The Syntax-Pragmatics Interface in North-Eastern Italian Consequences for the Geometry of the Left Periphery

A Thesis submitted to the University of Manchester for the degree of Doctor of Philosophy in the Faculty of Humanities

2018

Simone De Cia School of Arts, Languages and Cultures

Table of Contents

LIST OF TABLES 5

LIST OF FIGURES 5

LIST OF ABBREVIATIONS 6

ABSTRACT 7

DECLARATION 8

COPYRIGHT STATEMENT 9

DEDICATION 10

ACKNOWLEDGMENTS 11

1. INTRODUCTION 12

1.1. Theoretical and Empirical Scope 12

1.2. The Interplay between Information Structure and Syntax 17

1.3. Principal Claims 21 1.3.1. Wh-Typology and the Microparametric Variation in the Locus of SCLI 21 1.3.2. Residual V2 and Pragmatically Motivated V2 Effects 23 1.3.3. Root Contrastiveness as a Supra-Informational Status 26

1.4. Data Collection: Sample, Methodology and Database 27

1.5. Conventions for Glosses, Translations and Orthographic Representations 34

1.6. Organisation 37

2. NORTH-EASTERN : AN OVERVIEW 39

2.1. The Northern Italian Dialects 39

2.2. NEIDs: Overview and Genetic Affiliation 43 2.2.1. Friulian and Fornese 46 2.2.2. Lamonat and Sovramontino 49

2.3. : The Ambiguous Status of NIDs as Null-Subject Languages 54 2.3.1. Subject Clitics and the Null-Subject Parameter 55 2.3.2. SCLs across NEIDs: Preliminary Considerations 56 2.3.3. SCLs in Fornese and Sovramontino: Empirical Generalisations 57 2.3.3.1. SCLs in Declarative Clauses 59

2 2.3.3.2. SCLs in Negative Declaratives and Satellite Clitics 61 2.3.3.3. The Morpho-Phonetic Form of SCLs 64 2.3.3.4. SCLs in Root Interrogatives 65 2.3.3.5. SCLs in Root Negative-Interrogatives 68 2.3.4. SCLs in Different Syntactic Contexts 70 2.3.4.1. SCLs in Different Syntactic Contexts: Fornese and Sovramontino 70 2.3.5. The Morpho-Syntactic Status of SCLs: Further Assumptions 75

2.4. Conclusion 81

3. MICROPARAMETRIC VARIATION IN THE LOCUS OF SUBJECT INVERSION: APPARENT WH-IN-SITU 82

3.1. Introduction 82

3.2. Bellunese Wh-In-Situ: Existing Analyses 85

3.3. The Behaviour of Wh-Words 94 3.3.1 Apparently In-Situ Wh-Words 95 3.3.2 “Canonical” Wh-Words 101 3.3.3 On the Nature of Wh-Clitics 107 3.3.4 Cleft-Questions, Clitic-Wh-Questions and Tonic-Wh-Questions 115 3.3.5. Preverbal Tonic Wh-Elements: A D-Linked Interpretation 120

3.4. Microparametric Variation in the Locus of SCLI 121

3.5. Conclusion 129

4. FOCUS STRUCTURE, PRAGMATICALLY MOTIVATED V2 EFFECTS AND SUBJECTHOOD 131

4.1. Introduction 131 4.1.1. Overview and Structure 133

4.2. Focus Structure 138 4.2.1. Narrow Focus 139 4.2.2. Subjects in Narrow Focus: Lamonat vs. Sovramontino 141 4.2.2.1. Accommodating Focal Subjects: Object Topicalisation 146 4.2.2.2. Accommodating Focal Subjects: Cleft Structures 147 4.2.2.3. Subjects of Core Unaccusatives 150 4.2.2.4. Subjecthood in Lamonat and Sovramontino 151

4.3. The V2 Nature of Lamonat and Sovramontino 157 4.3.1 V2 Diagnostics 157 4.3.1.1. What Moves? The preverbal XP and the Clitic Status of Auxiliaries 158 4.3.1.2. Subject Inversion 163 4.3.1.3. Linear restrictions: V1 and V3 164 4.3.1.4. The Root vs. Embedded Asymmetry 168 4.3.2. The Typology of V2: High vs. Low V2 Languages 170 4.3.2.1. Lamonat and Sovramontino: A High-V2 System 171 4.3.3. Comparing and Contrasting Romance V2 in Synchrony and Diachrony 180

3 4.3.4. Informational Narrow Focus: Against an In-situ Analysis in Lamonat and Sovramontino 187

4.4. The Pragmatically Motivated V2 Mechanism 191

4.5. Three Focal Positions: Friulian Marked Informational Focus 204

4.6. Broad Focus and SVO 209

4.7. Conclusion 215

5. ROOT CONTRASTIVENESS AND V2: A SUPRA-INFORMATIONAL STATUS 216

5.1. Introduction 216 5.1.1. Contrastiveness and V2: An Overview 217 5.1.2 A Unified Treatment of Contrastive XPs: Cross-linguistic Evidence 218

5.2. The Data: Contrastive XPs in Lamonat and Sovramontino 220

5.3. The V2 Nature of Lamonat and Sovramontino 224

5.4. The Preverbal XP: The Contrastive Field 225 5.4.1. The Hierarchy of Contrast 225 5.4.1.1. Dominant Contrast 226 5.4.1.2. Membership in a Finite Set of Entities 227 5.4.1.3. Highlighting 229 5.4.2. Topical and Focal Contrastiveness: The Revised Hierarchy of Contrast 231

5.5. The Case of D-Linked Wh-Elements 234

5.6. Conclusion 238

6. CONCLUSION 239

6.1. Summary of Findings 239

6.2. Challenges, Future Objectives and Concluding Remarks 242

BIBLIOGRAPHY 244

WORD COUNT: 76, 276

4 List of Tables

1.1 Participant distribution per speech community. 29

2.1 Comparison between Friulian’s and Sovramontino’s plural nominal 51 and recostruction of Proto-Sovramontino’s plural morphology prior the loss of word-final –s. The masculine is ‘dog’ and the feminine noun is ‘chair’.

2.2 Tonic and atonic in Fornese. 57

2.3 Tonic and atonic pronouns in Sovramontino. 58

2.4 Renzi and Vanelli’s (1983) NID categories on the basis of the distribution of 59 obligatory SCLs across the different persons of the .

2.5 Negation and SCLs in Sovramontino and Fornese. 64

2.6 Atonic pronouns in Sovramontino. 66

2.7 Atonic pronouns in Fornese. 67

2.8 Present indicative conjugation of the verb ‘to eat’ in Friulian. 78

2.9 SCL forms vs. verb inflection forms in the relevant grammatical persons in 78 Sovramontino and Fornese.

3.1 Affirmative and interrogative SCLs in Lamonat and Sovramontino. 87

3.2 The internal makeup of the tonic wh-word aonde in Lamonat and Sovramontino 104 according to Munaro and Poletto (2014).

3.3 Tonic and clitic wh-elements in Lamonat and Sovramontino. 105

5.1 The typology of contrastive elements in Lamonat and Sovramontino with respect 231 to Molnár’s (2002, 2006) revised discourse-functions of contrastiveness.

5.2 Revised Hierarchy of Contrast for Lamonat and Sovramontino. 232

List of Figures

2.1 Geographical distribution of the four NEIDs under investigation and their 41 neighbouring Romance varieties.

4.1 Pitch contour of the root interrogative in (58) featuring an apparently in-situ 189 wh-element 4.2 Pitch contour of the root declarative in (60) featuring an XP in informational 189 narrow focus.

5 List of Abbreviations

CL clitic COND conditional DAT dative EXIST existential EXPL expletive F feminine FOC focal Frl. Friulian GER gerundive IMP imperfective Lam. Lamonat LOC locative M masculine NEG negation NEID North-Eastern Italian NID Northern Italian Dialect OBJ object OCL object clitic PRN PL plural PTCP past PST past REFL reflexive SCL subject clitic SCLI subject clitic inversion SG singular Sovr. Sovramontino SUB subjunctive SUBJ subject

6 Abstract

This thesis investigates a range of phenomena occurring at the syntax-pragmatics interface in four North Eastern Italian Dialects (henceforth NEIDs). The pragmatically motivated V2 system of Lamonat and Sovramontino provides valuable insights on the theoretical debate on the syntax-pragmatics interface, specifically: i) the make-up of the left periphery as the syntactic space that encodes discourse-pragmatic information cross-linguistically, and ii) the status of contrastiveness as a discourse-pragmatic notion. With respect to the first point, Lamonat and Sovramontino left peripheral space is less articulated or, put another way, more constrained than that of Friulian and Fornese, namely the other NEIDs under investigation in this thesis. The V2 constraint dictates certain restrictions in the make-up of the CP space: for example, the unavailability of clitic left dislocation of non-contrastive topics. In discourse-pragmatics terms, the left periphery seems to privilege elements that concern the management rather than the content of the interlocutors’ common ground. Despite the allegedly language-specific make-up of the C-domain, the actual left peripheral functional projections and their hierarchical order seem to be candidates for universal features of human language. As far as contrastiveness is concerned, Lamonat and Sovramontino show that it should be regarded as a complex discourse-pragmatic notion which is orthogonal to the topic-focus divide. More specifically, it should not be considered to be a discrete notion, but a continuum, in which the degree of contrastiveness is determined by the properties of the set containing the contrastive element with respect to the size and accessibility of alternatives. The present account offers a straightforward alternative for the analysis of wh-in- situ in Bellunese, whereby non-canonical constituent order in syntax is dictated by discourse-pragmatic needs. The present work also provides a new perspective on V2 as a syntax-pragmatics interface phenomenon: in the V2 system of Lamonat and Sovramontino, T-to-C movement is dictated by the discourse-pragmatic saliency of the fronted element. In fact, verb movement without the fronting of a salient XP in the C- space does not take place. Notwithstanding its discourse-pragmatic specialisation, the V2 system of Lamonat and Sovramontino is very likely a direct continuation of Medieval Romance V2. Finally, the V2 syntactic-pragmatic constraint of Lamonat and Sovramontino is not found in Friulian and Fornese. In these two NEIDs, residual V2 is in fact only manifested in T-to-C movement in root interrogatives and narrow focal element-verb adjacency in root declaratives.

7 Declaration

I declare that no portion of the work referred to in the thesis has been submitted in support of an application for another degree or qualification of this or any other university or other institute of learning.

8 Copyright Statement

The author of this thesis (including any appendices and/or schedules to this thesis) owns certain copyright or related rights in it (the “Copyright”) and s/he has given The University of Manchester certain rights to use such Copyright, including for administrative purposes.

Copies of this thesis, either in full or in extracts and whether in hard or electronic copy, may be made only in accordance with the Copyright, Designs and Patents Act 1988 (as amended) and regulations issued under it or, where appropriate, in accordance with licensing agreements which the University has from time to time. This page must form part of any such copies made.

The ownership of certain Copyright, patents, designs, trademarks and other intellectual property (the “Intellectual Property”) and any reproductions of copyright works in the thesis, for example graphs and tables (“Reproductions”), which may be described in this thesis, may not be owned by the author and may be owned by third parties. Such Intellectual Property and Reproductions cannot and must not be made available for use without the prior written permission of the owner(s) of the relevant Intellectual Property and/or Reproductions.

Further information on the conditions under which disclosure, publication and commercialisation of this thesis, the Copyright and any Intellectual Property and/or Reproductions described in it may take place is available in the University IP Policy (see http://documents.manchester.ac.uk/DocuInfo.aspx?DocID=2442 0), in any relevant Thesis restriction declarations deposited in the University Library, The University Library’s regulations (see http://www.library.manchester.ac.uk/about/regulations/) and in The University’s policy on Presentation of Theses.

9 Dedication

This thesis is dedicated to Anne Farrell, without her early inspiration and enthusiasm I would have never become a linguist.

