THE EFFECT OF PRIOR L1 KNOWLEDGE ON THE IMPLICIT AND EXPLICIT LEARNING OF L2 SYNTAX FROM READING A NOVEL By Jieun Ahn A DISSERTATION Submitted to Michigan State University in partial fulfillment of the requirements for the degree of Second Language Studies Doctor of Philosophy 201 9 ABSTRACT THE EFFECT OF PRIOR L1 KNOWLEDGE ON THE IMPLICIT AND EXPLICIT LEARNING OF L2 SYNTAX FROM READING A NOVEL By Jieun Ahn A growing number of researchers ha ve considered the acquisi tion of L2 grammar under incidental, or meaning - focused, learning conditio ns in order to understand the scope and possibilities of naturalistic adult L2 learning . However, the effect of the first language (L1) in the incidental learning of L2 syntax has no t been directly studied. To address this gap, in the present dissertation, I investigated the effect of prior L1 knowledge on the implicit and explicit learning of L2 syntax under incidental learning conditions . Forty L1 English (head - initial, right - bran ching) and forty L1 Korean (head - final, left - branching) speak ers read the novel The Mysterious Affair at Styles (Christie, 1920) rewritten in a semi - artificial language, Koreanish , which consisted of English vocabulary and head - final Korean syntax . The par eye movements were recorded during reading by usin g an EyeLink Portable Duo eye tracker ( SR Research, Canada) . From their eye movements, I derived two processing measures: - ti me responses to word order violations. After reading, the par ticipants were immediately tested with a surprise grammaticality judgment test (GJT) with source attributions, which was followed by the first part of the debriefing interview. Two weeks later, t he participants completed a delayed GJT and the second part o f the debriefing interview. Triangulation of the online and offline measures exhibited the significant and pervasive e ffect s of prior L1 knowledge on the incidental acquisition of L2 syntax. Du ring the exposure task (novel reading) , th e Korean experiment al (KE) group exhibited a faster initial decrease in sentence reading times than did the English experimental (EE) group. Furthermore, o nly the KE group showed online grammatical sensitivity . Par ticularly, t he Korean L1 - aware participants , w ho later became aware of the cross - linguistic similarity between the target language and their L1, showed robust sensitivity effect s, even before L1 awareness emerged. Subsequently , on the GJT , the KE group exh ibited strong er evidence of implicit and explici t knowledge than did t he EE group . The syntactic knowledge was significantly enhanced over time , which prompted them to outperform the EE group on the delayed GJT. These findings indicated that t he Korean speakers certainly had advantages in acq uiring the target word order, which followed their L1 Korean, and L1 awareness gave them an edge in knowledge development. In terms of online processing data, although both the KE and EE groups sped up over t ime while reading the exposure text, their learn ing curves had a different form. The EE the power law of practice (Anderson, 1982), whereas those of the KE group did not follow a clearly discernable pattern. The findings of this dissertation suggest that L2 learners with a different L1 background perform on an un equal footing because of their prior L1 experience, especially in naturalistic learning contexts. The two L1 groups, in fact, presented oppos ing direction s of knowled ge development. The Korean participants who had relevant prior L1 knowledge showed a progression from implicit to explicit knowledge under incidental exposure , whereas the English participants who lacked relevant prior L1 knowledge began with ex plicit know ledge and failed to reach implicit knowledge. This dissertation therefore elucidated the multifaceted aspects of L1 influence on L2 development, uncovering the complex nature of linguistic transfer, variability in L2 development, and the moderat ing role of awareness in meaning - focused, incidental learning condition s. Copyright by J IEUN AHN 201 9 v ACKNOWLEDGEMENT S Regarding my ability to achieve this wonderful accomplishment, I am greatly indebted to my dedicated advisor Dr. Aline Go dfroid for going above and beyond anything I could have expected from an advisor. I would like to express my deepest appreciation for h er persistent guidance and support through each stage of the process. Her enthusiasm and love for research are contagious . She opened a door to the world of eye - tracking for me, and it was like riding a magic carpet. Under her guidance, I successfully overcame many difficulties. Her faith in me str engthened my resolve, and s he has shown me what a good researcher and person s hould be. I have been extremely lucky to have her as my mentor. I would also like to extend my deepest gratitude to all my committee members for provid ing me with extensive personal and professional guidance. Without their constructive and intensive feedb ack, this dissertation would not have been possible. My earnest thanks go to Dr s . Sue Gass , Patti Spinner , and Shawn Loewen. Dr. Gass is an amazing role model of what it means to be a superior mentor, teacher, and leader. She is indeed a multitasking geniu s. Dr. Spinner nurtured my enthusiasm for syntax and always shared her expertise willingly. In her morphosyntax class, I was able to learn the beauty of syntax. Dr. Loewen believed in my abilities and provided me with continuous support and encouragement. He is very approachable and always gives care for hi s students. My sincere thanks also go to Drs. Zoltan Dienes and Patrick Rebuschat for their willing ness t o serve as external dissertation committee member s, which meant a great deal to me. Dr. Dienes is a statistics guru who provided a new angle for lookin g at my complicated eye - tracking data. I am extremely appreciat ive of h is insightful and valuable advice. I am also grateful to Dr. Rebuschat for his support throughout my research project . His vi expertise in implicit and explicit learning helped me develop my research design. Throughout this arduous work, I always had my family and friends to count on when times were t ough. Special thanks are due to my husband Dr. Joon Mahn Lee for his unparallel ed patien ce and support. Without him, I might never have reac hed this point. He was always beside me during the good and bad moments to push and motivate me. He has shown and taught me what true love is. I must also thank my lovely daughter Jiyoung Joan Lee for giv ing me the strength and patience to work through eve ry obstacle with which I was presented . She is the most beautiful gift that this world has g iven me. Special thanks are extended to my parents for their unconditional love and care. Their support was the source of my perseverance over the years. I am forever indebted to my parents for giving me the opportunities and experiences that have made me who I am. Many thanks also go to my parents - in - law for their incredible support and encouragement. They took car e of Jiyoung three times a week so that I could focus on writing my dissertation. Had it not been for their support, I could not have comp l eted th is work. My t hank s also go to everyone in the SLS family a group of genuinely nice people who want to help one an other. I cherish all the good times th at we have spent together. I also wish to thank all my friends (especially Saeyoon Kim, Soojin Kim, and Juwon Eum) who supported me during my Ph . D . journey . They were always around when I thought that it was impossi ble to continue. This journey would not have been possible if not for my family and friends, and I dedicated this milestone to them. vii TABLE OF CONTENTS LIST OF TABLES ................................ ................................ ................................ ...................... LIST OF FIGURES ................................ ................................ ................................ ...................... CHAPTER 1 ................................ ................................ ................................ ................................ ... 1 THE EFFECT OF PRIOR L1 KNOWLEDGE ON THE INCI DENTAL LEARNING OF L2 SYNTAX ................................ ................................ ................................ ................................ ....... 1 1 . 1 The P resent D issertation ................................ ................................ ................................ ........ 4 1 . 2 Overview of the D issertation ................................ ................................ ................................ . 5 1 . 3 Definition of K ey T erms ................................ ................................ ................................ ....... 5 CHAPTER 2 ................................ ................................ ................................ ................................ ... 9 OFFLINE MEAUSRE: GRAMMATICALITY JUDGMENTS ................................ .................... 9 2.1 Background ................................ ................................ ................................ ........................... 9 2. 1 .1 Development of implicit and explicit knowledge i n incidental learning conditions ..... 9 2.1.2 The effect of L1 on implicit and explicit knowledge development in incidental learning conditions ................................ ................................ ................................ ................ 1 2 2.1.3. Research questions ................................ ................................ ................................ ...... 1 6 2.2 M ethod s ................................ ................................ ................................ .............................. 1 7 2.2.1 Participants ................................ ................................ ................................ ................... 1 7 2.2.2 Targeted semi - artificial language: Kor eanish ................................ .............................. 1 9 2.2.3 Materials ................................ ................................ ................................ ...................... 2 2 2.2.4 Procedure ................................ ................................ ................................ ..................... 3 1 2.2.5 Analysis ................................ ................................ ................................ ....................... 3 3 2. 3 Results ................................ ................................ ................................ ................................ . 3 5 2. 3.1 Comprehension test ................................ ................................ ................................ ...... 3 5 2. 3.2 Overall GJT performance ................................ ................................ ............................ 3 5 2. 3 .3 Awareness reflected in retrospective verbal repo rts and GJT performance ................ 4 1 2. 3 .4 Awareness reflected in sour ce attributions and GJT performance .............................. 4 8 2. 3 .5 Summary of results ................................ ................................ ................................ ...... 5 3 2. 4 . Discussion ................................ ................................ ................................ .......................... 5 4 CHAPTER 3 ................................ ................................ ................................ ................................ . 6 1 ONLINE MEAUSRE: READING TIMES FROM EYE - TRACKING ................................ ....... 6 1 3.1 Background ................................ ................................ ................................ ......................... 6 1 3.1.1 Theoretical background on L2 developmental processes ................................ ............ 6 1 3.1.2 Incidental exposure and change in reading times ................................ ........................ 6 5 3.1.3 Incidental exposure and sensitivity to L2 violations ................................ ................... 6 7 3.1. 4 Research questions ................................ ................................ ................................ ....... 70 3.2 Method s ................................ ................................ ................................ ............................... 70 3. 2. 1 Participants ................................ ................................ ................................ ................... 70 3.2.2 Targeted s emi - artificial language: Koreanish ................................ .............................. 7 2 viii 3.2.3 Materials ................................ ................................ ................................ ...................... 7 4 3.2.4 Procedure ................................ ................................ ................................ ..................... 8 2 3.2.5 Analysis ................................ ................................ ................................ ....................... 8 3 3. 3 Results ................................ ................................ ................................ ................................ . 8 6 3. 3 .1 Comprehension test ................................ ................................ ................................ ...... 8 6 3. 3.2 Analyses of retrospective verbal reports ................................ ................................ ..... 8 6 3. 3 .3 Changes in residualized sentence reading time: Learning curve ................................ 8 9 3. 3 .4 Sensitivity to violations: Integrated knowledge ................................ ........................... 9 6 3. 3 .5 Summary of re sults ................................ ................................ ................................ .... 10 1 3. 4 Discussion ................................ ................................ ................................ ......................... 10 2 CHAPTER 4 ................................ ................................ ................................ ............................... 10 9 GENERAL DISCUSSION AND CONCLUSION ................................ ................................ .... 10 9 4.1 Summary of the Findings ................................ ................................ ................................ .. 10 9 4.1.1 English speakers ................................ ................................ ................................ ........ 10 9 4.1.2 Korean speakers ................................ ................................ ................................ ......... 10 9 4.2 General Discussion ................................ ................................ ................................ ............ 11 1 4.3 Limitation ................................ ................................ ................................ .......................... 11 5 4.4 Conclusion and Outlook ................................ ................................ ................................ .... 11 6 APPENDICES ................................ ................................ ................................ ............................ 117 Appendix A: Sample of exposure task (novel reading) ................................ .......................... 118 Appendix B : Sample of comprehension check questions ................................ ....................... 120 Appendix C : Sample of the grammaticality judgment test ................................ ..................... 121 Appendix D : Sample of the debriefing interview questions ................................ ................... 1 24 REFERENCES ................................ ................................ ................................ ........................... 1 26 ix LIST OF TABLES Table 1.1 Definition of key terms ................................ ................................ ................................ ................... 6 Table 2 .1 backgrounds and English proficiency ................................ .. 1 9 Table 2.2 The counterbalanced versions of the exposure task ................................ ................................ ..... 2 4 Table 2.3 Counterbalanced grammatical an d ungrammatical stimuli used in the testing set ....................... 2 9 Table 2.4 Mean accuracy (%) of grammaticality judgments for English and Korean speakers .................. 3 5 Table 2.5 scores of the grammaticality judgments for English and Korean speakers ............................. 3 7 Table 2.6 Mixed - design A NOVA for the scores on GJTs ................................ ................................ ....... 3 9 Table 2.7 ................................ ................................ ............................ 4 3 Table 2.8 v erbal r eport d ata ................................ ................................ ............................ 4 3 Table 2.9 GJT performance by awareness reflected in retrospective verbal report s ................................ ................................ ................................ ................................ ...................... 4 5 Table 2. 10 Th a ccuracy (%) across source attributions ..................... 50 Table 2. 1 1 .............................. 5 1 Table 3.1 ................................ .. 7 2 Ta ble 3.2 The counterbalanced version of the exposure task ................................ ................................ ....... 7 8 x Table 3.3 Regression output of the final model (M3) ................................ ................................ ................. 9 2 Table 3.4 Regression output for the English speakers ................................ ................................ ................. 9 3 Table 3.5 Regression output for the Korean speakers ................................ ................................ .................. 9 4 Table 3.6 The English spe ................................ ... 9 6 Table 3.7 ............ 9 9 Table 4.1 Summary of the findings ................................ ................................ ................................ ............ 11 1 xi LIST OF FIGURES F igure 2.1 Procedure in the experiment ................................ ................................ ................................ ......... 3 2 Figure 2 .2 Mean scores for Korean and English speakers ................................ ................................ ........ 3 7 Figure 2.3 Three - way interactions between Condition, Time, and L1 ................................ .......................... 3 8 Figure 2.4 Two - way interaction between Time and L1 - awareness for the KE group ................................ ... 4 8 Figure 2. 5 Two - way interaction between Attributions and L1 ................................ ................................ ...... 5 2 Figure 3.1 - log plot ................................ ................................ ............................... 90 Figure 3.2 - log plot ................................ ................................ ............................... 9 1 Figure 3.3 The log - log plot for unaware and VP - aware English speakers ................................ .................... 9 5 Figure 3.4 The log - log plot f or unaware, VP - aware, and L1 - a ware Korean speakers ................................ .. 9 5 Figure 3.5 ................................ 9 7 Figure 3.6 Two - way interaction between Block and Awareness for the Korean speakers ......................... 10 1 Figure 3.7 The English a residualized sentence reading times across blocks ....... 10 5 1 CHAPTER 1 THE EFFECT OF PRIOR L1 KNOWLEDGE ON THE INCIDENTAL LEARNING OF L2 SYNTAX A growing number of researchers ha ve considered the acquisition of L2 grammar under inci dental, or meaning - focused, learning conditions in order to understand the scope and possibilities of naturalistic adult L2 learning (Denhovsk a , Serratrice, & Payne, 2016; Godfroid, 2016; Grey, Williams, & Rebuschat, 2014; Kerz, Wiechmann, & Riedel, 2017; Kim & Godfroid, 2019; Leung & Williams, 2012, 2014; Miller & Godfroid, 2019; Morgan - Short, Deung, Brill - Schuetz, Farreta - Stutenberg, Wong, & Wong, 2015; Morgan - Short, Steinhauer, Sanz, & Ullman, 2012; Reb uschat & Williams, 2012; Robinson, 1995, 2005; Roger s, Revesz, & Rebuschat, 2015; Tagarelli, Ruiz, Moreno, & Rebuschat, 2016; Williams & Kuribara, 2008; Williams, 2011). This increased interest in incidental learning conditions reflects one of the core que stions in Second Language Acquisition ( SLA ) : H ow and to what extent is adult L2 acquisition comparable to child L1 acquisition? There is a broad consensus that incidental exposure to a first language ( L1 ) in naturalistic settings uniformly results in child consists of a complete set of intuitive, tacit, and implicit linguistic knowledge. Specifically, children incidentally pick up regularities from the L1 input without intending to learn and eventually master the comple x system of L1 without an awareness of what has been learned. In child L1 acquisition, implicit knowledge, which is naturally developed outside the classroom, underlies p such unconscious, implicit knowledge can be remark ably powerful (Dienes, 2008; Reber, 1993) and represents an elementary aspect of human cognition. In this sense, many SLA researchers have paid special attention to the possibility of 2 L2 learning because the goal of L2 instruction is t o develop implicit knowledge for fluent L2 use (Doughty, 2003). Of primary interest is the question of whether adult learners can also acquire a L2 without awareness under meaning - focused learning conditi ons, as children do in their L1 acquisition. Althoug h some evidence denies the possibility of learning without awareness (e.g., Hama & Leow, 2010; Leow, 2015), and not all studies report evidence for learning without awareness (e.g., Hamrick, 2013; Kim & G odfroid, 2019; Miller & Godfroid, 2019), there is ac cumulating evidence to support the opposite opinion: that adult learners can develop some unconscious, implici t k nowledge of L2 grammar under incidental learning conditions (e.g., Godfroid, 2016; Grey et al., 2014; Leung & Williams, 2011, 2012, 2014; Rebus chat & Williams, 2012; Williams, 2005, 2011; Williams & Kuribara, 2008). to acquire L2 knowledge under mean ing - focused, incidental learning conditions. A ltho ugh child L1 acquisition sets a good benchmark for adult L2 acquisition, it is imperative to note the differences between child L1 acquisition and adult L2 acquisition. Clearly, adult L2 learners bring pr ior L1 knowledge to their L2 learning. They already possess L1 representation s , which involve symbolic knowledge and linguistic notions, such as subject and predicate; hence, L2 learners are not empty vessels. G iven that adult L2 acquisition is built on well - established L1 knowledge and representations, the dominant L1 knowledge may give rise to L1 - based habits during L2 acquisitional processes, especially in the case of naturalistic L2 learning. Leung and Willia ms (2014), for example, demonstrated the effect of L1 on the implicit learning of L2 form - meaning mappings (classifiers) under meaning - focused conditions. However , the effect of L1 on other linguistic units, such as syntax, has not been fully examined un der incidental learning conditions. The role of prior L1 knowledge in the incidental 3 learning of L2 syntax therefore remains an open question in the field. By examining whether and how L1 knowledge influences L2 acquisition under meaning - focused conditions , this exploration has the potential to make a significant contribution to SLA. It adds to our un derstanding of whether and how prior L1 knowledge and the implicit and explicit L2 learning of syntax interact . To close this research gap in the present disse rtation, I aim to address how prior L1 knowledge affects the development of implicit and explicit syntactic knowledge from a purely meaning - focused task: reading a novel. Reading a novel had advantages over performing a highly controlled, sentence - level ta sk in the context of incidental learning. Specifically, it provided prolonged exposure to target patterns in a more ecologically valid research context and ensured that the participant s focus ed more closely on meaning. The following general questions gui de the present dissertation research: Does prior L1 knowledge affect the incidental learning of L 2 syntax consciously, unconsciously, or both? Does prior L1 knowledge affect learning processes as well as learning outcomes under incidental learning conditio ns? To answer these questions, I compared two L1 groups English (head initial, right - branching) a nd Korean (head - final, left - branching) that were learning a semi - artificial language under incidental learning conditions. The semi - artificial language, Korean ish , consisted of English vocabulary and head - final syntax had a pattern that was reverse to the target Koreanish syntax (i.e., a cross - linguistic Koreanish syntax (i.e., a cross - linguistic similarity). Further, I used multiple measures to tra trajectories: eye movements as an online measure of real - time processing and knowledge and grammaticality judgments as an offline kn owledge measure. These eye - movement measures were advantageous, in that they enabled me to employ a within - subject design (revealing how 4 - subjects design (showing how much extra read ing time the experimental groups spend compared to the control groups). I triangulated online mea sures derived from eye movements (residualized sentence reading times and grammatical sensitivity) with other offline knowledge measures ( grammaticality judgme nts) and with awareness measures (source attributions, retrospective verbal reports). Together, t hese methodological components have the potential to reveal dynamic relationships among prior L1 knowledge, awareness, learning, and knowledge under incidental learning conditions. 1 . 1 The P resent D issertation For the present dissertation, I investigate d the effect of prior L1 knowledge on the implicit and explicit learning of nonnative syntax under incidental exposure. I focused on how prior L1 knowledge not only constrains but also facilitates the development of implicit and explicit L2 knowledge. To d o so, I introduced methodological advances into this study. First, I used an authentic novel to incidentally expose participants to the target semi - a rtificial language. ts while they were reading a novel (Godfroi d, Ahn, et al., 201 8 ) and from their eye movement records I extracted online measures of learning and knowledge. A decrease in sentence reading times during training was used as an online measure of learning; an i ncreased sentence reading time to word orde r violations was used as an online measure of integrated knowledge. I was further interested in the shape of the learning trajectory (i.e., the decrease in sentence reading time) and whether or not it followed the power law of practice (Anderson, 1982; DeK eyser, 1997, 2015; Newell & Rosenbloom, 1981) . I triangulated the online measures derived from eye movements with an offline knowledge 5 measure (GJT) as well as two offline measures of awareness (Rebuschat, 2013; R ebuschat et al., 2015), namely, source attr ibutions (Dienes, 2004; Dienes & Scott, 2005; Dienes, 2008) and retrospective verbal reports. Third, this study is one of few to include a delayed offline GJT to capture the durability of the learning effect (Grey et al., 2014). These methodological device s shed light on the multif aceted effects of L1 transfer during the incidental learning of syntax. 1 . 2 Overview of the D issertation This dissertation is organized into four chapters. In chapter 1, I present the background and motivation for the dissertat ion and provide definitions of key terms. Chapter 2 focuses on the offline measure of grammaticality judgments, and chapter 3 concerns the online measures derived from the eye movements. Each chapter presents a li terature review, research questions, method s, results, and a discussion. Finally, in chapter 4, I summarize the findings of the study and present a general discussion. 