VERBAL MORPHOLOGY IN L2 SPANISH: MORPHOLOGICAL DEFICITS AND PROCESSING STRATEGIES By Suzanne Johnston A DISSERTATION Submitted to Michigan State University in partial fulfillment of the requirements for the degree of Second Language Studies—Doctor of Philosophy 2017 ABSTRACT VERBAL MORPHOLOGY IN L2 SPANISH: MORPHOLOGICAL DEFICITS AND PROCESSING STRATEGIES By Suzanne Johnston One of the fundamental questions in second language (L2) research asks whether L2 learners can access inflectional morphology in real time to process sentences in a native-like manner. More recently, research has indicated that the L2 parser is less efficient than the native parser, and only near native learners may be capable of processing sentences like native speakers (Hopp, 2010; Keating, 2009). Furthermore, L2 learners may continue to rely on nonnative-like processing strategies to make up for processing deficits in L2 comprehension, even after inflectional morphology has been acquired (Ellis & Sagarra, 2010; LoCoco, 1987; VanPatten, 1984, 1990, 1996). The current dissertation examines such behavior by examining the development and processing of subject/verb agreement in L2 Spanish. L1 English/L2 Spanish speakers completed two self-paced reading tasks that tested whether they perceived errors in morphological inflection, and whether they could reliably use morpho-syntactic cues during real time processing. A third self-paced reading task examined whether L2 learners’ processing was facilitated by the presence of overt subjects. Results suggest that while L2 learners showed sensitivity to some errors, the parser did not regularly process inflection during on-line comprehension. Furthermore, L2 learner processing was facilitated by the presence of an overt subject. These results are compatible with the prediction that L2 parsers are less efficient than L1 parsers, and that L2 learners may continue to rely on L1 or universal processing measures after morphological inflection has been acquired. Copyright by SUZANNE JOHNSTON 2017 ACKNOWLEDGEMENTS Nearly six years ago, I came to Michigan State University immediately after earning my Bachelor’s degree in Spanish. Perhaps like others who decide to pursue a Master’s degree, I did not have a clear career path upon entering the Applied Spanish Linguistics Program. It was during this time that I was fortunate to meet Dr. Bill VanPatten, who has guided me to become the teacher and researcher that I am today. Over the years, he has answered countless questions, provided copious amounts of advice, and has provided me with many opportunities to learn. For this, I am immensely grateful, and feel incredibly fortunate to have worked with Dr. VanPatten over the course of my six years at MSU. It is truly my privilege to be able to say that he is my mentor. I would also like to thank Dr. Patti Spinner and Dr. Aline Godfroid for their support, encouragement, and advice throughout these last four years. I have always appreciated their willingness to assist me with projects theoretically and statistically. I would also like to thank Dr. Paula Winke, who, despite this not being her area of expertise, agreed to serve on this committee, and provided valuable assistance in choosing materials for the proficiency test used in this dissertation. Andrew Dennhardt from the Center for Statistical Training and Consulting at MSU was also a source of immense help during the final stages of this dissertation. Without his help, the analyses of this study would not be as fine-tuned or as thorough. I thank him not only for sharing his knowledge with me, but also for being patient and kind throughout the process. I would also like to thank my colleagues and the other faculty in the Second Language Studies Program. The SLS program at MSU has been nothing but supportive over the course of iv the last four years, and I cannot imagine a more supportive group of colleagues. This being said, there were several friends who were my primary sources of support over the years. Changchang Yao’s friendship has made my time at MSU enjoyable, and it is my honor to call her one of my best friends. I have enjoyed our office conversations, movie nights, and puppy playdates immensely, and even in the craziest of times, she has been one of my dearest friends. I will miss seeing her every day when I move away. I will also miss Irina Zaykovskaya, who has also been an invaluable friend throughout this journey. Her support and staunch determination provided me with a rock when I needed it most. It is not often that someone—let alone another incredibly busy Ph.D. student—offers to come to your home to help organize and clean your bedroom, and Irina would not take no for an answer. It is for these small, yet significant, deeds that I feel honored to call Irina a friend. I would also like to thank all the Teaching Assistants in the thirdfloor Linguistics and Languages office. All of them have made working the long and tedious days enjoyable. I look forward to seeing all of them every day, and will miss them immensely. Finally, I would like to thank my family. My parents have always taught me the value of perseverance, and have believed in me even when I didn’t believe in myself. Their encouragement over the years has brought me to this point, and without it, I never would have succeeded. Nevertheless, the primary source of support throughout this journey is a family member, who came to my home only one year before I started my doctoral journey. Purchased on a whim, my slightly chunky French Bulldog, Moo, has been an integral source of support. He has sat by my side (quite literally) through sickness, exhaustion, frustration, and happiness. Our daily walks are a stern reminder of the value of slowing down, and taking time to (again, literally) sniff the flowers. Every single one. Just in case. v TABLE OF CONTENTS LIST OF TABLES ......................................................................................................................... ix LIST OF FIGURES ..................................................................................................................... xiv CHAPTER ONE ............................................................................................................................. 1 INTRODUCTION .......................................................................................................................... 1 Morpho-syntactic Features and Syntax ..................................................................................... 1 L2 Processing ............................................................................................................................ 6 L1 English/L2 Spanish as a Test Case ...................................................................................... 9 The Current Dissertation ......................................................................................................... 11 Overview of Dissertation ........................................................................................................ 12 CHAPTER TWO .......................................................................................................................... 13 EXPERIMENT ONE .................................................................................................................... 13 Background and Motivation ................................................................................................... 13 Spanish Verbal Morphology ............................................................................................. 13 Predictions about L2 Learner Behavior on L2 Verbal Morphology ................................. 16 VanPatten, Keating, and Leeser (2012) ............................................................................ 17 The Present Study ................................................................................................................... 20 Method and Procedure ............................................................................................................ 21 Participants........................................................................................................................ 21 Nonnative Speakers .................................................................................................... 21 Native Speakers .......................................................................................................... 21 Materials ........................................................................................................................... 22 Stimuli ......................................................................................................................... 22 Proficiency Test .......................................................................................................... 25 Procedure .......................................................................................................................... 26 Scoring and Analysis ........................................................................................................ 28 Results ..................................................................................................................................... 32 Native Speakers ................................................................................................................ 32 1sg/3sg Condition ....................................................................................................... 32 2sg/3pl Condition ........................................................................................................ 36 3sg/3pl Condition ........................................................................................................ 40 Summary of Results for the Native Control Group. ................................................... 44 Nonnative Speakers .......................................................................................................... 44 1sg/3sg Condition ....................................................................................................... 44 2sg/3pl Condition ........................................................................................................ 47 3sg/3pl Condition ........................................................................................................ 51 Summary of Results for the Nonnative Group ........................................................... 55 Discussion for Experiment One .............................................................................................. 56 Conclusion to Experiment One ............................................................................................... 60 vi CHAPTER THREE ...................................................................................................................... 61 EXPERIMENT TWO ................................................................................................................... 61 Background and Motivation ................................................................................................... 62 Sentence Processing in SLA ............................................................................................. 62 Topic/Subject Shifts in Spanish ........................................................................................ 66 Research Question for Experiment Two ........................................................................... 70 Method and Procedure ............................................................................................................ 71 Participants........................................................................................................................ 71 Nonnative Speakers .................................................................................................... 