The Syntax-Information Structure Interface: Clausal Word Order and the Left Periphery in Galician 9781614512059, 9781614512714

Citation preview

Timothy Gupton The Syntax-Information Structure Interface

Interface Explorations

Editors Artemis Alexiadou and T. Alan Hall

Volume 29

Timothy Gupton

The Syntax-Information Structure Interface

Clausal Word Order and the Left Periphery in Galician

ISBN 978-1-61451-271-4 e-ISBN (PDF) 978-1-61451-205-9 e-ISBN (EPUB) 978-1-5015-0030-5 ISSN 1861-4167 Library of Congress Cataloging-in-Publication Data A CIP catalog record for this book has been applied for at the Library of Congress. Bibliographic information published by the Deutsche Nationalbibliothek The Deutsche Nationalbibliothek lists this publication in the Deutsche Nationalbibliografie; detailed bibliographic data are available on the internet at http://dnb.dnb.de. © 2014 Walter de Gruyter GmbH, Berlin/Boston Typesetting: PTP-Berlin Protago-TEX-Production GmbH, Berlin Printing and binding: CPI books GmbH, Leck ♾ Printed on acid-free paper Printed in Germany www.degruyter.com

Between thought and expression lies a lifetime. – Lou Reed, Some Kinda Love

to Melissa, my love

Contents List of Tables and Figures | XIII Preface | XVII 1 1.1 1.2 1.2.1 1.2.2 1.3 1.4

Introduction | 1 A brief socio-linguistic history of Galician | 1 The sociolinguistic situation in Galicia | 3 Heritage-speaker bilingualism in Galicia | 4 Self-reported language competence ratings | 7 Interface instability in bilinguals | 11 Syntactic framework – assumptions | 13

2 2.1 2.2 2.3 2.4

2.9.1 2.9.2 2.9.3 2.10 2.11

The interaction between syntax and information structure | 18 Argument vs. Non-argument positions in syntax | 18 Syntactic tests for arguments and non-arguments | 21 The analysis of subjects as non-arguments in Spanish  | 24 Analyses of Spanish preverbal subjects as canonical arguments | 38 The debate on subjects in European Portuguese | 42 Analysis of EP subjects as non-arguments  | 42 Analyses of EP preverbal subjects as canonical arguments | 47 Taking stock of preverbal subject analyses | 61 Preverbal subjects in Galician  | 62 Syntactic structures in context | 66 When syntax meets discourse | 67 Defining Information Structure | 71 Definitions of Topic/Theme | 71 Definitions of Focus/Rheme | 73 Syntactic accounts of the syntax-information structure interface | 83 Casielles (2004) | 87 The Interface and Phases | 92 López’s (2009) interface model | 94 Establishing clausal structure in Galician | 105 Summary | 106

3 3.1 3.2

Methodology | 109 Preliminary concerns | 109 Participants: Tasks 1 and 2 | 111

2.5 2.5.1 2.5.2 2.6 2.7 2.8 2.8.1 2.8.2 2.8.3 2.8.4 2.9

X   

3.2.1 3.2.2 3.3 3.3.1 3.3.2 3.3.3 3.3.4 3.3.5 3.3.6 3.3.7 3.4 3.4.1 3.4.2 3.4.3 3.4.4 3.4.5 3.4.6 3.4.7 3.5 3.6 4 4.1 4.2 4.2.1 4.2.2 4.2.3 4.2.4 4.2.5 4.2.6 4.2.7 4.2.8 4.3 4.3.1 4.3.2 4.3.3 4.3.4 4.3.5 4.3.6 4.3.7

   Contents

Participant variables | 111 Procedures: Tasks 1 and 2 | 117 Task 1: Appropriateness Judgment Task | 118 Condition A | 119 Condition B | 119 Condition C | 121 Condition D | 123 Condition E | 125 Condition F | 126 Condition G | 128 Task 2: Word Order Preference Task | 129 Condition A | 130 Condition B | 131 Condition C | 132 Condition D | 133 Condition E | 134 Condition F | 135 Summary of quantitative experimental tasks | 136 Task 3: Recorded field interview  | 136 Conclusion | 138 Statistical analysis of quantitative and qualitative measures | 139 Introduction | 139 Task 1 | 139 Condition A | 140 Condition B | 141 Condition C | 142 Condition D | 143 Condition E | 144 Condition F | 144 Condition G | 148 Summary of Task 1 discourse conditions | 149 Task 2 | 150 Condition A | 150 Condition B | 151 Condition C | 152 Condition D | 153 Condition E | 155 Condition G | 156 Task 2 by language dominance | 157

Contents   

4.3.8 4.3.9 4.4 4.5 4.6 4.6.1 4.6.2 4.6.3 4.7 5 5.1 5.2 5.2.1 5.2.2 5.3 5.4 5.5 5.5.1 5.5.2 5.6 5.7 5.8 5.9 5.10 5.11

Task 2 statistical results by gender | 158 Task Two Summary | 158 Follow-up task for Task 2  | 159 Task 3 results | 162 Summary and discussion: quantitative measures | 171 Discussion  | 172 Subject CLLD within the SDRT notions of Subordination and Continuation | 175 A note on CLLD response ratings | 178 Final methodological considerations | 179 Toward a Left-peripheral Syntactic Analysis of Galician | 183 Introduction | 183 Main-clause cliticization in Galician | 185 A brief excursus on F | 192 Fernández-Rubiera’s (2009) syntactic account of clitic directionality in WIR | 193 Main clause preverbal elements in Galician | 196 Extending the left periphery: subordinate clause phenomena in Galician | 207 Extending the left-periphery: recomplementation | 213 Recomplementation in Galician | 218 Jussive/optative QUE (che2) | 223 Subject positions in Galician | 225 Affective phrases and preverbal subjects in Galician | 229 Subject positions and information structure in López (2009) | 233 The A vs. Ā debate revisited | 241 Summary | 242 Concluding remarks | 244

6

Appendix A Linguistic questionnaire for initial tasks | 248

7

Appendix B Task 1: Appropriateness Judgment Task | 252 Condition A: Thetic contexts | 252 Condition B: Discourse-old subject (subordination) | 253 Condition C: Discourse-old object (subordination) | 254 Condition D: Discourse-old subject (coordination) | 254 Condition E: Discourse-old object (coordination) | 256

7.1 7.2 7.3 7.4 7.5

   XI

XII   

7.6 7.7 8 8.1 8.2 8.3 8.4 8.5 8.6 8.7 9 9.1 9.2 10 10.1 10.2 10.3 10.4 10.5

   Contents

Condition F: Subject narrow-focus (rheme) | 256 Condition G: Object narrow-focus (rheme) | 257 Appendix C Task 2: Word order preference task | 259 Instructions and Practice Items | 259 Contexto 1: Universidade  | 260 Contexto 2: A entrevista sobre o restaurante  | 261 Contexto 3: Unha mudanza de pesadelo  | 262 Contexto 4: Unha noite de festa | 263 Contexto 5: O escándalo | 264 Contexto 6: A tenda de animais | 265 APPENDIX D Task 3: Recorded field interview | 267 Interview A: Questions for young participants | 267 Interview B: Questions for older participants | 267 Appendix E Follow-up WPT: for narrow-focus in Task 2  | 268 Cuestionario lingüístico | 268 Condition 1: Subject narrow focus (SV vs. VS) | 268 Condition 2: Object narrow focus (SVO vs. VSO) | 269 Condition 3: Object narrow focus (SVO vs. VOS) | 271 Condition 4: Object narrow focus (VSO vs. VOS) | 272

References | 274

List of Tables and Figures Table 1: Table 2: Table 3: Table 4: Table 5: Table 6: Table 7: Table 8:

Table 9: Table 10: Table 11: Table 12: Table 13: Figure 1:

Table 14: Table 15: Table 16: Table 17: Table 18: Table 19:

Self-reported first language(s) percentages by age group in Galicia. | 4 Self-reported habitual language use percentages by age group in Galicia. | 6 Percentages of Galicians surveyed who rated their Galician abilities high. | 7 Percentages of Galicians surveyed who rated their Spanish abilities high. | 7 Percentages of self-reported Galician-only speakers surveyed who rated their Galician abilities high. | 8 Percentages of self-reported Galician-only speakers surveyed who rated their Spanish abilities high. | 8 Percentages of self-reported predominantly Galician speakers surveyed who rated their Galician abilities high. | 9 Percentages of self-reported predominantly Galician speakers surveyed who rated their Spanish abilities as 3 or 4 (4=highest). | 9 Percentages of self-reported predominantly Spanish speakers surveyed who rated their Galician abilities high. | 9 Percentages of self-reported predominantly Spanish speakers surveyed who rated their Spanish abilities high. | 10 Summary of EP word orders and subject positions according to information structure  | 84 Interpretive values assigned by Pragmatics module | 95 [+urban] areas in Pontevedra province and current population figures. | 113 Sociolinguistic Map of Galicia. Percentages of participants who rated their Galician abilities as ‘high’ (3) or ‘very high’ (4) by education level. (González González et al. 2007: 225–228).  | 113 Summary of participant variables. | 114 Summary of gender, age and language-dominance variables | 117 Summary of information structure contexts appearing in tasks one and two.  | 136 Descriptive statistics, Task 1 Condition A (thetic sentences) | 140 Friedman statistics, Task 1 Condition A (thetic sentences) | 140 Descriptive statistics, Task 1 Condition B (common-ground subject – subordination) | 141

XIV   

   Contents

Table 20: Table 21: Table 22: Table 23: Table 24: Table 25: Table 26: Table 27: Table 28: Table 29: Table 30: Table 31: Table 32: Table 33: Table 34: Table 35: Figure 2: Figure 3 Figure 4: Figure 5: Figure 6:

Friedman statistics, Task 1 Condition B (common-ground subject – subordination) | 141 Descriptive statistics, Task 1 Condition C (common-ground object – subordination) | 142 Friedman statistics, Condition C (common-ground object – subordination) | 142 Descriptive statistics, Task 1 Condition D (common-ground subject – coordination) | 143 Friedman statistics, Task 1 Condition D (common-ground subject – coordination) | 143 Descriptive statistics, Task 1 Condition E (common-ground object – coordination) | 144 Friedman statistics, Task 1 Condition E (common-ground object – coordination) | 144 Descriptive statistics, Task 1 Condition F (subject narrow-focus/ rheme) | 145 Friedman statistics, Task 1 Condition F (subject narrow-focus/ rheme) | 145 Descriptive statistics, Task 1 Condition F with adjunct XP | 146 Friedman statistics, Task 1 Condition F with adjunct XP | 146 Descriptive statistics, Task 1 Condition F with DP object | 147 Friedman statistics, Task 1 Condition F word orders with DP object | 147 Descriptive statistics, Task 1 Condition G (object narrow-focus/ rheme) | 148 Friedman statistics, Task 1 Condition G (object narrow focus/ rheme) | 148 Summary of Task 1 word order preferences by discourse condition | 149 Task 2, Condition A mean word order preference proportions by token | 150 Task 2, Condition B mean word order preference proportions by token | 151 Task 2, Condition C mean word order preference proportions by token | 152 Task 2, Condition D mean word order preference proportions by token | 154 Task 2, Condition E mean word order preference proportions by token | 155

Contents   

Figure 7: Table 36: Table 37: Table 38: Table 39: Table 40: Table 41: Table 42:

   XV

Task 2, Condition G mean word order preference proportions by token | 157 Task 2, Condition D. Descriptive statistics for object narrow-focus (VSO vs. VOS) | 158 Summary of Task 2 conditions | 159 Follow-up WPT, Galician-dominant participant representation by age and gender | 160 Follow-up WPT, summary of object narrow-focus preferences  | 161 Revised summary of word order preferences, Task 2 and follow-up WPT  | 161 Summary of quantitative data measures  | 171 Summary of cliticization by clause type and preverbal element in Galician | 209

Preface The topic of this monograph has fascinated me since I first decided to look at the A versus Ā question for preverbal subjects in Galician as a paper topic in Bill Davies’s Advanced Syntax class at University of Iowa in spring 2006. At first, I didn’t expect that it would turn into such a captivating study, but the more I examined it, the more interesting it became. I remember the day that I told Paula Kempchinsky that that would be my dissertation topic. As I recall, she told me I would have to think bigger since so little had been written Galician, particularly from a generative theoretical perspective. I replied that I wanted to look at word order and information structure, but that I wanted to gather experimental data, too. I always needed more than just intuition-based grammar judgments anyhow. Now we were in business. Surely, I had no idea how much work was ahead of me, but I have never tired of it. Of course, there are inevitably loose ends for any study, for whatever reason. One such loose end in this investigation was the intonation of contrast in Galician. I hope my readers will be glad to know that I have continued to examine this topic (or is it focus?) in ongoing research, and look forward to presenting the initial findings of this research in the near future. I am collaborating with phonologists working on intonation, and am optimistic about the outcome. Another loose end is the number of participants in this study. To any who might wonder, I have been planning the follow-up study for this since I stopped gathering data for my dissertation back in 2009. It is on the way, but it will be quite an undertaking. That said, for those readers who do not know, gathering minority language data is a challenging enterprise. There are considerable challenges to gathering linguistic data when one is a relative outsider – not only to the country in question, but also to the language and culture in question. This is made rather more difficult when a minority language is involved, even when one investigates it within its majority context. First and foremost, I would like to thank Paula, my dissertation director, advisor, and mentor, who has been supportive and encouraging since the very first days of this project. I also thank my doctoral dissertation committee at University of Iowa: Jason Rothman, Bill Davies, Roumyana Slabakova, Mercedes Niño-Murcia, and Maria Duarte. I would also like to thank the following for comments and discussion of this manuscript at various stages of its development: Paula Kempchinsky, Julio VillaGarcía, Francesc González i Planas, Pilar Chamorro, Emily Sahakian, Fenton Gardner, Melissa Whatley, and my anonymous DeGruyter reviewer. All errors remain my own.

XVIII   

   Preface

I would like to thank a number of people for their assistance with grammaticality judgments. Of course, I accept full responsibility for all errors and or related misinterpretations. I thank the following in no particular order: my numerous anonymous informants throughout Galicia, Pilar García Mayo at UPV/EHU, Xosé Luís Regueira at USC, the bolseiros at Instituto da Lingua Galega, and Ana Rodríguez Rodríguez at University of Iowa. At the University of Vigo, which served as my base of operations in many respects during the development and data-gathering stages of this project, I would like to thank the following: the Área de Normalización Lingüística, Fernando Ramallo, Eva Garea, Javier Pérez-Guerra, Ana Luna Alonso, Patricia Sotelo, Lidia Xemma Fernández, Rolán Fernández, Marcos Marcote Márquez, Silvia Penas Estàevez, Daniel Barbero Patiño and family, Natalia García Blanco and family, countless kind people in Louredo (Mos) who invited me into their homes, and Elvira Riveiro Tobío and family. I apologize in advance to anyone I inadvertently left off this list. I would like to thank my family for their unflagging support, and lastly, I would like to thank my fiancée Melissa Whatley, for her love, encouragement, patience, and support over the course of the development of this monograph project.

1 Introduction 1.1 A brief socio-linguistic history of Galician The Galician language (Galego) is a Western Iberian Romance variety spoken primarily in the Spanish comunidad autónoma (autonomous community), or region of Galicia in the provinces of Pontevedra, A Coruña, Ourense, and Lugo.¹ It is also spoken in the western portion of the Asturian autonomy (Principality of Asturias) in the areas to the west and east of the Eo and Navia rivers, in O Bierzo and As Portelas in the eastern portions of the Castilla-León autonomy, and in parts of the autonomy of Extremadura in areas such as Cáceres and Val de Xálima. Similar linguistic varieties are also spoken in parts of Portugal adjacent to Galician-speaking areas of Spain, including but not limited to the erstwhile northern Portuguese provinces of Tras-Os-Montes and Minho, which coincide with the contemporary distritos (districts) of Bragança, Vila Real, Braga, and Viana do Castelo. Galician also boasts diaspora communities within other parts of Spain (e.g. Catalunya), and abroad in Buenos Aires (Argentina), Montevideo (Uruguay), La Habana (Cuba), Caracas (Venezuela), and Mexico City (Mexico). Following historical accounts (Freixeiro-Mato 1997; López Valcárcel 1991; Mariño Paz 1998; Monteagudo-Romero 1999), Galician and Portuguese were largely the same language – a language typically referred to as Galego-Portugués (Galician Portuguese) until around the 13th century. As evidenced in documents from the period, it was an everyday language among all social classes, and was the language of administrative, civil and ecclesiastical documents. It was also a literary language, and enjoyed prestige as a language of lyric poetry throughout the Iberian Peninsula. The political division of Portugal from present-day Galicia led to a differentiation of the two languages. Although the Galician language has enjoyed periods of resurgence from time to time over the years, including the present day, it has existed largely as a minority language for over five hundred years. Ramallo (2007) cites the thirteenth century as the origin of Spanish-Galician contact in Galicia, when the kingdom of Galicia became part of the Kingdom of Castile. Castilian Spanish gradually penetrated into Galicia, culminating in the sixteenth century with the installation of Castilian as the official language of the Kingdom of Castile and the disappearance of Galician from official documents (Monteagudo 1999). While Galician was still spoken in rural areas – arguably a

1 Note that the term Galegan is also used in English to refer to the same Romance variety spoken in the Spanish province of Galicia, owing to its greater similarity to the term galego in Galician and gallego in Spanish. I use Galician throughout this monograph.

2   

   Introduction

domain in which it enjoys majority status – it was not until the second half of the nineteenth century that Galician enjoyed a renaissance as a language of culture. Galician continued on this trajectory until the end of the Spanish Civil War and the rise of the Francisco Franco regime (1939–1975), during which Galician suffered a public disappearance. While speakers of Galician were not as viciously persecuted as their Catalan and Basque counterparts (perhaps owing to the fact that Franco himself was from Galicia), Galician was strongly discouraged and stigmatized as a rural, uneducated, and uncivilized language. Urbanization and modernity arrived comparatively late in Galicia, and Spanish was seen as part of this modernity, thus leading to a gradual rise in bilingualism that left Galician relegated to serving as a chiefly rural and familial language. Since the fall of the Franco regime and the transition to democracy, the minority languages of Spain have been legalized, revitalized, and standardized (Siguán 1992), but Galician has not enjoyed the prestige of Basque or Catalan, owing principally to Galicia’s (comparatively) weaker economy. While all official minority languages co-exist with Spanish in situations of diglossia² within the Spanish State, Galician differs in that Spanish is perceived to hold a high position of prestige in Galicia, while Galician is perceived to hold the low position. Galician is considered the language of the poorer, less-educated segment of society, while Spanish continues to be the language of social mobility and elite status (Murillo 1988; del Valle 2000). According to Ethnologue (Lewis, Simons, and Fennig 2013) estimates, there were 3,170,000 speakers of Galician in 1986, and 3,185,000 speakers worldwide. According to 2001 (census) estimates, there were roughly 2.2 million speakers of Galician, 1.4 million of whom reported always speaking Galician, and 780,000 of whom reported speaking Galician occasionally. According to the Instituto Galego de Estatística (IGE, Galician Statistical Institute: www.ige.eu), the population of Galicia in 2009 was 2,796,089. Among those who were self-reported habitual speakers of Galician 779,297 reported always speaking Galician and 687,618 reported speaking more Galician than Spanish. Due to the differences in data collection methodologies behind these figures, which are complicated by population migrations in the region, it is difficult to determine if language use is in decline. In the following section, I discuss the sociolinguistic context of Galicia in greater detail.

2 For both Murillo (1988) and del Valle (2000), only the Galician linguistic situation should be considered a case of diglossia, while Basque or Catalan are considered to be different, owing to reasons of prestige and use.

The sociolinguistic situation in Galicia   

   3

1.2 The sociolinguistic situation in Galicia According to the Ley Orgánica de Educación (Fundamental Law of Education), education within the Spanish state is free and compulsory from ages 6 to 16. The (re)establishment of Galician-language instruction in primary and secondary schools has been partly successful, as Del Valle (2000) reports that over 50% of speakers aged 16–25 speak only Spanish (17.7%), or more Spanish than Galician (35.7%). In 2007 the Galician Parliament issued a new decree (124/2007) requiring that a minimum of 50% of school instruction be conducted in Galician (LoureiroRodríguez 2009, but see also Regueira 2009 for a more in-depth history of the decree and reaction to it), a measure that was taken largely in order to address previous failures of compliance with laws in private schools and schools in urban areas (Ramallo 2007).³ According to Huguet (2004), linguistic normalization laws in the regional autonomies of Catalunya, Valencia, the Balearic Islands, Navarra, the Basque Country and Galicia are a response to a desire that students finish their years of compulsory education with a relatively balanced dominance of Spanish and the minority language in question. He argues that the result of bilingual education programs is that monolingual Basque (Euskera), Catalan or Galician speakers no longer exist, and that all native speakers of these languages also know and can use Spanish. In the subsections that follow, I will show statistical evidence suggesting that the majority of Galician speakers are also speakers of Spanish – especially among the younger segments of the population. This is of utmost importance with respect to the research detailed in this monograph, since many of the individuals who participated in the quantitative tasks that I describe in Chapter 3 reported being bilingual to greater or lesser degrees. It is also why I sought to determine participants’ native language in addition to their habitual language. In section 1.2.1, I discuss the bilingual situation in Galicia, presenting figures on self-reported native language and self-reported habitual language use. I also discuss how this situation relates to heritage-speaker phenomena such as attrition and incomplete acquisition. In section 1.2.2, I present figures on selfreported linguistic competence in Galicia, showing that those who claim to speak Galician to varying degrees report high levels of competence in both Spanish and Galician.

3 Among Castilian monolinguals and the Castilian-dominant community, this decree caused uproar and led to the creation of a protest group against the decree, ironically titled Galicia Bilingüe (Bilingual Galicia). See Regueira (2009), as noted above, for a review.

4   

   Introduction

1.2.1 Heritage-speaker bilingualism in Galicia According to 2008 data reported by the Galician Statistical Institute in Table 1, 73.52% of Galicians older than 65 years of age surveyed, and 56.3% of Galicians between the ages of 50 and 64 surveyed reported having learned only Galician as their first language. For speakers younger than 50, these figures fall below 50%. Interestingly, however, younger age groups reported higher levels of exposure to both Galician and Spanish as a child. According to 2004 data gathered by the Mapa Sociolingüístico de Galicia (MSG, Sociolinguistic Map of Galicia) study (González, Rodríguez, Fernández, Loredo & Suárez 2007: 281), over 70% of those surveyed learned or acquired Galician in the family environment. Table 1: Self-reported first language(s) percentages by age group in Galicia.

Language(s)

Galician Both Spanish Other Totals

Age group 40–49

15–29

30–39

50–64

65+

31.33 31.46 33.54 3.68

32.75 28.12 34.39 4.74

43.28 22.02 31.80 2.90

56.30 19.17 23.05 1.47

73.52 13.19 12.83 0.46

100.01

100.00

100.00

99.99

100.00

Source: Galician Statistical Institute (www.ige.eu). Santiago de Compostela, 2008.

Primary exposure to a minority language in a naturalistic setting such as the home is a typical hallmark of a heritage speaker of a language.⁴ Heritage speakers (see e.g. Montrul 2008; Rothman 2009 among numerous others) are adults who started as simultaneous bilinguals or child second language (L2) acquirers. Heritage speakers are native speakers of their first language (L1) in that they have acquired it naturalistically, but crucially, their L1 is not the dominant language of the society in which they live. The crucial factors involved in heritage language acquisition involve potential qualitative and quantitative differences in input, the influence of the societally dominant language, and difference in literacy aptitudes, potentially related to formal education in the heritage language. Typically, there is a distinct inequality between the support that the heritage language receives in the home, on the one hand, and the support

4 Rothman (2009) notes that the terms heritage speaker and heritage language are used principally in North America (see also Valdés 1995, 2000). In other areas of the world, heritage languages and speakers are referred to by terms such as minority, ethnic, or background, without much difference terminologically.

The sociolinguistic situation in Galicia   

   5

that the dominant language possesses or receives in the community at large, on the other. In comparison with native monolinguals, the combination of these factors can result in what may be interpreted as either arrested development or language attrition. Exposure to the societally-dominant L2 sets the stage for the onset of bilingualism. Sorace (2004) suggests that there is a direct association between incomplete learning and the onset of bilingualism as well as the onset of attrition. Montrul (2008) differentiates between incomplete L1 acquisition and L1 attrition as specific cases of intergenerational language loss. Montrul views L1 attrition as a phenomenon that can occur in childhood or adulthood. In attrition, a particular linguistic property that was mastered for some time with native-like proficiency is lost. For Montrul, incomplete acquisition happens in childhood when certain linguistic properties do not reach age-appropriate levels of proficiency due to intense exposure to the L2. Essentially, following childhood exposure, the L1 proficiency of the speaker fossilizes, or ceases to mature beyond a certain point as the individual matures. Although the definitions of these terms differ slightly, Montrul notes that they are not mutually exclusive concepts, as both may result simultaneously or even sequentially for any given linguistic property. There are also a variety of causes for which a certain property may not be acquired. For example, if a given property is lacking in the ambient input the child is exposed to, this can lead to incomplete acquisition. Rothman (2007) found that bilingual children who acquired Brazilian Portuguese in a purely naturalistic setting in the home but attended English-medium school did not acquire inflected infinitive forms. While one might interpret these results as suggestive that the consequence to not being exposed to schooling in a standard form of a language can lead to an incomplete grammar when compared to the grammar of an educated adult speaker, Pires & Rothman (2009) found that bilingual children who acquired European Portuguese in a similar naturalistic environment displayed evidence of having acquired inflected infinitives. The crucial difference between the groups in these studies lies in the input that the groups of children received. The monolingual vernacular that the Brazilian Portuguese-speaking children were exposed to and acquired lacked the feature of the grammar crucial to acquiring inflected infinitives. As for the children who were exposed to monolingual European Portuguese, they were exposed to such features and did acquire inflected infinitives in their speech. The results of Pires & Rothman (2009) suggest that the children who were exposed to and acquired Brazilian Portuguese in a naturalistic setting did not have incomplete grammars; rather, they fully acquired the grammar of the vernacular of the language that they were exposed to. Schmid & Köpke (2007) discuss four conditions generally considered necessary for L1 attrition to set in: 1) emigration, 2) extensive use of the L2 in daily life,

6   

   Introduction

3) extremely reduced use of the L1 in daily life, and 4) a fairly long time span (decades) during which the previous three occur. Although the first condition (1) does occasionally come into play in the Galician context, the current study does not examine Galicians outside of Galicia nor does it focus on Galicians who have spent extensive time abroad. Conditions (2), (3) and (4) also may be found in Galicia, these would be more serious factors for certain subsets of (typically) upwardly-mobile city dwellers. Social factors nothwithstanding, Galician is the majority language within Galicia. Therefore, while I grant that the presence of conditions (2), (3) and (4) provide minimal conditions for L1 attrition within Galicia, it is unclear to what extent attrition actively occurs among the general populace in Galicia. Given conditions (2) and (3) above, it is not a stretch to assume that language use plays a large role in language maintenance as well as in language loss. The Galician Statistical Institute statistics in Table 2 for selfreported habitual language use show that – even in the youngest age group – at least 74% of the population reported using Galician to some degree. Among individuals over age 65, approximately 78% reported using only Galician (always Galician or more Galician than Spanish), and roughly 66% of individuals aged 50 to 64 reported using only Galician or more Galician than Spanish. For individuals under age 50, however, usage figures for only Galician or more Galician than Spanish fall below 50%. For this age group, at least half of those surveyed reported using more Spanish than Galician or only Spanish. Table 2: Self-reported habitual language use percentages by age group in Galicia.

Habitual language 15–29

Age group 30–49 50–64

65+

always Galician more Galician than Spanish more Spanish than Galician always Spanish

18.59 24.37 31.58 25.47

22.06 26.59 26.87 24.48

34.26 32.67 17.30 15.77

52.90 25.60 10.52 10.98

Totals

100.01

100.00

100.00

100.00

Source: Galician Statistical Institute (www.ige.eu). Santiago de Compostela, 2008.

Examining the figures in Table 2 from a bilingual usage perspective, over half of those surveyed between the ages of 15 and 49 reported using both languages, and nearly half of those surveyed between the ages of 50 and 64 reported using both languages. For those over age 65, bilingual usage reported decreased to 36%. While these figures indicate a decrease in strictly monolingual Galician usage in younger generations, they also indicate an increasingly high level of bilingualism. Language use does not guarantee a high level of competence, however. In

The sociolinguistic situation in Galicia   

   7

the following section, I examine self-reported competences in the areas of oral comprehension, oral expression, written comprehension, and written expression for both Spanish and Galician.

1.2.2 Self-reported language competence ratings The following data from the 2004 Sociolinguistic Map of Galicia (henceforth MSG) in a number of competence areas show that Galicians estimate their language abilities high for both Spanish and Galician. According to the figures in Table 3, the majority of Galicians surveyed rated their Galician abilities high in areas of oral and written comprehension and expression (González et al. 2007: 171–174). In Galicia as a whole, the percentages of the same abilities in Spanish self-reported as high increased to near ceiling figures (cf. Table 3 & Table 4), with notable increases in the areas of oral expression, written comprehension and written expression (González et al. 2007: 225–228). Table 3: Percentages of Galicians surveyed who rated their Galician abilities high.

self-rating (4=highest)

oral comprehension

Galician abilities oral written expression comprehension

written expression

4

76.2

59.5

59.3

49.3

3

16.8

23.0

22.8

25.1

TOTAL

93.0

82.5

82.1

74.4

Source: González et al. 2007. Mapa Sociolingüístico de Galicia 2004, vol. 1: Lingua inicial e competencia lingüística en Galicia. A Coruña: Real Academia Galega.

Table 4: Percentages of Galicians surveyed who rated their Spanish abilities high.

self-rating (4=highest)

oral comprehension

Spanish abilities oral written expression comprehension

written expression

4 3

90.9 7.7

84.4 12.1

86.5 10.4

83.8 12.3

TOTAL

98.6

96.5

96.9

96.1

Source: González et al. 2007. Mapa Sociolingüístico de Galicia 2004, vol. 1: Lingua inicial e competencia lingüística en Galicia. A Coruña: Real Academia Galega.

8   

   Introduction

For self-reported habitual Galician monolinguals, self-rated oral abilities in Galician were rated highly more frequently than they were for Spanish (Cf. Table 5 & Table 6). Written comprehension abilities were rated highly nearly on par with Spanish abilities, but written expression abilities were rated highly less frequently than they were for Spanish written expression. Table 5: Percentages of self-reported Galician-only speakers surveyed who rated their Galician abilities high.

self-reported habitual language

self-rating (4=highest)

oral comprehension

Galician abilities oral written written expression comprehension expression

only Galician

4 3

86.3 12.5

81.5 16.1

67.8 21.3

58.3 25.0

TOTAL

98.8

97.6

89.1

83.3

Source: González et al. 2007. Mapa Sociolingüístico de Galicia 2004, vol. 1: Lingua inicial e competencia lingüística en Galicia. A Coruña: Real Academia Galega.

Table 6: Percentages of self-reported Galician-only speakers surveyed who rated their Spanish abilities high. self-reported habitual language

self-rating (4=highest)

oral comprehension

Spanish abilities oral written written expression comprehension expression

only Galician

4 3

83.4 13.0

63.0 25.0

72.3 18.1

67.7 21.4

TOTAL

96.4

88.0

90.4

89.1

Source: González et al. 2007. Mapa Sociolingüístico de Galicia 2004, vol. 1: Lingua inicial e competencia lingüística en Galicia. A Coruña: Real Academia Galega.

For speakers who reported speaking mostly Galician (i.e. more Galician than Spanish), all Galician abilities were rated high as frequently as they were for Spanish except for written expression abilities, which were rated high less frequently than they were for the same abilities in Spanish (Cf. Table 7 & Table 8). According to the MSG 2004 figures in Table 9, speakers who reported speaking more Spanish than Galician on a habitual basis rated their Galician abilities high nearly as frequently as those who reported speaking mostly Galician (Cf. Table 5) and only Galician (Cf. Table 7). In Table 9, we can see that most speakers rated their Galician abilities in the two highest rating categories in more than 80% of the responses. The only ability that rated high in less than 80% of responses was

The sociolinguistic situation in Galicia   

   9

written expression. In Table 10, we can see that these same speakers rated their Spanish abilities in the two highest categories nearly 100% of the time. Table 7: Percentages of self-reported predominantly Galician speakers surveyed who rated their Galician abilities high.

self-reported habitual language

self-rating (4=highest)

oral comprehension

Galician abilities oral written written expression comprehension expression

mostly Galician

4 3

87.0 10.9

78.0 18.3

68.9 21.3

58.4 26.1

TOTAL

97.9

96.3

90.2

84.5

Source: González et al. 2007. Mapa Sociolingüístico de Galicia 2004, vol. 1: Lingua inicial e competencia lingüística en Galicia. A Coruña: Real Academia Galega.

Table 8: Percentages of self-reported predominantly Galician speakers surveyed who rated their Spanish abilities as 3 or 4 (4=highest).

self-reported habitual language

self-rating (4=highest)

mostly Galician 4 3 TOTAL

oral comprehension

Spanish abilities oral written expression comprehension

written expression

87.8 10.0

77.6 17.8

81.9 14.0

77.8 16.7

97.8

95.4

95.9

94.5

Source: González et al. 2007. Mapa Sociolingüístico de Galicia 2004, vol. 1: Lingua inicial e competencia lingüística en Galicia. A Coruña: Real Academia Galega.

Table 9: Percentages of self-reported predominantly Spanish speakers surveyed who rated their Galician abilities high.

self-reported habitual language

self-rating (4=highest)

oral comprehension

Galician abilities oral written expression comprehension

written expression

mostly Spanish

4 3

77.2 17.5

56.2 28.8

61.3 24.4

50.9 26.7

TOTAL

94.7

85.0

85.7

77.6

Source: González et al. 2007. Mapa Sociolingüístico de Galicia 2004, vol. 1: Lingua inicial e competencia lingüística en Galicia. A Coruña: Real Academia Galega.

10   

   Introduction

Table 10: Percentages of self-reported predominantly Spanish speakers surveyed who rated their Spanish abilities high.

self-reported habitual language

self-rating (4=highest)

oral comprehension

Spanish abilities oral written expression comprehension

written expression

mostly Spanish

4 3

94.3 5.4

92.5 6.9

92.5 7.0

90.3 8.6

TOTAL

99.7

98.4

99.5

98.9

Source: González et al. 2007. Mapa Sociolingüístico de Galicia 2004, vol. 1: Lingua inicial e competencia lingüística en Galicia. A Coruña: Real Academia Galega.

It is a strong possibility that the lower Galician written expression ratings overall are related to the existence of multiple competing and revised orthographical norms for standard(ized) Galician in the post-Franco era, which in turn may have resulted in lower literacy levels in general in comparison with Spanish.⁵ It is worth noting, however, that in these self-reported ratings the majority of those who report using Galician to some degree rate their oral and written abilities as high. Assuming the MSG and Galician Statistical Institute figures to be a reliable estimate of these individuals’ competence in both Spanish and Galician, it would appear that Huguet’s (2004) description of language competences in (officially) multilingual Spanish regions like Galicia then is at least mostly accurate, thus providing further evidence suggesting that the majority of native speakers of Galician also know and use Spanish.⁶ This is of particular importance for the data reported on in the current monograph because, despite anecdotal reports that older, less literate speakers are the nearest approximation to monolingual speakers in Galicia (e.g. Siguán 1992), the majority of the Galician population is bilingual. As we saw in Table 2 above, with respect to usage, speakers who are predominantly monolingual speakers in any sense are also over the age of 65. Given that the Spanish Civil War ended in 1939, only those well over the age of 75 at the time this monograph goes to press could have possibly lived with little to no formal education in Spanish. Therefore, although it is important to bear in mind that a description of Galician spoken by Spanish-Galician bilinguals is not necessarily representative of Galician as it is/was known by those who are/were

5 The original norm was introduced in 1982 and made law in Galicia in 1983. The last reform to this norm took place in 2003. The competing orthographical norm is advocated by the AGAL (Associaçom Galega da Língua, or Galician Association of the Language), a group that seeks to re-align or re-integrate Galician with Portuguese. 6 Note however that the 2004 MSG (González 2007) only reported on participants aged 15 to 54.

