explicit knowledge than the compared groups both in orality and literacy (Bowles, 2011; Ellis, 2005; Montrul, 2008). These are presumably attributed to the combination of early AoO and ample input (Bley-Vroman, 1990; Lenneberg, 1967). In comparison to L1 speakers, HL learners do not receive enough or sustained input except for considerable initial exposure at a very early age in the context of family and community.  100 Adult late L2 learners usually do not receive input before age 14, which indicates a lack of input and late AoO. In the current study, a rough measure of the three groups self-reported input amounts1 from the bio-data questionnaire demonstrates that the L1 groups input amount was significantly larger than those of the HL group and L2 group (F=800.79, p<.001) throughout their lives, before age 14 (F=823.01, p<.001), after age 14 (F=340.86, p<.001), and after age 18 (F=340.86, p<.001). These results suggest that the two comparison groups unsustained input throughout their lives, and especially, the L2 groups late exposure to Korean might contribute to the different characteristics of linguistic knowledge (Muñoz, 2014).   the Fundamental Difference Hypothesis (FDH) (Bley-Vroman, 1990). According to the hypothesis, children acquire their L1s through Universal Grammar (UG) using domain-specific and deductive linguistic mechanisms, whereas L2 learners utilize domain-general and inductive problem-solving mechanisms due to a loss of plasticity and progressive brain lateralization (Birdsong, 2005). Therefore, in the current study, the result that there are significant differences in the amounts of two types of linguistic knowledge between L1 speakers and L2 learners arguably indicate that the two groups linguistic mechanisms might be different qualitatively as well as quantitatively. On the other hand, the L1 learners differences in linguistic knowledge from HL learners demonstrate the importance of amount and nature of input. In terms of both implicit and                                                  1 Table 27. The input amounts by groups (%) Age       Group L1(n=11) HL(n=38) L2(n=65) Total  100 27.21 7.05 Before 14  100 31.41 0 After 14  100 21.60 16.51 After 18  100 17.44 22.88 Note. This was measured roughly using their self-reports. The input amounts were converted in percentage scale by group. The question items in the questionnaire dealt with natural context such as input from parents, relatives, friends, and community schools. After the age of 18 in college, an additional question about regular class was added.   101 explicit knowledge, HL learners linguistic knowledge is significantly smaller than that of L1 speakers. Even if HL learners AoO is similar to L1 speakers, early AoO is not good enough to guarantee that HL learners sizes and ultimate attainments of implicit and explicit knowledge are comparable to L1 speakers. Actually, the HL learners input amount was significantly smaller than that of L1 speakers during every time period in question, emphasizing the importance of ample, sustained, and proper input for the two types of knowledge to develop throughout their life-long language learning process.  4.2.1.1. Input and age Comparisons between the HL and L2 group AoO, input amount, and size of linguistic knowledge suggest a more complex picture. In the current study, the total amount of input of HL learners was significantly larger than that of L2 learners (F=175.392, p<.001). If input amount is the only factor to influence two types of linguistic knowledge, HL learners both types of knowledge should be larger than those of L2 learners. However, this study showed that only implicit knowledge of HL learners was significantly larger than that of L2 learners, but not explicit knowledge. Therefore, this result suggests that the input amount does not explain every aspect of two types of linguistic knowledge between the two groups. In addition to input amount, AoO and instruction, should also be important factors for the explanation.   A review of the input table demonstrates that the HL learners received a significantly larger amount of input than the L2 learners before 14 years of agean approximate terminus among the ending ages in critical periods such as age 12 (Scovel, 2000), ages 12-13 (Lenneberg, 1967), and age 15 (Johnson & Newport, 1989). However, their input amounts decreased as they became more assimilated into the society where their dominant language was used. After age 18, the HL learners input amount became significantly smaller than the L2 learners, even though the HL learners implicit  102 knowledge was significantly larger than that of the L2 learners. This means that AoO base of implicit knowledge. Concerning the L2 learners, before age 14, their input amount was none. After age 14, the L2 learners input was still significantly smaller than that of HL learners (F=7.253, p=.008). However, after age 18, when the L2 learners received instruction and took courses in college, their input amount was significantly larger than that of the HL learners (F=5.304, p=.023). This means that the L2 learners were exposed to a considerable amount of input while they were living and learning in Korea. The L2 learners explicit knowledge was significantly larger than that of HL  where linguistic knowledge is concerned. Considering the facts that the L2 learners started learning Korean after age 14 and received their significantly large amount of input from instruction after age 18, the researcher suggests that as far as explicit knowledge is concerned, AoO is not as important as the amount and nature of input for both L2 and HL learners.  According to the FDH (Bley-Vroman, 1990), children access a domain-specific linguistic mechanism for their L1 acquisition, whereas adults access domain-general cognitive mechanisms for their L2. This idea has been developed with the neuro-cognitive approach (Paradis, 2004) suggesting that domain-specific linguistic mechanisms using UG lead to primarily implicit learning and, therefore, primarily implicit knowledge (Bley-Vroman, 1990; DeKeyser, 2000; Ellis, 2009). In comparison, domain-general cognitive mechanisms using L2 learners L1 inventory lead to an explicit learning process and primarily explicit knowledge (Bley-Vroman, 1990; DeKeyser, 2000; Ellis, 2009). The HL learners larger implicit knowledge than that of L2 learners attests that the HL learners learning process at an early age is basically the same as that of L1 speakers: natural and implicit learning due to their early AoO and a  103 significantly larger amount of input than L2 learners. However, sustained input afterwards decides the sizes and types of linguistic knowledge, guaranteeing the same ultimate attainment as native speakers. However, L2 learners larger explicit knowledge attests that their learning mechanism and input type of the learning are different from those of HL learners in nature. Thus HL learners larger implicit knowledge and L2 learners larger explicit knowledge are relevant to; 1) the FDH (Bley-Vroman, 1990), and 2) CPH. First, the FDH suggests that early simultaneous L2 learners domain-specific and UG-related learning mechanism concerns implicit knowledge, and late L2 learners domain-general and non-UG-related mechanism concerns explicit knowledge (Montrul, 2009). Second, CPH suggests that obtaining implicit knowledge requires ample amount of input during a proper time periodearly AoO (Lenneberg, 1967). Accordingly, the postulation is supported that unlike implicit knowledge, explicit knowledge does not require early AoO as much as ample analysis-focused input after puberty (DeKeyser, 2000).     4.2.1.2. Orality and literacy Modality should be emphasized more when the amounts of HL and L2 learner groups linguistic knowledge are compared due to the two groups different advantages in terms of listening and reading skills. Most studies using time constraints as a factor in written GJTs investigated L2 learners two types of knowledge, not HL learners (Ellis, 2005; Granena, 2012; Gutiérrez, 2012; Zhang, 2015). Unlike L2 learners, HL learners spend their early learning period with their Korean-speaking family members. They start to acquire the phonological segments of Korean in aural/oral modes. Thus, using a written mode might not let HL learners access their implicit knowledge as well as an aural mode, especially when they do not have enough literacy or code-focused explicit instruction. Actually, Bowles (2011) compared HL and L2 learner gro knowledge  104 using written GJTs with or without time constraints. In her study, the HL learners were taking college instruction just as their L2 counterparts. Therefore, the two groups reading skills of the target language were comparable. However, in the current study, 15 of the 38 HL learners have not received college instruction in Korean, whereas the L2 learners have considerable amounts of literacy from instruction at college. These unequal reading skills between the two groups might have confounded the comparison results in favor of L2 learners, rendering the two groups implicit knowledge unjustifiably unbalanced in favor of the L2 learners. In this context, the researcher presumed that using modality as a factor might let HL learners assess implicit knowledge with validity.  HL and L2 grou differences in their linguistic knowledge have been investigated in terms of one groups advantages over the other in different types of knowledge (Montrul, Foote, & Perpinan, 2008). That is to say that HL learners are better at oral tasks, whereas L2 learners are better at written tasks (Matsunaga, 2003; Montrul et al., 2008). This was substantiated in the results that demonstrate that HL learners showed significantly larger implicit knowledge than L2 learners. A post-hoc t-test using the results of the aural and written GJTs demonstrates the HL groups considerably smaller explicit knowledge than that of L2 learners. Even though the HL group difference between the written and aural GJT scores were not significantly different, M = -.038, p = .967, that of the L2 groups was significant, M = 5.95, p < .001. These results could be interpreted that HL learners, with limited explicit knowledge and more implicit knowledge, outperformed the L2 learners on the aural GJT. However, when the HL learner group took the written GJT, the extra time did not help them due to their lack of explicit knowledge to draw on. That the HL learners exhibited roughly the same scores on both GJTs confirms this argument. In comparison, the L2 learners, with  105 considerable explicit knowledge, were able to take advantage of the extra time to employ that their explicit knowledge, which was larger than that of HL learners. This means that L2 learners were able to supplement their smaller amount of implicit knowledge with their larger explicit knowledge in the untimed written GJT (Ellis, 2009; Loewen, 2009). This suggests that untimed written GJTs measure both implicit and explicit knowledge, making a coarse measure of explicit knowledge. Therefore, an additional factorsuch as grammaticalityshould be necessitated to measure explicit knowledge in a more valid way. The connection between oral tasks and a large amount of implicit knowledge can be explained through the interplay of input modes along with AoO and input amounts. Due to early AoO and home-based implicit learning context, L1 and HL learners are equipped with basically the same phonological inventory since birth as L1 monolinguals (Montrul, 2009). Unlike L1 speakers or HL learners well-established phonological knowledge, L2 learners tend to tap into L1 segmental and supra-segmental phonological features when they do not have proper L2 inventory to access (Abrahamsson, 2012; Bley-Vroman, 1990). This slows the processing of their L2 (Munro & Derwing, 1995), i.e., Korean, and makes L2 word-recognition difficult (Bradlow & Pisoni, 1999). Therefore, it could be inferenced from the data of the current study that their implicit linguistic knowledgemeasured using aurality/oralityshould be naturally smaller than that of HL learners of Korean. L2 learners advantage over HL learners in explicit knowledgemeasured using literacywas also confirmed from the results in the current study. The L2 learners significantly larger explicit knowledge could be due to 1) HL learners very limited amount of code-focused and literacy-centered explicit learning of Korean in established schools, and 2) L2 learners learning Korean through reading/writing-centered instruction in college after a critical period  106 when an explicit learning mechanism is more available than implicit learning. In short, AoO, input amount and modality are three major factors that explain the discrepancies between L1, HL and L2 groups two types of knowledge in Korean.  4.2.1.3. Learner variances When the test results of the HL and L2 learner groups were broken down, statistically significant differences were found only in the aural/oral EI and metalinguistic knowledge test, but not in the aural and written GJTs. This discrepancy might be explained from two perspectives concerning their task characteristics. First, the aural GJT in the current study did not require high cognitive load on the L2 learners. In previous research, L2 and HL  linguistic knowledge was measured using more cognitively taxing oral tasks such as an oral picture description task in Spanish (Montrul, Foote, & Perpinan, 2008), an oral production test in Japanese (Matsunaga, 2003), or a spoken word recognition test in French (Guillelmon & Grosjean, 2001). However, in the current study, the aural GJT requires neither correction nor production; it only requires receptivedecisionskills to process relatively shorter sentences consisting of 5.5 words on average. This is shorter than 6.3 Spanish words in a Spanish-as-a-foreign-language situation (Bowles, 2011), 8.1 English words in an ESL situation (Ellis, 2009), and 8.7 English words in an EFL situation (Zhang, 2015). The low processing load imposed on the L2 learners from judgment-oriented and manageable items in the aural GJTs could be the reason for a smaller difference found in the accuracy scores in the aural GJT between the two groups. Longer sentences with more words could have demonstrated the two groups linguistic knowledge significantly differently. Second, the L2 learners in the current study are not Korean as a Foreign Language (KFL) learners, but KSL, who have been exposed to listening to natural and casual Korean in their learning contexts. This could have made the L2 learners  107 correctness in aural GJT considerably high. For learners of Korean as a KFL, their aural GJT scores might not have been as high as those of L2 learners, resulting in a larger gap between the KFL and HL groups. In addition, the L2 learners proficiencies were relatively high, with their average instruction period being 19.75 months in Korean-as-a-second-language context. In sum, due to the short items in the aural GJT, the L2 learners long stay in Korea and high proficiency levels, the aural GJT might have been easy for the L2 learners.  