71 Native Speakers .......................................................................................................... 72 Materials ........................................................................................................................... 72 Stimuli ......................................................................................................................... 72 Procedure .......................................................................................................................... 74 Scoring and Analysis ........................................................................................................ 75 Results ..................................................................................................................................... 77 Native Speakers ................................................................................................................ 77 1sg/3sg Condition ....................................................................................................... 77 2sg/3pl Condition ........................................................................................................ 80 Summary of Results for the Native Speakers. ............................................................ 84 Nonnative Speakers .......................................................................................................... 85 1sg/3sg Condition ....................................................................................................... 85 2sg/3pl Condition ........................................................................................................ 88 Summary of Results for the Nonnative Speakers ....................................................... 92 Discussion of Experiment Two............................................................................................... 93 Conclusion of Experiment Two .............................................................................................. 94 CHAPTER FOUR......................................................................................................................... 95 EXPERIMENT THREE ............................................................................................................... 95 Background and Motivation ................................................................................................... 95 Summary of Results of Experiment Two.......................................................................... 95 Research Question for Experiment Three ......................................................................... 96 Method and Procedure ............................................................................................................ 97 Participants........................................................................................................................ 97 Nonnative Speakers .................................................................................................... 97 Native Speakers .......................................................................................................... 97 Materials ........................................................................................................................... 97 Stimuli ......................................................................................................................... 97 Procedure ........................................................................................................................ 100 Scoring and Analysis ...................................................................................................... 101 Results ................................................................................................................................... 102 Native Speakers .............................................................................................................. 102 1sg/3sg Condition ..................................................................................................... 102 2sg/3pl Condition ...................................................................................................... 107 Summary of Native Speaker Results ........................................................................ 112 Nonnative Speakers ........................................................................................................ 113 1sg/3sg Condition ..................................................................................................... 113 vii 2sg/3pl Condition ...................................................................................................... 117 Summary of Nonnative Speaker Results .................................................................. 122 Discussion of Experiment Three ........................................................................................... 123 Conclusion to Experiment Three .......................................................................................... 124 CHAPTER FIVE ........................................................................................................................ 126 GENERAL DISCUSSION AND CONCLUSION ..................................................................... 126 Summary of Findings of the Three Experiments .................................................................. 127 Experiment One. ............................................................................................................. 127 Experiment Two.............................................................................................................. 127 Experiment Three............................................................................................................ 128 Significance of Findings ....................................................................................................... 128 Limitations ............................................................................................................................ 132 Future Research .................................................................................................................... 133 Conclusion ............................................................................................................................ 134 APPENDICES ............................................................................................................................ 136 APPENDIX A: Experiment One Stimuli and Instructions ................................................... 137 APPENDIX B: Experiment Two Stimuli ............................................................................. 164 APPENDIX C: Experiment Three Stimuli ........................................................................... 181 APPENDIX D: Nonnative Background Questionnaire ........................................................ 201 APPENDIX E: Native Background Questionnaire .............................................................. 203 REFERENCES ........................................................................................................................... 205 viii LIST OF TABLES Table 1.1 Example of Spanish Verbal Paradigm for –ar Verbs in the Present Tense .................... 3 Table 1.2 Example of English Verbal Paradigm in the Present Tense ........................................... 3 Table 1.3 Example of English Verbal Paradigm in the Past Tense ................................................ 4 Table 1.4 Example of Spanish Verbal Paradigm for –ar Verbs in the Past Tense ......................... 5 Table 2.1 Summary of Person/Number (Participant/Individuation) Features in Spanish ............ 14 Table 2.2 Descriptive Statistics of Native Speakers for 1sg/3sg in Milliseconds ........................ 32 Table 2.3 Model Results Analyses for Native Speaker Target Region 1sg/3sg ........................... 33 Table 2.4 Model Results Analyses for Native Speaker First Spillover Region 1sg/3sg .............. 34 Table 2.5 Model Results Analyses for Native Speaker Second Spillover Region 1sg/3sg .......... 35 Table 2.6 Model Results Analyses for Native Speaker Third Spillover Region 1sg/3sg ............. 35 Table 2.7 Descriptive Statistics of Native Speakers for 2sg/3pl in Milliseconds ......................... 36 Table 2.8 Model Results Analyses for Native Speaker Target Region 2sg/3pl ........................... 37 Table 2.9 Model Results Analyses for Native Speaker First Spillover Region 2sg/3pl ............... 38 Table 2.10 Model Results Analyses for Native Speaker Second Spillover Region 2sg/3pl ........ 39 Table 2.11 Model Results Analyses for Native Speaker Third Spillover Region 2sg/3pl ........... 40 Table 2.12 Descriptive Statistics of Native Speakers for 3sg/3pl in Milliseconds ....................... 40 Table 2.13 Model Results Analyses for Native Speaker Target Region for 3sg/3pl .................... 41 Table 2.14 Model Results Analyses for Native Speaker First Spillover Region for 3sg/3pl ....... 42 Table 2.15 Model Results Analyses for Native Speaker First Spillover Region for 3sg/3pl ....... 43 Table 2.16 Model Results Analyses for Native Speaker Third Spillover Region for 3sg/3pl ...... 43 Table 2.17 Descriptive Statistics of Nonnative Speakers for 1sg/3sg in Milliseconds ................ 44 ix Table 2.18 Model Results Analyses of Nonnative Speakers for Target Region for 1sg/3sg ....... 45 Table 2.19 Model Results Analyses of Nonnative Speakers for First Spillover Region 1sg/3sg. 46 Table 2.20 Model Results Analyses of Nonnative Speakers for Second Spillover Region 1sg/3sg ....................................................................................................................................................... 46 Table 2.21 Model Results Analyses of Nonnative Speakers for Third Spillover Region 1sg/3sg 47 Table 2.22 Descriptive Statistics of Nonnative Speakers for 2sg/3pl in Milliseconds ................. 47 Table 2.23 Model Results Analyses for Nonnative Speaker Target Region 2sg/3pl.................... 48 Table 2.24 Model Results Analyses for Nonnative Speaker First Spillover Region 2sg/3pl ....... 49 Table 2.25 Model Results Analyses for Nonnative Speaker Second Spillover Region 2sg/3pl... 50 Table 2.26 Model Results Analyses for Nonnative Speaker Third Spillover Region 2sg/3pl ..... 51 Table 2.27 Descriptive Statistics of Nonnative Speakers for 3sg/3pl in Milliseconds ................. 51 Table 2.28 Model Results Analyses for Nonnative Speaker Target Region for 3sg/3pl .............. 52 Table 2.29 Model Results Analyses for Nonnative Speaker First Spillover Region for 3sg/3pl . 53 Table 2.30 Model Results Analyses for Nonnative Speaker Second Spillover Region for 3sg/3pl ....................................................................................................................................................... 54 Table 2.31 Post-Hoc Pairwise Comparisons of Morpheme/Grammaticality Interaction in Nonnative First Spillover Region for 3sg/3pl ............................................................................... 54 Table 2.32 Model Results Analyses for Nonnative Speaker Third Spillover Region for 3sg/3pl 55 Table 2.33 Review of Summary of Person/Number (Participant/Individuation) Features in Spanish .......................................................................................................................................... 57 Table 3.1 Descriptive Statistics of Native Reading Times for 1sg/3sg Condition in Milliseconds ....................................................................................................................................................... 