10 Acknowledgments

First, I would like to thank my supervisors Prof. Delia Bentley and Dr. Julio Villa- García for their help, guidance and feedback in the construction of this thesis. In the last two years of my Ph.D. programme, Dr. Julio Villa-García has been an excellent mentor: he taught me how to ‘think syntactically’ within generative syntax and made a ‘captive audience’ for my moments of syntactic euphoria. I have no words to express the sense of gratitude that I feel towards Prof. Delia Bentley. In the past eight years, she has been a role model and solid reference point in my life. I also want to thank my former co-supervisor Dr. Laurel MacKenzie, particularly for her guidance and help with data management. I am also very grateful to Prof. Eva Schultze-Berndt for her constructive criticism and frank feedback throughout my Ph.D. programme. Second, I would like to thank the speech communities of , Lamon and Sovramonte. Informants literally welcomed me in their homes and put up with my time-consuming interviews. In the community of Forni di Sopra, I am especially grateful to Alfio Anziutti for his help with the recruitment of informants, and Rossella Fachin for sharing her unpublished materials on Fornese. I also want to thank all the Fornese speakers that I interviewed, in particular Danilo, Reli and Cami. Special thanks go to Genni Sacchetti for her help in my pilot investigation of Fornese. In the community of Lamon, I want to thank the former council member Stefano Facchin for his irreplaceable help with the recruitment of informants. A big thank you goes to the local folk association ‘Drio le peche’ and its president Emma Gaio Maillard. I would like to thank all the Lamonat speakers who took part in the interviews, especially Maleta, Bacan, and Molina. In the community of Sovramonte, I would like to thank the Mayor Federico Dalla for his help in setting my fieldwork in motion. I would also like to thank Luciano Reato and the Pro-Loco Sovramonte for their help with the recruitment of informants. I am especially grateful to the family Morosoche for actively taking part in the interviews, namely Lorenzo, Giuliano, Luigino, Antonietta and Cirio. A huge thank you goes to my very own grandmother Elsa Facchin for being a particularly patient speaker of Sovramontino. I am also grateful to Pierin and Giamba who impeccably represented the tiny village of Aune. Finally, many thanks to the Friulian speakers in my hometown who gladly accepted to be interviewed, above all Gabriella Fabbro, Luisa Tami and Fausto Bertossi. Third, I want to thank all my PhD colleagues for their support and inspiration, particularly Hannah Booth, Stefano Coretta, Victoria Stampone and Jessica Iubini- Hampton. A big thank you goes to all the friends that I made along the way, especially Elena Spagnuolo, Laura Penna, Federica Coluzzi and Fabiana Marinaro. Special thanks goes to Stefano Locatelli, Ramona Ciuciani and Giulia Molteni for four unforgettable years at The Poplars. I am also tremendously grateful to all the friends who have always been there for me in these four years, in particular Franco Tosolini, Loreley Hann- Herrera, Maria Merelli, Cristina Fabro and Laura Percival. Finally, I am very grateful to Giovanni Villa for his patience and support. Last but not least, a special thanks goes to my parents, Marcella Fusillo and Giovanni De Cia, for their caring presence and support during my doctoral studies (unfortunately, they still struggle understanding what I do, but I guess it is common among linguists!). 11 1. Introduction

1.1. Theoretical and Empirical Scope

The primary aim of this thesis is to contribute to the study of the syntax– pragmatics interface in Romance through the fine-grained investigation of two understudied and closely related Northern Italian Dialects (henceforth abbreviated NIDs), Lamonat and Sovramontino: I will try to uncover the microparametric variation that characterises syntax-pragmatics interface phenomena within the Northern Italo- Romance family (Maiden & Parry 1997). To this end, I will compare and contrast Lamonat and Sovramontino with two other North-Eastern Italian Dialects (henceforth abbreviated NEIDs), Friulian and Fornese. The goal of the present project is to formalise the interplay between information structure and syntax within the cartographic framework (Cinque 1999, 2002; Rizzi 2004). More specifically, I will analyse the different syntactic strategies that these NEIDs employ to accommodate discourse-pragmatics into their syntax. The investigation of elements in narrow focus (in the sense of Lambrecht 1994) will be of paramount importance. This includes contrastive foci, wh-elements and clefted focal elements. The strategies that Lamonat and Sovramontino employ to incorporate narrow focus into the syntax suggest that: (i) some NIDs have retained a pragmatically motivated V2 system that dictates T-to-C movement with the internal merge (i.e. move) of a pragmatically salient constituent. This can be superficially appreciated in the Sovramontino sentence in (1) where the verb is fronted with the movement to the C-domain of the focal element na schafa, ‘a slap’:

(1) La ge a dat NA SCHAFA la so femena (Sovr.) 3SG.F.SCL DAT.CL have.3SG give.PTCP a slap the his wife

al so om to-the her husband ‘The wife gave a slap to her husband’

(ii) the syntactic-pragmatic behaviour of some NIDs is more in line with non-null- subject , like French and some varieties of Gallo-Romance (see Oliviéri 2004; Olviéri & Sauzet 2016), than null-subject Romance languages like Spanish and Standard Italian, as shown by the realisation of the Lamonat interrogative

12 in (2), where the subject chi, ‘who’, is interrogated and the question is only grammatical if the wh-subject is clefted:

(2) E-lo chi che dis mastela a motor? (Lamonat) be.3SG-3SG.M.SCL who that say.3SG basin with engine ‘Who says motorised basin?’

Both claims will be discussed in this thesis. Since NIDs have been traditionally considered null-subject and non-V2 languages, the two claims are relevant to an important theoretical point that has been debated since the onset of the study of Chomskian generative syntax: the mechanism of parameter (re-)setting (Holmberg & 2009). Note that, in the present thesis, I will often refer to (micro)parametric variation, but I will not employ parameters in their original forms (Chomsky 1981): I will instead assume features (Chomsky 1995) and use the term parameter as a purely descriptive label, namely the variation with respect to a specific syntactic property of a language. In this spirit, in Lamonat and Sovramontino, the V2 Parameter (see Bidese and Tommaselli 2007 for V2 in Cimbrian) has a negative value, but T-to-C movement takes place in root declaratives in tandem with the movement of a pragmatically salient constituent to the C-domain. Similarly, the Null-Subject Parameter (Chomsky 1981) has a positive value, but we will see that the NEIDs under investigation exhibit expletive- like elements, obligatory phonologically realised pronominal elements and a relatively rigid ban on subject fronting. The study of syntax-pragmatics interface phenomena allows us to gain a further theoretical and empirical insight on this longstanding debate: the explanation to these syntactic discrepancies, in fact, must be found in the interaction between the core parameters of a language and their dependent secondary micro- parameters, whose positive or negative values determine the licensing of different syntactic-pragmatic constructions. For instance, the licensing of focus-fronted subjects is contingent on the micro-parametric variation that regulates subject displacement, which, in turn, contributes to the positive or negative value of the Null-Subject parameter. Along the same lines, the overall value of the V2 parameter is determined by secondary micro-parameters that characterise the V2 phenomena, which are often detectable from a syntactic-pragmatic angle. For example, in a V2 system, which type of discourse-pragmatically salient constituents can be preposed in preverbal position depends on the micro-parametric variation that dictates which left-peripheral projection is targeted by V2 T-to-C movement. The data from NEIDs seem to support Bidese and

13 Tommaselli’s (2016, 2018) model of parameter (re)-setting: (i) if a core parameter loses its positive value, secondary micro-parameters are subsequently lost at a later stage; (ii) if a core parameter acquires a positive value, secondary micro-parameters have already been acquired at a previous stage. The data from Lamonat and Sovramontino seem to support this model with respect to: (i) the loss of Medieval Romance V2 (Benincà 1983; 2006) and residual V2 phenomena linked to the syntactic encoding of discourse– pragmatic information; (ii) the status of NIDs as non-consistent null-subject languages (Cardinaletti & Repetti 2010), which posit syntactic constraints on the realisation of subjects in narrow informational focus.

I will demonstrate that this microvariation can be constrained by the study of the syntax-pragmatics interface, more specifically by the investigation of the properties of the left periphery (Rizzi 1997), which is the portion of the clause that most prominently hosts discourse-pragmatic information. The present research aims to capture how micro- parametric variation is encoded in Rizzi’s (1997) split-CP model and what the consequences are for its geometry. The study of syntax-pragmatics interface phenomena will, in fact, provide us with further insights into the syntactic mechanisms at work in the CP-domain. The claims that I will put forward regarding the geometry of the left periphery must be interpreted within the cartographic framework (Rizzi 1997; Cinque 2002; Rizzi 2004). The present thesis assumes two key theoretical points which are fundamental to the cartographic framework1: (i) movement to the left periphery must be triggered by interpretative or discourse-related outcomes; (ii) functional heads within the C-domain are projected only when activated (see also Rizzi 2001b). In the next section, I will outline Rizzi’s (1997) split CP. Overall, I will show that the left periphery of these dialects is more constrained and, consequently, less rich than that of canonical pro-drop languages. More specifically, it can be reduced to two functional projections: ForceP, which traditionally encodes the illocutionary force of the clause (e.g. declarative or interrogative), and FocusP, which is the projection hosting focalised constituents (see Rizzi 1997). Nonetheless, I will argue that different features and null affixal-heads (a là Bošković 2001) spread from the aforementioned functional projections. For example, in Lamonat and Sovramontino, I will not argue in favour of a dedicated position for canonical wh-elements (e.g. Rizzi’s 2001b, 2013 IntP), but I will

1 The following principles are to some extent also pivotal in the Minimalist Programme (Chomsky 1995). Chomsky (2005) himself does not deny Rizzi’s (1997) model, acknowledging 14 still postulate the satisfaction of the [WH] feature in the left peripheral head hosting the wh-element. In addition to ForceP and FocusP, I will argue in favour of a high clause- external left-peripheral position, namely FrameP, which hosts frame-setting elements: XPs that provide temporal and/or modal restrictions to the circumstances of evaluation of the proposition expressed by the main clause (Haegeman 2000, 2007; Benincà & Poletto 2004; Poletto 2002) and can be regarded as independent speech-acts (Krifka 2001). Ultimately, the data from the NIDs under investigation suggest that Rizzi’s (1997) split-CP template is subject to (micro-)parametric variation: its make-up is language specific and, most crucially, shaped by the interaction between information structure and syntax. By the same token, the present research also suggests that the functional projections that make up the left periphery and their hierarchical order are good candidates to be universal features of human language.

As previously mentioned, for the purpose of the present investigation, I will focus on four NIDs that are located in the northeastern part of , namely Friulian, Sovramontino, Lamonat and Fornese. Lamonat and Sovramontino, being more syntactically distant from Standard Italian, will receive particular attention. Even though these four dialects are not fully representative of this linguistically rich and diverse part of Italy (see Maiden & Parry 1997), I will refer to them as North Eastern Italian Dialects (abbreviated NEIDs). In chapter 2, I will outline their main characteristics within the Romance varieties spoken in the . The present work aims to shed more light on the status of NIDs as a whole; I also hope to point out the micro-parametric variation across different NIDs with respect to the encoding of syntax-pragmatics information. From a generativist perspective, NIDs are optimal candidates for the study of the language faculty as a whole. In fact, they are non-codified language systems, which spontaneously developed from Vulgar . If we want to describe the properties of Universal Grammar as postulated by Chomsky (1986, 1995), these Romance varieties offer a valuable insight. This is especially true with respect to the investigation of the wh-parameter and the null-subject parameter (Chomsky 1981) along with the aforementioned resetting of the V2 parameter. In fact, not only do Lamonat and Sovramontino exhibit an ambiguous behaviour with respect to the subject parameter and residual V2 phenomena, but also wh-fronting does not superficially take place in root interrogatives. I will argue that the mechanisms behind these phenomena can be uncovered by investigating the syntax-pragmatics interface and can be accounted for

15 within the cartographic framework.

As for the empirical scope of the present research, it is worth noting that all the NIDs under investigation are vulnerable in terms of their language vitality (UNESCO 2003, Grenoble and Whalley 2006). The secondary goal of this research is hence to document part of the language system of these Romance varieties, before the on-going process of obsolescence definitely compromises their characteristic distinctive features. Intergenerational transmission of these Romance varieties is extremely weakened, leading to their very likely disappearance within the timeframe of the next two generations. As a consequence, the current younger generations of speakers are mainly receptive bilinguals and struggle in the active production of language, privileging the use of their regional Italian2. The pressure of Italian is very heavy on these NIDs: the current younger generations of speakers can hence be defined as semi-speakers (in the sense of Dorian 1977) of these indigenous Romance varieties. As Berruto (1987) points out, the linguistic situation in Italy is characterised by the mutual interaction of the dialects of Italy (sister languages of Italian) and Italian, which affects all the levels of the language system. Nonetheless, in more recent years, it is increasingly evident that Italian has greatly influenced the dialects of Italy especially among the younger generations, increasingly levelling their linguistically characteristic traits. Italian spread to the population of the Italian peninsula in the 60s, mainly with the introduction of compulsory education and the spread of mass media (De Mauro 1991). Berruto’s (1987, 1989) dilalia is the best term to describe the sociolinguistic situation in Italy: a type of diglossic situation (Ferguson 1959) whereby the high language, namely Italian, is not only spoken in formal language domains, but it is also used in informal domains alongside the low language, namely the dialects of Italy. The current linguistic situation is nonetheless unstable and will result in the loss of the low languages. Thus, the documentation of these language systems is perhaps the main empirical scope of this thesis: provide novel data that will enable us to enrich our current field of research. In fact, NIDs exhibit linguistic phenomena that deepen our understanding of the spectrum of possible linguistic structures in Romance. Given the importance of these first-hand data, I aim to make my data as accessible as possible to other researchers for future investigations. For this purpose, I will first try to introduce each linguistic phenomenon

2 By , I mean different varieties of Italian concerning the diatopic/geographic dimension of variation (see Berruto 1993) 16 under investigation in a framework-neutral way and only subsequently will I delve into the theoretical analysis of it. I hope this modus operandi will make the present research also appealing to those linguists working within different frameworks.