Likewise, the two comparison groups demonstrated a significant difference in mean scores in the metalinguistic test in favor of the L2 learners. However, in the written GJT, the L2 groups mean score was not significantly different from that of the HL group. This non-significance might be due to the postulation that decision-based written GJTs requires only moderately analyzed knowledge compared to production-based written tasks (such as a C-test), which needs a considerable amount of analyzed knowledge, attention to structure, and form-meaning coordination (Birdsong, 1989). Another reason can be the simple structures in the written GJT items, which consisted of 5.5 Korean words on average. Even though HL learners do not have an advantage over L2 learners in reading skills, reading the short sentences might not have been too difficult for the HL learners. In fact, all of the HL learners exposed to easy Korean literacy at an early age learned how to read and write Korean at the age of 6 on average. In other studies, HL learners literacy was not significantly different from that of L2 learners in reading comprehension (Matsunaga, 2003), gender agreement through written recognition and interpretation tasks (Montrul, Foote, & Perpinán, 2008), and gender agreement through untimed written GJT (Montrul, Foote, & Perpinán, 2008). Therefore, it suffices to say that the discrepancies between the oral EIT and aural GJT as well as between the metalinguistic test and untimed GJT would mean learner variances  108 from different tasks. In sum, different tasks measure different linguistic knowledges. Even in cases where several tasks are meant to measure the same type of linguistic knowledge, each task could demonstrate its own characteristics, depending on their cognitive processing load on specific learner groups. The more cognitive load learners experience, the more likely they are to tap implicit knowledge for spontaneous and effortless responses. The EIT requires more cognitive efforts for comprehension in aural mode and for production in oral mode. In comparison, the aural GJT requires only comprehension in aural mode, allowing more time to tap explicit knowledge with awareness. To compensate for this low cognitive load, GJTs should have proper time constraints for better measuring implicit knowledge, regardless of their modality (Kim & Nam, 2016). Time constrants could be essential when sentences are relatively short, as evidenced in the current study.  The HL learners knowledge reflects their position between L1 speakers and L2 learners. HL learners receive enough naturalistic input to develop their implicit knowledge as well as L1 speakers; demonstrate enough explicit knowledge so that this knowledge is as developed as L2 learners. Therefore, it seems that their primary advantage over L2 learners lies in implicit knowledge from their early AoO and considerable amounts of natural input. 4.2.1.4. C-test for various types of learners When researchers interpret various group results of a C-test as a proficiency test, care is due, as a C-test might not be an effective measure of proficiency for HL learners. Since the test measures primarily explicit knowledge, HL learners proficiency might result in very low compared to their true profiency when implicit knowledge is taken into account. In comparison, for L2 learners as a foreign language, a C-test might unwarrantedly reflect higher of profieicny. This is due to their form-focused instruction  109 in a visual mode, focusing on explicit knowledge in proficiency. Finally, a C-test for an L2 learner as a second language could present a more balanced indication of proficiency between explcit and implicit knowledge than that of an L2 learner of a foreign language due to the former learners balance between form-focused instruction in a visual mode and their exposure to natural learning contexts of L2.   In a comparison study between the three typles of L2 learners, a C-test alone might not be a good measure of proficiency because the results would be unjustly in favor of L2 learners and unfairly unfavorable to HL learners. In a foreign language context, L2 learners learn their L2s in primarily visual, written-modes. Conversely, in a second language context, L2 learners learn in written and spoken modes. Either way, L2 learners have an advantage over HL learners in a written-mode test because HL learners start to learn primarily in a speech mode. In addition, the reading passages in the C-test could be varied in regards to length, abstractness, and formality, making C-tests more difficult for HL learners whose learning is speech-oriented.  Using a C-test as a sole proficiency measurement for HL learners can be inherently misleading due to the tests written mode and HL learners non-homogeneous literacy skills. In Bowles study (2011), the HL and L2 learners were enrolled in Spanish and literature classes at a university in the US, guaranteeing homogeneity in the same levels between the two groups. In the current study, however, the two groups proficiency scores were significantly different (F=18.631, p<.001) in favor of the L2 learners. The reasons can be summed up as follows. The HL learners in the current study were from three different areas in the US, and 15 out of 38 of the participating HL learners (39.5 %) have not received any college-level instruction. These factors could have aggravated the HL learners potential heterogeneity in taking the C-test, especially because the HL learners who were without instruction learned only casual Korean from  110 family and relatives in aural/oral modes. Therefore, they were not familiar with the types of written passages in the C-test where considerably long and dense passages were used, and registers were different from their casual Korean. The majority of the HL learners debriefed that the topics for the third and fourth passages were unfamiliar and that they were more or less clueless in taking the proficiency test. This evidences that the reading discourse was not in accordance with their language-learning contextHL learners limited literacy development. In comparison, the L2 learners in this study were studying Korean as a second language at a university. In fact, the L2 learners were expected to obtain a proper level in the Test of Proficiency in Korean (TOPIK) to be admitted to a college or graduate program. This test includes a reading section, whose selected passages for reading and those selected for the C-test might not be very different from one another. This similarity might have enhanced the L2 learners scores of the C-test. In sum, the two group heterogeneous learning contexts would make it difficult to compare the two groups proficiency levels using only the C-test.  As a solution to the C-tests limitations as a proficiency test, a couple of alternative tests could be used. For instance, the TOPIK itself could be a candidate as it consists of listening, reading, and writing; the addition of a listening section might compensate the limitation of evaluating only explicit knowledge through a C-test. Yet, because the TOPIK is a test designed for L2 learners of Korean, its validity as a proficiency test for HL learners has yet to be proven. The caveat is that all of the L2 learners in the current study have taken the TOPIK at least once and, at most, six times. Therefore, L2 learners practice effect should be taken into consideration. Thus, designing proficiency tests for HL learners in comparison with L2 learners deserves more attention based on the nature of HL learners language-learning and linguistic knowledge.    111 Another alternative is combining HL learners accuracy scores on tests that measure implicit and explicit knowledge. From the perspective of implicit and explicit knowledge, this could be a more balanced measure of proficiency (Elder & Ellis, 2009; Philp, 2009). When scores of implicit and explicit knowledge are combined, the two groups proficiencies seem to be measured more comprehensively. For the purpose of comparing the linguistic knowledge of HL and L2 learners, researchers have utilized various proficiency tests such as vocabulary and cloze parts (Montrul et al., 2008, Montrul & Foote, 2014), a skill-balanced proficiency test designed for a specific language (Montrul, 2005; Montrul et al., 2008), a self-report (Montrul, 2005; 2006; Montrul et al., 2008), a C-test (Montrul, 2006), and the learners enrollment status (Bowles, 2011). From the perspective of balance between the implicit and explicit knowledge of various types of L2 learners, the proficiency test above should be carefully designed and utilized using appropriate proficiency measures undergo further investigation. As such, research comparing HL and L2 learners, especially concerning the balances between orality and literacy, as well as implicit and explicit knowledge.  4.3. Pedagogical implications  The current study provides pedagogical ideas about how the delayed acquisition of Korean influences the incomplete acquisition of two types of linguistic knowledge of naturalistic HL and instruction-oriented L2 learner groups. For HL learners, their delayed explicit learning of Korean in school and not-enough quality of input from limited learning context attributes to incomplete acquisition, especially regarding explicit linguistic knowledge. For L2 learners, incomplete acquisition via late AoO is ascribed to their delayed implicit and explicit learning. In addition, their lack of brain plasticity takes a toll on implicit learning as demonstrated in the results of this study.  112 The key issue, thus, involves how much HL and L2 learners incomplete acquisition can be solved through instruction, depending on their re-exposure effect (Montrul, 2008). Concerning HL learners, more instruction will facilitate their acquisition of explicit knowledge through analyzing skills they have with their majority language. On the other hand, it is assumed that L2 learners acquire their explicit knowledge of Korean through the same skills from their L1 explicit learning and knowledge. Concerning L2 learners implicit knowledge, exposure to more aural input might not remarkably facilitate L2 learners acquisition of implicit knowledge due to the interference of their phonological system in L1 and late AoOlack of brain plasticity. Accordingly, instruction at college should strike a balance between the two groups unequal, at the same time, incomplete linguistic knowledge and language skills. College curriculum and instruction should, thus, be modified based on the results in the current study. For the best effectiveness, college curriculum in teaching HL learners can be provided in two tracks via two separate courses for the two different groups. However, in cases where the two groups are forced into one class, assigning them different roles in pair activities could be a solution to the one-track learning situation.   113 CHAPTER 5. LIMITATIONS  The problem of unequal sample sizes in the current study is noteworthy. I think that collecting more data from HL learners to make the HL and L2 groups sample sizes equal would not influence the model fit indexes in the two-factor model. One potential reason is that HL learners' status of implicit and explicit knowledge will not change with a larger sample size. Logically, they should have a considerable amount of implicit knowledge acquired since birth and they should also have less explicit knowledge from their limited instruction. To prove this, a CFA was employed using only data collected from the HL learners with instruction in college. The two-factor model also fits the modified data well. Therefore, the researcher came to a temporary conclusion that the results might not change significantly, even if the sample sizes between the two groups become equal. In other words, the unequal sample sizes of more L2 learners and less HL learners might have influenced the original two-factor model in a negative way, compared to that of the one-factor model. Even when the group sample sizes were unfavorably unequal, the two-factor model still held. Thus, it is reasonable to think that more data from HL learners will not influence the validity of the two-factor model.   In actuality, the unequal sample sizes between the two groups might not be as important as the sample size of this study. A small sample size in a CFA could cause a false rejection of models based on TLI and RMSEA (Hu & Bentler, 1999), Heywood cases, and distorted statistical power and precision of the parameter estimates (Brown, 2015). Even if the sample size of the current study is acceptable (Mitchell, 19932; Stevens, 19963), from a more conservative view, a larger sample size is recommended                                                  2 Mitchell (1993) suggests that a desirable sample size is from 10 to 20 times of the number of observed variables.  3 Stevens (1996) suggests a 15-time sample size of the number of observed variables.  114 (Joreskog & Sorbom, 19894). Moreover, considering the recommendation that each latent variable should have a minimum of three indicators for best evaluating the acceptability of solutions (Brown, 2015), the current study might require more participantsa sample size of 200as well as more indicators.  Proficiency levels of the two comparison groups, which were measured through a C-test did not match. Therefore, the comparisons between the two groups implicit and explicit knowledge should be considered with due caution. It is noteworthy, nonetheless, that even though the HL groups proficiency level was significantly lower, the HL learners implicit knowledge was significantly larger than that of the L2 learners. More HL learners with higher proficiency scores on the C-test would produce higher accuracy rates in the written GJT and metalinguistic test, resulting in a smaller gap in explicit knowledge. In the EIT and aural GJT, the HL learners would obtain higher mean scores, widening the gap between the two groups. However, it is difficult to predict exactly how the two groups matching proficiency levels would influence the results of the CFAs, therefore, the measurement validity of the two-factor model. Considering that a C-test is likely to primarily measure the explicit knowledge of L2 learners due to its written modality and ample time component, proficiency levels of HL and L2 learner groups should be measured with caution.                                                   4 Joreskog & Sorbom (1989) present a conservative view concerning a desirable sample size. When the number of observed variables is less than 12, 200 participants are desirable.   115 CHAPTER 6. FUTURE STUDIES  Using various factors of HL and L2 learners will shed insight on the validity of measurements for implicit and explicit knowledge. For instance, dividing L2 learners based on their proficiency levels and running two separate CFAs might provide further ideas on studies about the interface between explicit knowledge and implicit knowledge. When L2 learners learn Korean in the Focus-on-Forms learning context, the low-proficiency learners have much more explicit knowledge than implicit knowledge. This would result in a one-factor modelexplicit knowledgefrom a CFA. However, for high-proficiency learners, the results might be that their implicit knowledge, if any, is comparable to their explicit knowledge. This would guarantee the two-factor models measurement validity. This differenceor even similaritybetween the results from the two CFAs using the two groups regarding low and high proficiency levels would give valuable suggestions regarding the possibility of interface between the two types of linguistic knowledge. Furthermore, adding another type of data from learners of Korean as a foreign language (KFL) could enrich the results of this study. Their major input is from code-focused instruction in literacy; therefore, the nature of their input would be different from that of learners of Korean as a second language (KSL). When three types of learners datai.e., HL, KSL, and KFL learnersare combined and if the two-factor model still holds, it would be concluded that the three groups of HL, KSL, and KFL learners do not demonstrate group variances for the tests. This invariance will confirm the construct validity of the measurements.  