Interface instability in bilinguals   

   11

monolingual Galician speakers, it is likely a more accurate representative of the Galician spoken by the majority of speakers in the present day  – and possibly even more representative of the Galician that will be spoken in the foreseeable future. Therefore, a crucial assumption that I will make without further discussion in this monograph is that the competencies of Spanish-Galician bilingual speakers are representative of Galician speakers in general. As previously mentioned, it is also debatable whether bilingual speakers of Galician fit the typical description of heritage speakers, as examined in section 1.2.1. Details of classification notwithstanding, Galicians are most certainly bilingual individuals who deal daily with questions of language maintenance and use, exposure, societal pressures, and literacy. As such, speakers of Galician may also be susceptible to incomplete acquisition, language attrition, and even language loss phenomena, which have been found to manifest themselves via interface instability in bilinguals. In the following section, I briefly examine interface instability as it pertains to bilinguals and the relevance that it may have for the linguistic interface that I examine.

1.3 Interface instability in bilinguals Research in child bilingualism (Genesee 1989; Meisel 1989; Genesee & Paradis 1996; Grosjean 1998; among numerous others) has posited that child bilinguals possess separate grammars from a very early stage, thus departing from the predominant view that bilinguals are the sum of two monolinguals. Therefore, acquiring two (or more) grammars differs from acquiring each language separately and sequentially (i.e. as a monolingual). Hulk & Müller (2000) argue that cross-linguistic interference, or interface vulnerability, can occur in bilinguals if an interface is involved and the languages in question overlap at the surface level. They argue that a second condition for interface vulnerability is satisfied if the input of one of the two languages incorrectly reinforces a seemingly possible structural (mis)analysis in the other first language (L1). Such language-internal factors are what cause cross-linguistic interference; not external factors such as language dominance. The Interface Hypothesis in its earlier incarnations is based on the assumption that, in accordance with Full-Transfer/Full-Access approaches to Second Language Acquisition (e.g. Schwartz & Sprouse 1996), narrow syntactic features instantiated in the lexicon of the second language (L2) are acquired without issue by the L2 learner (Sorace & Serratrice 2009; Belletti, Bennati & Sorace 2007; Sorace & Filiaci 2006; Sorace 2005), and that they are impermeable to cross-linguistic interference (Tsimpli 2007). Grammatical structures that involve an interface between syntax and other linguistic modules (e.g. seman-

12   

   Introduction

tics, phonology, discourse-pragmatics, and/or information structure), however, are vulnerable and may be subject to instability and residual optionality (Sorace 2005). Numerous psycholinguistic factors may affect the acquirability or stability of interface structures, including: 1) underspecification of features related to interfaces, which gives rise to optionality in mapping of such features from one module to another when one language instantiates a more complex setting than the other language; 2) processing limitations on the mapping and coordination of syntactic and contextual non-syntactic information; and 3) the quantity and quality of input received by bilinguals, which in turn may affect speed and accuracy of processing. It has been suggested that not all interfaces are equal. While some syntaxsemantics violations incur violations of ungrammaticality, and thus lead to more categorical intuitions among speakers, the syntax-pragmatics interface involves pragmatic conditions that determine appropriateness in context, thus often leading to gradient judgments regarding appropriateness. White (2008) calls the former ‘internal interfaces’, which are thought to be more readily acquired in an L2, and the latter ‘external interfaces’, which are problematic (though not inevitably so) even at very advanced stages of second language acquisition (e.g. Belletti, Bennati & Sorace 2007; Iverson, Kempchinsky & Rothman 2008; Rothman 2007a; Sorace & Filiaci 2006, Sorace 2000). It is worth noting however, that such categorization is not uncontroversial, as White (2010) has suggested that internal and external interfaces are subject to vulnerability in SLA. Sorace (2011, 2012) suggests disposing with the internal vs. external dichotomy entirely and taking what may be best described as a holistic approach to the interfaces, examining linguistic as well as processing (in)efficiencies in comparable bi-directional groups of bilinguals (e.g. Spanish L1/English L2 vs. English L1/Spanish L2). It is important to highlight that interface vulnerability is also found in cases of L1 attrition (e.g. Tsimpli et al. 2004) as well as in heritage speakers (e. g. Montrul 2002, 2004; Rothman 2007b). It has also been cited as a factor in developmental delays and general instability in language development in L1 acquisition (e.g. Schmitt & Miller 2007) and in child bilingual acquisition (e.g. Serratrice, Sorace & Paoli 2004). Many of these groups have been found to exhibit instability and optionality in their use and judgments of morpho-syntactic constructions whose distribution is governed by the syntax-discourse interface. There are notable exceptions, however, such as Slabakova, Rothman, and Kempchinsky (2011) and Slabakova, Kempchinsky, and Rothman (2012), which found that advanced and near-native L2 learners of Spanish demonstrated successful native-like acquisition of the syntax-pragmatics of clitic left-dislocation (CLLD) and clitic right-dislocation (CLRD) in Spanish. Montrul (2009) argues that a problem with the Interface Hypothesis is that it does not precisely define or predict which interfaces

Syntactic framework – assumptions   

   13

are complex or difficult to acquire. In fact, complexity is a nebulous term that is notoriously difficult to define or measure. In the following section, I introduce some of the theoretical syntactic assumptions underlying the empirical study in this monograph, as well as the particular interface of interest: the syntax-information structure interface.

1.4 Syntactic framework – assumptions Contemporary generative syntactic theory seeks to capture the largely unconscious internal knowledge that a speaker has of her/his language(s), or I-language. In turn, syntacticians propose a theoretical model that captures the abstract structure of this mental grammar. The challenge for Universal Grammar (UG) as a theory has always been to determine what it is that human languages share beyond an inherent propensity for language. According to current work in the Minimalist Program (in particular, Chomsky 1995, 2000, 2001, 2008), crosslinguistic differences result from differing features present (or absent) in the linguistic input that humans receive while acquiring language(s). Features are proposed to derive from a universal set of features, and are instantiated in the lexicon of a particular language. These features and the checking of these features are posited to determine the combinatory possibilities and impossibilities in a particular language. We may conceive the process underlying the creation of an utterance at the conceptual level as starting with the computational system of human language, or the core syntactic language module, which is proposed to select lexical items from the lexical array (i.e. the lexicon) for a given numeration, or sentence creation act. This process, from this point, to the point of spell-out, or enunciation, is referred to as the derivation. The derivation is a process of successive assembly, or merge, of lexical items, which continues until the numeration is exhausted. When lexical or functional items are selected from a given numeration for assembly, they are said to be externally merged in that in this moment they move into the abstract mental work space from without. When lexical or functional items are selected from within the work space for the purposes of feature-checking operations, they are internally merged, an operation that is at work in what is referred to as syntactic movement operations in previous iterations of formal generative theory. If all syntactic features are properly checked once the numeration is exhausted, the resulting syntactic object (i.e. clause) will converge; if not, it will crash, and thus be incomprehensible gobbledygook. Convergent syntactic objects will subsequently move on to interface with the ConceptualIntentional (semantic) module and the Articulatory-Perceptual module (physical realization) prior to Spell Out.

14   

   Introduction

Although the primary orientation of this monograph is situated within current generative theory, the treatment of clausal word order  – in particular that of subject positions – requires reference to notions that were of particular relevance to earlier generative approaches, such as Principles and Parameters Theory, or Government and Binding Theory (e.g. Chomsky 1981, 1982). One particular notion that I examine herein is that of A-positions and Ā-positions. Although I will discuss these notions in greater detail in Chapter 2, I would like to provide a brief introduction to what is of chief relevance to the current monograph. If we consider a language like Latin, for example, a language with flexible word order somewhat similar to Spanish and Galician, we find that a variety of word orders are possible (1 a-f) and that none of these affect the basic grammatical function of the subject, verb, or object. (1)

a.

Petrus amat Peter.NOM love.PRS.3SG ‘Peter loves Mary.’

Mariam. Maria.ACC

(SVO)

b. Mariam amat Petrus.

(OVS)

c.

Mariam Petrus amat.

(OSV)

d. Petrus Mariam amat.

(SOV)

e.

Amat Petrus Mariam.

(VSO)

f.

Amat Mariam Petrus.

(VOS)

Although the word orders obviously differ in the above example, it has been a matter of debate 1) what the base position of the clausal constituents are, in particular subjects, and 2) whether those constituents continue to appear in A-positions when word order varies. Even though the dominant word order in Latin was SOV (Posner 1996: 36) the above sentences represent convergent syntactic derivations generated by the numeration for Latin. However, it is not the case that every word order generated above would be an appropriate utterance in any given discourse context in Latin, which may choose to highlight one or more constituent in a variety of ways. However, it is unclear how discourse should be properly handled in the syntactic component. Assuming a basic Minimalist-era generative approach (Chomsky 1995, et seq.), the syntax alone can generate a number of orders via checking of optionally strong or weak features. If we consider a basic ditransitive syntactic structure (2), we see that the direct and indirect object XPs are generated low in the tree, in their thematic positions, where they are assigned thematic roles and perhaps check Case. From initial merge of V and DP,

Syntactic framework – assumptions   

   15

we construct the syntactic representation moving leftward and upward. Assuming the advances of the predicate-internal subject hypothesis (e.g. Kuroda 1988, Koopman & Sportiche 1991), subject DPs are generated in Spec, vP, attracted by a causative or agentive predicate that has moved from Vo to vo. (2)

TP T’ ?

T[EPP?, NOM?]  vP DPSUBJ

v’ v

VP XPIO

V’ V

XPDO

From this underlying SVO configuration, which is often assumed following Kayne (1994), the syntax may generate a variety of word orders via movement of constituents to other hierarchical positions. Movement is not assumed to be free, but rather very costly, and is to be avoided, if possible due to reasons of economy. Movement is assumed to be motivated by reasons of lexical featurechecking. Therefore, when we acquire words into our lexicon, we also learn something about their combinatorial properties. If we posit that verbs move further to T (generally based on adverb evidence), as in Irish or Arabic, for example, we may generate VSO word order. In English, it is generally assumed that SVO order is generated owing to a strong EPP feature on T and a need to check Nominative case, an operation that requires movement (or internal merge) of the subject DP to Spec, TP. However, if Nominative case checking is a universal requirement of subjects, as we generally assume, we have to ask how Case is checked on subjects in Irish or Arabic, whose subject DPs are never supposed to move to Spec, TP. Well, given that this appears to constitute a parametric difference between languages like English on the one hand and languages like Irish on the other, we may appeal to the notion that the EPP feature that must be checked is either strong (requiring movement) or weak (not requiring movement). However, the question remains what to do with languages like Spanish that allow a variety of word orders. If the EPP feature is a parametric choice, we don’t expect ‘both’ or ‘either’ to be an option available for the

16   

   Introduction

grammar. Additionally, the choice of word order in Spanish does not depend on purely lexical features related to either the subject or the predicate, which presents a problem for a claim suggesting that EPP features may be optionally strong or weak. Rather, word order appears to depend on the structure of the discourse at hand, or information structure. A crucial assumption of this monograph, following López (2009), is that the Information Structure is a module of the grammar that interfaces with the narrow syntactic module, an interaction which in turn determines which convergent outputs generated by the syntax will be appropriate in a given context in a given language. One of the primary goals of this monograph then is to reconcile Information Structure notions and their associated mechanisms with otherwise convergent syntactic objects. The earliest studies in generative grammar centered on the analysis of majority European languages such as English, French and German. Over the years, these studies have since expanded to include languages all over the world, thus continuously testing and advancing the theory. Minority languages have proven to be an extremely fruitful realm for syntactic research in this respect. For example, research on the so-called Italian dialetti, or dialects, within the Cartographic Program (e.g. Benincà 1983; Poletto 1993, 2000; Rizzi 1997 among numerous others) has been invaluable for the testing, as well as revising, of hypotheses relating to e.g. clitics. The myriad Italian varieties possess crucial differences from both standard Italian as well as from the “majority” Romance languages. It is in this spirit that I examine Galician in the current monograph. Galician is an ideal language to examine due to its linguistic proximity to both Spanish and Portuguese, and because it has not been extensively studied within the generative linguistic paradigm. Galician and Portuguese share a direct linguistic ancestor, and as such, share syntactic similarities such as enclisis, inflected infinitives, and a maintained (although restricted) use of the future subjunctive. Despite such similarities, Galician is a minority Romance language within the Spanish and Portuguese states; however, it is a majority language within Galicia. Not surprisingly, the sociolinguistic situation of Galicia plays an important role with respect to the sociolinguistic status of Galician, which in turn affects its use and maintenance. On the one hand, this monograph seeks to establish clausal word order preferences to inform the syntactic analysis of Galician. On the other hand, however, given the remarkable similarity of Galician to both Spanish and European Portuguese, it also seeks to provide insight to the syntactic debate introduced above for Spanish and European Portuguese. Therefore, the research questions guiding this monograph are the following:

Syntactic framework – assumptions   

   17

What is the basic word order in Galician?⁷ What is the preferred clausal structure in Galician for sentences in which grammatical subject or objects hold a privileged status (e.g. focus v. topic, theme v. rheme) within a larger discourse context? 3. What are the implications of the answers to questions 1 and 2 with respect to the syntactic analysis of Galician? 4. When clausal word order varies, do subjects in Galician display the hallmarks of appearing in A-positions, or Ā-positions? 1. 2.

I will further refine these questions as well as the notions of basic word order and privileged status in Chapter 2 as I examine the issues at stake with respect to the syntax-information structure interface as they relate to the investigation of Galician undertaken in this monograph. I will also establish the information structure assumptions that inform the empirical methodology that drives the analysis of Galician syntax that I propose.

7 Although Kayne’s (1994) antisymmetry proposes that all languages are considered to be underlyingly SVO, the fact remains that there are multiple ‘basic’ word orders attested cross-linguistically.

2 The interaction between syntax and information structure In chapter 1.4 I suggested that ‘free’ or ‘flexible’ word order is not quite what it seems to be and is dependent on a discourse context, which may deem it to be more or less appropriate. In the first part of this chapter, I discuss the syntactic analysis of different syntactic positions and the debate regarding its relevance for null-subject languages, in particular Spanish, European Portuguese, and Galician. As I will show, preverbal subjects in Galician present a number of complications for the A versus Ā debate, thus motivating the more in-depth treatment that I provide, taking into account the role of information structure. In the second part of the chapter, I review the literature on information structure, establishing the discourse contexts that are relevant to this monograph.

2.1 Argument vs. Non-argument positions in syntax A long-standing dichotomy within generative grammar involves that of arguments and non-arguments. Within the Government and Binding (GB) framework (e.g. Chomsky 1981, 1982) argument positions, or A-positions in Chomsky’s (1981: 47, 138 fn. 12) sense, are those in which an argument may appear and are potential θ-positions. Non-argument, or Ā-positions, are those in which an adjunct appears, especially wh- elements and topicalized constituents, which are assumed to move from their thematic position within the verb phrase and adjoin to C.⁸ In the simplest of terms, the subject and object arguments appear in A-positions, and thus are A-elements. These positions were of particular importance for A-binding (i.e. binding) relations, i.e. Principles A, B, and C of the Binding Theory. Prior to the advent of the predicate-internal subject hypothesis (e.g. Kuroda 1988, Koopman & Sportiche 1991), it was supposed that the canonical, base-generated position for subjects of transitive verbs was Spec, IP (later Spec, TP or Spec, AgrSP within the

8 See Rizzi (1997: 289–295) for a discussion of the characteristics differentiating Romance-type CLLD in Italian from fronted focus and from the syntactically moved topicalization structures found in English. Briefly summarizing, Romance CLLD involves a resumptive clitic for argumental XPs as opposed to a null anaphoric operator and allows for an “indefinite number of topics” (Rizzi 1997: 295) as compared to English topicalization, which may target at most one constituent. Rizzi follows Cinque (1990: Ch. 2) in assuming that Romance CLLD is base-generated and not a product of movement. See also Casielles-Suárez (2003: 327–330) for a comparison of Spanish CLLD and English topicalization.

Argument vs. Non-argument positions in syntax   

   19

Split-INFL framework of Pollock 1989).⁹ Notwithstanding, it is still rather standard practice to refer to the subject as the external argument, and the object(s) as the internal argument(s), in reference to the former syntactic positions of these arguments in relation to VP. Despite advances in syntactic theory, the notion of A- and Ā-positions has also endured, despite the fact that the formerly canonical, preverbal subject position (i.e. Spec, IP) is no longer strictly related to thematic role, a fact that creates a host of potential complications for the A vs. Ā distinction. However, a number of researchers, e.g. Koopman & Sportiche (1991), had noted complications arising with respect to the analysis of Ā-elements well before the Minimalist Program. Following the advances of the feature-checking syntactic model proposed by Chomsky (1995, et seq.), the notions of A- and Ā-positions have become more “intuitive” than well-defined, and have been re-cast in terms of “L-related” and “non-L-related” positions. By current Minimalist assumptions (e.g. Chomsky 1995, et seq. among others), the syntactic derivation selects and assembles two lexical items from the lexical array for first merge, thus forming a syntactic object. The relevant L(exical)-features, which motivate the merging of syntactic objects, are checked within the checking domain of a head with a matching L-feature. Lexical items are continuously merged (internally or externally) until the lexical array is exhausted, and all relevant morphological, semantic, and (perhaps) pragmatic features have been appropriately checked. The resulting C-I (conceptual-intentional) elements are sent, phase-by-phase, to the phonological component (i.e. the Articulatory-Perceptual system, or Phonological Form (PF) in earlier generative theory) for spell-out. Within this program, the external argument (i.e. the grammatical subject) merges in its vP-internal thematic position (Spec, vP), and then later checks features with T such as Tense, Case, and/or EPP-type features, an Agree operation that may involve syntactic movement to Spec, TP depending on the strength of the features in T. According to Chomsky (1995), narrowly L-related (non-adjoined specifier) positions have the basic properties of A-positions in previous frameworks.¹⁰ Therefore, a structural position that is not L-related (i.e. an adjunct) has the basic properties of Ā-positions. Within the later phases framework, Chomsky (2008: 150) defines an Ā-position as “one that is attracted by an edge-feature of a phase head; hence

9 Chomsky (1981: 47) notes that the subject position was not necessarily a θ-position. Rather this was dependent “on properties of the associated VP”, which accounts for e.g. unaccusative predicates, which are proposed to select a subject in the complement position of V. 10 Chomsky (1995: fn. 33) notes, however, that “if C has L-features, Spec, CP is L-related and would thus have the properties of an A-position, not an Ā-position.”

20   

   The interaction between syntax and information structure

typically in Spec, CP or outer Spec, v*P. Others are A-positions”.¹¹ The resulting issue then is the following: if Spec, vP is where subjects of transitive predicates are base-generated, it should be an A-position. However, then the question arises as to whether Spec, TP, the syntactic position in which preverbal subjects are supposed to appear, is an A-position or an Ā-position. Recent syntactic research has proposed that preverbal subjects in a wide variety of languages such as Greek (e.g. Alexiadou & Anagnostopoulou 1998), Spanish (e.g. Uribe-Etxebarria 1990, 1995; Ordóñez & Treviño 1999), European Portuguese (e.g. Barbosa 1996, 2000), Standard Arabic (Soltan 2007), and Kinande (Schneider-Zioga 2007) do not behave as A-elements, but rather as base-generated Ā-elements, thus creating doubt as to the nature of the syntactic position in which they appear. Such analyses are of theoretical relevance because they call into question the supposed universality of the EPP, or at least the mechanisms by which the EPP may be checked.¹² A number of other analyses, however, have disputed the non-argument hypothesis for preverbal subjects in Spanish (e.g. Goodall 2002, Suñer 2003) and European Portuguese (e.g. Costa 2004, Duarte 1997), presenting evidence suggesting that preverbal subjects are canonical subject arguments that behave as preverbal subjects do in numerous other languages.¹³ With respect to syntactic position, proponents of the preverbal-subjects-as-argument (A-element) hypothesis generally posit that canonical preverbal subjects appear in Spec, TP (3a); proponents of the preverbal-subjectsas-non-argument (Ā-element) hypothesis (e.g. Ordóñez and Treviño 1999) generally suggest that these elements appear in a structurally higher, left-peripheral – perhaps clitic left-dislocated (henceforth CLLD) – syntactic position such as Spec, TopicP (3b).

11 Note that [EPP] is also frequently referred to as an edge feature. Gallego (2005) has proposed that T is a locus of syntactic variation and that T is a phase head in Iberian Romance, thus suggesting that Spec, TP is an Ā-position. However, Kempchinsky (p.c.) notes that “it is not immediately clear that simply by virtue of its being a phase head, the Spec of T would then be an Ā position”. 12 For example, Alexiadou & Anagnostopoulou (1998) suggest that the EPP may be checked by V to T head movement in languages like Greek. 13 Note that, for Spanish, it has been argued that preverbal locatives in locative inversion structures (i) also appear in Spec, TP (e.g. Zubizarreta 1998, Ortega-Santos 2005, but see Kempchinsky 2002 for an alternative account). On this sort of analysis Spec, TP is not a syntactic position exclusively dedicated to preverbal subjects. (i)

Aquí ponemos  unas  mesas de bienvenida. here put.PRS.1PL  some  tables of welcome ‘We are putting some conference registration tables here.’

Syntactic tests for arguments and non-arguments   

(3)

a.

   21

preverbal subject as A-element  b. preverbal subject as Ā-element TP TopP DP SUBJ   T

T’

DP SUBJ  vP

  Top

Top’  TP

It is important to note that when proponents of the argument hypothesis refer to preverbal subjects as A-elements, this is a sort of shorthand for proposing that preverbal subjects appear in the canonical subject position (i.e. Spec, TP). Similarly, when those who advocate the non-argument hypothesis refer to preverbal subjects as Ā-elements, they make reference to either 1) a structurally higher Ā specifier position in the left periphery, such as TopP, or 2) a Spec, TP position that essentially behaves like an Ā-position (see fn. 11 above). The syntactic debate briefly introduced above is the point of departure for the current monograph. As noted by William Davies (p.c.), this is hardly a trivial problem; in fact, it is one that should give us pause. If it is the case that preverbal subjects in some languages don’t behave like those in other languages, it is imperative that we investigate the nature of these differences and discover the implications that they entail for hypotheses related to human language. Prior to a review of the arguments for and against the argument versus non-argument debate for preverbal subjects, I first examine some of the syntactic tests used in the literature for determining argument or non-argument status.

2.2 Syntactic tests for arguments and non-arguments Over the years, a number of syntactic tests have been formulated to determine the status of constituents within clauses. Phrases in A-positions have been found to be able to bind a strong reflexive pronoun (4a cf. 4b), they show evidence of reconstruction (4c from Sportiche 1999: 34, ex. 48a), they are hampered by intervening A-antecedents (4d), they do not display minimality effects when a constituent is extracted over them (4e), and they cannot license parasitic gaps (4f).¹⁴

14 Example (4f) comes from Haegeman (1994: 475, ex. 88).

22   

(4)

   The interaction between syntax and information structure

a.

Johni looked at himselfi in the mirror.

b. *Johni, Bill looked at himselfi in the mirror. c.

[Pictures of each other] seemed to the boys [t] to be fuzzy.]

d. *Johni thinks that Bill hurt himselfi. e.

[The magazine], John left [t] on the counter.

f.

*Poirot is a man [CP [who] [IP [t] runs away [when [you see [t] ]]]].

Phrases in Ā-positions are specifiers of an XP in which the specifier shares Ā/ Operator features with the X° head. They can license a parasitic gap (5a), are not hampered by an intervening A-antecedent (5b), and may reconstruct when Ā-moved (5c, d with fronted focus XPs cf. 5c’, d’ with topical XPs).¹⁵ (5)

a.

[Which book] did you leave [t] on the counter without reading [t]?

b. [Who]i did John say [t] saw himselfi? c.

[PER QUESTA RAGIONE]  ha   detto  che  se for this reason  have.PRS.3SG  say.PTCP  that  CL.SE ne  andrà [t]. CL.PART  leave.FUT.3SG ‘FOR THIS REASON he said that he will leave (not for that reason).’

c’. *[Per questa  ragione] ha detto  che  se  for this  reason  have.PRS.3SG say.PTCP  that  CL.SE  ne  andrà [t].  CL.PART  leave.FUT.3SG  ‘For this reason, he said that he will leave.’  Context: We consider Anna (to be) stupid. d. [GIANNI]  riteniamo [t] essere  stupido. Gianni  consider.PRS.1PL be.INF  stupid ‘GIANNI we consider stupid. (non Anna)’ d’. *?Gianni,  invece,  riteniamo essere  intelligente.  Gianni  on the contrary  consider.PRS.1PL be.INF  intelligent  ‘Gianni, on the contrary, we consider intelligent.’

15 Examples (4c–4d’) are taken from Cinque (1990: 64–67, ex. 13, 14, 25, 26).

Syntactic tests for arguments and non-arguments   

   23

According to Cinque (1990), the reconstruction in (5c) shows evidence that fronted focus elements are moved successive cyclically, while (5c’), which does not reconstruct, shows that topical XPs are CLLD elements (accompanied by the partitive clitic double) and thus base-generated. He also formulates the prediction that if CLLD moves in the same manner as wh- movement, it should be possible to dislocate the subject of the ECM predicate in (5d’), contrary to fact. This stands in contrast with a fronted-focus constituent (5d), which behaves like a moved element. These movement restrictions distinguish Ā-moved elements such as wh- elements and fronted focus elements, which involve an operator that can Ā-bind its trace, from base-generated CLLD elements, which cannot. One of the well-noted differences between CLLD elements and other (apparently) fronted elements in Romance is the requirement of a resumptive clitic double for dislocated object DPs (6a, cf. 6b).¹⁶ (6)

a.

Gianni, *(lo) vedrò Gianni CL.DAT.3SG.M see.FUT.1SG ‘Gianni, I will see tomorrow.’

domani. tomorrow

b. GIANNI (*lo) vedrò Gianni CL.DAT.3SG.M see.FUT.1SG ‘GIANNI I will see tomorrow.’

domani. tomorrow

Following Cinque’s argumentation, the CLLD construction in (6a) does not involve wh-type movement, and the obligatory resumptive clitic, or clitic double, is not a variable.¹⁷ Sentence (6b), however, involves movement akin to wh-movement as well as the creation of a variable in the gap created by movement. By Cinque’s analysis, Ā-moved elements have an operator in CP, while base-generated CLLD Ā-elements do not. Although the nature of CLLD constituents has been the subject of considerable debate in the literature with respect to CLLD constituents (see e.g. Cecchetto & Chierchia 1999, López 2009, Suñer 2006 for more on this debate), I

16 Note that in (6a) a comma follows the topicalized XP Gianni, as in his examples and in many others in the literature, but is not clear that presence of a comma necessarily implies a prosodic pause. Rizzi (1997) refers to CLLD topics as involving “comma intonation” in Italian. For Emonds (2004) comma intonation implies a prosodic pause or the potential for one; however, CasiellesSuárez (2003, 2004) notes that Spanish CLLD does not involve a prosodic pause. 17 Arregi (2003) proposes that clitics in CLLD are individual variables that are Ā-bound by the CLLD XP. He claims that both movement-based and base-generated accounts are compatible with such a proposal. I leave this issue for ongoing research.

24   

   The interaction between syntax and information structure

discuss it here briefly since Cinque draws a distinction among Ā-elements along these lines.¹⁸

2.3 The analysis of subjects as non-arguments in Spanish Uribe-Etxebarria (1990, 1995) sparked debate on the status of subject positions in Spanish by building upon Jaeggli’s (1985) observation of Superiority Effects in Spanish. Based on scope asymmetries between quantified DPs in preverbal versus postverbal positions, she proposes that preverbal subjects in Spanish are Ā-elements. In (7a, b), for example, a subordinate clause preverbal quantified DPs cannot take scope into the higher clause, over the wh- element. (7)

a.

[A quién] dices que cada senador amaba [t] ? to who say.PRS.2SG that each senator love.IMPFV.3SG ‘Who do you say that each senator loved?’ (*∀ > wh, wh > ∀)

b. [Qué] dices que todo dios ha comprado [t] ? what say.PRS.2SG that all God have.PRS.3SG buy.PTCP ‘What do you say that everybody bought?’ (*∀ > wh, wh > ∀) Therefore, the only possible responses to these questions are single answers which apply to all involved in the reply, as in (8).¹⁹ (8)

a.

It is John whom each senator loved.

b. It is this model computer that everybody bought. However, subordinate clause postverbal quantified DPs (9) can take scope over the matrix wh- phrase, thus allowing for ambiguity of scope interpretation. (9)

a.

[A quién] dices que amaba cada senador [t]? to who say.PRS.2SG that love.IMPFV.3SG each senator ‘Who do you say that each senator loved?’ (∀ > wh, wh > ∀)

18 For the moment, I will remain agnostic on CLLD items being Ā-moved or base-generated. A common assumption among treatments of CLLD is that resumptive clitics appear preverbally due to A-movement and, at the very least, A-bind a trace in thematic position. I return to CLLD elements in chapter 5 and their importance for the status of subject constituents. 19 Note that this possibility is not the result of coincidence in individual (pair-list) replies.

The analysis of subjects as non-arguments in Spanish    

   25

b. [Qué] dices que ha comprado todo dios [t] ? what say.PRS.2SG that have.PRS.3SG buy.PTCP all God ‘What do you say that everybody bought?’ (∀ > wh, wh > ∀) The additional scope interpretation that becomes available for a postverbal subject in the lower clause is a pair-list reading, thus admitting the replies in (10), in addition to those in (8). (10)

a.

Senator Smith loves Gary Cooper, Senator Brown loves Ava Gardner, …

b. Mary bought a book, Susan bought a computer, … Uribe-Etxebarria argues that the scope interpretation behavior of preverbal subjects in (8) is typical of Ā-elements, which may not move after S-structure, thus prohibiting further covert quantifier movement (i.e. to take scope over the previously-extracted wh- element). The preverbal subject, already in a peripheral position, may not acquire the additional scope interpretation available to postverbal subjects at LF.²⁰ The postverbal subject cada senador in (7a), however, is under no such limitation as an A-element, and may move further at LF to take scope over the wh- element. In light of the VP-internal subject hypothesis (e.g. Kuroda 1986, Fukui & Speas 1986, Koopman & Sportiche 1991), Uribe-Etxeberria speculates that this differences in scope interpretation in Spanish (as compared to e.g. English) may be due to a difference in Case-marking, which takes place in Spec, vP in Spanish, versus Spec, IP in English and other non null-subject languages. The implication of this conclusion is that preverbal subjects in Spanish appear there as a product of Ā-movement. Ordóñez & Treviño (1999, henceforth O & T) propose that preverbal subjects in Spanish are Ā-elements based on important symmetries between preverbal subject DPs and preverbal direct and indirect objects in clitic left-dislocation (CLLD) constructions, which are generally considered to be left-peripheral (i.e. Ā) elements. They suggest that these preverbal elements be uniformly analyzed due to their parallel behaviors with respect to remnant ellipsis, extraction possibilities for quantifiers and wh- elements, and preverbal quantifier scope.

20 In recent analyses such as Gallego (2005), the inability of preverbal subject quantifiers to move further (i.e. at LF) has been attributed to Criterial Freezing, which essentially makes this an Ā-position. Note, however, that Rizzi (2006) also suggests that Spec, TP is a Criterial Position. See Rizzi (2006) for more on Criterial Freezing.

26   

   The interaction between syntax and information structure

In the examples below, a preverbal subject (11a), a preverbal direct object (11b), and a preverbal indirect object (11c) may all stand as remnants of ellipsis in Spanish. (11)

a.

Él le dio unos libros a he CL.DAT.3SG give.PST.3SG some books to también [le dio unos libros also [CL.DAT.3SG give.PST.3SG some books ‘He gave some books to Pia and to Pepe, too.’

Pía Pia a to

b. Unos libros le dio a Pía some books CL.DAT.3SG give.PST.3SG to Pia cuadros también [le dio a books also [CL.DAT.3SG give.PST.3SG to ‘Some books, he gave to Pia and some pictures, too.’ c.

y and Pía] Pia]

Pepe Pepe

y unos and some Pía] Pia]

A Pía le dio unos libros y a Sara to Pia CL.DAT.3SG give.PST.3SG some books and to Sara también [le dio unos libros] also [CL.DAT.3SG give.PST.3SG some books] ‘To Pia, he gave some books and to Sara, too.’

O & T also show that these ellipsis remnants can be subordinated without issue (12). (12)

a.

Juan le dio unos libros a Pía y Juan CL.DAT.3SG give.PST.3SG some books to Pia and me parece que Pepe también CL.DAT.1SG seem.PRS.3SG that Pepe also [le dio unos libros a Pía] [CL-DAT.3SG give.PST.3SG some books to Pia] ‘Juan gave some books to Pia and it seems to me that Pepe did, too.’

b. Unos libros le dio a Pía y me some books CL.DAT.3SG give.PST.3SG to Pia and CL.DAT.1SG parece que unos cuadros también [le seem.PRS.3SG that some pictures also [CL.DAT.3SG dio a Pía] give.PST.3SG to Pia] ‘Some books, he gave to Pia and it seems to me that some books, too.’

The analysis of subjects as non-arguments in Spanish    

c.

   27

A Pía le dio unos libros y me to Pia CL.DAT.3SG give-PST.3SG some books and CL.DAT.1SG parece que a Sara también [le seem.PRS.3SG that to Sara also [CL.DAT.3SG dio unos libros] give.PST.3SG some books] ‘To Pia, he gave some books and it seems that to Sara, too.’

Notably, however, O & T do not consider cases in which a preverbal subject appears immediately adjacent to a preverbal direct object (13a, b) or indirect object (13c, d). (Note that constituents in these and subsequent examples throughout appear in bold for illustrative purposes only) (13)

a.

Juan, unos Juan some (me CL.DAT.1SG

libros le dio a Pía y books CL.DAT.3SG give.PST.3SG to Pia and parece que) unos cuadros también. seem.PRS.3SG that some pictures also

b. Unos libros Juan le dio a some books Juan CL.DAT.3SG give.PST.3SG to (me parece que) unos cuadros CL.DAT.1SG seem.PRS.3SG that some pictures c.

Pía y Pia and también. also

Juan, a Pía le dio unos libros y Juan to Pia CL.DAT.3SG give.PST.3SG some books and (me parece que) a Sara también. CL.DAT.1SG seem.PRS.3SG that to Sara also

d. A Pía Juan le to Pia Juan CL.DAT.3SG (me parece CL.DAT.1SG seem.PRS.3SG

dio unos libros y give.PST.3SG some books and que) a Sara también. that to Sara also

Examples (13b, d) show that a preverbal subject may follow a left-dislocated direct or indirect object. While the preverbal subject DPs are presumably leftdislocated in (13a, c) given their appearance to the left of the CLLD object, it is unclear that this is the case in (13b) and (13d) based on linear order alone. At bare minimum, the direct object (13b) and the indirect object (13d) appear in left-

28   

   The interaction between syntax and information structure

peripheral positions, but the status and position of the preverbal subjects in these examples remain unclear.²¹ O & T propose a uniform analysis for preverbal subjects, direct and indirect objects (14), by which a preverbal subject appears in a left-peripheral position, while pro appears in the canonical preverbal subject position. (14)

[YP DO/IO/SU [TP pro …no/también/tampoco/sí …]]

Despite the fact that the phonologically null element pro cannot be physically ‘seen’, O & T place it preverbally in (14) based on evidence in (15a) and (15b), which, following their analysis, suggests that an overt subject may not substitute preverbal pro.²² (15)

a.