1 . 3 Definition of K ey T erms In this section, I provide the definition s of key constructs in the dissertation and the correspond ing measurement variables used to operationalize them (Table 1.1). The key constructs include learning , power law of practice , knowledge , integrated knowledge , grammatical sensitivity, awareness , incidental learning , implicit knowledge , and explicit knowle dge . 6 Table 1.1 Definition of key terms Definition Measurement variable Learning The process of acquiring new knowledge from the input. A decr ease in sentence reading times during the exposure task. Power law of practice The ubiquitous decrease in th e time required to perform a task that follows a power function ( Y = X - n ) involv ing the process of automatization. A decrease in the sentence reading times that follows the power function. Knowledge The outcome of the learning process. (a) Online: gra mmatical sensitivity An increased reading time in response to syntactic violations during the exposure task. (b) Offline: grammaticality judgments Above - chance performance on grammaticality judgments after the exposure task. Integrated knowledge The m en tal representation of the knowledge that underlies automatic competence . An increased reading time in response to syntactic violations during the exposure task. Grammatical sensitivity The a bility to apply integrated knowledge of syntax in an automatic manner. An increased reading time in response to syntactic violations during the exposure task. 7 Awareness The conscious perception of what is being learned. (a) Offline: source attributions in the GJT Above - chance performance on grammaticality judgments attributed to recollection and rule knowledge. (b) Offline: retrospective verbal reports An a bility to verbalize the target rules. Incidental learning Learning conditions in which participants are exposed to target rules throug h a meaning - foc used activity without a prior notice of testing. Reading for comprehension a novel in which the target rules are embedded throughout the text. Implicit knowledge Unconscious knowledge that exists outside of the (a ) Online: sensi tivity to violations An unconscious slowdown in sentence reading times in response to syntactic violations during the exposure task. grammaticality judgments ; grammaticality judgments based on guessing and int uition A c ompar with that of the control group ; t above - chance performance . Above - chance performance on grammaticality judgments attributed to guess es and intuition. 8 Expli cit knowledge Conscious knowledge that the participants are aware that they know. (a) Offline: retrospective verbal reports A n a bility to verbalize the target rules. (b) Offline: grammaticality judgments based on recollection and rule knowledge Above - chan ce p erformance on grammaticality judgments attributed to recollection and rule knowledge. 9 CHAPTER 2 OFFLINE MEASURE: GRAMMATICAL ITY JUDGMENTS 2 . 1 Background 2 . 1 . 1 Development of implicit and explicit knowledge in incidental learning conditi ons Many studies targeting the incidental learning of non - native grammar have shown that learners can acquire some knowledge of the grammar without awareness ( Godfroid, 2016; Grey, Williams, & Rebuschat, 2014; Kerz, Wiechmann , & Riedel, 2017; Rebuschat & Williams, 2012; Robinson, 1995; Rogers, Revesz, & Rebuschat, 2015 ; Williams, 2011; Williams & Kuribara, 2008; but for an opposing view, see Hama & Leow, 2010; Leow, 2015b). In general, this body of work has exposed participants t o a semi - artificial languag e through a meaning - focused task (the training phase ) and then tested their knowledge of the target grammar without prior notice (the testing phase the target grammar rules , the re searchers investigated whet her the unaware participants showed learning effects on the knowledge measure, such as GJTs , to investigate the possibility of implicit learning. The findings from previous research have suggested that adult learning of non - nativ e syntax can take place in the absence of awareness of the target rules under incidental exposure conditions (but see Andringa, in press; Curcic, Andringa, & Kuiken, 2019; Hamrick, 2013; Kim & Godfroid, 2019; Miller & Godfroid, 2019 ), and that implicit learning without awareness is just slightly above chance on the knowledge measure. The latter result leads one to ask why the observed effect of implicit learning was so small. One answer may lie in the multiple mediating factors within the experimental de sign. One such factor that p ossibly affects incidental learning is the nature of exposure that learners receive during training. For example, recent studies found that the frequency 10 (Denhovska et al., 2016) and linguistic complexity (Tagarelli, Ruiz, Mor eno, & Rebuschat, 2016) of the input given to learners play a role in implicit learnin g. A related factor that has yet to receive ample attention is the type of exposure task. In general, researchers have used meaning - focused, psycholinguistic tasks as exp osure activities, such as plau sibility judgments (e.g., Rebuschat & Williams, 2012; Williams, 2011) or sentence - picture matching tasks (e.g., Godfroid, 2016; Leung & Williams, 2011). Such tasks can be advantageous because they allow researchers to exert ti ght control over the stimuli i n terms of sentence type or word frequency. However, the amount of input learners receive from such tasks may not be enough to engender robust learning. Furthermore, processing isolated single sentences is arguably a somewhat artificial task. As a growing body of language - processing research (Brennan, Hasson, Malach, Heeger, & Pylkkänen, 2012; Speer , Reynolds , Swallow, & Zacks, 2009; Nijhof & Willems, 2015; Willems, Frank, Nijhof, Hagoort, & Van den Bosch, 2015) and incidental vocabulary learning research ( Elgort, Brysbaert, Stevens, & Van Assche, 2018; Elgort & Warren, 2014; Godfroid , Ahn, et al., 2018; Mohamed, 2018; Pellicer - Sanchez & Schmidt, 2010) suggests, using m ore naturalistic stimuli in longer texts could overcome the problem of artificiality in ps ycholinguistic experiments. Reading a novel, for example , has three advantages for incidental learning over performing a highly controlled, sentence - level task. First, a novel is more engaging and contains more meaningful co ntent, which ensures a stronge r participant focus on meaning. Second, a novel provides exposure to target patterns over a greater length of time, and such rich experiences of the target patterns may increase the probability of learning. Third, reading a no vel represents a more natural reading task and resembles an everyday language learning activity; thus, it is ecologically valid (Godfroid et al., 2018). Hence, a longer text from a novel can be 11 considered a promising medium to convey the targeted syntax fo r the purpose of exploring the implicit and explicit learning of syntax. In light of this, I opted to use novel reading as an exposure task in this dissertation to train participants incidentally on target syntactic patterns in a more ecologically valid re search context. Another fact or that has been overlooked is the passage of time, that is, the delayed effect of incidental exposure (Grey et al., 2014; Morgan - Short et al., 2012; Robinson, 2002). Although Norris and Ortega (2000), in their meta - analysis o n the effectiveness of L2 inst ruction, reported a decline in the L2 learning effect from immediate tests to delayed tests (but for a contrasting view, see Spada & Tomita, 2010 ), this tendency apparently does not always occur , particularly in the incidental learning of L2 grammar. Morga n - Short et al. (2012), in an event - related potentials study, found positive evidence for L2 grammar development from implicit training, which was reflected in the consolidation of knowledge over several months. Both instructional groups showed increased na tivelike neural processing, even after a substantial, three - to - five - month, delay with no further exposure. Further, the implicit instruction group, who received incidental exposure to Brocanto2 o nly, demonstrated greater nativelike syntactic processing tha n the explicit learning group, who received incidental exposure and a metalinguistic explanation, before and after the delay. This indicated that the incidental exposure was likely linked to a mo re nativelike neurocognitive processing. Grey et al. (2014) demonstrated the delayed effects of incidental exposure on the implicit learning of L3 case marking and word order. In their study, the researchers found that learning from incidental exposure w as not only retained, but also showed improvement, in the del ayed test, which was administered two weeks after the immediate test. Additionally, awareness of the rules seemed to play a crucial role in acquiring word order. Specifically, awareness of the co rrect 12 rules appeared to increase the knowledge of word order, as demonstrated in the delayed test. Based on these results, the authors pointed out the importance of delayed testing to explore the effect of time on the incidental learning of grammar. Buildi ng on Morgan - Short et al. (2012) and Grey et al. (2014), in t his dissertation I too incorporate delayed testing to capture the changes in L2 grammar development over time. 2 . 1 . 2 The effect of L1 on implicit and explicit knowledge development in incidental learning conditions The role of L1 in L2 acquisition has b een studied extensively over three decades of SLA research (e.g., Clahsen & Felser, 2006; Gass, 1979, 1984; Ellis & Sagarra, 2011; Jarvis & Pavlenko, 2008; MacWhinney, 2005; Ringbom, 20 07; Tolentino & Tokowicz, 2011; White, 2000). For example, scholars tak ing a generative approach based on the linguistic theory of universal grammar (UG) (Chomsky, 1981) have had a longstanding interest in the interaction between UG and L1 transfer (for a review, see White, 2000). Recently, not only UG researchers but also us age - based researchers have begun to emphasize the role of L1 prior knowledge in implicit and explicit L2 learning (Brooks & Kempe, 2013; Godfroid, 2016; Leung & Williams, 2014; Onnis & Thiessen, 2013; Tolentino & Tokowicz, 2014; Williams, 2005; Williams & Kuribara, 2008). Usage - based researchers have increasingly acknowledged the importance of the L1 effect on L2 learning because L2 learners, especially in the case of natural L2 learning , bring prior L1 knowledge to their L2 learning, which may yield L1 - bas ed habits (e.g., Ellis & Sagarra, domain - specific, prior L1 knowledge could have an impact on learning that relies on domain - general learning mechanisms. The rol e of L1 in L2 syntactic acquisition is a fruitful avenue for research because it could open a productive dialogue between the usage - based and generative 13 approaches (VanPatten & Rothman, 2015; Williams & Kuribara, 2008). This exploration has the potential t o make a significant contribution to SLA theory because it helps researches to gain a better understanding of how and to what extent domain - general and domain - specific learni ng mechanisms can account for L2 learning. From the usage - based perspective, Ell is, Sagarra, and their colleagues (Cintrón - Valentín & Ellis, 2015; Ellis & Sagarra, 2010, 2011; Ellis, Hafeez, Martin, Chen, Boland, & Sagarra, 2014) viewed learned attention as a form of L1 transfer in L2 processing and acquisition. The central idea of th eir argument is that a prior L1 experience can direct L2 attentional processing, either positively or negatively. On one hand, L2 learners would pay more attention to familia r cues based on their prior L1 experience, processing them more fluently and autom atically. On the other hand, they would block their attention to unfamiliar and foreign cues, resulting in L2 processing biases. For example, Ellis and Sagarra (2010) Experim ent 2 and Ellis and Sagarra Experiments 2 and 3 consistently demonstrated how L1 b ackgrounds have led to attentional biases in L2 processing. They showed that when equal amounts of adverbial and verbal cues were presented in an input, Chinese speakers, who se L1 lacked morphological markings, experienced difficulties in acquiring verbal inflectional cues compared with Russian and Spanish speakers, whose respective L1s are rich in morphological markings. These findings indicated that learned attention in L1 m ay exert a major influence on subsequent L2 processing and acquisition. Researc hers have shown that L1 grammatical concepts that are relevant to L 1 form - meaning mappings can be transferred in L2 implicit learning (Brooks & Kempe, 2003; Leung & Williams, 2014; Williams, 2005; Williams & Lovatt, 2003). Williams (2005), in a seminal SLA study on learning without awareness, provided initial evidence of L1 influence in implicit learning. Participants whose L1 marked noun genders with determiners tended to be better at 14 learning targeted determiner - noun mappings in a semi - artificial language than participants whose L1 did not have a gender - marking system. Extending Williams (2005), Leung and Williams (2014) probed the L1 influence on the implicit learning of det erminer - noun mapping based on semantic relationships. In their study, native speak ers of English and Chinese participated in three reaction - time experiments that targeted semantic concepts such as animacy, the number of capital letters and strokes, and the long/flat distinction. Four artificial determiners were introduced as target form s: gi , ro , ul , and ne . Gi and ro were used with nearby objects, while ul and ne were used with faraway objects, and this mapping was explicitly described to participants. How ever, participants were not informed about an additional, hidden mapping rule. In particular, the determiners also varied by animacy (i.e., animate vs . inanimate) (Experiment 1), the number of capital letters and strokes (Experiment 2), and the long/flat d istinction (Experiment 3). An interesting crosslinguistic influence was found in E xperiment 3. In Experiment 3, the learning target (i.e., the long/flat distinction) was a concept exemplified in the Chinese classifier system. The Chinese speakers, whose L1 encoded this regularity in their grammar, slowed down their responses to article usage based on distance (near/far) when the hidden long/flat rule was violated. However, the English speakers, who did not have this distinction in their L1, did not show suc h sensitivity to the long/flat violation. Leung and Williams interpreted this as e vidence for the role of L1 constraints in the implicit learning of semantics. More generally, their results evidenced the importance of the L1 effect in implicit learning. Th ere is evidence for L1 transfer in the acquisition of semantics (Leung & Williams, 2014) and morphology ( Cintrón - Valentín & Ellis, 2015, Ellis, Hafeez, Martin, Chen, Boland & Sagarra, 2014; Ellis & Sagarra, 2011; Sagarra & Ellis, 2013; Williams, 2005 ) but not in other areas such as phonology, syntax, and pragmatics. To help close this g ap, in the present study, I investigate the effect of prior L1 syntactic 15 knowledge on the incidental learning of L2 syntax. To demonstrate L1 syntactic transfer, many resea rchers targeted word order and head - direction (Onnis & Thiessen, 2013; Williams & Kuribara, 2008; Williams, 2011). Williams and Kuribara (2008 ) used word order and head direction to investigate the characteristics of the initial stage of L2 UG - guided and L 2 frequency - guided learning. In their study, English native speakers, whose L1 is a head - initial language, were exposed to Japlish, a semi - artificial language consisting of English vocabulary and Japanese syntax. The participants completed a meaning - focuse d task without any instruction or feedback. The Japlish syntactic structures in th e exposure phase comprised canonical SOV sentences primarily, along with some scrambled OSV sentences that involved optional movement. After exposure, participants took a sur prise GJT on new sentences containing canonical and scrambled word orders. There w as evidence for the generalizability of knowledge of canonical patterns without awareness, but the picture regarding scrambled patterns was not so clear. Although a subset of participants who accepted trained scrambled structures also accepted unscrambled complex sentences, they failed to accept untrained scrambled simple sentences. This performance did not support the clustering effects that parameter resetting should entail. If the head direction parameter had been reset under the guide of a UG, all the u nderlying related structures should have been acquired, including scrambled structures. Based on these results, researchers concluded that the initial stage of adult L2 learn ing is not UG - guided. Within the field of cognitive psychology, Onnis and Thiess en (2013) documented English - induced biases on their statistical learning of sequential information. Statistical learning an environment. Implicit learn ing and statistical l earning share many similarities, in terms of 16 research area (i.e., how humans unconsciously gain information from the input) and research methodology (i.e., the use of the artificial grammar paradigm), and thus the two approaches are so metimes equated (Conw ay & Christiansen, 2006; Kuhn & Dienes, 2008; Monaghan, Schoetensack, & Rebuschat, 2019). Drawing on the fact that Korean is head - final and English is head - initial, Onnis and Thiessen (2013) conducted a corpus analysis on Korean and En glish corpora, which confirmed their prediction that the opposite head direction tendencies in the parsing behavior. With this crosslinguistic evidence establis hed, the authors comp ared how Korean and English native speakers sequence complex linguistic and non - linguistic stimuli. The results suggested that participants showed a parsing preference for patterns that matched t heir L1 word - order patterns in other wor ds, statistical learn characteristics. A question that remains is to what extent these findings from the statistical learning literature also apply to more natural linguistic materials (for cautionary results, see Kim & Godfroid, 2019 ). To expand our understanding of these issues, in the present study I compare the performance of English and Korean speakers in their learning of a semi - artificial language consisting of Korean word order and English vocabulary. 2 . 1 . 3 Research questions In this chapter, I will present the answers to the following research questions: RQ1 (i ncidental exposure ) Does incidental exposure facilitate the acquisition of syntactic knowledge? Do the experimental group and the control group di ffer in their perform ance o n the offline knowledge measure? RQ2 (prior L1 knowledge) Does L1 affect syntactic knowledge development under incidental exposure conditions? Do the 17 English experimental group and the Korean experimental group differ in their performance o n the offline knowledge measure? RQ3 (t ime ) n the offline knowledge measure change over time? RQ 4 (a wareness reflected in retrospective verbal reports) 4 - 1. Is there evidence of learning without awareness? Can unaware learners acquire syntactic knowledge from incidental exposure? 4 - 2. Do aware and unaware, English and Korean subgroups differ in their performance on the offline knowledge measure ? RQ 5 (awarenes s reflected in source attributions) 5 - 1. What is the nature of the acquired knowledge? To what extent is it implicit or explicit? 5 - 2. Do the English and Korean speakers differ in the development of implicit and explicit knowledge? 2 . 2 Method s 2 . 2 . 1 Participants In this study, I compared two language groups: English speakers, whose L1 has a head - initial structure, and Korean speakers, whose L1 has head - final word order. English speakers ( n = 40, M age = 23.71, SD = 3.90) and Korean speakers ( n = 40, M age = 25.25, SD = 4.42) were recruited from three universities: o ne large Mid - western university in the United States and two d ents in a university. 18 Each language group was subdivided into an experimental and a control group. This resulted in four subgroups: (a) an English experimental group (EE, n = 25), (b) an English control group (EC, n = 15), (c) a Korean experimental group (KE, n = 25), and (d) a Korean control group (KC, n = 15). The experimental groups, EE and KE, read the novel The Mysterious Affair at Styles (Christie, 1920) in a semi - artificial language that consisted of English v ocabulary and Korean word order. The co ntrol groups, EC and KC, read the English version of the novel; that is, the same novel with English vocabulary and English word order. The use of control groups was important because they yielded baseline data for th e GJT scores. As Hamrick and Sachs (201 7) rightly pointed out, the use of statistical chance as a comparison in incidental biases or to learn during the test ossibility that using chance as a baseline will pose a threat to the internal validity of the study. In this study, grammaticality judgments were employed as the baseline, which allowed a more accurate assessment of learning effects. The English speakers had no background in Korean or any oth er head - final languages, such as Japanese or Turkish. Further, the experimental groups did not have any knowledge of German, a language that was use d to create violation blocks in the tes ting phase during the exposure task (see below). The Korean speakers were unbalanced Korean - English bilinguals who were highly proficient in English (see Table 1). At the time of the research, 19 Korean speakers resid ed in the United States, whereas the re maining 31 Korean participants resided in South Korea. They had a mean iBT TOEFL score of 111.67 ( SD = 4.65) and a mean length of residence in an English - speaking country of 7.89 years ( SD = 3.27). All Korean speakers reported that their L1 was Korean and their L2 was English; however, they also reported that 19 they had no difficulty in reading English novels or communicating in English. Table 2.1 unds and proficiency levels. There was no difference between the KE and KC groups in their age of exposure ( U = 180.00, p = .847, r = 0.01), length of residence ( U = 174.50, p = .720 r = 0.05), TOEFL score ( U = 127.00, p = .474 r = 0.16), and self - rated pr oficiency ( U = 170.00, p = .639 , r = 0. 10). The participants were recruited by flyers, web postings, and word of mouth. They were compensated with $30 at the end of the experiment. Table 2.1. KE ( n = 25) KC ( n = 15) M S D M SD Age at testing 25.32 4.61 25.13 4.24 Age of exposure 5.80 1.76 5.93 1.67 Length of residence (years) 7.74 3.26 8.12 3.38 TOEFL score 112.22 4.99 110.69 3.97 Self - evaluation of proficiency Total 34.96 1.79 34.53 2.50 Listening 8.96 0.68 8.80 1.01 Reading 8.88 0.78 8.60 0.78 Speaking 8.56 0.82 8.47 1.13 Writing 8.56 0.87 8.67 0.98 2 . 2 . 2 Targeted semi - artificial language: Koreanish The present study adopted a semi - artificial language paradigm , in which vocabulary from 20 the participant wa s rearranged following the patterns in the target language . Specifically, the semi - artificial language in this study consisted of English vocabulary and Korean syntax. To create this hybrid language, which I will call Koreanish, I rear ranged the English words in the English version of the Agatha Christie novel The Mysterious Affair at Styles the words were familiar, but the syntax was an unkn own system that posed comprehension difficulty. The word orders in Korean and E nglish often look like mirror images of each other because the two languages order heads and complements differently. Syntactically, English is a head - initial language . T he he ads of phrases precede their complements, thus forming a right - branching structur e. Korean is a head - final language . H ead s follow complements, thus forming a left - and its complements are naturally located to because the preposition (head) comes at the end of the phrase; that is, Korean is postpositional. Based on Korean word order, I used four syntactic rul - finality to generate the sentence stimuli. First, in the verb phrase (VP), the order of elements is complement - verb. Second, in the postpositional p hrase (PP), the order of elements is noun phrase (NP) - postposition. Third, in the complementizer phrase (CP), the order of elements is clause - complementizer (e.g., relative pronoun, that , whether ). Fourth, in the NP, the order of elements is complement (e. g., appositive clause, relative clause) - noun. The following sentences are an actu al example from the novel and show how an English sentence (1), (3) can be rearranged according to Koreanish rules, as shown in (2), (4), respectively. 21 (1) English: A vague suspicion of everyone and everything filled my mind. (2) Korean: Everyone and e verything of a vague suspicion my mind filled. Everyone and everything of of s the head of the phrase and comes in final position . Second, the Third, the order. , First, in rela tive complementizer, is the head of the CP and should be placed at the end of the phrase . by preceding rather than following it. This results in the relative clause NP sequence the complex NP Hence , the full sentence reads I a nurse is who a cousin hav In this manner , Koreanish consistently features the head, 22 the main elemen t of the phrase, at the final position. 2 . 2 . 3 Materials The exposure task: Novel reading. The exposure text for training was the novel The Mysterious Affair at Styles by Agatha Christie. This novel provided incidental exposure to the semi - artificial word order system, Koreanish, to participants. The participants read the first two chapters, Chapters 1 and 2 (approximately 6,500 words), of The Mysterious Affair at Style s . The sam e novel was used in a series of eye - tracking studies that examined various aspects of English - (Cop, Dirix, Drieghe, & Duyck, 2017; Cop, Dirix, Van Assche, Drieghe, & Duyck, 2017; Cop, Drieghe, & Duyck, 2015; Cop, Keuleers, Drieghe, & Duyck, 2015). Cop and colleagues noted that this novel was selected based on its appropriate difficulty level for college level L2 English speakers (Flesch Reading Ease = 81.3, SMOG grade = 7.4) 1 as well a s its simi larity in word frequency distribution to natural language based on the Subtlex database (Van Heuven, Mandera, Keuleers, & Brysbaert, 2014) In other words, this novel is a suitable reading material for a L2 experiment because it is easy to read an d reflects real - life, natural language use. The participants were told that they were going to read the first two chapters of a detective novel (92 screens), and that for some of the participants the sentences would be presented with the words scrambled. The exper imental groups read the Koreanish version of the novel, consisting of English sentences with Korean word order, while the control groups read the 1 Flesch Reading Ease gives a scale between 0 (difficult to read) and 100 (easy to read); highe r scores reflect greater readability. The SMOG grade is an estimation of how many years of educations are required to understand the text. 23 main character s to help them understand the story. The participants were informed that there would be a practice session at the beginning and a short break approximately 15 screens later (6 breaks in total). At every other break, the participants were asked to answer co mprehension check questions. The purpose of the comprehension test was to keep participants engaged while reading the scrambled sentences for about an hour. The se comprehension test s , which contained eight simple true or false statements regardin g plot - spe cific information, w ere administered three times throughout the reading. The total number of questions was 24. One point was assigned for correct answers, and the total score of overall comprehension was 24. The overall reliability coefficients ( alpha) were 0.78 for the EE group and 0.71 for the KE group. Three counterbalanced versions of the exposure text were created to control for lexical and topical confounds (Table 2.2). 