Valid measurements of implicit and explicit linguistic knowledge are required in SLA. More measurements should be investigated concerning their validities and reliabilities. In the current study, as receptivedecisiontests, aural and written GJTs  116 have been utilized. However, as productive tests for explicit knowledge, a valid counterpart of an aural/oral EIT has not been developed. Results from an oral production task were compared with those from written recognition and comprehension tasks (Montrul et al., 2008), but oral production tasks were not directly compared with written production tasks. Since productive skills require more processing and attention than receptivedecisionskills, production tasks would be more affected by incomplete acquisition than by comprehension tasks. However, it is unknown how the receptive/productive dichotomy is related or unrelated to time pressure, grammaticality, and modality. In fact, the dichotomy of decision/production was also studied as another variable in investigating the characteristics of linguistic knowledge (Ellis, 2009) in addition to aforementioned factors. However, the dichotomy turned out to be invalid in the current study. More studies are due on this topic. The direction of future research should be two-fold. First, previous measures that have been found relatively valid should undergo more elaboration to be more refined in utilizing time constraint, modality, grammaticality, and decision/production. Second, more new measures should be developed and their relationship with the previous measures should be investigated from the perspective of the continuum between implicit and explicit knowledge. Each measures relative position should also be estimated in the continuum, not dichotomy (Hulstijn, 2015; Williams, 2005). In sum, the following depiction in Table 28 requires far more studies in the future to complete.       117 Table 28. Implicit and explicit knowledge and relevant measurements Speakers L1  L1  HL learners L2 as 2nd language L2 as Foreign language Constructs  Pure IM Stronger IM Weaker IM --- EX Pure IM (Less pure) IM - Automatized/proceduralized EX --- EX Measures ? EIT, Word monotoring test - Timed written, aural GJT Untimed  Aural GJT Untimed written GJT Ungrammatical items in untimed written GJT, C-test --- Meta linguistic or lingual test   118                     APPENDICES  119 APPENDIX A. Stimuli (Pilot study I)  A-1. Stimuli for Elicited Imitation Test A-2. Stimuli for the Timed and Untimed Grammaticality Judgment Test A-3. Stimuli for the Metalinguistic Knowledge Test  A-4. Background questionnaire  A-1. Elicited Imitation Test 1-1. *7ı9Ê':9Ê 9î  . (Children draw pictures well) 1-2. /Z8& 8^2² !Ž9Ê ,$9Ê GŒ%:. (Many beautiful flowers bloom in spring) 2-1. *:Û9Ê+Ò 3æ8& ˚V"ª .ö9Î%:. (People often cut their hands with a piece of paper) 2-2.  8WHª8Œ #Î)î+Ò :á7ıG²%:. (Koreans like movies and songs) 3-1. * G²˜Ã9®  7ž :á7®%:. (Korean economy slowed down yesterday) 3-2. ,$9Œ 7ı9Ê':9Ê 8z$î ˜K9&8&3r . (Many children exercise in the park today) 4-1. *.N˜Ã8&3r >ò>òG®˚â  G²%:. (You should drive slowly in the US) 4-2. Cª*^0ı:Z  BÂ˚â GÊ7Ò . (To listen, you should keep up the volume of TV) 5-1. *G²˜Ã,&9Œ  4Ó%:. (Korean is never easy) 5-2.  G²˜Ã -”Hª+Ò :ZHV ,þ+Î%:. (People never know Korean culture) 6-1. *,$9Œ   . (Many people live in Seoul) 6-2.   7ı;R ˚V ¢9"8ê. (It is very close from Seoul to Busan) 7-1. *8B ƒ8&  8z5C ˚r 9Þ%:. (Here are 50 people) 7-2. &ˆ3r˜V9Œ  9w @‰9Ê,Ê :á%:. (Usually a six-floor library is good) 8-1. *.N˜Ã9® 8z.j,˚$ê #^ @²˜Â9Ê%:. (Obama in the U.S. is your friend) 8-2.  G²˜Ã9®  8B9æ9Ê%:. (Now, the president of Korea is a woman) 9-1. *G¯339Œ G¯˜¦8& , ˜K/Ö+Ò G²%:. (Even though students go to school, they study) 9-2. 0ı˚V ,$9Ê , 4n8W9ı G²%:. (Even though it rains a lot, I swim) 10-1. * 52˚Z9Ê  ˚V ƒ+Ò  . (People do not know that time flies) 10-2. G¯33':9Œ 52H.8& 5:4nG² ˚Ù9ı 7ž%:. (Students know that they made mistakes in the exam) 11-1. * 3”˜ˆ 9ΘÂ˚V ,$9Ê $î8˚%:. (Next year, world population increased a lot) 11-2.  GÊ 9Œ >” . (The Olympics were very interesting last year) 12-1. *@²˜Â˚V ˜K/Ö 9î G®$ê ,¡9Þ%:. (Friends who study well are cool) 12-2. G²˜Ã  :á7ıG®$ê ˚Ù9Œ #Î)î9Ê%:. (What Korean people like is songs) 13-1. *:`9Œ9Ê':9Ê A:G™+Ò :á7ıG²%: 7ž%:. (I know that young people like coffee) 13-2. GZ)ç4ú˚V 9*'2AK8&3r  ˚Ù9Œ #Ö)ã%:. (It is surprising that France won the World Cup)  14-1. * G²˜Ã  9®GÊ ,"':8˚%:. (Smart phones were invented by Korean people)  120 14-2. 8ê;Þ9¢ 3r:f8&3r G²˜Ã>ł':9Ê 9î 9ÓIâ%:. (These days, Korean books read well in the bookstores) 15-1. * 8ˆ,˚˚V 7ı ƒ+Ò 9ê%:. (Usually mothers put their babies to sleep) 15-2.   &/338&˚â 8“9ı 9ÛIâ%:. (Usually an older sister dresses her younger sibling) 16-1. * .{9ı ,19Þ˚â ,‘$ê%:. (The grandfather eats boiled rice deliciously) 16-2. /Ö,þ%.!.3r$ê G²˜Ã8& ˜ˆ56%:. (My parents are in Korea) 17-1. *  ,þ9æ+Ò 56$ê%:. (People wear hats when it is cold in winter) 17-2.  Gß/KG®,Ê H²<3 9ˇ$ê%:. (People have big grins on their faces when they are happy)       A-2. Stimuli for the Timed and Untimed GJTs 1-1. *>ł .RH” 9Ó$ê%:. (Minho read a book) 1-2. * 8^2®˚â 7ı;R Hª9û9ı GÞ%:. (Sooni had beautiful makeup)  1-3. 9v.N˚V @²˜Â8&˚â  7J%:. (Yumi writes a letter to a friend) 1-4. .R4n˚V @²˜Â+Ò 8W;R8&˚â 3â˚rG²%:. (Minsu introduces his friend to Youngju) 2-1. *"þ&/339Ê  ,‘$ê%:. (My younger brother eats only apples) 2-2. *@Ò8&  9î)æ%:. (I cut an apple with a knife) 2-3. .R4n˚V @²˜Â8&˚â ,:%B 3v-™9ı . (Minsu sends a present to his friend every month) 2-4. 4n8W7~˚V 8z$î&ˆ 8WHª˜V8& ˚j%:. (Suyoung went to the theater)  3-1. *4n.N$ê  -¦ GÊ8ê? (What will Sumi do tomorrow?) 3-2. *"î$ê  4RG§9ı ˚j%:. (I went shopping yesterday) 3-3. 4nIÂ$ê 8z$î 7Ó3ã9Ê 9Þ%:. (Suhee has an appointment today) 3-4. 8‡GÊ .R4n$ê  .j2¶ ˚Ù9Ê%:. (Minsu will be busy this year) 4-1. *G²˜Ã G¯33':9Œ 8J5BG®˚â ˜K/ÖG²%:. (Korean students study hard) 4-2. *&ˆ3r˜V 7ž8&3r$ê :Æ8ÿ˚â ˜K/ÖG²%:. (People study in silence in the library) 4-3. 9Ê  0ı5N˚â . (These pictures sell at high prices)   4-4. 8B9æ˚V "þ9æ/J%: 0¶+Ê˚â ,&G²%:. (Women speak faster than men) 5-1. *9Ê >ł9Œ 8B˚Z . (This book is very difficult) 5-2. *4n.N$ê   8B9æ9Ê%:. (Sumi is never a woman who cries) 5-3. .R4n˚V  7ž . (Minsu has not come yet) 5-4. 9Ê ˜I8&$ê .N˜Ã  7ž ˚Z%:. (Only Americans go here) 6-1. *G²˜Ã  9Ò/N  %B7ı8ê. (Korean people and Japanese people are different) 6-2. *4n:k7~$ê 7ı@¾8& 10km+Ò ˚Î*B8ê. (Sujeong walks 10 kilometers in the morning)  6-3.   4Â9"8ê. (Korean is easy if you practice) 6-4.  7ı9Ê':9Ê  <˚*B8ê. (Young children cry out)  7-1. *9Ò9* 4ú-− %:3– 9Ò9Ê  339Ò9Ê%:. (My birthday is January 25)  7-2. *"î$ê 8J %:3– .Þ .Ú4ú+Ò . (I wait for a bus, number 15) 7-3. .R4n$ê A:G™ #z 9ê9ı ,˚3¾%:. (Minsu drank four cups of coffee) 7-4.   8J G² ˚r+Ò . (My older sister bought eleven apples)  121 8-1. *9Ê˚Ù':9Œ "î AJG>Cƒ8Œ . (These are my computer and printer) 8-2. *4n.N˚V #^ .{9ı ,‘$ê%:. (Sumi eats your boiled rice) 8-3.  N9®  8B ƒ3r ,Œ%:. (His company is far from here) 8-4.  @²˜Â9®  ˚j%:. (Yesterday I went to my friends wedding) 9-1. *   ˚V8ê. (Even though I learn Korean, I go to Seoul) 9-2. *.ƒ˚V 7ıEb3r 8WHª˜V8& ˚V8ê. (Because I am sick to my stomach, I went to the theater) 9-3.   '¢-”8&  7žGÊ8ê. (Because I live in a dormitory, I do not cook) 9-4. 0ı˚V 8Œ&ˆ 8BGß9ı ˚V8ê. (Even if it rains, I will go on a trip)  10-1. *.R4n$ê 52H.8&3r 5:4nGÞ ƒ+Ò 7¢7®%:. (Minsu knew that he had made a mistake on the exam) 10-2. *"î$ê  . (It is difficult for me to exercise) 10-3.  4o:r˚V ,$ ƒ '¢-”8&  9ê%:. (Minji cannot sleep because he has a lot of homework) 10-4. G¯33':9Œ  $^9Ê 8z ƒ+Ò . (Students wait for snow) 11-1. * 4n8ê9Ò8&$ê "ö7~˚V :á7ı8ê. (The weather was fine last Wednesday) 11-2. *  '¢   ˚Ù9Ê%:. (I will live in Seoul when I was a child) 11-3. 9ç#ı8&  ˚j%:. (I went to Seoul last year) 11-4. %:9¢ %B8& G²˜Ã8& 9Þ9ı ˚Ù9Ê%:. (I will be in Korea next year) 12-1. *9Ê˚Ù9Ê .R4n˚V  9¢53 ,‘8˚%:. (This is the food that Minsu ate yesterday) 12-2. *G¯339Ê  ˜K/ÖG®$ê 3æ9ı ':8˚%:. (The student who studies Korean raised their hands) 12-3.   9Ó$ê >ł9Œ ?îGû:Z9Ê%:. (The book that I am reading is Chun-hyang-jeon) 12-4. 9Ê˚Ù9Ê 8B+Ú8& G®8Œ9Ê8&3r 9Û9ı 8“9Ê%:. (These are the clothes that I will wear in Hawaii) 13-1. *"î$ê .R4n˚V  Œ@®+Ò :á7ıG® ƒ+Ò . (I did not know that Minsu liked Kimchi) 13-2. *"î$ê 4n.N˚V ˜K/ÖGÞ ƒ+Ò 7¢7®%:. (I knew that Sumi had studied) 13-3. 3⁄.N˚V >ł9Ó ƒ+Ò :á7ıG®$ê ˚Ù9ı 7¢7®%:. (I knew that Seongmi liked reading books) 13-4.  ƒ>~˚V 6528& 'ö"î$ê ˚Ù9ı . (I did not know that the train would leave at 6) 14-1. *BÆ 8“9Ê 9î Ef%:. (Large clothes sell well)  14-2. *.{9Ê ,19Þ˚â  N+Ý8& %J7®%:. (The boiled rice was put into the bowl deliciously)  14-3.  N <'9Ê .R4n8&˚â Ej*”%:. (The house was sold to Minsu) 14-4. &ˆ&§9Ê  9÷Hj%:. (The thief was caught by the police officer) 15-1. *.R4n˚V 4n.N8&˚â 3v-™9ı 7ž$ê%:. (Minsu pressed presents on Sumi) 15-2. *@²˜Â˚V "î8&˚â  /N%:. (My friend showed me a picture) 15-3.  &/338&˚â >ł9ı 9ÓIâ%:. (I have my younger sibling read a book) 15-4. $Z"î˚V 7ı ƒ8&˚â  ,‘9Î%:. (I feed the baby milk) 16-1. *4n.N7Ò, 9Ê˚Ù9ı  . (Sumi, give this to your grandfather) 16-2. *7ı.Ú%.!.3r  9î 9ê%:. (My father is sleeping well in the room)  122 16-3.  ˚V52,Ê :ZHªG®3”8ê. (Please call me when you get to Seoul) 16-4.  >fi-”9ı 8B5C%˚%:. (The grandfather opens the window)   17-1. *7ı9Ê':9Ê ,ÿ9Ê A::”3r ,&9ı 7ž . (Children do not listen to me because they get stubborn) 17-2. * N ,&9ı '9˜6  H²&/GÞ%:. (The words made my heart race) 17-3.  N$ê 9æ3”+Ò ˚V%:'B˜6 9ÎCƒ0F+Ò GÞ%:. (He interviewed after he became composed) 17-4.  N ,&9ı '9˜6 )N 'E9Ê  ƒ1:%:. (I jumped for joy to hear the news)    A-3. Stimuli for the Metalinguistic Knowledge Test *.R4n$ê ?ë˜Â+Ò GÊ3r 9î 9Î ƒ˚V ,$%:. (Minsu is popular because he plays soccer well) -The adverb should come in front of the verb that it is modifying. -The topic subject particle is wrong: .R4n˚V is correct.  -The conjunction 3r is awkward. The conjunction %˚ ¢ is better to express the reason or cause in the main clause.    *.R4n+Ò  ;R4ú˚V ,˚3ª8ê. (Minsu drinks orange juice)  -The particles after the two nouns are awkward. -Subjects should be followed by objects. Thus ;R4ú˚V should come first in the sentence. -The verb conjugation is awkward.  should not be contracted.   *"î$ê 8z$î8& G¯˜¦8& ˚V3r 52H.9ı &¦ ˚r ?:%:. (I took two exams at school today)  -When 8z$î is used as an adverb, -8& should be dropped. -The locative particle 8& should be 8&3r when the verb is go  -˜R,ÿ, a Sino-Korean numeral 9Ê should be used to denote two  *H:.c9>9Ê$ê  7N7N˚â . (Hemingway died a lonely death when he got old) -When an adjective ends with G®%:, the correct adverbializing suffix is IÞ. -For the past tense of ;S%:, the past tense marker 8˚ should be used, not ˚j. -The phrasal connector  is awkward; the connector - '¢ correctly describes when he was old. A²)Ò$ê   :á%:. (Coke is never good for health) -The adverb  requires a negative element within the same sentence. - The adverb  should come in the beginning of the sentence to put emphasis on the adverb.  -The particle 8& after  should be changed to a subject particle. 9æ&/>~˚V  3æ9÷9Ê+Ò !… 9æ9"8ê. (Hold onto the strap tightly because the car shakes)   -The verb in the main clause is a -regular verb. Therefore  is not weakened to a semivowel  []. -The phrasal connective %˚ is not correct here. A cause-effect connector should be used. -The adverb in the main clause modifies the whole clause. Therefore !… should come before 3æ9÷9Ê+Ò.  123 * ;R8& 529û8&3r ˚r &¦ ˚r+Ò . (I bought two dogs at the market) -˚r is a classifier for inanimate objects like apples. -The correct Korean number for two is &®.  -When a noun becomes an adverb for time, 8& is unnecessary.  *9Ê:r G² ˜â >ł9ı 3â˚rG¿%˚%:. (Now I introduce a book) -When a numeral is followed by a classifier, the genitive-particle deletion is not allowed. -In this sentence, the object particle is unnecessary.  -To begin a sentence, adverbs should be avoided.  *0ı˚V    . (I could not go mountain climbing because it rained) -Conjunctive  can never be preceded by a past tense marker. Therefore,  should be 8Œ3r. -When the verb is a transitive verb, the object has to have an object particle 9ı.  -It is correct to use another negative word 7ž to describe the subjects subjective decision.    *"î$ê  #Î)îG¾9Ê :á%:. (I particularly like singing) -When a verb ends with G®%:, the nominalization suffix should not be G¾  -The adverb should be placed right in front of the verb that it is modifying. -When the sentence has  a special adverb, the subject should have the generic subject particle. Therefore  is correct. * ;R,&8& .⁄Hª:f8& 4RG§9ı ˚V8ê. (I go shopping to the department store last weekend) -The tenses do not agree between the adverbial clause and the verb. -When the verb is go or come the correct locative particle is 8&3r, but not 8&. -4RG§˚V%: is a one word, which cannot be separated with the object particle, 9ı.  *9ÎH”$ê Eb$ê @²˜Â˚V   (Inho bought oranges that his friend was selling) - is modified by the relative clause, so the clause should be in the order of Subject , then Verb.  -The second ˚V should be $ê, another subject particle. -The relationship between 9ÎH” and @²˜Â is possession. Therefore, a possessive particle should be used like9ÎH”9® Eb$ê @²˜Â˚V. *:V$ê  . (I do not know how to drive) -To mean the know-how of a skill, the bound noun ;Z should be used. - is an object and this needs an object particle, +Ò. -:V is a honorific term so the verb  should be in agreement in terms of honorifics.  *Cö"Ò˚V  ,‘$ê%:. (Rabbits are eaten by lions) -Cö"Ò is the object and  is the agent of the verb. The verb should be changed to a passive voice.  -The subject is the topic of the sentence, so the subject particle should be the topic particle, $ê.  124 -The particle 8&˚â is a dative particle and should be changed to an inclusive particle, &ˆ. * 7ı ƒ8&˚â 7Ó9ı ,‘$ê%: (The doctor feeds the baby medicine) -The subject  makes the baby eat the medicine. Therefore the verb should have a causative suffix, 9Ê.  -The subject particle ˚V should be changed to !.3r, an honorific particle for mother. -The verb is a transitive verb which requires two objects (direct and indirect) and should be ordered with the direct object first and the indirect object second.  */Ö,þ%.!.3r 8z$î .{9ı 7ž ,‘$ê%:. (My parents do not eat boiled rice today) -The verb should have the honorific form for the subject. -The topic subject marker should be added to !.3r, resulting in !.3r$ê. -The negative marker  should be used instead of 7ž to describe a situation that does not allow them to eat rice.  *9Ê.