*A María, los niños le dieron un libro  to Maria the children CL.DAT.3SG give.PST.3PL a book  y  a Pía, Pedro también.  and  to Pia Pedro also  ‘To Maria, the children gave her a book, and to Pia, Pedro did, too.’

b. *A María, Juan le dio un libro y  to Maria Juan CL.DAT.3SG give.PST.1SG a book and  me han dicho que a Tomás, Tito también.  CL.DAT.1SG have-PRS.3PL say.PTCP that to Tomas Tito also  ‘To Maria, Juan gave her a book and I’ve been told that to Tomas, Tito did, too.’ Their other examples in support of this argument are unclear, and (16a), in particular, in which the predicate constituent in bold is assumed to be subject to ellipsis is of dubious grammaticality.

21 O & T do not specify the syntactic location of the elided constituent in the above examples. By current Minimalist assumptions, however, the elided element should be either T or v. Given that Spanish is generally assumed to have verb movement to T, I assume that the elided constituent is TP. It is worth pointing out as well that in the following analysis (14), nothing a priori requires that pro appear preverbally. Since it is phonologically null, one cannot be entirely sure. Here, I follow O & T in placing it preverbally, but see comments below regarding [EPP] checking and expletive pro. 22 Examples (14–16) taken from Ordóñez & Treviño (1999: 43–44, exs. 19–21).

The analysis of subjects as non-arguments in Spanish    

(16)

a.

   29

A ti la policía te va a detener, pero to you the police CL.ACC.2SG go.PRS.3SG to detain.INF but Pedro el juez no la va a detener. Pedro the judge not CL.ACC.F.3SG go.PRS.3SG to detain.INF ‘You, the police are going to detain, but Peter, the judge is not going to detain her.’

b. *A ti la policía te va a detener, pero  to you the police CL.ACC.2SG go.PRS.3SG to detain.INF but  a Pedro el juez no.  to Pedro the judge not  ‘You, the police are going to detain, but Pedro, the judge is not (going to).’ The issue here lies with the referent of the direct object clitic la, which in this example may only refer to the subject DP la policía in the superordinate clause. The fact that (16b) may not be derived from (16a) is supposed to demonstrate that a remnant with a preverbal DO cannot admit a preverbal subject. Assuming the elided constituent to be TP, this is supposed to be further support for the notion that preverbal subjects appear outside of TP. Assuming that el juez is not an appositive describing Pedro in (16b), but rather that Pedro and el juez are separate referents lends additional support to this point (17a cf. 17b). (17)

a.

A ti la policía te va a detener, pero to you the police CL.ACC.2SG go.PRS.3SG to detain.INF but el juez no [te va a detener] the judge not CL.ACC.2SG go.PRS.3SG to detain.INF ‘You, the police are going to detain, but the judge is not (going to).’

b. *A ti la policía te va a detener, pero  to you the police CL.ACC.2SG go.PRS.3SG to detain.INF but  a Pedro el juez no [te va a detener]  to Pedro the judge not CL.ACC.2SG go.PRS.3SG to detain.INF  ‘You, the police are going to detain, but Peter, the judge is not (going to).’ Despite the apparent person mismatch in the second-person accusative clitic te (which is not required for ellipsis, i.e. there are issues of strict versus sloppy identity), the point that the overt preverbal subject el juez in (17b) cannot be in the same position as pro becomes much clearer, thus supporting O & T’s argument that such substitution would predict the possibility of a two-constituent remnant.

30   

   The interaction between syntax and information structure

The truth is, however, that a two-constituent remnant is impossible regardless of whether a constituent in the second conjunct is a subject (18). (18)

*A María, los niños le dieron un libro y a Pía,  to Maria the children CL.DAT.3SG give.PST.3PL a book and to Pia un cuadro también. a picture also ‘To Maria, the children gave a book and to Pia, a picture, too.’

The deeper issue in the examples above lies not with the two-constituent remnant, but with the structure and order of the remnants. In (17a), the ellipsis construction establishes a contrast between the preverbal subjects la policía and el juez and the result is grammatical. Examples (17b) and (18) also attempt to do this, with a topicalized direct object in the former and a topicalized indirect object in the latter, but are ungrammatical due to the non-parallel remnant structures. If we maintain parallel remnant structures for these peripheral constituents between the two clauses (19), however, we obtain a grammatical result for a two-constituent remnant. (19)

La policía a TI te va a detener, pero the police to YOU CL.ACC.2SG go.PRS.3SG to detain.INF but el juez a PEDRO no. the judge to PEDRO not ‘The police are going to detain YOU, but the judge isn’t (going to detain) PETER.’

Example (19) then suggests that the possibility of a two-constituent remnant is not what is at stake, but rather the order and type of elements in opposition.²³ In (19) the (fronted focus) direct objects (a TI and a PEDRO) are in opposition with one another while the dislocated subjects (la policía and el juez), which precede each, are topical. Example (17b) compared with (19) suggests an asymmetry between subjects and direct objects. In (18), however, if we follow O & T in assuming that TP is elided then we also assume the abstract structure in (20), in which the elided elements in TP appear in brackets.

23 O & T’s (1999) point essentially hinges on assuming that ellipsis operates over TP. Therefore, a remnant subject, which is necessarily preverbal in Spanish, cannot simultaneously appear in Spec, TP and as a remnant. In (19), the preverbal subject, which appears to the left of the fronted focus XP [a TI/a PEDRO], is clearly a topical constituent. I leave a detailed examination of issues related to strict versus sloppy identity in ellipsis for future research as it would take the current discussion too far afield.

The analysis of subjects as non-arguments in Spanish    

(20)

   31

dieron un libro ] y a *A María, los niños [TP le  to Maria the children CL.DAT.3SG give.PST.3PL a book and to  Pía, un cuadro también [TP le dieron un libro]  Pia a picture also CL.DAT.3SG give.PST.3PL a book  ‘To Maria, the children gave a book and to Pia, a picture, too (gave them a book).’

In this case there are two competing direct objects in the second conjunct, a clear violation of the theta criterion. In (20), since un cuadro cannot be in opposition with un libro, which forms part of the elided constituent in each conjunct, it is forced into opposition with the preverbal subject los niños, which is impossible given their non-parallel argument status within the sentence (indirect object vs. subject). Therefore, the examples in (15) do not constitute clear proof that pro and overt subjects do not have the same distribution, but rather evidence suggesting an asymmetry between subjects and indirect objects when topicalized in the presence of a fronted focus element (17b cf. 19). This unexpected asymmetry then presents a problem for the proposed uniform analysis. O & T also discuss negative quantifier displacement. As is well known, in Spanish a preverbal subject may be a negative quantifier, but a subject may not appear following a direct object (21b) or indirect object (21c) negative quantifier in preverbal position, thus interrupting the adjacency of the negative quantifier and the verb. (21)

a.

Nadie le debe la renta a María. nobody CL.DAT.3SG owe.PRS.3SG the rent to Maria ‘Nobody owes rent to Maria.’

b. Nada (*Juan) le debe (Juan) a sus amigos. nothing Juan CL.DAT.3SG owe.PRS.3SG Juan to his friends ‘Juan owes nothing to his friends. ‘ c.

A nadie (*Juan) le debe (Juan) la renta. to nobody Juan CL.DAT.3SG owe.PRS.3SG Juan the rent ‘Juan owes rent to nobody.’

The impossibility of a preverbal subject following a negative quantifier suggests that either a) preverbal subjects are Ā-elements that incur a minimality violation when a negative quantifier is fronted over them, or b) the two elements compete for the same structural position. O & T reject the “dual hypothesis” for negative quantifiers, according to which a subject negative quantifier (21a) appears in a different structural position (e.g. Spec, TP) than a direct or indirect object

32   

   The interaction between syntax and information structure

negative quantifier (21b, c, respectively). By their analysis, if a preverbal subject appears in an A-position, the ungrammaticality of (22a) is unexpected. (22)

a.

*A nadie [Juan [le debe la renta]]  to nobody [Juan [CL.DAT.3SG owe.PRS.3SG the rent]]  ‘To nobody Juan owes rent.’

b. *Nadie [a Juan [le  nobody [to Juan [CL.DAT.3SG  Nobody to Juan owes rent.’

debe la renta]] owe.PRS.3SG the rent]]

Given that the preverbal subject Juan in (22a) and the preverbal indirect object a Juan in (22b) incur the same sort of blocking effect, O & T propose that both are Ā-elements.²⁴ Similar well-known effects are found in several structures in Spanish. The fronting of a simple wh- element (23a, b), a focus fronted XP (23c, modified from Leonetti & Escandell-Vidal 2007: 156, ex. 2), or a verum focus fronted XP (23d, from Leonetti & Escandell-Vidal 2007: 160, exs. 11a, 12a) triggers a subject-verb ‘inversion’ effect. (23)

a.

Qué (*Pedro) compró (Pedro) what Pedro buy.PST.3SG Pedro ‘What did Pedro buy at the market?’

en in

el the

mercado? market

24 Note, however, that in (i) when we reverse the order of the subject negative quantifier nadie and the indirect object a Juan we do not see the same intervention/blocking effect as in (22b). (i)

A Juan nadie le to Juan nobody CL-DAT.3SG ‘To Juan, nobody owes rent.’

debe la renta. owe-PRES.3SG the rent

O & T note that no such minimality effect occurs in the case of (topical CLLD) left dislocated XPs, for example under multiple dislocation. They suggest that the preposing of a negative quantifier (or wh- element) involves some sort of special (subject-verb) agreement. If such an agreement process takes place with NegP in (i), perhaps owing to verb adjacency, the prepositioning of A Juan creates no apparent obstacle. If it is the case that agreement is responsible for the ungrammaticality of the examples in (22), and not the A/Ā-status of the constituents in question then this would appear to be an independent phenomenon. For further discussion of such issues, see Rizzi (1991), Chung & Georgopoulos (1988) on wh-elements, and Laka (1990), Haegeman & Zanuttini (1991) on negative quantifiers.

The analysis of subjects as non-arguments in Spanish    

   33

b. A quién (*Susana) le dio (Susana) to whom Susana CL.DAT.3SG give.PST.3SG Susana el paraguas? the umbrella ‘To whom did Susana give the umbrella?’ c.

EL LIBRO (*yo) he terminado (no los artículos). the book I have.PRS.1SG finish.PTCP NEG the articles ‘I have finished THE BOOK (not the papers).’

d. Nada (*yo) tengo (yo) que añadir a lo que nothing I have.PRS.1SG I that add.INF to N.SG that ya dije... already say.PST.1SG ‘I have nothing to add to what I already said…’ If a preverbal subject appears in Spec, IP/TP, this inversion effect is unexpected. If preverbal subjects and negative quantifiers appear in a left-dislocated Ā-position, they are predicted to behave like topicalized XPs, which produce the same minimality effects when interacting with wh- fronting (24a, a’), focus fronting (24b, b’), and verum focus fronting (24c, c’).²⁵, ²⁶

25 Remaining issues for O & T’s analysis have to do with complex or ‘heavy’ wh- elements, as in (i), and long wh-extraction, as in (ii), neither of which triggers inversion effects. (i)

¿Qué tipo de literatura Octavio Paz nos  what type of literature Octavio Paz CL-DAT.1PL  What kind of literature does Octavio Paz suggest to us?

sugiere? suggest.PRES.3SG

(ii) ¿Qué dijiste que tus padres te iban a  what say.PST.2SG that your parents CL-DAT.2SG go.IMPFV.3PL to  What did you say that your parents were going to give you?

regalar? give.INF

For (i), O & T speculate that complex wh- elements are not really in Spec, CP, following Ordóñez (1997) and Rizzi (1997). Given the complexity of the expanded left periphery, it is unclear where such elements might move to. For (ii), O & T conclude that inversion effects depend on the syntactic nature of the moved wh- element, yet do not expand on this conclusion. See Richards (2010: Ch. 2) for a potentially productive alternative analysis involving syntactic distinctness. 26 Examples (24a, a’, b, b’) are taken from Casielles-Suárez (2003: 334, exs. 48–51). Examples (24c, c’) are taken and inferred from Leonetti-Escandell-Vidal (2007: 174, fn. 12, ex. (i)). By their analysis, verum focus, or ‘polarity focus’ (see also Féry 2007, Höhle 1992, Krifka 2007) fronting behaves like focus fronting syntactically in that it reconstructs and may not be accompanied by a clitic double. It differs, however, in that it does not force an interpretation requiring informational partition in a given clause, thus leading to focus covering the entire clause.

34   

(24)

   The interaction between syntax and information structure

a.

[El regalo] [¿quién] lo the gift who CL.ACC.3SG.M ‘The present, who has it?’

tiene? have.PRS.3SG

a’. *[Quién] [el regalo] tiene?  who the gift have.PRS.3SG b. [A sus padres] [MENTIRAS] les cuenta siempre. to his parents LIES CL.DAT.3PL tell.PRS.3SG always ‘He always tells his parents lies.’ b’

*[MENTIRAS] [a sus padres] les cuenta siempre.  LIES to his parents CL.DAT.3PL tell.PRS.3SG always

c.

[A ella] [poco] le puede haber to her little CL.DAT.3SG can.PRS.3SG have.INF ‘To her there is little that he can have said.’

contado. tell.PTCP

c’. *[Poco] [a ella] le puede haber contado.  little to her CL.DAT.3SG can.PRS.3SG have.INF tell.PTCP Whether we assume that pro stays in Spec, vP or moves to Spec, TP (e.g. to check EPP features) is irrelevant, as it triggers no such inversion effect.²⁷ Jaeggli (1987) notes that in situ subject wh- elements in Spanish may appear in postverbal position (25b), but not in preverbal position (25a). (25)

a.

*Qué dijiste que quién compró el otro día?  what say.PST.2SG that who buy.PST.3SG the other day  ‘What did you say that who bought the other day?’

b. Qué dijiste que compró quién el what say.PST.2SG that buy.PST.3SG who the ‘What did you say that who bought the other day?’

otro día? other day

O & T demonstrate that the same restriction holds for other preverbal elements, such as an indirect object (26), in Spanish.

27 O & T do not specify a syntactic location for pro. They contend that pro and lexical subjects have a different distribution based on examples like (15b, 16b, cf. 15a, 16a), thus ruling out pro as a verbal argument. Following Taraldsen (1992), they propose rather that a verbal agreement clitic is the true subject argument (see also Kato (2000) for a similar proposal for verbal agreement in Brazilian Portuguese). See O & T for more details of their analysis. I leave the issue of expletive pro for future research.

The analysis of subjects as non-arguments in Spanish    

(26)

a.

¿Quién crees que a ti te  who think.PRS.2SG that to you CL.DAT.2SG  a dar eso?  to give.INF that  ‘Who do you think is going to give you that?’

   35

va go.PRS.3SG

b. *Quién crees que a quién le va  who think.PRS.2SG that to whom CL.DAT.2SG go.PRS.3SG  a dar eso?  to give.INF that  ‘Who do you think (that) to whom is going to give that?’ c.

Quién crees que le va a who think.PRS.2SG that CL.DAT.2SG go.PRS.3SG to dar eso a quién? give.INF that to whom ‘Who do you think is going to give that to whom?’

Example (26a) shows that a regular, non-wh- indirect object may appear preverbally (in at least some embedded clauses), but that a wh- indirect object may not (26b). Comparing (26b) and (26c), we see that the indirect object wh- a quién in (26c) may appear in situ postverbally, but it may not move to a preverbal position in (26b), on a par with a subject wh- element, as in (25). O & T further base their uniform hypothesis on observations in Uribe-Etxebarria (1990, 1995), discussed in section 2.3 above, which demonstrate differing interpretations for preverbal versus postverbal subject universal quantifiers. Therefore, the parallel behaviors of preverbal subjects, direct objects and indirect objects with respect to remnant ellipsis, extraction possibilities for quantifiers and wh- elements, and preverbal quantifier scope serve as the empirical basis upon which they state their uniform (Ā) hypothesis. While O & T were not the first to claim that preverbal subjects are Ā-elements, their systematic approach reenergized the debate on preverbal subjects in Spanish, as well as in other null subject languages. As we have seen, however, at least some of this evidence is problematic. Camacho (2006) proposes that preverbal subjects sometimes behave as leftperipheral CLLD elements and other times as canonical subjects. He sums up the different preverbal subject positions as dependent on the following: i) subjecttype (e.g. lexical v. expletive); ii) verb position (e.g. modal v. non-modal); and iii) locality constraints (e.g. negative quantifier v. expletive). He rejects the uniform hypothesis for topical CLLD constituents and preverbal subjects owing to striking asymmetries that arise between the two with epistemic modals. Perhaps the most

36   

   The interaction between syntax and information structure

striking asymmetry he discusses is found in (27, examples (10a, b) and (11a) in Camacho, respectively): (27)

Background: Full-time doctors have little time to devote to each individual patient, and sometimes, patients suffer on account of this, but in this hospital, all patients seem to be very well attended to… a.

#Sí, a los pacientes, [los médicos residentes] los  yes to the patients the doctors resident CL.ACC.3PL.M  deben atender, por eso están tan bien.  must.PRS.3PL attend.INF for that be.PRS.3SG so well

b. Sí, a los pacientes, los deben atender yes to the patients CL.ACC.3PL.M must.PRS.3PL attend.INF los residentes médicos, por eso están tan bien. the resident doctors for that be.PRS.3SG so well ‘Yes, the patients must be taken care of by resident doctors, that’s why they are so well cared for’ c.

Sí, aquí los médicos residentes deben atender yes here the doctors resident must.PRS.3SG attend.INF a los pacientes por eso están tan bien. to the patients for that be.PRS.3SG so well ‘Yes, resident doctors must take care of patients here, that’s why they are so well cared for.’

For Camacho, the fact that a CLLD direct object (a los pacientes) and preverbal subject may not appear with an epistemic modal (27a) strongly suggests that the two do not always occupy the same left-dislocated syntactic position, which is unexpected under the assumption that left-peripheral topics (of any type) are recursive elements. He proposes that the topic a los pacientes in (27b) appears in the standardly-assumed CLLD position, but that the preverbal subject in (27c) moves to Spec, ModP from Spec, IP, rendering the projection agreement-active and thus L-related, which in turn attracts the modal to Modo.²⁸ In cases not involving a modal (28, ex. (26a) in Camacho), however, the order CLLD-preverbal subject is grammatical and both topic XP (a Marta) and preverbal subject (los niños) may appear in (Ā) CLLD positions. ²⁹

28 However, note that his would create complications of mixed A- and Ā-chains, as discussed in Chomsky (2008). 29 For preverbal negative quantifiers, the NPI is proposed to move to Spec, NegP. Camacho adopts Poletto’s (2000) cartographical clausal structure (i), with Number, Hearer, and Speaker

The analysis of subjects as non-arguments in Spanish    

(28)

   37

A Marta, los niños la saludan todos los días. to Marta, the children CL.DAT.3SG.F greet.PRS.3PL every the days ‘Marta, children greet her everyday.’

In summary, Camacho proposes that the only elements merged (internally or externally) in Spec, IP are pro, expletives, and (non-L-related) subjects that follow a modal and/or CLLD. The above analyses for preverbal subjects share the assumption that postverbal subjects are merged in Spec, vP and remain in situ in VSO order. While this is a common assumption in the literature, it is not universal. Zubizarreta (1998) argues that postverbal subjects occupy a position outside (i.e. higher than) the VP. This argument is based on the position of manner and aspectual (low VP) adverbs, which may appear between subjects and objects in VSO word order.³⁰ Zubizarreta (2007) proposes that postverbal subjects in VSO appear in Spec, TP, while the inflected verb moves to a higher Extended-I, or φ projection. Therefore, preverbal subject DPs in SVO must appear in positions higher than TP. For SVO, preverbal DP subjects are proposed to be base-generated in Spec, Ext-I and linked via Agree to a syntactic verbal agreement clitic (see Rizzi 1982, O & T 1999, Kato 2000 for similar concepts), which is linked to pro in argument position via Agree. The basics of Zubizarreta’s (2007) clausal structure proposal are summarized below (29) with subject positions in bold. (29)

[φP/Ext-IP [pre-V subj./XP] [φ’ [V+T+Agr] [TP [post-V subj.] [T’ [vP [v’ …]]]]]]

Zubizarreta proposes a separation of subject-verb agreement and nominative Case checking from the EPP-feature typically considered characteristic of the

projections, presumably because these are subject clitic (SCL) positions in her analysis, which I can only assume correspond with pro. (i)

[TopP Subj DPi [NegP [NumP ti [Num infl Vi ] [HearerP ti [Hearer tj [SpeakerP tj [TP]]]]]]

Poletto suggests that SCLs in the Northern Italian Dialects serve as a sort of substitute for verbal features when the main verb cannot support a given feature. This builds on her (1993) analysis in which she suggests that verbal auxiliaries do not have their own VP, and are instead inserted in a SCL head. 30 See also Ordóñez (2007) for a similar proposal for VSO in Spanish. By his analysis, post-verbal subjects may move from Spec, vP to the specifier of a neutral subject projection (NeutP), which appears hierarchically lower than T. This movement is proposed to be motivated by the fact that, unlike Catalan, French and Italian, not all Spanish postverbal subjects necessarily receive a default information focus reading. Therefore, an additional EPP feature is available for sentences with postverbal subjects receiving neutral readings.

38   

   The interaction between syntax and information structure

Spec, TP position. Spec, φP/Spec, Ext-IP is formalized as a subject of predication position in her analysis, a position which may or may not host a “logical” preverbal subject in the clause structure. In this analysis, pro only appears in sentences with preverbal subjects, moving from Spec, vP to Spec, TP. From there, it links (via Agree) with the preverbal subject in Spec, φP/Ext-I. Presumably then preverbal subjects inherit a theta-role via this Agree relation with pro. It is noteworthy that pro is lacking in postverbal subject sentences in this analysis. In VS sentences, the postverbal subject DP is proposed to undergo the same movement steps as pro in preverbal subject sentences, and thus directly receives theta-role assignment in v. Zubizarreta’s (2007) analysis differs from Camacho (2006) in that apparent left-peripheral elements do not move to the left periphery from non-peripheral positions (e.g. Spec, vP). There is undeniable similarity to O & T (1999) in that preverbal subjects are proposed to be base-generated; however, Spec, φP differs from Spec, TopP in O & T in that Zubizarreta’s subject of predication is related to feature checking, which figures importantly in Camacho (2006) as well.³¹

2.4 Analyses of Spanish preverbal subjects as canonical arguments A number of researchers have disputed the hypothesis that preverbal subjects in Spanish are non-arguments, presenting evidence in support of the notion that preverbal subjects are canonical subject arguments as in other languages. Goodall (2001, 2002) examines the analysis of O & T (1999), challenging the suggestion that preverbal subjects move to a left-peripheral position. Goodall shows that preverbal subjects do not appear in complementary distribution with either topic or focus elements. Importantly, he also points out the fact that preverbal subjects do not have the intonational qualities of either focused or topicalized left-peripheral elements. These are strongly indicative of Ā-fronted elements having a different informational status as compared to preverbal subjects. Goodall’s analysis takes into consideration the information structure status of preverbal subjects, the (im)possibility of bare nominals, and various wh- extraction contexts. He illustrates, following Fernández Soriano (1999), that in response to a neutral, thetic question preverbal subjects are possible (30a). In contrast, sentences with other fronted elements are infelicitous (30b) in the same context. 31 Ordóñez & Treviño (1999) do not explicitly discuss feature-checking as part of the derivation leading to DPs appearing in preverbal subject position, but it is my suspicion that they assume that such a mechanism is at play.

Analyses of Spanish preverbal subjects as canonical arguments   

(30)

   39

What happened? a.

Juan me regaló un anillo en el parque. Juan CL.DAT give.PST.3SG a ring in the park ‘Juan gave me a ring in the park.’

b. #En el parque Juan me regaló un anillo.  in the park Juan CL.DAT.1SG give.PST.3SG a ring  ‘In the park, Juan gave me a ring.’ For Goodall, elements within the CP layer necessarily have a specific informational role, such as topic or focus, while elements within the inflectional TP layer do not, which casts doubt on the idea that preverbal subjects such as Juan in (30a) and topicalized PPs such as en el parque in (30b) are moved to the same preverbal syntactic position. Goodall shows that preverbal subjects behave differently from Topic or Focus elements with respect to islands for extraction. Embedded topics (31a), embedded contrastive focus (31b), and embedded wh- elements (31c) all create islands for wh- extraction from an embedded clause, while preverbal subjects do not (31d). (31)

a.

*A quién crees que el premio se  to whom think.PRS.2SG that the prize CL.DAT.3SG  lo dieron?  CL.ACC.3SG.M give.PST.3PL  ‘The prize, who do you think that they gave it to?’

b. *A quién crees que EL CARRO le  to whom think.PRS.2SG that THE CAR CL.DAT.3SG  dieron (no la moto)?  give.PST.3PL not the motorcycle  ‘Who you think they gave THE CAR (and not the motorcycle)?’ c.

*A quién quieres saber cuál premio  to whom want.PRS.2SG know.INF which prize  le dieron?  CL.DAT.3SG give.PST.3PL  ‘To whom do you want to know which prize they gave?’

40   

   The interaction between syntax and information structure

d. A quién crees que Juan le dio to whom think.PRS.2SG that Juan CL.DAT.3SG give.PST.3SG el premio? the prize ‘To whom you think that Juan gave the prize?’ The fact that preverbal subjects do not create islands for wh-movement, in conjunction with the factors above, leads Goodall to conclude that preverbal subjects in Spanish must be merged lower than both FocP and TopP, in Spec, TP, (assuming the clausal hierarchy Top > Foc > TP) and that they are most probably A-elements. Suñer (2003) also takes the position that preverbal subjects are A-elements, and not left-dislocated or Ā-elements. She examines the distribution, interpretation, and binding facts of preverbal subjects in order to challenge the proposal made by Alexiadou & Anagnostopoulou (1998), who argue in a proposal similar to that of O & T (1999) that overt preverbal subjects in Greek have Ā-properties. Suñer proposes that preverbal subjects in Spanish move (from Spec, vP) to the outer Spec of TP (32a), following the specialization hypothesis of Cardinaletti & Roberts (1991) and Cardinaletti (1997) for subject positions.³² ³³ (32)

a.

[TP {Juan/Él} [XP parenthetical [TP {ello/pro} [T [V Vfinite ... ]]]]]

b. Juan/Él, a mi parecer, es muy simpático. Juan/He to my view be.PRS.3SG very nice ‘Juan/He, in my opinion, is very nice.’ c.

En esta clase/ Aquí, a mi parecer, faltan in this class/ here to my view lack.PRS.3PL ‘In my view, chairs are lacking here/in this class.’

sillas. chairs

d. Ello (*a mi parecer) no sería malo estudiar. it to my view not be.COND.3SG bad study.INF ‘It, in my opinion, would not hurt to study.’

32 Suñer (2003) examines an older version of the specialization hypothesis. In a newer version (e.g. Cardinaletti 2004), the different T positions are labeled SubjP (which hosts subject of predication features), EPPP (which hosts EPP features), and AgrSP (which hosts Case and φ-features). 33 Examples in (32a–e) come from Suñer (2003: 351–352). They are her examples (30c), (29a), (31a), (29b), and (31b), respectively.

Analyses of Spanish preverbal subjects as canonical arguments   

e.

Me (*a mi parecer) consta me.CL.DAT.1SG to my view be evident.PRS.3SG Mara estuvo ausente. Mara be.PST.3SG absent ‘It is evident to me, in my view, that Mara was absent.’

   41

que that

By this version of the specialization hypothesis, T has multiple specifier positions: the upper Spec, TP position is reserved for overt subjects (32a), strong pronouns (32b), and locative clitics (32c, following Fernández Soriano 1999), while the lower Spec, TP position is reserved for weak subject pronouns like Dominican expletives (30d, Henríquez Ureña 1939) and dative clitics (32e, Fernández Soriano 1999).³⁴ This hypothesis, which captures asymmetries between overt subjects and null subjects, is merged with the observation in Cardinaletti and Starke (1999) that subjects may be of different strengths, which are determined via their syntactic behavior. Burga (2007, 2009) examines the Spanish preverbal subject debate from an experimental perspective examining speaker judgments of scope. A & A (1998) observed that an SVO sentence such as (33a) may only have surface scope. They report the same scope facts for CLLD elements (which are base-generated in their analysis), as in (33b). (33)

a.

Un estudiante leyó cada libro. a student read.PST.3SG each book ‘A student read each book.’ (∃ > ∀ / *∀ > ∃)

b. Un libro lo a book CL.ACC.3SG.M ‘A book each student read.’

leyó cada estudiante. read.PST.3SG each student (∃ > ∀ / *∀ > ∃)

Given the parallel scope behaviors between CLLD elements and preverbal subjects, A & A conclude that preverbal subjects are also CLLD elements. However, Suñer (2003) presents the same scope data, but with differing judgments, to the effect that SVO sentences may have either surface or inverse scope interpretations. This formed part of Suñer’s proposal that preverbal subjects in Spanish do not occupy a left-peripheral position, but rather a canonical, Spec, TP position. Given this conflict in scope judgments in the limited data available, Burga gath-

34 In Caribbean varieties of Spanish, most notably Dominican Spanish, ello is an overt expletive. For a more in-depth discussion of ello and an alternative analysis, see Toribio (2000) and Gupton & Lowman (2013).

42   

   The interaction between syntax and information structure

ered impressionistic data from three naïve Peruvian informants testing surface scope as well as inverse scope interpretations.³⁵ She found that CLLD constructions allow for inverse scope readings, while SVO sentences do not. This led her to refute the left-peripheral subject hypothesis, and conclude that preverbal subjects are not left-peripheral elements, but rather canonical preverbal subjects (appearing in Spec, TP). While the analyses presented thus far propose that preverbal subjects in Spanish are either left-peripheral or canonical subjects, the debate is further complicated by regional differences (e.g. Burga 2007, 2009; Suñer 2003; UribeEtxebarria 1995; Zubizarreta 1998; see also note 35), which may in fact prove to be a source of the variation encountered by the above analyses. As I show in section 2.4, however, a comparatively limited geographical area such as Portugal presents complications as well, as a similar debate exists among researchers of European Portuguese.

2.5 The debate on subjects in European Portuguese Although there are numerous phonological and morphological differences between European Portuguese (henceforth EP) and Spanish, the main syntactic difference as related to clausal syntax is the possibility for proclisis and enclisis in finite clauses. Despite this rather dramatic difference, a similar debate over the status of preverbal subjects also exists in EP.

2.5.1 Analysis of EP subjects as non-arguments Barbosa (2000) strongly links her left-peripheral proposal for preverbal subjects to her phonological analysis of enclisis in EP. She suggests that enclisis results from a last-resort operation at PF which repairs any syntactic output which results in a clitic starting the first Intonational Phrase of a non-embedded clause (see also Uriagereka & Raposo (2005) for a similar proposal for EP and Galician). The Last Resort prosodic filter that she proposes in (34) prevents a clitic from starting an Intonational Phrase (i.e. the left-most position of a clause lacking an

35 Note, however, that numerous researchers have commented on dialectal variation in regard to scope interpretation (notably Suñer 2003). Others have commented to me anecdotally (Ordóñez, p.c., Zubizarreta, p.c.) that the left periphery appears to be much more active for peninsular Spanish speakers than for speakers from other areas.

The debate on subjects in European Portuguese   

   43

overt complementizer), thus ruling out (35a) based on the abbreviated Intonational Phrase structure in (35b).³⁶ (34)

*[IntP cl V…] IntP = Intonational Phrase

(35)

a.

*O viu  CL.ACC.M.3SG see.PST.3SG

o the

João. João

b. *[IntP o viu] c.

*O  the

João João

o CL.ACC.M.3SG

viu. see.PST.3SG

d. O João viu-o. / Viu-o o João. the João see.PST.3SG-CL.ACC.M.3SG ‘João saw it.’ The rule in (34) requires enclisis in most cases, including those with a preverbal subject (35c). This prosodic filter then provides a minimalist syntax-PF interface account for the well-known Wackernagel/Tobler-Mussafia Law in expressions that include a subject that does not belong to the subset of quantified expressions that trigger proclisis (see below). To quickly summarize, in EP matrix declaratives, a proclitic may not appear in the clause-initial position (35a), nor can it follow a preverbal subject (35c). In fact, the clitic must appear as an enclitic (35d) regardless of the position of the subject, a fact that suggests that preverbal subjects do not syntactically or prosodically ‘count’; otherwise we should expect to find proclisis to be acceptable in (35a, c), contrary to fact. Topicalization contexts also trigger enclisis (36a) or a topic-drop null clitic (36b) in EP.³⁷ (36)

a.

Esses livros, dei-os à those books give.PST.1SG-CL.ACC.1PL.M to_the ‘Those books, I gave (them) to Maria.’

Maria. Maria.

b. Esse livro, João já leu [e]. that book João already read-PST.3SG ‘That book, João already read.’

36 Examples in (35) are taken from Barbosa (2000: 35, exs. 11, 15), some with slight modification for the sake of uniformity. 37 Example (36a) taken from Barbosa (2000: 35, ex. 16), example (36d) from Barbosa (2000: 60, ex. 111b). Remaining examples in (36) are modified or created based in part on Barbosa’s presentation.

44   

   The interaction between syntax and information structure

c.

O MEU LIVRO, o Paulo leu (não o the my book the Paulo read.PST.3SG not the ‘Paulo read MY BOOK (not yours).’

d. ESSE LIVRO dou-lhe, mas este that book give.PRS.1SG-CL.DAT.3SG but this ‘I will give you THAT BOOK, but not this one.’ e.

O João, está a estudar na the João be.PRS.3SG to study.INF in_the ‘(As for) João, he is studying in the library.’

teu). yours

não. not

bibliotêca. library

Examples like (36c) would seem to demonstrate that contrastive focus also allows for topic-drop, but by Cinque’s (1990) analysis, contrastive elements do not allow for resumptive clitics. The fact that dar (‘to give’) in (35d) is a ditransitive predicate is more enlightening however, demonstrating that preverbal contrastive focus elements also trigger enclisis in EP.³⁸ Returning to the preverbal subject example in (35d) then, if the enclitic pronoun o were to refer to the topicalized o João, the result would be a Condition B violation in EP, as the appropriate reflexive pronoun is se. In fact, in EP subject CLLD, topicalized subject DPs as in (36e) are not accompanied by a resumptive subject clitic, but rather by pro. Turning to proclisis, Barbosa (2000) demonstrates that proclisis in EP occurs in embedded clauses (37a), when a bare QP precedes the verb (37b), in the presence of nonspecific indefinite QPs (37c), in the presence of affective operators such as negative QPs (37d) or DPs modified by a focus particle (37e), when accompanying a (fronted) wh- element (37f), in the presence of negation (37g), or following a preverbal aspectual adverb (37h).³⁹,⁴⁰

38 Note however that there is debate on this issue. According to Raposo & Uriagereka (1996), a contrastively focused subject triggers proclisis in EP. (i)

a. A MARIA me to Maria CL.ACC.1SG

criticou criticize.PST.3SG

(não not

Fernanda). Fernanda

b. *A MARIA criticou-me (não Fernanda).  to Maria criticize.PST.3SG-CL.ACC.1SG not Fernanda  ‘MARIA criticized me (not Fernanda).’ 39 Examples (37a–h) are taken unmodified from Barbosa (2000: 36, ex. 18–23). 40 Note that the EP quantified phrases in examples (37b–d) likely fall under the rubric of verum focus fronting (VFF, see note 26 as well as Leonetti & Escandell-Vidal 2007 for discussion). However, since they appear to behave in a similar fashion to Spanish and Italian, I will not discuss the behavior of VFF further. I leave the investigation of EP verum focus fronting for future research.