24 Table 2.2 The counterbalanced versions of the expos ure task Phase S yntax N of sentences Sentence ID Version A Version B Version C Practice English 44 544 583, 1 4 564 583, 1 24 1 44 Training Koreanish 479 5 483 25 503 45 523 Testing: Control_pre Koreanish 20 484 503 504 523 524 543 T esting: Violation German 20 504 523 524 543 544 563 Testing: Control_post Koreanish 20 524 543 544 563 564 583 25 The testing task: GJT . with source attributions was implemented after the exposure task. Importantly, the participants were not informed ahead of time that they would be tested. After reading the novel, they were t old that the scrambled order was not arbitrary but followed a complex system. They were asked to judge the grammaticality of the new sentences based on the system presented in the novel (see Appendix C). No feedback was provided regarding their answer. In the GJT, 80 new sentences were presented to the participants, distributed evenly between 40 grammatical and 40 ungrammatical items (Table 2.3). Five grammatical Koreanish patterns were created, including two simple and three complex patterns: (a) simpl e sentence: SOV; (b) simple sentence with postpositional phrase: SPP( postposition )V, (c) com plex sentence with a that - clause: S[SOV that ]V, (d) complex sentence with a relative clause: S[OV relative pronoun ]OV; (e) complex sentence with a subordinate clause: SOV subordinator , SOV. For each grammatical structure, two ungrammatical patterns were cr eated as direct counterparts: (a) *SVO and (b) *VSO for the SOV pattern; (c) *S[ that SOV ]V and (d) *SV[SOV that ] for th e S[SOVthat]V pattern; (e) *S[PP preposition ]V and (f) *SV[PP postposition ] for the S[PP postposition ]V pattern; (g) *S[ relative pronoun OV ]OV and (h)*SO[OV relative pronoun ]V for the S[OVrelative pronoun]OV pattern; (i) * subordinator SOV, SOV and (j) *SOV, S OV subordinator for the SOVsubordinator, SOV pattern. In the case of the complex sentences, one ungrammatical pattern contained the claus e - level error 26 (within the dependent clause) and the other ungrammatical pattern included the sentence - level error (with in the independent clause). First, the SOV pattern is a simple sentence with basic Koreanish word order and reflects the VP rule. Both *SVO and *VSO contrast with the SOV pattern. The *SVO pattern follows English word order, so it allowed me to test whet her the participants knew that English - like word order was not possible in Koreanish. The *VSO pattern also enabled me to assess whether participants could reject the cases that did not conform to the VP rule. Second, the S[PP( postposition )]V pattern rep resents a simple sentence that features a postpositional phrase. It tests knowledge of the PP and the VP rules. The ungrammatical pattern s for this structure are *S[PP( preposition )]V and *SV[PP( postposition )]. The *S[PP( pre position )]V pattern tests the PP rule by allowing me to determine whether participants knew that the apposition, the head of the PP, should come at the end of the phrase. The *S[PP( preposition )]V pattern allowed me to test the VP rule that the verb should come in final position in a ve rb phrase. Third, the S[SOV that ]V pattern is a complex sentence with a that - clause involving the CP and VP rule. The ungrammatical pattern s for this structure are *S[ that SOV]V and *SV[SOV that ]. *S[ that SOV]V assesses the CP rule applied at the clause leve l. It tests whether participants knew that the complementizer that should come at the end of the complement clause. *SV[SOV that ] measures th e VP rule at the sentence level, by testing whether participants kneow that the verb should be in sentence - final pos ition. Fourth, the S[OV relative pronoun ]OV pattern is a complex sentence that features a relative clause, which reflects the CP and the NP rules. The ungrammatical patterns for this structure are *S[ relative pronoun OV]OV and *SO[OV relative pronoun ]V. The *S[ relative 27 pronoun OV]OV pattern allowed me to measure knowledge of the CP rule at the clause level. It tests whether participants knew tha t the relative pronoun, the head of the CP, should follow complements and come at the final position in the relative clause. The *SO[OV relative pronoun ]V assesses the NP rule at the sentence level. It assesses whether participants knew that the relative cl ause, as a complement, should precede the head noun it modifies. Fifth, the SOV subor d inator , SOV pattern is a comp lex sentence with a subordinate clause. It involves the CP and VP rules. The ungrammatical patterns for this structure are * subordinator SOV, SOV and the *SOV, SOV subordinator . The * subordinator SOV, SOV pattern tests the CP rule at the clause level. Within the subordinate clause, the subordinator is the head that should follow complements. The *SOV, SOV subordinator pattern assesses the knowledg e of the VP rule at the sentence level. It enabled me to see whether participants knew that the verb in the main clause should come at the end of the sentence as the head of the VP. In addition to grammaticality judgments, participan ts were asked to indi cate the basis of their decision: guess , intuition , recollection , or rule knowledge . This source ratings were useful ( Dienes, 2004; Dienes & Scott, 2005 ; Rebuschat, 2013; Sp inner & Gass, 2019). Participants were informed to as well have flipped a coin. If a participant had some confidence in their decision and knew, to some degree , that the judgment was correct but could not describe why, they were told to opt for was based on the memory of specific sentences (o r parts of the senten ces) that they read in the followed a verbalizable rule when making their decision. Eight lists of the GJT were created, 28 counterbalanced for the g rammaticality and the presentation order of the stimuli. The reliability = 0.84 on the immediate test to =0.89 on the delayed test. 29 Table 2.3 Counterbalanced grammatical and ungrammatic al stimuli used in the testing set Sentence type Grammatical Pattern Ungrammatical pattern Target rule Simple, basic SOV ( k = 8) e.g., Joon the paper revised. *SVO ( k = 4) e.g., Joon revised the paper. VP rule *VSO ( k = 4) e.g., Revised Joon the paper. VP rule Simple, postposition S[PP postposition ]V ( k = 8) e.g., Max the classroom in studied. *S[ preposition PP]V ( k = 4) e.g ., Max in the classroom studied. PP rule *SV[PP postposition ] ( k = 4) e.g., Max studied the classroom in. VP rule Complex, that - clause S[SOV that ]V ( k = 8) e.g., The dean he his salary donated that lied. * S[ that SOV]V ( k = 4) e.g., The dean that h e his salary donated lied. CP rule *SV[SOV that ] ( k = 4) e.g., The dean lied he his salary donated that. VP rule Complex, relative clause S[OV relative pronoun ]OV ( k = 8) e.g., Jessie a blue coat wore who the man found. *S[ relative pronoun OV]OV ( k = 4) e.g., Jessie who a blue coat wore the man found. CP rule *SO[ OV relative pronoun ]V ( k = 4) e.g., Jessie the man a blue coat wore who found. NP rule 30 Complex, subordinate clause SOV subordinator , SOV ( k = 8) e.g., We dinn er enjoyed while, the band music played. *subordinator SOV, SOV ( k = 4) e.g., While we dinner enjoyed, the band music played. CP rule *SOV, SOV subordinator ( k =4) e.g., The band music played, we dinner enjoyed while. VP rule 31 Debriefing interv iew . levels of awareness of the targeted Koreanish syntactic patterns (see Appendix D). The interview was carried out over two sessions. At the end of Session 1, the first part of the debri efing participants whether they had noticed anything odd while reading the novel. In addition, I asked them to report whether they had noticed any particular rule or regul arity, to specify when they might have noticed it (i.e., during reading or on the test), and to describe what they believed they had noticed. In Session 2, the second part of the debriefing interview was conducted. I began by asking whether they eve r ind icated recollection of rule knowledge as a source of their grammaticality judgement. If so, they were asked to explain why and what they were thinking. In addition, I asked participants how they attempted to read the sentences with the words scrambled and whether they had tried to search for a pattern in the scrambled word order while reading. Moreover, I asked them to reflect specifically on the placement of words within the sentences and to recall any specific rule or regularity in an attempt to tap into levels. Finally, I asked them questions about their usual reading experiences. 2 . 2 . 4 Procedure The experiment was conducted in two sessions, Session 1 (90 minutes) and Session 2 (30 minutes). Session 1, which was carri ed out in a quiet, dimly - illuminated study room, included the exposure phase, the immediate testing phase, and the first part of the debriefing interview (short version). Session 2 comprised the delayed testing phase, the second part of the debriefing inte rview (long version), and the background questionnaires (see Figure 2.1). In Session 1, the participants first signed the consent f orm and then filled out a language 32 background questionnaire. After being told that they were going to read two chapters fro m a detective novel, a brief description of the main characters was provided. They were informed that there would be several comprehe nsion questions after reading each part (Appendix B); however, they were not told that there would be a GJT about the word order when they completed reading the novel. They silently read the first two chapters of The Mysterious Affair at Styles on a laptop screen while their eye movements were recorded by the Eyelink Portable Duo eye - tracking system (SR Research, Ottawa, Ontar io, Canada). The experimental groups read the novel in Koreanish and the control groups read the English version of the novel. After they finished reading the novel, a surprise GJT with source attributions was administered for immediate testing. This test was followed by the first part of the debriefing interview regarding awareness of the violation block in the reading (please see Tabl e 2.2, and the reading time data will be presented in chapter 3) and Koreanish word order. Figure 2.1 Procedure in the experiment Session 1 Background questionnaire Exposure task: reading a novel (with eye - movement recording) Testing task: Immediate GJT Debriefing interview (1st part) Session 2 (2weeks later) Testing task: Delayed GJT Debriefing interview (2nd part) Surveys on personality, impulsiveness, and cognitive style 33 Session 2 took place two weeks following Session 1 for delayed testing. To each participant, the researcher sent an e - mail containing the links to the delayed test, the second part of the debriefing interview, and the surveys regarding personal ity (Big Fiv e personality test, De Young, Quility, & Peterson, 2007), impulsiveness (the UPPS, Whiteside & Lynam, 2001), and cognitive style (Rational Experimental Inventory, Pacini & Epstein, 1999). The participants were asked to complete the test within three days. 2 . 2 . 5 Analysis All the data from the testing task ( grammaticality judgements) were entered into the Statistical Package for Social Sciences, version 25. Two participants, one from the EE group and one from the KE group, were excluded from t he analyses , leaving 24 EE and 24 KE. The excluded participants stated in their retrospective verbal reports that they intentionally searched for word order rules during reading, in violation of the incidental nature of the study. For the delayed GJT, anal yse s were conducted on 72 participants (23 EE, 14 EC, 22 KE, 13 KC) since 6 participants (1 EE, 1 EC, 2 KE, 2 KC) did not participate in Session 2. - prime ( d ) scores a sensitivity index that reflects participa nts Th e score is known to be a more accurate measure than raw accuracy in that it takes response bias into account based on hit s and false alarm rates (MacMillan & Creelman, 2005) ; t herefore, where appropriate, scores will be reported 2 . A score of zero is equivalent to chance performance and a of four is interpreted as near - perfect performance. That is, a positive 2 With hit rates and false alarms of 0 and 1, I did a standard correction to compute the scores. Given the maximum number of false alarms of 40, the extreme values (0 and 1) were strategically replaced with 1/(2 × 40) = 0.0125 and . 34 score indicate s above chance - level performance, whereas a negative indicates below chance - level performance. In this study, a higher ility to discriminate target - like word order in Koreanish. To investigate the effect of prior L1 knowledge on the incidental acquis ition of syntactic knowledge over time (RQs 1, 2, and 3), I ran a 2 × 2 × 2 mixed - design ANOVA on scores , with Time (Immediate, Delayed) as the within - subject factor and with Condition (Experimental, Control) and L1 background (English, Korean) as the between - subject factors. To confirm the - 1), I ran Mann Whitney U tests to compare the respective scores of the unaware group and the control group. Additionally, I ran one - sample t - d with 0 (i.e., chance) as the test value. To investigate the effect of ve rbal aware ness on the incidental acquisition of syntactic knowledge, I conducted a mixed - design ANOVA for each experimental group separately. For the EE group, I ran a 2 × 2 mixed - design ANOVA with Time (Immediate, Delayed) and Awareness (VP - aware, Unaware ) ; for the KE group, I performed a 2 × 3 mixed - design ANOVA with Time (Immediate, Delayed) and Awareness (L1 - aware, VP - aware, Unaware) . To probe the nature of the acquired syntactic knowledge (RQ 4 - 2), I ran one - sample t - tests on mean accuracy (%) by sourc e attributions, with 0.5 (i.e., chance) as the test value. Further, to investigate whether two L1 groups differed in terms of implicit and explicit knowledge devel opment (RQ 4 - 2), I ran a 2 × 2 × 2 mixed - design ANOVA on mean accuracy with Attributions (Imp licit, Explicit), L1 (English, Korean), and Time (Immediate, Delayed). To clarify the nature of any significant interactions, I followed up on the main model by us ing stepdown ANOVAs and compariso ns of simple effects . An alpha level of 0 .05 was used for al l statistical tests. 35 2 . 3 Results 2 . 3 . 1 Comprehension t est The overall comprehension scores of the EE and KE groups were 18.17 (SD = 2.51) and 18.13 (SD = 3.17) , respectively, which indicated that their comprehension level of the scrambled text was acce significant [t(46) = 0.31, p = .759, d = 0.09]. The overall comprehension scores for the control groups, the EC and KC groups, were 23.13 ( SD = 0.96) and 22.80 ( SD = 1.05), resp ectively. 2 . 3 . 2 Overall GJT p erformanc e Descriptive statistics of GJT accuracy scores (%) for the EE, EC, KE, and KC groups are presented in Table 2.4. Table 2.4 Mean accuracy (%) of grammaticality judgments for English and Korean speakers Immediate GJT Delayed GJT n M (%) SD 95% CI n M (%) SD 95% CI L1: English Experimental 24 56.93 9.80 [52.79, 61.07] 23 58.91 8.18 [55.37, 62.45] Control 15 45.08 4.64 [42.51, 47.65] 14 45.54 4.59 [42.88, 48.19] L1: Korean Experimental 24 69.64 11.70 [64.70, 7 4.58] 22 77.61 11.81 [72.38, 82.85] Control 15 49.04 9.44 [43.81, 54.27] 13 48.75 7.87 [43.99, 53.51] Note : Accuracy scores for the EE, EC, KE, KC groups were normally distributed according to the Shapiro - Wilk test ( p > .05) y was moderate [Immediate: M = 56.93, SD = 9.80; Delayed: M = 36 58.91, SD = 8.18] but exceeded a baseline of 50% [Immediate: t(23) = 3.46, p = .002, d = 0.71; Delayed: t was large [Immediate: M = 69.64, SD = 11.70; Delayed: M = 77.63, SD = 11.81] and was significantly greater than chance level [Immediate: t(23) = 8.22, p < .001, d = 1.68; Delayed: t(21) = 10.98, p < .001, d = 2.33]. The difference between the EE and the E C groups [Immediate: t (35.0 5) = 5.079, p < .001, d = 1.55; Delayed: t (34.862) = 6.365, p < .001, d = 2.02] and between the KE and the KC groups [Immediate: t(37) = 5.74, p < .001, d = 1.94; Delayed: t(33) = 7.82, p < .001, d = 2.88] was significant. Desc riptive statistics of d sc ores are reported in Table 2.5 , and the scores by group are depicted in Figure 2.2 . The difference between the EE and the EC groups was significant [I mmediate: U = 50.00, p < .001, r = 0.60; D elayed: U = 28.50, p < .001, r = 0.68 ], and the EE scores across the testing sessions [I mmediate: M = 0.37, SD = 0.74; D elayed: M = 0.54, SD = 0.62 ] were greater than 0 (i.e. chance) [I mmediate: t ( 23 ) = 2.43, p = .023, d = 0.50; D elayed: t ( 22 ) = 4.15, p < .001, d = 0.87 ) . The difference between the KE and KC groups w as also significant [immediate: U = 32.50, p < .001, r = 0.68; delayed: U = 1.00, p < .001, r = 0.82 ), and the KE (immediate: M = 1.13, SD = 0.81; delayed: M = 1.80, SD scores were als o significantly above 0 [I mmediate: t ( 23 ) = 6.84, p < .001, d = 1.40; D elayed: t ( 21 ) = 7.70, p < .001, d = 1.64 ) . 37 Table 2.5 grammaticality judgments for English and Korean speakers Immediate GJT Delayed GJT n n M SD 95% CI M SD 95% CI L1: English Exper imental 24 0.37 0.74 [0.06, 0.68] 23 0.54 0.62 [0.27, 0.81] Control 15 - 0.65 0.60 [ - 0.98, - 0.31] 14 - 0.57 0.48 [ - 0.85, - 0.30] L1: Korean Experimental 24 1.13 0.81 [0.79, 1.47] 22 1.80 1.10 [1.32, 2.29] Control 15 - 0.17 0.66 [ - 0.53, 0.19] 13 - 0.07 0.42 [ - 0.33, 0.18] Figure 2 . 2 Mean d scores for Korean and English speakers Note: Error bars represent 95% confidence intervals. 38 To investigate whether scores on the GJT differed as a function of Condition ( E xperimental , C ontrol), L1 (English/Korea n), and Time ( I mmediate , D elayed), I ran a 2 (Condition) × 2 (L1) × 2 (Time) mixed - design ANOVA. According to the Shapiro Wilk test, the distribution of scores was normal except for the EC scores. As can be seen in Table 2.6 , this analysis revealed significant main effects of Condition [ F ( 1,70 ) = 70.06, p < .001, = .51 ] , L1 [ F ( 1,70 ) = 27.36, p < .001, = .29 ] , and Time [ F ( 1,70 ) = 11.28 , p = .001, = .14 ] . A Time by L1 interaction was sign ificant [ F ( 1,68 ) = 4.83, p = .031 , = .07 ] , which was qualified by a borderline significant Time by L1 by Condition interaction [ F ( 1,68 ) = 3.84, p = .054, = .05 ] . Figure 2.3 represents the three - way interaction visually . Figure 2.3 Three - way interactions between Condition, Time, and L1 39 Table 2.6 Mixed - design ANOVA for the d scores on GJTs . + p < .06; * p < .05; ** p < .01; *** p < .0001 To follow up on the three - way interaction, I carried out a 2 (Time) × 2 (Condition) mixed - design ANOV A for each L1 group separately . Th e analysis for the L1 English group showed that the main effect of Condition was significant [ F ( 1,35 ) = 40.69, p < .001, = .54 ] , but the main effect of Time [ F ( 1,35 ) = 0 .79, p = .379, = .02 ] and the Time by Condition interaction [ F ( 1,35 ) = 0 .00, p = .986, = .00 ] were not significant. This indicated that incidental exposure through a novel was effective for English speakers to develop syntactic knowledge, but the amount of knowledge remained steady over time. The analysis for t he Korean speakers revealed significant main effects of Time [ F ( 1,33 ) = SS df MS F p power Main effect (within - subject variable) Time 2.56 1 2.56 11.28 .001** .14 .91 Main effect (between - subject variab le) Condition 59.52 1 59.52 70.06 < .001*** .51 1.0 L1 23.24 1 23.24 27.36 < .001*** .29 1.0 2 - way interaction Time x Condition 0.80 1 0.80 3.51 .065 .05 .46 Time x L1 1.10 1 1.10 4.83 .031* .07 .58 Condition x L1 1.62 1 1.62 1.91 .172 .03 .28 3 - way interaction Time x Condition x L1 .87 1 .87 3.84 .054+ .05 .49 40 14.75, p = .001, = .3 1] and Condition [ F ( 1,33 ) = 33.2 1 , p < .001, = .50 ] and, crucially, a significant Time by Condition interaction [ F ( 1,33 ) = 6.95, p = .013, = .17 ] . To understand the nature of this interaction, I ran simple effects of Condition for each time point as well as simple effects of Time for each Condition . First, the effect of Condition was significant for both the immediate [ F ( 1,33 ) = 20.95, p < .001, = .39 ] and the delayed tests [ F ( 1,33 ) = 34.55, p < .001, = .51 ] . Second, there was no effect of Time for the KC group [ F ( 1,33 ) = 0 .06, p = .814, = .00 ], while there was a strong effect for the KE group [ F ( 1,33 ) = 21. 49, p < .001, = .39 ] , which pointed to different amount s time. These results indicated that incidental exposure through a novel was effective for the Korean speakers to develop syntactic know ledge , and , importantly, there was a significant increase only in T o test the effect of L1 (RQ2), I ran additional 2 (L1) × 2 (Condition) factorial ANOVA s for each Time separately. Another two - way interaction was found only in the delayed test [ F ( 1,68 ) = 2.45, p = .042, = .059 ] , and here , I only report the results of simple effects of L1 for each Time . For the immediate test, the significant main effect of L1 [ F ( 1,74 ) = 15.77, p < .001, = .18 ] and the non - significant L1 by Condition interaction [ F ( 1,74 ) = 0 .70, p > .05, = .0 1] suggested that the Korean speakers outperformed the English regardless of condition. In other words, even without being exposed to Koreanish sentences, the Korean speakers were better at judging the grammaticality of Kore anish sentences than the English speakers. However , in the delayed test, there was a significant effect of L1 in the experimental condition [ F ( 1,68 ) = 31.39, p < .001 , = .32 ], but in the control condition, there was no effect of L1 [ F ( 1,68 ) = 2.9 7, p = .089, = .04 ]. This reflected the additional benefit that L1 Korean speakers derived from their native - language grammar for developing long - term syntactic knowledge of a new head - 41 final language . These findings suggest that the training, i .e., incidental exposure through a novel, facilitated the development of new grammatical knowledge for both the English and the Korean speakers (RQ1). The Korean experimental group weeks later, as reflected in the knowledge did not show evidence of enhancement (RQ3). Accordingly, the Korean experimental group outperformed the Engl ish experimental group at the delayed test (RQ2). This diverging pattern bet knowledge led to a clear advantage in the acquisition of Koreanish word order over time. 2 . 3 .3 . Awar eness reflected in retrospective verbal reports and GJT performance Analys is of retrospective verbal reports. The retrospective verbal report data from the debriefing interviews from Sessions 1 and 2 were awareness . The retrospective verbal report data from the debriefing interviews were analyzed to To enhance the inter - rater reliability in the coding, two independent rat ers (the researcher and a trained rater) coded the verbal report data, which was gathered from two debriefing interview sessio ns. The raters identified 1) what type of awareness the participants developed (unaware, VP - aware, L1 - aware), 2) when they became aware of the rules (exposure and testing), and 3) when they reported awareness (Session 1 and Session 2). The reliability coef .99, and for the identification of This analysis enabled me to identify subgroups of pa rticipants based on their reported awareness of the target rules . Note that only the experimental groups who read the Koreanish 42 ve rsion of the novel participated in the debriefing interview. From the verbal report data, I attempted to confirm the internal validity of the incidental learning condition (cf. Leow, 1997, Leow & Hama, 2013) by asking participants whether they tried to sea rch for patterns while they were reading. This inspection led me to exclude two participants one from each L1 group. Their ans wers, presented in Example ( 5 ), clearly showed that they intentionally searched for a pattern while reading, which means that their attention was oriented towards forms. Example ( 5 ) P119 (English participant) ern. But after the first break, I started to notice a pat tern , so I would try to figure it out and unscramble the sentences. I realized that the verb at the end of P304 (Korean participa nt) . were messed up regularly. I did try to find dicat ed that half of the participants 12 of 24 participants became aware of the VP rule. This was a partial awareness of Koreanish word order rules, since participants did not mention any other rules regarding head finality. Based on their awareness of the VP r ule, I categorized them as VP - aware. Among the 12 participants, 10 participants reported awareness of the VP rule at Session 1, during the short version of the interview, and two participants reported awareness at Session 2 (two weeks after Session 1) , dur ing the long version of the interview. All participants said they 43 became aware of the VP rule during the GJT. The other 12 participants did not report awareness of any target rules and were categorized as unaware. Table 2.7 report data n Timing of Awareness Comment Unaware 12 - I thought the words were randomly jumbled. V P - aware 12 Exposure phase I noticed the subject or noun was normally at the beginning of the sentence, and the verb was at the end. Unlike the anal verbal report data (see Table 2.8) exhibited three types of awareness: L1 aware ( n = 9), VP - aware ( n =10), and unaware ( n = 5). Table 2.8 n Timin g of Awareness Comment Unaware 5 - The rule I noticed is to place nouns at the beginning of the sentence. V P - aware 10 Exposure phase I recognized that one of the types is to place the verb in the end of the sentence. L1 - aware 9 Testing phase I've noticed that the order of some sentences was similar to my native language, Korean, at test. First, I found an interesting subgroup, the L1 - aware group, who reported their awareness of the 44 relationship between the target language and their L1 (Korean). These nine participants reported that they became aware that the language used in the novel foll owed Korean word order while they were taking the GJT. It is important that they were unaware of any rules ( n = 7) or only discovered the VP rule ( n = 2) during the exposure phase; it was in the testing phase that they figured out that the Koreanish senten ces in the novel matched their L1. In terms of the timing of reported awareness, among 9 L1 - aware participants, 2 participants reported their L1 awareness after taking the immediate GJT at Session 1; the rest, 7 participants, reported their awareness after taking the delayed GJT at Session 2. Second, 10 participants became VP - aware. They reported noticing the VP rule but did not notice other rules regarding head finality in the Koreanish text. All VP - aware participants reported their VP - awareness at Session 1 after taking the immediate GJT. Lastly, five participants remained verbally unaware and were categorized as unaware. Can learners acquire knowledge without awareness? Table 2.9 presents the d escriptive statistics for GJT performance by participant awar eness as reflected in the retrospective verbal reports. To find evidence of implicit learning, the unaware - prime scores were analyzed us ing one - sample t - tests and Mann U test. Although t he English unaware group ( n = 12) ou tperformed the EC group [I mmediate: U = 25.50, p = .002, d = 1.52; D elayed: U = 12.00, Z = - 3.57, p < .00 1 , d = 2.03 ], their performance on the immediate test w as not significantly above chance [ t ( 11 ) = 1.19, p = .259, d = 0.34 ] . T he Korean unaware group ( n = 5) , on the other hand, outperformed not only the KC group [I mmediate: U = 5.50, Z = - 2.79, p = .005, d = 1.60; D el ayed: U = .00, Z = - 2.94, p = .003, d = 2. 