Þ  7ı9Ê':9Ê ˚×9ı . (Children got scared because of this accident) -The predicate is a Korean collocation and ˚×9ı should be followed by ,‘8˚%:. -To denote the cause-effect relationship,  , accident, is described better using '¢-”8& than *². -The subject of the whole sentence is accident, . Therefore the particle *² should be either a topical subject marker $ê or a regular subject marker, ˚V  125 A-4. Background Questionnaire  1. Research code: ____________  Gender: ________     Age:  ____________ University: _____________  2. Current Korean class level or course name: __________________  - Class standing: Major _______________ Minor ______________ -The Korean language is your 1) heritage language______    2) foreign language______ -Specify any Chinese or Japanese courses you have taken so far ____________________________  3. Place of birth _______________ -If you were not born in the US, at what age did you come to the US: __________ years old -How long have you studied Korean as a heritage language or a foreign language ______ years ___ months since _______________ (year) -List all institutes (For example, Michigan State University) ______________________________________________________________________ -List all courses you took previously (For example: Korean 101) ______________________________________________________________________ -List all books you learn with (For example: Integrated Korean Beginning 1) ______________________________________________________________________  4. Check all the family members who you have lived with for over 6 months of your life that are native speakers of Korean. -Grandmother, grandfather, mother, father, brother(s), sister(s), relatives, spouse (significant other) -Biological Fathers ethnic background _______, Biological Mothers ethnic background ________  5. When you were a child, what was your first language (First language is the language your parents communicate with you before your age of 5?  -English______, Korean______, Both______ -Any other languages _____________________ -When growing up, I have received exposure of Korean language and culture. NA___ never____, rarely____, sometimes_____, often_____, almost always _______, 6. Have you ever lived in Korea for more than one month? -Yes____, No____ -If yes, at what age(s) ___________ For how long? _______ year(s) ______ month(s) 7. Have you ever studied in Korea? -If yes, list the school level or program (for example, K-12, college, summer school, study abroad, etc) -------------------- -For how long? _______ year(s) ________ month(s) 8. Have you ever studied Korean at other formal non-college institution in the US? (For example, Saturday school, Sunday school, high school extra curriculum activity) -Yes____, No____ -If yes, List institution _______ -At what age ______ for How long? _____ year(s) _____ month(s)  126 9. Have you visited Korea for less than one month? -Yes_____, No_____ -If yes, at what age(s) ___________ For how long? _______ year(s) ______ month(s)             Reason(s) of visit ________________________________________________  10. How often do you speak Korean with the following people?  NA 1.never 2. rarely 3.sometimes 4.often 5.always Friends        Significant other/spouse       Korean classmates / teacher       Grandparent(s)       Mother       Father       Sibling(s)       Relative(s)       Others (specify)          11. How often do the following people speak Korean to you?  NA 1.never 2. rarely 3.sometimes 4.often 5.always Friends        Significant other or spouse       Korean classmates or teacher       Grandparent(s)       Mother       Father       Sibling(s)       Relative(s)       Others (specify)         12. Please tick your other uses or exposure of Korean  1.never 2. rarely 3.sometimes 4.often 5.always I watch Korean/TV       127 movies I read books, newspapers, magazines in Korean      I write email, journals, in Korean      I listen to Korean music      Other(s) __________________        13. I am confident communicating in  1.strongly agree 2. agree 3.not sure 4.disagree 5.strongly disagree Korean      English      Other ___________________      Other ___________________       14. I am confident in the following Korean language skills  1.strongly agree 2. agree 3.not sure 4.disagree 5.strongly disagree Speaking      Listening      Reading      Writing      Grammar      Honorifics (polite style)      intimate/casual form (panmal)      Other(s) ___________________        15. If you have every taken a standardized test/exam of Korean,  -Name(s) of the test/exam ______________________________________________ -Score(s) ____________________________________________________________   16. What do you think of this study? Any comments and questions are welcome. ______________________________________________________________________ ______________________________________________________________________ ______________________________________________________________________ ______________________________________________________________________  128 ______________________________________________________________________ ______________________________________________________________________  17. How have you studied Korean grammar? Please share your strategies and experiences. ______________________________________________________________________ ______________________________________________________________________ ______________________________________________________________________ ______________________________________________________________________ ______________________________________________________________________ ______________________________________________________________________  18. How have your teachers of Korean taught Korean grammar? Please explain about that in detail. ______________________________________________________________________ ______________________________________________________________________ ______________________________________________________________________ ______________________________________________________________________ ______________________________________________________________________ ______________________________________________________________________  19. How are the tests, quizzes, and examinations for evaluating Korean grammar in class?  ______________________________________________________________________ ______________________________________________________________________ ______________________________________________________________________ ______________________________________________________________________ ______________________________________________________________________ ______________________________________________________________________ ______________________________________________________________________  20. Writing in Korean - When did you learn how to write Korean? At the age of _______ in the year of ________ - How did you learn how to write Korean? Who taught you to write Korean?  129 APPENDIX B. Stimuli (Pilot study II)   B-1. Stimuli for Elicited Imitation Test B-2. Stimuli for the Aural and Written Grammaticality Judgment Tests B-3. Stimuli for the Metalinguistic Knowledge Test (Sections 1 and 2) B-4. Oral Narrative Test   B-5. Background Questionnaire B-6. Proficiency Test  B-1. Stimuli for Elicited Imitation Test (The highlighted words are adverbs or adjectives that have been added due to the word count) 1.1-1. *7ı9Ê':9Ê 9î  . (Children draw pictures well) 2.1-2. /Z8& 8^2² !Ž9Ê ,$9Ê 9Þ%:. (There are many beautiful flowers in spring) 3.2-1. *3æ9Ê :Û9Ê+Ò ˚V"ª .ö9Î%:. (People often cut their hands with a piece of paper) 4.2-2. G²˜Ã  8WHª+Ò  :á7ıG²%:. (Koreans like movies very much) 5.3-1. * "ö7~˚V 7ž :á7®%:. (Yesterday the weather was not good) 6.3-2. 8z$î ˜K9&8&3r ,$9Œ 7ı9Ê':9Ê #Ò%:. (Many children hang out in the park today) 7.4-1. *.N˜Ã8&3r >ò>òG®˚â  G²%:. (You should drive slowly in the US) 8.4-2. 7ı9Ê':9Œ Cª*^0ı:Z  BÂ˚â G²%:. (Children keep up the volume of TV) 9.5-1. *   ,&9Ê%:. (Korean is never easy) 10.5-2.  G²˜Ã9®  Œ@®+Ò :ZHV ,þ+Î%:. (People never know Korean kimchi) 11.6-1. *,$9Œ   . (Many people live in Seoul) 12.6-2. G²˜Ã9®   7ı;R ˚V ¢9"8ê. (It is very close from Seoul to Busan in Korea) 13.7-1. *8B ƒ8&   ˚r 9Þ%:. (Here are 300 people) 14.7-2.   @¶ @‰8& 9Þ%:. (We are on the seventh floor now) 15.8-1. *8z.j,˚$ê #^ @²G² @²˜Â9Ê%:. (Obama is your close friend) 16.8-2.  G²˜Ã9® "ö7~$ê 7ı;R :á%:. (Now, the weather in Korea is very good) 17.9-1. *G¯339Œ G¯˜¦8& , ˜K/Ö+Ò G²%:. (Even though students go to school, they study) 18.9-2. 0ı˚V ,$9Ê , 4n8W9ı G²%:. (Even though it rains a lot, I swim) 19.10-1. * .z8&  . (People do not like exercising at night) 20.10-2. G¯33':9Œ 52H.8& 5:4nG² ˚Ù9ı 7ž%:. (Students know that they made mistakes in the exam) 21.11-1. * G²˜Ã  4n˚V $î8˚%:. (Next year, Korean population increased) 22.11-2.  GÊ Gfi9Œ >” . (The Olympics were very interesting last year) 23.12-1. *#Î)î+Ò 9î G®$ê ,¡9Þ%: @²˜Â˚V. (Friends who sing well are cool) 24.12-2. 8B9æ$ê  :á7ıG®$ê "þ9æ+Ò :á7ıG²%:. (Women like men who are good at sports)  130 25.13-1. *#^$ê G¯33':9Ê A:G™+Ò :á7ıG²%: 7ž%:. (I know that young people like expensive coffee) 26.13-2.  52˚Z9Ê ˚V$ê ˚Ù9ı ,þ+Î%: (People do not know that time goes)  27.14-1. * G¯33':8&˚â 9î Ef%:(Smart phones are sold well to students) 28.14-2. 8ê;Þ9¢ G²˜Ã >ł9Ê 9î 9ÓIâ%:. (These days, Korean books read well) 29.15-1. *8ˆ,˚˚V .z8& 7ı ƒ+Ò 9ê%:. (Mothers put their babies to sleep at night) 30.15-2.   &/338&˚â 8“9ı 9ÛIâ%:. (Usually an older sister dresses her younger sibling) 31.16-1. * .{9ı ,19Þ˚â ,‘$ê%:. (The grandfather eats boiled rice deliciously) 32.16-2.    G²˜Ã8& ˜ˆ56%:. (My father and mother are in Korea now) 33.17-1. * ?ë˜ÂG¶ '¢  )J%:. (Usually people run stealthily when they play soccer) 34.17-2.   ƒ/Ú9Ê :á9ı '¢ H²<3 9ˇ$ê%:. (People have big grins on their faces when they feel good)       B-2. Stimuli for the Aural and Written Grammaticality Judgment Tests (The highlighted words are adverbs or adjectives that have been added due to the word count and that could be removed for higher Cronbachs Alphas) 1.1-1. *8z$î ,$9Ê .RH”˚V >ł9ı 9Ó$ê%:. (Today Minho reads books a lot) 2.1-2. * 8^2®˚â 7ı;R Hª9û9ı GÞ%:. (Minji had very beautiful makeup)  3.1-3. 9v.N˚V @²˜Â8&˚â ,:9Ò  7J%:. (Yumi writes a letter to a friend every day) 4.1-4. .R4n˚V @²G² @²˜Â+Ò 8W;R8&˚â 3â˚rG²%:. (Minsu introduces his close friend to Youngju) 5.2-1. *"þ &/339Ê ˜6 ƒ8Œ  ,‘$ê%:. (My younger brother eats only meat and apples) 6.2-2. *  @Ò8&  9æ+Î%:. (Usually people cut an apple with a knife) 7.2-3. .R4n˚V @²˜Â8&˚â ,:%B 3v-™9ı . (Minsu sends a present to his friend every month) 8.2-4. 4n8W7~˚V 8z$î 7ı@¾8&&ˆ 8WHª˜V8& ˚j%:. (Suyoung went to the theater this morning, too)  9.3-1. *.R4n$ê  8WHª+Ò /R ˚Ù9Ê%:. (Minsu will watch a movie tomorrow) 10.3-2. *"î$ê  .⁄Hª:f8& 4RG§9ı ˚j%:. (I went shopping to a department store yesterday) 11.3-3. 4nIÂ$ê 8z$î ;g8êG² 7Ó3ã9Ê 9Þ%:. (Suhee has an important appointment today) 12.3-4. 8‡GÊ .R4n$ê  .j2¶ ˚Ù9Ê%:. (This year, Minsu will be busy in his company) 13.4-1. *G²˜Ã G¯33':9Œ  8J5BG®˚â ˜K/ÖG²%:. (Korean students usually study hard) 14.4-2. *7ı9Ê':9Ê &ˆ3r˜V 7ž8&3r :Æ8ÿ˚â ˜K/ÖG²%:. (Children study in silence in the library)  131 15.4-3. 9Ê  7ı;R 0ı5N˚â . (These pictures sell at very high prices)   16.4-4. 8B9æ˚V "þ9æ/J%: :Æ ^ 0¶+Ê˚â ,&G²%:. (Women speak a little faster than men) 17.5-1. *9Ê >ł9Œ G¯33':8&˚â 8B˚Z . (This book is not difficult for students at all) 18.5-2. *4n.N$ê  ,$9Ê  8B9æ9Ê%:. (Sumi is never a woman who cries a lot) 19.5-3. .R4n8Œ @²˜Â˚V  7ž . (Minsu and his friend have not come yet) 20.5-4. 8B ƒ$ê G²˜Ã  7ž 8~%:. (Only Koreans come here) 21.6-1. *G²˜Ã  9Ò/N  %B7ı8ê. (Korean people and Japanese people are different) 22.6-2. *4n:k7~$ê ,:9Ò 7ı@¾10km+Ò ˚Î*B8ê. (Sujeong walks 10 kilometers every morning)  23.6-3.  8J5BIÞ   4Â9"8ê. (Korean is easy if you practice diligently and interestingly) 24.6-4.  7ı9Ê':9Ê   <˚*B8ê. (Young children cry out everywhere)  25.7-1. *9Ò9* 4ú-− 9Ò9Ê  339Ò9Ê%:. (My birthday is January 20)  26.7-2. *"î$ê 8J %:3– .Þ .Ú4ú+Ò . (I wait for a bus, number 15) 27.7-3. .R4n$ê A:G™ #z 9ê9ı ,˚3¾%:. (Minsu drank four cups of coffee) 28.7-4.   8J G² ˚r+Ò . (My older sister bought eleven apples) 29.8-1. *9Ê˚Ù9Œ "î AJG>Cƒ8Œ . (These are my computer and printer) 30.8-2. *4n.N˚V #^ .{9ı ,19Þ˚â ,‘$ê%:. (Sumi eats your boiled rice deliciously) 31.8-3.  N9®  8B ƒ3r  ,Œ%:. (His company is very far from here) 32.8-4.  @²˜Â9®  0ıGß ƒ+Ò  ˚j%:. (Yesterday I flew to a friends wedding) 33.9-1. * ,   ˚Ù9Ê%:. (Even though I go to Seoul, I will learn Korean) 34.9-2. *.ƒ˚V 7ıEb3r 8WHª˜V8&  ˚V8ê. (Because I am sick to my stomach, I go to the theater quickly) 35.9-3.   '¢-”8&  7ž GÊ8ê. (Because I live in a dormitory, I do not cook) 36.9-4. 0ı˚V ,$9Ê 8Œ&ˆ 8BGß9ı ˚V8ê. (Even if it rains a lot, I will go on a trip)  37.10-1. *.R4n$ê 52H.8&3r 5:4n+Ò GÞ ƒ+Ò 7¢7®%:. (Minsu knew that he had made a mistake on the exam) 38.10-2. *"î$ê .z8&   . (It is difficult for me to exercise at night) 39.10-3.  4o:r˚V ,$ ƒ '¢-”8&  9ê%:. (Minji cannot sleep because she has a lot of homework) 40.10-4. 7ı9Ê':9Œ :Æ8ÿIÞ >ł 9Ó ƒ˚V  (It is difficult for children to read books quietly) 41.11-1. * 4n8ê9Ò8&$ê "ö7~˚V 7ı;R :á7ı8ê. (The weather is quite fine last Wednesday) 42.11-2. *  '¢   ˚Ù9Ê%:. (I will live in Seoul when I was a child)  132 43.11-3. 9ç#ı ˚V9ı8& 8z0¶G®˜6  ˚j%:. (I went to Seoul with my older brother last fall) 44.11-4. %:9¢ %B8& G²˜Ã8& 9Þ9ı ˚Ù9Ê%:. (I will be in Korea next month) 45.12-1. *9Ê˚Ù9Ê .R4n˚V  9¢53 ,‘8˚%:. (This is the food that Minsu ate yesterday) 46.12-2. *G¯339Ê  ˜K/ÖG®$ê 3æ9ı ':8˚%:. (The student who studies Korean raised their hands) 47.12-3.   9Ó$ê >ł9Œ . (The book that I am reading is Harry Potter) 48.12-4. 9Ê˚Ù9Ê 8B+Ú8& G®8Œ9Ê8&3r 9Û$ê 8“9Ê%:. (These are the clothes that I wear in Hawaii) 49.13-1. *"î$ê .R4n˚V  Œ@®+Ò :á7ıG® ƒ+Ò . (I did not know that Minsu liked Kimchi) 50.13-2. *"î$ê 4n.N˚V  ˜K/ÖGÞ ƒ+Ò 7¢7®%:. (I knew that Sumi studied last night) 51.13-3. 3⁄.N˚V >ł9ı :á7ıG®$ê ˚Ù9ı 7¢7®%:. (I knew that Seongmi liked reading books) 52.13-4.  ƒ>~˚V 6528& 'ö"î$ê ˚Ù9ı . (I did not know that the train would leave at 6) 53.14-1. *8B ƒ3r$ê BÆ 8“9Ê 9î Ef%:. (As for here, large clothes sell well)  54.14-2. * N 8^2² <'9Ê 4n.N8&˚â /J7®%:. (The pretty house was seen by Sumi)  55.14-3.  N #i9Œ <'9Ê .R4n8&˚â Ej*”%:. (The spacious house was sold to Minsu) 56.14-4.  N   9÷Hj%:. (The person was caught by the police) 57.15-1. *.R4n˚V 4n.N8&˚â ,$9Œ 3v-™9ı 7ž$ê%:. (Minsu presses many presents on Sumi) 58.15-2. *.RH”˚V @²˜Â8&˚â   /N%:. (Minho shows his friend a picture and a photo) 59.15-3.  &/338&˚â G²˜Ã >ł9ı 9ÓIâ%:. (I have my younger sibling read a Korean book) 60.15-4.  7ı ƒ8&˚â ,19Þ$ê  ,‘9Î%:. (Minji feeds the baby delicious milk) 61.16-1. *4n.N˚V  :á9Œ 3v-™9ı ;V%:. (Sumi gives a good present to her grandfather) 62.16-2. *3r  H™9æ 9î 9ê%:. (My father sleeps well in the room alone) 63.16-3.  3v33%.!.3r Bú˚V 7ı;R BÂ5C%˚%:. (The Korean teacher is very tall) 64.16-4.   <'8&3r  G®56%:. (My grandfather is exercising at home)   65.17-1. *7ı9Ê':9Ê ,ÿ9Ê A::”3r ,&9ı 7ž . (Children do not listen to me because they get stubborn) 66.17-2. * N ,&9ı '9˜6  . (The words made my heart race) 67.17-3.  N$ê 9æ3”+Ò ˚V%:'B˜6 9ÎCƒ0F+Ò GÞ%:. (He interviewed after he became composed) 68.17-4.  N ,&9ı '9˜6 )N 'E9Ê  ƒ1:%:. (I jumped for joy to hear the news)   B-3. Stimuli for the Metalinguistic Knowledge Test   133 Section 1.  