The debate on subjects in European Portuguese   

(37)

a.

Eu duvido que ele I doubt.PRS.1SG that he ‘I doubt that he saw her.’

a CL.ACC.3SG.F

b. Ninguén / Alguém o no one / someone CL.ACC.3SG.M ‘No one / Someone saw him.’ c.

Algum aluno se some student CL.REFL.3SG ‘Some student forgot the book.’

   45

visse. see.PST.SUBJ.3SG

viu. see.PST.3SG

esqueceu forget.PST.3SG

do of_the

livro. book

d. Nenhum aluno se no student CL.REFL.3SG ‘No student forgot the book.’

esqueceu do livro. forget.PST.3SG of_the book

e.

Só o Pedro o only the Pedro CL.ACC.3SG ‘Only Pedro saw him.’

viu. see.PST.3SG

f.

Quem o viu? who CL.ACC.3SG see.PST.3SG ‘Who saw him?’

g.

O João não o the João not CL.ACC.3SG ‘João didn’t see him.’

viu. see.PST.3SG

h. O Pedro já / nunca o the Pedro already / never CL.ACC.3SG ‘Pedro already / never saw him.’

viu. see.PST.3SG

The split in cliticization directionality properties in (36) and (37) drives Barbosa’s classification of preverbal subjects. She proposes that preverbal subjects, which trigger enclisis, are not arguments, and that the only true A-position for subjects in EP is the postverbal position.⁴¹ Therefore, in a VS sentence (38a), the verb has raised to Io and the subject remains within the VP (38a’).⁴²

41 Fernández-Rubiera (2009) takes a similar position regarding subjects in Asturian. 42 I take these examples direct from Barbosa (2000: 62–63, ex. (124a), (125)). Although the verb telefonar is unergative in these examples, I understand it to be her intention for this analysis to apply to transitive predicates as well given that both assign an agent theta-role.

46   

(38)

   The interaction between syntax and information structure

a.

Telefonou a telephone.PST.3SG the ‘Maria telephoned.’

Maria. Maria

a’. [IP [I’ telefonou [VP a Maria t ]] b. a Maria telefonou. the Maria telephone.PST.3SG ‘Maria telephoned.’ b’. [[A Maria] [IP telefonou proi ]] In the SV sentence (38b), however, she proposes that the subject Maria is a basegenerated CLLD adjunct that is doubled by pro, not an A-moved argument.⁴³ This preverbal subject is licensed by Chomsky’s (1977) Rules of Predication, which preceded an articulated left periphery. She follows Raposo (1996), in assuming that CLLD elements appear outside CP and are base-generated.⁴⁴ The XP predicated of the CLLD subject contains a resumptive clitic whose reference is fixed by the topic. According to Barbosa, CLLD subjects in EP do not obey subjacency, do not exhibit Weak Crossover (WCO) effects (see also Duarte 1987, Rizzi 1997, and Raposo 1996), and do not license parasitic gaps (see also Duarte 1987, Raposo 1996). Barbosa differentiates base-generated CLLD adjuncts such as fronted topics, which may have an enclitic clitic double (e.g. 35a), from Ā-moved elements such as quantificational operators which may not have a clitic double and may not serve as discourse links (39, ex. 104a in Barbosa 2000: 58). (39)

Nadai posso fazer eci por ti. nothing be_able.PRS.1SG do.INF for you ‘I can’t do anything for you.’

Unlike quantificational operators which trigger proclisis (36), CLLD subjects and objects obligatorily trigger enclisis (35), similar to preverbal subjects (34d). The symmetry between CLLD elements noted in Barbosa’s analysis resembles O & T’s analysis for CLLD elements in Spanish, thus casting doubt on the idea that there is a dedicated preverbal position for subjects (i.e. Spec, IP/TP following A-move-

43 In Barbosa’s analysis, pro remains in a postverbal position, and does not move to Spec, IP (e.g. to check an EPP feature), yet it is worth noting that nothing in her analysis necessarily rules out this possibility. 44 I present her analysis here, which does not assume Rizzi’s (1997) split-CP hypothesis. Within an articulated CP, a CLLD element would not be considered CP-external.

The debate on subjects in European Portuguese   

   47

ment). If this were the case, one would predict that subject DPs should behave differently from other fronted DPs, contrary to the facts presented by Barbosa. Preverbal subjects in EP may be Ā-moved QP elements, which cannot be discourse topics or discourse links (Vallduví 1990, 1992). These are a) bare QPs (ninguém, alguém), b) non-specific indefinite QPs (algum), c) affective operators such as negative QPs (nenhum) and DPs modified by a focus particle (só..., até...), and d) wh- phrases. All of these expressions trigger proclisis in EP. For Barbosa, CLLD is barred in these constructions, which also prevents doubling by a resumptive clitic.

2.5.2 Analyses of EP preverbal subjects as canonical arguments Duarte’s (1997) analysis of preverbal subjects in EP is a more strictly Minimalist proposal, by which the discourse function and syntactic position of subjects is driven by formal features in the grammar. She starts out with the generalization that preverbal subjects in EP are typically unmarked topics and express given information, cautioning that one cannot attribute the same syntactic structures to sentences with preverbal subjects as those with unmarked topics. Duarte shows that bare non-referential NPs in EP may not appear in a preverbal position (40a) unless they are dislocated marked topics (40b).⁴⁵ (40)

a.

*Mulheres adoram  women adore.PRS.3SG  ‘Women adore men.’

b. Mulheres, os homens women the men ‘Women, men adore.’

homens. men

adoram. adore.PRS.3SG

The above asymmetry suggests a non-uniform analysis for these constituents (e.g. Contreras 1986; Giorgi & Longobardi 1991; Longobardi 1994). However, as introduced in the previous section, negative QPs may not appear as marked topics in EP (41a cf. 41b).⁴⁶

45 Examples (40a, b) taken from Duarte (1997: 583, exs. 5a, 6a). 46 Examples (41a, b) taken from Duarte (1997: 583, exs. 5b, 6b).

48   

(41)

   The interaction between syntax and information structure

a.

Nenhuma pessoa conhecida viu os nossos no person known see.PST.3SG the our amigos na festa. friends in_the party ‘Nobody we know saw our friends at the party.’

b. *Nenhuma pessoa conhecida, os nossos amigos (não)  no person known the our friends (not)  viram na festa.  see.PST.3SG in_the party  ‘Nobody we know, our friends didn’t see them at the party.’ Contrary to other Romance languages like Italian (e.g. Belletti 1990), speakeroriented adverbs in EP may occur between the preverbal subject and the verb without necessarily creating a marked topic interpretation for the subject.⁴⁷ (42)

a.

O João... provavelmente ele vai chegar atrasado. the João probably he go.PRS.3SG arrive.INF late ‘João... probably he is going to arrive late.’

b. O João provavelmente vai the João probably go.PRS.3SG ‘João probably is going to arrive late.’ c.

chegar arrive.INF

Ninguém provavelmente vai chegar nobody probably go.PRS.3SG arrive.INF ‘Nobody probably is going to arrive late.’

atrasado. late

atrasado. late

For Duarte, the fact that a speaker-oriented adverb may appear to the right of a hanging topic (42a), a preverbal subject (42b) or a NegQP (42c) suggests that not all preverbal elements are left-dislocated.⁴⁸ She further notes that interjections like oxalá “would that/God willing” can appear between a marked topic and its comment (43b, cf. 43a), but not between a

47 Examples (42a–c) taken from Duarte (1997: 583, exs. 7a–c). 48 Note that for López (2009), a hanging topic in an HTLD context (see Cinque 1983/1997, 1990) is different from CLLD because HTLD can be doubled by a strong pronoun or an epithet (as is the case in 42a), while CLLD elements cannot. López follows Haegeman (1991) and Shaer & Frey (2005) in assuming that HTLD elements are “orphans” and should not be considered part of the clausal structure. Therefore, HTLD elements do not necessarily provide insight on the status and/or position of preverbal subjects. I return to a discussion of HTLD and its relation to preverbal subjects in Chapter 5.

The debate on subjects in European Portuguese   

   49

preverbal subject and its predicate (43c), which serves as further evidence against a uniform analysis for preverbal constituents.⁴⁹ (43)

a.

Oxalá que os alunos leiam esse livro oxalá that the students read.PRS.SBJ.3PL that book antes do exame! before of-the exam ‘Would that the students read that book before the exam!’

b. Esse livro, oxalá os alunos o leiam that book oxalá the students CL.ACC.M.3SG read.PRS.SBJ.3PL antes do exame! before of-the exam ‘That book, would (that) the students read it before the exam!’ c.

*Os alunos oxalá leiam esse livro antes  the students oxalá read.PRS.SBJ.3PL that book before  do exame!  of-the exam  ‘The students would that they read that book before the exam!’

By Duarte’s analysis, preverbal subjects and marked topics in EP appear in different structural positions: preverbal subjects in an IP-internal position and topics external to IP, adjoined either to IP or CP in the left periphery (see also Duarte (1987, 1997) for additional discussion). With respect to information structure, Duarte presents question-answer pairs to show that both preverbal subjects (44a) and marked topics (44b) must transmit given information, and are not felicitous replies to narrow-focus questions.⁵⁰ (44)

Who reacted badly to the test? a. #Os alunos reagiram mal ao teste.  the students react.PST.3PL badly to_the test  ‘The students reacted badly to the test.’

49 Examples (43a–c) taken from Duarte (1997: 583, exs. 8a–c). 50 Examples (44a, b) are slight modifications of Duarte (1997: 584, exs. 9a, b). Contextualizing questions have been translated to English.

50   

   The interaction between syntax and information structure

What do women adore? b. #Perfumes, as mulheres adoram.  perfumes the women adore.PRS.3PL  ‘Perfumes, women adore.’ These infelicitous replies to subject narrow-focus (44a) and object narrow-focus (44b) questions support her proposal that preverbal subjects and topics in EP typically represent discourse-known information, and thus cannot appear preverbally when they are newly-acquired (i.e. narrow-focused) information. Duarte distinguishes between informational focus, the part of a sentence with the greatest level of novelty, and identificational focus, which identifies to the exclusion of other elements (e.g. Szabolcsi 1981, Kiss 1995). In (45a, b) contrastive focus is infelicitous as a reply to a narrow focus question, thus distinguishing it from informational focus.⁵¹ (45)

Who does the king govern? a. #AO POVO governa o  to_the people govern.PRS.3SG the  ‘THE PEOPLE governs the king.’

rei. king

Who is it that little is known about? b. #DELE se sabe  of_him CL.PASS.3SG know.PRS.3SG  ‘ABOUT HIM little is known. ‘

pouca coisa. little thing

However, even when an appropriate context for contrastive focus is created (46a, b), it doesn’t always result in a felicitous reply.⁵² (46)

In medieval society, the king governs the aristocracy. a. #Não, AO POVO governa o rei e não À  no to_the people govern.PRS.3SG the king and not to_the  ARISTOCRACIA.  aristocracy  ‘No, THE PEOPLE the king governs and not THE ARISTOCRACY.’

51 Examples (45a, b) are slight modifications of Duarte (1997: 584, exs. 11a, b). Contextualizing questions have been translated to English. 52 Examples (46a, b) are taken from Duarte (1997: 585, exs. 13a, b).

The debate on subjects in European Portuguese   

   51

…but little is known about the life of Laurence Olivier and Vivien Leigh b. #DELE se sabe pouca coisa e não  of_him CL.PASS.3SG know.PRS.3SG little thing and not  DELA.  of_her  ‘ABOUT HIM little is known and not ABOUT HER.’ For Duarte, the absence of WCO effects in (47a) further demonstrates that this dislocated element differs from identificational focus (47b), which displays WCO effects and lacks quantificational force.⁵³ (47)

a.

[Ao povo]i governa o seui legítimo rei. to_the people govern.PRS.3SG the POSS.M.3SG legitimate king ‘The people, its legitimate king governs.’

b. *É [ao povo]i que governa o seui  be.PRS.3SG to_the people that govern.PRS.3SG the POSS.M.3SG  legítimo rei.  legitimate king  ‘It is the people that its legitimate king governs.’ Duarte uses the terms categorical judgments and thetic judgments (predication and presentational relations, respectively, in Guerón’s (1980) terminology). By her analysis, in which X represents a variable with the information status of given/ known, categorical judgments follow an (X)-S-V word order pattern (48a), thetic judgments follow a pattern of (X)-V-(Y)-S (48b), and D-linked thetic judgments follow X-V-(Y)-S word order (48c).⁵⁴

53 Examples (47a, b) are taken from Duarte (1997: 585, exs. 14b, c). 54 These examples are taken from Duarte (1997: 585, exs. 15a–c). This use of thetic judgment appears to differ from Kuroda’s (1972) definition and from Zubizarreta’s (1998) description of out-ofthe-blue sentences, by which neither subject nor object holds a privileged status informationally, and all information involved is new with respect to the common ground. This is how Costa (2004, see (57a) below) operationalizes the term, and it is also how I employ the term in this and subsequent chapters. It is unclear to me how a sentence may be at once discourse-linked and thetic/ out-of-the-blue. Duarte does not discuss what modifications in these patterns are necessary (e.g. substitution of S by pro) when a subject is D-linked. Note also that chegar (48a, b) is an unaccusative predicate, which is standardly analyzed as underlyingly VS. However, nothing crucially hinges on this assumption since it appears to be a predicate that could classify as an alternating agentive predicate, given its animate subject (see discussion in Perlmutter 1978: 162–165).

52   

(48)

   The interaction between syntax and information structure

a.

O João chegou. categorical the João arrive.PST.3SG ‘João arrived.’

(X)-S-V

b. Chegou o João.

thetic

(X)-V-(Y)-S

c.

D-linked thetic

X-V-(Y)-S

Ao povo governa o rei.

The variable X is D(iscourse)-linked (e.g. Pesetsky 1987) in Duarte’s analysis, and its status of given or not given is unknown to the computational system. Subjects of predication (49a, b) and D-linked elements (50a,b) do not exhibit the same binding behaviors in the contexts below.⁵⁵ (49)

a.

Por for os the

isso, [os meninos]i prometeram [PROi arrumar that the children promise.PST.3PL clean up.INF brinquedos]. toys

b. ??Por isso prometeram [os meninos]i [PROi arrumar os brinquedos].   ‘For that reason, the children promised to clean up the toys.’ (50)

a.

Por isso, o Joãoi pensa nelei/j com preocupação. for that the João think.PRS.3SG in-him with worry

b. Por isso pensa [o João]i nelej/*i com preocupação. ‘For that reason, João worries about him.’ Example (49a), which involves subject control, displays a downgrade in acceptability when the matrix subject appears postverbally (49b). Additionally, a preverbal subject may bind a PP-internal pronoun (50a), but a postverbal subject may not (50b). For Duarte, the above facts strongly suggest that preverbal and postverbal subjects appear in different structural positions within the syntax. Duarte assumes a hierarchy of syntactic projections with both AgrSP and AgrOP (AgrSP > TP > AgrOP > VP). Since EP has visible V-to-T movement, nominative Case is checked by a DP moving to Spec, TP, which is proposed to be the syntactic position where postverbal subjects appear. Duarte proposes that the formal features of heads AgrS and T codify discourse properties (thetic judgments or categorical judgments, respectively). Strong vs. weak EPP is the feature of AgrS that codifies this distinction. She also proposes that in languages like English, an 55 Examples (49a, b) are taken from Duarte (1997: 586, exs. 16a, 17a). Examples (50a, b) are Duarte’s examples (16b) and (17b).

The debate on subjects in European Portuguese   

   53

argumental or expletive DP in Spec, AgrSP does not encode a distinction between predications and presentations, or the status (e.g. topic) of a DP. The relevant feature for AgrSP is [+D] in English (51). (51)

English-type languages:⁵⁶ SVO [[Spec, AgrSP DP] [AgrSP [+D] [Spec, TP t [TP [VP V]…]]]]

(52)

EP-type languages: SVO [[Spec, AgrSP DP] [AgrSP [+EPP] V+T [Spec, TP t [TP [VP t ]…]]]] VSO [[Spec, AgrSP pro] [AgrSP [-EPP] V+T [Spec, TP DP [TP [VP t ]…]]]]

According to Duarte’s analysis of EP (52), a DP in Spec, AgrSP encodes the predicational status of the sentence and consequently, the topic status (given or not) of the DP in question. Thus, the relevant feature of AgrSP in EP is [+EPP]. A subject DP moved to Spec, AgrSP (attracted by a strong [EPP] feature) results in a predicational status for the sentence, and with it, a categorical judgment. Thus the preverbal subject is interpreted as an unmarked topic. Essentially, [±EPP] determines informational status and whether a subject moves past the verb to Spec, AgrSP. So, if a [-EPP] head is in AgrS, the subject DP merges in Spec, TP (at Spell Out), and expletive pro is inserted in Spec, AgrSP.⁵⁷ The resulting derivation is presentational (i.e. a thetic judgment), and as expected, the DP subject doesn’t receive a topical interpretation, and is considered new information. This proposal gives formal form and content to the categorization of discourse-oriented language (see also e.g. Raposo 1986; Duarte 1987). She proposes that discourse-oriented languages, via formal features that characterize their corresponding functional heads, make use of features that encode information that is relevant at the level of discourse structure. On the other hand, a discourse-configurational language makes use of specialized functional heads to encode relevant behavior(s) at the level of discourse structure (e.g. Kiss 1995). For Duarte, EP does not have specialized heads to encode discourse structure. This is somewhat of a paradox, however, because it is basically doing the same thing as specialized heads with pre-existing “core” heads by assigning them special, potentially ad hoc, features to encode discourse structure within the core syntax. Costa (2004) argues that preverbal subjects in EP are canonical elements that appear in Spec, TP. An important asymmetry between preverbal subjects (53a)

56 Note that these structure typology summaries are mine based on the article. Duarte’s analysis seems to suggest verb movement for English – perhaps as high as AgrS. As she does not specify the details, I avoid committing to this portion of the analysis. 57 Note that if AgrSP is in fact [-EPP], the merge of an expletive pro in VSO in (52) is superfluous.

54   

   The interaction between syntax and information structure

and other left-peripheral elements (53b) arises with a fronted subordinate clause wh-element. ⁵⁸ (53)

a.

Perguntei [que livro] o Pedro leu [t]. ask.PST.1SG what book the Pedro read.PST.3SG ‘I asked what book Pedro read.’

b. *Perguntei [que livro], á Maria, lhe deram [t].  ask.PST.1SG what book to_the Maria CL.DAT.3SG give.PST.3PL  ‘I asked what book, to Maria, they gave.’ If the subject o Pedro in (53a) appears in the same left-dislocated position as the fronted indirect object á Maria in (53b), it should incur the same minimality violation, contrary to fact. Additionally, the non-obligatory nature of so-called subjectverb inversion in (53a) suggests that the wh- element que does not compete with the preverbal subject for the same preverbal position. In EP, multiple constituents may appear in a dislocated position (54a) and without ordering restrictions (54b). However, such flexibility does not extend to preverbal subjects (55a, b), as the preverbal subject appears to require adjacency to the verb (55b).⁵⁹ (54)

a.

Aos alunos, sobre sintaxe, o Rui falou. to_the students about syntax the Rui speak.PST.3SG ‘To the students, about syntax, Rui spoke.’

b. Sobre sintaxe, aos alunos, o Rui falou. about syntax to_the students the Rui speak.PST.3SG ‘About syntax, to the students, Rui spoke.’ (55)

a.

Esse bolo, o Paulo comeu-o. that cake the Paulo eat.PST.3SG-CL.ACC.M.3SG ‘That cake, Paulo ate.’

b. ??O Paulo, esse bolo, comeu-o.   the Paulo that cake eat-PST.3SG-CL.ACC.M.3SG   ‘Paulo, that cake, ate.’

58 Following examples are Costa’s (2004: 15, exs. 9a, b, respectively). See Costa (2004) for additional data on Portuguese-based Creoles and dialectal Portuguese. 59 These examples are Costa’s (2004: 13, exs. 5a, b, 6a, b, respectively).

The debate on subjects in European Portuguese   

   55

If preverbal subjects were dislocated elements like those in (54), we would predict (55b) to be as acceptable as (55a), contrary to fact. Following Costa’s assumption that Ā-movement reconstructs and that A-movement cannot, a preverbal subject in Spec, IP (56a) should not be able to reconstruct, while a left-dislocated subject (56b) should be able to, a prediction which bears out. ⁶⁰ (56)

a.

Três livros foram lidos por dois estudantes. three books be.PST.3PL read by two students ‘Three books were read by two students.’ (S > agent; *agent > S)

b. Três livros, dois estudantes leram-nos. three books two students read-PST.3PL-CL.ACC.M.3PL ‘Three books, two students read them.’ (S > O; O > S) By Costa’s analysis, the inability of the preverbal passive subject of (56a) to reconstruct rules out an interpretation in which the agent takes scope over the subject theme. In (56b) however, the left-dislocated object can reconstruct to a postverbal position, thus allowing for scope ambiguity.⁶¹ For Costa, another problematic aspect of the left-peripheral subject analysis for EP has to do with information structure. In thetic, or “out of the blue”, sentences SVO (57a) is the only felicitous word order possible.⁶² (57)

What happened? a. O Pedro partiu o braço. the Pedro break.PST.3SG the arm

60 The inability of A-movement to reconstruct is not without controversy (e.g. Sportiche 1999). These examples are Costa’s (2004: 15, exs. 10a, b, respectively). 61 Example (56a) is hardly ideal for showing reconstruction; for even if we admit A-reconstruction, the reconstruction site for the subject theme would be structurally higher than the PP adjunct por dois estudantes. Example (56b) would be more enlightening if it were compared with Dois estudantes leram três livros as this would indicate the scope possibilities of a sentence that is canonically ordered. Even with A-reconstruction, one would predict S > O scope, but not O > S. Regardless, following Cinque (1990), if the left-dislocated direct object três livros is base-generated, the apparent reconstruction effect is created by the direct object clitic double –(n)os. 62 This example and (58) below are Costa’s (2004: 16). Both are numbered (11) in his monograph, perhaps owing to the fact that both start with the same question in Portuguese O que é que aconteceu?, but their replies differ.

56   

   The interaction between syntax and information structure

b. #Partiu o Pedro o braço.  break.PST.3SG the Pedro the arm  ‘Pedro broke his arm.’ c.

#O braço, o Pedro  the arm the Pedro  ‘His arm, Pedro broke (it).’

partiu-o. break.PST.3SG-CL.ACC.M.3SG

For Costa, the facts in (57) are problematic for the uniform, left peripheral account for two reasons. If o Pedro were a dislocated element in (57a), we would have no straightforward explanation for why (57c), with a dislocated direct object, is an infelicitous response to an out-of-the-blue question. Secondly, if we assume that the subject in (57a) is left-dislocated, we then have to explain why it cannot stay in its initial merge position (i.e. Spec, vP), as in (57b).⁶³ Costa proposes that (57b) is not infelicitous due to the exhaustive nature of the subject in inversion constructions, primarily because its behavior is the same as unergative predicates (58a, b) and alternating mono-argumental intransitive predicates (58c, d). (58)

What happened? a.

O João espirrou. the João sneeze.PST.3SG ‘João sneezed.’

b. #Espirrou  sneeze.PST.3SG  ‘João sneezed.’ c.

o the

João. João

O João viajou. the João travel.PST.3SG ‘João traveled.’

d. #Viajou o  travel.PST.3SG the  ‘João traveled.’

João. João

63 If Barbosa (2000) is on the right track and preverbal subjects of this flavor (i.e. as in 57a) are CLLD elements, they should be base-generated in the left-periphery and co-indexed with pro by most accounts, and not as the result of movement. See the discussion on López (2009) however for an alternative account by which CLLD elements are a product of movement.

The debate on subjects in European Portuguese   

   57

His claim is that if the predicate-internal A-position (Spec, vP) were the only position for the subject, we would predict a preference for VS order with these predicates as well. Preverbal constituents with raising predicates in EP (59a, b, examples 12a, b in Costa 2004: 17) are typically thought to involve left-dislocation, and not movement to Spec, IP since EP lacks overt expletives. Movement from the subordinate clause to the superordinate Spec, IP position followed by movement to a leftperipheral position would constitute super-raising. (59)

a.

O homen parece que viu um monstro. the man seem.PRS.3SG that see.PST.3SG a monster ‘The man it seems that he saw a monster.’ b. O João parece que está parvo. the João seem that be.PRS.3SG foolish ‘João it seems that he is foolish.’

Costa notes that there is also a definiteness effect in this construction since indefinite preverbal subjects (60, ex. 13a in Costa 2004: 17) cannot be left-dislocated. (60)

*Umas meninas parece que  some girls seem.PRS.3SG that  ‘Some girls seems that (they) are sick.’

estão doentes. be.PRS.3PL sick

For Costa, this definiteness effect is expected, since normally left-dislocation can only affect definite XPs. Preverbal subjects, however, exhibit no such definiteness effects in passive sentences (61a, a’), with stative predicates (61b, b’), or with transitive predicates (61c, c’) – even with a bare nominal.⁶⁴ (61)

a.

O homem foi assassinado. the man be.PST.3SG assassinated ‘The man was assassinated.’

a’. Un homem foi a man be.PST.3SG ‘A man was assassinated.’

assassinado. assassinated

64 Examples in (61) from Costa (2004: 17, exs. 14a–c’).

58   

   The interaction between syntax and information structure

b. As meninas estão the girls be.PRS.3PL The girls are sick.

doentes. sick

b’. Umas meninas estão doentes. some girls be.PRS.3PL sick ‘Some girls are sick.’ c.

As baleias comem the whales eat.PRS.3PL ‘The whales eat fish.’

c’. Baleias comem whales eat.PRS.3PL ‘Whales eat fish.’

peixe. fish

peixe. fish

Therefore, for Costa, the above facts suggest that preverbal subjects are not leftdislocated elements. Otherwise, if preverbal subjects were necessarily left-dislocated, the facts in (61a’-61c’) would go unexplained. Costa notes Barbosa’s (1995) observation that some null-subject Romance languages require subject clitic-doubling, evidence which works in favor of his left-peripheral subject analysis. Costa illustrates that in EP, doubling is possible (62a) with subjects, but not required (62b). In thetic contexts, however, subject doubling is extremely downgraded (63b cf. 63a). ⁶⁵ (62)

Who read what? a. O João, ele leu o livro. the João he read.PST.3SG the book ‘João, he read the book.’ b. O João leu o the João read.PST.3SG the ‘João read the book.’

(63)

livro. book

What happened? a. O João leu o livro. b. ??*O João, ele leu o livro.

65 Examples (62–63) are Costa’s (2004: 18, exs. 15–16).

The debate on subjects in European Portuguese   

   59

By Barbosa’s (1995) analysis, a left-dislocated preverbal subject can always be doubled, but in light of the data above, this does not appear to be the case for EP. If we take seriously the possibility that EP has subject clitics then we also have to consider their interaction with other pronouns and clitics. The doubled pronoun/ clitic ele may appear in a preverbal (64a) or postverbal position (64b).⁶⁶ (64)

a.

(O the

João,) João

leu read.PST.3SG

ele he

o the

livro. book

b. (O João,) ele leu o livro. the João he read.PST.3SG the book ‘(João,) he read the book.’ The trick now is determining which ele is a pronoun and which is a subject clitic. If we consider the possibility that ele is a subject pronoun in (64b), and we continue to assume that preverbal subjects are left-peripheral elements then both O João and ele must be left-peripheral elements. Costa correctly admits that nothing a priori rules out such an analysis since (65, 19 in Costa 2004: 19)) is a grammatical construction. (65)

O João, a ele, vi-o the João to him see-PST.1SG-CL.ACC.M.3SG ‘João, him I saw at the theater.’

no in_the

cinema. cinema

In (65) both o João and a ele are arguably left-peripheral elements as well. The resumptive object clitic o (enclitic on vi-o) co-occurs with multiple fronting, which leads to the prediction that the subject clitic ele should be able to co-occur with the dislocated preverbal DP subject O João and the preverbal pronoun ele, contrary to fact (66).⁶⁷

66 Examples (64a, b) are Costa’s (2004: 19, ex. 18a, b). 67 Example (66) taken from Costa (2004: 20, ex. 20). This analysis makes crucial (and necessary) rhetorical leaps in regards to cliticization. It is worth pointing out that Barbosa’s (1995) analysis is the only one I am aware of that treats EP pronouns as subject clitics. Note that in the following examples, EP does not allow both a preverbal pronoun and a “subject clitic” to appear adjacent to (i), or separated by the verb (ii) without dislocation of O João. (i)

*Ele ele leu o livro.

(ii) *Ele leu ele o livro.

60   

(66)

   The interaction between syntax and information structure

O João, ele leu (*ele) o livro. the João he read.PST.3SG he the book ‘João, he read (*he) the book.’

Example (66) not only suggests that ele is a subject pronoun, but that preverbal ele appears in a canonical (Spec, TP-type) position, thus explaining why it may not appear concurrently in a postverbal and preverbal position. As Costa shows, EP allows for complementizer-less if-clauses with conterfactual hypotheticals, but such constructions only allow postverbal subjects (67a, cf. 67b).⁶⁸ (67)

a.

Tivesse o João ido ao have.PST.SBJV the João go.PTCP to_the ‘Had João gone to Brazil...’

b. *O João tivesse  the João have.PST.SBJV  ‘João had gone to Brazil...’

ido go.PTCP

Brasil… Brazil

ao Brasil… to_the Brazil

Following Costa, the fact that only postverbal subjects are possible in these constructions is indicative of I-to-C movement in EP. A preverbal subject appearing in Spec, IP then correctly predicts (67a), since the auxiliary verb ter moves across the preverbal subject to the C realm. If the preverbal subject were left-dislocated, however, we might expect (67b) to be grammatical, contrary to fact. However, as (68) shows, a preverbal subject in a left-dislocated position is not prohibited as long as an overt subject appears lower than the verb (68). (68)

O João, tivesse ele ido the João have.PST.SBJV he go.PTCP ‘João, had he gone to Brazil...’

ao to_the

Brasil... Brazil

Importantly, example (68) shows again that even though a preverbal subject may appear in a left-peripheral position, it is not required to appear in the left periphery. Costa’s analysis also examines child acquisition data for EP. If preverbal subjects are left-dislocated, the prediction is that VSO should be the unmarked word order in child production. Adragão (2001) showed, however, that VS order is marked in early child production, appearing in only 7% of a total of 1060 sen-

68 Examples (67a, b) taken from Costa (2004: 19, exs. 17a, b).

Taking stock of preverbal subject analyses   

   61

tences. Most of the VS sentences produced were passives, unaccusative, and predicative structures, which have unmarked VS order in adult production also. That very little left-dislocation appears in the data is claimed to be the result of the late acquisition of clitics in EP (e.g. Duarte & Matos 2000), thus further indicating that preverbal subjects are unmarked in EP. For this reason, Adragão & Costa (2003) predict that only once children master left-dislocation should they produce SV orders. Friedmann & Costa (2011) also examined childhood L1 acquisition data, finding that L1 child speakers of EP and Hebrew produce only SV with transitive predicates, while L1 child speakers of Arabic, a VS(O) language, and Argentine Spanish produce VS with transitive predicates. They link this VS tendency in Spanish to the availability of (accusative) clitic doubling in Argentine Spanish and Palestinian Arabic, which has been found to be mastered early in Spanish-acquiring children (at age 1;07 according to Torrens & Wexler (1996)). This is taken to suggest that the first spell-out phase in Spanish and Palestinian Arabic is vP, while in Hebrew and EP, this phase is IP/TP. This conclusion at once suggests that Spec, TP in EP is an A-position, while the same position in (at least Argentine) Spanish is an Ā-position. Despite the indications of Costa and colleagues’ tests in favor of a canonical, preverbal subject as argument analysis, Costa (2004) suggests that the results of many of the syntactic tests in that work do not automatically extend to other null subject languages, i.e. there is a need for thoroughness when examining the data, and that a few syntactic or pragmatic tests alone may prove to be misleading. I return to Costa’s (2004) analysis of the syntax-information structure interface in EP below, where I revisit analyses that specifically treat the syntax-information structure interface.

2.6 Taking stock of preverbal subject analyses As we have seen thus far this chapter, there are substantive arguments for and against each side of the argument/non-argument debate in both Spanish and European Portuguese. It appears that the only definitive conclusion that one may arrive at after considering the above analyses is one noted above in (68), based on data in Costa (2004): that preverbal subjects may appear in a dislocated position, albeit not necessarily so. López (2009: 109, ex. 124) arrives at the same conclusion for Catalan, citing sentences like (69a) in which a preverbal subject may clearly be left-dislocated. The Spanish equivalent of this example (69b, translation mine) shows that the same holds for Spanish.

62   

(69)

   The interaction between syntax and information structure

a.

El Joan, jo no crec que pugui ensenyar the Joan I not think.PRS.1SG that can.PRS.SBJ.3SG teach.INF un curs d’àlgebra. a course of algebra ‘Joan, I don’t think he’s able to teach an algebra class.’

b. Juan, (yo) no creo que pueda enseñar un curso de álgebra. However, López (2009: 109, ex. 122a) also shows that instances of subject raising are clear cases of A-movement with preverbal subjects in Catalan (70a), which also holds in the Spanish equivalent (70b, translation mine):⁶⁹ (70)

a.

El Joan em sembla ser the Joan CL.DAT.1SG appear.PRS.3SG be.INF ‘Joan seems to be intelligent.’

intel.ligent. intelligent

b. Juan me parece ser inteligente. The data above is quite interesting indeed, and serves to show that sometimes preverbal subjects are a product of A-movement and sometimes they are a product of Ā-movement. If it were as simple as this, this would be a much shorter monograph; however, I will show in the next section that many of the test data presented for EP and Spanish do not behave the same way in Galician. These results are perhaps not entirely surprising, especially in light of Costa’s (2004) warning that cross-linguistic factors may limit the validity of the syntactic tests examined in the literature. Burga (2007, 2009) also suggests that dialectal factors may influence the outcome of such tests, which makes sense since she addressed the issue of preverbal quantifier scope via data from another variety of Spanish. As we will see, data from Galician hardly clarify the issue regarding preverbal subjects in null-subject languages.

2.7 Preverbal subjects in Galician As we have seen thus far, there are worthy arguments for both sides of the debate related to preverbal subjects as canonical vs. non-canonical / argument vs. nonargument constituents, especially in Spanish and European Portuguese. This debate is of great theoretical importance, and holds implications for the universal69 See Torrego (2002) and López (2007) for further discussion of these examples and of interesting asymmetries related to the presence or absence of a clitic pronoun.