04 ] but also the English unaware group [I mmediate: U = 10.00, Z = - 2.11, p = .035, d = 1.19; delayed: U = .00, Z = - 2.87, p = .004, d = 2.21 ]. This finding indicated that the Korean unaware 45 group was better at learning without awareness than the English un aware group under incidental exposure. Table 2.9 n Accuracy (%) M SD 95% CI M SD 95% CI L1: English Unaware Immediate 12 54.27 7.32 [49.62, 58.92] 0 .21 0.62 [ - 0.18, 0.61] Delayed 11 58.07 8.22 [52.54, 63.59] 0.43 0.44 [0.14, 0.72] V P - aware Immediate 12 59.58 11.49 [52.29, 66.88] 0.53 0.85 [ - 0.01, 1.06] Delayed 12 59.69 8.43 [54.33, 65.04] 0.64 0.76 [0.16, 1.12] L1: Korean Unaware Immediate 5 6 6.25 9.84 [54.03, 78.47] 1.02 0.66 [0.20, 1.84] Delayed 4 72.19 1.20 [70.28, 74.09] 1.20 0.08 [1.07, 1.33] V P - aware Immediate 10 65.63 11.12 [57.67, 73.58] 0.80 0.79 [0.24, 1.36] Delayed 9 70.44 7.44 [64.69, 76.14] 1.17 0.49 [0.79, 1.54] L1 - aware Im mediate 9 75.97 11.56 [67.09, 84.85] 1.56 0.79 [0.95, 2.17] Delayed 9 87.22 11.56 [78.34, 96.10] 2.70 1.17 [1.80, 3.60] Further, considering that the VP - aware group reported awareness only regarding the VP rule, I tested their performance on other rul es, for which they did not mention in their verbal reports. The relative clause sentence, one of the target structures which involved the NP and CP rules, was a suitable structure to test the acquisition of the CP and NP rules outside of verbal 46 awareness. The KE group [Immediate: M = 0.40, SD = 0.75, 95% CI = 0.09, 0.72; Delayed: M =1.15 , SD = 1.93, 95% CI = 0.29, 2.01] outperf ormed the KC group [Immediate: M = - 0.24, SD = 0.92, 95% CI = - 0.75, 0.27] [Immediate: U = 112.50, p = .049, r = 0.31; Delayed: U = 55.00, p =.003, r = 0.51], which reflected their learning of complex structure for which they did not verbalize a rule. In M = 0.06, SD = 1.02, 95% CI = - 0.37, 0.49; Delayed: M = - 0.29, SD = 0.75, 95% CI = - 0.61, 0.04] d scores [Immediate: M = - 0.09, SD = - 0.79, 95% CI = - 0.53, 0.35; Delayed: M = - 0.50, SD = 0.88 , 95% CI = - 1.00, 0.01] [Immediate: U = 173.50 p = .849, r = 0.06; Delayed: U = 149.00, p = .701, r = 0.06 ]. T aken together, the Korean speakers showed stronger evidence of learning without awareness than the E nglish speakers. Relationship between awareness reflected in retrospective verbal reports, L1, and GJT performance. To answer Research Question 4 - 2 , which investigates the relationship between awareness based on retrospective verbal reports and grammatical knowledge development under incidental learning conditions, I carried out a mixed - design ANOVA with Time as a within - subject variable and Awarenes s as a between - subject variable on the scores for each L1 group separately. The main effect and interactions with Awareness are of interest here . The Kolmogorov - scores were normally distributed ( p > .05). For the EE group, a 2 × 2 mixed - design ANOVA with Awareness (V P - awar e, Unaware) and Time (Immediate, Delayed) was carried out. The analysis showed that t he main effect of Awareness [ F ( 1,21 ) = 0 .6 8 , p = .420, = .0 3] and the main effect of Time [ F ( 1,21 ) = 0 .44, p = .514, = .02 ] were not significant. The Awareness by Time interaction was also not significant [ F ( 1,21 ) = 0 .0 2 , p = .895, = .00 ]. This indicated that for English speakers, there 47 was no evidence showing that awareness affected grammatical knowledge development ove r time. For the KE group, a 3 × 2 mixed - design ANOVA with Awareness (L1 - aware, VF aware, Unaware) and Time ( I mmediate, D elayed) was performed. The analysis reveale d significant main effects of Awareness [ F ( 2,19 ) = 6.95, p = .005, = .42 ] and Time [ F ( 1,19 ) = 18.05, p < .001, = .49 ] . The interaction between Time and Awareness was not significant [ F ( 2,19 ) = 2.83, p = .084, = .23 ] . Post difference between the L1 - aware an d the u naware groups (mean difference = 1.13, 95% CI = 0 .06, 2.21, p = .038) as well as the L1 - aware and the VF - aware groups (mean difference = 1.14, 95% CI = 0 .30, 1.98, p = .007). This indicated that the L1 - aware group excelled in the GJT across the testing sessions, which demonstrated that L1 - awareness boosted test performance across time points. To delve into the special role of conscious L1 transfer, I recategorized the Korean expe rimental group into the L1 - aware and L1 - unaware groups. The L1 - unaware group combined participants from the VF - aware and the Unaware groups into a single category. A 2 × 2 mixed - design ANOVA with Time (Immediate, Delayed) and L1 - awareness (L1 - aware, L1 - una ware) revealed significant main effects of Time [ F (1,20) = 18.99, p < .001, = .23] and L1 - awareness [ F (1,20) = 14.62, p = .001, = .42] and a significant Time by L1 - Awareness interaction [ F (1,20) = 5.93, p = .024, = .23]. I then analyzed t he two - way interaction (Figure 2.4) by examining the simple effec ts of Time for each awareness group. There was no effect of Time in the L1 - unaware group [ F (1,20) = 3.21, p = .088, = .14], while there was in the L1 - aware group [ F (1,20) = 21.71, p < . 001, = .52]. The results indicated that the L1 - awar e group improved significantly in the two - week time period following the training intervention, which 48 suggested that L1 - awareness led to improved knowledge consolidation. Figure 2.4 Two - way interaction between Time and L1 - awareness for the KE group 2 . 3 .4 Awareness reflected in source attributions and GJT performance Is acquired knowledge implicit or explicit? The E proportions and accuracy (%) across source attributi ons are presented in Table 2.10 . For the analysis of source attributi ons, accuracy scores were used instead of scores because most of the participants did not have complete data for sixteen cells (source x grammaticality x correctness) to get the inform ation (hits and false alarm rates) necessary to calculate scores. Therefore, mean accuracy scores by source attributions will be used in all subsequent analyses. Also, the not attribute their decision on recollection or rule without being exposed to the target language. I combined Guess and Intuition attributions to calculate the total proportion of responses 49 based on unconscious, implicit knowledge; I also added Recollection and Rule knowledge attributions to genera te the t otal proportion of responses based on conscious, explicit knowledge. If a participant showed above chance performance even when they judged the grammaticality of a given sentence based on Guess or Intuition, this was considered evidence of unconsci ous, imp licit knowledge (Dienes & Scott, 2005). Table 2.11 and Figure 2.5 present the descriptive implicit knowledge on the immediate test [ M = 53.79, SD = 11.35] was not significantly above chance [ t (23) = 1.64, p =.115, d = 0.33] but on the delayed test [ M = 61.59, SD = 17.70] it was significantly above chance [ t (21) = 3.07, p = .006, d = 0.66]. Unlike the English speakers, the ribution s [Immediate: M = 61.83, SD = 19.00; Delayed: M = 66.80, SD = 21.41] were significantly above chance for both the immediate [ t (23) = 3.05, p = .006, d = 0.62] and delayed [ t (19) = 3.51, p = .002, d = 0.78] tests. In case of the explicit knowledge, both the across the two testing sessions. Therefore, the results indicated that the Korean experimental group exhibited both implicit and explicit knowledge. Meanwhile, the English experi mental g , as there was only evidence of implicit knowledge in the delayed test. 50 Table 2.10 Proportion (%) Accuracy (%) M SD M SD 95% CI L1: English Guess Immediate 16.51 17.11 59.03 20.20 [49.58, 68.48] Delayed 24.67 21.62 55.96 22.61 [45.07, 66.86] Intuition Immediate 45.47 25.36 54.09 13.98 [48.18, 59.99] Delayed 35.92 23.78 61.17 16.77 [53.54, 68.81] Recollection Imm ediate 12.97 11.41 61.5 0 21.65 [51.42, 71.68] Delayed 13.26 20.93 73.56 18.92 [63.83, 83.29] Rule knowledge Immediate 25.05 27.50 59.90 24.18 [49.18, 70.62] Delayed 26.14 34.12 61.33 33.94 [44.45, 78.21] L1: Korean Guess Immediate 23.19 27.21 58.3 8 28.80 [45.62, 71.15] Delayed 26.22 33.60 68.46 25.65 [54.26, 82.66] Intuition Immediate 28.39 24.86 60.74 24.34 [49.95, 71.53] Delayed 32.35 33.57 66.77 25.24 [54.22, 79.32] Recollection Immediate 25.50 25.31 77.46 16.41 [69.98, 84.94] Delayed 14 .44 23.73 82.20 20.11 [70.60, 93.81] Rule knowledge Immediate 22.92 27.90 85.03 13.96 [77.30, 92.76] Delayed 26.99 38.57 84.19 11.82 [76.25, 92.13] 51 Table 2.11 Implicit attributions (based on Guess or Intuition) Explicit attributions (based on Recollection or Rule) n M SD 95% CI n M SD 95% CI L1: English Immediate 24 53.79 11.35 [49.00, 58.58] 24 61.83 14.22 [55.83, 67.84] Delayed 22 61.59 17.70 [53.75, 69.44] 22 67 .77 18.99 [59.35, 76.19] L1: Korean Immediate 24 61.83 19.00 [53.81, 69.86] 22 79.59 13.32 [73.68, 85.50] Delayed 20 66.80 21.41 [56.78, 76.82] 18 83.78 12.12 [77.15, 89.80] Figure 2.4 attributions 52 Relationship between awareness reflected in source attributions, L1, and GJT performance. To answer Research Question 4 - 4, which investigates whether the English and the Korean experimental groups differ in their development of implicit and explicit knowledge, I conducted a 2 x 2 X 2 mixed - design ANOVA on GJT mean accuracy with Attributions (implicit, explicit) and Time (immediate, delayed) as a within - subject variable and L1 (English, Korean) as a between - subject variable. According to the Shapiro - Wilk test, the distribution of accuracy scores by implicit and explicit attributi ons was normal except for the English and the Korean The main effects of Attributions [ F (1,35) = 32.6 8, p < .001, = .48], L1 [ F (1,35) = 4.97, p = .032, = .12] , and Time [ F (1,35) = 5.73, p = .022, = .14] were significant, as were the Attributions by L1 interaction [ F (1,35) = 4.70, p = .037, = .14]. This Attributio ns by L1 interaction is illustrated in Figure 2.5. Figure 2.5 Two - way interaction between Attributions and L1 53 To follow up on the two - way interaction, the simple effects of L1 were analyzed for each attribution separately. The Korean experimental grou p outperformed the English experimental group when they relied on explicit sources [ F ( 1,35) = 8.59, p = .006 , = .20], but there was no relied on implicit sources [ F (1,35) = 0 .78, p = .383, = .02] . The results thus showed that the Korean experimental group developed more explicit knowledge than the English experimental group under incidental exposure, whereas there was no difference between the two L1 groups in the development of implicit knowledge . Therefore, this suggested that the Korean experimental gro up s prior L1 knowledge had a facil itative effect on the development of explicit knowledge under incidental exposure . 2 . 3 .5 Summary of r esults Both the English and Korean experimental groups reliably learned the target grammar after incidental exposure through a novel. After two weeks, which prompted them to outperform the English experimental group more at the delayed GJT. This enhanced knowledge consolidation was led by the Korean L1 - aware group, who noticed at test (not during reading) that the target grammar matched their L1 grammar, Korean. The Korean experimental group provided stronger evidence of implicit learning and implicit knowledge than the English experimental group. T licit learning and implicit knowledge was comparatively limited. The Korean experimental group also developed more explicit knowledge than the English 54 experimental group. 2 . 4 Discussion Multiple lines of evidence in the present study demonstrated the impact of prior L1 knowledge on the implicit and explicit learning of non - native syntax under incidental learning conditions. Although the English and Korean experimental groups consistently outperformed the c ontrol groups, the differential patterns of the - language similarity facilitated the development of implicit and explicit knowledge . First, the Korean experimental group showed strong evidence of implicit knowledge over time, while the English experimenta l group did not provide reliable evidence of implicit knowledge. Sec ond, the Korean experimental group acquired more explicit knowledge than the English experimental group, largely through their aw areness of the connection between their L1 and the target l anguage. These findings extend previous literature in important ways, being the first to reveal a dynamic between L1 prior knowledge, awareness, and time in incidental syntax learning. The present investigation provided not only the supporting evidence of the possibility of learning L2 grammar without awareness (Godfroid, 2016; Kerz, Wiechman, & Riedel, 2017; Robinson, 1995; Rogers et al., 2016; Williams, 2011; Williams & Kuribara, 2008; Williams & Rebuschat, 2012; Tagarelli et al., 2016), but also the com pelling evidence for syntactic transfer of L1 word order knowledge in incidental learning conditions. Interestingly, L1 transfer occurred at both conscious and unconscious levels. First, the Kore indicative of unconscious L1 transfer. Two awareness measures retrospective verbal reports and source attributions suggested that while the English experimental group only showed evidence of implic it 55 knowledge at the delayed test, the Korean experimental g roup reliably exhibited a considerable amount of implicit knowledge across both testing sessions. A follow - up investigation of performance on the relative clause, a complex structure that combined two rules of which most participants remained unaware (i.e. , NP and CP), showed that only the Korean speakers judged these structures with above - chance accuracy, whereas the English speakers did not. In other words, although no Korean speaker mentioned rul es or patterns regarding relative clauses, they still learn English speakers for implicit learning of head - final syntax, most likely due to unconscious transfer of L1 word order rules. The Korean speakers may have unconsciously rel ied on their deeply entrenched L1 word order, boosting their GJT accuracy on responses based on guess and - general lear ning mechanisms (at play in the current learning task) inte racted with their domain - specific L1 knowledge (cf. Leung & Williams, 2014) and produced superior learning outcomes as a result. Second, another noteworthy finding of this study concerns the con scious transfer of L1 knowledge . We can reasonably assume t hat the L1 - aware group, those who noticed the connection between the target language and their L1, used their unconscious L1 knowledge strategically in the present task. This conscious L1 transfer provided the L1 - aware participants a critical advantage ove r their L1 - unaware peers. It also led them to produce significant gains in GJT scores from the immediate to the delayed test . On the debriefing interview, L1 - aware participants reported experiences rewrote the sentence in Korean in my mind I read the given test sentence s in t he way I read Korean sentences developing mor e explicit knowledge than the English speakers. 56 L1 awaren ess appears to be very unique in its nature. They developed conscious judgmental knowledge based on their unconscious structural knowledge (Dienes & Scott, 2005). L1 awareness involves the delibera te activation of a full set of implicit L1 knowledge, but i t certainly cannot be equated to the development of L2 implicit knowledge. Rather, it is the activation of a given body of knowledge for strategic use. More importantly, L1 awareness also cannot be viewed as the development of explicit knowledge; that is, although the participants were aware of the importance of their L1, they could not verbalize the specific rules in the L1 and the target language. One possibility is that the participants did not h ave the metalinguistic terminology needed to describe the r egularities in the target language (none of the participants were linguistic majors), particularly regarding the rules for complex structures. Hence, some caution is needed when interpreting the no n - verbalization as a lack of explicit linguistic knowledge. One could imagine a situation in which a tennis player tries playing badminton. Although he or she has not swung a badminton racket before, he will be able to hit the shuttlecock quickly when his body unconsciously remembers the body movement used in tenn is. Once he figures out the similarities between tennis and badminton, he would be able to master badminton with the help of his experience in tennis, even if he has never taken a single tennis les son in his life. In this way, it is probable that the Korea n speakers did better than the English speakers on the GJT with the help of their deeply entrenched Korean grammar. Traditionally, SLA researchers have examined two levels of awareness by means o f debriefing interviews ( Schmidt , 1995, 2001): noticing (i. e., a low level of awareness that involves a conscious registration of stimuli) and understanding (i.e., a high level of awareness that requires the recognition of a rule ) . Such a dichotomy may be too simple to capture the multifaceted cognitive layers of awareness in some cases. Keeping this in mind, some 57 researchers recently attempted to establish more detailed categories of awareness through a thorough debriefing interview (Curcic, Andringa, & Ku iken, 2019; Godfroid, 2016; Leow, 2015a, 2015b; Rebuschat e subgroup (i.e., participants who noticed the important role of verbs) showed mixed characteristics of aware and unaware learner s. Using a visual - world paradigm, Curcic et al. (2019) foun determiners helped them comprehend at test) showed successful L2 predictive processing. It should be highlighted that the se types of awareness commonly enabled learners to engage i n a strategic and deliberate behavior that contributed to their excellent test performance, similarly to the L1 - awareness observed in this study. However, unlike Godfroid (2016) and Curcic et al. ( 2019), in this study I used an untimed, written GJT as asse ssment and L1 - awareness emerged only during the GJT. The question for the Korean speakers, then, is how was the representation of Koreanish established? The shared syntax model may provide an exp lanation for this representational question (e.g., Hartsuik er, Pickering, & Veltkamp, 2004; Bernolet, Hartsuiker, & Pickering, 2013; Hartsuiker & Bernolet, 2017 ). Using evidence of crosslinguistic priming, this model proposed that bilinguals have shared sy ntax representations, in which all L2 structures similar to L1 merge with the corresponding L1 node and thus establish one final state of abstract representations. Based on this model, the extensive amount of Koreanish input may have resulted in shared rep resentations of Korean and Koreanish in the Korean speakers , given that the word orders of the two languages were identical. The Korean speakers could have accessed their existing L1 syntactic nodes, combined the Koreanish structures with the L1 nodes, and then connected the English vocabulary to the shared syntax . 58 From a methodological perspective, this study presented two advantages: the use of a naturalistic task reading a novel during training and the use of a delayed test. First, the training effect of the naturalistic task reading a novel showed an interact ion with prior L1 knowledge. The amount of input (497 sentences) achieved through novel reading during the training was almost four times greater than that achieved in previous studies [ Grey et al. (2014) : 128 sentences; Kim & Godfroid , in press : 120 sente nces; Miller & Godfroid , (2019) : 120 sentences; Rogers et al. ( 2016 ) : 144 sentences ; Tagarelli et al. ( 2016 ) : 120 sentences; Williams & Rebuschat ( 2012 ) : 120 sentences ] . Moreover, reading an aut hentic novel might have simulated more natural reading process es than reading isolated sentences or unconnected, short texts. However, depending on the L1 background, such abundant exposure involved in a naturalistic task resulted in differential learning rates. On one hand, the English experimental group showed simi lar learning rates (immediate: 56.93%; delayed: 58.91%) as those in previous studies that utilized shorter and more controlled exposure tasks [ e.g., Grey et al. ( 2014 ) : immediate 57.5%, delayed 59.8%; Kim & Godfroid (in press) : 55.6 %; Miller & Godfroid ( 2 019 ) : 53.4 %; Rogers et al. ( 2016 ) : 55.44%; Tagarelli et al. ( 2016 ) : 55.53%; Williams & Rebuschat ( 2012 ) : 54.6% ] . On the other hand, the learning effects were stronger (immediate: 69.64%; delaye d: 77.61%) in the Korean experimental group than in the study groups in previous studies. This di screpancy indicated that the four - time input in fact did not produce superior learning gains when participants were not aided by prior L1 knowledge. That is, t he extensive input under incidental exposure was likely to be beneficial only when a positive L1 tran s fer occurred. As to why more input did not lead to greater learning without the help of L1 knowledge remains a question . Plausibly , the naturalistic input from the novel was cognitively too demanding for beginner lea rners because of the complexity of the unmodified sentences. In terms of maintain ing the 59 ecological validity of the study , I had no control over the sentences from the original material (i.e. the novel); for example, the type and token frequency of the tar get structures, which were shown to mediate the effectiveness of incidental learning (Denhovska, Serratrice, & Payne, 2016) , were not controlled . T here might have been too much syntactic complexities in the input (e.g., too many rules and examples), and su ch input was introduced at once without any instructions, which might have caused cognitive burden for beginner learners wh o could not rely on prior L1 knowledge. Second, the use of delayed tests revealed an intriguing pattern of knowledge development ov er time. Consistent with Grey et al. (2014) and Morgan - Short et al. (2012), incidental learning of syntax was durable regar dless of L1 background. Moreover, the Korean speakers showed gains in GJT scores over time without additional exposure, and these gai ned stemmed primarily from the L1 - aware group . This finding could be accounted by the benefits of sleep in the enhancement of implicit and explicit memories (Batterink, Oudiette, Reber, & Paller, 2014; Fischer, Drosopolous, Tsen, & Born, 2006; Plihal & Bor n, 1997; Wagner, Gais, Haider, Verleger, & Born, 2004). Psychological evidence suggests that sleep has facilitative effects on the conversion of implicitly acquired information into explicitly available knowledge, possibly through an interaction between im plicit and explicit memory processes. For example, Batternink et al. (2014) provided neurolinguistic evidence showing that sleep contributes to the stabilization of new linguistic rule knowledge due to the reactivation of linguistic regularities during sle ep. Likewise, in the present study, during the two - week span between the immediate and delayed tests, the participants had multiple nights of sleep, which presumably benefitted their memory consolidation of newly acquired syntactic knowledge. Given that th e majority of Korean L1 - aware participants (seven out of nine) became aware at delayed testing, there is a possibility that 60 sleep could have given rise to L1 awa reness, resulting in their improved performance at delayed testing. Had it not been for the delayed tests, such interaction between time, L1, and awareness could not have been demonstrated . To obtain a more complete picture of the time - course of implicit a nd explicit learning, more research is needed that uses delayed testing, with varying degrees of delay built into the resea rch design. 61 CHAPTER 3 ONLINE MEASURE: READING TIMES FROM EYE - TRACKING 3 . 1 Background 3 . 1 . 1 Theoretical backgro und on L2 developmental processes A wide array of SLA theories share a goal of explaining how L2 develops. One way to observe the acquisitional processes in L2 development is to draw on a cognitive psychological perspective, particularly through the lens of domain - general learning mechanisms. From this cognitive perspectiv 1998, p. 637), such as reasoning, motor activity, and visual perception. This approach has enriched the field by making it possible to explore empirically whether L2 development is comparable to other cognitive development as well as the ways in which it is unique (Ellis, 2006; DeKeyser, 2007; MacWhinney, 1997; Ullman, 2005). In this section, I will briefly introd uce four cognitive - psychological frameworks that provide insights into the present study, with regard to L2 knowledge development and representation: the two - dimensional model of language proficiency (Bialystok, 1994), the representational redescription mo del (Karmiloff - Smith, 1992), the radical plasticity thesis (Cleeremans , 2007, 2011), and skill acquisition theory (Anderson, 1993; DeKeyser, 2003, 2015). Bialystok (1994) put forward a two - dimensional framework for explaining cognitive aspects of L2 prof iciency development. Specifically, she posited analysis and control as two processing constructs that jointly lead to an increase in L2 proficiency. First, analysis is the (knowledge of the world) become rearranged into explicit representati 159). Through analysis, implicit knowledge becomes explicit knowledge, and this increases the accessibility to knowledge. Second, control is the real - essing choice about where 62 igh level of control, an ability to selectively allocate attentional resources, is necessary for learners to develop fluency or automaticity (also see Segalowitz, 2010). According to Bia are applied, mental representations undergo a qualitative change, creating more analyzed representations that can be attended more efficiently and selectively. However, these two abilities might not always go hand in hand. It is possible some learners may have a higher level of analysis but a lower level of control. Karmiloff - itive development. In this model, implicit knowledge can be gradually redescribed into explicit knowledge, going through four stages of representational redescription. The first stage is the Implicit level, which involves data - driven learning and procedura l representations. Next, at Explicit Level 1, knowledge is abstracted from the procedural representation. This may yield inflexible behavior because the child makes error corrections to adapt his or her knowledge to a new code. Knowledge is still unconscio us at this point. The third stage is Explicit Level 2, in which the co nscious manipulation of knowledge is possible but verbalization of knowledge is not available yet. Due to increased understanding, performance may show improvement during this period. Fi nally, at Explicit Level 3, a full set of explicit knowledge is formed , which appears to be conscious, verbalizable, and flexible. According to Karmiloff - explicit knowledge is likely to show a U - shaped pattern, be cause learners make errors and error corrections at Explicit Level 1 a nd subsequently show improvement at Explicit Level 2. The radical plasticity thesis (Cleeremans, 2007, 2011) regards awareness as evidence of learning. Drawing on a connectionist appro ach, Cleeremans accounted for how the brain 63 formulates awareness throu gh learning. The crux of his thesis is that awareness is the outcome of representational system. By processing information over long time scales, learners acc umulate implicit knowledge, which results in a weak knowledge representation. Then, as learning , Dienes and Perner, 1999), resulting in a stronger mental representat ion with explicit knowledge. Awareness plays an essential role in learning because it gives adaptive, flexible control over behavior. The involvement of awareness will decrease until the learner reaches a very high - quality of representation; this is the po int where automatic behavior emerges. Cleeremans pointed out that automatic behavior may not be unconscious (but for a different account, see also Williams, 2009); rather, awareness is o ptional in automatic behavior, because automatic behavior Skill acquisition theory (Anderson, 1982, 1993) draws on the power law of practice as a key property of human learning mechanisms, which is ubiquitously applicable to a wide array of cognitive (e.g., reading) and psychomotor skills (e.g., telegraphy). When humans practice a of knowledge through initial changes in behavior to eventual fluent, spontaneous, largely effortless, and highly - linguistic domains. In 1981, Newell and Rosenbloom presented a se minal study on automaticity, in which they maintained the power functi on to be the central law of human learning. Because practice resulted in a specific shape of learning curve, namely the power function (i.e., speci ), across diverse tasks, this mathematic al function became referred to as the power law of practice. In the ma thematical formula, Y represents response times (RTs) or error rate; X represents the amount of practice; the 64 exponent n represents the learning rate at which performance improves with p ractice. Importantly, the power learning curve of RT and error rate in volves an initial quick and drastic declarative knowledge ) to procedural knowledge ). This learning p eriod is then followed by a slow and gradual plateau, which is an inde x of the automatization of procedural knowledge (Anderson, 1982, 1993). According to this view, for various human skills, practice and the power law of practice entail different stages t hat lead a qualitative change over time. The speed - up observed in the power learning curve can be considered as the adaptation to the statistical structure of the input (Anderson & Schooler, 1991). 