1.*.R4n$ê 9î ?ë˜Â+Ò GÊ3r 9Î ƒ˚V ,$%:. (Minsu is popular because he plays soccer well) -The adverb 9î should come in front of the verb that it is modifying. -The correct adverb is :ZHV to emphasize the adjective ,$%:.  -The correct position of adverb 9î is between 9Î ƒ˚V and ,$%:.  -When the conjunction 3r is modified by 9î,  3rshould be changed to to express the reason or cause in the main clause.     2.*.R4n$ê  ;R4ú9Ê ,˚3ª8ê. (Minsu drinks orange juice) -;R4ú is an object. Therefore, an object particle is necessary.  -When the noun ends with a vowel, the correct particle is ˚V, not 9Ê. -Subjects should come first in a sentence. Thus ;R4ú9Ê should come before .R4n$ê. -The verb ,˚3ª8ê requires a locative particle 8&3r.   3.*"î$ê 8z$î8& G¯˜¦8& ˚V3r 52H.9ı &¦ ˚r ?:%:. (I took two exams at school today)  -When 8z$î is used as an adverb, -8& should be dropped. -The temporal particle 8& should be 8&3r when the verb is go  -8z$î8& is not correct because the sentence is about the past tense, ?:%:. -8z$î8& is an adverb and should be in front of the subject "î$ê.  4.* N$ê  7N7N˚â . (He died a lonely death when he got old) -When an adjective ends with G®%:, the correct adverbializing suffix is IÞ. -7N7N˚â should be placed in the beginning of the sentence because 7N7N˚â modifies the whole sentence. -Because the tense is the past, 7N7N˚â should be conjugated accordingly. -The phrasal connector  should be $ï8˚9ı '¢ when followed by an adverbial expression.   5.*A²)Ò$ê :ZHV  :á%:. (Coke is never good for health) -The adverb :ZHV requires a negative element within the same sentence. -The adverb :ZHV should come in the beginning of the sentence to put emphasis on the subject.  -:ZHV requires an object. Therefore, the particle 8& after  should be changed to an object particle 9ı. -When a sentence has :ZHV, the topic marker should not be used.   6.*  3æ9÷9Ê+Ò !… 9æ9"8ê. (Hold onto the strap tightly because the car shakes) -The verb 9æ9"8ê is a -regular verb. Therefore  should be kept in the stem, 9÷. -The conjugation should be in deferential style using 5C528z when this is announced in a subway station.  134 -9æ9"8ê should be a passive voice in the sentence. 9÷HV8ê is correct. -9æ9"8ê is a verb. Therefore this does not require an adjective, !….  7.* ;R8& 529û8&3r ˚r+Ò &¦ ˚r . (I bought two dogs at the market) -˚r is a classifier for inanimate objects like apples.  -The correct Korean number for two is &®.  -The correct Sino-Korean number for two is 9Ê. -&¦ ˚r is a subject that is associated with , a verb. Therefore &¦ ˚r˚V is correct.   8. Low reliability *9Ê:r G² ˜â >ł9ı 3â˚rG¿%˚%:. (Now I introduce a book) -When a numeral is followed by a classifier, the genitive-particle deletion is not allowed. -In this sentence, the object particle is unnecessary.  -To begin a sentence, adverbs should be avoided. -The numeral G² modifies >ł, the noun. Therefore G² should be followed by the noun instead of the classifier.  Modified  *9Ê:r >ł9ı 9Ò ˜â 3â˚rG¿%˚%:. (Now I introduce a book) -When you count an object, a different numeral than 9Ò should be used. -The order of the words is awkward. >ł9ı should come after ˜â, because ˜â modifies the noun, >ł. -The object particle should come after ˜â, because ˜â is the object of the transitive verb, 3â˚rG¿%˚%:. -The classifier is not correct. A different classifier should be used for >ł.   9.* 0ı˚V    . (I could not go mountain climbing yesterday because it rained) -Conjunctive  can never be preceded by a past tense marker. Therefore, 3r should be 8Œ3r. - is a transitive verb. Therefore 0ı  requires an object marker.  -Logically it is correct to use a negative word 7ž in front of  in this sentence.  -Conjunctive  should be replaced  to mean even though.     10.*"î$ê  #Î)îG¾9Ê :á%:. (I particulary like singing) -When a verb ends with G®%:, the nominalization suffix should not be G¾  -:á%: is a transitive verb and #Î)îG¾ should be an object of the verb. 9ı should be used instead of 9Ê. -To agree with , #Î)îG¾9Ê should be #Î)îG¾9Œ to make this a topical expressions. -:á%: is an adjective. Therefore we should use a verb :á7ıG²%:   11.* ;R,&8& .⁄Hª:f8& 4RG§9ı ˚V8ê. (I got shopping to the department store last weekend) -The tenses do not agree between the adverbial clause  ;R,& and the verb. -When the verb is go or come the correct locative particle is .⁄Hª:f8&3r.   135 -4RG§˚V%: is a one word, which cannot be separated with the object particle, 9ı.  -4RG§8& ˚V8ê is the correct collocation. Accordingly .⁄Hª:f8&3r should  be .⁄Hª:f8&to mean a destination.   12.*9ÎH”$ê Eb$ê @²˜Â˚V  . (Inho bought oranges that his friend was selling) - is modified by the relative clause, so the clause should be in the order of Subject , then Verb.  -The  ˚V should be $ê, another subject particle. -The ˚V should be 8&˚â3r to refer to a person that 9ÎH” buys oranges from.  -The relationship between 9ÎH” and @²˜Â is possession. Therefore, a possessive particle should be used like9ÎH”9® Eb$ê @²˜Â˚V.  13. Low reliability *:V$ê   . (I do not know how to drive yet) -To mean the know-how of a skill, the bound noun ;Z should be used. - is an object and this needs an object particle, +Ò. - The verb  should be,þ+Ê3”8ê because :V is a humble word.  - requires an object such as >~+Ò or .Ú4ú+Ò in front of it.   Modified  *:V$ê   . -To mean the know-how of a skill, the bound noun ;Z should be used. Therefore :ZG¶ ;Z9ı is correct - is an object in itself. Therefore, this does not need an object particle, +Ò. - The verb  should be,þ+Ê3”8ê because :V is a humble word.  - requires an object such as >~+Ò or .Ú4ú+Ò in front of it.   14.* ;¦˚V ˜67ç9Ê8&˚â ,‘$ê%:. (Usually mice are eaten by cats) -˜67ç9Ê is the agent of the verb. The verb should be changed to a passive voice.  -˜67ç9Ê is the topic of the sentence, so the particle should be the subject  particle. -The particle 8&˚â is a dative particle and requires a causative verb, ,‘9Î%:. -Usually, mice do not eat cats. Therefore, ;¦ and ˜67ç9Ê should be switched.  15.* 7ı ƒ8&˚â 7Ó9ı ,‘$ê%:. (The doctor feeds the baby medicine) -The subject  makes the baby eat the medicine. Therefore the verb should have a causative suffix, 9Ê.  -The subject particle ˚Vmust be changed to !.3r, an honorific particle for a doctor. -The verb is a transitive verb which requires two objects (direct and indirect) and should be ordered with the direct object first and the indirect object second.  - To denote the passive relationship between the doctor and the baby, the verb ,‘$ê%: should be a passive verb.  16.*/Ö,þ%.!.3r 8z$î .{9ı 7ž ,‘$ê%:. (My parents do not eat boiled rice today)  -The verb ,‘$ê%: should have the honorific form with the correct word for the subject.  136 -The topic subject marker should be added to !.3r, resulting in !.3r$ê. -The negative marker  should be used instead of 7ž to describe a situation that does not allow them to eat rice.  -This statement should be in deferential style due to the subject. Therefore, ,‘9™5C%˚%: is correct.   17.*9Ê.Þ  ,$9Œ 7ı9Ê':9Ê ˚×9ı %:. (Children got scared because of this accident) -The expression is a Korean collocation and ˚×9ı should be followed by ,‘8˚%:. -To denote the cause-effect relationship,  , accident, is described better using '¢-”8& than *². -The subject of the whole sentence is accident, . Therefore the particle *² should be either a topical subject marker $ê or a regular subject marker, ˚V - is the past tense. Therefore 9Ê.Þ should be :V.Þ to denote the past.  (Children were scared of this accident)                                   137 Section 2.  In the text, find ONE example for each item (on the next page).  Underline the word(s) and put the number of the item below the line. Example:       0.  Noun                       "î$ê G²˜Ã 9Ê%:.                               0 Text     "î$ê ;g˜Ã 8&8ê.   G²˜Ã8& . G²˜Ã9Œ ˜68êG² 7ı@¾9® "î)Ò)Ò˜6 ,&GÊ:”8ê. "î$ê  7ı@¾ 8528&  3”4n+Ò G®˜6 7ı@¾9ı ,‘. 952>: G²˜Ã  G¯˜¦8& ˚V8ê. "î$ê 3#ı :Z/ÖCƒ G²˜Ã .ƒ96. G²˜Ã    7€7ı8ê. G²˜Ã 4n8˙9Ê "³"î,Ê 8zIˆ 5 &ˆ3r˜V8&3r ˜K/ÖGÊ8ê.  &ˆ3r˜V8&3r  4o:r+Ò G®˜6 .   3v33%.9Ê "î8&˚â 9Ê 4o:r+Ò 52Bú3¾. &ˆ3r˜V9Œ ;¦ ;S9Œ 'E :Æ8ÿGÊ8ê.      8z$î9Œ  ^8ê9Ò9Ê)Ò3r 8zIˆ8& 4n8˙9Ê 8˝.  ^8ê9Ò :V#Š8&$ê G²˜Ã @²˜Â':9ı ,""î3r ˚o9Ê 53 G²˜Ã 8WHª+Ò /f8ê. "î$ê G²˜Ã 8WHª˚V  ˚Ù9ı 9î 7¢7ı8ê.  @²˜Â':9Œ  9î , G²˜Ã,&*² 9Ê7Ò ƒG®,Ê3r .N9Þ˚â 52˚Z9ı . 8ê;î G²˜Ã,&9ı 9î G®$ê @²˜Â': '¢-”8& G²˜Ã G²˜Ã -”Hª8& %VG®8B ,$9Ê .ƒ9"8ê.   G²˜Ã,&9ı 9î  ,",  9î G®˜6 5L.  N)î3r G²˜Ã 8WHª+Ò 9æ,˜ 8˝9Ê /R ˚Æ8^8ê.  1. Topic particle (=topic marker) 2. Object particle 3. Genitive particle (=possessive particle) 4. Locative particle  5. Sino number 6. Noun classifier 7. Noun as adverb  8. Adverbial suffix  9. Conjunctive, conjunction 10. Transitive verb  11. Nominalization 12. Relativization 13. Complementation 14. Passives 15. Causatives 16. Honorific expression  17. Collocation       138 B-4. Oral Narrative Test     Oral Narrative Test  Read the following text twice very carefully. When you are ready recall the story in Korean orally in three minutes. Try to recall as much as possible.      .R4n7~$ê (company)8&3r 9ÒG¿%˚%:. 7ı@¾8& .R4n7~$ê 7528& %˚%:. 7ı@¾9™*² A:G™ G² 9ê˜R 0Ë9ı %:.   Hâ ˚g%˚%:. <'8&3r  30/Ú . .R4n7~$ê 8z:Z 9528& 8& ˚V3r 8zIˆ 5 9ÒG¿%˚%:. 3æ%.(customer)9ı ,""ö '¢&ˆ 9Þ˜6,  9ÒG® ƒ&ˆ G¿%˚%:. .R4n7~$ê 9ÒG® ƒ˚V 5A9™,Ê, :Æ8ÿIÞ H™9æ(alone) 9Ê*]˚â ,&G¿%˚%:. 5 9&(five hundred million won)9Ê 33 ƒ,Ê  7ž ˚^ ˚Ù9Ê%:.      N*F%Æ  .R4n7~8&˚â (surprising) 9Ò9Ê .       7ı@¾ .R4n7~˚V  ˚V$ê  ”8&  (wallet)9ı 1˚r /J7®.  7ž8&$ê &˚ 5," 9&˜R /K˜â(lottery ticket) 19û9Ê 9Þ8˚. 8&3r AJG>Cƒ*² /K˜â(lottery ticket)9® .ÞH”(number)+Ò >fl7ı %˚%:. .R4n7~$ê 9æ ƒ $^9ı  .U9ı 4n˚V .  N /K˜â9Œ 1.      .R4n7~$ê  N /K˜â9ı  ˚Ù9ı 33˚W.   N /K˜â9ı ;R9Î(owner)9Î  Œ3v338&˚â &"*z;R$ê ˚Ù&ˆ . 9ö52 Iˆ8& .R4n7~$ê ˚g˜R /K˜â9ı  Œ3v338&˚â &"*z;R ƒ*² .  C352*²  Œ3v339® <'8& .       N <'8& ˚j9ı '¢ G¶,”%˚!.3r >ò>òIÞ -”9ı  .  Œ 3v33%. ˜ˆ5C%˚ ¢ .R4n7~$ê G¶,”%˚!.  . G¶,”%˚!.3r %˚%:.  Œ 3v339Œ  7ı':9Û%˚%:. .R4n7~$ê  Œ3v338&˚â  /K˜â9ı &"*z %:.  /K˜â9Ê 1'G9Ê)Ò˜6 .        /K˜â9Ê 1'G9Î ˚Ù9ı 8ê.  Œ3v339Œ /K˜â9Ê 1'G9Î ˚Ù9ı  .U9ı 4n˚V .  >fl7ı ;R3ª3r . 3v-™(present)*² 5 %:. .R4n7~8&˚â 5 .  N*F%Æ  9Ò9Ê . .R4n7~$ê  7ž ˚V&ˆ ,  %ª ˚V˜6 %:. .R4n7~˚V  :á7ıG®˚â &r ˚Ù9Û%˚%:.          Min-su is an office worker. Every morning, he gets up at 7:00 am. For breakfast, he has a cup of coffee with milk in it and bread. And he goes to his company on foot. It takes twenty minutes from his house to the company. He arrives at his company at 8:00 am and works till 6:00 pm. He sometimes meet his customers or works at his desk. When he gets tried, he quietly says to himself, If I have 1 million dollars, I will quit and take a rest at home.  139         Yesterday, something surprising happened to Min-su.         Yesterday morning, Min-su found a wallet that was dropped on his way to the company. There were 30000 won in cash and one lottery ticket. He brought the wallet with the ticket and checked the ticket numbers on the computer. He could not believe his eyes. Surprisingly, that was the winning ticket last week.                Min-su thought about having the ticket. He also thought about returning the ticket to the owner. After a while, he decided to return the wallet and the ticket to the owner. He caught a taxi and went to the owners house.          He pressed the bell on the door of the house and an old lady answered the door slowly. Is Mr. Kim here Min-su asked the old lady.  She answered, Mr. Kim is my son After a while, the owner appeared. Min-su returned the wallet and the ticket. After that, he also told the owner the good news about the winning ticket.         I did not know that I won a lottery. Mr. Kim could never believe that he won the lottery. I thank you for being honest and returning my wallet. As a reward, Id like to give you 5 hundred million won. Now Min-su has 5 hundred million won. However something surprising happened. Even though he does not have to go to the company, he feels like hed love to now. It is not surprising that Min-su has 5 hundred million dollars. What is really surprising is that he comes to like his company.                         140 B-5. Background Questionnaire                             Linguistic Background Questionnaire  1. Basic bio-data questions  -Age:  ____________ (year of birth: _________)  Gender _______  -Are you a student now? Yes ____(University:   ) No ____(Graduated school_________)  - Class standing: Major _______________ Minor ______________ -Specify any Chinese or Japanese courses you have taken so far __________________________  -Place of birth _______________ -If you were not born in the US, at what age did you come to the US: __________ years old -Check all the family members who you have lived with for over 6 months of your life that are native speakers of Korean. -Grandmother, grandfather, mother, father, brother(s), sister(s), relatives, spouse (significant other) -Biological Fathers ethnic background _____, Biological Mothers ethnic background ________ -Specify if applicable.        -Foster Fathers ethnic background _____, Foster Mothers ethnic background ________       -The age when they adopted you from Korean to the US __________  -When you were a child, what was your first language (First language is the language your parents communicate with you before your age of 5?  -English______, Korean______, English & Korean______,  -Any other languages _____________________  2. Input-interaction-output experience Generally speaking, when growing up I have received exposure of Korean _______ 0% 10 20 30 40 50 60 70 80 90 100% never hardly seldom rarely occasionally sometimes often frequently normally usually always  1. Age-based self-report (Use the following scale) 0% 10 20 30 40 50 60 70 80 90 100% never hardly seldom rarely occasionally sometimes often frequently normally usually always    -50%, from the source during the age period.  