77 Table 3.2 Model Results Analyses for Native Speaker Target Region for the 1sg/3sg Condition78 Table 3.3 Model Results Analyses for Native Speaker First Spillover Region for the 1sg/3sg Condition ...................................................................................................................................... 79 Table 3.4 Model Results Analyses for Native Speaker Second Spillover Region for the 1sg/3sg Condition ...................................................................................................................................... 80 x Table 3.5 Descriptive Statistics of Native Reading Times for 2sg/3pl Condition in Milliseconds ....................................................................................................................................................... 81 Table 3.6 Model Results Analyses for Native Speaker Target Region for the 2sg/3pl Condition ....................................................................................................................................................... 82 Table 3.7 Model Results Analyses for Native Speaker First Spillover Region for the 2sg/3pl Condition ...................................................................................................................................... 83 Table 3.8 Model Results Analyses for Native Speaker Region Second Spillover Region for the 2sg/3pl Condition .......................................................................................................................... 84 Table 3.9 Descriptive Statistics of Nonnative Reading Times for 1sg/3sg Condition in Milliseconds .................................................................................................................................. 85 Table 3.10 Model Results Analyses for Nonnative Speaker Target Region for the 1sg/3sg Condition ...................................................................................................................................... 86 Table 3.11 Model Results Analyses for Nonnative Speaker First Spillover Region for the 1sg/3sg Condition ...................................................................................................................................... 87 Table 3.12 Model Results Analyses for Nonnative Speaker Second Spillover Region for the 1sg/3sg Condition ......................................................................................................................... 88 Table 3.13 Descriptive Statistics of Nonnative Reading Times for 2sg/3pl Condition in Milliseconds .................................................................................................................................. 89 Table 3.14 Model Results Analyses for Nonnative Speaker Target Region for the 2sg/3pl Condition ...................................................................................................................................... 90 Table 3.15 Model Results Analyses for Nonnative Speaker First Spillover Region for the 2sg/3pl Condition ...................................................................................................................................... 91 Table 3.16 Model Results Analyses for Nonnative Speaker Second Spillover Region for the 2sg/3pl Condition .......................................................................................................................... 92 Table 4.1 Descriptive Statistics of Native Reading Times for 1sg/3sg Condition in Milliseconds ..................................................................................................................................................... 103 Table 4.2 Model Results Analyses for Native Speaker Target Region for the 1sg/3sg Condition ..................................................................................................................................................... 104 Table 4.3 Model Results Analyses for Native Speaker First Spillover Region for the 1sg/3sg Condition .................................................................................................................................... 105 xi Table 4.4 Model Results Analyses for Native Speaker Second Spillover Region for the 1sg/3sg Condition .................................................................................................................................... 106 Table 4.5 Model Results Analyses for Native Speaker Third Spillover Region for the 1sg/3sg Condition .................................................................................................................................... 107 Table 4.6 Descriptive Statistics of Native Reading Times for 2sg/3pl Condition in Milliseconds ..................................................................................................................................................... 108 Table 4.7 Model Results Analyses for Native Speaker Target Region for the 2sg/3pl Condition ..................................................................................................................................................... 109 Table 4.8 Post-Hoc Pairwise Comparisons of Morpheme/Match Interaction in Native Target Region for 2sg/3pl....................................................................................................................... 110 Table 4.9 Model Results Analyses for Native Speaker First Spillover Region for the 2sg/3pl Condition .................................................................................................................................... 110 Table 4.10 Model Results Analyses for Native Speaker Second Spillover Region for the 2sg/3pl Condition .................................................................................................................................... 111 Table 4.11 Model Results Analyses for Native Speaker Third Spillover Region for the 2sg/3pl Condition .................................................................................................................................... 112 Table 4.12 Descriptive Statistics of Nonnative Reading Times for 1sg/3sg Condition in Milliseconds ................................................................................................................................ 113 Table 4.13 Model Results Analyses for Nonnative Speaker Target Region for the 1sg/3sg Condition .................................................................................................................................... 114 Table 4.14 Model Results Analyses for Nonnative Speaker First Spillover Region for the 1sg/3sg Condition .................................................................................................................................... 115 Table 4.15 Model Results Analyses for Nonnative Speaker Second Spillover Region for the 1sg/3sg Condition ....................................................................................................................... 116 Table 4.16 Model Results Analyses for Nonnative Speaker Third Spillover Region for the 1sg/3sg Condition ....................................................................................................................... 117 Table 4.17 Descriptive Statistics of Nonnative Reading Times for 2sg/3pl Condition in Milliseconds ................................................................................................................................ 118 Table 4.18 Model Results Analyses for Nonnative Speaker Target Region for the 2sg/3pl Condition .................................................................................................................................... 119 xii Table 4.19 Post-Hoc Pairwise Comparisons of Morpheme/Match Interaction in Nonnative Target Region for 2sg/3pl ...................................................................................................................... 120 Table 4.20 Model Results Analyses for Nonnative Speaker First Spillover Region for the 2sg/3pl Condition .................................................................................................................................... 120 Table 4.21 Model Results Analyses for Nonnative Speaker Second Spillover Region for the 2sg/3pl Condition ........................................................................................................................ 121 Table 4.22 Model Results Analyses for Nonnative Speaker Third Spillover Region for the 2sg/3pl Condition ........................................................................................................................ 122 xiii LIST OF FIGURES Figure 2.1 Sample Plot of Residuals for Nonnative Target Region ............................................. 31 Figure 2.2 Sample Log-Transformed Residuals for Nonnative First Person Singular/Third Person Singular Target Region ................................................................................................................. 32 xiv CHAPTER ONE INTRODUCTION One of the fundamental questions in second language acquisition (SLA) is how second language (L2) learners acquire and process inflectional morphology in the L2. Various hypotheses have been proposed that range from proposing that L2 learners are unable to acquire inflectional morphology that does not exist in the L1 (see Hawkins, 2005; Hawkins & Chan, 1997) to proposing that inflectional morphology is completely acquirable, and that any apparent deficits are due to either processing failures, poor instruction, or lack of access to input (Hopp, 2010; Roberts, Gullberg, & Indefrey, 2008; Schwartz & Sprouse, 1994; 1996). The current dissertation’s purpose is three-fold. The first is to examine the extent to which L2 subject/verb agreement morphology has been acquired by B1 level L2 Spanish/L1 English speakers. Subject/verb agreement was chosen, because according to current hypotheses, this inflectional morphology should be acquirable for the targeted language learners. The second purpose is to examine whether L2 learners process morphological inflection reliably within a sentence by testing their ability to detect topic/subject shifts when a shift is indicated by morphological inflection. The third is to examine the extent to which L1 English/L2 Spanish processing is facilitated by overt subjects in the context of a topic/subject shift. Morpho-syntactic Features and Syntax Universal Grammar (UG) posits that languages mark lexical items with a certain set of linguistic properties called features. Traditionally, features have been divided into phonological, semantic, and morpho-syntactic features. The current dissertation is concerned with morphosyntactic features only, which are features that are realized through the syntax and can determine the shape that a word or phrase takes (Adger, 2003). Features can be further divided into 1 interpretable or uninterpretable features. Interpretable features are features that are assigned contextually or are inherent to lexical items (e.g., person, number, gender on nouns), while uninterpretable features appear through an agreement processes (e.g., subject verb agreement on verbs). Within this categorization, features can also be considered strong or weak. Strong features trigger syntactic operations, while weak features do not (Adger, 2003). For example, both English and Spanish have some form of verb agreement. However, Spanish is considered to have a strong V feature, which means that there is perceivable verb movement, which can be seen through the placement of adverbs. In (4), the verb and its object are separated by the adverb ‘often’, because the verb has risen to a position above the adverb. (4) [CP[CJuani][TP[Tbebej][VP[ADVa menudo][V’[Vtj][NPcafé]]]] Juan bebe a menudo Juan drink.3rd.sg. often café. coffee. ‘Juan drinks coffee often.’ This can be contrasted with English, which has a weak V feature for lexical verbs, and demonstrates no perceivable verb movement. Because of this, example (5) is ungrammatical in English. (5) *Juan drinks often coffee. For auxiliary verbs and copulas, however, English exhibits some strong V features and shows signs of verb movement (see (6a-6c)). (6) a. Juan is really active. b. Juan doesn’t really have a choice. c. Juan has actually done the homework. 2 In addition to Spanish and English differing syntactically between strong and weak V movement, they also differ in richness of their morphological inflections on verbs. To be considered morphologically rich, a language must mark more than one feature on the same lexical item (e.g., person, number, tense) (Bobaljik, 2006). For example, the Spanish verb hablas (‘You speak’) in Table 1.1 carries a verb class marker (-a), a marker for second person and singular (-s), and is marked for the present tense (null). Because all Spanish verbs are marked for person, number, and tense, Spanish is considered morphologically rich. Table 1.1 Example of Spanish Verbal Paradigm for –ar Verbs in the Present Tense Singular First person hablo speak.1sg Plural hablamos speak.3pl Second person hablas speak.2sg habláis speak.2pl Third person habla speak.3sg hablan speak.3pl In contrast, English is considered morphologically poor. One such way of seeing this is that, except for the copula be, English does not mark verbs for more than one feature, and, absent the third person singular, has no other inflection for lexical verbs in the present tense (Table 1.2). Table 1.2 Example of English Verbal Paradigm in the Present Tense Singular First person I speak Second person Third person Plural We speak You speak You all speak He/She speaks They speak 3 Furthermore, for lexical verbs, tense and person/number agreement are also in complementary distribution (see 7a-7b). In (7a), the verb walks is marked with third person singular -s, and is not marked for tense. In the past tense (7b), however, it is marked for tense (-ed), but is not marked for person or number. (7a) John walks to school. (7b) John walked to school. This pattern is also true for auxiliary verbs do and have (see 8a-9b). (8a) He does go to school. (8b) He did go to school. (9a) He has a car. (9b) He had a car. In each of these cases, when the verb is in the past tense, it no longer marks for person/number agreement as it does in the present tense. Thus, apart from the copula be, English does not show any signs of subject/verb agreement outside of the present tense (Table 1.3). Table 1.3 Example of English Verbal Paradigm in the Past Tense Singular First person I spoke Second person Third person Plural We spoke You spoke You all spoke He/She spoke They spoke This can be contrasted with Spanish, which maintains subject/verb agreement across tenses. Table 1.4, for example, demonstrates subject/verb agreement in the preterit tense. As shown, 4 despite the change in tense from present (Table 1.2) to past, Spanish maintains distinct person/number morphemes, while English does not (1.3). Table 1.4 Example of Spanish Verbal Paradigm for –ar Verbs in the Past Tense Singular First person hablé speak.1sg.past Plural hablamos speak.3pl.past Second person hablaste speak.2sg.past hablasteis speak.2pl.past Third person habló speak.3sg.past hablaron speak.3pl.past Because English does mark for person and number for third person singular subjects, English carries remnants of subject/verb agreement on lexical verbs. However, it is unlikely that the syntactic processes through which it is realized mirrors Spanish morphologically or syntactically. For example, because English does not mark person and number on any lexical verbs other than for third person singular subjects in the present tense, Bobaljik (2006) proposes that English has only one for lexical verbs under which either tense or aspect is assigned, but never both. Spanish, on the other hand, has two syntactic nodes under which tense and aspect are assigned respectively, and is the reason why Spanish exhibits such rich morphological inflection. Thus, it cannot be said that English lacks subject/verb agreement and Spanish has subject/verb agreement, as this analysis is overly simplistic. Instead, it should be argued that the way that English marks subject/verb agreement differs from the way it is marked in Spanish. For this reason, Lardiere (2008, 2009) proposes that features that are shared across languages, however minimally, must be reconstructed, rather than acquired in L2 acquisition. This process occurs as nonnatives are exposed to meaningful and comprehensible input, and build mental representations of the morphophonological units that encompass the features of a given 5 language. At any point, nonnative speakers may run into difficulty acquiring morphemes for various reasons, such as features being represented on different types of lexical items (prefix, suffix, infix), numerous features being accounted for on one morphological item versus there being only one feature per morphological item, or the failure to appropriately parse morphosyntax (Spinner, 2013), etc. During this process, nonnative speakers may show sensitivity to some features, but not to others, depending on the features encoded and where learners are in the reconstruction process (Dekydtspotter & Renaud, 2009). Thus, nonnative-like performance is predicted at lower-levels of acquisition, as features are reassembled. In the case of L1 English/L2 Spanish subject/verb agreement, nonnative speakers must not only reassemble individual morphological features for person and number (Lardiere, 1998, 2000, 2007, 2008), they also must acquire additional syntactic structures to allow verb movement and for these features to be realized (VanPatten, Keating, & Leeser, 2012). Thus, at different points in acquisition, L2 learners may demonstrate knowledge of some morphological features and syntactic structures while simultaneously showing a lack of knowledge of others (Dekydtspotter & Renaud, 2009; McCarthy, 2008, 2012; VanPatten et al., 2012). L2 Processing In addition to acquiring the individual morphemes for subject/verb agreement, L2 learners must also be able to access these morphemes in real time. This is often referred to as linguistic performance, which is the ability to draw on linguistic competence during real-time processing. If learners have not acquired or reconstructed a morpheme, or it is not robustly represented in the morpho-lexical store, it is unlikely that L2 learners will demonstrate the ability to process this morpheme during linguistic performance. Processing can be defined as the moment by moment computation of language that allows speakers to make connections between 6 meaning and form during real time comprehension (VanPatten, 2012). However, the reverse may also occur: L2 learners may not process inflection during comprehension despite having a robust representation within the morpho-lexical store. This observation has been made in several L2 studies that have investigated L2 mental representation and processing behavior (Hopp, 2010; Keating, 2009; Prévost & White, 2000; Rothman & Iverson, 2007; Rothman, 2009; White, 2003; White, Valenzuela, Kozlowska-MacGregor, & Leung, 2004). Such studies have hypothesized that L2 learners may not process inflection because they cannot access these inflections in realtime. That is, L2 learners may have native- or near-native-like morphemes available to them, but the linguistic mechanism responsible for accessing these representations (the parser) does not have enough time or resources to do so during on-line tasks. One hypothesis that was derived from early research regarding this behavior was the Missing Surface Inflection Hypothesis (MSIH) (Prévost & White, 2000). The MSIH posited that the pressures of on-line production cause L2 learners to have difficulty accessing appropriate inflectional morphology in real time (Herschensohn, 2001; Prévost & White, 2000; White, 2003). Because the parser cannot access the appropriate inflection in real time, L2 learners were prone to what is called defaulting behavior, or the placement of underspecified forms where more specified forms are called for (McCarthy, 2008; Lopez Prego, 2012). While the MSIH seemingly holds promise in its predictions, it may only be able to account for the behavior of advanced-level nonnative (VanPatten et al., 2012). Because the MSIH was originally argued to be applied only to L2 production, it is speculative whether the predictions it makes can also be applied to L2 comprehension. The reason for this is that more recent studies have tended to find that L2 learners are also display the behaviors predicted by the MSIH in on-line comprehension. Hopp (2010), for example, found that L2 German speakers were more likely to fail to process case cues 7 when the task or sentence was more complex. This was also affected by the L1 of the participants (English, Dutch, or Russian), and their proficiency level. As such, Hopp concludes that the L2 parser is less capable of integrating syntactic, lexical, and morpho-syntactic information simultaneously, but does not predict defaulting behavior. Similarly, Keating (2009) found that nonnative speakers showed evidence of having acquired new features (in this case grammatical gender), but they were not sensitive to errors in gender agreement outside of the local domain (i.e. the noun phrase). Keating therefore argued that linguistic competence was not impaired, but rather linguistic performance, or the processing of the stimuli presented to the L2 learners during comprehension, had failed to compute grammatical gender agreement in some conditions. McCarthy (2008) also examined grammatical gender, and found that during on-line comprehension, L2 learners also demonstrated defaulting behavior on feminine-marked clitics. The results of these studies are also echoed in others, such as McCarthy (2012) and Lopez Prego (2012), suggesting that while the predictions of the MSIH may explain L2 learner behavior, it may extend also to on-line comprehension. As an extension to the MSIH, current research has posited that L2 learners may be affected by cognitive factors, such as working memory (e.g., McDonald, 2006) or slower processing mechanisms (e.g., Hopp, 2010; Scherag, Demuth, Rösler, Neville, & Roder, 2004; McDonald, 2000), that affect their ability to access the inflectional morphology, despite having a native-like L2 mental representations of inflectional morphology. In cases where the parser fails to access inflection during on-line comprehension, nonnative speakers may turn to L1 and/or universal default processing strategies to comprehend aural or written stimuli. One such strategy is that L2 learners may show a reliance on adverbs for determining tense, as was observed in Ellis and Sagarra (2010). The authors for this study argued that influence from the L1—in this case Chinese—blocked the cues from the L2 (Latin), because 8 the Chinese-speaking participants had learned to pay attention to lexical items in their L1 instead of inflectional morphology. However, other researchers argue that the concept of blocking or learned attention (Ellis & Sagarra, 2010) are inaccurate, because L2 learners are relying on universal default processing strategies that allow them to comprehend sentences in real time with a parser that is limited by the amount of information it can take in (VanPatten, 2012). For example, one such strategies called the First Noun Principle argue that nonnatives may assign the role of subject to the first noun that appears in the sentence, regardless of its actual role and regardless of L1 influence (see LoCoco, 1987; VanPatten, 1984, 1990, 1996, 2003). Another universal processing strategy, the Lexical Primacy Principle, argues that L2 learners may rely on lexical items rather than inflectional morphology. That is, between these two principles, nonnative speakers may show a preference for attaching a verb to the nearest overt reference despite the cues provided by morphological inflections (Mandell, 2013). For example, in (10) (Mandell, 2013, p. 22), the verb is marked with 2sg, which indicates that a new subject (i.e. not ‘Franco’) has been introduced. However, if the processing of inflection of ‘hablas’ does not occur, the verb may be automatically co-referenced with ‘Franco’, because it is the nearest available overt referent. (10) Franco dice que hablas bien el español. Franco says that speak.2sg. well the Spanish. ‘Franco says you speak Spanish well.’ L1 English/L2 Spanish as a Test Case L1 English/L2 Spanish participants provide the opportunity to examine the development of morpho-syntax and processing behavior in several ways. First, L1 English learners must reassemble features that are minimally present in their L1 to match the expected features of L2 9 Spanish, as well as the necessary syntactic structures (not tested in the current dissertation) (McCarthy, 2012; VanPatten et al., 2012). This is required for Spanish not only because Spanish is morphologically rich, but also because Spanish is a null subject language that permits subjects to be represented with an unspoken pro. This means both native speakers and L2 learners must rely on verbal morphology to reliably ascertain subjects. For example, in sentence (11), the only indication that the verb in the second clause is 2sg is the morpheme –s. (11) Franco dice que hablas bien el español. Franco says that speak.2sg. well the Spanish. ‘Franco says you speak Spanish well.’ Second, L1 English/L2 Spanish learners must also be able to access this information during real time processing by developing native-like processing heuristics. However, the research investigating L2 development has tended to indicate that even after inflection has been acquired, L2 learners may not reliably process inflection in real time due to cognitive constraints (McDonald, 2000, 2006; Hopp, 2010; LoCoco, 1987; VanPatten, 1984, 1990, 1996, 2003). Finally, due to a failure to acquire L2 processing heuristics or limited processing capabilities, L1 English/L2 Spanish participants may revert to L1 or universal processing strategies, even if a robust representation of morphological items has developed (Hopp, 2010; Roberts et al., 2008). In the case of subject/verb agreement, this may take the form of reliance on or a preference for overt subjects over null subjects. Thus, the examination of L1 English/L2 Spanish allows me to investigate various aspects of morphosyntactic and processing development. 10 The Current Dissertation The current dissertation has three purposes. The first is to establish the extent to which intermediate-level L1 English/L2 Spanish learners have developed robust mental representations of inflectional morphology in L2 Spanish. This is necessary before addressing the second purpose, which is to examine whether L2 learners can reliably access mental representations within the morpho-lexical store while processing sentences. The third purpose is to examine the extent to which L1 English/L2 Spanish processing is facilitated by overt subjects. To address the first purpose, L1 English/L2 Spanish learners read Spanish sentences with grammatical and ungrammatical subject/verb agreement, and their reading times were collected. Reading times on grammatical and ungrammatical sentences were compared to examine whether participants showed sensitivity to errors in subject/verb agreement by reading ungrammatical sentences longer than grammatical sentences. This portion of the dissertation is a partial replication of VanPatten et al. (2012), and tests learners’ sensitivity to errors in person and number. VanPatten et al. (2012) also tested learner sensitivity to verb raising and adverb placement, which is omitted in the present study. If learners have morphemes robustly represented within the morpho-lexical store, they should slow down while reading sentences that contain errors in subject-verb agreement. A second set of bi-clausal sentences was created to test whether L2 learners can reliably access these morphemes while processing sentences, when they have robust representations within the morpho-lexical store. In these sentences, the first clause contained an overt subject, and the second clause contained a null subject that either matched the subject in the first clause or did not. These sentences examined whether L2 learners had the ability to process verbal morphology reliably, as they should slow down when a new subject is introduced in mismatched 11 clauses. If L2 learners cannot reliably access morphemes within the morpho-lexical store, they should read both sets of sentences for approximately the same amount of time. Such behavior may also indicate a reliance on overt subjects for determining the subject of a sentence and for feature checking procedures. Finally, the third set of stimuli that tested L2 learner reliance on overt pronouns to process subject/verb agreement was the same as the second set, but with overt pronouns in both clauses. In this case, L2 learners may demonstrate more native-like behavior when an overt subject marks a topic/subject shift compared to their behavior on sentences in which a topic/subject shift is marked only by morphological inflection. Overview of Dissertation This dissertation is structured in the following way. Chapter Two presents the results of an experiment designed to test whether L1 English/L2 Spanish had developed robust enough representations of individual morphemes to show sensitivity to errors in subject/verb agreement in L2 Spanish. Chapter Three presents the results of an experiment designed to the examine whether L1 English/L2 Spanish participants could reliably access those representations to process topic/subject shifts in Spanish. Chapter Four presents the results of an experiment designed to test whether overt subjects facilitated L2 learners’ processing of topic/subject shifts. The final chapter presents a general discussion and conclusion based on the three experiments. 12 CHAPTER TWO EXPERIMENT ONE The purpose of the present experiment is to determine L1 English/B1-level L2 Spanish speakers have acquired the necessary morphological inflections to process subject/verb agreement in Spanish. In a partial replication of VanPatten et al. (2012), the current experiment uses self-paced reading experimental methodology to examine whether L2 learners detect errors in 1sg, 2sg, 3sg, and 3pl inflectional morphology during self-paced reading task. When features are shared across languages, however minimally, these features must be reassembled over time to match the expectations of the L2 (Lardiere, 2008, 2009). However, this process is slow, and occurs as learners are exposed to meaningful and comprehensible input in the L2. Furthermore, this process may be subject to parsing errors (Spinner, 2013), and prone to errors as reassembly occurs (Dekydtspotter & Renaud, 2009). Thus, the current experiment establishes a baseline for what present-tense morphemes have been reassembled by L1 English/L2 Spanish learners at the CEFR B1-level of acquisition. The current study also examines whether the acquisition of these features follows proposed feature specification hierarchies proposed by Harley & Ritter (2000, 2002) that seem to predict L1 acquisition of inflection morphology. Background and Motivation Spanish Verbal Morphology In Spanish, each verbal morpheme carries features that mark it for person, number, tense, and how it attaches to verbs (_V) (Table 2.1). For example, the verb hablo (‘I speak’) is marked with the morpheme (-o), which carries features for first person [-addressee] and singular [-group]. It is also attached to the end of the verb (habl-), fulfilling the (_V) placement of the morpheme. Although one morpheme may be marked for several features, each feature operates 13 independently of the others (Bejar, 2003). The targeted features for this dissertation are the phifeatures person and number on present tense verbal morphology. Each morpheme carries a degree of specification (or underspecification), which is both independent of and relative to other inflectional morphemes (Hanson, 2000; Harley & Ritter, 2000, 2002). Crucial to the current dissertation are two nodes proposed by Harley and Ritter (2000, 2002): Participant and Individuation. Person features are encoded under Participant, and Number features are encoded under Individuation. Under this analysis, third person features (Table 2.1) are the absence of a person (Participant) feature, and is marked only with marker of number (Individuation) (Harley & Ritter, 2002). First person, on the other hand, is represented by a bare, underspecified Participant node, and receives the default interpretation of Speaker [-addressee]. It is also marked with a [-group] feature specification. Second person also contains a Participant node, although this node is specified, and receives the marked interpretation of Addressee. Thus, the implicational hierarchy for person is argued to be 3sg/1sg < 2sg (Austin, 2015; Hanson, 2000; Harley & Ritter, 2002). Number is similarly constructed with singular forms being marked with a bare Individuation node, whereas plural’s Individuation node is encoded with the feature for [+group]. Again, as singular is the bare Individuation node, the implicational hierarchy of these forms is singular < plural (Harley & Ritter, 2002). Table 2.1 Summary of Person/Number (Participant/Individuation) Features in Spanish Singular Plural First person -o -mos [-addressee, -group] [-addressee, +group] Second person Third person -s [+addressee, -group] -Ø [-group] 14 -is [+addressee, +group] -n [+group] Due to these geometries, Hanson (2000) and Harley and Ritter (2002) propose that 1sg and 3sg will emerge simultaneously in first language production as default forms of their specific categories: Participant and Individuation respectively. They are followed by second person forms. Similarly, singular forms are expected to be acquired before plural forms. A summary of these features, along with their respective inflections can be seen in Table 2.1. Features marked with a + value are considered specified in that category. The geometries proposed by Harley and Ritter (2000, 2002) are of interest to the current study, because it has been shown to predict the order of emergence of inflectional morphology in L1 Spanish acquisition. For example, Austin (2015) found that in adult L1 Spanish, verbal morphemes tended to be used with the following relative frequencies when speaking with children: 3rd sg > 2nd sg > 3rd pl > 1st sg > 1st pl > 2nd pl. However, beyond 3rd singular emerging first, there was no other correlation between the adults’ and children’s speech. 1sg emerged nearly simultaneously with 3sg, and was subsequently followed by 2sg (e.g., 3sg/1sg > 2sg). These findings suggest that L1 acquisition primarily occurs along the lines of Harley & Ritter’s (2000; 2002) proposed feature geometries, rather than input frequency (Clahsen, Aveledo, & Roca, 2002; Aguado-Orea & Pine, 2015). Such results suggest that in L1 acquisition, children choose a “base” or “default” morpheme that serves as a base for determining the structure of the linguistic paradigm (Bybee, 1985; Albright, 2002). 