93), DeKeyser (1997) conducted the first study to e mpirically investigate whether L2 skill development followed the power law of practice and whether a skill can be transferred to other skills, from comprehension to production and vice versa. In this longi tudinal study, the participants explicitly learned grammar rules and vocabulary items of an artificial language before they engaged in computer - administered comprehension (sentence picture matching) and production tasks (picture description) over 8 weeks. The results confirmed the prediction of skill acqui sition theory, by showing that both longitudinal RT data and error rates followed the power function. The learning curves for both L2 comprehension and production revealed fast proceduralization followed b y slow automatization. This finding, therefore, pro vided evidence showing that L2 skills, like other cognitive and psychomotor skills, can be automatized with practice, undergoing two important stages proceduralization and automatization. The review of t hese theoretical explanations of domain - general pri nciples of learning provides important lessons to understand the development of L2 knowledge. First, repeated 65 exposure and extensive experience bring about a qualitative change in mental representations. S econd, the accumulation of implicit cognition may b e necessary for further knowledge development. Third, the involvement of awareness is likely to aid knowledge development, by providing control over behavior. Lastly but most importantly, knowledge develop ment can be represented in a speedup in behavior, w hich can be highly informative of how language users can automatically and fluently access their mental representations. As Bialystok (1994) rightly pointed out, essentially, L2 acquisition can be differ ent from other developmental processes, including L 1 acquisition. This is because adult L2 learners need to build a new representation on a fully elaborated L1 linguistic system. Of note, adults already e and more well - developed procedure for directing a would be an interesting exploration to see how prior L1 knowledge, which may or may not be domain - general, affects L2 d evelopment, which is hypothesized to draw on domain - general learning mechanisms, such as the power learning curve. In light of this, in the present study I aim to explore empirically how L1 grammatical knowledge affects the change in real - time processing o f syntactic structures in incidental learning condi tions. 3 . 1 .2 Incidental exposure and changes in reading times Recently, eye - tracking methodology emerged as a powerful tool to explore cognitive processes that underlie L2 development in diverse learnin g contexts (e.g., vocabulary: Godfroid, Ahn, et al. , 2018; Elgort et al., 2018; grammar: Indrarathne, Ratajczak, & Kormos, 2018). Eye - movement recording can offer detailed information about how the learning process unfolds over time (Godfroid, 2020; Godfro id & Winke, 2015; Leow, Grey, Marijuan, & Moorman, 2014 ) because it gives a spatial and temporal reflection of attentional processes in real - time. That is, as 66 an online processing measure, eye movement records can provide a window into peripheral, selectiv e, and focal attention (Godfroid, 2019). One of the benefits of eye - movement data is that it offers research participants a fairly natural reading experience (but see Spinner, Gass, & Behney, 2013), as an eye - tracking experiment usually does not require a secondary task (e.g., pressing a button) that may i nterfere with normal reading processes (Dussias, 2010; Godfroid & Spino, 2016 ) . Thus, for reading long texts, such as novels, eye - tracking may well be the preferred methodology to observe moment - by - moment cognitive processing. Combining natural reading ma terials with an incidental learning experiment, Godfroid, Ahn, et al. (2018) used eye - tracking methodology to investigate changes in reading times for unfamiliar vocabulary that occurred repeatedly in the novel and how the reading times related to vocabula ry learning. The participants read five chapters of an authentic English novel, in which target Farsi - Dari words were naturally embedded, while their eye - movements were recorded. Using growth - curve modelli ng, the authors found that with repeated exposure, processing time (i.e., total reading time) on the target words decreased in a non - linear fashion, generating an S - shaped pattern. The authors proposed that the speed - up in lexical processing over time coul d al build - up and specification of a new word Using an eye - tracking methodology, Indarathne et al. (2018) focused on the cognitive proce ssing during implicit and explicit learning of a sy ntactic construction. Similar to Godfroid, Ahn, et al. (2018), the authors also found an S - shaped curve in their eye - movement data. The initial decrease in total fixation duration in the early stage of lea rning was considered a reflection of habituation; t the other hand, the sharp decrease in total fixation duration during the last stage of learning, 67 which was shown only in the explicit instruction conditions, was taken as evid ence of increased unannounced post - test. Taken together, a decrease in eye fixation times (i.e., speed - up) can be used as evidence of a change in mental representati on, such as proceduralization and automatization. Eye - tracking data in these studies provided a window into cognitive changes during learning processes. To obtain a multi - faceted understanding of real - time L2 processing and learning, researchers could furt her triangulate eye - movements with other measures of knowledge and awareness (Godfroid & Winke, 2015; Rebuschat et al., 2013). 3 .1. 3. Incidental exposure and sensitivity to L2 violations Recently, many SLA researchers have begun to pay special attentio n to online sensitivity to grammatical violations as a reflection of integrated L2 knowledge (Godfroid, 2016; Granena, 2013, Jiang, 2007, 2012; Leung & Williams, 2011, 2012, 2014; Sanz & Grey, 2015; Spinne r & Foote & Upor , 2017 ; Spinner & Jung, 2018; Suz uki, 2017; Suzuki & DeKeyser, 2015, 2017; Vafaee et al., 2016). Integration of knowledge into the linguistic system is likely to enable automatic competence, which involves spontaneous language use in both receptive and productive tasks (Jiang, 2007). If L 2 knowledge is integrated, learners can engage in automatic and fluent processing without attending to grammatical accuracy. A typical example of integrated knowledge would be the L1, in the sense that L1 can be automatically retrieved and fluently produce d without attending to grammatical accuracy. Researchers and practitioners ultimately aim for learners to achieve this type of integrated knowledge because it is automatic in its activation and functioning and underlies fluent language use (Doughty, 2003 ). Grammatical sensitivity can be indexed by an increased processing time in response to 68 L2 violations (e.g., Godfroid, 2020). which can be assess ed by psycholinguistic techniques, such as RTs (e.g., Leun g & Williams, 2011), eye movements (e.g., Keating, 2009), and event - related brain potentials (ERP) (e.g., Tokowicz & MacWhinney, 2005). Researchers can compute sensitivity by subtracting processing times for grammatical sentences from those for matched, un grammatical sentences; the difference reflects the degree of sensitivity for each participant. The core assumption of this methodology is that someone with integrated knowledge will unconsciously sl ow down to a grammatical violation in the sentence while p rocessing. However, if a participant has incomplete or little implicit knowledge, then his or her processing time for violations and non - violation sentences would not be significantly different. Adopting a RT methodology, Leung and Williams (2011) bolste findings of implicit learning of form - meaning mappings by introducing a real - time component. They found online sensitivity increased RTs in the violation block among participa nts who remained verbally unaware. Using a pretest - treatmen t - posttest design, Godfroid (2016) also matching task) but also at the posttest (a word monitoring tas k) that followed the intervention. In these studies, an unc onscious slowdown to grammatical errors was interpreted as the automatic activation of implicit knowledge (Godfroid, 2016; Leung & Williams, 2011). Interestingly, online sensitivity appears to be (Foucart & Fren ck - Mestre, 2011; Jiang, 2007, 2011; Keating, 2009; Lim & Christianson, 2015; Sagarra & Ellis, 2013; Tokowicz & MacWhinney, 2005; Tolentino & Tokowicz, 2011, 2014). To illustrate, Jiang (2007) used RTs from a self - paced reading task to reveal how online sen sitivity, which was hypothesized to reflect integrated L2 knowledge, can be selectively activated depending on the target structure. Native speakers of English and Chinese English - as - 69 a - second - langu age (ESL) speakers read grammatical and ungrammatical Engli sh sentences containing two target structures: verb categorization and plural - s. Through a comparison of reading times between English and Chinese speakers, he found out that L2 knowledge integrat ion was dependent on the type of structure. While the Engli sh speakers showed sensitivity to both target forms, the Chinses speakers were sensitive only to verb categorization; they did not show a delay when reading sentences containing errors (omissions o f) plural - s. One possible explanation for this phenomenon was the transfer of L1 knowledge because in Chinese, plural marking is rare while the verb categorization system is similar. An ERP study by Tokowicz and MacWhinney (2005) documented the effect o f L1 - L2 similarity on online grammatical sensitivity. Engli sh learners of Spanish performed a GJT while their brain responses were recorded. Importantly, there were three types of target forms: a) tense marking (L1 - L2 similar form), b) determiner number ag reement (L1 - L2 dissimilar form), c) determiner gender agree ment (L2 unique form). Although GJT accuracy was only near chance - level, the researchers found online sensitivity, evidenced by a difference in brain responses to ungrammatical and grammatical sent ences, for tense marking and determiner - gender agreement, b ut not for determiner - number agreement. In other words, participants showed differential sensitivity to syntactic anomalies depending on the L1 - L2 similarity. Specifically, they showed a P600 effec t 3 when the L2 form was similar to the L1 and when it was u nique to the L2, but not when it was dissimilar. Based on these findings, the authors emphasized the importance of cross - language similarity in implicit processing. However, it remains unknown how awareness played a role during online processing because no awareness measure was employed in their study. An open question, therefore, is whether participants were aware of the L1 - L2 relationships to which 3 P 600 is a brain wave form that takes place to syntactic anomalies. 70 they showed sensitivity. 3 . 1 . 4 Research q uestio ns In this chapter, I will present the answers to the fol lowing research questions: RQ1: Learning curve 1 - 1. (prior L1 knowledge and learning curve) sentences cha nge according to the power law of practice? 1 - 2. (awarenes s and learning curve) Do the aware and the unaware groups differ in terms of changes in residualized sentence reading times over time? RQ2: Sensitivity 2 - 1. (prior L1 knowledge and sensitivity) Do the English and Korean speakers show online sensitivity by slowing down their reading times for sentences with word order violations? 2 - 2. (awareness and sensitivity) Do the aware and unaware groups differ in terms of online sensitivity to word order violations? 3 . 2 Methods 3 . 2 . 1 Participants The same g roup of English speakers and Korean speakers who participated in the study reported in Chapter 2 also participated in this study. English speakers ( n = 40, M age = 23.71, SD = 3.90) and Korean speakers ( n = 40, M age = 25.25, SD = 4.42) were recruited from a large, M id w estern university in the United States and two universities in South Korea. Those invited to 71 participate held at least a a degree - granting university program at the time of the experiment. Each group was subdivided into two: an experimental and a control group . This resulted in four subgroups: (a) an English experimental group (EE, n = 25), (b) an English control group (EC, n = 15), (c) a Korean experimental group (KE, n = 25), and (d) a Korean control gro up (KC, n = 15). The experimental groups, EE and KE, read the novel The Mysterious Affair at Styles (Christie, 1920) in a semi - artificial language that consisted of English vocabulary and Korean word order. The control groups, EC and KC, read the English v ersion of the novel; that is, the same novel, but with English vocabulary and English word order. The English speakers did not have any knowledge of Korean or any other h ead - final languages, such as Japanese or Turkish. Further, the experimental groups h ad no experience in German, the language used to create ungrammatical sentences in the violation blocks during the testing phase (see section 3 - 2 - 3). The Korean speakers we re Korean - English bilinguals who were highly proficient in English, such that they h ad no difficulty in reading L2 English novels. At the time of the research, 19 Korean speakers resided in the United States, whereas the remaining 31 Korean participants re sided in South Korea. Their mean iBT TOEFL score was 111.67 ( SD = 4.65) and their me an length of residence in an English - speaking country was 7.89 years ( SD = 3.27). Tabl e 3.1 lev els. The KE and KC groups showed no difference in their age of exposure ( U = 180.00, p = .847, r = 0.01), length of residence ( U = 174.50, p = .720 r = 0.05), TOEFL score ( U = 127.00, p = .474 r = 0.16), or self - rated proficiency ( U = 170.00, p = .639, r = 0.10). The recruitment process involved flyers, web postings, and word of mouth, an d compensation took the form of a $30 reward at the end of the experiment. 72 Table 3.1 KE ( n = 25) KC ( n = 15) M SD M SD Age at testing 25.32 4.61 25.13 4.24 Age of exposure 5.80 1. 76 5.93 1.67 Length of residence (years) 7.74 3.26 8.12 3.38 TOEFL score 112.22 4.99 110.69 3.97 Self - evaluation of proficiency Total 34.96 1.79 34.53 2.50 Listening 8.96 0.68 8.80 1.01 Reading 8.88 0.78 8.60 0.78 Speaking 8.56 0.82 8.47 1.13 Writing 8.56 0.87 8.67 0.98 3 . 2 .2 Targeted semi - artificial language: Koreanish For t he present study , I adopted a semi - artificial language paradigm, in which vocabulary or second language was rearranged to follow the targ patterns. Specifically , the semi - artificial language involved English vocabulary and Korean syntax. To create this hybrid language, hereafter named English version of the Agatha Christie novel, The Myst erious Affair at Styles , were rearra nged according to Korean word order. The words were thus familiar to the participants, but the syntax formed an unknown system that created difficulty in comprehension. W ord order in Korean and English seem to mirror e ach other, because each language has a different system of ordering heads and complements (Onnis & Thiessen, 2013). Syntactically, 73 English is a head - initial language in which heads of phrases precede their complements, thus forming a right - branching struct ure; Korean, on the other hand, is a head - final language, wherein heads follow complements to form a left - branching structure. To illustrate, the English sentence in Korean, because the verb (head ) comes at the end of the phrase, an The following sentences are excerpts from the novel used in the experiment and demonstrate how an English sentence (1) and (3) can be rearranged according to Koreanish rules, as shown in (2) and (4) respectively. (1) English: A vague suspicion of eve ryone and everything filled my mind. (2) Korean: Everyone and everything of a vague suspicion my mind filled. the apposition, is the head of the phrase and is moved to the final pos ition. Second, everyone of a vague suspicion, serves as the head of the phrase. filled. filled, " and follows SOV word order. 74 First, t the r elative by preceding rather than following it. This results in the relative clause usin. thus forming the verb phrase urse is who a cousin have , the head, migrates to the end of the phrase. Hence, the full sentence reads, In this manner, Koreanish consistently features the head, the main element of the phrase, in the final position. 3 . 2 . 3 Materials The exposure task: Novel reading. The exposure text for training, The Mysterious Affair at Styles by Agatha Christie, provided incidental exposure to the semi - artificial word order system, Koreanish, to participants. The participants read the first two chapters, Chapters 1 and 2 (approximately 6,500 words), of The Mysterious Affair at Styles . T he same novel has been used in a series of eye - Dutch - Van Assche, Drieghe, & Duyck, 2017; Cop, Drieghe , & Duyck, 2015; Cop, Keuleers, Drieghe, & Duyck, 2015). Cop and her colleagues noted that this novel was selected based on its appropriate difficulty level for college level L2 English speakers (Flesch Reading Ease = 81.3, SMOG grade = 7.4), 4 as we ll as i ts similarity in word frequency distribution to natural language based on the Subtlex database (Van Heuven, Mandera, Keuleers, & Brysbaert, 2014). In other 4 Flesch Reading Ea se gives a scale between 0 (difficult to read) and 100 (easy to read); higher scores reflect greater readability. The SMOG grade is an esti mation of how many years of educations are required to understand the text. 75 words, this novel represents suitable reading material for an L2 experiment because it is eas y to re ad and reflects real - life, natural language use. The participants were told that they were going to read the first two chapters of a detective novel (92 screens), and that for some of the participants the sentences would be presented with the wor ds scra mbled. Before reading, the participants were given a brief groups read the Koreanish version of the novel, consisting of English sentences with Korean wor d order , while the control groups read the English version. The participants were informed that there would be a practice session at the beginning and a short break approximately 15 screens later (total 6 breaks). At every other break, the participants wer e asked to answer comprehension check questions. The purpose of the comprehension test was to keep participants engaged while reading the scrambled sentences for about an hour. These comprehension tests, which contained eight simple true or false statement s regar ding plot - specific information, were administered three times throughout the reading. The total number of the questions was 24. One point was assigned for correct answers, and the total score of overall comprehension was 24. The overall reliability coeffic KE group. It is important to note that the novel was employed not only to introduce Koreanish oreanis h during training. Unbeknownst to participants, the exposure text comprised three phases: (a) a practice phase, (b) a training phase, and (c) a testing phase. In the practice phase, all sentences were presented in English. I selected a practice pass age fro m the English version of the same novel and presented it to the participants in order to familiarize them with reading on the screen 76 and moving from one screen to the next. After that, an introductory part of the novel was presented, which provided partici pants with the context of the story and helped familiarize them with the plots. Next, the training phase involved an extensive amount of Koreanish sentences ( n = 479) that flooded participants with the target syntactic input. Lastly, the testing p ha se was used sentences each: (a) a control_pre block, (b) a violation block, and (c) a control_post block. The control_pre and control_post blocks followed Koreani sh word order and were used as baselines to calculate the increase in reading times in the violation block. Sentences in the violation block were ungrammatical, in that they did not follow the Koreanish word order rules. Rather, they followed the German wo rd orde r (Rebuschat & Williams, 2012). Therefore, in the violation block, English sentences were rearranged in accordance with German syntactic rules, in which the v erb is placed at the first, second , or final position depending on the clause type and sequ ence. F f the s ame sentence, violation block were ungrammatical not only by Koreanish rules but also English ones; however, sometimes the violation sentences, which w ere rea rranged in German word order, also followed English word order and were identical to the original English sentences. Those sentences were undesirably affect t he read ing times for these items, which may obscure any sensitivity effects. Three counterbalanced versions of the exposure text were created to control for lexical and topical confounds (see Table 3.2). The sentences in the testing phase rotated between the 77 co ntrol_pre, violation, and control_post blocks, as shown in Table 3.2. For example, 20 sentences, from Sentence 524 to Sentence 543, formed the violation block in Version B, the control - post block in Version A, and the control pre - block in Version C. Due to this rotation, the starting point of the training phase, in which Koreanish sentences were presented, varied depending on the text version. However, regardless of which version participants saw, they received the same amount of training, since they read t he same amount of Koreanish sentences. To even out the number of presented sentences across all participants, participants who read Version A and Version B read an extra 40 (Sentence 544 Sentence 583) and 20 sentences (Sentence 564 Sentence 583) res pective ly, which were taken from the testing phase in Version C. Those extra sentences were presented in regular English syntax as a practice passage during the practice session, which enabled me to control the content familiarity with the exposure text. 5 5 For Version C, participants read an a dditional passage selected from chapter 4 (25 sentences) before they started reading the main text. 78 Table 3.2 The counterbalanced version of the exposure task Phase Syntax N of sentences Sentence ID Version A Version B Version C Practice English 44 544 583, 1 4 564 583, 1 24 1 44 Training Koreanish 479 5 483 25 503 45 523 Testing: Control_pre Koreanish 20 484 503 504 523 524 543 Testing: Violation German 20 504 523 524 543 544 563 Testing: Control_post Koreanish 20 524 543 544 563 564 583 79 Online measures of learning and knowledge: Eye - movements. movements during the exposure task, I calculated two online measures of learning and knowledge: changes in sentence reading times as a learning measure (DeKeyser, 1997; Godfroi d, Ahn, et al., 2018; Indarathne et al., 2018 ) and increased reading times of (i.e. sensitivity to) violations as a knowledge measure (Godfroid, 2016; Jiang, 2007; Leung & Williams, 2011; Suzuki & DeKeyser, 2015, Suzuki, 2017; VanPatten & Smith, 2015). The se eye - movement measures were advantageous, in that they enabled me to employ a within - subject design - subjects design (showing how much extra reading time the experim ental groups spend compared to the control gro ups). From the eye - tracking data in the exposure task, I derived changes in reading times over time and from the testing phase, I calculated changes (increases) in reading times in response to violations. Cha nges in sentence reading times: learning curve . Changes in residualized sentence reading times were used as evidence of learning. To calculate the residualized sentence reading ading time for the corresponding Koreanish sentence. Thus, the residualized sentence reading time represents the extra processing time participants spend when reading Koreanish word order, compared to English word order. This measure yielded one difference value per sentence and enabled me to conduct a sentence - level analysis (Cop et al., 2015). With the residualiz ed sentence reading times, I focused on how and in what form the processing of Koreanish sentences changed over time. Specifically, I investiga ted whether the power law of learning can be applied to changes in residualized sentence reading times. A 80 learnin g curve described by a power function involves two critical components: a rapid decrease in the early stage of learning, followed by a gradual plateau in the later stage. If the changes in residualized sentence reading times of Koreanish are governed by th e power function (i.e. Y = X - n ) , where Y is the residualized sentence reading time and X is Time, this suggests that participants started proce ssing Koreanish sentences more automatically and fluently as a result of repeated exposure (Anderson, 1982, 1993; DeKeyser, 1997). Sensitivity to violations: integrated knowledge. S ensitivity to word order violations was employed as the concurrent measur e of integrated knowledge (Jiang, 2007). Integrated knowledge brings about automatic competence in processing (Jiang, 2007). If participants succeeded in integrating Koreanish word or der in their mental representation, it was expected that they would autom atically slow their reading speed when processing ungrammatical sentences in the violation block, even though they were focusing on comprehension rather than form. To compute sensi tivity, I subtracted the average residualized reading times for the contr ol blocks from those for the violation block; the difference in reading times reflects the degree of sensitivity. To better understand the nature of this measure and its underlying co nstruct, I triangulated the sensitivity data with one of the awareness me asures, verbal reports. Immediately after completing the exposure task, participants were asked if they noticed anything odd during reading. If they said no and showed an unawareness of the target rules, this was taken as evidence supporting the developmen t of unconscious, integrated knowledge of Koreanish word order during learning. Debriefing interview . levels of aw areness of the targeted Koreanish syntactic patterns (see Appendix D). Th e interview 81 was carried out over two sessions. At the end of Session 1, after the post - test (see Chapter 2), I conducted the first part of the debriefing interview to examine particip violation block. The goal was to determine whether participants had noticed the change in word order in the violation block. I asked the participants whether they had noticed anything odd while reading the novel. In addition, I asked them to report whether they had noticed any particular rule or regularit y, to specify when they might have noticed it (i.e., during reading or on the test), and to describe what they believed they had noticed. In Session 2, the second part of the debrie fing interview was conducted online. I began by asking whether participan ts ever indicated recollection or rule knowledge as a source of their grammaticality judgement. If so, they were asked to explain why and state what they were thinking. In addition, I asked participants whether they had tried to search for a pattern in the levels, I asked them to reflect specifically on the placement of words within the sent ences and to recall any specific rule or regularity. Finally, I asked the m questions about their usual reading experiences. Apparatus . Eyelink Portable Duo (SR R esearch, Canada) laptop - mounted eye - tracking camera with a sample rate of 2,000 Hz. In this study , reading was binocular, but only the right eye was recorded. A chin - rest was used to minimize head movement. The text of the novel was presented in the form of paragraphs in 16 - point Consolas font, triple - spaced, on a light gray background. At a 63 - seatin g distance, one letter subtended 0.37 degrees of visual angle. A maximum of ten lines were presented on the screen, and a drift check was performed between screens. The eye - tracker wa s recalibrated three times per participant, and more frequently if I deem ed it necessary. 