INPUT Ages (references)  Sources Put the percentage of the scale above 0-5  (birth-before Parents Siblings and grandparents  F______/M_____ S_____/G_____  141 elementary school) Cousins and relatives  Entertainment materials, e.g., CDs/tapes for kids  Korean books, magazines, newspapers  Korean comic books  Korean dramas Korean movies Korean songs  Korean friends in community  Korean friends on the web  Korean friends at school  Saturday or Sunday school  School extra curriculum activity Any other source (e.g. caretaker) __________ C _____/R_____ ____________ ____________ ____________ ____________ ____________ ____________ ____________ ____________ ____________ ____________ ____________ ____________ 6-8 (elementary school-lower grades) Parents Siblings and grandparents  Cousins and relatives  Entertainment materials, e.g., CDs/tapes for kids  Korean books, magazines, newspapers  Korean comic books  Korean dramas Korean movies Korean songs  Korean friends in community  Korean friends on the web  Korean friends at school  Saturday or Sunday school  School extra curriculum activity Any other source (e.g. caretaker) __________ F______/M_____ S_____/G_____ C _____/R_____ ____________ ____________ ____________ ____________ ____________ ____________ ____________ ____________ ____________ ____________ ____________ ____________ 9-10 (elementary school-higher grades) Parents Siblings and grandparents  Cousins and relatives  Entertainment materials, e.g., CDs/tapes for kids  Korean books, magazines, newspapers  Korean comic books  Korean dramas Korean movies Korean songs  Korean friends in community  Korean friends on the web  Korean friends at school  Saturday or Sunday school  School extra curriculum activity Any other source (e.g. caretaker) __________ F______/M_____ S_____/G_____ C _____/R_____ ____________ ____________ ____________ ____________ ____________ ____________ ____________ ____________ ____________ ____________ ____________ ____________ 11-13 (junior high school) Parents Siblings and grandparents  Cousins and relatives  Entertainment materials, e.g., CDs/tapes for kids  Korean books, magazines, newspapers  Korean comic books  F______/M_____ S_____/G_____ C _____/R_____ ____________ ____________ ____________  142 Korean dramas Korean movies Korean songs  Korean friends in community  Korean friends on the web  Korean friends at school  Saturday or Sunday school  School extra curriculum activity Any other source (e.g. caretaker) __________ ____________ ____________ ____________ ____________ ____________ ____________ ____________ ____________ ____________ 14-17 (high school) Parents Siblings and grandparents  Cousins and relatives  Entertainment materials, e.g., CDs/tapes for kids  Korean books, magazines, newspapers  Korean comic books  Korean dramas Korean movies Korean songs  Korean friends in community  Korean friends on the web  Korean friends at school  Saturday or Sunday school  School extra curriculum activity Any other source (e.g. caretaker) __________ F______/M_____ S_____/G_____ C _____/R_____ ____________ ____________ ____________ ____________ ____________ ____________ ____________ ____________ ____________ ____________ ____________ ____________ 18-22 (College) Parents Siblings and grandparents  Cousins and relatives  Entertainment materials, e.g., CDs/tapes for kids  Korean books, magazines, newspapers  Korean comic books  Korean dramas Korean movies Korean songs  Korean friends in community  Korean friends on the web  Korean friends at school  Saturday or Sunday school  School extra curriculum activity Regular class Any other source (significant other) ____ F______/M_____ S_____/G_____ C _____/R_____ ____________ ____________ ____________ ____________ ____________ ____________ ____________ ____________ ____________ ____________ ____________ ____________ ____________ 23-30 (After graduation, workplace)  Parents Siblings and grandparents  Cousins and relatives  Entertainment materials, e.g., CDs/tapes for kids  Korean books, magazines, newspapers  Korean comic books  Korean dramas Korean movies Korean songs  F______/M_____ S_____/G_____ C _____/R_____ ____________ ____________ ____________ ____________ ____________ ____________  143 Korean friends in community  Korean friends on the web  Korean friends at school  Saturday or Sunday school  School extra curriculum activity Regular class in college or culture center Any other source (significant other) ____ ____________ ____________ ____________ ____________ ____________ ____________ ____________  OUTPUT Ages (references)  Sources Put the percentage of the scale above 0-5  (birth-before elementary school)   Parents Siblings and/or grandparents  Cousins and relatives  Korean entertainment songs for kids  Korean songs  Korean friends in the community Korean friends on the web  Korean friends at school  Saturday school or Sunday school  School extra curriculum activity Writing Korean alphabet, email, journal, Any other source (e.g. caretaker) _________ F______/M_____ S_____/G_____ C _____/R_____ ____________ ____________ ____________ ____________ ____________ ____________ ____________ ____________ ____________ 6-8 (elementary school-lower grades)  Parents Siblings and/or grandparents  Cousins and relatives  Korean entertainment songs for kids  Korean songs  Korean friends in the community Korean friends on the web  Korean friends at school  Saturday school or Sunday school  School extra curriculum activity Writing Korean alphabet, email, journal, Any other source (e.g. caretaker) _________ F______/M_____ S_____/G_____ C _____/R_____ ____________ ____________ ____________ ____________ ____________ ____________ ____________ ____________ ____________ 9-10 (elementary school-higher grades)   Parents Siblings and/or grandparents  Cousins and relatives  Korean entertainment songs for kids  Korean songs  Korean friends in the community Korean friends on the web  Korean friends at school  Saturday school or Sunday school  School extra curriculum activity Writing Korean alphabet, email, journal, Any other source (e.g. caretaker) _________ F______/M_____ S_____/G_____ C _____/R_____ ____________ ____________ ____________ ____________ ____________ ____________ ____________ ____________ ____________ 11-13 (junior high school) Parents Siblings and/or grandparents  F______/M_____ S_____/G_____  144   Cousins and relatives  Korean entertainment songs for kids  Korean songs  Korean friends in the community Korean friends on the web  Korean friends at school  Saturday school or Sunday school  School extra curriculum activity Writing Korean alphabet, email, journal, Any other source (e.g. caretaker) _________ C _____/R_____ ____________ ____________ ____________ ____________ ____________ ____________ ____________ ____________ ____________ 14-17 (high school)   Parents Siblings and/or grandparents  Cousins and relatives  Korean entertainment songs for kids  Korean songs  Korean friends in the community Korean friends on the web  Korean friends at school  Saturday school or Sunday school  School extra curriculum activity Writing Korean alphabet, email, journal, Any other source (e.g. caretaker) _________ F______/M_____ S_____/G_____ C _____/R_____ ____________ ____________ ____________ ____________ ____________ ____________ ____________ ____________ ____________ 18-22 (College)   Parents Siblings and/or grandparents  Cousins and relatives  Korean entertainment songs for kids  Korean songs  Korean friends in the community Korean friends on the web  Korean friends at school  Saturday school or Sunday school  School extra curriculum activity Writing Korean alphabet, email, journal, Regular class at college Any other source (significant other) ______ F______/M_____ S_____/G_____ C _____/R_____ ____________ ____________ ____________ ____________ ____________ ____________ ____________ ____________ ____________ ____________ 23-30 (and after) (After college graduation, workplace)    Parents Siblings and/or grandparents  Cousins and relatives  Korean entertainment songs for kids  Korean songs  Korean friends in the community Korean friends on the web  Korean friends at school  Saturday school or Sunday school  School extra curriculum activity Writing Korean alphabet, email, journal, Regular class at college or culture center Any other source (significant other) ______ F______/M_____ S_____/G_____ C _____/R_____ ____________ ____________ ____________ ____________ ____________ ____________ ____________ ____________ ____________ ____________   145 3. Korean instruction  3-1. Korean class level or course name you have taken at the latest: __________________ (For example, KR 101, KR 102, KR 201.KR 402, Business Korean)   -I have studied Korean as heritage language / foreign language For ______ years ___ months since the age of _________ (the year of: _____________) -List all institutes (For example, Michigan State University) ______________________________________________________________________ ______________________________________________________________________ ______________________________________________________________________  -List all courses you took previously (For example: Korean 101) ______________________________________________________________________ ______________________________________________________________________ ______________________________________________________________________  -List all books you learn with (For example: Integrated Korean Beginning 1) ______________________________________________________________________ ______________________________________________________________________ ______________________________________________________________________  3-2. How have you studied Korean grammar? (Tick as many as appropriate) ___ I ask for help from parents, ____ help from teachers, _____ help from friends, ____ study with grammar books, ____ from natural communication in speaking, ____from natural communication in listening, ____ from natural communication in writing, ____ from natural communication in reading, _____ from text books in class, ______ from lectures,  Specify if there is any other way ___________________________________________  3-3. How have your teachers of Korean taught Korean grammar? (Tick as many as appropriate) _____ lecture or explicit explanation, ____ class activities, _____ authentic materials such as newspapers and movies, _____ homework, ______ grammar-focused test, quizzes, and examinations,  Specify if there is any other way ____________________________________________ ______________________________________________________________________  3-4. How would you describe your Korean teachers teaching style in your Korean class? (Tick every style that is relevant) ______Meaning-oriented, ______communicative-activity oriented, ______grammar-oriented, ______teacher-centered instruction, ______student-centered instruction, ______textbook-oriented instruction, ______listening-speaking focused class, ______writing-reading focused class, Any other styles you want to specify _________________________________________  3-5. How are the tests, quizzes, and examinations for evaluating Korean grammar in class?  ____ written tests, quizzes, and examinations have items or sections for grammar, ____  146 grammar is tested through oral tests, ____ grammar is tested through written tests, ____ grammar is not tested explicitly, _____ grammar is usually emphasized in evaluations,  Specify if there is any other point _________________________________________ ______________________________________________________________________ ______________________________________________________________________  3-6. Have you acquired Korean at other formal non-college institution in the US? (For example, Saturday school, Sunday school, extra curriculum activities in elementary, junior high, or high schools) -Yes____, No____ -If yes, list institutions ____________________________________________________ -At what age ______ for how long? _____ year(s) _____ month(s) -What book(s) did you learn with? __________________________________________ -Was the instruction style different from the instruction in college? _______________  If yes, how would you describe the difference? ________________________  3-7. Reading and Writing Reading - When did you learn how to read Korean? At the age of ______ in the year of ______ - How did you learn how to read Korean? Who taught you to read Korean? ______________________________________________________ Writing - When did you learn how to write Korean? At the age of _____ in the year of ________ - How did you learn how to write Korean? Who taught you to write Korean? _______________________________________________________    4. Studying abroad in Korea & Living in Korea 4-1. Have you ever visited Korea? -Yes____, No____ -If yes, at what age(s) ___________ for how long? _______ year(s) ______ month(s) _______ week(s) _______day(s) -Reason(s) for being in Korea ______________________________________________  If you visited Korean multiple times, please keep answering. -2nd visit  At what age(s) ___________ for how long? _______ year(s) ______ month(s) _______  week(s) _______day(s)  Reason(s) for being in Korea ______________________________________________  -3rd visit  At what age(s) ___________ for how long? _______ year(s) ______ month(s) _______  week(s) _______day(s)  Reason(s) for being in Korea ______________________________________________  -4th visit  At what age(s) ___________ for how long? _______ year(s) ______ month(s) _______  week(s) _______day(s)  Reason(s) for being in Korea ______________________________________________  147  4-2. Have you ever studied Korean in Korea? -Yes____, No____ -If yes, at what age(s) ___________ for how long? _______ year(s) ______ month(s) _______ week(s) _______day(s)  -If yes, list the school level or program (for example, K-12, college, summer school, study abroad, etc) ______________________________________________________________________ ______________________________________________________________________ ______________________________________________________________________  5. Tasks in this study Complete the following based on the difficulties you felt during this study. Names of the tasks: Imitation, Narrative, Aural task, Written task, Multiple Meta, Example Meta,  (Most difficult)__________>___________>__________>__________>__________>_________(Easiest)    If you have any ideas, please contact me heoyeon7@gmail.com! I wish you the best!                             148 B-6. Proficiency Test (Reprinted and adapted from Lee-Ellis, 2009)   Korean Proficiency Test  This is a test of how well you comprehend and produce Korean. You will read four short passages. In each, parts of some words are missing.  Study each passage and write in the missing letters. Each line represents one syllable.   -No negative point will be deducted for a wrong answer.  -Spelling will not be assessed as long as the words are identifiable.   Example: 7ž#«___ ___ ___. :r 9Ê___ ___   Œ >ö4n9Û%˚%:.  Your job is to complete the test as:            7ž#«G® 3” 8ê.   :r 9Ê+Ú 9Œ   Œ >ö4n9Û%˚%:.  -Notice that partial points are available. If you know only part of the  missing parts, fill in the part that you know instead of skipping the  :r 9Ê __ 9Œ  Œ >ö4n9Û%˚%:.credit.)  You will have 20 minutes. This test is designed for all ranges of proficiency (i.e., from beginning to near-native), so it will seem challenging to many of you. However, please do your best until the end, and try to work on all four passages if you have time.    Passage 1 7ž#«G®3”8ê. :r 9Ê+Ú9Œ  Œ >ö4n9Û%˚%:. :V$ê %VG¯___ ___ %:%/%˚%:. 7ı@¾8& ___ ___ G¯˜¦ ___ ___ ˚g%˚%:. ___ ___ ___ 9ı G¿___ ___.  G² %:___ ___ 7ı@¾9ı . 7ı@¾9Œ  ƒ4o___ 53%O8&3r . :V$ê %VG¯___ ___ ___  .ƒ___ ___ ___.  4n___ ___ ,:9Ò 8z___ 10528& 529ç___ ___ ___.  7F___ ___ ,&G® ƒ˚V ___ ___ ___ ___.  '9___ ___ 9Ó ƒ$ê 4Ó___ ___ ___.  .ƒ___ ___ ˚Ù9Ê >” ___ ___ ___ ___. ;R,&8&$ê @²___ ___ ___ ˚o9Ê  O___ ___ ___ 8WHª+Ò /[%˚%:. 8WHª+Ò ___ Iˆ8& G²˜Ã 53%O8&3r :V___ ___ . G²˜Ã 53___ ___  O9û .j___ 8\8& . /Þ˜6___ ___ .  Œ@®>b˚r$ê %˚%:.   149 Passage 2 8‡ 8B+Ú8&$ê ˚V:Ç':˜R G¾!. :r;R&ˆ8& 8BGß9ı ˚V*z˜6 GÊ8ê. :r;R&ˆ$ê G².n___ ___  9Þ___ 3‡9Ê8&8ê. G²˜Ã9® G®8Œ9Ê)Ò /Þ___ ___. :r;R&ˆ$ê 9æ___ ___ 7ı+Ú%:9"3r 56H™___ ___ 9û3â*² 9Î___ ___  ,$___ ___. 8z$î9Œ 8BGß___ ___ :ZHª+Ò ˚Î___  :r;R&ˆ ___ ___ 8«/K 0ıGß___ ___ ___ #z 9û 8^___ ___ ___ ___. 8BGß___ ___ ___ H”Cª&ˆ 3â˚r___  H”Cª9Œ  7ž :k___ ___ ___. 9ÎCƒ#“9™*² :k___ ___ %ª >fl7ı /J___  H”Cª9Ê :á___ 7¢7ı /J___ ___ GÊ8ê. 8ê___ ___ 9ÎCƒ#“9Ê 9Þ___ ___ H”Cª 2& 7ı___ ___ 9v,ÛG² ˜V___ ,Û3â8Œ ,1___ ___ 53%O&ˆ >fl7ı /R ___  >” .  Passage 3 7ž#«G®3”8ê.  7´8& 9Z@®G²  .⁄Hª:f9Û%˚%:. :VI .⁄Hª:f___ ___ ___  ,4___  "ò___ :rFÞ9ı 3”9Ò___ ___ %˚%:.  