Preverbal subjects in Galician    

   63

ity of the EPP and its associated feature(s). Gupton (2006, 2010, 2014a, b) applies many of the tests examined earlier in this chapter to Galician, but nets inconclusive results, suggesting that the analysis of preverbal subjects depends on more than syntactic tests of A- or Ā-status. This inconclusiveness is the impetus for the additional quantitative and qualitative data-gathering measures employed and reported on in chapters 3–5. As introduced earlier in this chapter, studies that propose that preverbal subjects are Ā-elements (e.g. O & T 1999 for Spanish) center on important similarities between preverbal subject DPs and preverbal direct and indirect objects, which are generally considered to be left-peripheral (i.e. Ā) elements. Such analyses suggest that preverbal elements be uniformly analyzed due to parallel behaviors in a variety of syntactic tests, many of which prove to be problematic and inconclusive indicators upon closer inspection. I provide examples from Galician below as they pertain to some of the principal issues of debate. An example claimed to be indicative of “uniform” behavior for preverbal subjects and (direct and indirect) objects is the preverbal fronting of negative quantifiers (71) and whelements (72).⁷⁰,⁷¹ Consider the following examples in Galician.

70 Note that previous studies have also examined similar behavior in contrastive fronted focus phrases. Contrastive focus phrases in Galician (i) behave differently than negative quantifiers and wh-elements in that 1) a preverbal subject may intervene, and 2) they trigger enclisis rather than proclisis. (i)

A CENORIA o coello comeuna /*a comeu (non the carrot the rabbit eat.PST.3SG-CL.ACC.3SG.F not ‘The rabbit ate THE CARROT (not the apple).’

a the

mazá). apple

Given this behavior, in Gupton (2010, 2012, 2014a), I discuss the status of focus fronting in Galician, proposing that contrastive fronted focus elements in Galician are in fact contrastive topics. It is worth pointing out that Fernandez-Rubiera (2009) presents judgments from eastern Galician suggesting that proclisis is preferred with contrastive fronted focus. I have consulted with more than ten native, habitual speakers from western and central Galicia, all of whom roundly reject proclisis with contrastively fronted elements. While these speakers claim that production of contrastive focus with proclisis is indicative of a novofalante, or L2 speaker of Galician, I suspect that further investigation will determine such data to be representative of interdialectal variation. 71 Note that the presence of wh-elements, negation, and negative quantifiers trigger proclisis in Galician. While nothing appears to be structurally wrong with (71c) with a postverbal subject, one of my informants in particular found it to be “very unnatural, almost poetic”. This is why I classify it as downgraded (?). Potential asymmetries between negative quantifier types (subjects and direct objects v. indirect objects) are certainly worthy of further investigation, but are beyond the scope of this monograph. I discuss direction of cliticization in greater detail in Chapter 5.

64   

(71)

   The interaction between syntax and information structure

a.

Ningúen lle debe o aluguer a María. nobody CL.DAT.3SG owe.PRS.3SG the rent to Maria ‘Nobody owes rent to Maria.’

b. Nada (*Xoán) lle debe (Xoán) aos seus nothing (Xoán) CL.DAT.3SG owe.PRS.3SG Xoán to_the his amigos. friends ‘Xoan owes nothing to his friends.’

(72)

c.

A ninguén (*?Xoán) lle debe (?Xoán) o aluguer. to nobody (Xoán) CL.DAT.3SG owe.PRS.3SG (Xoan) the rent ‘Xoan owes rent to nobody.’

a.

Quen lle debe o aluguer who CL.DAT.3SG owe.PRS.3SG the rent ‘Who owes rent to Maria?’

a María? to Maria

b. Que (*Xoán) lle debe (Xoán) aos seus nothing (Xoán) CL.DAT.3SG owe.PRS.3SG Xoán to_the his amigos? friends ‘Xoan owes nothing to his friends.’ c.

A quen (*Xoán) lle debe (Xoán) o aluguer? to nobody (Xoán) CL.DAT.3SG owe.PRS.3SG (Xoan) the rent ‘To whom does Xoan owe rent?’

The fact that the above fronting phenomena do not tolerate intervening topics or subjects (i.e. triggering subject-verb “inversion”) suggests that the preverbal position – be it Spec, TP or something else entirely – is not reserved exclusively for subjects in Galician. Zubizarreta (1998) proposes that Spec, TP is a syncretic head that may host a variety of features, a proposal that is tantamount to suggesting that Spec, TP is at once an A and Ā-position, available for subjects, negative quantifiers, wh-elements, fronted focus, and CLLD XPs. Goodall (1993), however, speculates that such elements simply transit through Spec, TP toward an eventual Ā-position, thus making Spec, TP unavailable, perhaps even for preverbal subjects. Another debated example involves scope relations. Jaeggli (1985) and UribeEtxebarria (1992, 1995) examine subject quantifier scope freezing effects in Spanish, a phenomenon that has also been claimed to suggest that preverbal subjects behave like Ā-elements, since quantified preverbal subjects in Spanish cannot

Preverbal subjects in Galician    

   65

undergo covert quantifier movement to take scope over the already extracted whelement (73a). No such effects materialize in Galician, however (73b).⁷² (73)

a.

A quién dices que cada senador amaba t? (Wh->∀,*∀>Wh-) to who say.PRS.2SG that each senator love.IMPFV.3SG

b. A quen dis que cada senador amaba t? (Wh->∀,∀>Wh-) to who say.PRS.2SG that each senator love.IMPFV.3SG ‘Who is it you say that each senator loved?’’ According to López (p.c.), however, freezing effects even in Spanish are weak at best, and disappear when a wh-element and a quantified subject appear in the same clause (74). (74)

En qué hotel descansa cada senador? in which hotel rest.PRS.3SG each senator ‘Which hotel is each senator resting in?’

(Wh->∀, ∀>Wh-)

In (74) the postverbal subject may take scope over the adjunct wh-element, and is not frozen in postverbal position. The fact that such “freezing effects” disappear in these environs weakens not only the Ā-position argument for Spanish preverbal subjects, but also the empirical value of this test overall. As seen above in section 2.4.2, Goodall (2001) argues against a left-peripheral analysis for preverbal subjects in Spanish, maintaining that analogues of (75) suggest an asymmetry between preverbal subjects and left-peripheral preverbal objects in Spanish. The same holds for Galician: a subordinate clause topicalized direct object (o premio in 75a) triggers island effects, but a subordinate clause preverbal subject (Xoán in 75b) does not. (75)

a.

*A quen cres que o premio lle deron?  to whom think.PRS.2SG that the prize CL.DAT.3SG give.PST.3PL  ‘To whom do you think they gave the prize?’

72 The judgments of one Galician speaker consulted suggest that only postverbal subjects in Galician undergo scope freezing. This phenomenon deserves further attention, but is not pursued in this monograph.

66   

   The interaction between syntax and information structure

b. A quen cres que Xoán lle deu to whom think.PRS.2SG that Xoán CL.DAT.3SG give.PST.3SG o premio? the prize ‘To whom do you think Xoan gave the prize?’ If topics and preverbal subjects are Ā-elements, both should constitute an island for wh- movement, which is clearly not the case in (75b). Goodall’s analysis of Spanish considers this positive syntactic evidence for preverbal subjects appearing in Spec, TP, which is generally assumed to be an A-position. The inconclusive results briefly introduced in this section suggest that the analysis of preverbal subjects may depend on more than syntactic tests of A- or Ā-status for insight, especially when so many produce questionable or conflicting indications. In fact, as previously mentioned Costa (2004) has suggested that certain tests may lack cross-linguistic validity. Although the data presented to this point are not uniformly in favor of preverbal subjects appearing in either A or Ā-positions, the data lean more toward them being L-related A-positions since preverbal subjects never behave as islands when verb-subject “inversion” phenomena are not involved, a position which assumes that CLLD constituents are base-generated.⁷³

2.8 Syntactic structures in context In this section, I discuss the importance that the information structure can hold for the acceptability and the analysis of certain word orders. I will show that languages like Spanish appear to allow for a large degree of flexibility with respect to word order, but once certain convergent word orders are placed within certain communicative contexts, they cease to be acceptable. I also discuss many of the basic concepts and dichotomies associated with analyses of information structure in the literature such as topic/theme and focus/rheme, and discuss the assumptions that I make with respect to each in this monograph. I also discuss some of

73 The issue of base-generation versus movement is crucial with respect to the presence or absence of reconstruction effects; however, an in-depth investigation of the base-generation versus movement debate for CLLD would take the current discussion too far afield. Note, however, that López’s (2009) approach to the left-periphery, discussed in greater detail later in this chapter, treats both focus fronting and CLLD as products of syntactic movement. I leave a more in-depth treatment of this debate, which is far from uncontroversial, for ongoing research on contrast in Galician and for future research.

Syntactic structures in context   

   67

the formal syntactic accounts for information structure, as well as the theoretical problems and limitations associated with such analyses. In particular, I examine López’s (2009) proposal for the syntax-information structure interface.

2.8.1 When syntax meets discourse According to Casielles (2004), SVO word order in Spanish (76d) is felicitous in reply to a variety of question types. It can answer a thetic question (76a), a wide (predicate) focus question (76b), or a direct object narrow-focus (rheme) question (76c). (76)

a.

¿Qué pasó?  what happen.PST.3SG  ‘What happened?’

b. ¿Qué hizo Juan?  what do.PST.3SG Juan  ‘What did Juan do?’ c.

¿A quién llamó  to whom call.PST.3SG  ‘Who did Juan call?’

Juan? Juan

d. Juan llamó a su mujer. Juan call.PST.3SG to his wife ‘Juan called his wife.’ For Zubizarreta (1998) SVO is not a felicitous response to a subject narrow-focus (rheme) question (77a). A reply such as (77a) with main prosodic prominence (indicated by underscore) on the unknown, narrow-focused element (i.e. the subject) would be appropriate in English, but inappropriate in Spanish.⁷⁴ Zubizarreta claims that such prominence in Spanish forces a contrastive, emphatic

74 However, see e.g. Hoot (2012) for experimental evidence from native and heritage speakers of Mexican Spanish indicating that this order is acceptable in these varieties.

68   

   The interaction between syntax and information structure

interpretation on the subject (77b), and that VOS (77c) is the only appropriate reply to a subject narrow-focus question.⁷⁵,⁷⁶ (77)

Who ate an apple? a.

#Juan comió  Juan eat.PST.3SG

una an

manzana. apple

b. #JUAN comió una manzana  JUAN eat.PST.3SG an apple  ‘JUAN ate an apple (not Pedro).’ c.

Comió una manzana eat.PST.3SG an apple ‘Juan ate an apple.’

(no not

Pedro). Pedro

Juan. Juan

She proposes that main prominence in Spanish lies on a phrase-internal constituent (i.e. Juan in 77c), which renders SVO an incompatible word order with noncontrastive information focus (i.e. rheme) on the subject in the question Who ate an apple? in (77). The same can be seen in a direct object narrow-focus questionanswer (78). ⁷⁷

75 Note that for Ordóñez (1997) VOS is an infelicitous reply for thetic questions (ia), but VSO (ib) is not. Ordóñez notes that VSO improves with a preverbal time adverbial (ib). It is also worth noting that, in my experience, the acceptability of VSO in transitive declaratives is subject to interspeaker variation. (i)

What happened yesterday? a. #(que) ayer ganó la lotería  that yesterday win.PST.3SG the lottery  ‘(that) yesterday John won the lottery.’

Juan John

b. Ayer ganó Juan la lotería. yesterday win.PST.3SG John the lottery ‘Yesterday John won the lottery. ‘ 76 Example (77a, b) are examples (51a, b) in Zubizarreta (1998: 20). Note that she marks (77a) as *, but I mark it # since it is not strictly ungrammatical outside of the context provided. Example (77c) is Zubizarreta’s (1998: 22, ex. 57a). 77 Example (78a) is Zubizarreta’s (1998: 22, ex. 58a). Examples (78b, c) are (53a, b) in Zubizarreta (1998: 21). Note that Zubizarreta marks (78b) as * and (78c) as grammatical, although this was to illustrate the reading imposed by the prosody of (78b). As neither are felicitous replies to the context, I mark them #.

Syntactic structures in context   

(78)

   69

What did Maria put on the table? a.

María puso sobre la mesa el libro. Maria put.PST.3SG on the table the book ‘Maria put the book on the table.’

b. #María puso el  Maria put.PST.3SG the c.

libro book

sobre la on the

mesa. table

#María puso el LIBRO sobre la mesa (no la revista).  Maria put.PST.3SG the BOOK on the table not the magazine  ‘Maria put THE BOOK on the table (not the magazine).’

For Zubizarreta, in (78) only the narrow-focused new information (i.e. that which resolves the variable opened by the question) can appear at the rightmost clausal edge, which rules out (78b) as a felicitous reply; (78a) satisfies this requirement. As in (77), placing main prominence on a phrase-internal constituent (78b) forces a contrastive reading (78c), which is incompatible with the direct object narrowfocus question. Therefore, main prominence on the object makes SVO completely compatible with object narrow focus questions. By Zubizarreta’s analysis, rightmost prominence is a result of the interaction of the Nuclear Stress Rule (NSR, see Chomsky & Halle 1968 for its original formulation) and the Focus Prominence Rule (FPR) in Spanish. Zubizarreta (1998: 76, ex. 127) reformulates the NSR for Spanish and French as in (79a): (79)

Spanish and French NSR a.

Given two sister nodes Ci and Cj, the one lower in the asymmetric c-command ordering is more prominent.

b. All phonological material is metrically visible for the NSR in Spanish. The additional statement in (79b, example 128 in Zubizarreta 1998: 76) sets Spanish apart from French, in which defocalized and anaphoric constituents are metrically invisible for the NSR. The NSR interacts with the FPR (80, 56 in Zubizarreta 1998: 21). (80)

Focus Prominence Rule (FPR) Given two sister nodes Ci (marked [+F]) and Cj (marked [-F]), Ci is more prominent than Cj.

70   

   The interaction between syntax and information structure

Zubizarreta makes use of the diacritic [+F] to mark constituents that form part of the focus, and [-F] to mark constituents that form the presupposition or part of the presupposition.⁷⁸ To avoid conflicts between the NSR and the FPR, [-F] constituents undergo movement. It is important to note that this movement is not motivated by feature checking, but rather Last Resort scrambling which ensures that a focalized constituent marked [+F] is in position to receive prominence via the C-NSR. Zubizarreta calls such movement in languages like Spanish and Italian p-movement, or prosodic movement, which is absent in languages like German, French or English.⁷⁹ She traces the differences in the output of p-movement in Spanish and Italian to a difference in the preverbal field, specifically on T, which she suggests is a syncretic category in Spanish, and which may check the discourse-based functional features focus, emphasis or topic.⁸⁰ The results differ because in Italian, VOS is proposed to be derived by (remnant) movement of VO around S, which has moved from Spec, TP to Spec, FocP, and this movement is subject to the Relative Weight Constraint. Recall that Spec, TP is also claimed to be the syntactic position for preverbal subjects according to proponents of the canonical subject argument. However, whether this is the true position for preverbal subjects is not of immediate importance. What is of relevance, however, is that there is clearly a relation between the clausal structure of a sentence or phrase and its appropriateness within the discourse-information structure in (76–78). This relation must be taken into account in order to describe the clausal word orders that are operative and, more importantly, appropriate in Galician. Among the few extant works available that discuss the discourse configurationality of Galician (e.g. Freixeiro Mato 2006a, b), infelicitous word orders are rarely discussed. Only once we have established which word orders are accepted and preferred can we return to the implications that the syntactic analysis of Galician may have for the larger debate regarding argument positions. In the following sections I discuss the information structure considerations that serve as the foundation for the methodology that I employ to gather such judgment data for Galician.

78 Note that for Zubizarreta, the feature [F] is not a lexical feature, but a derived phrase marker, which remains undefined until after Σ-structure, which essentially constitutes a pre-PF interface level. It is at this level that p-movement also takes place in her analysis. 79 See Gupton & Leal Méndez (2013), Hoot (2012), Leal-Méndez & Shea (2012), Leal Méndez & Slabakova (2011), for important objections to p-movement. 80 See also Zagona (2002: 206–219) for an interesting review of similar proposals for Spanish.

Syntactic structures in context   

   71

2.8.2 Defining Information Structure The division of sentences into informational units dates back to Weil (1879 [1844]: 29), in which he proposes an informational split distinct from that of subjectpredicate. He describes this split as starting with “the ground upon which the two intelligences (speaker and hearer) meet”. From this point, the statement, or énonciation proceeds. Vallduví & Engdahl (1996: 460) describe information packaging as the “structuring of sentences by syntactic, prosodic, or morphological means that arises from the need to meet the communicative demands of a particular context or discourse.” It has been well noted in the literature on information structure that sentences consist of a less informative part (Topic or Theme), and a more informative part (Focus or Rheme). Many different dichotomies have been proposed to account for this split in the information structure: Theme-Rheme, Topic-Comment, Topic-Focus, Focus-Presupposition, Focus-Open Proposition, and Focus-Background (Hockett 1958, Kuno 1972, Halliday 1967, Erteschik-Shir 1986, 1997, Prince 1986, Rochemont 1986, Ward 1988, Vallduví 1990, among others). Vallduví (1990) observes that, while such proposals share the fundamental idea of a split, they diverge in where the split occurs. For example, among the many proposals that separate the Topic/Theme from the rest of the sentence there is disagreement on how to identify the topic. The same frequently results from definitions of focus or rheme. The paradoxical result is that, despite ample disagreement regarding the precise definitions of these terms, they are frequently taken for granted.

2.8.3 Definitions of Topic/Theme Descriptions of theme, or topic, in the literature generally share notions of aboutness, discourse-oldness, shared knowledge, or discourse salience. Mathesius (1975 [1961]: 81) describes the theme as “the element about which something is stated”. Firbas (1964: 272) proposes that, within the approach of communicative dynamism (CD), the theme possesses the lowest level of CD. For him, the theme “need not necessarily convey known information or such as can be gathered from the verbal and situational context. It can convey even new, unknown information.” Contreras (1976: 16) finds the notion of CD too arbitrary, and proposes rather that “the theme contains those elements which are assumed by the speaker to be present in the addressee’s consciousness”. Hockett (1958: 201), who introduced the dichotomy Topic-Comment, describes the topic as “what the speaker is going to talk about”. In a similar vein, Halliday (1967: 212) describes the theme as “what I am talking about now”, or the point of departure for the clause in question.

72   

   The interaction between syntax and information structure

Gundel’s (1988) description of topic characterizes it as what the sentence is about as well, but specifies that it is associated with given (non-focal) information in the sentence, and never has primary stress. Sgall et al. (1986: 80) describe the topic as “the items the speaker supposes to be activated in the hearer’s memory at a given point of time”. Erteschik-Shir (1997), among others, describes topics as old or presupposed information; however, Reinhart (1981) disagrees with the classification of topics as old information. Building on Stalnaker (1978), she characterizes discourse as a joint procedure of building a “context set” consisting of subsets of propositions. Within the context set, sentence topics are one such subset: a way in which one classifies referential entries. Dahl (1974) argues for a tripartite structure involving the two separate articulations: topic-comment and focus-background. Vallduví (1990) refines this idea, noting an overlap in Dahl’s (1974: ex. 3) two dichotomies in (81), in which John is both topic and background, and drinks forms part of both the comment and background. (81) What does John drink? –

topic  comment John drinks beer. background focus

He resolves this redundancy in Dahl by proposing a trichotomy. For Vallduví, topics are called links: they only appear in a sentence-initial position and serve to activate the hearer’s ‘knowledge store’. By this proposal, the sentence is divided first into Focus-Ground, and, from there, the Ground is separated into Link-Tail. In this proposal, Focus is “the only nonelidable part of the sentence” (p. 57), and the Tail is at once non-topic and non-focus. By this analysis then, John drinks forms the link, and beer forms the focus.⁸¹ Lambrecht (1994) takes a view of topics in line with aboutness. He argues that, since pronominal elements may be topical, Vallduví’s sentence-initial characterization of topics is inappropriate. Lambrecht restricts topics to discourse referents. In his classification, such referents are what a proposition is about. Others have characterized topics in different terms such as presupposition, background, or open proposition. Jackendoff (1972) describes presupposition as information assumed by the speaker to be shared with the interlocutor. Dahl (1974) describes the background as the non-focused part of the sentence, a description quite similar to Vallduví’s (1990: 58) description of ground as “the complement of the focus”. Ward (1988) describes an open proposition in terms of its salience assumed by the speaker in the discourse.

81 I discuss the various characterizations of focus in the literature in the following section.

Syntactic structures in context   

   73

Aside from the number of ways in which topical elements have been described, there is the added complexity of the number of flavors topics come in. Languages mark topics in a variety of ways: phonologically, morphologically, lexically, and syntactically. Because of the degree of cross-linguistic variation in how languages may mark topical elements, Casielles (2004) proposes that the definition of topic in one language may be insufficient for describing and characterizing a topic in another language. She suggests that to understand the nature of topics in a language, one must consider the specific characteristics of that language, and how it encodes topical behavior. I examine her treatment of topics in greater detail in section 2.9.1. The basic definition of topic that I assume in subsequent chapters is a rather vanilla one based on the notions of “aboutness” or “discourse old” discussed above. Topical, discourse-old elements tend to appear in preverbal or CLLD positions in Romance. I return to CLLD, in particular the limitations involved in discourse-oldness for CLLD discussed in López (2009), and its relevance for the syntax-information interface briefly in section 2.8.4, and then in greater detail in section 2.9.3. As most accounts of topic occur in a dichotomy with terms like comment or focus, they inevitably involve an accompanying definition of the term focus as well. In the following section, I present some of these characterizations of focus in the literature.

2.8.4 Definitions of Focus/Rheme Informally, focus or rheme elements are “more informative and less topical” elements (Casielles 2004:127) and, generally speaking, are prosodically prominent. They are often assumed to represent new information within the discourse, and also tend to occur toward the rightmost edge of a sentence. Perhaps expectedly so, there are exceptions and complications to this definition. Rochemont (1986) disagrees with the notion that focus correlates with new information based on the existence of focused pronouns (82), which for him are old information (focused constituents appear in capitals in this and subsequent examples unless otherwise noted). (82)

Who did they call? Pat said they called HER.

In (82) the object pronoun her must refer to known, old information. Therefore, since her may be focused, Rochemont claims that the correlation between focus and new information does not hold. Casielles (2004) points out that even though

74   

   The interaction between syntax and information structure

a pronoun may refer to a prominent, discourse-old entity, it does not preclude it from being new information, i.e. the focus of a sentence. In (82), the informative element is the direct object, which is active enough in the discourse to warrant being expressed by the pronoun her. Rochemont’s initial definition of focus (new information) is based on the notion of c-construability. A c-construable element has a semantic antecedent in the discourse δ. Focus elements are not c-construable. A problem noted for such an analysis is the case of a focused subject pronoun, as in (83). (83)

John hit Mary, then SHE hit HIM.

As (83) suggests that a discourse element can be both focused and c-construable, Rochemont proposes abandoning a definition of focus based on new information, and instead proposes two types of focus: Presentational and Contrastive. Presentational focus is described as non-rightmost, non-contrastive stress, and is used with verbs like appear, which in the unmarked variety, have an accented subject (84a). (84)

a.

The case was judged. Then a LAWYER appeared.

b. The case was judged. Then a lawyer APPEALED. By his analysis, verbs like appear differ from others like appeal in that they seem to transfer their status as the focus of new information to their subject. However, Casielles (2004) notes that verbs like appear coincide with the set of unaccusative verbs, which need not necessarily affect focus projection.⁸² For Rochemont, contrastive focus is defined by a rather complex calculus (85).

82 Chomsky (1971) proposed that focus may project in sentences like (i) in which any bracketed elements may be part of the focus in English. However, Casielles observes that this is only possible with rightmost focus in English. Focus that does not appear in a rightmost position in a sentence (ii) cannot project, and must remain narrow. (i)

He was (warned (to look out for (an ex-convict (with (a red SHIRT))))).

(ii) Laurie followed RALPH into the bedroom. For Rochemont, nonrightmost accented elements are not Contrastive focus elements. The notions of marked and unmarked accent are crucial for focus projection (see e.g. Cinque 1993, Reinhart 1995, Nash 1995, Zubizarreta 1998). Unmarked accent is generated by the grammar, falls on the rightmost constituent, and identifies the unmarked focus of the sentence. Only this type of focus may project.

Syntactic structures in context   

(85)

   75

An expression P is a Contrastive Focus in a discourse δ, δ = {φ1, ... φn}, if, and only if, (i)

P is an expression in φi, and

(ii) if P/φi is the result of extracting P from φi, then P/φi is c-construable, and φi is not c-construable. Casielles takes issue with Rochemont’s definition of c-construability because it would treat certain reflexive pronouns (e.g. 86) as contrastive. (86)

Who did John hit? He hit HIMSELF.

While (86) does not appear to be contrastive – at least lacking further information related to the discourse context – the fact that himself has a semantic antecedent would make it c-construable, and, by Rochemont’s proposal, therefore contrastive. For Casielles, an additional complication to this analysis is that certain expressions may be both presentational and contrastive focus. Following the definition in (85), both (87) and (88) should be contrastive since, in her view, the non-focused part is c-construable in both examples. (87)

A: Bill’s financial situation is a source of constant concern to Mary. B: Bill’s financial situation is a source of constant concern to BILL.

(88)

John hit Mary, and then he KICKED her.

However, the only focus that is c-construable by Rochemont’s analysis is the one in (87) because it has an antecedent. The verb KICKED in (88), however, qualifies as contrastive focus in that the non-focused portion of the expression is c-construable and the whole sentence is not. It qualifies as presentational focus in that the focus is not c-construable. While Rochemont finds this overlap in focus types a desirable consequence of his calculus, Casielles finds this problematic, not only because only-Contrastive, only-Presentational, and Contrastive and Presentational are not clearly defined, but also because the additional concept of direct and indirect c-construability are introduced as important for non-focused material (Rochemont 1986: fn. 103). In the end, Casielles rejects the notion of a proposed division of focus which allows for most presentational foci to be contrastive foci at the same time.

76   

   The interaction between syntax and information structure

Gundel (1994) discusses three ways focus has been used in the literature, describing three types of focus: psychological, semantic, and contrastive focus. Psychological focus refers to the center or focus of attention (AI focus in Hajičová (1987)), which would be topical by many of the analyses of topic discussed in section 2.8.3. Casielles (2004) refers to psychological focus as the current center of attention in a discourse, which is also more akin to topical elements. Semantic focus refers to new information being asserted, or “the part of the sentence that answers the relevant wh-question (implicit or explicit) in the particular context in which the sentence is used” (Gundel 1994: 461). This semantic focus can be marked by pitch accent, word order (including special focus positions in the syntax), focus-marking particles, or any combination of these. Semantic focus includes context-active, discourse-old elements such as the pronoun SHE in (89). (89)

Mary said it was SHE (=Mary) who called.

By this analysis then, semantic focus may fall on a previously mentioned or evoked constituent without affecting its status as focus. Therefore, according to this definition, not all semantically focused material need be entirely new to a discourse. This is the concept of focus that Casielles (2004) adopts. Gundel’s contrastive focus (CF) differs from that of Rochemont in that her CF refers to a strategy (phonological or syntactic) for making an element prominent in order to focus an interlocutor’s attention on said element. Due to the fact that Gundel’s CF falls primarily on topics, and due to the potential confusion that can result from such a definition, Casielles prefers to call Gundel’s CF “emphatic stress”. As previously discussed, Vallduví (1990) separates the sentence into Focus and Ground. Ground represents the unfocused portion of the sentence, and is further divided into link and tail. For him, focus is the informative part of the sentence, and is the only part of the sentence that may not be elided. I will not go into the detailed particulars of Vallduví’s Information Packaging calculus, but within his system, Focus comes in two varieties: Retrieve-add focus and Retrievesubstitute focus. When a sentence lacks a tail in its information structure, the relevant information is retrieved by adding focus (thus retrieve-add), and when it has a tail, information is retrieved by substituting focus in the relevant position within the structure (retrieve-substitute). Structures of the latter type with tails correspond to narrow focus, and structures lacking them (retrieve-add) correspond to wide focus. Casielles (2004) takes issue with Vallduví’s treatment of structures with tails, particularly when focus is either retrieve-substitute focus, or narrow focus. She claims that all are Focus-Background structures, which in her terms is FocusGround, viewing the distinction between link and tail as irrelevant in these cases.

Syntactic structures in context   

   77

She argues that when one has an instance of narrow focus, Vallduví’s distinction does not take into account that the rest of a sentence is necessarily part of the background (Vallduví’s ground). Kiss (1998) bases her analysis on Hungarian, proposing two types of focus: Identificational focus and Information focus. By her analysis, identificational focus possesses syntactic and semantic properties lacking in information focus sentences. The following are the basics of her proposal for focus types, as presented in Casielles (2004). (90)

Information Focus a. merely marks the non-presupposed nature of the information b. allows for any type of phrase c.

does not take any scope

d. does not involve any movement e. (91)

can be either smaller or larger (i.e. it can project)

Identificational Focus a. expresses exhaustive information b. does not allow for all kinds of phrases (excluding universal quantifiers, also-phrases, and even-phrases) c.

takes scope

d. moves to the specifier of a functional projection e.

is always coextensive with an XP available for operator movement (does not project), although it can be iterated

Casielles (2004) notes a similarity in this division with Rochemont’s Presentational vs. Contrastive Focus, and Vallduví’s Retrieve-add vs. Retrieve-substitute system, both of which (ignoring some crucial differences) essentially draw a line between narrow and wide focus. She also points out that Spanish allows for universal quantifiers (92) and even-phrases (93) in identificational focus contexts. (92)

TODOS LOS SOMBREROS quería llevarse la niña. all the hats want.IMPFV.3SG take.INF-SE.3SG the girl ‘The girl wanted to take ALL THE HATS.’

78   

(93)

   The interaction between syntax and information structure

HASTA UN SOMBRERO quería until a hat want.IMPFV.3SG ‘She wanted EVEN A HAT.’

llevarse. take.INF-SE.3SG

As previously stated, Kiss's characterizations of focus are based on Hungarian, which possesses important differences from other languages with respect to focus. Although (92) and (93) bear similarities to identificational focus in Hungarian, they do not express the same propositions: for example, hasta un sombrero in (93) is hardly exhaustive (i.e. she bought a number of other things, including a hat), while it is necessarily exhaustive in Hungarian. Furthermore, identificational focus in Spanish, English, and Catalan does not allow for iteration, thus prohibiting multiple narrow foci, a structure allowed in Hungarian. Information focus (what I will later refer to as narrow focus) does not require movement, either in Spanish or English (or in Hungarian, for that matter); however, Spanish is more restrictive with respect to narrow focus (95) than English (94). (94)

What did Mary buy her sister? a. She bought a HAT for her sister. b. She bought her sister a HAT.

(95)

Qué le compró María what CL.DAT.3SG buy.PST.3SG Maria ‘What did Maria buy her sister?’ a.

a su hermana? to her sister

#?Le compró un SOMBRERO a su hermana.  CL.DAT.3SG buy.PST.3SG a hat to her sister  (no una mochila)  not a backpack

b. #Le compró un sombrero  CL.DAT.3SG buy.PST.3SG a hat c.

a su hermana. to her sister

Le compró a su hermana un sombrero. CL.DAT.3SG buy.PST.3SG to her sister a hat ‘She bought her sister a hat.’

In English (94), narrow focus may or may not involve movement since emphatic stress is used to identify narrow-focused information. Although the Spanish examples may involve movement (e.g. prosodic movement), this movement is distinct from the type of movement typical of identificational focus in Hungarian. As previously discussed in examples (77) and (78), Spanish has been

Syntactic structures in context   

   79

claimed (by e.g. Zubizarreta (1998)) to not allow emphatic stress (95a) without forcing a contrastive focus interpretation. Assuming that nuclear stress falls on the rightmost edge (as in e.g. Zubizarreta (1998)), (95b) inappropriately places narrow focus on her sister, and is therefore infelicitous in this question context. Only when the informational focus coincides with the rightmost edge (95c) does a felicitous reply result. Casielles points out that the fact that Identificational focus takes scope may be related to its position, and not necessarily to the fact that it is narrow focus. She also suggests that wide and narrow foci are not necessarily different types of focus. For Lambrecht (1994), topic and focus do not form a dichotomy; rather, they are separate relations. For Lambrecht, topic has to do with the aboutness of a proposition, while focus has to do with the conveying of new information (his Pragmatic Assertion). For him, all declarative sentences convey information: therefore, all declaratives have a focus, but not all have topics. Focus is information that is added to, not superimposed upon, a pragmatic presupposition: “The focus is, therefore, the element of information whereby the presupposition and the assertion differ from each other... It is the unpredictable element in the utterance”(op. cit.:158–159). The types of focus functions in his analysis are Predicate-Focus (96a), Argument-Focus (96b), and Sentence-Focus (96c). These focus types correspond to the sentence types Topic-comment, Identificational, and Event-reporting, respectively, as noted below (here caps indicate prosodic focus, not contrast). (96)

Context: What did the children do next? a. The children went to SCHOOL. Predicate-focus/Topic-comment Context: Who went to school? b. The CHILDREN went to school.

Argument-focus/Identificational

Context: What happened? c. My CAR broke down.

Sentence focus/Event-reporting

In predicate-focus sentences, the predicate forms the focus. In Argument-focus sentences, the focus is the missing argument. In Sentence-focus sentences, the focus includes the subject and the predicate. Casielles (2004) disagrees with Lambrecht’s characterization of Sentence-focus, preferring to incorporate this sentence type with his Predicate-focus type, as they are both wide-focus in nature. She further disagrees with Lambrecht’s proposal that Sentence-Focus lacks a topic, in line with Erteschik-Shir’s (1997) “here and now” stage-topic (discussed further below).

80   

   The interaction between syntax and information structure

Reinhart (2006) proposes that focus is coded in the phonological component (PF). She suggests that the identification of a focus unit may be determined for each derivation via a set of possible pragmatic assertions (PPA). In the case of foci, at the point where syntax and stress are visible, at the interface between syntax and pragmatics, a reference set of possible foci are generated. At this point the discourse selects the member appropriate to the given context. Reinhart formalizes this proposal (97) for a stressed object in English (bold-face in (98a, b) indicates a prosodically stressed constituent). While in theory any of the members of the set in (98c) may be chosen as the focus of the utterance, at the interface, only one may be chosen. At that point, the relevant discourse conditions will choose which set(s) are appropriate. (97)

Focus Set The focus set of a derivation D includes all and only the constituents that contain the main stress of D.

(98)

a.

[IP Subject [VP V Object]]

b. [IP Subject [VP Object V]] c.

Focus set: {IP, VP, Object}

Given the difficulties in arriving at a consensus on the meaning of focus, I follow López (2009) in defining regular focus as in Jackendoff (1972): focus resolves a variable left open in the previous discourse. López provides the following example (p. 28, ex. 31). (99)

What did John bring? [x | John brought x] John brought the wine. [x=the wine, ‘the wine’ is focus]

The initial discourse in (99) creates and leaves open the variable x, which requires resolution. Therefore, the part of the question that resolves this variable (=the wine) is the focus/rheme of the sentence reply. In the following chapters, I frequently make use of the term narrow focus, which I use when either the subject or an object is the unresolved variable. In many Romance languages a focused element can be displaced to the front of a clause or sentence, as in the Catalan example ((100), from López 2009: 28, ex. 32)).

Syntactic structures in context   

(100) [Context: You gave him the spoons.] ELS GANIVETS li vaig the knives CL.DAT.3SG PST.1SG ‘THE KNIVES I gave him.’