82 3 . 2 . 4 Procedure The experiment was conducted in two sessions, Session 1 (90 minutes) and Session 2 (30 minutes). Session 1, which was carried out in a quiet, dimly - illuminated study room, included the exposure phase, the immediate testin g phase, and the first part of the debriefing interview (short version). Session 2 comprised the delayed testing phase, the second part of the debriefing interview (long version), and the background questionnaires. In Session 1, the participants first si gned the consent form and then filled out a questionnaire regarding their language background. I explained that they were going to read two chapters from a detective novel (Appendix A ) and provided a brief description of the main characters. The participan ts were informed that there would be comprehension questions after reading each section (Appendix B). Then the practice session started, which allowed them to become accustomed to reading on a laptop with an eye - tracker recording their eye movements. The e ye - tracker was recalibrated three times per participant (at every other break) and more frequently if I deemed it necessary. They silently read the first two chapters of The Mysterious Affair at Styles on a laptop screen while their eye movements were reco rded by the Eyelink Portable Duo eye - tracking system. The experimental groups read the novel in Korean ish, and the control groups read the English version of the novel. After they finished reading the novel, a surprise GJT with source attributions was admi nistered for immediate testing (see Chapter 2). This test was followed by the first part of the debrie awareness of the violation block during reading and Koreanish word order. Session 2 took place two weeks after Sess ion 1 and involved delayed testing, along with the second part of the debriefing interview. To each pa rticipant, the researcher sent an e - mail containing the links to the test and the debriefing interview questions. The participants were asked to complete the test 83 within three days. 3 . 2 . 5 Analysis In this study, I took a chronometric approach to mental p rocessing by analyzing changes in residualized sentence reading times over repeated exposure to nonnative syntax. Before the analysis, the eye - movement re cordings were screened for data quality. At first, I removed short fixations under 80 milliseconds usi ng a built - in program in Data Viewer (SR research, Canada) par ticipant recordings on a trial - by - trial basis. After a careful inspection, I manually corrected fixati ons for drift, which is a discrepancy between the recorded eye gaze location and a errors (e.g., participant movement) or technical errors (e.g., track loss). If it was too difficult t o judge the true locations of drift fixations, I deleted the corresponding sentences to ensure the recording accuracy, which amounted to 1.3% of the entir e dataset. I also decided to exclude the recordings from one Korean participant and one English partic ipant because they skipped large parts of the reading. After cleaning the eye - movement data, I computed the residualized sentence reading times by subtrac experimental grou p for the Koreanish text. I did this for the English speakers (EE EC) and the Korean speakers (KE KC) separately. This new measure, which reflected se ntence - level extra processing time for Koreanish sentences, was used as a dependent variable, sometime s with log - transformation if necessary. All the data from the exposure task (novel reading) were entered into the Statistical Package for the Social Scien ces, version 25. The significance level alpha was set at 0.05. Research question 1. To investigate the real - time learning effect reflected in the changes 84 in sentence reading time during the training phase (Blocks 1 24), I explored whether the eye - moveme nt data followed the power law of practice (DeKeyser, 1997). Combining visual and statistical evidence , I specifically examined the effects of L1 and awareness on the learning curve. The mathematical function that describes the learning curve is a power fu nction: Y = X - n . The most popular approach for testing the power learning curve is to perform a logari thmic transformation. The logarithmic transformation is applied to the independent and the dependent variables. This will render the relationship between log X and log Y (i.e., the learning curve) linear if it follows a power law. Thus, a straight line in a log - log plot can be taken as evidence of the power function. Furthermore, such linearity can be statistically tested by means of a linear regression mod el. Consequently, to answer research question 1 - 1, which investigated the effect of L1 on the learni ng curve, I first created log - log plots for each L1 group. On the x - axis, log Block was displayed as the time - course variable. On the y - axis, log Residual ized sentence reading time was shown. Then, I used a linear mixed - effect model to test whether the two L1 groups differed in their linear relationship on the log - log scale, for instance, whether they had a different slope. The dependent variable was log Re sidualized sentence reading time. The independent variables were log Block (within - subject, time - cours e variable) and L1 (between - subject, group - difference variable). To answer research question 1 - 2, which explored the effect of awareness on the learning c urve, I ran separate linear mixed effects models for the English and the Korean speakers because the t wo groups had different levels for the Awareness variable. The dependent variable was the log Residualized sentence reading time. Log Block and Awareness were entered as the independent variables. For all the analyses, the main effects and the interactions with log Block were of interest. I fitted the models with random by - subject intercepts and random slopes 85 for log Block to account for individual variatio ns in how experimental participants processed the Koreanish sentences initially and over time. Resea rch question 2. To investigate the effect of L1 and awareness on online sensitivity during the testing phase (Blocks 25 27), I performed a series of mixed - design ANOVAs. Online sensitivity was assessed by means of the differences between the violation blo ck (Block 26) and the control - pre block (Block 25) and the control - post block (Block 27). The main effects and the interactions of Block were of interest. To answer research question 2 - 1, which focused on the effect of L1 on sensitivity, I carried out a 2 x 3 mixed - design ANOVA on residualized sentence reading times with L1 (English, Korean) as a between - subject variable and Block (control_pre block, violat ion, control_post block) as a within - subject variable. Additionally, I performed one - sample t - tests on A positive sensitivity score indicates that the participant is sensitive to grammatical violat ions because of their integrated, implicit knowledge of the target syntax ; a negative sensitivity scor e could signal reverse sensitivity, which may be indicative of the activation of non - targetlike grammar rule s . In order to answer research question 2 - 2, w hich examined whether subgroups with different awareness levels differed in sensitivity, I performed a mixed - design ANOVA for each L1 group separately. The analyses were conducted on Residualized sentence reading times with Awareness (English: unaware, VP - aware; Korean: unaware, VP - aware, L1 - aware) as a between - subject variable and Block (control pre_block , violation, control post_block) as a within - subject variable. 86 3 .3 Results 3 . 3 .1 Comprehension t est The overall comprehension scores of the EE and K E groups were 18.17 (SD = 2.51) and 18.13 (SD = 3.17), respectively, which indicated that their comprehension level of the scrambled significant [t(46) = 0.31, p = .759, d = 0.09]. The overall comprehension scores for the control groups, the EC and KC groups, were 23.13 ( SD = 0.96) and 22.80 ( SD = 1.05), respectively . 3 . 3 . 2 Analyses of retrospective verbal reports The retrospective verbal report data from the debri efing interviews from Sessions 1 and 2 were . To enhance the inter - rater reliability in the coding, two independent raters (the researcher and a trained rater) coded the verbal report data, which was gather ed from two debriefing interview sessions. The raters identified 1) what type of awareness the p articipants developed (unaware, VP - aware, L1 - aware), 2) when they became aware of the rules (exposure and testing), and 3) when they reported awareness (Session 1 and Session 2). The reliability coefficient (kappa) for the identification of awareness type was This analysis enabled me to identify subgroups of participants based on their reported awareness of the targe t rules. Note that only the experimental groups, who read the Koreanish version of the novel, participated in the debriefing interview. From the verbal report data, I attempted to confirm the internal validity of the incidental learning condition (cf. Le ow, 1997; Leow & Hama, 2013); that is, whether participants engaged in primarily meaning - focused processing as intended. To this end, I asked the p articipants 87 whether they tried to search for patterns and focused on word order while they were reading. This inspection led me to exclude two participants one from each L1 group who engaged in intentional learning. Their answers, presented in Example ( 6 ), clearly showed that they intentionally searched for a pattern while reading, which meant that their attentio n was primarily oriented toward forms. Example ( 6 ) P119 (English participant) first break, I started to notice a pattern, so I would try to figure it out and unscramble the sentences. I real ized that the verb at the end of P304 (Korean participant) d that the order of noun, verb, and etc. were messed up regularly. I did try to find the order of the words in t The verbal data of the remaining 48 participants (English: n = 24; Korean: n = 24) did not sh ow evidence of such rule - data indicated that half of the participants 12 of 24 participants became aware of the VP rule. This was a partial awareness of Koreanish word order rules, since p articipants did not mention any other rules regarding head finality. Based on their awareness of the VP rule, I categorized them as VP - aware. Among the 12 participants, 10 participants reported awareness of the VP rule at Session 1, during the short versio n of the interview, and two participants reported awareness at Session 2 (two weeks after Session 1), during the long version of the interview. All participants 88 said they became aware of the VP rule during the GJT. The other 12 participants did not report awareness of any target rules and were categorized as unaware. verbal report data exhibited three types of awareness: L1 aware ( n = 9), VP - aware ( n =10), and unaware ( n = 5). First, I found an interesting subgroup, the L1 - aware group, who reported their awareness of the relation ship between the target language and their L1 (Korean). These nine participants reported that they became aware that the lan guage used in the novel followed Korean word order while they were taking the GJT. It is important that they were unaware of any rul es ( n = 7) or only discovered the VP rule ( n = 2) during the exposure phase; it was in the testing phase that they figured o ut that the Koreanish sentences in the novel matched their L1. In terms of the timing of reported awareness, among 9 L1 - aware partic ipants, 2 participants reported their L1 awareness after taking the immediate GJT at Session 1; the rest, 7 participants, re ported their awareness after taking the delayed GJT at Session 2. Second, 10 participants became VP - aware. They reported noticing th e VP rule but did not notice other rules regarding head finality in the Koreanish text. All VP - aware participants reported t heir VP - awareness at Session 1 after taking the immediate GJT. Lastly, five participants remained verbally unaware and were categori zed as unaware. Additionally, I also aimed to ensure the nature of online sensitivity (i.e., conscious or unconscious) from the retrospective verbal reports. Given that online sensitivity can be evidenced by an automatic slowdown in processing of violatio ns without awareness (Godfroid, 2016; Jiang, 2007), which may take place outside awareness, I asked the par ticipants whether they noticed anything odd or changes in the pattern while they were reading the two chapters of the novel. This information enabled me to exclude four participants (two from each L1 group), 89 who noticed the presence of the violation block during the testing phase (see Example 7 ). Example ( 7 ) P 117 (English participant) ared to go back to a normal P 326 (Korean participant) As shown in Example 4, some participants noticed that the word o rder changed at the violation block, and I decided to exclude these violation finders from the main sensiti vity analyses. These four violation finders showed a lower sensitivity score ( M = - 55, SD = 346) than the rest of the participants (see Table 3.6, En glsih: M = - 12, SD = 483; Korean: M = 158, SD =292). This indicated that after noticing the change in the w ord order, they read the violation sentences, which followed German word order, faster than the Koreanish sentences. 3 . 3 . 3 Changes in residualized s entence reading time: Learning curve The visualizations of the log - residualized sentence reading times are presented in Figures 3.1 and 3.2, respectively. In these graphs, log Block is on the x - axis, and log Residualized sentence re ading time is on the y - axis. The two log - the linear relationship, I carried out a linear mixed - effect model. The dependent variable was log Residualized sentence reading time. The independent variables were the log Block as a time - 90 course variable and L1 as a grouping variable. I tested the follow ing series of models to determine whet her the addition of fixed effects and interactions improved the model fit: M0, an intercept - only model; M1, a model adding a fixed effect for log Block; M2, a model adding a fixed effect for L1; M3, a model adding an interaction term between log Block and L1. Figure 3.1 - log plot All models, except for M0, included random slopes for log Block and the random by - subject intercepts. M0 included only the random by - subject intercept. The mo del fit was assessed by evaluating the change in chi - goodness of fit (Baayen, 2008). Each consecutive model yielded a significantly better fit than the 91 p p p < .01]. Figure 3.2 - log plot The final model (M3) (Table 3.3) revealed that L1 [ F (1, 46) = 6.56, p = .014], log Block [ F (1, 46) = 15.87, p < .001], and the interaction between L1 and log Block [ F ( 1, 46) = 9.10, p = .004] significantly predicted log Residualized sentence reading time. The relationship between log Block and log Residualized sentence reading time showed significant variance in the p < .01] 46.71, p < .01] across the participants. Crucially, the two - way interaction indicated that the two L1 groups followed different learning trajectories. Next, I fitted separate multilevel models to the English and the Korean speakers to 92 understand the nature of the interaction. In the separate models, I retained log Block as a predictor; additionally, to answer research question 1 - 2, I included the main effect of Awareness in the models. For the English speakers, Aw areness was a two - level factor (unaware and VP - aware); for the Korean speakers, it was a three - level factor (unaware, VP - aware, and L1 - aware). The visualizations of the log - residualized sentence reading times by awareness are presented in Figures 3.4 and 3.5, respectively. Table 3.3 Regression output of the final model (M3) b SE p Fixed effects Intercept 7.66 0.10 < .001 Log block - 0.17 0.03 .497 L1: English 0.37 0.14 .014 Log block x L1 - 0.11 0.04 .004 Random effects by subject Within - person variance 0.06 2.68E - 3 < .001 Between - person intercept 0.21 0.05 < .001 Between - person slope: log block 0.01 3.10E - 3 < .001 For the English speakers, the log block significantly predicted log Residualized sentence reading time [ b = - 0.13, SE = 0.03, t (23) = - 4.03, p = .001]. However, the main effect of Awareness [ F (1, 23) = 0.02, p = .089] and the interaction between Awareness and log Block 93 [ F (1, 23) = 0.01, p = .097] were not significant (Table 3.4). Thus, there was a significant linear relationship between log Residualized sentence reading time and log Block, which supported the entence reading times decreased according to a power function. However, for the Korean speakers, the main effects of log Block [ F (1, 23) = 0.58, p = .455], Awareness [ F (2, 23) = 0.36, p = .702], and the interaction between Awareness and log Block [ F (2, 23) = 0.22, p = .803] were all not significant (Table 3.5), which indicated that their learning curve on the log - log scale was essentially flat. Table 3.4 Regression output for the English speakers b SE p Fixed effects Intercept 8.02 0.11 < .001 Log block - 0.13 0.03 .001 Awareness: VP - aware 0.02 0.16 .888 Log block x awareness - 1.47E - 3 0.45 .974 Random effects by subject Within - person variance 0.03 1.93E - 3 < .001 Between - person intercept 0.13 0.04 .002 Between - person slope: log block 9.03E - 3 3.49E - 3 .005 94 Table 3.5 Regression output for the Kor ean speakers b SE p Fixed effects Intercept 7.74 0.25 < .001 Log block - 0.05 0.06 .434 Awareness: L1 - aware 5.17E - 3 0.31 .987 Awareness: VP - aware - 0.20 0.31 .530 Log block x L1 - aware 0.05 0.07 .515 Log block x VP - aware 0.03 0.07 .716 Random effects by subject Within - person variance 0.09 5.67E - 3 < .001 Between - person intercept 0.28 0.09 .002 Between - person slope: log block 0.01 5.18E - 3 .022 95 Figure 3.3 The log - log plot for unaware and VP - aware English speakers Figure 3.4 The l og - log plot for unaware, VP - aware, and L1 - aware Korean speakers 96 To sum up, the significant two - way interaction between log Block and L1 and the follow - up analyses showed that the two L1 groups followed a different learning curve. On the one hand, the Engl e power learning curve; clearly discernible pattern. For both L1 groups, there was no significant relationship between awareness of the target syntax and residuali zed sentence reading times. Therefore, this diverging pattern suggested that prior L1 knowledge affected the learning curve under incidental exposure (RQ 1 - 1), regardless of whether participa nts later developed conscious knowledge about the target syntax ( RQ 1 - 2). 3 .3.4 Sensitivity to violations: Integrated knowledge sentence reading times on the testi ng block are presented in Table 3.6 and visualized in Figure 3.5. Figure 3.5 showed that the two L1 groups responded very differently to the violations; the Korean speakers slowed down, whereas the English speakers sped up. Table 3.6 The English speakers sensitivity to violations L1: English ( n = 21) L1: Korean ( n = 21) M SD 95% CI M SD 95% CI Control_pre 497 533 [254, 739] 713 718 [386, 1040] Violation 387 537 [143, 632] 825 847 [440, 1211] Control_post 302 657 [2, 6 01] 625 750 [283, 967] Sensitivity score - 12 483 [ - 231, 208] 157 292 [24, 290] Note . The sensitivity score was computed as RSRTviolation ((RSRTcontrol - pre + RSRTcontrol - post)/2) 97 Figure 3.5 he testing phase To investigate whether prior L1 knowledge would affect the development of sensitivity to violations under incidental exposure, a 2 (L1) x 3 (Block) mixed - design ANOVA was conducte d. The Shapiro Wilk test showed that the distributions of the English and the Korean residualized sentence reading times by block were all normal ( p > .05). This analysis showed the significant main effect of Block [ F (2, 80) = 3.23, p = .045, = 0.08], but the main effect of L1 [ F (1, 40) = 2.74, p = .106, = 0.06] and the interaction between L1 and Block [ F (2, 80) = 1.47, p = .235, = 0.04] were not significant. The post - hoc paired t - tests showed that the significant d ifference stemmed from the difference between the control pre - block and the control post - block [ t (41) = 2.47, p = .018, d = 0.38]. The difference between the violation and the 98 control post - block was marginally significant [ t (41) = 1.98, p = .055, d = 0.35 ] , while the difference between the violation and the control - pre block was not significant [ t (41) = - 0.03, p = .978, d = 0.00]. Therefore, the ANOVA results did not provide evidence of online sensitivity because it suggested that the participants did not s how a robust effect of a slowdown in reading times during the violation block. Additionally, I ran one - sample t - test value (English: M = - 12; Korean: M s core was not significantly different from zero [ t (20) = - 0.11, p = .913, d sensitivity score was positive [ t (20) = 2.46, p = .023, d = 0.54]. However, the latter result should be interpreted with caution because as se e n in the ANOVA results, this positive score was primarily due to the decrease in reading time during the control - post block rather than the increase during the violation block. Thus, it could be concluded that there was some evidence for the Korean speake r sensitivity. The descriptive statistics of the English and the Korean speakers sensitivity to violations by awareness are presented in Table 3.7. To answer research question 2 - 2, w h ich investigated the relationship between awareness and sensitivity, I ran separate analyses on each L1 group. The reason was that each L1 group had different levels of Awareness (English: 2 levels; Korean: 3 levels). For the English experimental group, a 2 (Awareness) x 3 (Block) mixed - design ANOVA Wilk test, the distributions of the residualized sentence reading times for each block were all normal ( p > .05). The a nalysis showed that the main effect of Block [ F (2, 38) = 1.86, p = .170, = 0.09], awareness [ F (1, 19) = 0.14, p = .712, = 0.01], and the block by awareness 99 interaction [ F (1, 19) = 0.01, p = .918, = 0.00] were all not significa nt, which meant that there their sensitivity to Koreanish word order violations. Table 3.7 The and the sensitivity to viol ations by awareness Control_pre Violation Control_post Sensitivity score n M SD M SD M SD M SD L1: English Unaware 12 514 563 470 552 312 642 57 503 VP - aware 9 473 523 276 528 287 715 - 104 469 L1: Korean Unaware 4 630 644 296 196 413 362 - 225 22 9 VP - aware 8 472 643 566 714 275 665 192 264 L1 - aware 9 964 871 1292 939 1030 802 295 191 For the Korean experimental group, the Shapiro Wilk test indicated that the distributions of the residualized sentence reading times for each block were all nor mal ( p > .05), except for the unaware group - post block. A 3 (Awareness) x 3 (B lock) mixed - design ANOVA showed a significant main effect of Block [ F (2, 36) = 3.58, p = .038, = 0.17], which was qualified by an Awareness by Block interaction [ F (4, 36) = 2.94, p = .033, = 0.25] (Figure 3.6). The main effect of Awareness was not significant [ F (2, 18) = 2.25, p = .134, = 0.02]. To follow up on the two - way interaction, I analyzed the simple effect of Block for each 100 Awareness group separately. The follow - up tests revealed a significant effect of Block for th e L1 - aware group [ F (2, 16) = 4.87, p = .022, = 0.38] but not for the VP - aware group [ F (2 , 14) = 2.75, p = .098, = 0.28] and the unaware group [ F (2, 6) = 1.96, p = .221, = 0.40]. Although there was no significant effect of Block for the unaware group, the effect size was large ( = 0.40), which may indicate a power issue due to a small sample size ( n = 4). In fact, Table 3.7 and Figure 3.6 suggested that the Korean un sentence reading times showed the opposite tendency as the VP - aware and the L1 - aware groups during the testing phase. The una their reading time for the co ntrol_pre (630 ms) and the control_post (413 ms) blocks, whereas the L1 - aware (1292 ms) and the VP - n block w ere slower than the ir reading times for the control_pre (L1 - aware: 964 ms; VP - aware: 472 ms) an d the control_post blocks (L1 - aware: 1030 ms; VP - aware: 275 ms). For the L1 - aware group, the mean sentence reading time differences between the violatio n block and the control pre - block (328 ms) as well as between the violation block and the control - post b lock ( - 262 ms) were statistically significant (control - pre block: t (8) = 3.22, p = .012, d = 1.08; control - post block: t (8) = - 3.05, p = .016, d = 1.01) . These significant differences indicated that the L1 - aware group increased their reading times for ungrammatical Koreanish sentences and then decreased their reading times for grammatical Koreanish sentences. Although the L1 - aware group did not be come awa re of the relevance of the L1 until the grammaticality judgment posttest and, hence, was unaware of the relationship between Koreanish and their L1 (Korean) during reading, they unconsciously exhibited a large amount of sensitivity during the criti cal viol ation block. Moreover, the L1 - aware participants were the slower, more careful readers at a descriptive level (see Table 3.7 and Figure 3.6). 101 Figure 3.6 Two - way interaction between B lock and A wareness for the Korean speakers 3 . 3 .5 Summary of results T pattern. The English speakers did not show sta tistically reliable se nsitivity to word order violations. T he Korean L1 - aware group members, who later became aware of the cross - linguistic similarity between the target language an d their L1, showed rob ust sensitivity to syntactically ungrammatical sentences in the input. The Korean VP - aware group showed a response pattern that was consistent with the L1 - 102 aware group. The Korean unaware group showed a qualitatively different respons e. 3 . 