8B___ ___ 9ZG² 8B3⁄/K Aª#^___ ___ 8B3⁄ :k___˜R  3ã___ ___ 50 GZ*² 3”9ÒG®˜6 9Þ___,  @‰ 7ı&/___ Aª#^8&3r&ˆ Aª___, , 9û___ 'G9® ˚þ___  ˚W 30GZ*²  G¶___ ___ ˚V˚ÿ8& Ef___ ___ ___ . @¶ ___ 8&3r$ê <' 7ž9ı 'ƒ___ ___ ___ GÊ ;Z :Z___ IÞCƒ8Œ ˚V4ú "ò___ 'G %:7ç___  ˚V___ :rFÞ9ı ___ ___ ___ . :V___  .⁄Hª:f˜R G¾___  "î ƒ ;V___ ___ 529çG®3”8ê. ˜6˚s 8B*B/Ú9® ,$9Œ 3⁄9& %:. .  Passage 4 &ˆ529® ˚V9û BÆ -”:r:f9Ê)Ò,Ê -¦%˚-¦%˚GÊ&ˆ  -”:r˚V :r9Ò BÂ%:. &ˆ*²8&3r$ê ˜¦___  9Î___ ___  52___ ___ .  ?ò___ ___ 52˚Z8&$ê ___ ___ >~*˜9Ê 9Ò:r___  &ˆ___ ___ 7ı;R /K___ ___ ___. ˚â%:˚V $Ê8ë ˚o9Œ %V&ˆ___ ___ ;R>~"ò9Œ ,:___ 5B˚WG² 4n;V___ ___. 9æ&/___ ___ :f:f ,$7ı___ ___ .n,Ê ;R___ ˜K˚Z9Œ :r___ ___ ___ 9Þ ƒ '¢___ ___ ;R>~"ò9Ê 33___ ___. ;R>~9û9Ê /Ö:ÇG®,Ê ___ ___ ;RC3˚V ˜>___9Ê"î  ;R>~+Ò G®___  ,$%:. 9Ê*]˚â /Þ___9™*² ;R___ ___ >~*˜9Œ (f %:52  H™___ ___ 9&9Î9Ê &n___ %ª 5B___ ___  9ı 9Ò9™Bþ%:. 'ƒ)Ò3r  -”:r+Ò  ƒ 9ZGÊ3r$ê 9æ˚V8ÿ/J%:$ê .Ú4ú"î  ,$9Ê 9Ê8ÿGÊ7Ò G¶ ˚Ù9Ê%:.         Passage 1 Hello, my name is Cheol-soo Kim. I go to college. After I get up in the morning, I go to the gym for exercise. After that, I have breakfast at a dormitory cafeteria. I  150 learn Korean at school. The class starts at 10 am. Writing and speaking in Korean are difficult, but listening and reading are easy. It is fun to learn Korean. During the weekends, I watch movies with my friends. After that, I have dinner in a Korean restaurant, which is next to the movie theater. Bulgogi is delicious. Kimchi stew is hot.   Passage 2 I will go to Je-joo Island with my family this summer. Je-joo Island is located in the south of Korea. It is called Hawaii of Korea. It is very popular as a spot for newlyweds due to its beautiful landscapes. Today I will call the tour agency to get four round tickets from Seoul to Jejoo Island. The agency also recommenended a hotel, but I have not decided yet. I will get more information on the web. The internet makes it very easy to find tourist attractions and restaurants as well as hotels.    Passage 3 Hello. This announcement is from Seoul Department Store located in front of Seoul Station. Approaching winter, we have winter clothes and heating appliances on sale. In the section for career women, suits and underwears are 50% off. In the section for children on the third floor, winter merchandise such as coats, scarfs, and gloves are 30% off. On the seventh floor, heating appliances and gas stoves for homes are on special sale. Please prepare for this winter with us, Seoul Department Store. Thank you for your attention.     Passage 4 The most serious problem in a city is traffic, most of all. People waste energy and time on the roads due to traffic jams. The roads become crowded during rush hours due to loads of vehicles. In addition, parking issues in a metropolitan city like New York are very serious. They are from increasing numbers of cars while the parking spaces are limited. In this case, people park their cars on the alleys in residential areas and even on the roads. These illegally parked cars aggravate the confusion from traffic jams, which causes more severe traffic issues. Therefore, people should use buses and the subway more in order to solve these problems.     151 APPENDIX C. Item total statistics (Pilot study I)  Table 29. Item total statistics for each test focusing on cronbachs alphas if item deleted (Pilot study I)  Cronbach's Alpha if Item Deleted Elicited Imitation Test Timed GJT Untimed GJT Metalinguistic Test Item 1 .987 .963 .952 .841 Item 2 .987 .963 .952 .843 Item 3 .987 .963 .952 .841 Item 4 .987 .964 .952 .821 Item 5 .987 .963 .951 .830 Item 6 .988 .963 .951 .820 Item 7 .987 .965 .954 .842 Item 8 .987 .964 .952 .841 Item 9 .987 .964 .952 .852 Item 10 .987 .963 .952 .832 Item 11 .988 .963 .951 .831 Item 12 .988 .963 .951 .850 Item 13 .987 .964 .952 .833 Item 14 .987 .964 .952 .827 Item 15 .987 .964 .951 .837 Item 16 .987 .963 .951 .815 Item 17 .987 .963 .952 .835 Item 18 .987 .965 .952  Item 19 .987 .964 .951  Item 20 .987 .963 .951  Item 21 .987 .963 .952  Item 22 .987 .964 .952  Item 23 .988 .963 .952  Item 24 .987 .963 .952  Item 25 .987 .963 .952  Item 26 .987 .964 .951  Item 27 .987 .963 .951  Item 28 .987 .964 .952  Item 29 .987 .964 .952   Item 30 .987 .964 .952  Item 31 .987 .963 .952  Item 32 .987 .964 .951  Item 33 .987 .963 .952  Item 34 .987 .963 .952  Item 35  .964 .951   152 Table 29. (contd) Item 36  .963 .950  Item 37  .964 .951  Item 38  .964 .952  Item 39  .963 .952  Item 40   .963 .951  Item 41  .964 .953  Item 42  .964 .952  Item 43  .965 .954  Item 44  .963 .951  Item 45  .963 .951  Item 46  .964 .951  Item 47  .963 .951  Item 48  .963 .951  Item 49  .963 .952  Item 50  .964 .952  Item 51  .963 .951  Item 52  .964 .951  Item 53  .963 .951  Item 54  .963 .951  Item 55  .963 .951  Item 56  .963 .952  Item 57  .963 .951  Item 58  .963 .952  Item 59  .963 .952  Item 60  .964 .952  Item 61  .964 .952  Item 62  .964 .951  Item 63  .964 .952  Item 64  .963 .951  Item 65  .963 .951  Item 66  .964 .951  Item 67  .964 .952  Item 68  .965 .952            153 Appendix D. Item total statistics (Pilot study II)  Table 30. Item total statistics for each test focusing on cronbachs alphas if item deleted (Pilot study II)  Cronbach's Alpha if Item Deleted Elicited Imitation Test Aural GJT Written GJT Metalinguistic Test1 Metalinguistic Test2 Item 1 .877 - - .710 .864 Item 2 .875 .869 .853 .710 .872 Item 3 .883 - .847 .672 .858 Item 4 .877 .872 .841 .664 .910 Item 5 .880 .867 .838 .672 .858 Item 6 .880 .873 .844 .671 - Item 7 .879 - - .690 - Item 8 .885 - .842 .738 - Item 9 .881 .868 .844 - .858 Item 10 .882 .867 .835 .697 .864 Item 11 .884 - - .668 - Item 12 .885 .870 .847 .674 - Item 13 .884 .866 .836 .802 - Item 14 - .877 .840 .697 - Item 15 .875 - - .674 - Item 16 .879 .873 .851 .643 .878 Item 17 .883 .865 .853 .703 - Item 18 .883 .867 .840   Item 19 .876 .875 .847   Item 20 .887 - .845   Item 21 .874 .869 .838   Item 22 .879 .868 .847   Item 23 .882 .872 .844   Item 24 - - -   Item 25 .888 .865 .843   Item 26 .879 .870 .850   Item 27 .880 .873 .843   Item 28 .883 .873 .851   Item 29 .873 .867 .839    Item 30 .878 .865 .845   Item 31 .870 .869 .848   Item 32 - .872 .840   Item 33 .880 .870 .849   Item 34 .885 .873 .854   Item 35  .874 -    154  Table 30. (contd) Item 36  - .847   Item 37  .866 .836   Item 38  .876 .850   Item 39  - -   Item 40   .872 -   Item 41  .869 .843   Item 42  .869 .844   Item 43  .870 .846   Item 44  - .847   Item 45  .865 .847   Item 46  .869 .848   Item 47  .879 -   Item 48  .869 -   Item 49  .865 .847   Item 50  .865 .851   Item 51  - .847   Item 52  - -   Item 53  .865 .841   Item 54  .869 .855   Item 55  .870 .847   Item 56  .873 .856   Item 57  .872 .849   Item 58  .864 .843   Item 59  .880 .860   Item 60  .873 .845   Item 61  .872 .840   Item 62  .867 .837   Item 63  .878 .840   Item 64  .871 .848   Item 65  .864 .846   Item 66  .869 .840   Item 67  .868 .843   Item 68  .869 .843    Note. - means that the variable has zero variance, which means that the 10 participants answered correctly or incorrectly. However, this does not guarantee the same results when this study includes L2 learners.      155 APPENDIX E. Summary of all CFAs Table 31. Summary of the important indices from all CFAs  Observable  variable (O.V.)  Factor        Models INDEX   NFI ( .90) RMSEA (.00RMSEA  .05)  (p  .05) df CMIN/DF (0MIN/DF2) AIC (smaller than comparison model) 4 OVs  EIT,   AuralGJT,  WrittenGJT,  Metalinguistic 2 Figure 1. 2F-4-Default. 4 OVs Default Implicit vs. explicit .998 .000 0.410 (p=.522) 1 .410           Heywood case 26.410 Figure 2. 2F-4-Rival 1. 4 OVs Ungrammatical items in written GJT .998 .000 0.483 (p=.487) 1 .483 26.483 2F-4-Rival-2. Grammatical vs. ungrammatical items  Unidentifiable -    1 1F-4-Default. 4 OVs Default Explicit .914 .302 20.573 (p=.000) 2 10.286 44.573 Figure 3. 1F-4-Rival-1. 1 Covaried error terms btw imitation and auralGJT .998 .000 0.410 (p=.522) 1 .410 Heywood case 26.410 Figure 4. 1F-4-Rival-2. 1 Covaried error terms btw writtenGJT and metalinguistic test .998 .000 0.410 (p=.522) 1 .410 26.410 5 OVs   EIT,  AuralGJT  Narrative test  WrittenGJT  Metalinguistic 2 2F-5-Narrative-Default. Implicit vs. explicit  .840 .400 51.983 (p=.000) 3 17.328 Heywood case 85.983 2F-5-Narrative-Rival-1. Ungrammatical items in writtenGJT .870 .309 32.302 (p=.000) 3 16.767 66.302 2F-5-Narrative-Rival-2 2 Covaried error terms btw auralGJT and narrative, imitation and narrative .985 .095 3.846 (p=.146) 2 1.923  39.846 2F-5-Narrative-Rival-3 Grammatical vs. ungrammatical items  Unidentifiable     1 1F-5-Narrative-Default. 5 OVs Default Explicit .756 .381 78.914 (p=.000) 5 15.783  1F-5-Narrative-Rival-1. 2 Covaried error terms btw  metalinguistic test and writtenGJT, imitation test and narrative  .969 .152 10.095 (p=.018) 3 3.365  1F-5-Narrative-Rival-2. 3 Covaried error terms btw metalinguistic test and writtenGJT, imitation test and narrative, auralGJT and narrative  .991 .065 2.849 (p=241) 2 1.414 39.630   156 Table 31. (contd) 5 OVs   EIT  AuralGJT  WrittenGJT  Metalinguistic  C-test 2 2F-5-C-test-Default 5 CVs Default: Implicit vs. explicit .927 .269 25.201 (p=.000) 3 8.400 59.201 Figure 5. 2F-5-C-test-Rival-1.   2 Covaried error terms btw writtenGJT and metalinguistic test, metalinguistic test and C-test .998 .000 0.656 (p=.418) 1 .656 38.656 Figure 6. 2F-5-C-test-Rival-2. Ungrammatical items in writtenGJT .972 .096 7.749 (p=.101) 4 1.937  41.749 Figure 7. 2F-5-C-test-Rival-3.   2 Covaried error terms btw writtenGJT_ungra and metalinguistic test, metalinguistic test and C-test .998 .000 0.677 (p=.411) 1 .677 38.677 2F-5-C-test-Rival-4 Grammatical vs. ungrammatical items  Unidentifiable     2F-5-C-test-Rival-5 Orality+grammatical vs. literacy+ungrammatical  .964 .109 6.645 (p=.084) 3 2.215 40.645 1 1F-5-C-test-Default  .885 .261  39.761 (p=.000) 5 7.952 69.761 1F-5-C-test-Rival-1. 1 Covaried error term btw Metalinguistic test and C-test .940 .202 20.640 (p=.000) 4 5.160 52.640 1F-5-C-test-Rival-2. 2 Covaried error terms btw Metalinguistic test and C-test, imitation and auralGJT .981 .106 6.451 (p=.092) 3 2.150 40.451 6 OVs  EIT  Narrative test  AuralGJT  WrittenGJT  Metalinguistic  C-test 2 2F-6-Default. 6 OVs Default: Implicit vs. explicit .817 .295 79.080 (p=.000) 8 9.885 117.080 2F-6-Rival-1. 3 Covaried error terms btw auralGJT and narrative, narrative and imitation, metalinguistic test and C-test .980 .084 8.619 (p=.125) 5 1.724 52.619 2F-6-Rival-2. Ungrammatical items in writtenGJT .873 .217 46.283 (p=.000) 8 5.783 84.283 Figure 8. 2F-6-Rival-3. Ungrammatical items in writtenGJT.  3 Covaried error terms btw auraGJT and narrative, narrative and imitation, metalinguistic and C-test .986 .019 5.191 (p=.393) 5 1.038 49.191 2F-6-Rival-4. Grammatical vs. ungrammatical items   Unidentifiable    Figure 9. 2F-6-Rival-5. Orality+grammatical vs. Literacy+ungrammatical  .953 .064 11.339 (p=.183) 8 1.417 49.339    157 Table 31. (contd)   Figure 10. 2F-6-Rival-6. Orality+grammatical vs. Literacy+ungrammatical. 1 covaried error terms btw metalinguistic test and C-test .966 .043 8.323 (p=.305) 7 1.189 48.323 1 F-6-Default. 6 OVs Default .752 .328 107.478 (p=.000) 9 11.942 143.478 1F-6-Rival 1. 6 OVs 2 Covaried error terms btw metalinguistic test and C-test, imitation test and narrative  .934 .174 28.546 (p=.000) 7 4.078 68.546 1F-6-Rival 2. 6 OVs 4 Covaried error terms btw metalinguistic test and C-test, metalinguistic test and writtenGJT, imitation test and narrative, auralGJT and narrative  .972 .119 12.228 (p=.032) 5 2.446 56.228        158             REFERENCES              159 REFERENCES  Abrahamsson, N. (2012). Age of onset and nativelike L2 ultimate attainment of morphosyntactic and phonetic intuition. Studies in Second Language Acquisition, 34, 187-214.  Abrahamsson, N., & Hyltenstam, N. (2008). The robustness of aptitude effects in near-native second language acquisition. Studies in Second Language Acquisition, 30, 481-509. American Counsil on the Teaching of Foreign Languages. (2015). ACTFL proficiency guidelines. Retrieved from http://www.actfl.org/publications/guidelines-and-manuals/actfl-proficiency-guidelines-2012 Ahn, J. (2010). The use of Korean honorifics for the heritage Koreans in the US. Korean English Linguistics and Literature, 36(3), 235-254. Anderson, J. (1983). The architecture of cognition. Cambridge, MA: Harvard University Press.  Allen, R., and Reber, A. (1980). Very long-term memory for tacit knowledge. Cognition, 8, 175-185. Ardila, A. (2003). Language representation and working memory with bilinguals. Journal of Communication Disorders, 36, 233-240.  Au, T., Kinghtly, L., Jun, S., & Oh, J. (2002). Overhearing a language during childhood. Psychological Science, 13, 238-243. Baddleley, A. (2003). Working memory and language: An overview. Journal of Communication Disorders, 36, 189-208.  Bialystok, E. (1979). Explicit and implicit judgments of L2 grammaticality. Language Learning, 29, 81-103. Bialystok, E. (1982). On the relationship between knowing and using linguistic forms. Applied Linguistics, 3(3), 181-206.  Bialystok, E. (1994). Analysis and control in the development of second language proficiency. Studies in Second Language Acquisition, 16, 157-168.  Birdsong, D. (1989). Metalinguistic performance and interlinguistic competence. Berlin: Springer. Birdsong, D. (2005). Interpreting age effects in second language acquisition. In J. Kroll & A. De Groot (Eds.), Handbook of bilingualism: Psycholinguistic approaches (pp. 109-127). Oxford: Oxford University Press. Bley-Vroman, R. (1989). What is the logical problem of foreign language learning? In S. Gass & J. Schachter (Eds.), Linguistic perspectives on second language  160 acquisition (pp. 41-68). Cambridge: Cambridge University Press.   Bley-Vroman, R. (1990). The logical problem of foreign language learning. Linguistic Analysis, 20, 3-49. Bley-Vroman (2009). The evolving context of the Fundamental Difference Hypothesis. Studies in Second Language Acquisition, 31, 175-198. Bowles, M. (2011). Measuring implicit and explicit linguistic knowledge: What can heritage language learners contribute? Studies in Second Language Acquisition, 33, 241-271. Bradlow, A., & Pisoni, D. (1999). Recognition of spoken words by native and non-native listeners: Talker-, listener-, and item-related factors. The Journal of the Acoustical Society of America, 106(4), 2074-2085. Byrne, B. (2010). Structural equation modeling with AMOS. New York: Routledge. Cho, M. (1982). G²˜Ã7ı&/9®  Hã'3 8F˜Â:  [A study on Korean acquisition of Korean children: Strategy model]. Seoul: Seoul National University Press. Cho, Y., Lee, H., Schulz, C., Sohn, H., & Sohn, S. (2010). Integrated Korean: Beginning 1, 2. Honolulu: University of Hawai Cole, R., & Pickering, S. (2015). Phonological and visual similarity effects in Chinese and English language users: Implications for the use of cognitive resources in short-term memory. Bilinguialism: Language and Cognition, 13(4), 499-512. Crowell, S. (2004). The neurobiology of declarative memory. In J. Schumann, S. Crowell, N. Jones, N. Lee, & S. Schuchert (Eds.), The neurobiology of learning: Perspectives from second language acquisition (pp.75-109). Mahwah, NJ: Lawrence Erlbaum.  Danks, J., & End, L. (1987). Processing strategies for reading and listening. In R. Horowitz, and Samuels (Eds), Comprehending oral and written language. (pp. 271-294). UK: Emerald Group Publishing Limited.  