1sg serves as default Participant node and 3sg serves as the default Individuation node. The construction of a morphological paradigm in L1 acquisition is important to consider when examining the acquisition of verbal morphology in the L2, as it may serve as blueprint for what can be expected. If L2 acquisition is not fundamentally different from L1 acquisition, and if 15 the hypotheses of Bybee (1985) and Albright (2002) are correct, then L2 learners’ behavior should mimic the behavior of L1 learners. Predictions about L2 Learner Behavior on L2 Verbal Morphology Based on the feature hierarchies proposed by Harley and Ritter (2000) and L1 Spanish speaker data, L2 learners can be expected to make two types of errors if incomplete acquisition is observed. The first type of error is called feature clash, and occurs when an over-specified morpheme is used where a less-specified form is required (McCarthy, 2006, 2008) (1). (1) *El chico hablas español. The boy Spanish. speak.2sg ‘The boy speaks Spanish.’ In the given example, the 2sg form (-s) is used where 3sg (-a) is required. Because third person singular subjects only are marked only with [-group] features, the specified [+addressee] feature of 2sg clashes with features of third person singular. This is an error of feature clash. Feature default errors, on the other hand, are errors that occur when a less-specified morpheme is used in place of a more specified morpheme (Lopez Prego, 2012; McCarthy, 2008) (see example 2). (2) *Tú habla con tus amigos. You speak.3sg with your friends. ‘You are speaking with my friends.’ In example (2), the third person singular null morpheme [-group] is used in a context that requires the second person singular morpheme (-s) [+addressee, -group]. This type of error is called a default error, because the verb (habla) does not carry any features that clash with the features present on the subject. 16 Of the two different error types, L2 learners tend to commit defaulting errors with more frequency than feature clash errors (Bartning, 2000; Lopez-Prego, 2012; McCarthy, 2008, 2012; Prévost & White, 2000; Renaud, 2010, 2014; White, 2003; White et al., 2004). There are several possible reasons that default errors are more common than feature clash errors. The first reason is that classroom-based L2 learners may be exposed to default morphology more frequently in the input (VanPatten et al., 2012). As such, L2 learners may use default forms instead of more marked forms, simply because these forms are the most robustly represented in the input. A second reason is based on what is seen in L1 child acquisition, and related to the feature hierarchy (Harley & Ritter, 2000). As mentioned previously, L1 Spanish children choose the morpheme with the least amount of specification as a “base” to build representations of features present in their language (Albright, 2002; Austin, 2015; Bybee, 1985). Because of this, L1 children tend to make errors of underspecification, and produce not specified or underspecified inflections in contexts where more specified inflections are called for. L2 learners may make similar errors of underspecification, and may also routinely use such inflections as a default when no other forms are either readily available or known (McCarthy, 2008; 2012). Each of these explanations is mutually compatible with the observation that L2 learners may not show sensitivity to errors like those in (2), because they are not sensitive to feature default errors. VanPatten, Keating, and Leeser (2012) To test whether L2 learners show sensitivity to errors in subject-verb agreement and thus some underlying knowledge of the relationship between subjects and verbal inflection in Spanish, the current study partially replicates VanPatten et al. (2012). The experiment conducted in this study addressed two aspects of L2 Spanish acquisition: the acquisition of syntax and the acquisition of morphology. Specifically, the researchers asked whether intermediate level 17 learners’ behavior was the result of representation of verbal morphemes in the morpho-lexical store (i.e., mental representation) or whether problematic processing behavior, caused by difficulties with real-time access. The participants in their study were L1 English/L2 Spanish learners, whose Spanish proficiency was described as non-advanced. Furthermore, the participants were primarily classroom-based learners of Spanish, as they had not studied abroad or had extensive contact with the L2 outside of the classroom. To test whether L2 learners had acquired L2 Spanish inflectional morphology, and to ensure that representation was not the cause of nonnative-like performance, the researchers created a set of segmented experimental quadruplets that tested for sensitivity to errors in subject/verb agreement on present tense -ar verbs (3) (VanPatten et al., 2012, p. 120) (3) a. Ahora Pedro now Pedro.3sg b. *Ahora Pedro now Pedro.3sg c. Ahora yo now I.1sg d. *Ahora yo now I.1sg toma el refresco en el salón. drinks.3sg the soft drink in the living room tomo el refresco drinks.1sg the soft drink in the living room en el salón. tomo el refresco en el salón. drinks.1sg the soft drink in the living room toma el refresco drinks.3sg the soft drink in the living room en el salón. ‘Right now, Pedro/I drinks/drink the soft drink in the living room.’ In example (3), sentences A and C are correct, because the verb is marked with the inflectional morphology that matches the sentence subject. Sentences B and D are incorrect, because the verb is incorrectly conjugated for the sentences’ subjects. The authors created quadruplets that alternated 1sg (-o) and 3sg (-Ø) and 2sg (-s) with 3pl (-n) inflectional morphology so that the 18 length of the verb was maintained. The experimental quadruplets were segmented as shown above, and presented in a non-cumulative, self-paced reading procedure. Participants read each sentence at their own pace before being presented with a comprehension check in English to ensure that participants understood the sentences. The results of this study indicated that L2 Spanish learners had not yet acquired robust enough representations within the morpho-lexical store to process inflection reliably. The nonnative participants did not show any sensitivity to any error in morphological inflection. They also did not show any sign of defaulting behavior (the placement of a less specified form in a context where a more specified form is called for), which would also be predicted if a processing issue were taking place. Thus, VanPatten et al. (2012) concluded that intermediate L2 learners may not yet have had enough L2 exposure to build robust representations of the morphology itself. There are several benefits to a replication of VanPatten et al. (2012) for the current dissertation. First, this study deviates from traditional measures (e.g., paper and pencil-based tests) of measuring the acquisition of verbal morphology by using an on-line processing measure (i.e., self-paced reading). As such, it is more likely address the issue of mental representation than off-line tasks that invite reflection and explicit knowledge (e.g., Bruhn de Garavito, 2003; McCarthy, 2008). Second, although the nonnative participants did not show sensitivity to errors in VanPatten et al. (2012), the same methodology used with a more advanced group with controlled proficiency may demonstrate different results. Finally, replicating this study allows for the examination of inflectional morphology across similar conditions. That is, through a replication of this study, it is possible to the examine whether L2 learners show sensitivity to 19 errors in a simple, on-line context, before asking whether L2 learners can reliably process morphology in the discourse. The Present Study The present study partially replicates VanPatten et al. (2012) by replicating the portion of their study that examined inflectional verbal morphology, and examined whether L2 learners demonstrate sensitivity to errors in the present tense. The present student does not replicate a portion of the study by VanPatten et al. that examines verb raising and adverb placement. Specifically, the research question for this section is: 1. Do L2 Spanish speakers at B1 proficiency show sensitivity to errors on present tense verbal morphology on regular –ar verbs as measured by a self-paced reading task for the following morphemes: a. 1st person singular –o b. 2nd person singular –s c. 3rd person singular –Ø d. 3rd person plural –n? If L1 English/L2 Spanish learners have built robust enough mental representations of inflectional verbal morphology, this should show up as a slowing down in reading when encountering a mismatch between subject and verbal inflection. If they have not, then L2 learners should not slow down when encountering such mismatches. It is also equally possible that L2 learners will show robust representations of some morphemes, but not others. In other words, they may show incomplete acquisition of person-number morphology due to their proficiency level. 20 Method and Procedure Participants Nonnative Speakers. Thirty-nine nonnative Spanish speakers were recruited from upperlevel Spanish language classes at Michigan State University. All participants spoke English as a first language. Participants who spoke heritage languages that had rich verbal morphology or that allowed for the use of null subjects were excluded, which lead to one participant being removed. Each participant was given a reading proficiency test in Spanish to ensure that they could read at the B1 proficiency level. Only participants who score more than 60% (significantly above and different from chance) on the test were included in the final analyses. Based on the reading proficiency test, sixteen additional participants were removed from the analyses, leaving a total of n = 22 L1 English/L2 Spanish speakers. The remaining 22 participants had to score above 75% accuracy on comprehension questions during the self-paced reading task to be included in the final data set. This ensured that participants were not only reading each stimulus sentence for meaning, but also that they understood what they were reading. All 22 remaining participants scored above this threshold. There were 19 females and 3 males in the final 22 participants. They were an average of M = 20.67 years old (SD = 1.63 years), and 11 of them had studied abroad for six weeks or more. Native Speakers. Seventeen native speaking participants were recruited from Michigan State University. Two participants were eliminated, because they were heritage speakers of Spanish, and had been raised in the United States. The remaining fifteen participants had all grown up in a Spanish-speaking country until age 12, and all had Spanish as an L1 with no childhood bilingualism. They were an average of M = 32.60 years old (SD = 5.93 years). The 21 average age of arrival in the United States for participants was M = 24.20 years (SD = 7.06 years). Other than English, five participants claimed to speak an additional L2. Materials As part of this dissertation, participants completed a self-paced reading task, and took one proficiency measure. Each of these will be discussed in turn. Stimuli. One-hundred and twenty experimental quadruplets were created for this experiment (see Appendix A), and were created based on VanPatten et al.’s (2012) stimuli. All sentences were segmented for a non-cumulative, moving-window, self-paced reading procedure in SuperLab 5.0. Forty of these stimuli were created to test participants’ sensitivity to errors in person (1st person singular-3rd person singular) as below: (4) A. Ahora \ yo \ limpio \ la \ casa \ de \ la cliente. B. *Ahora \ yo \ limpia \ la \ casa \ de \ la cliente. C. Ahora \ la criada \ limpia \ la \ casa \ de \ la cliente. D. *Ahora \ la criada \ limpio \ la \ casa \ de \ la cliente. ‘Right now, I/the maid am/is cleaning the client’s house.’ 