4 Discussion U sing an eye - tracking methodology in this study, I demonstrated the effects of prior L1 knowledge on online learning (i.e., a decrease in sentence reading times) and knowledge (i.e., sensitivity to violations) of nonnative syntax. The eye - movement data reve aled an intriguing interaction between L1 knowledge and online L2 development. On the one hand, changes in the the training phase; however, these participants fa iled to show sensitivity to violations during the times did not fit the power learning curve, but the Korean L1 - aware group showed sensiti vity to word order vio lations and the Korean VP - aware group (but not the unaware group) showed a on the two online measures captured multifaceted aspects of L1 influence on L2 dev elopment, uncovering the complex nature of linguistic transfer, variability in L2 development, and the moderating role of awareness in meaning - focused, incidental learning conditions. Residualized sentence reading times reflect the amount of extra atten tion needed for processing Koreanish input, and changes in this measure could reveal online learning processes. the power learning curve (De Keyser, 1997), which i s known to be a hallmark of domain - general learning (Anderson, 1982, 1993; Newell & Rosenbloom, 1981). Given that the power function has been commonly found in intentional learning contexts (DeKeyser, 2015), I extended 103 DeKeyser (1997) by observing the the power function even under incidental learning conditions. Ahn, et al., 2018; Indrarathne et al., 2018) likely reflects a gradual withdrawal of attentional proces sing. Regardless of whether they obtained conscious rule knowledge (i.e., awareness), the English speakers showed a lower need for controlled attentional processing over time. This gradual removal of attention can be taken as an indic ation of increased pro cessing efficiency. - reported reading patterns enabled me to infer how they controlled their attentional processing. In the debriefing interview, I asked the participants what they were thinking when r eading the assigned no vel and how they read the sentences with the words scrambled. The analyses of their answers led me to conclude that twice the number of English speakers ( n = 19) than Korean speakers ( n = 9) tended to use self - developed reading strate gies for comprehension . Their verbal reports revealed two broad types of reading strategies meaning - oriented and form - oriented ones. First, four English speakers used a meaning - oriented strategy by focusing on keywords in the sentences. They reported that they attention to important information to understand the content of the story. Second, mor e than half of the English speakers (15 of 23) used a form - focused strategy by searching for the subject and the verb in each sentence to rearrange the structure. After realizing that the verb was placed at the end of the sentence, th form - focused strategy could have contributed to the acceleration of their reading sp eed (as they became fa ster at locating the verb and the subject), it also suggested that they were still 104 processing the Koreanish sentences based on their L1 (English) representation. A similar L1 - based parsing tendency was also shown in Onnis and Thiessen er, it also revealed that the incidental, meaning - focused condition elicited some conscious, deliberate processing strategies as well. This might be the reason that the power law of practice was applicable even in the incidental conte xt. Crucially, the E nglish speakers failed to show sensitivity to word order violations, suggesting that they had not integrated their new syntactic knowledge yet. Incidental exposure from reading a novel was probably insufficient for English speakers to develop a complete se t of integrated, implicit knowledge (Williams & Kuribara, 2008), primarily due to the dissimilarity between English and Korean word order. Because the Koreanish word order had exactly the reverse pattern of their L1 English, they woul d perhaps need more ex plicit instruction, such as rule - search instruction or metalinguistic explanations, to master the L2 rules that differed from their L1 rules (Indrarathne & Kormos, 2017). For example, a recent study (McManus & Mardsen, 2019) found tha t L2 practice, combine d with explicit instruction about the L1 L2 differences in form - meaning mapping, significantly facilitated L2 speed and accuracy, which pointed to the pedagogical value of instruction on the L1 L2 differences for L2 grammatical develo pment. In this sense, L2 teaching would benefit greatly from an in - depth understanding of the L1 exist. In other words, in an L2 classroom, it should be kept in mind t hat the starting point of L2 acquisition is the L1. sentence reading times did not fit the power function. What then do the changes in their sentence reading times indic ate? A visual represen 105 reading times by block (Figure 3.7) suggests that the Korean speakers showed a sharper initial decrease during the early stage of learning compared with the English speakers. Figure 3.7 The English and To test the group difference shown in the early stage of reading, I conducted an additional 2 (L1) x 24 (Block) mixed ANOVA on log - transformed residualized sentence reading times. 6 T he 6 Log - transformation was used to meet the normality assumption. According to the Shapiro Wilk test, the distributions of both the English and the Korean sp p < 0.01). However, with the log - ed sentence reading time was normal ( p = p < 0.01). 106 results 7 indicated that the Korean speakers read faster than the English speakers during the very early stage of reading the novel (from Sentence 20 to Sentence 60), but after that, reading times for the two groups could not be distinguished statistical ly. Thus, above all, the Korean speakers showed a steeper initial decrease than the English speakers did. knowledge of the Korean word order. Gi ven that the Korean speake rs in this study were Korean - English bilinguals, they possibly could have spontaneously accessed their L1 (Korean) while reading the vocabulary in L2 English because among bilinguals, both their languages are known to be automatic ally co - activated even in a task performed in one language (Hatzidaki, Branigan, & Pickering, 2011; Sanoudaki & Thierry, 2015). Perhaps the cross - language activation helped the Korean speakers become familiar with the Koreanish syntax more rapidly than the English speakers, acceler ating their speed of reading Koreanish sentences. However, despite the sharp drop - off in their reading times, they did not show a consistent and gradual decrease for the rest of their reading as the English speakers did. Rather, t hey showed subsequent peak s and valleys in processing patterns could be accounted for by the representational redescription model ( Karmiloff - Smith, 1992). Acco rding to this developmenta l framework, when learners build their abstract rule knowledge from the procedural representation (i.e., Explicit Level 1), they tend to make more errors as the knowledge is redescribed into a more efficient format. During this 7 The analyses showed a significant main effect of Block [ F (9.38, 412.53) = 5.84, p = .020, p 2 = 0.12], which was qualified by a significant L1 by Block interaction [ F (9.38, 412.53) = 2.63, p = .005, p 2 = 0.06]. The simple effect test of L1 for each Block showed that this effect was significant only for Block 2 [ F (1, 44) = 4.52, p = .039, p 2 = 0.09] and Block 3 [ F (1, 44) = 6.20, p = .017, p 2 = 0.12]. 107 pe riod, their performance ma y present a U - shaped learning pattern. In this regard, it could be error - correcting behaviors, which led them to formulate unconscious st ructural knowledge. Anot her notable finding of this study is that the Korean speakers, especially the L1 - aware group, exhibited integrated knowledge of Koreanish as reflected in their sensitivity to word order violations. The L1 - aware group members, who later recognized the c ross - language similarity after finishing the exposure task, already showed an automatic slowdown in their reading times for violations prior to that. Such a sensitivity effect indicated that they developed integrated knowledge of the Koreanish word order u nder incidental exposure conditions. It is notable they did not report any awareness of the violations in their verbal reports, which suggest that their slowdown in processing of violations might have taken place outside awareness; therefore, it can be arg ued that their integrated knowledge was also implicit and unconscious. The Korean L1 - - established L1 (Korean), which could be taken as evidence of unconscious L1 transfer . Th ey in turn developed metacognition of the cross - language similarity, which I referred to as L1 awareness in this study. In this way, the L1 - interesting connection between unconscious and conscious access to pri or L 1 knowledge. This finding appears to be in line with the radical plasticity theory proposed by Cleeremans (2007, 2011). According to Cleeremans (2011), unconscious knowledge, which can give rise to sensitivity effects, will shape higher - order conscious ness as a result of learning. When the brain unconsciously processes information and learns about its own unconsciously accrued representations, one will eventually go through conscious experience of the unconscious knowledge and become aware of what we kn ow. This theory could explain the L1 - 108 performance; the ones who showed evidence of unconscious access to Koreanish (i.e., integrated knowledge) subsequently gained conscious access to the metacognitive knowledge of the Korean Koreanish relatio nshi p. In this way, the L1 - interesting progression from unconsciousness to consciousness. T he methodological approach in this study resulted in a more solid understanding of the nature of grammar development. The us e of control groups was advantageous , because they provided baseline eye - tracking data, which enabled the researcher to more accurate ly assess extra processing times. This methodological feature enabled me to employ a within - subject design (revealing how i ndiv between - subjects design (showing how much extra reading time the experimental groups needed when compared to the control groups). In addition, triangulating online measures with awareness me asur es was highly informative , particularly when making infer ences about cognitive processes in incidental learning conditions (Rebuschat, 2013; Rebuschat et al., 2015). This triangulation allowed me to probe the complex, multi - faceted effects of prior L1 know ledge during processing , which could not have been possible with traditional accuracy measures. 109 CHAPTER 4 GENERAL DISCUSSION AND CONCLUSION 4 . 1 Summary of the F indings 4 . 1 . 1 English speakers Online measures. During the exposure task, the Engli residualized sentence reading times followed the power function. This learning curve, which was governed by the power law of practice, was indicative o f learning processes that involved proceduralization and automatization (Anderson, 1982, 1993; DeKeyser, 1997, 2015). grammatical sensitivity was not statistically s ignificant (no slowdown). This suggested that, in spite of their readin g time gains, the English speakers did not show evidence of integrated knowledge of head - final word order. There were no significant differences in reading patterns or grammatical sensi tivity for the awareness subgroups. Offline measures. The analysis of the English experimental group was able to acquire the target grammar after incidental exposure through a novel. The English experimental group exhibited strong evidence of explicit knowledge; however, ev idence of their implicit knowledge was weak and limited . The effect of awareness was not significant for syntactic knowledge development. 4 . 1 . 2 Korean speakers On line measures. During the exposure task, the Korean experimental group showed a strong init ial decrease in residualized sentence reading time; however, their changing reading times did not follow a clearly discernible pattern. Crucially , there was a sugges tion of the Korean integrated, implicit 110 knowledge of the target syntax. T he Korean experimental group tended to unconsciously slow down their reading during the violation block . Inter grammatical sensitivity differed depending on the levels of awareness. T he L1 - aware participants , who later became aware of the cross - linguistic similarity between the target language and their L1, showed robust se nsitivity to syntactically ungrammatical sentences in the input. The Korean VP - aware group showed a response pattern that was consistent with the L1 - aware group. The Korean unaware group showed a qualitatively different response. Offline measures. On the GJT, the Korean experimental group showed robust learning of the target grammar after inci dental exposure through a novel. Furthermore, a fter two weeks, to outperfor m the English experimental group on the delayed GJT. This enhanced knowledge consolidation was led by the Korean L1 - aware group, who se members noticed during the immediate GJT (not during reading) that the target grammar matched their L1 grammar, Korean. T wo weeks later, at delayed testing, these participants judged sentences with 87.22 % accurac y, compared to 58.91 % accuracy for the English speakers. Notably, u nlike the English experimental group, the Korean experimental group acquired not only explicit but also implicit knowledge. The Korean experimental group clearly demonstrated their advantag e over the English experimental group in developing both implicit and explicit knowledge. This became apparent in the triangulation of GJT scores with the two offline awareness measures. First, a triangulation of GJT performance with retrospective verbal r eports showed that the Korean unaware group developed more implicit knowledge than the Eng lish unaware group. Second, the analysis of GJT performance as a function of source attributions revealed that the Korean experimental group developed more explicit k nowledge 111 than the English experimental group. Table 4.1 summarize s the findings from both the online and offline measures for the two participant groups . Table 4.1 Summary of the findings L1 Decrease in reading times Power law of practice Sensitivity Grammaticality judgments Implicit Explicit English Korean (L1 - aware) 4 . 2 General D iscussion I n the present dissertation, I observed stark o verall differences between the English and language learning task. T he e ffect of prior L1 knowledge on the acquisition of L2 syntax was pervasive throughout the experiment. During the exposure task (reading a novel), the Korean group exhibited a faster initial decrease in sentence - reading times than did the English group. Further more , only the Korean group showed grammatical sensitivity while reading as evidenced by increased sentence - reading times in response to syntactic violations. Subsequently , on the testing task (GJT), the Korean experimental group demonstrated reliab le implicit learning and knowledge across time while the English experimental group did not. Moreover, the Korean experimental group acquired more explicit knowledge than did the English experimental group, largely through their awareness of the con nection between their L1 and the target language. L1 awareness , the conscious perception of cross - similarity between the L1 (Korean) and the target Kore a nish 112 syntax , contributed to memory consolidation over time. In this sense, the Korean experimental grou p consi stently had advantages over the English experimental group. In particular, the triangulati on of diverse measures uncovered intriguing patterns of convergence as well as divergence in terms of how speakers with different L1 backgrounds interacted with t he task . An interesting pattern of convergence concerns the online and offline knowledge measures . T he Korean speakers had advantages over the English speakers on both the online violation block and the offline GJT . This pattern supported the idea t hat cro ss - language similarity boosts L2 learning (Ringbom, 2007). On the online knowledge measure, the Korean experimental group, especially those who later noticed the cross - language similarity between their L1 and the target language (L1 - aware participan ts), sh owed evidence of grammatical sensitivity to violations, whereas the English experimental group did not. This may reflect the Even before awareness of the simila rity between L1 and the target language emerged, the L1 - aware group reliably showed evidence of implicit knowledge of the target word order. On the offline knowledge measure, the Korean experimental group outperformed the English experimental group , and it exhibited strong evidence of implicit knowledge while the English experimental group did not ( as seen in the comparison of the English and Korean unaware s as well as in the above - chance performance on grammaticalit y judgments attributed to guess and intuition across testing sessions ). Moreover, the L1 - aware group excelled over all others and showed excellent performance on the GJT based primarily on their conscious knowledge of L1 - L2 cross - linguistic similarities. I n this way, both knowledge measures (sensitivity and GJTs) s howed that the Korean L1 - aware group led the way in the acquisition of the head - final target syntax . These converging findings from both the online 113 and off line measures provide compelling evidence of the significant L1 influence on the development of implicit syntactic knowledge under incidental exposure . The Korean speakers certainly had advantages in acquiring the target word order, which followed their L1 Korean, and the conscious registration o f L1 - awareness gave them an edge in knowledge development. However, unlike the knowledge measures, the learning measure the learning curve revealed a different picture. That is, the lea r ning and knowledge measures diverged in terms of L1 influence. Altho ugh both the Korean and English experimental groups sped up over time during reading the exposure text (i.e., evidence of learning), their learning curves had a different ng data followed the power function (Ande rson, not. This was quite an unexpected but very interesting diverging pattern. This divergence could be somewhat resolved if we posit that the English experimental gr was a reflection of automatization of explicit knowledge of the target word es followed the power function, it is pla usible that incidental exposure through a novel was insufficient for them to integrate the target syntactic knowledge into their mental representation. Therefore, on the violation block at the end of the reading tas k, they were not sensitive to the word or der violations. Similarly , they were unable to show evidence of implicit knowledge on the GJT; however, they reliably showed a fair amount of explicit knowledge on the GJT. Such evidence of explicit knowledge on the GJT may support the idea that th e Englis h speakers were automatizing declarative, explicit knowledge of Korean word order during learning. It may be that without the guidance of prior L1 knowledge, they needed more exposure and practice to develop unconsc ious grammatical reflexes to word order v iolations. Further more , one possib le 114 explanation for did not show a power function is that there was more heterogeneity in this group. However, this account must remain speculative for now . The s ubgroup analyses for the th ree awareness groups did not show marked differences between the subgroups, but there may be other sources of variation that account for the atypical learning curves. Future researchers could explore this possibilit y in greater depth. Changes in the Kore (steeper decrease) at the initial stage of reading. Moreover, their L2 development showed a progression from unconscious, implicit knowledge to conscio us, explicit knowledge (Cleeremans, 2007, 2011; Karmiloff - Smith, 1992), which was seen in its full form in the L1 - aware subgroup. Conversely, the English speakers, who did not have relevant prior knowledge, started off with explicit knowledge and fell shor t of achieving implicit knowledge (Jiang, 2007, 2011; Keating, 2009; Tokowicz & MacWhinney, 2005). Thus, the two L1 groups evidenced different learning trajectories, owing most likely to the prior L1 knowledge that they brought to the task. The findings of th is dissertation demonstrate the sign ificant and durable effects of prior L1 knowledge on the incidental acquisition of L2 syntax. First, L1 syntactic transfer occurred both consciously and unconsciously under incidental learning conditions ; i ndeed, 15 of the 24 Korean experimental participan ts remained unaware of the cross - language similarity in the experiment. Second, L1 syntactic transfer occurred at both online and the offline knowledge measures. Third, L1 awareness, which is the availabili ty of conscious, metacognitive knowledge of the L1 / L2 relationship, functioned as a catalyst for the development of implicit and explicit L2 syntactic knowledge. Thus, a reasonable conclusion of the present dissertation is that L2 learners with different L 1 backgrounds perform on an un equal footing due to their prior L1 experience, 115 especially in naturalistic learning contexts. The two L1 groups, in fact, presented oppos ing direction s of knowledge development. The Korean participants who had relevant prior L 1 knowledge progress ed from implicit to explicit k nowledge under incidental exposure while the English participants who lacked relevant prior L1 knowledge began with explicit knowledge and failed to reach implicit knowledge. 4 . 3 Limitation There were some limitations in the present study. First, the s ample size for the awareness subgroups was quite small. For example, I could not have anticipated that there would be five participants in the Korean unaware subgroup. This could not be control led for because I was not able to predict how many participant s would become aware of the rules prior to the actual experiment. Therefore, future research with a larger number of participants would be needed to confirm the findings regarding the relationsh ip between awareness and syntactic knowledge development. Seco nd, the two L1 groups had different levels of familiarity with the vocabulary that made up the Koreanish text. English speakers were reading the Koreanish text in their L1, English, whereas the Korean - English bilinguals, though highly fluent in English, we re reading the words in the text in their L2. This meant their speed of lexical processing was likely slower and some English vocabulary may have been unfamiliar to them. Any difficulties in com prehending the text, however, would have been relatively small , considering the two experimental groups performed on a par on the reading comprehension tests. Third, we cannot exclude the possibility of the materials effects for the reading task. Unlike se ntence - processing experiments that allow tight control over th e input, the order of the sentence presentation for the reading material in the study the unmodified, authentic novel was not random and the 116 possibilities for counterbalancing the text were limi ted. In other words, every participant read the sentences in a predetermined, fixed order during the exposure phase. Such fixed order and limited counterbalancing of the sentences could have introduced noise in the data that may have obscured some of the l earning effects. Hence, the nonsignificant effect of power law and the flat slope for the log - log plot for the Korean speakers (see Figure 3.4) should be interpreted with caution. Given that increased noise levels may be a trade - off when using authentic ma terials, future research should attempt to seek an optimal bal ance between ecological validity and experimental control depending on its research goals. 4 . 4 Conclusion and O utlook I introduced notable methodological advances in the present dissertatio n. First, I used an authentic novel to incidentally expose par ticipants to the target semi - artificial language , creating a more ecologically valid research context . A coherent, longer text from the novel was likely to ensure a stronger participant focus on meaning , and it stimulated more natural reading processes tha n isolated sentences or unconnected short texts. Second, I employed both online and offline measures : eye movements as an online measure of real - time processing and knowledge a s well as gramm aticality judgments as an offline knowledge measure. This allowed me to triangulate online measures d erived from eye movements (residualized sentence - reading times and grammatical sensitivity) with offline knowledge measures ( grammaticality judgments) and with awareness measures (source attributions, retrospective verbal reports). Third, I used delayed te sting to track changes in knowledge development over time. These methodological components contributed to a comprehensive exploration of the pervasive and durable transfer effects of L1 in L2 syntax 117 acquisition. Overall, the L1 Korean (head - final) speakers had crucial advantages over the L1 English (head - initial) speakers when learning a new head - final language under meaning - focused, incidental learning cond itions. L1 syntactic transfer occurred at both the conscious and unconscious levels , and L1 syntactic transfer occurred both online and in the offline knowledge measures. L1 - awareness, defined as the availability of conscious, metacognitive knowledge of th e L1 - L2 relationship, aided the development of implicit and explicit L2 syntactic knowledge. Despite the extensive amount of input, the English speakers, who lacked the guidance of cross - language similarity , did not show strong learning effects. In light of this, future researchers could incorporate different treatment conditions, such as enhancement and explicit instruction, to examine whether such treatments can diminish the L1 influence on the development of L2 syntactic knowledge (for example, see McMa nus & Marsden, 2019). T he interaction between treatment conditions and L1 influence certainly merits further investigation. Moreover, in follow - studies, it would be interesting to examine individual differences in cognitive style and personality to see who becomes aware and who does not. Further, given that the present dissertation employed only receptive tasks as a knowledge measure, future research would benefit from using both receptive and productive tasks to expand the scope of observation concerning i ncidental learning. Another The in - depth investigation of the effect of prior L1 knowledge on implici t and explicit L2 learning will help researchers and educators both to better understand the variations in L2 acquisitional processes and to design tailore d instructions for diverse learner populations. 118 APPENDICES 119 Appendix A: Sample of exposure task (novel reading) Task instruction Welcome to the study! Thank you for your participation. In this experiment you will read a detective novel. The goal is to read this text as natural as possible. You are going to read the first two chapters of the detective novel. For some of you, the sentences will be presented with the words scrambled. There will be practice screens at the beginning, which will be followed by a short break, and a break approximately every 15 screens afterwards. You will have six breaks throughout the experiment. You will be asked to answer simple comprehension questions at every other break. There will be 92 screens i n total. To move from one screen to the next, press any button on the controller in front of you. Try not to move your head during the experiment please. Before each slide, you will be asked to look at a dot in the upper left - hand corner of your screen to continue. If you need to stop at any time during the experiment, please stop only when the dot is in the upper left - hand corner of the screen and notify the researcher. Sample of the p ractice phase There used to be intense public interest in what was kn people have now begun to forget about it. Nevertheless, because of that previous interest, both by my friend Poirot and the family themselves asked me to write an account of the whole story. This, we all hope, will silence the sensational rumors that still persist. I will therefore briefly write of the circumstances which led to my being involved in the affair. 120 Sample of the t raining phase I injury due to the b attlefield from home had been sent. And, rather a depre ssing rehab facility in John Cavendish ran across when what do to my mind make up to was trying. I the past few years in him much had not seen. Indeed, I him part icularly well never had known. He his forty - five years than younger looked though, he me than a good fifteen years was older. A boy as I Essex Sample of the violation block I was anxious John to get a hol d of, but he was nowhere to be seen. Evidently had something very momentous that afternoon occurred. I tried the few words to forget I had overheard. But, no matter how much I tried, I could them not altogether from my mind dismiss. What was Mary Cavendis h's concern in the matter? Mr. Inglethorp was in the parlor, when I to supper down came. His face was impassive as ever, and the strange unreality of the man struck me again. Mrs. Inglethorp came last down. She looked still agitated, and during the meal wa s there a somewhat awkward silence. As usual, surrounded he his wife with little attentions, placing her a cushion at back, and playing altogether the role of the devoted husband. 