   81

donar. give

In (100), the context does not leave a variable open to be resolved. Rather, focus fronting (FF) then creates this variable (λx you gave him/her x), which in turn opens up the set {x | x=things I may give him/her}. At the same time, FF provides this value for x (=the knives), thus creating a contrast with the immediately preceding context. The interpretive import of focus fronting (FF) then is contrastive, and it may not answer a wh- question, either explicit or implicit. Crucially, this definition of contrast departs from other definitions of contrast discussed earlier in this chapter. For López, the difference between contrastive focus in FF and regular (information) focus, or rheme, is the type of discourse that each may felicitously integrate with. López proposes that regular focus always occurs in situ, while contrastive focus is always fronted.⁸³ This goes against the standard assumption in Romance linguistics that in situ focus may be contrastive. López notes that FF differs from CLLD in that it may not be doubled by a resumptive clitic. Additionally, FF may not be followed by an emphatic particle like sí (que) (101b) while CLLD may (101a).⁸⁴ (101) a.

Las judías sí me las he comido. the beans yes SE.1SG CL.ACC.3PL.F have.PRS.1SG eat.PTCP ‘The beans indeed I have eaten them all.’

b. *LAS JUDÍAS sí me he  the beans yes SE.1SG have.PRS.1SG *’THE BEANS indeed I have eaten them all.’

comido. eat.PTCP

With respect to interpretation, López notes that CLLD is necessarily linked to an antecedent in the discourse, while FF is not. Both FF and CLLD open a variable and close it, however CLLD differs in that it must open a variable that stands in a (minus transitive) poset relation with its antecedent. Consider the following example (53 in López, 2009: 37).

83 See e.g. Zubizarreta (1998), Brunetti (2007) for more on this possibility. 84 See Arregi (2003), Garrett (2013) for more discussion related to this test. Although this example appears for Spanish, (López 2009) shows that the same test also holds for Catalan.

82   

   The interaction between syntax and information structure

(102) [Context: Joan brought the furniture.] a.

La LLET va portar, res mes. the milk PST.3SG bring.INF nothing more ‘He brought the MILK, nothing else.’

b. #La llet, la  the milk CL.ACC.3SG.F  ‘The milk, he brought…’ c.

va PST.3SG

portar... bring

Les cadires les va portar el Joan, pero les taules... the chairs CL.ACC.3PL PST.3SG bring the Joan but the tables ‘Joan brought the chairs, but the tables...’

In (102), FF in (102a) opens up the predicate [x | Joan brought x], whereas CLLD in (102b) opens up the more restrictive [x | xR{furniture} & Joan brought x], in which R refers to a relation between x and its antecedent. López notes that this notion of a poset relation for CLLD is not quite exact because a poset relation is transitive. The relation between an antecedent and anaphor, however, is not transitive. Therefore a (minus transitive) poset relation would rule out (103). (103) [Context: What did you do with the furniture?] #Les potes de les cadires les vaig deixar al  the legs of the chairs CL.ACC.3PL PST.3SG leave to_the  magatzem.  storage area  ‘The legs of the chairs I left in the storage area.’ In (103), the chairs belong to the set of furniture, and chair legs forms part of chairs, but chair legs do not belong to the set of items of furniture. Therefore, furniture cannot be the antecedent of chair legs. López notes that this limitation is crucial because without it, any number of potential anaphoric relationships could be conceived such that any notion of discourse coherence would be lost altogether. I discuss in greater detail the framework that López adopts for determining felicitous discourse relations for CLLD elements in section 2.9.3. In this section, I have provided some examples of the various notions of focus that have been presented in the literature. Although this review has been far from exhaustive, I have discussed some of the important issues that must be taken into consideration. The notion of focus that I adopt in this monograph follows the very basic notion assumed by López: regular focus provides resolution of a variable left open in the preceding discourse. The constituent providing

Syntactic accounts of the syntax-information structure interface   

   83

resolution for this variable is also referred to as the rheme (especially in López 2009). When I use the terms subject narrow-focus or object narrow-focus, I am referring to a particular variable in the discourse that has been opened in the preceding question and then is resolved in the reply. Furthermore, I adopt a view of focus such that focus fronting should be distinguished from regular focus in that contrastive focus fronting simultaneously opens a variable and resolves it, but without the anaphor-antecedent restrictions proposed to exist for CLLD. Note that the experimental methodology that I describe in Chapter 3 and report on in Chapter 4 do not make use of focus fronting. However, I return to a discussion of focus fronting in my syntactic analysis of cliticization and recomplementation in Galician in Chapter 5. In the following section, I discuss some of the chiefly syntactic accounts of the interaction between syntax and information structure in the literature.

2.9 Syntactic accounts of the syntax-information structure interface A number of researchers have described the syntax-information structure relation in purely syntactic terms. As we saw earlier in this chapter, Costa (2004) provides some basic generalizations for clausal word orders in EP as they relate to information structure. According to his analysis, SVO word order with transitive verbs in EP may appropriately answer sentence-focus (i.e. all-focus or thetic-evoking) questions (104a), VP-focus questions (104b), or object-focus questions (104c). (104) a.

What happened?

b. What did Paulo do? c.

What did Paulo break?

VSO word order is appropriate for answering object-focus questions or subjectand object-focus questions (e.g. Who broke what?). VOS word order is claimed to be appropriate only for subject-focus in EP. By Costa’s analysis, VOS involves object-scrambling over the subject, presumably to a vP-adjoined outer specifier position since he maintains that postverbal subjects in VOS remain in situ, just as they do in VSO sentences. I summarize the above information below in Table  11, along with the subject positions that Costa proposes. Crucially, there is no indication of what word orders are preferred in EP. Clearly, SVO or VSO are both appropriate for object-focus sentences, but there are no indications of which word order(s) are preferred in these situations, either.

84   

   The interaction between syntax and information structure

Table 11: Summary of EP word orders and subject positions according to information structure

word order information status

syntactic position of subject

SVO

Spec, IP

sentence-focus VP-focus object-focus subject- & object-focus object-focus (only) subject-focus

VSO VOS

Spec, VP Spec, VP

Source: Costa, João. 2004. Subject Positions and Interfaces: The Case of European Portuguese. Berlin & New York: Mouton de Gruyter.

But what of structures involving left-peripheral elements? Costa uses the term old information to refer to topics. According to Costa, old information “has to be either topicalized or defocused”. He uses the term defocused in a similar manner to Zubizarreta’s (1998) [-F]-marking, which provides a likely explanation for preverbal subjects in SVO in contexts that evoke a predicate-focus or object-focus reply. However, his use of topicalization merits brief mention. Costa refers to (105a) as an example of topicalization, and (105b) as clitic-left dislocation. (105) a.

O bolo, o Pedro the cake the Pedro

comeu. eat.PST.3SG

b. O bolo, o Pedro comeu-o. the cake the Pedro eat.PST.3SG-CL.ACC.3SG.M ‘The cake Pedro ate (it).’ Although example (105a) might appear more similar to focus-fronting of the type found in Italian and Spanish given that it lacks a resumptive clitic, it appears that it is more similar to the type of topicalization found in English (see e.g. Villalba 2000 for an analysis of English topicalization). According to Costa, EP lacks focus fronting of the type found in Spanish. Despite his claim that such focus preposing is ungrammatical in EP (106a), in other examples (106b), Costa presents contexts which would seem to be comparable to exactly the type of preposing found in Spanish. (106) a.  

*ESSE LIVRO, o João  that book the João ‘THAT BOOK João read.’

leu. read.PST.3SG

Syntactic accounts of the syntax-information structure interface   

b. O MEU LIVRO, o Paulo leu the my book the Paulo read.PST.3SG ‘Paulo read MY BOOK (not yours).’

(não not

o the

   85

teu) yours

As (106b) lacks a resumptive clitic, it appears to be focus fronting of the type attested in Spanish. There is a great amount of variation and debate as related to focus fronting in EP (see also note 38). Due to this disagreement, I will not discuss if further for now; rather, I focus our attention on the linear orders indicated for EP, which appear to be more or less analogous to those of Spanish. Subjects representing new information (i.e. information focus) appear at the right edge of the clause, which in Costa’s analysis appear in situ (i.e. post-verbally) in Spec, vP, as in the Romance Nuclear Stress Rule (R-NSR) reformulation discussed by Casielles (2004) and Zubizarreta (1998) for Spanish. As previously mentioned in our discussion of proposals for Spanish, the object in VOS is proposed to scramble in EP, presumably adjoining to vP to escape the projection of focus. This appears quite similar to Zubizarreta’s and Casielles’s respective analyses of Spanish. Importantly, however, Costa’s (2004) analysis does not propose how syntax, information structure and phonology should interface. Rizzi (1997) proposed the expanded CP field as an interface layer between the propositional content (IP/TP) and the superordinate structure. For embedded clauses, the superordinate structure is the higher clause; for matrix (root) clauses however, this is the discourse. This proposal incorporates Topic and Focus both as features and as labels of heads and projections in the narrow syntax.⁸⁵ Rizzi proposed the following structure for the left-periphery (107): (107)   FceP  TopP* FocP  TopP*  FinP TP

85 Note that in Rizzi (1997, 2004), focus and topic are referred to as criterial features.

86   

   The interaction between syntax and information structure

By this analysis, the topic projection TopP is a recursive element which may appear prior to or following a focused element. As many languages only allow for one focus element per sentence (Hungarian is a notable exception in this respect), FocP is not afforded an asterisk for recursivity. Topic and Focus projections are only activated when their corresponding features are present in a given numeration as a phonologically null lexical item. When this happens, the feature occupies the head of the projection, and the topicalized or focused element occupies the specifier, as in (108b). By this analysis then, the phonologically null Topic feature is posited to be at bare minimum an interpretable feature, and maximally, as part of the lexicon, and is internally merged in the head of Topic projection that receives the label of the functional feature. (108) a.

The book I gave to John.

b.

TopP  DP

  Top’

The book  Topo   [+Top]  [+Top]

  …

The lexical item the book, which is marked with a [+Top] feature is attracted to Spec, TopP by the corresponding [+Top] feature on the Top head, and thus checks this feature. Therefore, for any proposed left-peripheral Topic or Focus element, two crucial generative assumptions must be made: 1) that the features [+Top] and [+Foc] must exist in the lexicon as phonologically null lexical items, and 2) that some subset of phonologically realized lexical items may be [+Top]or [+Foc]-marked in the Numeration prior to entering the syntactic derivation. An alternative to option 2 is that some number of phonologically realized lexical items also have a corresponding [+Top]- or [+Foc]-marked entry within the lexicon. Theoretically, many would consider this an undesirable scenario since it would triple the lexical learning burden on the part of a child acquiring the language. In Jackendoff (1972), one the earliest generative analyses, F-marking was proposed as an “artificial construct” to account for focused elements. Pollock (1989) proposed the functional projection FocP and a corresponding [+F] feature. However, the existence of this syntactic feature [+F] (Focus) has been challenged in the literature. Despite making use of a [+Focus] feature in her analysis, Zubizarreta (1998) observes that [±F] as a lexical feature is conceptually problematic since it would violate Chomsky’s (1995: Ch. 4) Inclusiveness Condition (see also

Syntactic accounts of the syntax-information structure interface   

   87

Szendrői 2001, 2004).⁸⁶ As mentioned above, however, in the Zubizarreta proposes (109), [F] not as a lexical feature, but as a derived phrase marker, which remains undefined until after Σ-structure. (109) Σ-structure

(sets of phrase markers, feature checking) (unique phrase marker) (F-marking, NSR, FPR, p-movement)

 LF PF

Assertion structure

In this model, phrase markers remain essentially inert at the stage in which features are checked. It is after Σ-structure (and prior to LF and PF) that F-marking, the NSR, the FPR, and p-movement take place in her analysis. Despite the debate surrounding topic and focus features, numerous analyses have made use of them, which I can only assume is for lack of a more explanatorily attractive alternative. Casielles (2004), which I discuss in the next section, is one such analysis.

2.9.1 Casielles (2004) Casielles (2004) examines the information structure dichotomies discussed above (e.g. New–Old information, Topic–Focus, Topic–Comment, Theme–Rheme, etc.), and after thorough analysis, she arrives at two basic dichotomies (110) that become the backbone of her proposal: Sentence Topic (STopic)–Focus and Focus–Background. Casielles draws a division between topic and background based on the following phonological, syntactic, and discourse features.

86 Inclusiveness involves the manners by which a node may acquire a feature – in this case, the discourse feature Focus. Following Chomsky (1995: 228), a non-terminal node inherits features from its daughter, while a terminal node may be assigned a feature from the lexicon. Therefore, the assignment of [+F] features to a constituent would have to happen in the lexicon, and not after it enters the derivation.

88   

   The interaction between syntax and information structure

(110) Sentence Topic (STopic) + single + sentence-initial + referential ± discourse-old ± unaccented

Background ± single ± sentence-initial ± referential + discourse-old + unaccented

Although both STopics and Background elements are topical in Casielles’s analysis, she suggests that STopics could (and perhaps should) be referred to as preverbal subject topics, and Background as wide topics (p. 99, fn. 41). Unlike Background, STopics may be [- discourse-old], suggesting that STopics are present in thetic, “out of the blue” sentences, a categorization which bears similarity to Cardinaletti’s (2004) subject of predication.⁸⁷ A crucial difference between STopics and Background has to do with their syntactic status and the syntactic positions available to them. The pre-verbal subject position to which STopics move in Casielles’s analysis results from a VP-internal subject moving to the pre-verbal specifier position Spec, TP. This movement of a subject DP is motivated by checking of a [+topic] feature. Crucial to her analysis is that Spanish STopics may only be DPs, and not NPs, a distinction motivated by constraints on the distribution of bare nominals, which may only appear post-verbally (cf. 111a and 111b) in Spanish. (111)

a.

Jugaban niños en play.IMPFV.3PL children in

la calle. the street

b. *Niños jugaban en la calle.  children play.IMPFV.3PL in the street  ‘Children were playing in the street.’

87 Casielles admits trouble classifying thetic sentences (Kuroda 1972; Sasse 1987) by the two dichotomies she proposes. She notes that Lambrecht (1994) has a third sentence type for thetic sentences called Event-reporting, by which the whole sentence is focused when answering the question “What happened?”. Another possibility that Casielles considers is that they are STopic-Focus sentences with a null STopic, but she notes that Erteschik-Shir (1997) disagrees with such a notion, and instead posits a “here and now” stage-topic to describe such sentences (which is supposed to correspond with Kratzer’s (1989) spatio-temporal argument). The limit to this possibility, however, is that only stage-level predicates can be stage-topics in such a system (and would thus exclude individual-level predicates – perhaps not a problem since they are not eventive). Casielles also hypothesizes that thetic sentences may be instances of STopic-Comment structures, a structure that does not appear within her classification of sentence types.

Syntactic accounts of the syntax-information structure interface   

   89

Bare nominals in their grammatical, post-verbal position (111a) may only receive an existential reading in Spanish, and not a generic reading, which is more easily exemplified with the use of a present tense verb (112). (112)

Juegan niños (generalmente) en la play.PRS.3PL children generally in the √ ‘(Generally) there are children playing in the street.’ * ‘Children generally play in the street.’

calle. street

In Spanish, a bare nominal cannot receive a generic interpretation, which, by her analysis, is why it may not appear in a preverbal position. To account for this in Spanish, Casielles builds on Diesing’s (1992) mapping hypothesis proposing the Bare Noun Movement Constraint (BNMC), according to which only DPs can escape the VP and move to Spec, TP.⁸⁸ Since bare nominals are not DPs, but rather NPs, they cannot escape the VP to move to preverbal Spec, TP. While bare nominals may be focused (113a), focus movement is proposed to be more akin to wh- movement (following Cinque 1990), which requires an operator, thus distinguishing this sort of movement from DP-movement. Bare nominals may appear pre-verbally as Topics (113b), but these are considered to be clitic leftdislocated (CLLD) elements (bold-face is used in this and subsequent examples for expository purposes). (113)

a.

LANGOSTAS destruyeron las grasshoppers destroy.PST.3PL the ‘GRASSHOPPERS destroyed the crop.’

b. Dinero tengo money have.PRS.1SG ‘Money I have.’

cosechas. crops

yo. I

The main difference then between a topicalized bare nominal in pre-verbal position and a pre-verbal subject has to do with Casielles’ categorization of topicalized information. The dislocated bare nominal in (113b) is Background, or nonfocused information, and crucially not a topic. Pre-verbal subjects are Sentence Topics (STopics), which may not be bare nominals. Since a bare nominal subject

88 Following Diesing (1992) it is in Spec, IP that a DP is mapped into the Restrictor and therefore bound by the Generic Operator, which allows DPs to receive a generic interpretation. Since bare NPs cannot reach the Restrictor in Spanish, they may not receive a generic interpretation.

90   

   The interaction between syntax and information structure

may only appear pre-verbally as a focused or (dislocated) topicalized element, it therefore must not appear in Spec, TP. Other non-DP elements such as a PP (114a), an AP (114b), or a VP (114c) may appear preverbally, and any number of them may appear. (114) a.

De la conferencia no he oído nada. of the conference not have.PRS.1SG hear.PTCP nothing ‘About the conference I haven’t heard anything.’

b. Listo no lo clever not CL.ACC.3SG.M ‘Clever he doesn’t seem.’ c.

parece. seem.PRS.3SG

Estudiando nunca está. study.PROG never be.PRS.3SG ‘Studying he never is.’

Casielles suggests that these non-DP preverbal elements are Background elements which appear in a CLLD position (at times with a null resumptive clitic). Yet it still remains to be explained how such moved elements escape the VP to arrive in their CLLD location. Although Spanish CLLD is sensitive to island constraints (115a), the dislocated BNs in (115b) and (115c, from Rivero 1980) do violate the wh-island constraint. (115)

a.

*A Carlos conozco solo a las personas que  to Carlos know.PRS.1SG only to the people that  le gustan.  CL.DAT.3SG please.PRS.3PL  ‘To Carlos I know only the people that appeal.’

b. Dinero te pregunta (que) por qué no tiene money CL.DAT.2SG ask.PRS.3SG that why not have.PRS.3SG ‘Money she asks you why she doesn’t have.’ c.

Dinero dicen que cree que tiene money say.PRS.3PL that believe.PRS.3SG that have.PRS.3SG ganas de ahorrar desire of save.INF ‘Money, they say that he believes that he has the desire to save.’

Given the BN data above in (115b–c), combined with the fact that BNs may not appear in a preverbal subject position, Casielles suggests that these CLLD items in

Syntactic accounts of the syntax-information structure interface   

   91

(114) and (115) are base-generated, as in Cinque (1990).⁸⁹ An attractive facet of this possibility is that if CLLD items do not involve movement, they are not subject to Casielles’s BNMC, thus causing no problem for her proposal. Although Casielles’s analysis creates an adequate calculus by which focus prominence interacts with information structure to appropriately configure word order, it makes use of a computational system that is dependent on the features [+topic] and [+focus], which, as it has been suggested, violates the Inclusiveness Condition (see note 86). Additionally, while her view of the interface rules out certain word orders (e.g. with preverbal bare NP subjects), few comments are made with respect to the acceptability of word orders according to information structure context. The lone exception to this is where she notes that an SVO sentence (116a), which she classifies as a case of STopic-Focus, does not have any specific discourse requirements. (116) a.

Juan llamó a Juan call.PST.3SG to ‘Juan called his wife.’

su mujer. his wife

b. ¿Qué pasó?  what happen.PST.3SG  ‘What happened?’ c.

¿Qué hizo Juan?  what do.PST.3SG Juan  ‘What did Juan do?’

d. ¿A quién llamó  to who call.PST.3SG  ‘Who did Juan call?’

Juan? Juan

As seen at the start of this chapter, Casielles (2004: 207, ex. 53) claims that (116a) may appropriately answer any of the three questions in (116b–d), but does not comment on the appropriateness of other word orders such as VSO or VOS, nor whether other such word orders are more or less appropriate an answer than SVO to such questions in Spanish. She does, however, point out that BackgroundFocus structures are inappropriate replies to start a discourse, or to reply to a narrow-focus question. Yet the argument could be made that the subject and verb

89 The base-generation vs. movement approach to CLLD is not uncontroversial. Note that the assumption of movement for CLLD items is crucial according to López’s (2009) analysis, which I examine below in section 2.9.3, CLLD is a product of movement.

92   

   The interaction between syntax and information structure

in (116a) are Background elements in a reply to an object-narrow-focus question (116d). The analyses examined thus far largely predate or do not take into account more recent developments in generative syntax. With the advent of multiple spell-out (Uriagereka 1999) and phase theory (Chomsky 2008), the possibility of derivational “pauses” has come about. It is at such pauses, or phase edges that interface relations such as PF-syntax or syntax-pragmatics have been proposed to take place. In the following section, I examine some of the interface phenomena that have been proposed to occur at these points.

2.9.2 The Interface and Phases Parafita Couto (2005) examines the interface of information structure and syntax as it pertains to focus in Galician. Due to the existence of sentences such as (117), she, too, suggests that the [+Focus] feature must exist in the grammar. (117)

a.

Para for as the

TI ires you go.SBJV.FUT entradas ben tickets well

ó partido, tiñan que ser to_the game have.IMPFV.3PL that be.INF baratas. cheap

b. Para ires ó partido TI, tiñan que ser for go.SBJV.FUT to_the game you have.IMPFV.3PL that be.INF as entradas ben baratas. the tickets well cheap ‘For YOU to go to the game, the tickets must have been cheap.’ For Parafita Couto, movement of the type in (117b) is rightward p(rosodic)-movement to a phase edge. In this proposal, each phase edge is the locus for focus encoding. PF and semantics have access to the syntactic module at these phaseedges. Such access is necessary to ensure that the emerging structure meets the demands of the unfolding (umbrella-type) discourse. By her analysis then, phase edges are landing sites for p-moved XPs. This proposal is attractive in that it obeys Chomsky’s (2008) notion of phases, which allows for multiple Spell-out over the course of a given derivation, thus granting PF cyclic access to non-phase-edge material at the end of each syntactic phase.⁹⁰ However, this analysis runs into

90 Note that similar, but unrelated proposals possess interesting similarities in this respect. In Steedman’s (2000) Combinatory Categorial Grammar (CCG) approach to the interface, for exam-

Syntactic accounts of the syntax-information structure interface   

   93

the same limitations and theoretical objections as previous studies that attempt to account for the syntax-discourse interface: positing the existence of a [±focus] feature in the lexicon, a crucial assumption of this proposal. Szendrői (2001, 2004) argues that the inclusion of such pragmatic features in the lexicon violates the Inclusiveness Condition in Chomsky (1995) since, for a [+Focus] feature to be assigned to a constituent in a given Numeration, it would have to be a feature on that lexical item (see also comments in note 86). Szendrői insists that there is no way in which this could be the case, thus suggesting that [+F] is no more than a diacritic inserted to account for characteristics unrelated to a lexical property of a lexical item (see also Brunetti (2004), Emonds (2004) and Reinhart (2006) for critiques along similar lines). She proposes that Focus denotes and encodes an information status relation of constituents relative to the rest of a given utterance, and the same holds for Topic. However, she insists that the encoding of this relation via diacritics or features may not occur in the syntactic computation without violating Inclusiveness. Revithiadou & Spyropoulos (2004, 2009) and Spyropoulos & Revithiadou (2007) focus on reconciling phase theory and Multiple Spell-Out (Uriagereka 1999; see also Kratzer & Selkirk 2007 for a similar prosody-syntax interface treatment). They examine left-peripheral clitic-doubled objects and preverbal subjects in Greek, showing that prosodic islands match syntactic islands in the case of clitic-doubled objects, thus suggesting a natural syntax-prosody interface point. Crucially, however, this island correspondence does not hold for subjects. Therefore, they propose that clitic-doubled objects are part of what Uriagereka terms separate derivational cascades, assembled and spelled-out before they reach the main derivational cascade. Curiously, preverbal subjects in Greek may be extracted from and are susceptible to prosodic restructuring, facts which motivate their proposal that preverbal subjects in Greek form part of the main syntactic and prosodic derivation. The preceding interface analyses share in the intuition that some sort of syntactic interface coincides with phase edges. This concept is also central to López’s (2009) analysis of the syntax-pragmatics interface in Spanish and Catalan. A crucial difference to his proposal, however, lies in the model of the grammar that he proposes, which makes more concrete (and, as we will see, testable) predictions regarding grammaticality and acceptability.

ple, intonational boundaries coincide with major syntactic boundaries (see also Selkirk 1990 for a similar approach). Within this particular framework, surface structure, information structure, and intonation coincide within a given clause. Much recent work in the Cartographic Program (e.g. Frascarelli 2004, Frascarelli & Hinterhölzl 2007) also assumes a correlation between information structure, intonation, and syntactic projection position.

94   

   The interaction between syntax and information structure

2.9.3 López’s (2009) interface model López (2009) proposes that discourse relations are determined by their syntactic configuration. Within this model of the grammar (118), information structure functions are assigned by a module called pragmatics, which “inspects” the syntactic structure at each phase end and assigns pragmatic values to constituents in certain syntactic positions. (118)   LEXICON   ⇩ CHL  ∑  PRAGMATICS   ⇩  ∑[p]   ⇩ DISCOURSE According to his proposal, pragmatic values may only be altered within the boundaries of the phase; otherwise, they are unaffected by further syntactic movement. Once assigned, a pragmatic value stays with a constituent as it continues to move (or not) throughout the derivation of the clause. The resulting values of a given constituent at Spell-out (i.e. phase end) dictate its discourse-pragmatic interpretation. By this analysis, which centers on Catalan, the pragmatics module assigns two basic interpretive values: (discourse) anaphoricity and contrastiveness. An attractive facet of this proposal is that these pragmatic features [±a] (anaphoric) and [±c] (contrastive) are not assigned to lexical elements in the numeration as they enter the derivation, which would lead to the objections raised earlier based on Inclusiveness; rather, these features are assigned derivationally as the pragmatics module “reads” the output from the syntactic module. Therefore, constituents appearing in certain structural positions at phase-end get assigned interpretive pragmatic features [±pf]. The possible combinations of these pragmatic features determine the discourse function of a constituent. For López, they are the following, as in Table 12.⁹¹

91 Note that in this table, López uses the term rheme. He uses this term interchangeably to refer to regular focus, a term that encompasses narrow-focus. Recall that, for López, regular focus differs from contrastive focus, primarily in syntactic terms. Contrastive focus is fronted, while regular focus occurs in situ.

Syntactic accounts of the syntax-information structure interface   

   95

Table 12: Interpretive values assigned by Pragmatics module

+c

-c

+a CLLD CLRD -a FF Rheme Source: López, Luis. 2009. A Derivational Syntax for Information Structure. Oxford: Oxford University Press.

Assignment of the [+a] feature is assigned to a clitic X, which López assumes to already be in a feature-dependency Agree relation with the verb prior to phaseend.⁹² I take a step backward here briefly to describe the Agree relations at play in (119) prior to assignment of pragmatic features. (119)

  vP XP

v’ Subj

 v’ v

X[uf]

 VP

 v[f][uφ] XP[uf]   V’

 Agree  Agree The feature [f] on v is proposed to be akin to Case, and is valued by the clitic X. The object XP then does not have its [uf] satisfied yet. Following merge of the external argument, the remaining unvalued feature on the object XP triggers movement of the object XP to the outer Spec of vP, which allows it to have its features checked/ valued.⁹³ This in turn creates a local dependency between the clitic and verbal argument (object). This dependency relation is crucial with respect to the assign-

92 For López, the clitic X is a feature matrix that merges early with v (see also Bonet 1995). This feature matrix later gets spelled out as the clitic in the Morphology module. Note that if the clitic X adjoins to the verb as low as little v, it will have to excorporate following v-to-T movement in order to net proclisis. 93 Note that López assumes feature checking to be a very local process that may only occur within the c-command domain of the probe (i.e. the feature that requires checking). This assumption is crucial in motivating the movement (by Attract) of the doubled XP to (outer) Spec, vP.

96   

   The interaction between syntax and information structure

ment of [+a] features. When [+a] features are assigned to the (anaphoric) clitic X by the pragmatic module at the end of the vP phase, Spec, vP also becomes [+a], as in (120). (120)

  vP XP[+a]

v’

Subj

 v’  v+X[+a]

  VP[−a]

Agree The VP complement of X is then assigned [-a], which matches well with the fact that information focus elements are non-anaphoric. Note that the Agree relationship between the clitic X and the clitic double is crucial for López’s proposal, as elements that do not enter into such a relationship with the clitic X (e.g. fronted focus, which does not have a clitic double) cannot be marked as [+a] by the pragmatics module. On the one hand, this prevents a number of constituents (e.g. the external argument, elements that will be focus fronted, and nonD-linked phrases) that also stop in Spec, vP on their way to higher positions, from being marked with the [+a] feature; on the other hand, it does not prevent the complement of Spec, vP from being marked [-a].⁹⁴ Constituents that are [+a]marked then are peripheral elements, which either remain in Spec, vP for CLRD, or CLLD elements that later move higher in the structure for another interpretation. The latter constituents are those that move to Spec, FinP, the edge of the second phase, and are assigned [+c] by the pragmatics module at phase-end. Let us briefly examine how [±a]-assignment would work in Galician with contexts that do not involve structures in the higher, left-peripheral realm, and the sort of pragmatic predictions that it makes.⁹⁵ Consider SVO (121a), VSO (121b) and VOS (121c) word orders.

94 D-linked wh-determiners (e.g. which in English, quin in Catalan) are inherently [+a] by López’s analysis (i.e. they enter the derivation already so marked). Therefore, they are not susceptible to the rules on [±a] assignment. 95 Note that this is one of the motivations behind CLLD involving movement in López’s analysis.

Syntactic accounts of the syntax-information structure interface   

(121)

   97

[Context: What did John eat?] a.

[TP (Xoán) [T’comeu [vP [v’ [VP Xoán eat.PST.3SG ‘Xoán ate an apple.’

unha an

mazá[-a] ...]]]]] apple

b. [T’comeu [vP (Xoán) [v’ [VP unha mazá [-a] ...]]]]] c.

# [T’comeu [vP unha mazá [v’ Xoán [-a] [VP...]]]]]

For the context provided in (121), both (121a) and (121b) are felicitous replies (with or without the overt subject DP Xoán) to the object narrow-focus question since, in both of these, unha mazá as lowest complement of the initial phase head, gets marked [-a] for regular focus/rheme at phase-end. In (121a), the external argument is not in an Agree relation with a clitic prior to moving on to Spec, TP, and therefore is unaffected by [+a]-marking at the end of the vP phase. The same applies in (121b), but Xoán does not continue to move higher. Because the direct object has moved to Spec, vP in (121c), Xoán is marked [-a], thus correctly predicting its infelicitousness for this context. However, if we alter the question context to Who ate an apple? then only (121c) is appropriate, as it is the only configuration in which Xoán is marked [-a]. López’s analysis runs into complications for all-focus, or thetic sentences because in an out-of-the-blue sort of context, the whole sentence should be marked [-a], or non-anaphoric. To deal with this outcome, he suggests that subjects can also bear an additional feature, which he calls [ud]. The interpretable counterpart of this feature is [d], which appears on Fin. Unvalued φ-features on Fin allow it to probe and trigger movement of the subject DP to Spec, TP. This portion of the proposal is somewhat problematic, namely due to the [d] feature that he proposes to initiate a new discourse. If [d] is a discourse feature like [±a] and [±c], it is unclear why this particular feature would be purely syntactic and not be involved with the pragmatics module. I return to this issue and its importance for the analysis of Galician that I propose in Chapter 5. For the moment, I turn to assignment of the [c] pragmatic feature. As a point of departure for [±c]-assignment, let us examine the structure that López proposes for the left periphery. Since he does not assume topic and focus as lexical or discourse features, his simplified left periphery consists of only Force and Fin (122) projections.

98   

   The interaction between syntax and information structure

(122)

ForceP Force

FinP

CLLD/FF/wh-    Fin’   Fin

  TP

This left peripheral syntax posits that wh-phrases, focus phrases and clitic-dislocated phrases occupy recursive specifier positions of FinP. When more than one of these is present they appear as stacked specifiers of FinP. Notably, preverbal subjects do not appear in Spec, FinP, but rather in Spec, TP. Moving on, I now examine how [+c] assignment is proposed to work if we modify our context from (121) as in (123). (123) [Context: Who ate the apple?] a.

A mazá comeuna Xoán. the apple eat.PST.3SG Xoán ‘The apple, John ate.’

b. [FinP a mazá[+a,+c] [Fin’[TP[T’ comeuna [vP [v’ Xoán [-a] [VP...]]]]]

In a CLLD reply (123a) to the context in (123), a mazá first moves to Spec, vP, after which Pragmatics marks it [+a], owing to its agree relation with the clitic a (the epenthetic consonant /n/ is the phonological result of the clitic attaching to a diphthong), also marked [+a], while still in v. The [+a]-marked direct object a mazá then later moves to Spec, FinP (123b). At the end of the phase, the pragmatics module marks the constituent in Spec, FinP as [+c], while the complement of FinP (i.e. the remaining structure) gets marked [-c]. Fronted focus elements are proposed to make the same movement steps as CLLD save for [+a] marking in Spec, vP, which does not apply due to the lack of a clitic-double dependency. A particularly attractive aspect of this proposal is that it examines in detail how to determine the appropriateness of CLLD elements.⁹⁶ López proposes that

96 See López (2009) for an in-depth discussion of the mechanisms responsible for CLRD elements. As CLRD phenomena are orthogonal to this monograph, I omit a review of this portion of his proposal.

Syntactic accounts of the syntax-information structure interface   

   99

CLLD requires a particular type of discourse relation in order to be felicitous and appropriate. Analyses of discourse generally assume a hierarchical structure for discourse (see e.g. Hobbs 1985, Polanyi 1988, Grosz & Sidner 1986, Mann & Thompson 1987, Asher 1993, van Kuppevelt 1995). Asher & Vieu (2005), working in the Segmented Discourse Representation Theory (SDRT) framework, distinguish between two particular types of discourse structure relations: coordination and subordination. The two are summarized as in (124). (124) Coordination: narration, background, result, continuation, parallel, contrast, question-coordination, correction Subordination: elaboration, instance, topic, explanation, precondition, commentary, question-answer pairs The relations in (124) describe particular types of structural relations possible for a given proposition in relation to the preceding discourse. In the case of thetic (Kuroda 1972), or out of the blue (Zubizarreta 1998) sentence contexts, which lack a strict connection or relation to a preceding discourse, I assume that these are not strictly categorized as either subordinating or coordinating, since such relations are always defined relative to the initial lead sentence in Asher & Vieu’s (2005) analysis.⁹⁷ According to these relations, both narrow- and wide-focus (López’s regular focus or rheme) question-answer pairs qualify as subordination discourse relations. López adopts Jackendoff’s (1972) definition of a focus reply as “provid[ing] a resolution for a variable left open in the preceding discourse” (López 2009: 28), yet follows Vallduvì and Vilkuna (1998) in using the terms regular focus (narrow focus), and rheme interchangeably. Foci expressing contrast, however, qualify as coordinating discourse relations for López, thus differentiating regular focus from contrastive (i.e. fronted) focus. Given that topic(al) has been used to refer to aboutness (e.g. Lambrecht 1994), discourse-old information (e.g. Erteschik-Shir 1997), or discourse-active elements (e.g. Halliday 1967; Sgall et al. 1986), and other theme-related notions as previously discussed in section 2.8.3, López proposes a departure from these notoriously problematic terms in his approach, assuming instead that topics are simple constituents that describe ce dont on parle, or “what one is speaking about”, following Busquets et al. (2001). Although López follows Asher & Vieu (2005) in considering topic to be a subordinating discourse relation, it is important to point

97 Note that while thetic, or “out of the blue” contexts can answer the question “What happened?”, they may also initiate a discourse lacking a question eliciting such a context. Asher & Vieu (2005) provide only one example with such a question, and in that example, they define discourse relations starting with the first sentence of the reply.