DeKeyser, R. (1997). Beyond explicit rule learning: Automatizing second language morphosyntax. Studies in Second Language Acquisition, 19, 195-221.  DeKeyser, R. (2000). The robustness of critical period effects in second language acquisition. Studies in Second Language Acquisition, 22, 499-533.  DeKeyser, R. (2003). Implicit and explicit learning. In C. Doughty & M. Long (Eds.), The handbook of second language acquisition (pp. 313-348). Malden MA: Blackwell. DeKeyser, R. (2007). Skill acquisition theory. In B. VanPatten & J. Williams (Eds.), Theories in second language acquisition: An introduction (pp. 97-113). Mahwah, NJ: Lawrence Erlbaum.   161 DeKeyser, R. (2009). Cognitive-psychological processes in second language learning. In M. H. Long & C. J. Doughty (Eds.), The handbook of language teaching (pp. 119-138). Chichester: Wiley-Blackwell. Douglas, D. (2001). Performance consistency in second language acquisition and language testing research: A conceptual gap. Second Language Research, 17(4), 4442-456. DÖrnyei, Z. (2009). The psychology of second language acquisition. Oxford: Oxford University Press. Ellis, N. (1994). Implicit and explicit language learning: An overview. In N. Ellis (Ed.), Implicit and explicit learning of languages (pp. 1-31). New York: Academic Press. Ellis, N. (2002). Frequency effects in language processing: A review with implications for theories of implicit and explicit language acquisition. Studies in Second Language Acquisition, 24, 143-188. Ellis, N. (2005). At the interface: Dynamic interactions of explicit and implicit language knowledge. Studies in Second Language Acquisition, 27, 305-352. Ellis, R. (2004). The definition and measurement of L2 explicit knowledge. Language Learning, 54(2), 227-275.  Ellis, R. (2005). Measuring implicit and explicit knowledge of a second language: A psychometric study. Studies in Second Language Acquisition, 27, 141-172.  Ellis, R. (2006). Modeling learning difficulty and second language proficiency: The differential contributions of implicit and explicit knowledge. Applied Linguistics, 27(3), 431-463.  Ellis, R. (2009). Measuring implicit and explicit knowledge of a second language. In R. Ellis, S. Loewen, C. Elder, R. Erlam, J. Philp, and H. Reinders (Eds.), Implicit and explicit knowledge in second language learning, testing, and teaching. (pp. 31-64). New York: Second Language Acquisition. Ellis, R., & Barkhuizen, G. (2005). Analysing learner language. Oxford: Oxford University Press.  Ellis, R., Loewen, S., Elder, C., Erlam, R., Philp, J., & Reinders, H. (2009). Implicit and explicit knowledge in second language learning, testing, and teaching. New York: Multilingual Matters.  Eretin, G., & Alptekin C. (2013). The explicit/implicit knowledge distinction and working memory: Implications for second-language reading comprehension. Applied Psycholinguistics, 34, 727-753.   Erlam, R (2009). The elicited oral imitation test as a measure of implicit knowledge. In R. Ellis, S. Loewen, C. Elder, R. Erlam, J. Philp, and H. Reinders (Eds.), Implicit and explicit knowledge in second language learning, testing, and teaching. (pp. 65-93). New York: Second Language Acquisition.  162 Godfroid, A., Loewen, S., Jung, S., Park, J., & Gass, S. (2015). Timed and untimed grammaticality judgments measure distinct types of knowledge: Evidence from eye-movement patters. Studies in Second Language Acquisition, 37, 269-297. Granena, G. (2012). Age differences and cognitive aptitudes for implicit and explicit learning in ultimate second language attainment. PhD dissertation. University of Maryland.   Guillelmon, D. & Grosjean, F. (2001). The gender marking effect in spoken word recognition. The case of bilinguals. Memory and Cognition, 29, 503-511. Gutiérrez, X. (2012). Implicit knowledge, explicit knowledge, and achievement in second language Spanish. The Canadian Journal of Applied Linguistics, 15(1), 20-41. Horowitz, R., & Samules, J. (1987). Comprehending oral and written language: Critical contrasts for literacy and schooling. In R. Horowitz, and Samuels (Eds), Comprehending oral and written language. (pp. 1-52). UK: Emerald Group Publishing Limited.  Hulstijn, J. (2002). Towards a unified account of the representation, processing and acquisition of second language knowledge. Second Language Research, 18, 193-223.  Hulstijn, J. (2005). Theoretical and empirical issues in the study of implicit and explicit second language learning. Studies in Second Language Acquisition, 27, 129-140. Hulstijn, J. (2015). Explaining phenomena of first and second language acquisition with the constructs of implicit and explicit learning: The virtues and pitfalls of a two-system view. In P. Rebuschat (Ed), Implicit and explicit learning of languages, (pp. 25-46). Amsterdam: John Benjamins.    Isemonger, I. (2007). Operational definitions of explicit and implicit knowledge: Response to R. Ellis (2005) and some recommendations for future research in this area. Studies in Second Language Acquisition, 29, 101-118.  Jeong, J. (2008).    9æ9¢  8F˜Â [A study on the acquisition of Korean consonants by Chinese learners: Focused on perception and production of plosives].  ˜¦9w%VG¯9& . Jeong, K., Pae, S., & Kim, K. (2006). 12, 18, 24˚r9* 8W9v7ı9®   [The early phonological development of Korean Children who are 12, 18, and 24 months old]. 9û7¶8F˜Â, 11(3), 1-15. Johnson, J., & Newport, E. (1989). Critical period effects in second language learning: The influence of maturational state on the acquisition of English as a second language. Cognitive Psychology, 21, 60-99.  Jöreskog, K. G., & Sorbom, D. (1989). LISREL 7: A guide to the program and applications (2nd edition). Chicago: SPSS.   163 Karmiloff-Smith, A. (1979) A functional approach to child language. Cambridge: Cambridge University Press.  Kim, J., & Nam, H. (2016). Measures of implicit knowledge revisited: Processing modes, time pressure, and modality. Studies in Second Language Acquisition, 1-27.  Kline, R. (2015). Principles and practice of structural equation modeling. New York: The Guilford Press.   Krashen, S. (1981). Second language acquisition and second language learning. Oxford: Pergamon. Kwon, K., Lee, Y., & Sohn, M. (1979). G²˜Ã 7ı&/9®  (I) [Development of Korean phonemes of Korean children (I)]. Seoul: Korean Educational Development Institute. Kim, H., & Kang, N. (2010).   -”.ë   8F˜Â (I): -˜6 8Œ -+Ò ;g5B9™*² [A study on grammar acquisition of Korean learners I: Focused on -ko and -eosseo. , 132, 292-323.  Kim, J., Park, D., Lee, B., Lee, H., Jeong, H., Choi, J., and Heo, Y. (2005). Korean grammar I: Structures. Seoul: Communication Books. Kim-Renaud, Y. (2009). Korean: An essential grammar. New York: Routlege.  Kim, S. (2006).  9æ9¢8& %VG² ;g˜Ã9Î9®  .ê;R8Œ  8F˜Â [Chinese learners perceptual range and acquisition of Korean consonants]. %VG¯˜¦ . Ko, Y., & Ku, B. (2008).  -”.ë*¶ [A review of Korean grammar]. Seoul: <'-”%O.  Matsunaga, S. (2003). Instructional needs of college-level learners of Japanese as a heritage language: Performance based analysis. Heritage Language Jouranl, 1(1), 58-75. Laufer, B., & Goldstein, Z. (2004). Testing vocabulary knowledge: Size, strength, and computer adaptiveness. Language Learning, 54(3), 399-436.  Lee-Ellis, S. (2009). The development and validation of a Korean C-Test using Rasch Analysis. Language Testing, 26(2), 245-274.  Lee, I., & Ramsey, R. (2000). The Korean language. New York: State University of New York Press. Lee, H., & Lee, S. (2010). G¯*‘ :Z 7ı&/9®  .r%B8& ˜VG² :Û%> 8F˜Â [A longitudinal study of Korean preschoolers development of cohesive devices]. G²˜Ã , 9(2), 19-32.  164 Lee, H., Jang, Y., Choi, Y., & Lee. S. (2008). /Ö,þ/J˜6+Ò  :Û%> ˜V>ƒG² G²˜Ã 7ı&/9® ?^ ƒ -”.ëHkC23â  [Acquisition of grammatical function words in Korean children]. , 13(2), 159-173. Lee, J. (2003). Research on errors committed by learners of Korean. Seoul: Pakeejeong. Lee, Y., Kim, H., Kong, D., Hong, J., & Long, M. (2005). Variation of the linguistic profiles of advanced English-speaking learners of Korean. Language Research, 41(2), 437-456.  Lenneberg, E. (1967). Biological foundations of language. New York: Wiley. Loewen, S. (2009). Grammaticality judgment tests and the measurement of implicit and explicit L2 knowledge. In R. Ellis, S. Loewen, C. Elder, R. Erlam, J. Philp, and H. Reinders (Eds.), Implicit and explicit knowledge in second language learning, testing, and teaching. (pp. 94-112). New York: Second Language Acquisition Loewen, S. (2003). Grammaticality judgment tests: What do they really measure? Paper presented at the Second Language Research Forum, Tucson, AZ. MacLaughlin, B., Rossman, T., & McLeod, B. (1983). Second language learning: An information-processing perspective. Language Learning, 33, 135-158.  Mitchell, R. (1993). Path Anaysis: Pollination. In SM Schneider, and J. Gurevitch (Eds.), Design and analysis of ecological experiments. (pp. 211-231). NY: Chapman and Hall. Montrul, S. (2004). Subject and object expression in Spanish heritage speakers. A case of morpho-syntactic convergence. Bilingualism: Language and Cognition, 7, 125-142. Montrul, S. (2005). Second language acquisition and first language loss in adult early bilinguals: Exploring some differences and similarities. Second Language Research, 21(3), 199-249. Montrul, S. (2006). On the bilingual competence of Spanish heritage speakers: Syntax, lexical-semantics and processing. International Journal of Bilingualism, 10(1), 37-69. Montrul, S. (2008). Incomplete acquisition in bilingualism. Amsterdam: John Benjamins. Montrul, S. (2009). Re-examining the fundamental difference hypothesis: What can early bilinguals tell us? Special issue of Studies in Second Language Acquisition, 3(2), 225-257. Montrul, S. (2010). How similar are adult second language learners and Spanish heritage speakers? Spanish clitics and word order. Applied Psycholinguistics, 31, 167-207.                   Montrul, S., Foote, R., & Perpinan, S. (2008). Gender agreement in adult second  165 language learners and Spanish heritage speakers: The effects of age and context of acquisition. Language Learning, 58(3), 503-553. Munro, M., & Derwing, T. (1995). Processing time, accent, and comprehensibility in the perception of native and foreign-accented speech. Language and Speech, 38, 289-306.   Muñoz, C. (2014). Contrasting Effects of Starting Age and Input on the Oral Performance of Foreign Language Learners. Applied Linguistics. 35(4), 463482.  OGrady, W., Lee, M., & Choo, M. (2003). A subject-object asymmetry in the acquisition of relative clauses in Korean as a second language. Studies in Second Language Acquisition, 25, 433-448.  Oh, J., Jun, S., Knightly, L., & Au, T. (2003). Holding onto childhood language memory. Cognition, 86, B53-B64. Paradise, M. (1994). Neurolinguistic aspects of implicit and explicit memory: Implications for bilingualism and second language acquisition. In N. Ellis (Ed.) Implicit and explicit language learning (pp. 393-419). New York: Academic Press. Paradise, M. (2004). A neurolinguistic theory of bilingualism. Amsterdam: John Benjamins.  Paradise, M. (2009). Declarative and procedural determinants of second languages. Amsterdam: John Benjamins. Philp, J. (2009). Pathways to proficiency: Learning experiences and attainment in implicit and explicit knowledge of English as a second language. . In R. Ellis, S. Loewen, C. Elder, R. Erlam, J. Philp, and H. Reinders (Eds.), Implicit and explicit knowledge in second language learning, testing, and teaching. (pp. 194-215). New York: Second Language Acquisition Pienemann, M. (1998). Language processing and second language development: Processibility theory. Amsterdam: John Benjamins. Polinsky, M. (2011). Reanalysis in adult heritage language: New evidence in support of attrition. Studies in Second Language Acquisition, 33, 305-328.    Reber, A. (1993). Implicit learning and tacit knowledge; An essay on the cognitive unconscious. Oxford: Oxford University Press.  Rebuschat, P. (2013). Measuring implicit and explicit knowledge in second language research. Language Learning, 63(3), 595-626. Rebuschat, P. & Williams, J. (2009). Implicit learning of word order. Proceedings of the 31th Annual Conference of the Cognitive Science Society, 425-430. Robinson, P. (2005). Cognitive abilities, chunk-strength, and frequency effects in implicit artificial grammar and incidental L2 learning: Replications of Reber,  166 Walkenfeld, and Hernstadt (1991) and Knowlton and Squire (1996) and their relevance for SLA. Studies in Second Language Acquisition, 27, 235-268. Roehr, K., & Gutiérrez, G. (2009). The status of metalinguistic knowledge in instructed adult L2 learning. Language Awareness, 18(2), 165-181. Scheffler, P. (2011). Rule difficulty: teachers intuitions and learners performance. Language Awareness, 20(3), 221-237. Scovel, T. (2000). A critical review of the critical period research. Annual Review of Applied Linguistics, 20, 213-223. Seo, H., & Lee, S. (1999). 2-53”  7ı&/9®  .r%B [Development of conjunctives of normal children aged from ages of 2 to 5]. 8F˜Â, 4, 167-185. Seol, W. (2010). ;g˜Ã9Î    -”.ë  ˜VG² 8F˜Â: -8&, -8&3r, -*²+Ò ;g5B9™*² [A study on Korean grammar acquisition of Chinese learners: Focused on -e, -eseo, and ro. 8W"þ%VG¯˜¦ . Sohn, H. (1999). The Korean language. Cambridge: Cambridge University Press. Sogang Korean Institute (2008). Sogang Korean New 2(A). Seoul: Sogang International Culture Educational Institute.    Song, M., OGrady, W., Cho, S. & Lee, M. (1997). The learning and teaching of Korean in community schools. In Korean Language in America 2, Y, Kim (ed.), 111-127. American Association of Teachers of Korean. Spada, N., Shiu J., & Tomita, Y. (2015). Validating an elicited imitation task as a measure of implicit knowledge: comparisons with other validation studies. Language Learning, 65(3), 723-751. Stevens, J. (1996). Applied multivariate statistics for the social sciences (3rd ed.). New Jersey: Lawrence Erlbaum Associates. Suzuki, Y., & DeKeyser, R. M. (2015). Comparing elicited imitation and word monitoring as measures of implicit knowledge. Language Learning, 65(4), 860-895. Ullman, M. (2001). The neural basks of lexicon and grammar in first and second language: The declarative/procedural model. Bilingualism: Language and Cognition, 4, 105-122.   Ullman, M. (2004). Contribution of memory circuites to language: The declarative/ procedural model. Cognition, 92, 231-270. Ullman, M. (2005). A cognitive neuroscience perspective on second language acquisition: The declarative/procedural model. In C. Sanz (Ed.), Mind and context  167 in adult second language acquisition: Methods, theory and practice (pp. 141-178). Washington, DC: Georgetown University Press.   Um, I. (1989). Acquisition order of Korean particles and influence of tension on second language acquisition. , 1, 169-203. Vafaee, P., Kachinske, I., & Suzuki, Y. (2016). Validating grammaticality judgment tests: Evidence from two new psycholinguistic measures. Studies in Second Langauge Acquisition, 1-37. Valdés, G. (2001). Heritage language students: Profiles and possibilities. In J. Peyton, J. Ranard, & S. McGinnis (Eds.), Heritage languages in America: Preserving a national resource (pp. 37-80). McHenry, IL: The Center for Applied Linguistics and Delta Systems.  Zhang, R. (2015). Measuring university-level L2 learners implicit and explicit linguistic knowledge. Studies in Second Language Acquisition, 1-30.