1sg (-o) and 3sg (-Ø) were contained in one experimental quadruplet (example (4)) to maintain the length of the verb in the target region (bolded for demonstrative purposes in example (4)). In other words, whenever a 1sg subject was presented with an ungrammatical verb, the verb was always marked with 3sg morphology, as in sentence B in example (4). Whenever a 3sg subject was presented with an ungrammatical verb, as in sentence D in (4), this verb was always marked with 1sg morphology. Another forty sentences were created to test participants’ sensitivity to errors in person and number agreement (2sg-3pl) as below: 22 (5) A. Ahora \ tú \ preparas \ la \ cena \ para \ los niños. B. *Ahora \ tú \ preparan \ la \ cena \ para \ los niños. C. Ahora \ las madres \ preparan \ la \ cena \ para \ los niños . D. *Ahora \ las madres \ preparas \ la \ cena \ para \ los niños. ‘Right now, you/the mothers are preparing dinner for the children.’ 2sg-marked verbs (-s) alternated with 3pl-marked verbs (-n) to maintain the length of the verb in the target region (bolded for demonstrative purposes in example (4)). Whenever a 2sg subject was presented with an ungrammatical verb, that verbal was always marked with 3pl morphology, as in sentence B in example (5). Whenever a 3pl subject was presented with an ungrammatical verb, that verb was always marked with 2sg morphology, as in sentence D in example (5). The remaining forty sentences were created to test participants’ sensitivity to errors in number agreement (3rd singular-3rd person plural) as below: (6) A. Ahora \ el cliente \ considera \ la \ oferta \ de \ los jefes. B. *Ahora \ el cliente \ consideran \ la \ oferta \ de \ los jefes. C. Ahora \ los clientes \ consideran \ la \ oferta \ de \ los jefes. D. *Ahora \ los clientes \ considera \ la \ oferta \ de \ los jefes. ‘Right now the client/the clients considers/consider the bosses’ offer.’ In these sentences, whenever a 3sg subject was presented in an ungrammatical sentence, as in sentence B in (6), the verb was marked with 3pl morphology. Similarly, whenever a 3pl subject was presented in an ungrammatical sentence, as in sentence D in (6), the verb was marked with 3sg morphology. Each stimulus consisted of seven separate regions. Following VanPatten et al. (2012), the first region always contained an adverb (Ahora ‘now’) or phrase (En este momento ‘at this 23 moment’) that indicated that the sentence in the present tense. This was done so that participants had to interpret the sentence as being in the present tense, and to avoid ambiguity with the morphological inflection -ó (3sg, past) in the case of the 1st person singular/3rd person singular alternations in example (4). The second region always contained an overt pronoun or noun phrase that was the subject of the sentence. The sentences with 3sg or 3pl subjects all consisted of noun phrases, such as el chico/los chicos (‘the boy’/‘the boys’), to eliminate any issues with the ambiguity of subjects such as él/ellos (‘he’/‘they’), given that these pronouns do not normally occur in the discourse without a referent. The third, bolded region is the target region. This region contains the verb, which is correctly conjugated in sentences A and C, and incorrectly conjugated in sentences B and D in each of the experimental quadruplets. If L2 learners are sensitive to errors in subject-verb agreement, reading times for sentences B and D are predicted to be longer than A and C. This is because the ungrammaticality of B and D should force longer processing times as participants’ parsers are forced to reconcile the ungrammaticality. These quadruplets were blocked to create a total of four participant groups. Finally, only -ar verbs with regular endings were used to avoid any possible confounds between -ar and -er/-ir verbs. Each target region was followed by three spillover regions that consisted of a determiner or preposition in the fourth region, and a noun in the fifth region. The sixth region contained another preposition. The final region was not controlled, and various phrases appeared in this position. Each sentence was followed by a comprehension check in English (7). English was chosen as the language of the comprehension checks to avoid giving participants additional cues related to Spanish verbal morphology. Each comprehension check was in the form of a statement that participants decided was true or false based on the sentence they just read. All sentences 24 targeted the context of the sentence, and did not directly address the issue of the subject. These questions were made so that half of the comprehension checks were true (correct answer: ‘A’) and half of the comprehension checks were false (correct answer: ‘B’). In (7), the answer is ‘True’. (7) Ahora \ las madres \ preparan \ la \ cena \ para \ los niños. ‘Right now, the mothers prepare dinner for the children.’ Somebody in this sentence is cooking for someone else. A. True B. False Finally, the stimuli from experiments two and three served as distractors for Experiment One. These stimuli will be discussed in their appropriate chapter. Proficiency Test. As part of the experiment, all native and nonnative participants were also asked to take a proficiency test. The test was a subsection of example materials from the reading portion of the Diploma de Español Como Lengua Extranjera (DELE) exam, taken from the Instituto de Cervantes DELE exam website (https://examenes.cervantes.es/es/dele/prepararprueba). The test consisted of total of 18 questions from three separate tasks. The first task consisted of six questions that asked learners to read descriptions and match movie-goers to the movie that best suited their wants. The second task consisted of six questions and asked participants to read the profiles of three athletes, and to subsequently check which question applied to which athlete. The final task was a cloze test, and consisted of six questions. This task asked participants to read an email sent to the director of a company about an error that occurred online. The test contained blank spaces and participants had to choose the word that best filled in the blank from three possible options. These subsections of the test targeted reading at the B1 25 Spanish proficiency level, and were chosen based on the piloting of the exam with a subset of twelve Spanish language students from upper-level language courses. This B1 level, as described by the Common European Framework, consists of learners who able to understand the ideas of clear texts that are related to hobbies with which participants are familiar. These learners can also handle most everyday interactions that can occur during travel in Spanish-speaking countries (Instituto Cervantes, 2016). This reading test was chosen, because the task that participants were asked to complete was a relatively simple reading task. Additionally, only three subsections were chosen to limit the amount of time that participants spent in the lab to a 90-minute maximum. The proficiency test was found to be reliable with a Chronbach’s alpha of a = .938, indicating that the test was reliable. In general, the native control group scored higher than the nonnative group (M = 17.20, SD = 1.15), as expected. The lowest score for the native group was 11 and the highest score was 18. The nonnative group had a mean of M = 13.45 (SD = 1.85). The lowest score for the nonnative group was 11 and the highest was 16. These results were submitted to a two-sample t-test, which confirmed that the nonnative group was significantly different from the native control; F(1, 36) = 48.74, p < .001. This indicates that the nonnative group was less proficient than the native control, although they could read at an intermediate (B1) proficiency level. Procedure Participants came to a lab where they sat comfortably in a chair in front of a computer with a response pad with seven keys to start the experiment. Up to four participants were seated in the lab at any given time, and were paid $20 each for their time. After reading the consent form and indicating their consent, participants began the experiment. The instructions for the experiment informed participants that they would be reading sentences that would be presented 26 in segments, and that segments would be revealed as they pressed the middle green button on the response pad in front of them. This dissertation utilized a moving-window, non-cumulative, selfpaced reading procedure. This means that as participants pressed the middle green button, the sentence fragment that was on the screen disappeared and the following segment appeared. Participants could not go back to reread the sentence. During the instructions, participants practiced with this procedure with five Spanish sentences. These sentences used the first person plural verb form -mos to avoid providing participants with additional cues on the targeted verbal morphology as seen in (8). (8) Normalmente \ vamos \ a \ la \ universidad \ por \ la \ mañana. ‘Normally, we go to the university in the morning.’ The people in this sentence are probably students. A. True B. False Each practice question was also presented with a true/false comprehension check to allow participants to practice with the seven-button response pad. To indicate a ‘True’ response, participants were told to press the button furthest to the left marked ‘A’. To indicate a ‘False’ response, participants were told to press the button furthest to the right marked ‘B’. To separate each stimulus and give participants the opportunity to pause, a focus slide with a plus sign (+) in the center of the screen appeared after each comprehension check question. During the instructions participants were told to pause only when they saw the plus sign on the screen. After practicing with the procedure, participants began the experiment. Participants moved through the stimuli sentences at their own pace. Half-way through, the experiment was programmed to pause, and participants were instructed that they should stop 27 to take a proficiency test. Participants then took the 18-question proficiency test in paper form. When they finished the proficiency test, they continued to the second section of the experiment. They again continued at their own pace until the experiment’s end. Once they finished the second section, they were asked to fill out a background questionnaire (either Appendix D or E, depending on the participant’s group), and were paid for their time. The total experiment time was approximately 60-90 minutes in length. After receiving payment, participants were asked what they thought the experiment was about, and were debriefed if they liked. Scoring and Analysis The participants’ reading times were collected for the target region and the three spillover regions. Consistent with self-paced reading methodology, reading times that were less than 100ms were removed from the analysis, because reading times under 100ms may signify that participants did not consciously read the words in the targeted region, although priming may occur. Removing these times consisted of elimination of .004% of the data for this experiment. To analyze the data, linear mixed models were conducted for each of the three conditions (i.e., first person singular/third person singular; second person singular/third person plural; third person singular/third person plural) regions. Linear mixed models were chosen because of their ability to deal with robust, correlated, non-evenly distributed data (Baayen, Davidson, Bates, 2008). The traditional way of analyzing reading times has been ANOVAs, which assume that data are independent and lack the ability to deal with outlying reading times (Jegerski, 2014; Keating & Jegerski, 2015). The traditional analyses can also cause data loss, as the method employed to deal with outlier reading times is to replace them with the participant’s mean for that condition. Thus, linear mixed models were chosen for their ability to account for multiple observations, outlying reading times, and random factors, such as participant and item. 28 Each linear mixed model had two random and two fixed factors. The two random factors were participant and item. The two fixed factors were morpheme (-o or –-Ø, -s or –n, and –a and –n) and grammaticality (i.e., grammatical or ungrammatical). There were two possible ways to analyze the data. The first option was to hold the verb form constant, so that sentence A in example (9) was compared to sentence D, and sentence B was compared to sentence C. The second option was to hold the subject constant, so that sentence A was compared to sentence B, and sentence C was compared to sentence D. (9) A. Ahora \ yo \ limpio \ la \ casa \ de \ la cliente. B. *Ahora \ yo \ limpia \ la \ casa \ de \ la cliente. C. Ahora \ la criada \ limpia \ la \ casa \ de \ la cliente. D. *Ahora \ la criada \ limpio \ la \ casa \ de \ la cliente. ‘Right now, I/the maid am/is cleaning the client’s house.’ To determine which analyses should be used, the native control group was examined. As natives, they were expected to show sensitivity to errors in every condition; however, this was not the case. In fact, natives demonstrated no sensitivity to third person plural in the first set of analyses, where the inflection marked on the verb was maintained (e.g., *tú hablan vs. los chicos hablan). They also showed a pattern of reading 3pl subjects longer than 2sg forms. This demonstrates a possible confound in the data due to longer subjects in the area preceding the target region. In the second analyses, however, native speakers behave in the expected manner consistently across conditions with no significant confounds due to the length of the subject region. For this reason, the results from the second set of analyses were considered more controlled and more reliable, and ultimately used. 29 For the 3sg/3pl condition (10), only one set of analyses was possible due to the length of the target region. That is, to ensure that one inflection was not more salient than another, each morpheme was compared to the same form but in an ungrammatical context (e.g., *el cliente consideran vs. los clientes consideran). For these stimuli, sentence A was compared to sentence D, and sentence B was compared to sentence C. The native control group performed as expected in this condition, and slowed down significantly when presented with an ungrammatical sentence compared to a grammatical sentence. (10) A. Ahora \ el cliente \ considera \ la \ oferta \ de \ los jefes. B. Ahora \ el cliente \ consideran \ la \ oferta \ de \ los jefes. C. Ahora \ los clientes \ consideran \ la \ oferta \ de \ los jefes. D. Ahora \ los clientes \ considera \ la \ oferta \ de \ los jefes. ‘Right now, the client/the clients considers/consider the boss’s offer.’ For the linear mixed effects models to be valid, several steps were taken to ensure that the model used was the best fit. First, the covariance structure of the model was first tested using two covariance structure matrixes due to the correlated nature of the data: compound symmetry and unstructured. To determine which covariance structure was best, the model’s AICs were compared. For determining model fit, the lower the AIC, the better the model fits. For each model, the unstructured covariance structure had the lowest AIC, indicating that this structure best fit the data. In addition to using the AIC score to determine the appropriate covariance structure, the residuals were plotted for each region; an example of which can be seen in Figure 2.1. 30 Figure 2.1 Sample Plot of Residuals for Nonnative Speaker Target Region Residuals that fit the model well are those that do not cluster around the line marking zero, nor diverge from the line in a linear fashion. As seen in Figure 2.1, the residuals based on reading times for the target region are not ideally distributed, because they diverge from the zero line linearly, and are also clustered together. Subsequent regions also had similar patterns in their residuals. Because of this, all data were log-transformed. When transformed into logs, the residuals were more appropriately distributed and the model fit the data better, as shown in the Figure 2.2, which represents the log-transformed data for the target region. Finally, another important factor in ensuring that the model is valid is ensuring that all fixed factors placed in the model are significant predictors. Using the unstructured model, each fixed factor was examined, and it was found that each fixed factor was significant in the model. 31 Figure 2.2 Sample Log-Transformed Residuals for Nonnative First Person Singular/Third Person Singular Target Region Results Native Speakers 1sg/3sg Condition. Descriptive statistics for the native control group can be found in Table 2.2. As for the nonnative group, linear mixed models with an unstructured covariance structure. Table 2.2 Descriptive Statistics of Native Speakers for 1sg/3sg in Milliseconds Morpheme Target Spillover Region One Grammatical 411.90 372.48 (yo hablo) (279.13) (130.45) First Person (yo) Ungrammatical 440.67 423.95 (*yo habla) (313.91) (180.51) Third Person (el chico) Spillover Two 427.62 (165.94) Spillover Three 371.73 (103.17) 444.86 (253.11) 418.55 (331.18) Grammatical (el chico habla) 420.51 (181.33) 388.59 (166.86) 443.37 (235.22) 397.04 (120.76) Ungrammatical (*el chico hablo) 435.44 (201.54) 438.14 (265.80) 449.04 (200.13) 391.55 (146.87) In the target region (Table 2.3), neither morpheme, grammaticality, nor any interaction between them were significant predictors in the model for native speakers. 32 Table 2.3 Model Results Analyses for Native Speaker Target Region 1sg/3sg Meaning Fixed Effects Estimate Mean log reading Intercept 2.592 time of the fixed effects included in the model Std. Error 0.013 p value <.001 Effect of the morpheme being first person compared to third person Morpheme -0.019 0.017 .234 Effect of the sentence being ungrammatical versus grammatical Grammaticality 0.013 0.015 .394 Effects of the morpheme being ungrammatical and first person Morpheme*Grammaticality -0.003 0.021 .862 In the first spillover region (Table 2.4), however, grammaticality was a significant predictor in the model (estimate = 0.037, p = .021), indicating that ungrammatical sentences were read longer than grammatical sentences. 33 Table 2.4 Model Results Analyses for Native Speaker First Spillover Region 1sg/3sg Meaning Fixed Effects Estimate Std. Error Mean log reading Intercept 2.567 0.010 time of the fixed effects included in the model p value <.001 Effect of the morpheme being first person compared to third person Morpheme -0.016 0.015 .270 Effect of the sentence being ungrammatical versus grammatical Grammaticality 0.037 0.016 .021 Effects of the morpheme being ungrammatical and first person Morpheme*Grammaticality 0.009 0.023 .700 In the second spillover region, there are no significant predictors or interactions (Table 2.5). In the third spillover region (Table 2.6), there is a trend toward significance for both morpheme (estimate = -0.023, p = .060) and the interaction between morpheme and grammaticality (estimate = 0.033, p = .060). 34 Table 2.5 Model Results Analyses for Native Speaker Second Spillover Region 1sg/3sg Meaning Fixed Effects Estimate Std. Error Mean log reading Intercept 2.606 0.014 time of the fixed effects included in the model p value <.001 Effect of the morpheme being first person compared to third person Morpheme -0.002 0.015 .853 Effect of the sentence being ungrammatical versus grammatical Grammaticality 0.014 0.015 .352 Effects of the morpheme being ungrammatical and first person Morpheme*Grammaticality -0.006 0.020 .761 Table 2.6 Model Results Analyses for Native Speaker Third Spillover Region 1sg/3sg Meaning Fixed Effects Estimate Std. Error Mean log reading Intercept 2.581 0.009 time of the fixed effects included in the model p value <.001 Effect of the morpheme being first person compared to third person Morpheme 0.024 0.013 .060 Effect of the sentence being ungrammatical versus grammatical Grammaticality -0.009 0.014 .527 Effects of the morpheme being ungrammatical and first person Morpheme*Grammaticality 0.033 0.017 .060 35 2sg/3pl Condition. The descriptive statistics for the 2sg/3pl condition for the native control group are presented in Table 2.7. In general, native speakers read ungrammatical sentences longer than grammatical sentences. Table 2.7 Descriptive Statistics of Native Speakers for 2sg/3pl in Milliseconds Morpheme Target Spillover Region One Grammatical 388.41 365.87 Second person (tú hablas) (170.91) (110.09) (-as) Ungrammatical 487.14 459.27 (*tú hablan) (365.64) (231.80) Third person plural (-an) Spillover Two 438.88 (223.05) Spillover Three 387.70 (128.90) 444.23 (218.72) 408.31 (317.23) Grammatical (los chicos hablan) 510.34 (606.06) 401.95 (150.77) 443.91 (194.69) 413.33 (130.89) Ungrammatical (*los chicos hablas) 571.50 (424.13) 474.27 (216.22) 483.83 (289.78) 393.55 (111.65) The results for the target region analysis can be seen in Table 2.8. In this region, morpheme is a significant predictor, indicating that second person singular forms are read faster than third person plural forms (estimate = -0.077, p < . 001). In addition to morpheme being a significant predictor, grammaticality was also a significant predictor in the model (estimate = 0.046, p = .032), with ungrammatical sentences being read longer than grammatical sentences as expected. 36 Table 2.8 Model Results Analyses for Native Speaker Target Region 2sg/3pl Meaning Fixed Effects Estimate Mean log reading Intercept 2.639 time of the fixed effects included in the model Std. Error 0.016 p value <.001 Effect of the morpheme being second person compared to third person Morpheme -0.077 0.017 <.001 Effect of the sentence being ungrammatical versus grammatical Grammaticality 0.046 0.021 .032 Effects of the morpheme being ungrammatical and second person Morpheme*Grammaticality 0.004 0.026 .887 In the first spillover region (Table 2.9), morpheme is, again, a significant predictor, indicating that second person singular forms were read faster than third person plural forms (estimate = 0.034, p = .006). Grammaticality is also a significant predictor in the model (estimate = 0.061, p <.001), with ungrammatical sentences being read longer than grammatical sentences. 37 Table 2.9 Model Results Analyses for Native Speaker First Spillover Region 2sg/3pl Meaning Fixed Effects Estimate Std. Error Mean log reading Intercept 2.581 0.011 time of the fixed effects included in the model p value <.001 Effect of the morpheme being second person compared to third person Morpheme -0.034 0.017 .006 Effect of the sentence being ungrammatical versus grammatical Grammaticality 0.061 0.014 <.001 Effects of the morpheme being ungrammatical and second person Morpheme*Grammaticality 0.012 0.021 .553 In the second spillover region (Table 2.10), the grammaticality predictor trends toward significance consistent with the pattern observed in the previous regions (grammatical < ungrammatical). 38 Table 2.10 Model Results Analyses for Native Speaker Second Spillover Region 2sg/3pl Meaning Fixed Effects Estimate Std. Error Mean log reading Intercept 2.611 0.014 time of the fixed effects included in the model p value <.001 Effect of the morpheme being second person compared to third person Morpheme -0.008 0.017 .631 Effect of the sentence being ungrammatical versus grammatical Grammaticality 0.031 0.017 .065 Effects of the morpheme being ungrammatical and second person Morpheme*Grammaticality -0.018 0.023 .418 In the final spillover region (seen in Table 2.11), morpheme continues to be significant as a predictor (estimate = -0.027, p = .028), while grammaticality trends toward significance (estimate = -0.019, p = .074). In both cases, the patterns of the previous regions are maintained (2sg < 3pl; grammatical