100   

   The interaction between syntax and information structure

out that the term topic within the SDRT framework differs from its counterpart in the syntactic literature in that it “bears a particular relation to [a] constituent [in a segmented discourse representation structure]” (Asher 1993: 267). López proposes that CLLD, which is typical of (syntactically) topical constituents, requires discourse subordination in addition to a discourse antecedent in the superordinate sentence in order to be felicitous and appropriate. CLLD items are assigned [+a], [+c] by Pragmatics in this framework, and as anaphoric elements, also require a discourse antecedent in the superordinate sentence in order to be felicitous and appropriate. Possible discourse antecedent relations have the following characteristics (125) in López’s proposal (following observations in Villalba 2000): (125) Characteristics of antecedent discourse relations (López 2009) 1.

the antecedent relation may involve strict identity, or, more commonly, a partial identity (e.g. set-subset, part-whole) relation exists between antecedent and CLLD item,

2.

immediate discourse adjacency is not required, and

3.

there is symmetry of contrast between the CLLD item and a similarlyrelated item

Consider the continuations of the context established in (126a, b) in Catalan from López (2009: 41), which establishes a coordinating discourse relation for the following sentences. (126) a.

Joan brought the furniture in a truck.

b. He opened the truck and... c.

#la taula de fòrmica, la va portar a  the table of formica CL.ACC.3SG.F AUX.PST carry.INF to  la cuina.  the kitchen  ‘...the formica table, he carried to the kitchen.’

d. va portar la taula de fòrmica, a la cuina. AUX.PST carry.INF the table of formica to the kitchen ‘...carried the formica table to the kitchen.’ The narration relation created by (126b) relative to (126a) is an example of a coordination discourse relation between sentences. In both (126c) and (126d), there is

Syntactic accounts of the syntax-information structure interface   

   101

an appropriate antecedent for the CLLD phrase the formica table, found in the furniture in (126a) (a formica table is an article of furniture following (125)), but the CLLD sentence (126c) is infelicitous because it continues the discourse initiated in (126b). As continuation is a coordination relation, la taula de fòrmica may not appear as a CLLD element in (126c), but may appear felicitously in situ in (126d). However, if the proper subordination discourse context is created (127, also from López 2009: 41), CLLD is appropriate. (127) a.

Joan brought the furniture in a truck.

b. He opened the truck and started carrying them into the house. c.

La taula de fòrmica, la va portar the table of formica CL.ACC.3SG.F AUX.PST.3SG carry.INF a la cuina... to the kitchen ‘The formica table, he took to the kitchen...’

Sentence (127b) continues the narration begun in (127a), and (127c) elaborates on the discourse continued in (127b). Following (124), elaboration is a subordinating discourse relation, which according to López, creates an appropriate context for CLLD in (126c). As in (125), the formica table in (127c) also has an appropriate anaphoric discourse antecedent, since it is a subset of (non-adjacent) the furniture in (127a). For López, CLLD is most naturally found when symmetrically contrasted with another similarly antecedent-related CLLD element. Therefore, (127c) could be followed with a sentence like and the chairs, he left in the front room, but crucially not with #and the fish, he put in the refrigerator, since fish do not form a subset of possible articles of furniture. Example (127c) could likely stand alone in a given discourse; however, its utterance would create an expectation of implicit contrast with other similar elements relevant in the discourse. The antecedent for CLLD need not be immediately adjacent, but must be found within the preceding discourse in the appropriate subordinating relation, as we see in (128). (128) a.

John brought in the food.

b. He put the fish in the fridge. c.

It was an excellent fish

d. and he didn’t want it to go bad.

102   

   The interaction between syntax and information structure

e.

De la carn, se’n va oblidar of the meat CL.REFL.3SG-CL.PART AUX.PST forget.INF completament. completely ‘He completely forgot about the meat.’

Above, (128c) and (128d) are subordinate to (128b), but this does not block the connection between the meat in (128e) and its antecedent the food in (128a). In (129) the relations between the propositions (Π) in (128) are represented graphically. Coordinating relations are horizontal lines, and subordinating relations are angled vertical lines. (129)

ΠaAntec Πb  Πc

 ΠeCLLD

  Πd

In (128e), CLLD is appropriate, and thus unaffected by the non-adjacency of propositions (a) and (e). López presents another example (130) to additionally illustrate that discourse structure is more important than adjacency. (130) a.

John brought in the food.

b. He put the fish in the fridge c.

and went out to the terrace for a second to smoke.

d. Then he spoke to Theresa on the cell for a little while. e’. De la carn, se’n of the meat CL.REFL.3SG-CL.PART completament. completely ‘He completely forgot about the meat.’

va oblidar AUX.PST forget.INF

e’’. #a l’entrar, la carn la va posar  to the-enter.INF the meat CL.ACC.3SG.F AUX.PST put.INF  al congelador  to_the freezer  ‘When he came in, he put the meat in the freezer.’

Syntactic accounts of the syntax-information structure interface   

   103

Following López’s analysis, (130e’) is able to create a symmetrical contrast between the fish and the meat, irrespective of (130c) and (130d) which intervene. In (130e’’), beginning with when he came in makes it the next in a series of events that started with (130b). This sequence of events makes (130e’’) a coordination context, which is not amenable to CLLD of la carn. Examples (131) and (132) graphically represent the two discourse progressions from above. (131)

  ΠaAntec Πb → Πc → Πd

Πe’ CLLD

 contrast (132) ΠaAntec   Πb → Πc  → Πd → Πe’’ In (132), no subordination relation can be made between proposition (a) and (e’’), which rules out CLLD according to his analysis. The coordinating discourse relation result serves the function of consequence, as in (133), modified for Galician from López (2009: 40). (133) a.

Xoán esqueceu coller as chaves Xoán forget.PST.3SG grab.INF the keys ‘Xoán forgot the keys.’

b. e como resultado non as tiña and as result not CL.ACC.3PL.F have.IMPFV.3SG cando chegamos ao aparcadoiro. when arrive.PST.1PL to_the parking garage ‘and as result, he didn’t have them when we arrived at the parking garage.’ b’. #e como resultado, as chaves non as  and as result the keys not CL.ACC.3PL.F  tiña cando chegamos ao aparcadoiro.  have.IMPFV.3SG when arrive.PST.1PL to_the parking garage  ‘and as result, the keys, he didn’t have them when we arrived at the parking garage.’

104   

   The interaction between syntax and information structure

Following López, sentences like (133a) and (133b) are coordinated in the sense that (133b) is the (conjoined) consequence of (133a). (133b’) is infelicitous because the CLLD object as chaves does not have an appropriate antecedent in the superordinate discourse (i.e. prior to the utterance of (133a)). Therefore, the lack of discourse subordination in the immediately preceeding conjunct results in a repetitive reading of as chaves, and thus, incoherence. Narration and continuation discourse relations are very similar to each other in that they both continue a narrative. Narration involves an order of events (134), while continuation continues the discourse with a conjoined clause (135) (both examples modified for Galician from López 2009: 40). (134) a.

Xoán levou os mobles ao novo piso e Xoán take.PST.3SG the furniture to_the new apartment and meteunos dentro. put-PST.3SG-CL.ACC.3PL.M inside ‘Xoán took the furniture to the new apartment and took it inside.’

b. #Despois, os mobles, limpounos para  after the furniture clean.PST.3SG-CL.ACC.3PL.M for  quitarlles o po.  remove.INF-CL.DAT.3PL the dust  ‘Afterward, the furniture, he cleaned to get the dust off (it).’ b.’ #Despois, as cadeiras, limpounos para  after the chairs clean.PST.3SG-CL.ACC.3PL.M for  quitarlles o po.  remove.INF-CL.DAT.3PL the dust  ‘Afterward, the chairs, he cleaned to get the dust off (them).’ b.’’ Despois, limpounos para quitarlles o po. ‘Afterward, he cleaned it to get the dust off .’ (135) a.

O can mordeulle the dog bite.PST.3SG-CL.DAT.3SG ‘The dog bit Silvia’s hand…’

a the

man hand

a to

Silvia Silvia

b. e o gato esgaduñoulla. and the cat scratch.PST.3SG-CL.DAT.3SG-CL.ACC.3SG.F ‘…and the cat scratched it.’

Establishing clausal structure in Galician   

   105

b’. #e a man o gato esgaduñoulla.  and the hand the cat scratch.PST.3SG-CL.DAT.3SG-CL.ACC.3SG.F  ‘…and, her hand, the cat scratched it.’ b’’. #??e o matapiollos o gato  and the thumb the cat  esgaduñoullo.  scratch.PST.3SG-CL.DAT.3SG-CL.ACC.3SG.M  ‘…and, her thumb, the cat scratched it.’ As in (133b’), the ensuing sentences in (134b’) and (135b’, 135b’’) are infelicitous because, in coordinating discourse relations, the CLLD direct objects do not have an anaphoric antecedent in the superordinate discourse. In other words, the lack of discourse subordination is what creates a repetitive, infelicitous reading for the CLLD direct object XP. López’s proposal then provides not only a mechanism by which the pragmatic module interfaces with the syntax in order to assign discourse features to a variety of sentence elements, but also provides metrics for predicting discourse appropriate clitic-left dislocations. The calculus López proposes was crucial in designing discourse contexts to test for discourse appropriateness in Galician. As this proposal informs the methodology that I discuss in Chapter 3, the results that I gather will help to determine the preferred, appropriate word orders for Galician and also test the adequacy of López’s analysis for the Galician language.

2.10 Establishing clausal structure in Galician As previously mentioned, works related to syntax interfaces involving pragmatics, information structure, and or discourse in Galician are in extremely short supply. Freixeiro Mato (2006a, b) is one of the few extant works that makes reference to information structure notions such as theme and rheme, but no reference is made to word orders that are more or less appropriate in a given context. I therefore seek to enrich the literature on the syntax-information structure interface in Galician by establishing clausal word order preferences and judgments for a variety of constructed information structure contexts. At the same time I wish to further inform the syntactic debate on the preverbal field in Galician using quantitative experimental data gathered using naïve Galician-speaking informants. Although SVO word orders have been claimed to be acceptable in all discourse contexts in Spanish and EP, it is unclear if SVO is preferred to other word orders like VSO and VOS in all of these discourse contexts. In order to determine

106   

   The interaction between syntax and information structure

word order preferences, I have created discourse contexts and tested them with native speakers of the target language.⁹⁸ However, there are certain challenges present in creating discourse contexts and potential replies. In particular, subjects can be problematic in languages like Spanish, EP, Catalan, and Galician in part because they are null-subject languages. The Spanish question-answer pair in (136) typifies the phenomenon I am referring to. (136) a.

¿Qué  cocinó Juan?  what  cook.PST.3SG Juan  ‘What did Juan cook?’

b. ((Juan) cocinó) una tortilla. Juan cook.PST.3SG a tortilla ‘((Juan) cooked) a tortilla.’ In (136b), the subject or the subject and verb may be lacking, elided in a reply to the object-narrow-focus question in (136a). In fact, the shortest reply consisting of just two words and no predicate would likely be the most common reply to this question. However, in order to effectively establish word order preferences for the subject, verb, and object, all of these constituents must be present. Therefore, for most of the discourse contexts that I examine, all possible replies are necessarily full sentences including a DP subject, verb and DP object. As previously discussed, since there are many more possible word orders in a language like Galician (as compared to a language like English or French), particular care must be taken since word orders like SVO, VSO and VOS are never ungrammatical; rather they may be more or less dispreferred. Therefore, the goal of eliciting judgments will be establishing word order judgments and preferences according to information structure context. The result may be that not all well-constructed syntactic structures are acceptable at the level of discourse structure. Determining these preferences then will assist in describing the syntax-information structure interface for Galician. I describe the quantitative and qualitative data-gathering methods that I employ to determine these preferences in Chapter 3.

2.11 Summary In this chapter, I have reviewed the literature on the syntactic analysis of preverbal subjects as either arguments or non-arguments. I have also reviewed

98 Recall the limitations related to monolingual Galician discussed in chapter 1.

Summary   

   107

many of the definitions of topic and focus in the literature. Additionally, I have discussed some of the analyses in the literature that have proposed to account for the syntax-information structure interface. In the methodology that follows, I adopt López’s (2009) analysis of information structure in Spanish and Catalan for two reasons: 1) information structure features are not directly assigned in the syntax, thus avoiding theoretical problems related to Inclusiveness, and 2) his proposal makes clear predictions related to the acceptability of CLLD, which allows this study to simultaneously inform questions related to word order preference as well as questions related to preverbal subject position in Galician. To recap, I briefly summarize the information structure assumptions that inform the following chapter. The definition of thetic contexts that I assume follows Zubizarreta’s (1998) characterization of these contexts as “out of the blue”, or what many others have called “all focus”. Such sentences may either initiate a discourse or provide replies to questions like “What happened?”. As previously discussed, the basic definition of topic that I assume is based on the notions of “aboutness” or “discourse old” in the literature outlined in section  2.8.3. However, since I examine CLLD topics in particular, I assume López’s (2009) analysis for CLLD, as discussed in section 2.9.3. According to López’s proposal, CLLD requires discourse subordination, which consists of a discourse antecedent and a subordinating discourse context as defined by Asher & Vieu (2005). In Chapter 3, I design task conditions to test the validity of this proposal for Galician. The definition of focus that I assume follows López (2009), which assumes that the difference between regular focus (rheme) and contrastive focus depends on how the focus integrates into the existing discourse. While regular focus resolves an open variable in the discourse to be resolved in the reply, contrastive focus simultaneously opens and resolves a variable. Regular focus provides an answer to an explicit or implicit wh-question, while contrastive focus cannot answer a wh-question. The regular focus contexts I will employ are what I refer to as narrow focus, whereby a wh- question elicits either the subject or object. I do not employ any contrastive focus or focus fronting contexts in the quantitative data that I gather. In Chapter 3, I also detail more precisely how the above information structure assumptions are tested for in the quantitative tasks and conditions that I employ in my investigation of clausal word order in Galician. Via this research methodology I collected descriptive corpus data on the clausal structure of Galician not only for the sake of documenting this minority language, but also for the implications that this data may have for the theoretical analysis of similar Romance

108   

   The interaction between syntax and information structure

Languages like Spanish and Portuguese. The chief questions to be addressed in the remaining chapters of this monograph are the following: i. What is the preferred clausal structure in Galician for out-of-the-blue, thetic, sentences? ii. What is the preferred clausal structure in Galician for sentences in which the grammatical subject represents topical, common-ground information? iii. What is the preferred clausal structure in Galician for sentences in which the grammatical object represents topical, common-ground information? iv. What is the preferred clausal structure in Galician for sentences in which the grammatical subject is narrow-focused (rheme)? v. What is the preferred clausal structure in Galician for sentences in which the grammatical object is narrow-focused (rheme)? vi. Do CLLD elements in Galician conform to López’s (2009) analysis of CLLD? vii. How does the data collected contribute to the overall analysis of clausal structure in Galician? viii. What does the data obtained for Galician imply for previous analyses of clausal structure in Spanish and Portuguese?

3 Methodology 3.1 Preliminary concerns As we have seen thus far, there are a number of difficulties related to grammatical subjects (i.e. the external argument in the case of transitive predicates) in a number of Romance languages, and Galician is no exception. However, the case of Galician is complicated not only by the data examined in Chapter 2.7, but also by the lack of extant claims regarding Galician clausal structure or information structure. As discussed, I follow López (2009) in assuming that the syntactic module assembles the lexical building blocks of the clause, which in turn feed into the pragmatic module. Since not all well-constructed syntactic structures are acceptable at the level of discourse pragmatics, the syntax–information structure interface is a crucial step in the building of an acceptable clause. Accurately describing this interface requires systematic, empirical data on what types of sentence structures are (more) appropriate in certain kinds of discourse contexts to serve as a guide. Therefore, I have designed quantitative and qualitative experimental tasks to create a variety of pragmatic contexts geared to elicit various types of production data for Galician, which will in turn provide the empirical base for my analysis of Galician clausal structure. The quantitative data is largely based on solicited intuition production, while the qualitative data is solicited, semi-informal speech production. For well over two decades now, experimental measures supplemented by quantitative statistical analyses have formed the basis of a large body of generative SLA research investigating theoretical issues. Despite the fact that the vast majority of such research also gathers quantitative data from native speakers for statistical comparison with a wide variety of L2/Ln learner levels, a relatively small number of researchers in theoretical syntax have adopted such experimental measures. Informally, experimental syntactic research is characterized by the administration of a clearly defined psycholinguistic task – often in questionnaire form with a randomized and counterbalanced task design – ideally completed by participants who are naïve to the prescriptive rules and theoretical questions at stake. Task completion is followed by quantitative statistical analysis on the part of the researcher. Quantitative questionnaires of this type can be used to gather data on grammaticality, acceptability, unacceptability, and even ungrammaticality judgments (e.g. Goodall 2008). Some examples of such studies (as cited in Myers 2009a) are Bernstein, Cowart & McDaniel (1999), Featherston (2005a, b), Clifton, Fanselow & Frazier (2006), and Sprouse (2007). The most obvious benefit to experimental research is that it avoids the experimenter bias inherent in much

110   

   Methodology

intuition-based theoretical linguistic research, for which the chief informant is typically the author. Myers (2009b: 411) succinctly summarizes the benefits of experimental methodology: “[w]hile informal judgments often happen to be reliable, only sampling and statistical analysis allow reliability to be assessed.” This is not to say that experimental data holds a privileged status in comparison with other types of linguistic data, e.g. corpus or intuitive data (see González-Vilbazo, Bartlett, Downey, Ebert, Heil, Hoot, Koronkiewicz & Ramos 2013 for a more comprehensive methodological data model), but rather that quantitatively elicited grammar judgments can help provide the theoretical researcher with a more reliable and complete picture of native-speaker intuitions and competencies (Penke & Rosenbach 2004). In the ideal case, experimental data will be complemented by other types of data-gathering measures (Geeslin 2010). Henry (2005) in particular found follow-up consultation with individual speakers to be especially insightful with respect to speaker judgments of (non-standard) Belfast English. Although questionnaire design, data collection and statistical analysis can be a time-consuming pursuit, the data gathered can be invaluable in supporting (or questioning) corpus data, oral/written speech samples, fieldwork, and individual grammar judgments. The data gathered in experimental measures can also be used to confirm or disconfirm linguistic hypotheses, e.g. Davies’s (1996) testing of the Morphological Uniformity Hypothesis. Clearly, there are practical and logistical limitations with respect to using pencil-and-paper or Internet-based questionnaires for research on endangered or minority languages that lack an established written norm (or a written form altogether), but the limitations of a research tool or method – limitations and dangers which are omnipresent for a wide variety of trusted methods (see e.g. Schütze 2005) – should not be taken to suggest that such methods and tools are invalid. The experimental measures detailed in this chapter seek to gather data on acceptability judgments, preferences, and uses of a variety of word orders for a variety of information structure contexts. The verbs used to elicit judgments and preferences in these tasks are all agentive transitive predicates and were chosen in order to guarantee the presence of a subject and object argument in the clausal structure. I leave intransitive predicates, which lack a grammatical object, for future research. I designed these question contexts bearing in mind in particular the proposals of Zubizarreta (1998) for Spanish and Italian, Costa (2000, 2004) for European Portuguese, Casielles (2004) for Spanish, and López (2009) for Spanish and Catalan, all of whom have examined in some detail the syntax-discourse interface. In the following section, I describe the subjects who participated in the first phase of quantitative data collection in this investigation, examining in turn the variables under investigation and the particular experimental data-gathering procedures I employed.

Participants: Tasks 1 and 2   

   111

3.2 Participants: Tasks 1 and 2 Prior to completing tasks 1 and 2, participants completed a linguistic history questionnaire, which included a maximum of 63 questions about age, place of birth, current place of residence, experience living outside of Galicia, the language(s) that participants spoke and learned as a child, as well the language(s) that they speak at home, at school, at work, with friends and colleagues, and with family relations. The linguistic questionnaire also included questions about experience with Galician-language education at the primary, secondary and tertiary (university) level. These questions and the response options were very similar or identical to questions included in questionnaires used by the Sociolinguistic Map of Galicia (MSG, González González 2007). I have included the full text of the Galician linguistic history questionnaire in Appendix 1 for reference. As with the experimental procedure detailed below, this questionnaire was administered in a computer-assisted module using the WebSurveyor Internet survey interface via Information Technology Services at University of Iowa. I report on participant responses to linguistic history questions below in my description of variables that I had originally planned to examine, and explain the decisions taken to remove such variables from consideration.

3.2.1 Participant variables For this investigation, I recruited 54 male and female participants currently living in the Galician province of Pontevedra. Pontevedra is located on the western Atlantic coast of Spain, just north of the border with Portugal. Among the 54 who started the Internet-based quantitative tasks, the results from 34 of these participants were removed from statistical consideration following initial analysis. 28 of these did not complete the second battery of tasks. The remaining 6 belonged to the 31–49 year-old age group, which I examine in Gupton (2014b). Of the 20 remaining participants, 10 were male and 10 were female, all of whom reported using Galician on a daily basis. According to 2007 Internet population figures from the Galician Statistical Institute (www.ige.eu), the province of Pontevedra had a population of 947,639, of which 48.4% was male and 51.6% female. The two age groups examined in this study are ages 18–30, and older than 50. The initial plan was to consider groups differing by one generation, and therefore detect potential intergenerational grammar differences, which might provide evidence of potential linguistic changes in progress.

112   

   Methodology

The next variable I consider is primary living environment. The Sociolinguistic Map of Galicia (henceforth MSG) divides Galician population centers into the following four groups according to population size: i. fewer than 5,000 inhabitants over 15 years of age ii. 5,000 to 10,000 inhabitants over 15 years of age iii. 10,000 to 50,000 inhabitants over 15 years of age iv. greater than 50,000 inhabitants over 15 years of age Group four corresponds to the seven major Galician cities: Vigo, A Coruña, Santiago de Compostela, Lugo, Ferrol, Ourense and Pontevedra. The least populous urban environment among the seven major Galician cities is the city of Pontevedra, with 80,202 inhabitants as of 2007. Taking into account economic environment, the MSG further divides these groups into eight categories: urban-1, urban2, suburban, vila-1, vila-2, vila-3, rural-1, rural-2. Urban and suburban areas are those with 40,000 or more inhabitants, vila (small town) areas are those with between 2,000 and 39,999 inhabitants, and rural areas are those with fewer than 2,000 inhabitants.⁹⁹ These classifications are not without debate, however; Lois González (1992) defines vila more conservatively, having between 2,000 and 15,000 inhabitants, thus leaving a grey area of some 60,000 between the categories of vila and urban. In fact, Lois González argues that inhabitants of vilas such as Carballo, Oleiros, Nalón or Vilagarcía de Arousa with populations of between 30,000 and 40,000 should also be considered as urban speakers owing to the lifestyle and industries present in these areas. Ramallo (p.c.) agrees with this assessment, stating “vilas like Vilagarcía de Arousa have all of the characteristics of an urban environment.” For the purposes of this study, I adopt the binary variable [±urban] to describe the primary living environment of Galician-speaking participants. I define [+urban] as speakers who primarily reside in areas of greater than 30,000 inhabitants, thus including a subset of speakers classified as vila-1 in the Fernández & Rodríguez (1994) portion of the MSG. Therefore, I consider speakers in this investigation from the areas of Pontevedra (city), Vigo and Vilagarcía de Arousa to be [+urban]. I define [-urban] as speakers who primarily reside in areas of fewer than 30,000 inhabitants, which includes all remaining areas of Pontevedra province which do not lie in the immediate surroundings of the urban centers above. See Table 13 below for a summary of the [+urban] population centers involved in this investigation.

99 It is unclear to me why these MSG population number categories do not coincide. However, as I discuss below, these categories have been actively debated in the literature.

Participants: Tasks 1 and 2   

   113

Table 13: [+urban] areas in Pontevedra province and current population figures.

Area

Population

Vigo Pontevedra Vilagarcía de Arousa

294.772 80.202 36.743

Source: Galician Statistical Institute (www.ige.eu). Santiago de Compostela, 2007.

The remaining participant variable that I consider in this investigation is level of education. The MSG reports very similar dominance percentages for reading, writing, speaking and comprehension for those with secondary and university education (Figure 1). ¹⁰⁰

Figure 1: Sociolinguistic Map of Galicia. Percentages of participants who rated their Galician abilities as ‘high’ (3) or ‘very high’ (4) by education level. (González González et al. 2007: 225–228).

100 The MSG differentiates between three educational levels that I collapse under the label “secondary”. These are ESO = Educación Secundaria Obligatoria (Obligatory Secondary Education), FP = Formación Profesional (Vocational-technical Development), and bacharelato (universitypreparatory baccalaureate). This is because all three of these groups have minimally completed ESO. Percentages are means of ratings of the three groups.

114   

   Methodology

All four language abilities asre rated higher with higher formal education levels. There is a notable difference in reported reading and writing abilities between those with and without a primary education, which the MSG attributes to the fact that, as a general rule in Spain, students become more familiar with (formal) written language and texts at the higher education levels. Curiously, the group without formal studies rates their speaking abilities lower than the more educated groups, which is quite surprising, since this skill does not require schooling, strictly speaking. Due to the successes of compulsory education, [-educated] speakers in the 18–25 age group exist in extremely limited numbers. Participants among all age groups who reported having no formal education in González González et al. (2007: 153, table 2.1) was extremely low (1.4%), while those who reported only primary studies was rather higher (18.2%). While exact counts are not provided, the totals for these groups comprise less than 20% of the population, assuming this study to be representative of Galicia. For the purposes of this work, I define [+educated] as Galician speakers who have completed at least some level of secondary education.¹⁰¹ Owing to the fact that Ramallo (p.c.) reports that Galician speakers whose university majors in some way involve Galician (e.g. Translation, Galician Philology, etc.) possess a greater metalinguistic awareness than those whose major studies do not, I attempted to include a wider spectrum of [+educated] speakers. A summary of the participant variables under consideration to this point, as well as their numerical representation in the questionnaire results, appears in Table 14 below. As can be clearly seen, not all of the variables are represented in the participant population that successfully completed both days of the Internetbased tasks. Table 14: Summary of participant variables.

Male 18–30 50+ [+urban]/[+educated] n=5 [+urban]/[−educated] [−urban]/[+educated] [−urban]/[−educated]

n=3 n=1

Female 18–30 50+ n=6 n=3

n=1 n=1

101 Following the 1990 educational reform law, or LOGSE (Ley de Ordenación General del Sistema Educativo), secondary education is compulsory until 16 years of age, as compared to 14 years of age prior to the law.

Participants: Tasks 1 and 2   

   115

Over half of the participants came from a [+urban], [+educated] background, and the majority of those belong to the 18–30 year-old group. There were very few participants from [-urban] backgrounds, yet among those, there were no participants from [-urban], [-educated] backgrounds. As many of the above variables were not sufficiently represented in the results gathered, I only report on the variable of age as it relates to participant preferences and ratings. While my particular interests lay in the Galician-dominant portion of the population (i.e. native and habitual speakers of Galician), post hoc examination of participant responses to the linguistic history indicated differing levels of experience with and use of the Galician language in daily life. Therefore, based on participant replies to the linguistic history questionnaire, I separated participants into three language-dominance groups: Galician-dominant, Spanish-dominant, and dual dominance based on their replies. The definition of the term Galiciandominant that I adopt for the purposes of this monograph includes speakers who replied to language use questions with either only Galician or more Galician than Spanish. These answer options match the classifications used in the Sociolinguistic Map of Galicia (MSG), 1992–1996 (Fernández & Rodríguez 1994), the Sociolinguistic Map of Galicia (MSG) 2004 (González González 2007), and those in ReiDoval (2007) which examined a subset of data from the MSG of urban speakers from the seven major urban areas of Galicia. According to Rei-Doval’s (2007) figures for only Galician and more Galician than Spanish usage groups, 45.7% of urban Galician males are Galician-dominant speakers compared to 40.3% of Galician females. Among these speakers, however, literacy figures are quite low, owing to the fact that Galician has been largely orally transmitted for the past five hundred years among a mostly rural population. According to the MSG, only about half of Galician-dominant speakers reported to be competent reading Galician, and only a quarter reported to be competent writing in Galician – a figure which steeply decreases in older age groups.¹⁰² All of the individuals that I classified as Galician-dominant individuals reported Galician as their first language, and reported using only Galician or more Galician than Spanish in the majority of the family, social, and work contexts included in the linguistic questionnaire.

102 This appears to be a pessimistic reading of the figures. Examining the original MSG 2004 figures in González González (2007: 173–174), which presents figures on written expression (writing) and written comprehension (reading), we find a slightly different scenario. 49.3% of respondents rated their writing abilities as ‘4’, and 25.1% rated them as ‘3’, which taken together comprises 74.4%. 59.3% of respondents rated their reading abilities as ‘4’, and 22.8% rated them as ‘3’, which totals 82.1%.

116   

   Methodology

The group that I classify as dual-dominance based on their linguistic history questionnaire replies reported “both Galician and Spanish” as their first languages and reported everyday use of Galician as either only Galician or more Galician than Spanish in no less than half of the questions related to family, social, and work. The group that I have classified as Spanish-dominant based on their linguistic history questionnaire responses reported their first language as Spanish. Among these individuals, the frequency with which they reported speaking only Galician or more Galician varied more widely than in the other two language dominance groups. These speakers reported that they started to learn the language as early as six years old, and as late as 40 years old.¹⁰³ Although approximately 90% of Spanish-dominant speakers report being able to understand Galician, and approximately 50% report being able to speak it (Fernández & Rodríguez 1994), Galician is typically perceived as the low-prestige variety, while Spanish is the high-prestige variety within Galicia.¹⁰⁴ An extremely large percentage of the population is exposed to Spanish in school at a very young age. In urban environments with greater levels of domestic and foreign immigration and industry such as Vigo or A Coruña, the level of exposure to Spanish in society is even greater. This is reflected in habitual language preference among the younger (16–25) urban generation, among whom 87.6% prefer to use Spanish, compared to 12.4% who prefer to use Galician (Ramallo 2007). Del Valle (2000), however, argues that such analyses of prestige and preferential use are based on a ‘linguistic culture of monoglossia’ prevalent in Western thought. By his analysis, the “language attitudes and linguistic behavior of Galicians are grounded in the linguistic culture of heteroglossia: acceptance of multiple norms and resistance to convergence.” (Del Valle 2000: 130) The group that I have classified as Galician-dominant consists of six individuals between the ages of 19 and 28 at the time of data collection in 2008. The dual-dominance group consists of eight participants between the ages of 18 and 29, and one individual who was 57. The Spanish-dominant group consists of six people between the ages of 50 and 57, and one who was 24. I summarize these numbers in Table 15 below.

103 These ages represent the extremes reported. Most of these speakers reported starting to learn the language between the ages of 12 and 15. 104 Note that as with the term Galician-dominant, I have collapsed the two MSG speaker categories of Only Spanish and More Spanish than Galician into the category Castilian-dominant. When I use the term Castilian-dominant, I am referring to Castilian Spanish speakers. I use these terms interchangeably.

Participants: Tasks 1 and 2   

   117

Table 15: Summary of gender, age and language-dominance variables

language dominance

Male Female Total 18–30 50+ 18–30 50+

Galician dual Spanish

2 2 1

0 1 4

4 5 0

0 0 1

 6  8  6

TOTAL

5

5

9

1

20

Due to the numerical presence of each age and language dominance, separating the age variable from language dominance presents definite complications. I discuss this issue in greater detail in Chapter 4 when I present the results of the study.

3.2.2 Procedures: Tasks 1 and 2 Tasks 1 and 2 were quantitative data-gathering tasks administered by computerassisted Internet questionnaire modules using the WebSurveyor interface on University of Iowa servers. Participation was entirely voluntary and anonymous. Monetary compensation or otherwise was not provided. Subjects were given the Internet address of the survey and participated from their homes and without supervision by the investigator. These tasks consist of a total of 65 questions completed over the course of two separate website visits to avoid participant fatigue and linguistic saturation. In each visit, each participant responded to either 33 or 32 linguistic task questions, which involved 17–18 questions from task 1, and 15 questions from task 2. Question stimuli were previously randomized by an online random number generator (http://stattrek.com) prior to creation of modules in WebSurveyor. Participants were randomly assigned to two groups, thus varying the order in which the quantitative linguistic tasks as well as the items within the tasks were presented in order to avoid any task effect. In the case of task 1, which included audio prompts, the gender of the interlocutors in the task items also varied between the two groups. For Task 1, there were seven different condition types, and five tokens for each condition, yielding a total of 35 tokens total. There were six different condition types in task two, and five tokens for each condition type, netting a total of 30 tokens. As Task 2 was composed of six different extended discourse contexts, participants completed three discourse contexts per visit in order to avoid task-type fatigue as well as overall saturation fatigue. Context presentation order for Task 2 was pseudo-randomized, since total randomization would have adversely affected the construction of the information structure contexts within them.

118   

   Methodology

3.3 Task 1: Appropriateness Judgment Task Task 1 is based on the grammaticality judgment task (Bley-Vroman & Yoshinaga 1992, among numerous others) used in Second Language Acquisition research to test the acceptability of sentences in learner grammars, results from which are typically compared with native-speaker results. In this task however, participants first read a conversational context and then provided an appropriateness rating for (syntactically) grammatical sentences that varied in clausal structure. For each conversational context, a triad of possible sentence responses with varying word orders was provided following Kallestinova’s (2007) methodology employed for gathering data on clausal variants in Russian. Participants were instructed to click on a speaker icon ( ), which led to audio hyperlinks so that participants would listen to each possible response in the triad and then rate them on a scale of acceptability. Audio recordings, recorded by native speakers of Galician, were provided following Kitagawa & Fodor (2006), who suggest that misjudgments can arise when prosody is “left to the imagination.” Therefore the presence of audio recordings helps to control for the intonation properties of the clausal structures presented, and to prevent accommodation of a word order with an unintended reading. As all clausal word order options were grammatical, this task scale utilized an ordinal scale of 1 to 5, (1 meaning ‘not acceptable’, and 5 meaning ‘totally acceptable’), following recommendations in Schütze (1996) and White (2003), and thus differed from Bley-Vroman & Yoshinaga’s (1992) original scale, which used a five-point ordinal scale from -2 to 2. This scale accompanied each task 1 triad for the participants’ reference. 1 = non aceptable 2 = pouco aceptable 3 = máis ou menos aceptable 4 = bastante aceptable 5 = totalmente aceptable (preferible)

not acceptable marginally acceptable more or less acceptable rather acceptable totally acceptable (preferable)

These ratings provide insight on clausal word order acceptability among participants according to the information structure context in question. Instructional sample tokens preceded the task so that participants would know that there may be more than one acceptable manner in which to conclude the discourse situations in the task, (i.e. that two or more sentences may receive the same acceptability rating). This was an important design factor, as it was previously unknown if more than one word order would be deemed acceptable for a given context. Pragmatic conditions elicited in the stimuli included thetic (out-of-the-blue) situations, subject narrow-focus or rheme, object narrow-focus or rheme, subject

Task 1: Appropriateness Judgment Task   

   119

arguments as topical information, and object arguments as topical information, each of which I discuss below.

3.3.1 Condition A Condition A tokens sought to establish a clausal word order preference for thetic, out-of-the-blue situations. These do not presuppose knowledge of the subject, verb, or object, but only that something occurred (i.e. they ask the basic question “What happened?”) These questions provide the following possible word order responses: Subject-Verb-Object (SVO), Verb-Subject-Object (VSO), and VerbObject-Subject (VOS), as in (137). (137) Context: Xoán and Iago are friends. They are talking about the weekend. Xoán – Que fas esta noite? what do.PRS.2SG this night Xoán – ‘What are you doing tonight?’ Iago – Por que? Que pasa? why what happen.PRS.3SG Iago – ‘Why? What’s up?’ a.