121 Appendix B: Sample of comprehension check questions 1. The narrator (Hastings) is recovering from an injury. ( True False ) 2. 3. John Cavendish has a younger sister name d Jessica. ( True False ) 4. ( True False ) 5. Evie Howard is a delicate and pale lady . ( True False ) 6. endish, is a beautiful woman. ( True False ) 7. The group eats outside. ( True False ) 8. Alfred Inglethorp is a welcoming and charming host. ( True False ) 122 A ppendix C : Sample of the grammatical ity judgment test The scrambling of the sentences in the novel (Task 1) was not arbitrary but followed a complex system. This was actually based on the grammar of a real language. In the second part of the experiment, you will read 80 new sentences. 40 of these sentences were generated by means of the same language grammar as in t he previous part of the experiment. The sentences follow the system presented in the novel. These sequences are called GRAMMATICAL. The other 40 sentences were generated randoml y, i.e., they do NOT confirm to the same language grammar presented in the no vel. The sentences do not follow the system presented in the novel. These sequences are called UNGRAMMATICAL. Your task is to decide which ones are grammatical and ungrammatical. Try to respond as quickly and accurately as possible. Simply pick whatever c omes to your mind first. In addition to deciding on whether each new sentence follows the complex system of the previous sentences, we will also ask you what the b asis of your decision was. * 1: GUESS Your decision was based on a true guess, i.e., you might as well have flipped a coin. * 2: INTUITION Your decision was based on intuition, i.e., you feel that your decision is correct but you have no idea why. * 3: RECOLLECTION Your decision was based on the recollection of specific sentences (or parts of sentences) that you have read before that seemed similar; or you failed to recollect a specific sentence that was similar. * 4: RULE KNOWLEDGE Your decision w as based on rule knowledge, i.e., you followed a rule when making the decision and you are able to describe the rule at the end of the experiment. From now on, you will see two practice items with the researcher first. If you have any questions, please le t the researcher know. [Practice] 123 1. My student Spanish learned. ( ) grammatical ( ) ungrammatical What is the basis of your decision? Please write down the number to give your response (1: Guess, 2: Intui tion, 3: Recollection, 4: Rule knowledge) _______________________ 2. The teacher who ask ed many questions the student liked. ( ) grammatical ( ) ungrammatical What is the basis of your decision? Please w rite down the number to give your response (1: Guess, 2: Intuition, 3: Recollection, 4: Rul e knowledge) _______________________ Now you are ready to start the main part of Task 2. Your task is: 1) to decide whether it is grammatical (follows the word order system presented in Task 1) or ungrammatical (does NOT follow the word order system presented in Task 1) , 2) and to report what the basis of your judgment was. Don't forget to leave your answer in the comments. If you have any questions, plea se let the researcher know. If not, please proceed. 1. Susan the table touched. ( ) gra mmatical ( ) ungrammatical What is the basis of your decision? Please write down the number to give your response (1: Guess, 2: Intuition, 3: Recollection, 4: Rule knowledge) _______________________ 2. Until Meg ( ) grammatical ( ) ungrammatical What is the basis of your decision? Please write down the number to give your response (1: Guess, 2: Intuition, 3: Recol lection, 4: Rule knowledge) _______________________ 124 3. Julie Mr. Lee the prize won that believed. ( ) grammatical ( ) ungrammatical What is the basis of your decision? Please write down the number to give y our response (1: Guess, 2: Intuition, 3: Recollection, 4: Rule knowledge) _______________________ 4. Betsy the coffeeshop at waited. ( ) grammatical ( ) ungrammatical What is the basis of your decision? P lease write down the number to give your response (1: Guess, 2: Intuition, 3: Recollection, 4: Rule knowledge) _______________________ 5. My father his driveway shoveled who the neighbor thanked. ( ) grammatical ( ) ungrammatical What is the basis of your decision? Please write down the number to give your response (1: Guess, 2: Intuition, 3: Recollection, 4: Rule knowledge) _______________________ 125 Appendix D : Sample of the debriefing interview questions [Session 1: short version of the debriefing interview] 1. While reading the novel, did you notice any rules or regularity? If yes, please indicate what you have noticed and during when you noticed it. 2. Please describe what you were thinking d uring reading. 3. Did you notice anything odd at the end of the novel? If yes, please indicate what you believe you have noticed. [Session 2: long version of the debriefing interview] 1. Have you e ver indicated rule knowledge as a source as a basis for your decisions? If so, please describe what you had been thinking. Why did you select rule knowledge category? If not, please share any other ways in which you made your choices. 2. As mentioned in th e experiment, the scrambling of the sentences was not a rbitrary. Instead, the word order in the sentences was based on a complex system. Reflecting now specifically on the placement of words within the sentences, can you recall any specific rule or regular ity? 3. In Session 1, how did you read the sentences wi th the words scrambled? P lease describe your own reading processes. 4. In Session 1, did you try to find a pattern from the scrambled word order during reading? Did you focus on the word order during reading? If so, please describe how you tried to find t he pattern . 5. How much do you like reading a detective novel? 126 6. How many novels do you read a year? study? 127 REFERENCES 128 R EFERENCES Anderson, J. R. (1982). Acquisition of cognitive skill. Psychological review , 89 (4), 369. Anderson, J. R. (1993). Rules of the mind . Hillsdale, NJ: Erlbaum. Anderson, J. R., & Schooler, L. J . (1991). Reflections of the environment in memory. Psychological science , 2 (6), 396 - 408. Andringa , S. (in press, 2020). The emergence of awareness in uninstructed L2 learning: A visual world eye tracking study. Second Language Research . Andringa, S. J., & Rebuschat, P. (2015). New directions in the study of implicit and explicit learning. Studies in Second Language Acquisition , 37, 185 196. Batterink, L. J., Oudiette, D., Reber, P. J., & Paller, K. A. (2014). Sleep facilitates learning a new linguistic rule . Neuropsychologia , 65 , 169 - 179. Bialystok, E. (1994). Analysis and control in the developmen t of second language proficiency. Studies in second language acquisition , 16 (2), 157 - 168. Brennan, J., Nir, Y., Hasson, U., Malach, R., Heeger, D. J., & Pyl kkänen, L. (2012). Syntactic structure building in the anterior temporal lobe during natural story listening. Brain and language , 120 (2), 163 - 173. Brooks, P. J., & Kempe, V. (2013). Individual differences in adult foreign language learning: The mediating effect of metalinguistic awareness. Memory & cognition , 41 (2), 281 - 296. Chomsky, N. (1981). Lectures on government and binding . Dordrecht: Foris. Christie, A. (1920). The mysterious affair at S tyles . N.p.: John Lane. Retrieved on 11 October, 2017 , from P roject Gutenberg, www.gutenberg.org Cintrón - Valentín, M., & Ellis, N. C. (2015). Exploring the interface: Explicit focus - on - form instruction and learned attentional biases in L2 Latin. Studies in Second Language Ac quisition , 37 (2), 197 - 235. Clahsen, H., & Felser , C. (2006). Grammatical processing in language learners. Applied psycholinguistics , 27 (1), 3 - 42. Conway, C. M., & Christiansen, M. H. (2006). Statistical learning within and between modalities: Pitting abs tract and against stimulus - specific represntation s. Psychological Science , 17 , 905 912. Cop, U., Dirix, N., Drieghe, D., & Duyck, W. (2017). Presenting GECO: An eye - tracking corpus 129 of monolingual and bilingual sentence reading. Behavioral Research , 49 , 6 02 615. Cop, U., Dirix, N., Van Assche, E., Dri eghe, D., & Duyck, W. (2017). Reading a book in one or two languages? An eye movement study of cognate facilitation in L1 and L2 reading. Bilingualism: Language and Cognition , 20 (4), 747 769. Cop, U., Dri eghe, D., & Duyck, W. (2015). Eye movement patter ns in natural reading: A comparison of monolingual and bilingual reading of a novel. PLoS ONE , 10 , e0134008. doi:10.1371/journal.pone.0134008 Cop, U., Keule ers, E., Drieghe, D., & Duyck, W. (2015). Frequenc y effects in monolingual and bilingual natu ral reading. Psychonomic Bulletin & Review, 22, 1216 1234. Cleeremans, A. (2007). Consciousness: the radical plasticity thesis. Progress in brain research , 168 , 19 - 33. Cleeremans , A. (2011). The radical plastici ty thesis: how the brain learns to be conscious. Frontiers in psychology , 2 , 86. Curcic, M., Andringa, S., & Kuiken, F. (2019). The role of awareness and cognitive aptitudes in L2 predictive language processing. Language L earning , 69 , 42 - 71. Doughty, C. (2003). Instructed SLA: Constraints, compensation, and enhancement. In C. Doughty & M. H. Long (Eds.), The handbook of second language acquisition (pp. 256 - 310). Oxford, UK: Blackwell Publishing. DeKeyser, R. M. (1997). Be yond explicit rule learning: Auto matizing second language morphosyntax. Studies in second language acquisition , 19 (2), 195 - 221. DeKeyser , R. (2007). Skill acquisition theory. In B. VanPatten & J. Williams (Eds.), Theories in second language acquisition (pp. 97 112). Mahwah, NJ: Erlbaum. Denhovska, N., Serratrice, L., & Payne, J. (2016). Acquisition of second la nguage grammar under inciden tal learning conditions: The role of frequency and working memory. Language Learning , 66 (1), 159 190. Denhovska, N., & Serratrice, L. (2017). Incidenta l learning of gender agreement. Journal of Psycholinguistic Research , 46 , 1187 1211. DeYoung, C. G., Q uilty, L. C., & Peterson, J. B. (2007). Between facets and domains: 10 aspects of the Big Five. Journal of Personality and Social Psychology , 93 (5), 880 - 896. doi:10.1037/0022 - 351493.5.880 Dienes, Z. (2004). Assumptions of subjective meausres of unconsciou s mental states: Higher order thoughts and bias. Journal of Consciousness Stu d ies , 11 , 25 45 . Dienes, Z. (2008). Subjective measures of unconscious kno wledge. Progress in Brain Research , 130 168, 49 64. Dienes, Z., & Scott, R. (2005). Measuring unconscious knowledge: Distinguishing structural knowledge and judgment knowledge. Psychological Research , 69 , 338 351. Dienes, Z., & Perner, J. (1999). A theory of implicit and explicit knowledge. Behavioral and brain sciences , 22(5), 735 - 808. Dussias, P. E. (2010). Uses of eye - tracking data in second language sentence processing research. Annual Review of Applied Linguistics , 30 , 149 - 166. Elgort, I., Brysbaert, M., Stevens, M., & Van Assche , E. (2018). Contextual word learning during reading in a second language: An eye - movement study. Studies in Second Language Acquisition , 40 (2), 341 - 366. Elgort, I., & Warren, P. (2014). L2 vocabulary learning from reading: E xplicit and tacit lexical know ledge and the role of learner and item variables. Language Learning , 64 (2), 365 - 414. Ellis, N. C. (1996). Sequencing in SLA: Phonological Memory, Chunking and Points of Order. Studies in Second Language Acquisition , 18 , 91 - 12 6. Ellis, N. C. (1998). Emergentism, connectionism and language learning. Language learning , 48 (4), 631 - 664. Ellis , N. C . (2006) . SLA: The Associative Cognitive CREED. In B. VanPatten & J. Williams; A. F. Williams (Eds.). Theories in second language acqu isition: An introduction . Mahwah NJ: Erlbaum . Ellis, N. C., & Schmidt, R. (1997). Morphology and longer distance de pendencies: Laboratory research illuminating the A in SLA. Studies in Second Language Acquisition , 19 (2), 145 - 171. Ellis, N. C., & Sagarra, N. (2010). Learned attention effects in L2 temporal reference: The first hour and the next eight semesters. Languag e Learning , 60 , 85 - 108. Ellis, N., & Sagarra, N. (2011). Blocking and learned attention in language acquisition: A replication and generali zation study. Studies in Second Language Acquisition , 33 , 589 624. Ellis, N. C., Hafeez, K., Martin, K. I., Chen, L., Boland, J., & Sagarra, N. (2014). An eye - tracking study of learned attention in second language acquisition. Applied Psycholinguistics , 3 5 (3), 547 - 579 . Fischer, S., Drosopoulos, S., Tsen, J., & Born, J. ( 2006). Implicit learning explicit knowing: a role for sleep in memory system interaction. Journal of Cognitive Neuroscience , 18 (3), 311 - 319. 131 Foucart, A., & Frenck - Mestre, C. (2011). Gramm atical gender processing in L2: Electrophysiological evidence of the effect of L1 L2 syntactic similarity. Bilingualism: Language and Cognition , 14 (3), 379 - 399. Ga s s, S. (1979). Language transfer and universal grammatical relation s. Language learning , 29 ( 2), 327 - 344. Gass, S. (1984). A review of interlanguage syntax: Language transfer and language universals. Language learning , 34 (2), 115 - 132. Godfroid, A. ( 2016). The effects of implicit instruction on implicit and explicit knowledge development. Studi es in Second Language Acquisition, 38 (2), 177 215. - tracking data (pp. 44 - 57). In R. P. Leow (Ed.), The Routledge handbook of second language research in class room learning . New York: Routled ge. Godfroid, A. (in press, 20 20 ). Eye tracking in second language acquisition and bilingualism: A research synthesis and methodological guide . New York: Routledge. Godfroid , A., Ahn, J., Choi, I., Ballard, L., Cui, Y, Jo hnston, S., Lee, S., Sarkar, A., & Yoon, H. (2018). Incidental vocabulary learning in a natural reading context: an eye - tracking study. Bilingualism: Langu a ge and Cognition , 21 (3), 563 - 584. Godfroid, A., & Spino, L. A. (2015). Reconceptualizing reactivi tracking: Absence of evidence is not evidence of absence. Language Learning , 65 (4), 896 - 928. Godfroid, A. , & Winke, P. (2015). Investigating implicit and explicit processing using L2 learners' eye - movement data. In P. Rebuschat (Ed.), Implicit and explicit learning of languages (pp. 325 - 348). Philadelphia, PA: John Benjamins. Grey, S., Williams, J. N., & Rebuschat , P. (2014). Incidental exposure and L2 learning of morphosyntax. Studies in Second Language Acquisition , 36 , 1 35. Hama, M., & Leow, R. P. (2010). Learning without awareness revisited: Extending Williams (2005). Studies in Second Language Acquisition , 3 2 , 465 491. Hamrick, P. (2013). Development of conscious knowledge during early incidental learning of L2 syntax. G eorgetown University, Washington, DC. Hamrick, P., & Sachs, R. (2017). Establishing evidence of learning in experiments employing artific ial linguistic system. Studies in Second Language Acquisition , 1 17. Hatzidaki , A., Branigan, H. P., & Pickering, M . J. (2011). Co - activation of syntax in bilingual language production. Cognitive psychology , 62 (2), 123 - 150. 132 Indrarathne, B., & Kormos, J. (2017). Attentional processing of input in explicit and implicit conditions: An eye - tracking study. Studies in Second Language Studies , 39 , 401 430. Indrarathne, B., Ratajczak, M., & Kormos, J. ( 2018). Modelling c hanges in the c ognitive p rocessing of g rammar in i mplicit and e xplicit l earning c onditions: Insights f rom an e t racking s tudy. Language Learning , 68 (3), 669 - 708. Jarvis, S., & Pavlenko, A. (2008). Crosslinguistic influence in language and cognition . Routledge. Jiang, N. (2007). Selective integration of linguistics knowledge in adult second language learning. L anguage Learning , 25 , 603 - 634. Jiang, N., Novokshano va, E., Masuda, K., & Wang, X. (2011). Morphological congruency and the acquisition of L2 morphemes. Language Learning , 61 (3), 940 - 967. Karmiloff - Smith, A. (1992). Beyond modularity: A developmental p erspective on cognitive science . Cambridge, MA: MIT Pre ss. Keating, G. D. (2009). Sensitivity to violations of gender agreement in native and nonnative Spanish: An ey e - movement investigation. Language Learning , 59 (3), 503 - 535. Kerz, E., Wiechmann, D., & Riedel, F. B. (2017). Implicit learning in the crowd: Investigating the role of awareness in the acquisition of L2 knowledge. Studies in Second Language Acquisition , 39 (4), 711 - 734. Kim, M., & Godfroid, A. (in press, 2019). Should we listen or read? Modal ity effects in implicit and explicit L2 knowled ge of grammar. The Modern Language Journal , 103 (3). Kuhn, G., & Dienes, Z. (2008). Learning non - local dependencies. Cognition , 106 (1), 184 - 206. Leow, R. P., Grey, S., Marijuan, S., & Moorman, C. (2014). Concu rrent data elicitation procedures, processes, a nd the early stages of L2 learning: A critical overview. Second Language Research , 30 (2), 111 - 127. Leow, R. P. (2015a). Explicit learning in the L2 classroom: A student - centered approach . New York: Routledge. Leow, R. P. (2015 b ). The roles of attention and (un) awareness in SLA: Conceptual replication of N . C . Ellis & Sagarra (2010a) and Leung & Williams (2012). Language Teaching , 48 (1), 117 - 129. Leung, J. H. C., & Williams, J. N. (2011). The implicit learn ing of mappings between forms and con textually - derived meanings. Studies in Second Language Acquisition , 33 , 33 55. Leung, J. H. C., & Williams, J. N. (2012). Constraints on implicit learning of grammatical form - 133 meaning connections. Language Learning , 62 , 634 662. Leung, J. H. C., & Williams, J. N. (2014). Crosslinguistic differences in implicit language learning. Stu dies in Second Language Acquisition , 36, 634 662. Lim, J. H., & Christianson, K. (2015). Second language sensitivity to agreement errors: Evidence from eye movements during comprehension and translation. Applied Psycholinguistics , 36 (6), 1283 - 1315. MacWh inney, B. (1997). Second language acquisition and the competition model. In A. M. B. de Groot & J. F. Kroll (Eds.), Tutorials in bilingual ism: Psycholinguistic perspectives (pp. 113 144). Mahwah, NJ: Erlbaum MacWhinney, B . (2005). A unified model of language acquisition . In J . F. Kroll & A . M . B . de Groot (eds.). Handbook of bilingualism: Psycholinguistic approaches . Oxford: Oxford Universi ty Press . MacWhinney, B. (2008). A unified model . In N. Ellis & P. Robinson (Eds.), Handbook of cognitive lingu istics and second language acquisition (pp. 341 371). London, UK : Routledge. McManus, K. & Marsden E. (2019). Signatures of automaticity durin g practice. Explicit instruction about L1 processing routines can improve L2 grammatical processing. Applied Psyc holinguistics , 40 (1), 205 - 234. Miller, Z., & Godfroid, A. (in press , 2019 ). Emotions in incidental language learning: An individual differences approach. Studies in Second Language Acquisition. Mohamed, A. A. (2018). Exposure frequency in L2 reading: An eye - movement perspective of incidental vocabulary learning. Studies in Sec ond Language Acquisition, 40 (2), 269 - 293. Monaghan, P., Schoetensack, C., & Rebuschat, P. (2019). A single paradigm for implicit and statistical learning. Topics in Cogn itive Science . Morgan - Short, K., Deng, Z., Brill - Schuetz, K. A., Faretta - Stutenberg, M., Wong, P. C., & Wong, F. C. (2015). A view of the neural representation of second language syntax through artificial language learning under implicit contexts of expos ure. Studies in Second Language Acquisition , 37 (2), 383 419. Morgan - Short, K., Finger , I., Grey, S., & Ullman, M. T. (2012). Second language processing shows increased native - like neural responses after months of no exposure. PLoS One , 7 (3), e32974. Morg an - Short, K., Sanz, C., Steinhauer, K., & Ullman, M. T. (2010). Second langua ge acqui s ition of gender agreement in explicit and implicit training conditions: An event - 134 related potential study. Language Learning , 60 , 154 193. Nijhof, A. D., & Willems, R. M. (2015). Simulating fiction: individual differences in literature comprehension rev e aled with fMRI. PLoS One , 10 (2), e0116492. Norris, J. M., & Ortega, L. (2000). Effectiveness of L2 instruction: A research synthesis and quantitative meta analysis. Language learning , 50 (3), 417 - 528 . Onnis , L., & Thiessen, E. (2013). Language experience changes subsequent learning. Cognition , 126 , 268 284. Pacini, R., & Epstein, S. (1999). The relation of rational and exper imental information processing styles to personality, basic beliefs, and the ratio - bias phenomenon. Journal of Personality and Social Psychology , 76 (6), 972 - 987. Pellicer - Sán chez, A., & Schmitt, N. (2010). Incidental Vocabulary Acquisition from an Authentic Novel: Do" Things Fall Apart"?. Reading in a Foreign Language , 22 (1), 31 - 55. Perruchet, P., & Pacton, S. (2006). Implicit learning and statistical learning: One phenomenon , two approaches. Trends in Cognitive Sciences , 10 (5), 233 238. Plihal, W., & Born, J. (1997). Effects of early and late nocturnal sleep on declarative and procedural memory. Journal of cognitive neuroscience , 9 (4), 534 - 547. Rayner, K. (1998). Eve moveme nts in reading and information processing: 20 years of research. Psychological Bulletin, 124 (3), 372 422. DOI: 10.1037/0033 - 2909.124.3.372 Rebuschat, P. (2013). Measuring implicit and explicit knowledge in second language research. Language Learning , 63, 595 626. Rebuschat, P., Hamrick, p., Riestenber, K., Sacks, R., & Ziegler, N. (2015). Triangulating measures of awareness: A contribution to the debate on learning without awareness. Studies in Second Language Acquisition , 37, 299 334. Rebuschat , P., & Williams, J. N. (2012). Implicit and explicit knowledge in second language acquisition. Applied Psycholinguistics , 33 , 829 856. Rebuschat, P., & Williams, J. N. (2012). Introduction: Statistical learning and language acquisition. In P. Rebuschat, & J. N . Williams (Eds.), Statistical learning and language acquisition. (pp. 1 - 12). Berlin: Mouton de Gruyter. Ringbom, H. (2007). Cross - linguistic similarity in foreign language learning . Clevedon: Multilingual Matter. 135 Robinson, P. (2002). Learning co ndition s, aptitude complexes, and SLA. In P. Robinson (Ed.), Individual differences and instructed language learning (pp. 113 - 133). Philadelphia, PA: Benjamins. Robinson, P. (2005). Cognitive abilities, chunk - strength, and frequency effects in implicit artifici al grammar and incidental L2 leaning: Replications of Reber, Walkenfeld, and Hernstadt (1991) and Knowlton and Squire (1996) and their relevance for SLA. Studies in Second Lang uage Acquisition , 27, 235 268. Rogers, J., Reves, A., & Rebuschat, P. (2016). Implicit and explicit knowledge of inflectional morphology. Applied psycholinguistics , 37, 781 812. Sanoudaki, E., & Thierry, G. (2015). Language non - selective syntactic acti vation in early bilinguals: the role of verbal fluency. International Journal of Bilingual Education and Bilingualism , 18 (5), 548 - 560. Sanz, C., & Grey, S. (2015). Effects of conditions on L2 development: Moving beyond ac cu racy. In P. Rebuschat (Ed.), I mplicit and explicit learning of languages (pp. 301 - 324). Amsterdam: John Benjam ins. Spada, N., & Tomita, Y. (2010). Interactions between type of instruction and type of language feature: A meta - analysis. Language learning , 60 (2), 263 - 308. Spinner, P., Foote, R., & Upor, R. A. (2017). Gender and number processing in second language Swahili. Linguistic Approaches to Bilingualism , 8 (4), 446 - 476. Spinner, P., & Gass, S. M. (2019). Using judgments in second language acquisition research . Routle dge : New York. Spinner, P., Gass, S. M., & Behney, J. (2013). Ecological validity in eye - tr acking: An empirical study. Studies in Second Language Acquisition , 35 (2), 389 - 415. Spinner, P., & Jung, S. (2018). Production and Comprehension in Processability Theory: A s elf - Paced Reading Study. Studies in Second Language Acquisition , 40 (2), 295 - 318. Suzuki, Y. (2017). Validity of new measures of implicit knowledge: Distinguishing implicit knowledge from automatized explicit knowledge. Applied Psycholinguistic s , 38 (5), 1229 - 1261. Suzuki, Y., & DeKeyser, R. M. (2015). Comparing elicited imitation and word monitoring as measures of implicit knowledge. Language Learning , 65 (4), 860 - 895. Suzuki, Y., & DeKeyser , R. M. (2017). The interface of explicit and implicit knowledge in a second language: Insights from individual differences in cognitive aptitudes . Language Learning , 67 (4), 747 - 490. 136 Speer, N. K., Reynolds, J. R., Swallow, K. M., & Zacks, J. M. (2009). R eading stories activates neural representations of visual and motor experiences. Psychological science , 20 (8), 989 - 999. Tagarelli, K., Ruiz, S., Vega, J., & Rebuschat, P. (2016). Variability in seco nd language learning: The roles of individual differences , learning conditions, and linguistic complexity. Studies in Second Language Acquisition , 38 (2), 293 316. Tokowicz, N., & MacWhinney, B. (2005). Implicit and explicit measures of sensitivity to violations in second language grammar: An event - related potential investigation. Studies in second language acquisition , 27 (2), 173 - 204. Tolentino, L. C., & Tokowicz, N. (2011). Across language, space, and time. Studies in Second Language Acquisition , 33 , 91 - 125. second language grammar instruction. Language Learning , 64 (2), 279 - 309. Ullman, M. T. (2005). A cognitive neuroscience perspective on second language acquisition : The declarative/procedural model. In C. Sanz (Ed.), Mind and context in adult second language acquisition: Methods, theory, and practice (pp. 141 178). Washington, DC: Georgetown University Press. Vafaee, P., Suzuki, Y., & Kachisnke, I. (2017). Validati ng grammatical judgment tests. Studies in Second Language Studies , 39(1), 59 - 95. Van Heuven, W. J., Mandera, P., Keuleers, E., & Brysbaert, M. (2014). SUBTLEX - UK: A new and improved word frequency database for British English. The Quarterly Journal of Ex perimental Psychology , 67 (6), 1176 - 1190. VanPatten, B., & Smith, M. (2015). Aptitude as grammatical sensitivity and the initial stages of learning Japanese as a L2. Studies in Second Language Acquisition , 37 , 135 - 165. VanPatten, B., & Rothman, J. (2015) . What does current generative theory have to say about the explicit - implicit debate? In P. Rebuschat (Ed.), Explicit and implicit learning of languages (pp. 89 - 116). Amsterdam: John Benjamins. Wagner, U., Gais, S., Ha ider, H., Verleger, R., & Born, J. (2004). Sleep inspires insight. Nature , 427 (6972), 352. Wh ite, L. ( 2000 ) . Second language acquisition: from initial to final state. In J. Archibald (ed.), Second language acquisition and linguistic theory (pp. 130 155). Blackwell. Whiteside, S. P., & Lyn am, D. R. (2001). The five factor model and impulsivity: Using a structural model of personality to understand impulsivity. Personality and Individual Differences , 30 (4), 669 - 689. doi:10.1016/S0191 - 8869(00)00064 - 7 . 137 Willems, R. M., Frank, S. L., Nijhof, A. D., Hagoort, P., & Van den Bosch, A. (2015). Prediction during natural language comprehension. Cerebral Cortex , 26 (6), 2506 - 2516. Williams, J. N. (2005). Learning without awareness. Studies in Second Language Acqui sition , 27 , 269 304. Williams, J. N. ( 2009). Implicit learning in second language acquisition. In W. C. Ritchie & T. K. Bhatia (Eds.), The new handbook of second language acquisition (pp. 319 353). Bingley, UK: Emerald Press. Williams, J. N., & Lovatt, P. (2003). Phonological memory and rule learning. Language learning , 53 (1), 67 - 121.