The last ambitious question that I will posit at the end of this work is: what is the driving force that triggers syntactic change? I will not offer a satisfactory answer, but, based on this work, I will make a strong case for discourse-pragmatics as a co-occurring force that determines syntactic change. Research on syntactic change has identified in the weakening the morphology of a language the cause of the rigidification of its constituent order (e.g. Latin vs. Italian). We will, however, see that, despite the strong correlation between a rigid syntax and a poorer morphology, morphology alone is not enough to account for syntactic change. If we take Friulian as an example, this language exhibits a very rich morphology within the Romance family, yet its syntax is more rigid than that of Italian. Surely, syntactic change is not the result of a single linguistic factor, but I claim that discourse-pragmatics and, more specifically, the way in which a language unpacks information has a significant impact on the syntactic structures that are licensed in that language.

In the next sections of this introductory chapter, I will further discuss each theoretical and empirical point that I have raised so far. By doing so, I will first condense the major claims put forward by the present research and stress their importance with respect to the overall scope of this thesis. Second, I will engage in the discussion of language documentation and the practices that accompany it, providing a transparent window on how I collected the data that will be presented throughout the thesis.

1.2. The Interplay between Information Structure and Syntax

In the last two decades, the interplay between information structure and syntax has been extensively investigated. Linguists have tried to capture and represent syntax- pragmatics interface phenomena in a of different frameworks (see Lambrecht 1994; Kiss 1998; Van Valin 1999; Frascarelli 2000; Belletti 2004; López 2009 and many others). In Romance, much of the recent work on information structure revolves around challenging and refining the traditional notions of focus and topic. The present

17 project does not aim to investigate exhaustively information structure per se across the NEIDs, but the strategies that the syntax adopts to accommodate it and encode different discourse-pragmatic interpretations. Although some relatively recent work has pioneered this path (see Benincà & Poletto 2004 on NIDs; Cruschina 2006 and Bentley 2008 on Sicilian) more research on the Italian dialects is needed to gather a full picture of how syntax-pragmatics interface phenomena vary across Romance.

The interaction between information structure and syntax determines whether or not the output of the derivation of an utterance is pragmatically felicitous. The wrong configuration between syntax and information structure gives rise to a pragmatically infelicitous sentence, as shown in (3):

(3) CONTEXT: Who has arrived? (Italian) a. E’ arrivato MARIO be.3SG arrive.PTCP Mario ‘Mario has arrived’

b. #MARIO è arrivato Mario be.3SG arrive.PTCP ‘Mario has arrived’

(3b) is pragmatically infelicitous (hence marked with #) because the subject of the clause, namely Mario is focal (hence written in capital letters). (3a) and (3b) therefore fit different types of discourse: their use is not random, but constrained by specific discourse-pragmatic interpretations. In postverbal position, the subject is marked as focal, whereas, in preverbal position, it is marked as topical. The infelicitous outcome of (3b) is due to the fact that this sentence exhibits a subject in topical configuration, but the discourse-pragmatic context demands it be in a focal configuration. In order to gain a better understanding of the data in (3), let us briefly define the notion of topic and focus. The grammatical means that differentiates the assertion in an utterance from the presupposed or topical part of the same utterance is called the focus structure of the sentence. This label is primarily associated with the work of Lambrecht (1986; 1987; 1994; 2000). Focus refers to new or non-presupposed information that was prompted in discourse, as shown by Mario in the question-answer pair in (3). The topical part of the utterance or topic (Lambrecht 1994, adopted from Gundel 1988) is traditionally considered the presupposed-known portion of the sentence. In discourse-semantics, the notion of topic can be defined as the subject of the predication; in other words, what the

18 sentence is ‘about'. This information status is primarily associated with given or old information that is easily retrievable from discourse (Givon 1976, Lambrecht 1994, Kiss 1995, Erteschik-Shir 2007). As suggested in the example in (3), the syntax often finds a way to accommodate these different informational statuses, giving rise to what are known as syntax-pragmatics interface phenomena. Discourse functions like focus and topic are integrated into the speech act through information structure, which is, in turn, represented in the syntax as discourse representation structures (see Kamp and Reyle 1993). The syntax-pragmatics interface ultimately determines the felicitousness of an utterance in a given discourse-pragmatics context.

In the cartographic enterprise, the role of interface properties is encoded in an articulated system of functional projections (Cinque 1999; Rizzi 2004). EPP (or edge) features are inherently associated with the given functional heads and trigger obligatory movement in order to check and eliminate (or value) them to deliver a specific interpretative outcome (Chomsky 2001). The portion of the sentence that is responsible for encoding of discourse-pragmatics properties of an utterance is the left periphery or C-domain. Starting with Rizzi (1997), many generative studies on the left peripheral area of the clause, traditionally labelled CP, have shown its richness (Poletto 2000; Benincà 2001; Benincà and Poletto 2004; Poletto and Pollock 2004; Ledgeway 2005; Villa-García 2015 and many others). Most syntax-pragmatics interface phenomena involve this area of the clause, which is schematically represented in (4) below:

(4) ForceP

TopicP

FocusP

FinP

…

In the model of the left periphery as proposed by Rizzi (1997), the CP-space is made up of four functional projections in the following hierarchical order: (i) ForceP is responsible for the typing of of the clause in terms of its illocutionary force (e.g. interrogative or declarative); (ii) TopP hosts left-dislocated constituents in its specifier position; (iii) similarly, FocusP hosts focalised constituents in its specifier position; and finally (iv) FinP encodes information regarding the finiteness and the mood

19 specifications of the clause. The general idea is that an agreement relationship is established between the head of a phrase and its specifier. For instance, as for focal or topical readings, head and specifier agree with respect to a [FOCUS] or a [TOPIC] feature, which is responsible for the correct interpretation of the informational status of the given element. The interpretation as focus or topic of an element in the left peripheral space is automatically given by the element filling different functional projections. The relation between syntax and the pragmatic-semantic component responsible for the interpretation of the utterance is expressed in a simple way: the discourse-pragmatic interpretation of an element is the direct consequence of its syntactic configuration.

In the present research, I will try to uncover the mechanisms behind the interaction of information structure and syntax across NEIDs by investigating the pragmatic-syntactic constraints that are relevant to the CP-domain. I will hence try to formalise the interaction between discourse-pragmatic constraints and syntactic constraints, which generates the appropriate syntactic-pragmatic configurations and, ultimately, discourse-pragmatically felicitous outcomes. To this end, I will show that the adoption of focus and topic as primitive informational statuses is not sufficient to account for the interface phenomena investigated in this thesis: I will hence support a further internal differentiation between different types of focal and topical elements (Kiss 1998; Bianchi & Frascarelli 2010), which are encoded through different functional features.

Cross-linguistically, the left peripheral space of the clause is more or less fit for the encoding of discourse-pragmatic information. For example, in English the interaction between information structure and syntax has developed in a way that discourse-pragmatic functions cannot fully rely on the left peripheral area of the clause for their realisation, resorting to prosodic cues. This is due to the fact that SVO constituent order is relatively rigid in English and impedes the free realisation of pragmatically salient information in the C-domain: discourse-pragmatic information is hence realised in its canonical position through prosody. In Italian, on the other hand, the same discourse-pragmatic information can be realised in the left peripheral space of the clause, as the basic constituent order of the language is less rigid (see Van Valin 1999; Bentley 2008 for a comparison between English and Italian). In the case of the

20 NIDs, the accessibility of the left peripheral area to discourse-pragmatic information is subject to micro-parametric variation. In some NEIDs, like Lamonat and Sovramontino, the C-domain is more restricted due to an allegedly on-going rigidification of their basic SVO constituent order. Thus, as we will see in more depth in the next chapter, the hybrid status of NIDs as non-consistent null-subject languages (Cardinaletti & Repetti 2010) gives rise to peculiar syntactic configurations that reflect the syntax-pragmatics interface properties of these Romance varieties.

1.3. Principal Claims

Having set the theoretical and empirical scope behind the present research, let us now see what main claims are put forward in this thesis.

1.3.1. Wh-Typology and the Microparametric Variation in the Locus of SCLI

The first claim concerns the syntax of root interrogatives across the NEIDs. I claim that there is (micro)parametric variation in the locus of subject clitic inversion (henceforth abbreviated as SCLI). SCLI is the result of T-to-C movement (movement of the inflected verb), which is necessary for the well-formedness of root interrogatives (see Rizzi 1991), as shown by the root declarative in (5) and its interrogative counterpart in (6) below:

(5) Las boisas as mangia (Fornese) the.PL.F girl.PL.F SCL.3PL.F eat.3PL ‘The girls eat’

(6) Mangi-las las boisas? eat.3PL- SCL.3PL.F the.PL.F girl.PL.F ‘Do the girls eat?’

I claim that T-to-C movement across the NIDs is not homogeneous with respect to the functional head targeted in the C-domain. I formulated this hypothesis on the basis of the investigation of the distribution of wh-elements across these Romance varieties. Among the different NEIDs under investigation, we identified two subgroups of dialects that vary with respect to how high the verb moves in the C-domain in root interrogatives. I claim that T-to-C movement in Friulian and Fornese targets a low

21 functional projection in the left peripheral space, namely FinP, whereas, in Lamonat and Sovramontino, T-to-C movement targets a higher functional projection in the C-domain, namely ForceP. This microparametric variation is key in my research because it directly interacts with discourse-pragmatics. Syntax-pragmatics interface phenomena observe this microparameter determining specific configurations in the encoding of focal and topical elements in each subgroup of dialects. As I will claim in the next subsection, the (micro)parametric variation in the locus of SCLI is a reflex of a previous V2 stage of these languages during the Medieval Romance period (Benincà 1983; Benincà 2006).

I claim that the micro-parametric variation that is relevant to the C-domain of root interrogatives depends on which left peripheral head contains the silent particle Q (Baker 1970; Chomsky 2000) that is primarily responsible for the well-formedness of questions. Since Q is affixal in nature, it must have an overt host and attracts T to adjoin to it. Once the T head containing the clitic and the inflected verb moves to the target position adjoining to Q, SCLI takes place due to the morphological rearrangement of the null affix, the inflected verb and the SCL. The left peripheral functional head in which Q is licensed and SCLI takes place is what I labelled the locus of SCLI. In Lamonat and Sovramontino T-to-C movement targets a higher left peripheral projection, namely ForceP, whereas, in Fornese and Friulian the targeted functional projection is FinP. The underlying syntax-pragmatics interface phenomenon that results from this micro-parametric variation is the distribution of wh-elements. Across NEIDs, the position of wh-items depends on their morpho-syntactic status and discourse- pragmatic status. In Lamonat and Sovramontino, a wh-element can occupy three distinct syntactic positions as shown by the data in (7):

(7) a. Sa-a-lo fat Toni? (Lamonat) what-have.3SG-3SG.M.SCL do.PTCP Toni ‘What did Tony do?’

b. A-lo fat che Toni? have.3SG-3SG.M.SCL do.PTCP what Toni ‘What did Tony do?’

c. Che casa a-lo fat su Toni? What house have.3SG-3SG.M.SCL do.PTCP up Toni ‘Which house did Tony build?’

22 In the two NEIDs under investigation, discourse-pragmatically unmarked tonic wh-elements must appear postverbally, as in (7b), whereas discourse-pragmatically marked wh-elements such as wh-XPs that bear a D-linked interpretation (Pesetsky 1987) must appear preverbally, as in (7c). As for (7a), its discourse-pragmatic interpretation is equivalent to (7b), but its morpho-syntactic status makes it appear proclitically in preverbal position: sa- in (7a) is in fact a wh-clitic element. In the present research, I will argue that their distribution, along with the distribution of tonic wh-elements, is the result of the interaction between syntax and information structure: more specifically, the interplay between the micro-parametric variation in the locus of SCLI, the properties of narrow focal elements and discourse-linked XPs. Finally, with my investigation, I hope to shed light on the long-standing debate on the nature of wh-in-situ in Bellunese (Munaro 1999; Poletto & Pollock 2004; Poletto & Pollock 2009; Manzini & Savoia 2011).

1.3.2. Residual V2 and Pragmatically Motivated V2 Effects

I will claim that Lamonat and Sovramontino have retained a Medieval Romance- like V2 system (see Benincà 1983) that is key for the realisation of pragmatically salient constituents. I argue that the V2 constraint is activated if the C-domain is lexicalised: T- to-C movement hence takes place in root interrogatives and root declaratives that feature an XP in narrow focus and/or bearing a contrastive reading. In the syntactic model that I will assume, V2-motivated T-to-C head movement assures both the right pragmatic interpretation and the adequate phonological realisation of the informational units. The claim expands on the proposal that the microparametric variation in the locus of SCLI across NEIDs is determined by the V2 nature of Lamonat and Sovramontino, whereby V2 has been reinterpreted as a crucial syntactic strategy to encode salient pragmatic information. This line of research helps us to account for the pragmatic- syntactic differences that distinguish the two groups of NEIDs identified in this thesis: Friulian-Fornese, on one hand, and Sovramontino-Lamonat, on the other. I will consequently reach the conclusion that T-to-C V2-motivated head movement is what ultimately differentiates Lamonat and Sovramontino from Friulian and Fornese.