Xoán – Carlos Carlos

vai celebrar o seu aniversario. go.PRS.3SG celebrate.INF the his birthday

b. Xoán – Vai celebrar Carlos o seu aniversario. go.PRS.3SG celebrate.INF Carlos the his birthday c.

Xoán – Vai celebrar o seu aniversario Carlos. go.PRS.3SG celebrate.INF the his birthday Carlos Xoán –‘Carlos is going to celebrate his birthday.’

3.3.2 Condition B Condition B items sought to determine the preferential position of the subject when it is topical, common-ground information, as discussed in Chapter 2. To avoid a null subject as well as unnatural repetition of an overt subject in these items, a switch reference or paraphrase of the subject referent is used (e.g. in

120   

   Methodology

(138), the switch from “your daughter” to “Belén”).¹⁰⁵ The possible word order responses examined for this condition are SVO, VSO, and VOS. (138) Context: Xoán and Felipe are friends. They are talking about Felipe’s daughter. Xoán – Cantos anos ten a túa filla máis how_many years have.PRS.3SG the your daughter more pequena? small Xoán – ‘How old is your youngest daughter?’ a.

Felipe – Pois, Belén ten well Belén have.PRS.3SG

cinco anos. five years

b. Felipe – Pois, ten Belén cinco anos. well have.PRS.3SG Belén five years c.

Felipe – Pois, ten cinco well have.PRS.3SG five Felipe – ‘Well, Belén is five years old.’

anos Belén. years Belén

Condition B contexts were all Question-Answer (Q-A) pairs, which, as previously discussed in chapter 2.9.3, are classified as subordination contexts according to Asher & Vieu’s (2005) analysis of discourse relations in the Segmented Discourse Representation Theory (SDRT) framework (Asher 1993). According to López (2009), discourse subordination creates a context appropriate for clitic left-dislocation of previously introduced referents. Since in this condition the subject has already been introduced within the discourse, the candidate word order with the highest acceptability rating for this condition will provide data to inform this question. Recall that, as discussed in Chapter 1, whether preverbal subjects are left-dislocated is a matter of considerable debate; however in a given SVO sentence it is impossible to detect whether a preverbal subject appears in a canonical or left-peripheral position. Assuming that CLLD is appropriate in subordination discourse contexts following López (2009), a response pattern favoring SVO replies for Condition B may or may not indicate that preverbal subjects are CLLD elements. Although the discourse relation here is appropriate for a CLLD

105 It was clear to the author a priori that common-ground subjects typically surface most naturally as null subjects in extended, non topic-switched discourse in Galician. However, as the goal of this study was to establish the precise position of the subject relative to other clausal constituents, a certain level of naturalness had to be sacrificed.

Task 1: Appropriateness Judgment Task   

   121

preverbal subject, pauses and intonation were controlled for in the design of this task, and these tokens did not have the intonation characteristic of subject CLLD. Therefore, assuming that participants completed the condition tokens faithfully and listened to each response in the triads, the interpretation of preverbal subjects as CLLD elements is unlikely. The other possibilities in this condition, VSO and VOS, are not expected to be preferred word orders since, according to the Romance Nuclear Stress Rule (see Zubizarreta 1998), subject and object rheme constituents should appear in rightmost, clause-final position. The availability of CLLD is also of relevance for direct objects, which are examined in Condition C.

3.3.3 Condition C Condition C tokens were designed to establish the appropriateness of a number of clausal structures when an object is topical, common-ground information within the context presented. These contexts also involved yes-no Q-A pairs with elaboration replies, thus serving as a discourse subordination counterbalance for the Condition B tokens above. These items sought to test the availability and relative appropriateness of clitic left-dislocation for Galician clausal structure. The response triads in this condition have a bi-clausal structure, placing the first conjunct with CLLD in opposition to the second conjunct. Recall that this is supposed to create an anaphoric relation appropriate for opposition with a CLLD constituent. Following (Villalba 2000), an appropriate relation between a CLLD constituent and its antecedent must belong to one of the following categories: subset, superset, set-membership, or part-whole. This creates a contrast between each answer and its corresponding clause in reply to the context question. According to López (2009), such a contrast is the most natural way of using CLLD in Romance. Furthermore, the CLLD object in the first conjunct will create a strong preference for a parallel CLLD structure in the second one. Note that the resumptive object clitic in these tokens is enclitic on the verb since Galician requires enclisis with CLLD. The word order possibilities for the first conjunct in these triads are CLLDSVcl (clitic left-dislocated object followed by a preverbal subject and a verb with enclisis), CLLDVclS (clitic left-dislocated object followed by a verb with enclisis and a postverbal subject), and SVO, as in (139).

122   

   Methodology

(139) Context: Moving day. Carlos and Patricia are a couple. When they enter their new apartment, there is a ton of furniture inside. Patricia – Caramba! Meteu a túa familia todos os mobles caramba put.PST.3SG the your family all the furniture xa? already Patricia – ‘Wow! Your family already brought in all the furniture?’ a.

Carlos – As mesas os meus tíos metéronas the tables the my uncles put.PST.3PL-CL.ACC.F.3PL pero but as cadeiras deixáronas no portal the chairs leave.PST.3PL-CL.ACC.3PL.F in_the door

b. Carlos – As mesas metéronas os meus tíos the tables put.PST.3PL-CL.ACC.F.3PL the my uncles pero but as cadeiras deixáronas no portal the chairs leave.PST.3PL-CL.ACC.3PL.F in_the door Carlos – ‘The tables my uncles brought in, but the chairs they left in the entryway.’ c.

Carlos – Os meus tíos meteron as cadeiras pero as the my uncles put.PST.3PL the chairs but the cadeiras chairs deixáronas no portal leave.PST.3PL-CL.ACC.3PL.F in_the door Carlos – ‘My uncles brought in the tables, but the chairs they left in the entryway.’

Word orders did not vary in the second conjunct in this condition, and all included an appropriately anaphoric CLLD object according to the prerequisites introduced above.¹⁰⁶ Following López (2009) then, the prediction is that the object should receive higher appropriateness ratings in a CLLD position in the first conjunct. An 106 Note that the preverbal subject os meus tíos in examples (139a, c) is also a viable candidate for left-dislocation as well since it, too, has an appropriate discourse antecedent in a túa familia. I return to this issue below.

Task 1: Appropriateness Judgment Task   

   123

object appearing in canonical position is predicted to violate the parallel structure of CLLD between the two conjuncts. This condition also sought to determine whether a preverbal or postverbal subject would be available or preferred in the first (CLLD) conjunct; however, CLLDVclS was not an expected word order for reasons related to the Romance NSR, since it was not the case that the postverbal subjects in this condition represented narrow-focus (rheme) response information. SVO was not a predicted word order either, since the direct object, which represents discourse-anaphoric information in these tokens, should scramble to escape the scope of narrow focus according to the Romance NSR. CLLDSVcl was expected to be the highest rated word order for this condition because the direct object and preverbal subject always belonged to the set of previously evoked (i.e. anaphoric) referents in the discourse, either directly or indirectly. Note that the second conjunct in all response options in (139) involves a null subject. All subjects in the response triads were subsets, recasts, or switch references of the subject in the elicitation context in order to avoid 1) repetitiveness, which would create an unnatural response, and 2) null subjects, since a null-subject response in the first conjunct would not provide any indications about the position of the subject and its acceptability, as discussed in section 3.3.2, above.

3.3.4 Condition D Condition D tokens were designed to determine clause structure acceptability for contexts in which an appropriately anaphoric subject appears in a continuation or result relation to the discourse elicitation context. This information structure condition serves as a counterbalance for Condition B above, which included a previously-evoked subject in a subordinating relation to the preceding discourse. It is also a counterbalance for Condition E below, which includes a common-ground direct object in a continuation or result relation. Recall that both of these discourse relations are classified as coordination contexts by Asher & Vieu (2005), which by López’s (2009) analysis, should prohibit a CLLD subject.¹⁰⁷ The word order options provided in the response triads were SVO, VSO, and VOS, as in (140). 107 Note that the implicit assumption here is that both subject and object CLLD will be infelicitous in coordination discourse contexts. I return to this issue in Chapter 4, showing evidence suggesting that subject and object CLLD do not have the same discourse properties. It is worth noting that only two of the tokens in this condition involved result contexts. Note also, for Asher and Vieu (2005), result has a “chameleon-like quality”. They propose that result is a coordinating discourse relation by default, but it may also be a subordinating relation if the surrounding context supports it (see discussion in Asher and Vieu 2005: 605–606). Individual results for these tokens do not indicate divergent behavior as compared to continuation contexts.

124   

   Methodology

(140) Manuel – Escoitaches? hear.PST.2SG Manuel – Did you hear? Agustín – O que? the what Agustín – What? Manuel – Samuel Sánchez gañou unha Samuel Sánchez win.PST.3SG a Manuel – Sanuel Sánchez won a gold medal!

medalla medal

de ouro! of gold

Agustín – Si? Que ben! yes how good Agustín – ‘Yeah? That’s so great!’ a.

Manuel – Pois, si. E well yes and conseguiu get.PST.3SG un bo contrato de a good contract of

como resultado, o rapaz as result the young_man

publicidade. publicity

b. Manuel – Pois, si. E como resultado, conseguiu o well yes and as result get.PST.3SG the rapaz young_man un bo contrato de publicidade. a good contract of publicity c.

Manuel – Pois, si. E como resultado, conseguiu un well yes and as result get.PST.3SG a bo contrato good contract de publicidade o rapaz. of publicity the young_man Manuel – ‘Well, yes. And as result, the fellow got a good advertising contract.’

The availability of CLLD in this discourse condition is crucial in regards to the preferred clausal position of the subject as well as the suggestion that preverbal subjects and CLLD objects behave in a syntactically uniform manner. If López (2009) is correct in proposing that coordination discourse contexts prohibit CLLD

Task 1: Appropriateness Judgment Task   

   125

in Galician as it is posited to do in Spanish and Catalan, SVO should be an impossible word order in Condition D and accordingly receive low acceptability scores. VSO and VOS however should receive higher acceptability ratings on the acceptability judgment task for Condition D; however, since the subject was anaphoric with respect to the preceding discourse in these tokens, it should not appear at the rightmost edge. Therefore, VSO was predicted to be the highest rated word order for this condition. Nevertheless, a high acceptability ratings for SVO clausal word order in this condition would provide indirect evidence suggesting that subjects in preverbal position are not (necessarily) left-peripheral elements.¹⁰⁸ That is, if discourse-old preverbal subjects appear to be non-peripheral elements, this suggests additional evidence against left-peripheral accounts of preverbal subjects.

3.3.5 Condition E Condition E contexts also sought to verify the availability and appropriateness of dislocation of objects for Galician clausal structure. In these discourse contexts, the direct object represented appropriately anaphoric discourse-old, commonground information, as in Condition C; however, the contexts employed in the superordinate discourse in these response triads were narration, and therefore discourse coordination contexts, as in Condition D above.¹⁰⁹ The word orders examined in this item were CLLDSVcl, CLLDVclS, and SVcl, as in (141). Three questionnaire items of this type included PP adjuncts, and two included adverbial adjuncts, neither of which was predicted to affect the appropriateness of the replies in the triads. (141)

Context: Pedro saw Maria at the library on Saturday. a.

O domingo, a rapaza viuna Marco the Sunday the girl see.PST.3SG-CL.ACC.F.3SG Marco alí tamén. there also

108 See López (2009) also for a similar observation to the effect that just because preverbal subjects may be dislocated does not mean that they must be. 109 As a methodological note, conditions D and E would have been better counterbalanced for CLLD subjects and objects had both consisted of narration, continuation or result discourse contexts, and not an asymmetrical combination of these. However, as coordinating discourse relations, the same CLLD felicity predictions should apply (with the possible exception of result, as noted in note 107).

126   

   Methodology

b. O domingo, a rapaza Marco viuna the Sunday the girl Marco see.PST.3SG-CL.ACC.F.3SG alí tamén. there also ‘On Sunday, the girl, Marco saw (her) there also.’ c.

O domingo, Marco viuna alí tamén. the Sunday Marco see.PST.3SG-CL.ACC.F.3SG there also ‘On Sunday, Marco saw her there also.’

If López (2009) is correct in that coordination discourse contexts prohibit CLLD, as in Condition D, then both CLLD word order responses (CLLDSVcl and CLLDVclS) in these response triads should receive low acceptability ratings for this condition. If preverbal subjects in SVO (here SVcl) are CLLD elements, they too should receive low acceptability ratings; however if preverbal subjects are not CLLD constituents, they should receive very high acceptability ratings.

3.3.6 Condition F Condition F tokens were designed to establish a preference for Galician clausal architecture when there is narrow focus on the subject (i.e. the subject is the rheme, thus resolving the variable opened in the discourse). Recall that question-answer pairs such as narrow-focus are classified as discourse subordination contexts following Asher & Vieu (2005), and, following López (2009), create an appropriate context for CLLD. However, CLLD material is typically topical, common-ground material in some sense, even when it is in a subset/poset relation with a preceding discourse constituent. Therefore, the prediction is that a narrow-focus subject should appear to the right of topical information, which dovetails nicely with Zubizarreta’s (1998) reformulation of the Nuclear Stress Rule for Romance (R-NSR), by which new information (i.e. narrow-focus constituents) should appear at the rightmost edge of the clause. For her, p-movement, or prosodic movement, is the result of a constituent scrambling to the left of other constituents and escaping vP in order to avoid the projection of (information) focus. Condition F is a counterbalance for Condition G below, in which objects appear in narrow-focus eliciting contexts. Three of the five tokens in this condition involved permutation of an adverbial XP adjunct with the direct object appearing as a clitic pronoun, as in (142). The word order permutations that appeared in triads of this type are SVclX, VclXS and VclSX.

Task 1: Appropriateness Judgment Task   

   127

(142) Context: Sandra and Beatriz are friends. Sandra asks Beatriz about a box in her apartment. Sandra – Que é iso? Que bonito! what be.PRS.3SG that how pretty Sandra – ‘What’s that? How pretty?’ Beatriz – Iso? Pois, é un regalo que chegou that well be.PRS.3SG a gift that arrive.PST.3SG por correo para o meu aniversario. by mail for the my birthday Beatriz – ‘That? Well, it’s a gift that arrived in the mail for my birthday.’ Sandra – Quen cho who CL.DAT.2SG-CL.ACC.M.3SG Sandra – ‘Who sent it to you?’ a.

Beatriz – A miña irmá the my sister a semana pasada. Non the week past not para a miña festa. for the my party

enviou? send.PST.3SG

envioumo send.PST.3SG-CL.DAT.1SG-CL.ACC.M.3SG vai poder vir go.PRS.3SG be_able.INF come.INF

b. Beatriz – Envioumo a send.PST.3SG-CL.DAT.1SG-CL.ACC.M.3SG the a semana pasada. Non vai poder the week past not go.PRS.3SG be_able.INF para a miña festa. for the my party c.

miña irmá my sister vir come.INF

Beatriz – Envioumo a semana send.PST.3SG-CL.DAT.1SG-CL.ACC.M.3SG the week pasada a miña irmá. Non vai poder vir past the my sister not go.PRS.3SG be_able.INF come.INF para a miña festa. for the my party Beatriz – ‘My sister sent it to me last week. She won’t be able to come to my party.’

If Zubizarreta’s (1998) configuration of the R-NSR for Spanish also applies to Galician, it is predicted that VclXS replies such as (142c) will be rated highly. If adjuncts scramble to avoid the scope of narrow-focus, a reply like (142b) with

128   

   Methodology

VclSX order should not be rated highly. A reply such as (142a), with SVclX word order should also receive low ratings since the subject would fall outside of the scope of narrow-focus, and additionally, could be inappropriately construed as topical CLLD material. The remaining two stimuli included DP objects, with the possible word order responses SVO, VSO, VOS. For these contexts, VOS should be the highest-rated word order among participants.

3.3.7 Condition G Condition G tokens counterbalance those of Condition F, seeking data on the preferred clausal structure for Galician when the information structure involves a narrow-focus (rheme) direct object. As discussed in section 3.3.6, questionanswer pairs create an appropriate context for CLLD. Therefore, a high acceptability ratings for SVO could suggest the presence of a CLLD subject. The word orders provided as responses to these object-focus triads were SVO, VSO and VOS, as in (143). (143) Context: Afonso, Mateo and Beto are friends. Afonso and Beto are talking about a drawing that took place recently Afonso – Gañou Mateo algunha cousa no sorteo? win.PST.3SG Mateo some thing in_the drawing Afonso – ‘Did Mateo win anything in the drawing?’ Beto – Pois, si. well yes Beto – ‘Well, yes.’ Afonso – Entón? Que gañou? then what win.PST.3SG Afonso – ‘So? What did he win?’ a.

Beto – O the

cabronazo gañou un bastard win.PST.3SG a

b. Beto – Gañou win.PST.3SG c.

o the

televisor! television

cabronazo un televisor! bastard a television

Beto – Gañou un televisor win.PST.3SG a television Beto – ‘The bastard won a television!’

o cabronazo! the bastard

Task 2: Word Order Preference Task   

   129

As previously mentioned, if Zubizarreta (1998) is on the right track, narrow-focus constituents should appear to the rightmost edge of the clause. If this is the case, experiment participants should rate VSO and SVO highly in this condition. VOS is not expected to be rated highly, as this would suggest that the subject and object are within the scope of narrow-focus.

3.4 Task 2: Word Order Preference Task Since Task 1 allowed for the possibility that participants may rate differing clausal structures with identical or very similar acceptability ratings, Task 2 was designed to encourage participants to choose a clear preference from two word order possibilities. If it turns out that two word orders are equally acceptable, this task will also provide confirmation of such a preference. Although Task 2 response options did not include accompanying audio, they appeared in an extended pragmatic context of three to seven stimuli in a connected, continuous discourse. Task two items included most of the same information structure contexts as in task one, with the exception of subordination versus coordination items testing the availability of clitic left-dislocation. Three sample questions preceded the items in task two in order to provide instructions to participants on how the contexts would be presented and how they could be answered. These instructions highlighted to participants that while more than one word order may be possible, one or another may not be appropriate depending on the context. The three sample contexts manipulated Adjective–Noun order possibilities, and were included to encourage participants to pay particular attention to the discourse contexts provided. As is well known, Spanish adjectives may precede or follow a noun. In the case of some adjectives, the placement of the adjective determines one or another specific meaning (see e.g. Bernstein 2001, among many others). The same applies to Galician (144).

(144) My friend Pedro has a lot of money, but he has very bad luck in life. a.

Pedro é un home Pedro be.PRS.3SG a man ‘Pedro is a poor man.’

pobre. poor

b. Pedro é un pobre home. Pedro be.PRS.3SG a poor man ‘Pedro is an unfortunate man.’

130   

   Methodology

c.

As dúas posibilidades. the two possibilities Both possibilities.

In option A, home pobre means ‘poor (not rich) man’, while pobre home in option B means ‘unfortunate man’. Therefore, only Noun–Adjective order, option B, is appropriate in this particular context. In the first sample token only NA order was appropriate, in the second sample only AN order was appropriate, and in the third sample, both AN and NA orders were possible and appropriate. This highlighted to participants that all three possible answers provided over the course of the task were possible response choices. Therefore, if it were the case that two word order options were equally possible and appropriate in a given context, participants would realize that it was appropriate to indicate such a preference.

3.4.1 Condition A Condition A involved thetic, or out-of-the-blue, contexts. Participants had to choose the appropriate word order, with the subject preceding or following the verb. Therefore the possible word orders in this item type were SV or VS (145): (145) Cristiano – Que aconteceu? what happen.PST.3SG Cristiano – What happened? Samo – Non se sabe not CL.PASS know.PRS.3SG Samo – ‘Nobody knows exactly.’ a.

exactamente. exactly

Samo – A señora da limpeza the lady of_the cleaning muller morta. woman dead

encontrou a find.PST.3SG the

b. Samo – Encontrou a señora da limpeza a find.PST.3SG the lady of_the cleaning the muller morta. woman dead Samo – ‘The cleaning lady found the wife dead.’

Task 2: Word Order Preference Task   

c.

   131

As dúas posibilidades. the two possibilities Both possibilities.

SVO and VSO orders were predicted to be the leading candidates for most acceptable clausal word order for thetic contexts. This item was designed to elicit a preference between these two word orders in case it turned out that both of these responses in the triads received statistically similar ratings in the corresponding Task 1 condition.

3.4.2 Condition B Condition B tokens involved a subject DP that would be established as commonground in the preceding context. The interested reader can find the full text of Task Two in Appendix C. The possible word orders in this context were SVO or VSO as well: (146) Veciño – Quen foi o outro rapaz que vin neighbor who be.PST.3SG the other fellow that see.PST.1SG hai pouco? haber.PRS.3SG little Neighbor – Who was the other fellow that I saw a little bit ago? Manuel – Foi o seu amigo, Fran. Veu be.PST.3SG the his friend Fran come.PST.3SG para xogar ao baloncesto esta tarde, for play.INF to_the basketball this afternoon Manuel – ‘That was his friend, Fran. He came to play basketball this afternoon...’ a.

pero non sabía but not know.IMPFV.3SG exame mañá. exam tomorrow

que that

Daniel ten un Daniel have.PRS.3SG an

b. pero non sabía que ten Daniel un but not know.IMPFV.3SG that have.PRS.3SG Daniel an exame mañá. exam tomorrow ‘...but he didn’t know that Daniel has an exam tomorrow.’

132   

   Methodology

c.

As dúas posibilidades. the two possibilities Both possibilities.

3.4.3 Condition C In Condition C, the subject represented the rheme within the extended discourse provided. The word orders participants were able to choose from in this condition were SVcl or VclS, as in (147). (147) Uxío – Quen che suxeriu esta empresa? who CL.DAT.2SG suggest.PST.3SG this business Uxío – ‘Who recommended this business to you?’ a.

Henrique – O meu irmán the my brother suxeriuma suggest.PST.3SG-CL.DAT.1SG-CL.ACC.F.3SG amigo do dono. friend of_the owner

porque because

é be.PRS.3SG

b. Henrique – Suxeriuma o meu suggest.PST.3SG-CL.DAT.1SG-CL.ACC.F.3SG the my irmán porque é amigo do dono. brother because be.PRS.3SG friend of_the owner Henrique – ‘My brother suggested it to me because he’s a friend of the owner.’ c.

As dúas posibilidades. the two possibilities Both possibilities.

If Galician clausal structure adheres to Zubizarreta’s (1998) configuration of the Nuclear Stress Rule for Romance (R-NSR), we predict that participants will prefer VS replies, which conform by placing narrow-focus (i.e. rheme) subjects in a postverbal position.

Task 2: Word Order Preference Task   

   133

3.4.4 Condition D Condition D presented an object narrow-focus (rheme) information structure context. These tokens counterbalance Condition C, eliciting participants to choose between VSO and VOS order, as in (148). (148) Xulia – Está lista xa be.PRS.3Sg ready already Xulia – ‘Is your sister ready yet?’

a túa irmá? the your sister

Noelia – Case, case... almost almost Noelia – ‘Almost, almost...’ Xulia – Que  busca? what  look_for.PRS.3SG Xulia – ‘What is she looking for?’ a.

Noelia – Busca unha toalla a  pobriña para look_for.PRS.3SG a towel the  poor for secar o pelo. dry. INF the hair

b. Noelia – Busca a  pobriña unha toalla para look_for.PRS.3SG the  poor a towel for secar o pelo. dry. INF the hair Noelia – ‘She’s looking for a towel to dry her hair (with).’ c.

As dúas posibilidades. the two possibilities Both possibilities.

The prediction here is that if VOS in Galician implicates scrambling of the object to avoid the rightward projection of information focus as previously discussed, the object will remain in situ, therefore favoring a choice of VSO order in these discourse contexts. As the grammatical subject in this condition represents common-ground information within the context of the story provided, it was not assumed that SVO would be a preferred word order in the corresponding Task 1 condition, and therefore was not included in this task. As we will see in the following chapter, however, SVO was highly rated, suggesting that object DPs may remain in their thematic position. Therefore, SVO should have been a word order

134   

   Methodology

under consideration in this condition. This methodological shortcoming was remedied in the Task 2 follow-up condition, which gathered preference data for all three of the word orders examined in Task 1 for object narrow-focus contexts. I provide the methodological details of this follow-up as well as its results in Chapter 4 , following the results for Task 2.

3.4.5 Condition E Condition E items are a counterbalance for Condition B. Within these contexts, the object represents common-ground information within the discourse context, therefore satisfying the anaphoric requirements of CLLD, as previously introduced for Condition C in Task 1. The two possible word orders in this item type were CLLDSVcl and CLLDVclS, with the discourse-old object clitic left-dislocated, as in (149). (149) Uxío – E onde  deixaron os outros mobles? and where  leave.PST.3PL the other furniture Uxío – ‘And where did they leave the rest of the furniture?’ a.

Henrique –Uff! O  sofá a miña   uff the  sofa the my   encontrouno   find.PST.3SG-CL.ACC.M.3SG

filla daughter no balcón. on_the balcony

b. Henrique – Uff! O  sofá encontrouno a miña  uff the  sofa find.PST.3SG-CL.ACC.M.3SG the my  filla no balcón.  daughter on_the balcony Henrique – ‘(Sigh) The sofa my daughter found (it) on the balcony.’ c.

As dúas posibilidades. the two possibilities Both possibilities.

As mentioned above, the response options for this token did not include SVO for two reasons: 1) it was not assumed that SVO would be a preferred word order when an object is discourse-old, and 2) I had previously predicted that CLLD word orders would receive higher ratings in Task 1.

Task 2: Word Order Preference Task   

   135

3.4.6 Condition F To ensure that participants faithfully performed the tasks in Task 2, five distractor items were included, which make up Condition F. The distractors were of the same type in the instruction section of Task 2, and elicited a choice between Noun–Adjective and Adjective–Noun word orders, as in (150). (150) Xulia – E non levas a túa mochila? and not carry.PRS.2SG the your backpack Xulia – ‘And you are not taking your backpack?’ Noelia – Que dis? what say.PRS.2SG Noelia – ‘What are you saying?’ a.

Noelia – A  mochila azul? Non. Perdina the  backpack blue no lose.PST.1SG-CL.ACC.F.3SG na praia hai un ano. in_the beach haber.PRS.3SG one year

b. Noelia – A azul mochila? Non. Perdina the blue backpack no lose.PST.1SG-CL.ACC.F.3SG na praia hai un ano. in_the beach haber.PRS.3SG one year Noelia – ‘The blue backpack? No. I lost it at the beach a year ago.’ c.

As dúas posibilidades. the two possibilities Both possibilities.

These questions served not only as distractors, but also to ensure that participants completed the tasks properly. It is important to point out that the word order options in conditions A through E do not incur sharp ungrammaticality depending on the choice of clausal word order; however, a choice of AdjectiveNoun order, as in (150b), does. Three different reply contexts were included: two in which only Adjective-Noun order was possible, two in which only NounAdjective order was possible, and one in which either Noun-Adjective or Adjective-Noun order was possible. If a participant chose an ungrammatical response, their questionnaire results were highly scrutinized, as it was taken as a potential indication that they did not complete the task(s) in a faithful or mentally-focused manner.

136   

   Methodology

3.4.7 Summary of quantitative experimental tasks In Table 16, I summarize the information structure contexts used in the experimental tasks designed to gather quantitative data on word order acceptability (task 1) and word order preference (task 2). Table 16: Summary of information structure contexts appearing in tasks one and two.

Task 1

code Task 2

code

thetic common-ground subject (subord.) subject narrow-focus object narrow-focus common-ground object (subord.)

A B F G C

A B C D E G

common-ground subject (coord.) common-ground object (coord.)

D E

thetic common-ground subject subject narrow-focus object narrow-focus common-ground object N-Adj. order

As we can clearly see, these tasks do not entirely overlap in their coverage. Task 2 was not designed to gather preference data on subordinating versus coordinating discourse, and Task 1 did not include distractor stimuli. In the following section, I describe the qualitative data gathered during this investigation in Galicia in 2008.

3.5 Task 3: Recorded field interview Researchers in sociolinguistics have found that obtaining linguistic evidence from minority, non-prestige varieties presents unique challenges. Cheshire and Stein (1997) claim that the “fluid” nature of non-prestige varieties makes grammaticality judgments a difficult task, and that only standardization and the establishment of a grammar make such judgments possible. However, Henry (2005) points out that such ‘standard forms’ of a variety can also cloud the matter since speakers may consider certain standardized forms or uses to be ‘incorrect’ in their variety, and therefore ‘ungrammatical’. When such standardized forms are reinforced by an education system, it can lead to otherwise grammatical structures being highly stigmatized. Henry’s (2005) examination of Belfast English found that follow-up interviews with participants uncovered subtleties in acceptability judgments from linguistic questionnaires that otherwise would have been overlooked. She suggests, therefore, that linguistic questionnaires alone often cannot provide a complete picture of a speaker’s grammar. She notes that paper-and-

Task 3: Recorded field interview    

   137

pencil questionnaires typically make use of a standard variety of a minority language which is often unfamiliar to older or less literate speakers of non-prestige languages. As a result, such individuals are frequently not receptive to written questionnaires. This is an important consideration because, in the Galician context, these individuals represent the nearest approximation to monolingual speakers of the language (Siguán 1992). The third task consisted of 19 field interviews recorded with an Olympus DS-40 digital voice recorder connected to a Crown flat microphone. Although participants volunteered to be interviewed, the flat microphone was used to lessen the potential of recording anxiety. A small subset of the interviewees also participated in Tasks 1 and 2, but since anonymity was preserved, it is unknown which of the interviewees they were. A Galician-speaking male conducted seven interviews in Vigo, and a Galician-speaking female conducted 11 interviews in Louredo (Mos), a rural village roughly 45 minutes from Vigo. A third Galicianspeaking female conducted the lone interview recorded in Pidre, a small village located roughly 3.5 miles from the city of Pontevedra. Galician-speaking individuals conducted the interviews in order to guarantee comprehension on the part of the interviewer, and to avoid simplification of linguistic structures for the benefit of the author. Although the author and Primary Investigator (PI) was present for all of the recordings, he did not conduct the interviews. Given that the PI’s participation was minimal during the recordings, he had familiarized the interviewers with the questions and expectations of the interview beforehand. The interviewees were acquaintances, family or friends of all the interviewers, which greatly assisted in gathering more natural, relaxed, and informal data. The interview questions for task 3 included questions about family, hometown, opinions about older and younger generations, opinions about the Galician language, and thoughts about the future of Galicia (see Appendix D for the full text of these questions). All participants were also asked to tell an anecdote from their youth or from their hometown. The older participants were additionally asked about their experiences during the transition to democracy after the death of Francisco Franco, former dictator of Spain. Task  3 was designed and structured to be as similar as possible for all interviewees in an attempt to ensure analogous, and therefore comparable, responses. The goal of these short interviews was also to gather a qualitative sample of spontaneous speech with which to compare the results of the quantitative tasks introduced above.

138   

   Methodology

3.6 Conclusion The methodology detailed in the three tasks detailed above solicited and gathered indications of word order acceptability, preferences, and usage in Galician in a variety of pragmatic contexts. Bearing in mind the above limitations and difficulties in eliciting grammaticality judgments from native speakers of non-prestige minority languages, great care was taken in designing the tasks in this chapter. All of the tasks detailed above were created with the consultation and advice of native Galician speakers working in sociolinguistics and Galician education in order to assure that these tasks would focus more on Galician speakers’ estimations of what is pragmatically acceptable and appropriate, and less on what is lexically appropriate.¹¹⁰ Statistical results from tasks 1 and 2, which appear in the following chapter, indicate the clausal preferences of Galician speakers for these contexts. The spontaneous production data gathered in task 3 help to shed further light on how common such word orders are in the elicited speech of interviewees. A crucial generative linguistic assumption that I also make with respect to the recorded corpus is that a lack of certain word orders should not be taken as an indication that such word order(s) are lacking in the grammars of these speakers. The combination of the indications gathered in these tasks provide vital qualitative and quantitative evidence guiding the syntax-information structure analysis that I propose for Galician in Chapter 5.

110 Note that the Galician speakers who advised me in this capacity also piloted the Internetbased tasks and conditions. While they did not provide ratings to the stimuli in tasks 1 and 2 (thus precluding the possibility of a true pilot test), they were exhaustive in their commentaries and suggestions with respect to the discourse contexts and response options.

4 Statistical analysis of quantitative and qualitative measures 4.1 Introduction In this chapter, I present the results of the quantitative and qualitative tasks described in Chapter 3. In section 4.2, I present the statistical results for each Task 1 condition. In section 4.3, I present the statistical results for each Task 2 discourse condition. In section 4.4, I report on a follow-up quantitative task that I carried out for two conditions whose results in Task 2 did not indicate a clear word order preference. I describe the follow-up task and present the statistical results gathered for this task. I report on the word orders attested and their accompanying discourse contexts in the recorded field interviews in section 4.5. In section 4.6, I make closing comments on this chapter and the tasks reported on within. The number of participants in the quantitative tasks described below did not approach or surpass the critical number of 30 required to analyze the data with standard parametric statistics (Student 1908). Additionally, initial data inspection for both tasks indicated that participant responses were not normally distributed. Rather, they were quite skewed. The five-point rating scale by which the word order options in Task  1 were rated for appropriateness is ordinal and not scalar (i.e. a rating of 4 is not necessarily twice as high as a rating of 2). The choices available in Task 2 were also ordinal, as they only provided a word order preference choice of a, b, or c. Therefore, when inter-group comparisons can be made, I have analyzed the data using the Friedman test, which is a non-parametric statistical alternative to either an ANOVA or a two-tailed t-test. In section 4.2, I detail the statistical results for Task 1. Sections 4.2.1 through 4.2.7, I provide descriptive statistics for the word order triads for each condition, and then provide statistical comparisons of the word order responses for each triad. I summarize these results in 4.2.8. In section 4.2.9 I provide statistical comparisons for Task 1 conditions by dominant language.

4.2 Task 1 Recall that Task 1 was the Appropriateness Judgment Task, in which participants rated continuation/response sentence triads to a variety of information structure contexts on a five-point Likert scale. For each condition in Task 1, I present descriptive statistics and discussion below. I follow the data descriptions with Friedman ranks of means measures to discover statistical differences. Additional comparisons follow when significant differences are detected.

140   

   Statistical analysis of quantitative and qualitative measures

4.2.1 Condition A For thetic contexts, SVO word order received a mean acceptability rating near ceiling (4.91), as reported in Table 17. In comparison with the mean ratings for VSO (3.14) and VOS (2.63), these statistics suggest that SVO is the preferred response word order. The ratings for VSO and VOS display greater individual variation than SVO (minimum 4, maximum 5), as both received a range of ratings as low as 1 and as high as 5. Table 17: Descriptive statistics, Task 1 Condition A (thetic sentences)

word order N

Mean

std. dev.

min.

max.

SVO VSO VOS

4.91 3.14 2.63

0.28762 1.08265 1.26055

4 1 1

5 5 5

100 100 100

Not surprisingly, the small range of variation in ratings for SVO is reflected in its small standard deviation (0.28762). Despite the wide range of ratings given for VSO and VOS, they do not exhibit very large standard variations (1.08265 and 1.26055, respectively). Table 18: Friedman statistics, Task 1 Condition A (thetic sentences)

word orders

N

χ2

df p-value

SVO v VSO v VOS SVO v VSO SVO v VOS VOS v VSO

100 100 100 100

145.531 87.044 93.000  7.667

2 1 1 1

< 0.001 < 0.001 < 0.001 0.006

The p-value of the Friedman non-parametric test ranking the means for all three word orders (p