In Lamonat and Sovramontino, the V2 constraint is pragmatically motivated: the current V2 system has retained the pragmatic saliency of the preverbal position from

23 Medieval Romance and further developed to become a syntactic means of assuring linear adjacency with focal elements. As Salvi (2016) points out, the V2 system in place nowadays in some NIDs is different from that of Medieval Romance. The same is true for Lamonat and Sovramontino’s V2 system. I will argue that (i) the V2 constraint has been reinterpreted as the lexicalisation of a functional head in the C-domain, and (ii) the traditional preverbal position solely hosts constituents bearing a contrastive reading. In fact, in the two NEIDs under investigation, it seems that the [EDGE] feature that, in V2 languages, is responsible for the presence of a constituent in preverbal position (see Vikner 1995) has pragmatically specialised, becoming [CONTRASTIVE]. Lamonat and Sovramontino’s V2 system therefore fundamentally differs from that of Medieval Romance as the latter could host any pragmatically salient XP in preverbal position (see Benincà 2006). In this respect, it is not a fully-fledged V2 system, but a collapsing one, whereby the V2 constraint is a residual syntactic trait of the language. I will carefully explore the differences between the NEIDs under investigation and the contemporary Romance V2 languages, reaching the conclusion that Lamonat and Sovramontino’s V2 system must find an adequate description in light of its close interaction with discourse- pragmatics. Following Wolfe’s (2016) V2 typology, the synchronic data from Lamonat and Sovramontino would suggest that these two varieties were once fully-fledged High- V2 or Force-V2 systems. Making reference to Rizzi’s (1997) split-CP model, V2- motivated T-to-C movement targeted the highest left peripheral head, namely Force°, and it was accompanied by A’-movement of another constituent in its specifier position. In synchrony, I argue that, while verb movement to Force° is still in place, the generalised A’-movement of a preverbal constituent has been lost, but mandatorily retained for a reduced set of pragmatically salient elements: contrastive foci and contrastive topics. In the next subsection, I will provide further insights on this last claim.

From this angle of investigation, wh-in-situ in Lamonat and Sovramontino can be resolved by postulating that the verb crosses the fronted wh-element by moving as high as Force° due to the V2 constraint. Rizzi (1996, 2001), in his discussion of the wh- criterion, argues that, in Romance, evidence of previous V2 systems is indeed often manifest in verb movement in root interrogatives. In Lamonat and Sovramontino’s collapsing V2 system, the displaced verb acts as a boundary that marks different discourse functions; in particular, it sharply sets the boundary between contrastive vs.

24 non-contrastive informational units (see Kuno 1976; Kiss 1998; Buring 1999), as shown respectively in (8a) and (8b):

(8) a. CONTEXT: How much money did you give to Vittoria? (Sovramontino) Ghe o dat DIESE SCHEI mi a Vittoria DAT.CL have.1SG give.PTCP ten money I to Vittoria ‘I gave ten euros to Vittoria’

b. CONTEXT: Did you give twenty euros to Vittoria? No. DIESE SCHEI ghe o dat mi a Vittoria NEG ten money DAT.CL have.1SG give.PTCP I to Vittoria ‘No. I gave ten euros to Vittoria’

I will also claim that the aforementioned pragmatically motivated V2 system has direct consequences for the mapping of the left periphery, which Rizzi (1997) argues to be the syntactic locus in which informational-structural categories are encoded. For instance, Lamonat and Sovramontino bar clitic left dislocation (in the sense of Cinque 1990) as the head of TopP is unsuitable to host head movement. As a result, the C- domain of these NEIDs lacks the TopP layer, but, in turn, features an abundant use of scene-setting topical elements (see Poletto 2000; Benincà & Poletto 2004; Greco & Haegeman 2016 a.o.) which, I will argue, are realised in the left-most projection of the left peripheral space, namely FrameP.

Finally, Friulian and Fornese are not V2 systems synchronically. Nonetheless, I will investigate some Friulian data that suggest that the NEID in question exhibits a similar interaction between V2 and focus structure. My data, however, overall show that Friulian has lost the V2 constraint: I will claim that Friulian has only partially retained V2-motivated T-to-C movement in order to fulfil a specific pragmatic function, namely that of conveying marked narrow informational focus (see Benincà & Poletto 2004, Bianchi & Francarelli 2010, Cruschina 2012 a.o.). This syntactic strategy seems to be less frequently adopted by the new generation of speakers, suggesting that Friulian is aligning with SVO Romance languages, like Italian, in accommodating pragmatic functions into the syntax. To conclude, the importance of the present research lies in the analysis of the V2 constraint as a syntax-pragmatics interface phenomenon, whereby an XP is not by default moved to SpecCP to satisfy a mere syntactic requirement, but is constrained by discourse-pragmatics.

25 1.3.3. Root Contrastiveness as a Supra-Informational Status

In Romance and Germanic V2 (Holmberg 1986; Holmberg & Platzack 1995; Vikner 1995; Vilkuna 1995; Poletto 2002; Benincà 2006 a. o.) there exists a crucial relationship between the V2 phenomenon and information structure: the fronted XP is pragmatically salient by virtue of encoding some specific discourse-pragmatic function. I will claim that, in Lamonat and Sovramontino, the V2 linear order is not obtained by preposing just any pragmatically salient XP, but specifically contrastive XPs. This suggests that the syntax-pragmatics interface realisation of contrastiveness as a discourse-pragmatic feature is a key language component in the dialects under investigation. In Lamonat and Sovramontino, the retention of Medieval Romance V2 (see Benincà 1983) is crucial for the realisation of contrastive XPs, lending support to the claim that a pivotal bidirectional relation between information structure and the V2 constraint is in place. As previously mentioned, V2-motivated T-to-C movement ensures (i) adjacency of the verb to the pragmatically salient element that, hence, receives discourse prominence as well as (ii) the correct interpretation of contrastive elements, which are structurally realised in the preverbal field.

Contrastiveness is traditionally considered a property of foci. There is general agreement in favour of two distinct dedicated structural positions that encode either contrastive or informational focus (Kiss 1998; Rizzi 1997; Belletti 2001, 2004; Benincà & Poletto 2004; Bianchi & Francarelli 2010; Cruschina 2012; Bianchi & Bocci 2012; Bocci 2013; Rizzi & Bocci 2017 and many others). There is nonetheless mounting evidence that contrastiveness is independent of focus and also extends to topics: topics can also bear a contrastive reading and occupy a specialised syntactic position (Szabolcsi 1981; Gundel 1988; Frascarelli and Hinterhölzl 2007; López 2009; Neeleman et al. 2009; Bianchi & Francarelli 2010). In Lamonat and Sovramontino, I claim that contrastiveness has the ability of superseding topic or focus status, exhibiting a sharp distinction between the syntactic treatment of contrastive and non-contrastive XPs. I will show that a constituent bearing contrastiveness is banned from the TP layer and must therefore be realised in left-peripheral position.

Contrastive elements, be they topics or foci, are limited to one occurrence per sentence. I will claim that the preverbal position occupied by contrastive XPs is filled

26 through the activation of the V2 constraint: more specifically, the [EGDE] feature that in Medieval Romance V2 was responsible for the presence of an overt XP in the specifier position of the C° targeted by the moved verb has undergone pragmatic specialisation and become [CONTRASTIVE]. My analysis of contrastive elements will hence treat contrastive topics and foci as occupying the same highest structural position in the clause in the SpecCP position of the C° hosting the moved verb. The present research proposes a unitary syntactic analysis of this discourse-pragmatic category in the two NEIDs under investigation: contrastiveness should be regarded as an independent discourse-pragmatic status that supersedes topic and focus. In this regard, crucially, I will argue that contrastiveness should not be considered a discrete notion, but a continuum in which the degree of contrastiveness is determined by the properties of the set containing the contrastive element (Molnár 2002, 2006).

Having briefly outlined the major claims put forward in this thesis, I will now discuss data collection. I will (i) try to further justify the empirical scope of the present work, (ii) describe the sample of speakers that took part in the study, (iii) assess the methodology used throughout the data collection, and (iv) point out the follow-up work that is still necessary to make my data widely accessible to the linguistics research community.

1.4. Data Collection: Sample, Methodology and Database

Given the endangered status of the varieties under investigation (see Moseley 2012; De Cia 2013), not only did my data collection have the aim to provide suitable answers to my research questions, but also the ethical imperative to produce viable material to counteract language endangerment in case the speech communities have the interest to do so (see Nathan 2006). To this end, in my data collection, I observed the common practices of language documentation (Hale et al. 1992; Woodbury 2003; Himmelman 2006). As Himmelman (1998: 166) points out:

“The aim of a language documentation is to provide a comprehensive record of the linguistic practices characteristic of a given speech community. Linguistic practices and traditions are manifest in two ways: (1) the observable linguistic behaviour, manifest in everyday interaction between members of the speech community, and (2) the native speakers’ metalinguistic knowledge, manifest in their ability to provide interpretations and systematizations for linguistic units and events.”

27 Even though the output of my data collection is far from being an adequate and complete documentation of the varieties under investigation, it is worth pointing out which documentation practices I adopted in my data gathering in order to assess its strengths and limitations.

The data were collected in several field trips between 2015 and 2017. I recorded approximately 35 hours of interviews: 12 hours for Lamonat, 9 hours for Sovramontino, 8 hours for Fornese and 6 hours for Friulian. All participants gave oral or written consent. Data gathered through an ethnographic approach through participant observation were also included in my research. This is especially true for Sovramontino and Friulian as, during my fieldwork, I physically spent more time in the municipalities of Sovramonte (BL) and (UD). In line with the ethical procedures on data gathering and management, all data were anonymised and pseudonyms were given to core consultants so that if they wanted to opt out from the study, I would be able to identify their contribution and delete it. Data were collected from a total of 51 speakers between the age of 32 and 94 among the four different speech communities. A distinction must be drawn between core and satellite participants. Core participants: (i) were interviewed more than once; (ii) were interviewed in groups of two, three or four; (iii) the interview setting was formal and interviews were always recorded. Satellite participants: (ii) were interviewed only once or very rarely more than once; (ii) were often interviewed individually; (iii) interviews were not always formal and not always recorded.

The bulk of my observations is based on the evidence provided by 28 core participants. The function of satellite participants was two-fold: (i) double-check linguistic trends and findings found among core participants, functioning as a control group; and (ii) fill in possible gaps in a quick and efficient manner throughout the data gathering process. Satellite participants were generally more meta-linguistically aware, higher educated and slightly younger. Satellite participants were also key in the initial stage of the data-collection as, being generally dynamic figures in their speech communities, they helped me with the recruitment of core participants and with the translation of some of the questionnaires into the dialects. Paradoxically, by the same token, they were not suitable candidates: their metalinguistic awareness, in fact, often resulted in prescriptivist attitudes towards their native language. This is not uncommon

28 in fieldwork situations: highly educated speakers may in fact already have pre- established ideas of how language works (see Chelliah 2001; Abbi 2001) which at times could be misleading for linguists carrying out their investigations. While recruiting core participants, I excluded two other speaker categories, namely young and very old speakers: all core participants were between the age of 60 and 78. Young speakers tend to be dominant in the regional variety of Italian spoken in the area (see Berruto 1989, 2005). They are often semi-speakers of the dialect (in the sense of Dorian 1977): their native tongue is regional Italian, but they have mainly a passive knowledge of their dialect. By contrast, very old speakers tend to be monolingual in the dialect. Even though their language is virtually uncontaminated by regional Italian, it is very difficult to carry out a formal interview with them. Several problems arise: i) difficulty communicating in the , namely Italian; ii) difficulty understanding stimuli in stimuli-driven and elicitation tasks (Chelliah & De Reuse 2011); iii) physical impediments such as poor hearing, compromised vocal tract, fatigue while speaking and reduced mobility. I, nonetheless, managed to record two speakers over 90 in two of the four speech communities (Sovramontino and Lamonat). The language of very old informants is extremely valuable with respect to narratives and naturally occurring data. The data from old monolingual and young bilingual speakers are only partially used in the present research: they are considered as produced by satellite participants.

Table 1 below shows in detail the number of consultants interviewed per speech community, drawing the distinction between core and satellite participants.

Core Participants Satellite Group 1 Group 2 Group 3 Participants Fornese 2 2 3 3 Friulian 3 3 N/A 8 Lamonat 4 3 N/A 3 Sovramontino 2 3 3 9 Table 1. Participant distribution per speech community

Core participants were interviewed in small groups, which ranged from a maximum of four units to a minimum of two units. Group interviews have the advantage of favouring the collection of naturally occurring data along with elicited data from questionnaires and semi-staged activities. In light of the language documentation objective of the present research, working with small groups allows for an array of different data

29 gathering techniques. In the case of understudied languages, in order to put forward a sound generalisation about the investigated language system, it is crucial that different methodological approaches yield findings that point to the same underlying generalisations. To this end, I employed the three types of Himmelman’s (1998) communicative events, namely: participant observation, staged-communication and elicitation. In each interview all three techniques were adopted. Acts of staged- communication and data elicitation took up two thirds of the interview time, but naturally occurring speech was also abundantly gathered. Speakers were strictly instructed not to speak the lingua franca, namely Italian, with the interviewer and among each other, but their native NID: naturally occurring data, therefore, were recorded throughout the interview. A considerable amount of naturally occurring data was collected during breaks and at the beginning or end of each session. A break would be taken every 45 minutes (normally interviews lasted 90 minutes). During breaks, I would leave the room so that the informants would find themselves in a more intimate and colloquial setting, in which they could freely speak to one another; they were nonetheless aware that the microphone was recording and advised not to release any sensitive information. Each session was designed as suggested by Chelliah & De Reuse (2011): it included (i) a warm up stage; (ii) a selection of different tasks and (iii) a closing stage. In the warm up stage, 15 to 20 minutes, I freely interacted with the speakers and let them interact with each other. Again, note that the speakers were instructed only to speak in their native NEID. The warm-up stage conversation topics varied from the weather to the events of the week. In each speech community, we regularly met weekly for a month. In this first stage, I would also enquire about any aspect of the previous interview that was unclear or generated new questions. The central part of the interview generally lasted 45 to 50 minutes with a break after the first 25 minutes and consisted of a selection of tasks: different types of elicitation, grammaticality judgements and semi-staged acts of communication. The tasks were mainly questionnaire-driven, but the use of a questionnaire was not crucial, as it mainly functioned as a roadmap for the interview. In the closing stage, I generally thanked the speakers, shared some of the findings of previous interviews, and planned the next session(s).

Before going into detail on the different tasks used to elicit data, let us briefly cover the recording practice that I adopted in order to produce high quality recordings. I

30 recorded in stereo mode WAV using a Zoom H4 handy recorder. As Nathan (2006) points out, a recording made with a stereo microphone allows the listener to better separate the background noise from the actual conversation, making the recording more comfortable to listen to and easier to transcribe. In addition, I tried as much as possible to carry out the interviews in a quiet setting, facing away from the outside noise in order not to compromise the “listenability” of the recordings (Nathan 2006: 281). Interviews took place in different settings to accommodate the needs of the speakers, this ranged from public spaces to private houses: in each setting, I tried to achieve a good “signal/noise ratio” (Ladefoged 2003: 17) by organising the available space in relation to the location of the microphone and the informants. As back up, all interviews were also recorded with a second lower quality MP3 recorder. Finally, the main recorder was monitored with the aid of closed headphones in its making (Austin 2006 on tips on this type of practices).

Data collection other than naturally occurring speech through participant observation was carried out through different data gathering techniques: various types of elicitation, grammaticality judgements and semi-staged acts of communications. Participants observation through the recording of natural occurring speech is by all means the most desirable data gathering technique, especially given the focus of my research; nonetheless, collecting an adequate amount of data in a relatively short period of time through participant observation only is extremely hard to achieve: I hence had to resort to elicitation and acts of staged-communication. I will now point out which techniques were used and assess their limitations. In the preliminary stages of my investigation, I heavily relied on elicitation, especially on translation questionnaires in which speakers were presented with a questionnaire in the lingua franca, namely Italian, and asked to provide a suitable translation into the dialect. All core participants were in fact active bilingual speakers in the dialect and in Italian. This first stage was necessary as I only had a limited knowledge of the dialects under investigation. The data gathered though translation questionnaires were key to better design the following interviews. Nonetheless, as Chelliah and De Reuse (2011) point out, translation tasks are often not entirely reliable, especially when eliciting syntactic constructions, as there is the peril of informants engaging in word-to-word translation and producing calques of their contact language into their native language. This technique is also problematic as sentences are often out of context: hence determining their discourse-pragmatic

31 interpretation proves almost impossible (Chelliah 2001). In the second stage of my data gathering, I resorted to safer data gathering techniques, namely target language interrogation elicitation and target language manipulation elicitation (see Chelliah & De Reuse 2011). In the first type of elicitation, I used the dialect under investigation to come up with contexts or questions to which the informants had to reply to. In devising these questionnaires in the target NID, the help of satellite participants was crucial. In the second type of elicitation, I manipulated the syntactic constructions in the target language, namely the dialect, and asked the informants to react to them. They provided grammatical judgements that ranged from completely acceptable to completely unacceptable. Alternatively, in a finite set of sentences with different constituent orders, I would ask the informants which constituent combination sounded the best and which sounded the worst. With target language manipulation elicitation, I generally provided a suitable context in which the sentence would be uttered. In this respect, I also tried to gather data through semi-staged acts of communication by (i) providing a context to which speakers had to react and (ii) taking advantage of the fact that I was simultaneously interviewing more than one speaker and they could actually recreate the settings in which speakers would be more likely to utter the target syntactic construction. The difficulty with this type of activity is that some of the speakers were not able to abstract away from the interview setting and fully engage in the role play; nonetheless, the fact that at I was working with a small group of people positively helped in devising such activities.

The use of different elicitation methods to target the same syntactic constructions was relatively successful; nonetheless, at times, this modus operandi gave rise to inconsistencies, whereby speakers would give different versions of the same syntactic construction while engaging in different elicitation techniques. Such cases were overcome (where possible) numerically and acknowledging the superiority of natural occurring data over other elicitation techniques. Intra-speaker variation due to the pressure of the dominant language, Italian, surely played a role in determining such inconsistencies. By checking the speech of the oldest speakers, namely those speakers less proficient in Italian, I was able to judge whether or not the elicited syntactic construction was influenced by Italian or was authentic to the Romance variety under investigation. In any case, I tried to verify the syntactic constructions encountered though elicitation by also looking at naturally occurring data.

32

As far as the investigation of syntax-pragmatics interface phenomena is concerned, the main challenge in eliciting syntactic structures that serve different discourse-pragmatic functions is determining the extent to which shared knowledge plays a role in the informant’s selection of a particular syntactic construction over another. This is particularly true in those situations in which informants are performing a semi-staged act of communication and are asked to react to a specific communicative context given to them. Another challenge lies in overcoming the informants’ bias when giving grammaticality judgements in the evaluation of discourse-pragmatic data: the general tendency of speakers is to dislike marked word orders that convey specific discourse-pragmatic interpretations, even when they were caught using them in naturally occurring speech. Speakers’ unconsciousness and naturalness in the use of these syntactic-pragmatic constructions must in part be the reason why informants had a strong bias against them in an artificial setting. Also, the fact that the interviewer had a non-native accent while uttering these syntactically marked clauses deprived them of their naturalness, rendering them somehow ungrammatical to the ear of the informants. In fact, if the given marked structure was uttered by a native speaker of the dialect, informants were much more positive and lenient in their grammaticality judgement. Speakers, if asked for a judgment, had a bias towards a default SVO constituent order: the most neutral and unmarked word order that is found across many Romance varieties (see Cruschina 2016). The data gathering experience revealed that syntactic constraints that lead to ungrammaticality are much easier to test compared to discourse motivated pragmatic-syntactic configurations. Nonetheless, being aware of these problems, to the best of my abilities, I tried to overcome them and gather a bulk of data that, as closely as possible, resembles the way in which the NEIDs under investigation encode discourse-pragmatic information.

The interviews were subsequently stored in the University of Manchester p-drive system, transcribed with ELAN and subsequently exported as excel files. Not all the interviews were transcribed entirely, but, only those segments that contained relevant data for the purpose of the present research. Nevertheless, given the language documentation objective of this project, a considerable amount of data can still be extracted from the original recordings. The data that was transcribed was gathered in an excel database per dialect, containing the following elements: (i) the token (generally in

33 sentences), (ii) its unique ID, (iii) the recording from which the token was extracted, (iv) the translation into Italian and, finally, () a code for the main morpho-syntactic phenomenon that the token features. More work is definitely needed to turn these Excel spread sheets into proper corpora: I hope to undertake this work in the near future and allow other researchers to freely use my data for their research in the field.

To conclude, these materials could be a solid basis for a proper documentation project, if, in the near future, I or other researchers were given the financial and labour support to document these Romance varieties. Such a project may lead to the creation of several spoken corpora of these endangered Romance varieties. This goal could not have been achieved for all four Romance varieties in the setting of the present research, although I have taken a first step in that direction.

1.5. Conventions for Glosses, Translations and Orthographic Representations

This section has been included to guide the reader in the understanding and interpretation of the data provided throughout the thesis. I will first discuss the orthographic representation of the data, its strengths and limitations. It is important to note that that the NEIDs under investigation, except Friulian, do not have a written standardised form. I will then move to the illustration of the conventions used for the glosses and for the free translations from the dialects into English.

Complex writing systems are developed to satisfy the necessity of managing bureaucracy and, therefore, power relations (Cardona 1986). Not all speech communities have this necessity and have developed a written form of their language. This is indeed the case with Lamonat, Sovramontino and Fornese. All speech communities that did not develop a written form, founded their cultures on oral tradition. Given its nature, orthography includes many aspects that are ‘extra-linguistic’ and carries cultural, historical and symbolic values and implications (Gelb 1993). Hagège (1985) argues that the normalization of a language and its orthography is a political property, whereby the union of the language and its orthography is very appealing to the power, whereas language and orthography variation are elements of disturbance of the solidity of power. Given these considerations, adopting an

34 orthographic system to represent the NEIDs under investigation is no easy task.

All the investigated speech communities have nonetheless some awareness of the uniqueness of their native Romance variety and made various attempts at representing their own languages. The only fully codified NEID is Friulian through the work of the Friulian Philological Society (Societât Filologjiche Furlane) and the Regional Agency for the (Agjenzie Regjonâl pe Lenghe Furlane - ARLeF). Friulian enjoys institutional support and has been the object of various linguistic studies (see Heinemann & Melchior 2011). The other NEIDs under investigation, however, do not enjoy the same institutional support and no codified orthography exists for these varieties. Among the existing attempts to the standardisation of these languages, the following are worth mentioning: the Vocabolario Sovramontino by Vittorio Dalla Torre and Eugenio Bottegal (2018), Il Dialetto di Lamon: Cultura nelle Parole by Dr. Loredana Corrà (2001), and the Grammatica Fornese: Forni di Sopra by Rosella Fachin (unpublished book). The development of an orthography must be appealing to the speech community, especially when dealing with endangered languages, as the success of a language documentation or revitalisation project may depend on it (Seifart 2006). Codifying the orthography of the NEIDs under investigation is beyond the purpose of the present research. Nonetheless, I will also not represent the data from the NEIDs under investigation using the IPA symbols: I want to give these varieties the right status as languages and not simply as systems of sounds. To this end, I will try to devise an orthographic system that possesses the following characteristics: it must be an alphabetical representation of a phonographic system, whereby the basic units of the correspond to the elements of the sound structure of the language (Rogers 2005). Such an orthographic system is common across standardised Romance languages, but the extent to which phonemes correspond to graphemes varies from language to language in terms of consistency and completeness (Katz & Frost 1992). For geographical and political reasons, I will adopt the orthographic system of Italian and apply it to the NIDs under investigation. In addition, I will try to represent those sounds that do not belong to the Italian phonemic inventory with their own grapheme. In choosing the appropriate grapheme, I will consider the attempts to codify these languages by the various speech communities so that the result is an orthographic system that is not too far from the speakers’ expectations.

35 As far as Friulian is concerned, I will simply adopt the existing standardised orthography for the language. I will now point out a few features of the Friulian orthography that depart from Italian to help the reader process the Friulian data. In Italian, and in the proposed orthography for Lamonat and Sovramontino, the sound that is represented with the grapheme c is either pronounced /k/ or / ʧ/. The latter pronunciation is triggered when c is followed by the /e/ or /i/, for example in words like Italian cera (wax), cinema (cinema), bacio (kiss). In Friulian, the voiceless /ʧ/ is represented with grapheme when it precedes the vowel /a/, /o/ and /u/ and when it is in word-final position, as in çavate (slipper) and leç (law) respectively (Vicario 2005). Another Friulian grapheme that is not present in Italian is , which is the orthographic representation of the front semi-vowel /j/. This sound is represented with j only when it appears at the beginning of a word as in jet (bed) and joibe (Thursday), whereas, when the /j/ sound is present word-medially or word-finally, as in ploie (rain) and lui (July), is represented with the grapheme i (Vicario 2005). The semi-vowel /j/ is also present in Italian, but it is always orthographically represented with the grapheme i. In Lamonat and Sovramontino, I will hence follow the Italian representation of /j/. In Friulian, vocalic length is a contrastive phonetic feature and different minimal pairs can be identified depending on vowel length, for example short /a/ in lat (milk) and long /a:/ in lât (lake) (Vicario 2005). Vocalic length is not a contrastive phonetic feature in Italian, in Lamonat or in Sovramontino. In Friulian, long are marked with accent: â, ê, ô, û and î. Finally, Friulian exhibits the palatalization of the Late Latin velar /k/ and /g/ in front of the low-mid vowel /a/. In Friulian orthography, these palatal plosives are respectively represented as cj and gj as in gjat

(cat) and blancje (white).

Fornese shares most of its linguistic morpho-phonological properties with Friulian, therefore the same orthographic rules will be used. Lamonat and Sovramontino, on the other hand, show a somewhat different set of phonemes, which must be mapped onto their own orthographic representations. Lamonat exhibits the front rounded vowels /y/ and /ø/. Following the speech community’s orthographic representation for these marked vowels, I will respectively use the graphemes ü as in dür (hard) and ö as in schös (snail). Given that these vowels are not uniformly used by all the Lamonat interviewed speakers, but some speakers simply produce /u/ and /o/ instead, the ü and ö orthographic representation will only be used if /y/ and /ø/ were actually uttered by the speaker. The voiceless affricate sound /ʧ/ is orthographically

36 represented as in front of /a/, /o/, /u/. Instead the velar /k/ is orthographically represented as c in front of /a/, /o/, /u/, and as ch in front of /e/ and /i/. Lamonat and Sovramontino exhibit the dental sounds /θ/ and /ð/: I will respectively represent them as th and dh, for example dent (tooth) vs. dhent (people) or thoc (log) vs. toc (piece).

Now that I have schematically outlined the orthographic code used to represent the data from the NEIDs under investigation, I will explain how the examples will be organised. First, if relevant to the discussion, sentences are accompanied by a suitable discourse-pragmatic context (marked as CONTEXT:) in which they can be uttered. If not clear from the discussion, the name of the NEID discussed in the example will be given in brackets. Ungrammatical examples will be marked with an asterisk (*), sentences that are pragmatically infelicitous in the provided discourse-pragmatic context will marked with hash (#), and sentences that are considered grammatical by only a part of my informants will be marked with a question mark (?). If not stated otherwise, the portion of the sentence that bears discourse-pragmatic focus will be represented in capitals. Note that no distinctions will be made for regular topical elements, frame- setters or hanging topics. Scene-setting elements are generally followed by a pause, which I will mark with a comma; nonetheless, acoustically speaking, this is not always the case. As far as the glosses are concerned, they will be presented in a version of the Leipzig glossing rules. For convenience, some of the most common or non-standardised abbreviations used in this thesis are listed at page 6.

Finally, all grammatical and, at times, non-grammatical examples will be accompanied by a free translation into English (i.e. paraphrase). This translation is not word-by-word, but tries to convey the closest discourse-pragmatic meaning in English. In this respect, free translations are by no means reliable sources of syntactic comparison between the dialects and English.

1.6. Organisation

In the next chapter, the reader can find an overview of the dialects under investigation. This will not be exhaustive and has the sole aim of sketching the characteristic linguistic traits of these Romance varieties. In the second part of the next chapter I will talk more about the ambiguous status of the NEIDs under investigation as 37 non-consistent null-subject languages (Cardinaletti & Repetti 2010) and put forward some considerations on the nature of the traditional null-subject parameter (Chomsky 1981). In chapter 3, I will investigate the behaviour of wh-elements in Lamonat and Sovramontino and compare them with that of Friulian wh-items. I will also outline the mechanisms involved in root interrogatives in the satisfaction of the structural requirement of questions (Rizzi 1991) by investigating subject clitic inversion (SCLI). In chapter 4, I will discuss the complex relation between narrow focus and residual V2. In the first part of the chapter I will discuss focus structure (Lambrecht 1994) and the difference in the syntactic-pragmatic treatment between narrow focus and broad focus in the dialects under investigation. In the second part of the chapter, I will point out the syntactic characteristics that make Lamonat and Sovramontino pragmatically motivated V2 languages along with the discussion of V2 types and V2 in synchrony and diachrony in Romance. In chapter 5, I will discuss contrastiveness in the NEIDs under investigation. More specifically, I will look into how contrastive XPs are encoded in the syntax of Lamonat and Sovramontino. I will put forward a scalar typology of discourse- pragmatic contrast, whereby the degree of contrastiveness carried by an XP vastly depends on the properties of the set from which the contrastive element is drawn and contrasted (Molnár 2006). I will also discuss the behaviour of d-linked wh-elements and outline the similarities between these elements and contrastive XPs. Finally, in chapter 6, I will draw some concluding remarks and point out some areas left for future research.

38 2. North-Eastern Italian Dialects: An Overview

In this chapter, I will provide an overview of the four NEIDs under investigation: Fornese, Friulian, Lamonat and Sovramontino. More specifically, in the first part of the chapter, I will provide: (i) a brief description of each identified dialect subgroup, namely Lamonat-Sovramontino and Friulian-Fornese, and (ii) an informed discussion on their genetic affiliation. In the second part of the chapter, I will outline and briefly discuss a characteristic trait of the NIDs: their subject clitic (SCL) system. I will reach the conclusion that the morpho-syntactic behaviour of SCLs cannot be accounted for as a unitary phenomenon across the NIDs. Nonetheless, the development of SCLs correlates with the general rigidification of the basic SVO constituent order across the NIDs. This is important because word order rigidity affects the encoding of discourse- pragmatic information into the syntax. In this respect, I will point out why the NIDs are considered to be non-consistent null-subject languages (Cardinaletti & Repetti 2010) and, most importantly, spell out the underlying assumptions about the SCL system of the NEIDs under investigation. This last point will be key to understand the relationship between the syntax of SCLs and the interpretation of syntax-pragmatics interface phenomena in the remainder of the thesis.

2.1. The Northern Italian Dialects

A fine-grained description of the four NEIDs under investigation goes beyond the purpose of the present research. In this section, I will point out the general characteristics of Northern Italian dialects with respect to their classification within the Romance family and the main unitary traits of North-Eastern Italian Dialects. The NEIDs under investigation, namely Fornese, Friulian, Lamonat and Sovramontino can be divided into two subgroups: Fornese-Friulian and Lamonat-Sovramontino. In subsection 2.2.1 and 2.2.2, I will briefly discuss the main characteristics of the two groups, paying particularly attention to the microvariation across the languages of each group.

Historically, the communities where Fornese, Friulian, Lamonat and Sovramontino are spoken have never enjoyed a unitary political status, except under the

39 domination of the between the 15th and 17th centuries. The lack of political unification has meant considerable linguistic variation in the area. Undoubtedly, the influence of Venetian is an identifying trait of two of the four NEIDs under investigation: Lamonat and Sovramontino. At the same time, the mountainous and isolated location of these speech communities has favoured the retention of relatively conservative linguistic traits compared to their neighbouring NIDs. As far as their genetic affiliation is concerned, in broad terms, all four varieties belong to Northern Romance. At the same time, they exhibit morpho-phonological features of Western Romance (Zamboni 1998, drawing on La Fauci 1988; Andreose & Renzi 2015). More specifically, the Friulian variety of Tricesimo and Fornese (Forni di Sopra) exhibit the charactering traits of the Rhaeto-Romance (Haiman & Benincà 1992). The Friulian variety of Tricesimo is a dialect of Eastern-Central Friulian, whereas Fornese shares its linguistic traits with Carnic or Northern Friulian (Benincà & Vanelli 2016) and, to a lesser extent, Cadorino Ladin (Pellegrini 1979; Zamboni 1984; Salvi 2016). The classification of Lamonat and Sovramontino within Northern Italo-Romance is less straightforward: Venetan, Ladin and Gallo-Italic features (see Maiden & Parry 1997) are virtually equally present in the two dialects. The morpho-phonological make-up of these two NEIDs seems to point to the conclusion that they are Ladin-Venetan varieties with Gallo-Italic influence due to the close proximity to Western Trentino. The map in Figure 1 shows the four NEIDs under investigation and their neighbouring Romance varieties. Note that the map is simplified and only represents the North-Eastern part of Italy3.

3 Map based on Benincà, Parry and Pescarini (2016) and Istituto Ladin de la 40 Figure 1 Geographical distribution of the four NEIDs under investigation and their neighbouring Romance varieties.

In identifying the genetic affiliation of the four NEIDs, we must briefly introduce the (Ladin Dispute), which concerns the status of Rhaeto-Romance (Gartner 1883) as a language family. In his investigation of Romance, Ascoli (1873) originally acknowledged Rhaeto-Romance as a single variety, which he labelled Ladino: he identified a group of linguistically peculiar dialects spoken in the Alpine arch that stretches from to the North-Eastern border of Italy. Rhaeto- Romance has been subsequently divided into three main dialect areas: Romansh in Switzerland, Ladin in Trentino-Alto-Adige and in some parts of the province of Belluno, and Friulian in -Venezia-Giulia (Haiman and Benincà 1992). Rhaeto- Romance has received a great deal of attention, which culminated with the pivotal work of Haiman and Benincà in 1992. These scholars argue that Friulian, Ladin and Romansh lack a real basis for unity and autonomy, reaching the conclusion that Rhaeto-Romance is more of a scholarly invention than an actual fully-fledged language family. The partial documentation of Sovramontino, Lamonat and Fornese which I have conducted in my PhD programme indeed seems to weaken the claim that the Rhaeto-Romance

41 family is a unitary and autonomous language family even further. The data from these three varieties, along with the data from Friulian, seem to support Pellegrini’s (1972) view that the linguistic features that differentiate Rhaeto-Romance from the rest of the NIDs were once also common to most Gallo-Italic and Venetan dialects. In other words, the Rhaeto-Romance varieties are a conservative group of Romance languages, which are characterised by a higher number of shared retentions than shared innovations. This is particularly evident in Lamonat and Sovramontino, which exhibit features that are too innovative to be Rhaeto-Romance, but too conservative to be classified as Gallo-Italic: they offer a synchronic insight into the diachronic development of contemporary NIDs.

Before looking in detail at the characteristics of each dialect, let us briefly point out the common features shared by all the NEIDs under investigation. Despite their non-straightforward sub-classification, the fours NEIDs exhibit clear morpho- phonological features that place them within Northern and Western Romance. With respect to diachronic phonological changes, a characteristic trait is the of final unstressed vowels except –a. For example, the Latin word for ‘night’ NOCTE(M) has lost its final unstressed vowel –e resulting in Friulian [ɲot], Fornese [nuot], Lamonat and Sovamontino [not]. Conversely, final –a has been maintained as in Friulian [piere], Sovramontino [piera] < Latin PETRA(M) ‘stone’. Benincà, Parry and Pescarini (2016) point out that this rule affects all NIDs, but the number of morpho-phonological contexts to which it applies is dialect-specific.4 Note that this diachronic process clearly distinguishes Standard Italian from NIDs: Italian has in fact retained final unstressed vowels as in notte ‘night’ and pietra ‘stone’, yielding a different classification within Romance. Another diachronic phonological process that affected the NIDs is the degemination of long consonants, including those formed by the assimilation of Latin clusters: Friulian [tiare] and Lamonat [tera] < Latin TERRA(M) ‘earth’. The last crucial diachronic process worth mentioning and typical of NIDs and, more generally, Western Romance is the voicing of intervocalic voiceless stops, which have subsequently undergone to different degrees: for example, Latin ROTA(M) ‘wheel’ and SAPONE(M) ‘soap’ have become Friulian [ruede] and [savoŋ], and Sovramontino [roða] and [saoŋ]. At the morpho-syntactic level, the four NEIDs under investigation follow the NIDs and, more generally, the tendency of Northern Romance

4 In Southern Venetan and Ligurian, for instance, the diachronic rule only applies to a limited number of morpho-phonological contexts. 42 to replace synthetic with analytic forms. This is particularly evident with respect to the aoristic drift of the present perfect, whereby the analytic perfect has replaced the synthetic across virtually all NIDs. A last feature worth mentioning and, perhaps, the most characteristic feature of NIDs is the presence of subject clitics (henceforth abbreviated SCLs): phonetically realized subject pronouns with variable degrees of obligatoriness and a fixed position within the inflectional domain (Renzi & Vanelli 1983; Poletto 1993, 2000; Goria 2004; Manzini & Savoia 2005). In the second part of the chapter (section 2.3), I will succinctly outline the SCL system of the dialects under investigation: the general morpho-syntactic behaviour of SCLs across NIDs and their dialect-specific characteristics. The empirical generalizations describing the behaviour of SCLs and the theoretical assumptions behind their analysis will be important to inform the discussion of syntax-pragmatics interface phenomena in the subsequent chapters. In fact, the manifestation of SCLs within the clause offers precious insights in the investigation of the interaction between syntax and discourse-pragmatics.

In the next section, I will point out the main characteristics of Fornese, Friulian, Lamonat and Sovramontino. I will do so in support of the preliminary genetic sub- affiliation provided above and to show the reader that cross-dialectal microvariation is the result of the independent diachronic development of these varieties directly from their ancestor. In fact, notwithstanding their misleading label, the NEIDs are sister languages of Standard Italian.

2.2. NEIDs: Overview and Genetic Affiliation

Fornese, Friulian, Lamonat and Sovramontino are no satisfactory representative sample of the complex array of languages that identify the linguistically rich north- eastern part of Italy. Leaving aside the non-Romance languages spoken in the area (German, Slovenian, Cimbrian and Resian among other Slavic and Germanic dialects), NEIDs can be described as a that includes varieties of Friulian, Ladin, Venetan and Gallo-Italic (Maiden & Parry 1997).

With the exception of the Friulian variety of Tricesimo, the NEIDs under scrutiny are situated in transitioning areas between different dialect sub-groups. Fornese neatly marks the end of the Friulian linguistic area and the beginning of the Cadorino Ladin

43 one, displaying characteristics of both dialect groups. Lamonat and Sovramontino exhibit linguistic features of all main dialect groups in the region except Friulian, namely Ladin, Gallo-Italic and Venetan. Besides the relative proximity of Lamonat and Sovramontino to the Ladin and Gallo-Italic dialect areas, I believe that the reason behind this amalgam of features is rooted in the geopolitical history of the speech communities of Lamon and Sovramonte. The villages of Lamon and Sovramonte are located in a mountainous area on both sides of the valley that links the cities of Primiero and Feltre. Their isolated mountainous location (away from the main trade routes linking highly populated towns) allowed Lamonat and Sovramontino to maintain and further develop their pre-Venetan linguistic features. In 1420, Lamon and Sovramonte, along with the territories of the city of Feltre, were incorporated into the Most Serene Republic of Venice (Corrà 2001; Zorzi 2001): from that date to the end of the 18th century, Lamon and Sovramonte constituted border territories of the Republic of Venice. The mountainous and peripheral geo-political location may indeed have resulted in less sustained contact with Venetan5, which favoured the retention of conservative Gallo- Italic/Ladin features (more specifically, the language stage from which Gallo-Italic and Ladin branched off), reflecting de facto a frozen stage in the development of contemporary Gallo-Italic.

The label NEIDs solely denotes a geographic area as no specific set of linguistic features clearly separates NEIDs from NIDs. NEIDs, nonetheless, display some unitary traits that are common to most dialects spoken in the area. First, the first person plural verb ending is either –on or –emo (Pellegrini 1956; Benincà, Parry & Pescarini 2016), as Sovramontino [rið-on] ‘we laugh’ and Venetan [(a)nd-emo] ‘we go’. Note that Friulian first person plural ending is –ín [an-ín] ‘we go’. Second, SCLs, if present,

5Sovramontino exhibits a greater number of Venetan features than Lamonat. It is very plausible that Sovramontino was in more sustained contact with Venetan than Lamonat. Nonetheless, Sovramontino’s Venetan features are more archaic than Lamonat’s. In this respect, it seems that Lamon entered into relatively sustained contact with Venetan at a later stage and Sovramontino remained isolated from that time onwards. From a speculative point of view, this period of isolation might have started in 1631 when Sovramonte was plagued by the Black Death (De Bortoli 2016). This historical event might have dictated a deviation in the trade route that linked Feltre to the Primiero valley (Province of Trento) towards the Lamon territories: Lamonat hence started being in regular contact with a more innovative variety of Venetan. This hypothesis is solely based on the known morpho-phonological diachronic processes that took place in Venetan and are found in the two varieties (see section 2.2.2). Further linguistic and, crucially, historical evidence is needed to test this hypothesis. 44 invert with the inflected verb in root questions, as shown in (1b), the interrogative equivalent of (1a). This is called subject clitic inversion (abbreviated SCLI).

(1) a. Mario al a cognût fa-lu (Fornese) Mario 3SG.M.SCL have.3SG need.PTCP do.INF-OCL.M.3SG ‘Mario had to do it’

b. Mario, a-lu cognût fa-lu? Mario have.3SG-3SG.M.SCL need.PTCP do.INF-3SG.M.OCL ‘Mario, did he have to do it?’

In Central and Western NIDs, SCLI (subject clitic inversion) is less commonly found in root interrogatives. Note that SCLI is optional in Fornese: speakers claim that the interrogative force of the clause can also be conveyed through intonation alone. Third, NEIDs privilege weak preverbal negation over postverbal negators. This type of negation, which generally precedes the SCL except the vocal clitic a (Zanuttini 1997; Poletto 2000; Calabrese & Pescarini 2014), sharply distinguishes Eastern and Western NIDs. In most Western NIDs, (2) would be grammatical without the preverbal negation no. In the NEIDs under investigation, if, in negative declaratives, ‘western-style’ post- verbal negators are present, they need to be licenced by the preverbal negation. Across NEIDs, postverbal negative elements like mia/miga/mica do not hence behave like true negators, but rather like negative polarity items (NPIs), as in (2):

(2) Me pare *(no) l a mia magnà l pom (Sovramontino) My father NEG 3SG.M.SCL have.3SG NEG eat.PTCP the apple ‘My father did not eat the apple’

Generally speaking, NEIDs, with the exception of Venetan, are more conservative than the other NIDs. Ladin, Friulian and their peripheral varieties seem to have independently maintained linguistic stages once shared with all other NIDs. Salvi (2016: 154) talks about ‘the survival of an archaic phase of Cisalpine Latinity’ that became lost in the rest of the territory of due to the expansion of linguistic innovations.

In the next two subsections, I will (i) discuss the microvariation between the two dialect sub-groups identified in the present research (Friulian-Fornese and Lamonat- Sovramontino) and (ii) compare and contrast the two NEIDs that make up each group. Note that the following subsections do not aim to be an exhaustive dialectological

45 description of the four NEIDs under investigation, but to highlight those linguistic traits that are key to set the two different dialect areas apart.

2.2.1. Friulian and Fornese

The two NEIDs investigated in the present research are spoken in the municipalities of Tricesimo and Forni di Sopra (Province of ). The number of speakers is circa 6000 for Tricesimo Friulian and 1000 for Fornese. The Fornese spoken in Forni di Sopra must not be confused with Fornese spoken in the neighbouring municipality of (see Manzini & Savoia 2005). The latter is a relatively pure variety of Carnic or Northern Friulian (Benincà & Vanelli 2016), whereas the former shows considerable influence of Cadorino Ladin. As previously mentioned, the Friulian variety of Tricesimo is a standard example of Eastern-Central Friulian (Benincà & Vanelli 2016).

Both Friulian and Fornese show the palatalization of the Latin velar plosives /k/ and /g/ before /a/ (Vicario 2005; Benincà & Vanelli 2016; Loporcaro 2011). For example, Latin CANE(M) ‘dog’ has become Friulian [can] and Fornese [ʧan], and Latin GALLU(M) ‘rooster’ has become Friulian [ɟal], Fornese [ʤal]. This diachronic phonological change has been considered one of the defining traits of the Rhaeto- Romance family (Haiman & Benincà 1992). Nonetheless, there is evidence that this phonological rule once affected the whole northern Italian territory, but survived only in Ladin, Friulian and Romansh (Pellegrini 1991; Tuttle 1997; Benincà, Parry & Pescarini 2016). In the palatalization process, Fornese seems to be a step ahead with respect to Friulian with the Friulian palatal plosives /c/ and /ɟ/ being realized as the /ʧ/ and /ʤ/. In this respect, Fornese is closer to Cadorino Ladin than Northern Friulian (Pellegrini 1979, 1992). No palatalization is instead found in Lamonat and Sovramontino in this context: see Lamonat [kan] ‘dog’ and Sovramontino [gal] ‘rooster’. Another common trait between Friulian and Fornese is the retention of word- final –s. This is another feature that has been proposed as a unifying trait of the Rhaeto- Romance putative language family. The suffix –s marks the plural on , and determiners, as shown in (3b), the plural of (3a):

46 (3) a. Chista biela femina (Fornese) This.F beautiful.F woman.F ‘This beatutiful woman’

b. Chistas bielas feminas This.F.PL beautiful.F.PL woman.F.PL ‘These beautiful women’

In line with Western Romance, the suffix –s also marks the second person singular of virtually all tenses and moods, except the imperative (Benincà & Vanelli 2016), as shown in (4):

(4) To tu mangjis mase (Friulian) you 2SG.SCL eat.2SG too-much ‘You eat too much’

In all other NIDs, word-final –s has been lost. Lamonat and Sovramontino have also lost the suffix –s, but have not yet developed another plural marking strategy, showing invariant singular and plural forms for masculine nouns (see section 2.2.2). The last feature that has been proposed as a distinctive trait of Rhaeto-Romance and is present in both Friulian and Fornese is the retention of the liquid in the cluster C+[l] (Haiman & Benincà 1992). In Lamonat and Sovramontino, along with the other NIDs, it has been palatalised. For instance, Latin CLAVE(M) ‘key’ retains the C+l cluster in Friulian [kla:f], but has undergone palatalisation in Lamonat [ʧao]. The same is true for Latin GLACIE(M) ‘ice’ and UNGULA(M) ‘fingernail’, which have respectively become Sovramontino [ʤaθ] - Friulian [glaʧe], and Sovramontino [onʤa] - Fornese [oŋgla]. These three morpho-phonological processes, namely (i) palatalisation of velar plosives, (ii) retention of word-final –s and (iii) retention of the cluster C+[l] have been put forward as evidence in support of the putative Rhaeto-Romance language family. As previously mentioned, these seem to constitute independent shared retentions rather than common shared innovations. I hence agree with Haiman and Benincà (1992) in considering Rhaeto-Romance more of a scholarly invention, rather than a proper language family.

Moving away from morpho-phonological diachronic processes, another common feature between Friulian and Fornese is the development of double compound tense forms, featuring two auxiliaries, as shown in (5) and (6):

47

(5) Lu ai vût comprât al marcjât (Friulian) 3SG.M.OCL have.1SG have.PTCP buy.PTCP at-the market ‘I bought it at the market’

(6) I ai vût stât a ciasa so (Fornese) 1SG.SCL have.1SG have.PTCP be.PTCP to house his ‘I went to his house’

An invariant past participle form of the auxiliary have appears between the inflected auxiliary have and the past participle of the lexical verb. There is general agreement that this tense is used to refer to some point anterior to a reference point retrievable from the discourse (Poletto 2009; Marcato 1986). Melchior (2012) argues that it has an experiential flavour, whereby the speaker underlies the case that an event or situation has already occurred, rather than indicating an actual occurrence of the event or situation at a specific reference point in time. Lamonat and Sovramontino do not show trace of this type of double compound tense forms.

As for the microvariation between Friulian and Fornese, I will now point out some of the main differences between the two NEIDs at the morpho-phonological level. Note that the comparison is by no means exhaustive. A characteristic trait of Friulian is vowel length opposition: vowel duration generates phonemic contrasts like Friulian [lat] ‘milk’ and [la:t] ‘lake’. The devoicing of an originally intervocalic consonant along with the apocope of the final untressed vowel determined the lengthening of the preceding vowel allegedly through a process of (Vicario 2005; Benincà & Vanelli 2016). Long vowels are hence only found in final closed with a single consonant coda (Finco 2007). In Fornese and, more generally, in Northern Friulian, lengthened /e:/ and /o:/ have further developed into the falling /ej/ and /ow/. For instance, Latin FOCU(M) ‘fire’ and NIUE(M) ‘snow’ have developed into Friulian [fo:k] and [ne:f], but Fornese [fowk] and [nejf]. In Lamonat and Sovramontino vowel length does not generate phonemic contrasts. Another evident morpho-phonological difference between Fornese and Friulian concerns final atonic vowel –a. Fornese has maintained Latin –a, whereas in the Friulian variety of Tricesimo and, more generally, in Eastern-Central Friulian, Latin final unstressed –a has become – e. Let us consider the Friulian feminine [blance] ‘white’ from Germanic BLANK as an example. In [blance], the diachronic process that led to the change of final atonic –a into –e is particularly evident; in fact, the /k/ in

48 [blanka] has undergone palatalization, meaning that atonic final –e must have been –a for the latter phonological process to have taken place. This is indeed confirmed by Fornese [blanʧa] ‘white’, where final atonic –a has been retained. As for Lamonat and Sovramontino, similarly to Fornese, they have maintained Latin final unstressed –a.

Before moving on to the discussion of Lamonat and Sovramontino, I will end this section by pointing out a peculiar difference between Friulian and Fornese, which concerns the string of sounds that make up basic affirmative ‘yes’. Friulian, similarly to Italian and Italo-Romance in general, employs sì, whereas in Fornese sì is non-existent and speakers utter aì instead. Interestingly, some elderly Sovramontino speakers from the village of Aune during an interview reported that affirmative sì was not common in their grandparent generation, but they instead used eì, now obsolete in contemporary Sovramontino. Whether or not Fornese aì and Sovramontino archaic eì are related and share the same etymology is a matter of further research; nonetheless, it is important to note that the isolated mountainous locations of these speech communities has favoured the retention of conservative Romance traits and, at the same time, further dialectalization.

2.2.2. Lamonat and Sovramontino

Sovramontino and Lamonat are spoken in the south-eastern part of the province of Belluno in the municipalities of Sovramonte and Lamon. The two speech communities include roughly 1000 speakers and 2500 speakers respectively. These two Northen- Romance varieties exhibit phonological, morpho-syntactic and lexical characteristics of all main dialect areas of the region, namely Venetan, Ladin and Gallo-Italic. As argued, the reason behind such an amalgam of features must be found within the geo-political history of these speech communities. I will now provide a partial description of Lamonat and Sovramontino, highlighting their challenging sub-classification within Northern Italo-Romance.

As far as their phonology is concerned, Lamonat and Sovramontino are characterised by the extensive presence of the dental [θ] and [ð]. These are the result of different diachronic processes of lenition. One such process is the aforementioned lenition of the intervocalic stops /t/ and /d/. Note that originally

49 intervocalic Latin /d/ in word-final position is also lenited in Lamonat, but retained in Sovramontino6. For instance, Latin NIDU(M) ‘nest’ has become Lamonat [nið] and Sovramontino [nid]. In an identical fashion, Latin VIRIDE(M) ‘green’ has become

Lamonat [virð] and Sovramontino [verd]. Another process of lention that led to the dental fricatives [θ] and [ð] regards Latin reflexes of CI, CE, GI and GE word-initially and post-consonantally (Zamboni 1974; Tuttle 1997). For example, Sovramontino [θendre] < Latin CINERE(M) ‘ash’; Sovramontino [ðenoʧo] < Latin GENUCULU(M) ‘knee’; and Lamonat [molθer] < Latin MULGERE ‘to milk’. Both Lamonat and Sovramontino exhibit apocope of [v] before a velar vowel. This diachronic outcome is also robustly attested across Ladin varieties (Salvi 1997). For example, Sovramontino [oltar] < Latin VOLGERE ‘to turn’ and Lamonat [ostro] < Latin VOSTRU(M) ‘your (plural)’. Palatalization of stem-final coronal consonants to mark the plural was once very productive across NIDs (Benincà, Parry & Pescarini 2016). In Lamonat and Sovramontino, this is now limited to nouns ending in –l and –on. For instance, Lamonat singular [kaval] ‘horse’ becomes plural [kavaj] and Sovramontino [porton] ‘gate’ becomes plural [portuj]. In Friulian and Fornese, plural palatalization is limited to masculine nouns ending in –l. As far as the phonemic inventory of the two NEIDs is concerned, Lamonat exhibits rounded front vowels [y] and [ø] of which there is no trace in Sovramontino. For instance, Lamonat [skøla] ‘school’, [sʧøs] ‘snail’, [dyr] ‘hard’, [myr] ‘wall’. This feature is typical of the western-most Rhaeto-Romance (Haiman & Benincà 1992) and Gallo-Italic varieties. Rounded front vowels [y] and [ø] are hence a known feature of NIDs, but are unexpected in NEIDs. It is virtually impossible to establish whether they constitute an independent relatively recent innovation or a retention from a common Gallo-Italic stage. Note that this peculiar Lamonat trait is dying out: [y] and [ø] are being replaced by [u] and [i], but they are still relatively robust among speakers aged 70+.

As for the morpho-phonological characteristics of the two NEIDs, the loss of word-final –s has affected both Lamonat and Sovramontino. Most masculine plural nouns have the same form as their singular counterparts: Lamonat singular [ranʧ] and plural [ranʧ] ‘spiders’, Sovramontino singular [pom] and plural [pom] ‘apples’.

6 Note that the loss of final inherited Latin unstressed vowels (except /a/) does not apply to Venetan borrowings (for example, Sovramontino [viaðo] ‘trip’). Venetan borrowings came into these NEIDs at a later stage once this particular phonological process ceased to be productive.

50 Differently from Venetan varieties, Lamonat and Sovramontino have not yet developed a strategy to differentiate the plural and the singular forms of masculine nouns after the loss of word-final -s. In this respect, they reflect a frozen stage between Rhaeto- Romance and Gallo-Italic. Not surprisingly, Sovramontino and Lamonat’s plural nominal morphology resembles that of Friulian, but without word-final –s, as shown in Table 1 below:

Masculine Masculine Feminine Feminine Singular Plural Singular Plural Friulian [can] [can-s] [cjadre-e] [cjadre-i-s]

Proto- *[kan] *[kan-s] *[karieg-a] *[karieg-e-s] Sovramontino

Sovramontino [kan] [kan-ø] [karieg-a] [karieg-e-ø]

Table 1. Comparison between Friulian’s and Sovramontino’s plural nominal morphology and recostruction of Proto-Sovramontino’s plural morphology prior the loss of word-final –s. The masculine noun is ‘dog’ and the feminine noun is ‘chair’.

Without written records of Lamonat and Sovramontino, it is virtually impossible to reconstruct the plural morphology of the NEIDs under scrutiny before the loss of plural –s. For instance, we cannot say whether or not the feminine plural ending in –e has developed before or after the loss of –s. Invariant singular and plural forms are common in the Bellunese area. As far as inflectional morphology is concerned, Sovramontino and Lamonat seem to closely resemble Livinallese: the less-conservative Ladin variety spoken in the north-western part of the Province of Belluno (Salvi 2016). As far as Ladin-like morphology is concerned, Lamonat and Sovramontino’s first person singular takes the ending –e in the present and imperfective of both the indicative and subjunctive moods. Similarly, Lamonat and Sovramontino exhibit the first person plural ending in –on as opposed to Venetan -emo. In Lamonat, the first singular ending is nonetheless absent as a result of the general diachronic phonological rule that deleted all final unstressed vowels except –a. In Sovramontino, the ending has being reintroduced, as in many other NIDs, due to paradigm levelling (Benincà, Parry & Pescarini 2016). For example, (i) Lamonat [vag] and Sovramontino [vae] ‘I go’; (ii) Lamonat [dig] and Sovramontino [dige] ‘I say’.

51 At the morpho-syntactic and lexical levels, Lamonat and Sovramontino exhibit clear Venetan characteristics such as: (i) the existential proform ghe, as in (7); (ii) the invariant third person dative clitic ghe, as in (8); (iii) the forms of the first and second singular tonic subject pronouns mi and ti, which derive from the respective Latin oblique pronouns (see Zamboni 1974).

(7) Ghe n-è do femene (Sovramontino) EXIST.CL PART.CL-be.3SG two women ‘There are two women’

(8) Mi ghe l è dit (Lamonat) I DAT.CL 3SG.M.OCL have.1SG say.PTCP ‘I said it to him/her/them’

As for the other tonic pronouns, the first and second plural forms derive from reflexes of Latin NOS/VOS compounded with ALTERI: Lamonat and Sovramontino noaltri and oaltri. Lamonat has also retained the Ladin/Friulian-like first person tonic pronoun no(e). The past participle forms are also Venetan, but rather archaic. For second and third verb conjugations, Sovramontino past participle form is –est/ist, whereas Lamonat exhibits a relatively less archaic form in –ù/–ì (see Tuttle 1997). For example, Sovramontino [skonðest] and Lamonat [skonðù] ‘hidden’. This difference seems to suggest two probable scenarios: (i) Lamonat went into sustained contact with Venetan at a later stage than Sovramontino, (ii) after initial sustained contact with Venetan, Sovramontino suffered some degree of isolation, which resulted in the retention of its archaic features. Other differences between Lamonat and Sovramontino seem to point in that direction. One striking difference has to do with the Lamonat formation of a restricted set of plural nouns. Possibly due to the later contact with Venetan and the relative distance between the two varieties at the time of contact, Lamonat has borrowed from Venetan the entire plural forms (as opposed to just the plural morphology) of a limited number of nouns like singular [sor] ‘sister’ and plural [sorele] ‘sisters’. The singular is more similar to Friulian [su:r] or Gardenese Ladin [sor], hence pre-Venetan, whereas the plural [sorele] comes from the Venetan singular form [sorela]. In Sovramontino, which was allegedly exposed to Venetan at an earlier stage (and for a longer time), both lexeme forms have been replace by Venetan singular [sorela] and plural [sorele]. Another interesting phenomenon in Lamonat, possibly contact-induced with Venetan, is the double plural marking in –i on singular nouns ending –l whose plural form is already marked by the palatal sound –j: for example, Lamonat singular

52 form [porθel] ‘pig’ and plural form [porθjei] ‘pigs’. Sovramontino, on the other hand, only adopts the archaic pre-Venetan palatalisation strategy to mark the plural on masculine nouns ending in –l: singular [porθel] ‘pig’ and plural [porθej] ‘pigs’. Note that further investigation is needed to define the exact nature of this morpho- phonological change, which may as well be an instance of independent of the Venetan influence. A straightforward non-Venetan morpho-syntactic feature that is attested in Lamonat and Sovramontino regards auxiliary selection: reflexive and se ‘self’ trigger have auxiliary selection, as shown in (9) and (10), (11):

(9) La se a rot an brath (Sovramontino) 3SG.F.SCL 3SG.RFLX have.3SG break.PTCP an arm ‘She broke an arm’

(10) Me o petenà i cavej mi mi sola 1SG.RFLX 1SG.have comb.PTCP the hair.PL I I alone ‘I combed my hair own my onw’

(11) Me è scatì le cavele da mi sola (Lamonat) 1SG.RFLX 1SG.have comb.PTCP the hair.PL by I alone ‘I combed my hair own my own’

Have auxiliary selection as opposed to be auxiliary selection in this morpho-syntactic context (as in Italian, Friulian and Fornese) is a robustly documented feature of Ladin (Salvi 1997). Finally, the Lamonat variety of the village of San Donato exhibits an invariant clitic a (Poletto 2000). In the next section, I will explain the morpho-syntactic properties of SCLs. The peculiarity of San Donato clitic a is that it is obligatory with the first person singular and plural, but does not behave like a subject clitic. This can be seen in (12), (13), (14) and (15):

(12) a. A vag a nothe (Lamonat – San Donà) CL go.1SG to wedding ‘I go to a wedding’

b. *Vag a nothe go.1SG to wedding

(13) a. A bal e cant CL dance.1SG and sing.1SG ‘I dance and sing’

53 b. *A bal e a cant CL dance.1SG and CL sing.1SG

(14) a. Noe a no canton We CL NEG sing.1PL ‘We don’t sing’

b. *Noe no a canton We NEG CL sing.1PL

(15) A cant-e? CL sing.1SG-1SG.SCL ‘Shall I sing?’

Differently from SCLs, clitic a (i) cannot be repeated in a coordinated structure (cf. 13), (ii) must precede the preverbal negation (cf. 14), (iii) does not undergo inversion with the inflected verb in questions, but co-occurs with the first person singular or plural SCL that surfaces in root interrogatives and undergoes SCLI. In section 2.3, I will discuss the morpho-syntactic behaviour of SCLs across NIDs with reference to the NEIDs under investigation and the peculiar behaviour of clitic a in San Donà will hence become more transparent.

To conclude, the genetic sub-classification of Lamonat and Sovramontino is less straightforward than that of Friulian and Fornese. The above description of the four NEIDs under investigation seems to further weaken the unity and the autonomy of Rhaeto-Romance as a language family. The data from Fornese, the Friulian variety of Tricesimo, Lamonat and Sovramontino support the view that Ladin, Friulian and their peripheral dialect areas are conservative or ‘crystallised’ NIDs: they ultimately offer precious insights on a full array of frozen developmental stages of Gallo-Italic.

2.3. Subject Clitics: The Ambiguous Status of NIDs as Null-Subject Languages

In this section I will outline the morpho-syntactic properties of SCLs across the NEIDs under investigation. The aim of this section is two-fold: (i) to provide useful empirical data towards the on-going debate on Northern Italian dialects as non- consistent null-subject languages (Cardinaletti & Repetti 2010) and (ii) to spell-out the

54 theoretical assumptions on the morpho-syntactic behaviour of SCLs I will adopt in the discussion of syntax-pragmatics interface phenomena in the later chapters of this thesis.

2.3.1. Subject Clitics and the Null-Subject Parameter

The set of properties that constitute the Null Subject Parameter (Chomsky 1981) has been a debated topic in syntactic theory. It is a binary parameter, which divides the world’s languages into null-subject languages (henceforth abbreviated as NSLs) and non-null-subject languages. Chomsky (1981) claims that NSLs have subjects that are not assigned a phonetic representation, and hence are not realised phonetically. The subject category is therefore empty, but syntactically filled by pro, which ensures the fulfilment of Chomsky’s Extended Projection Principle [EPP]. Northern Italian dialects have posed a serious challenge to the syntactic characterization of the Null Subject Parameter. More specifically, NIDs challenge its binary nature: they exhibit obligatory overt clitic subject pronouns at the onset of the verbal phrase, but, as I will show in the next subsections, their presence depends on grammatical person and syntactic context.

SCLs are phonetically realised pronominal elements that, if required by the grammatical person or the syntactic context, obligatorily accompany finite verbs, as shown in (16) below:

(16) a. A ì bruta (Fornese) 3SG.F.SCL is ugly.F ‘She is ugly’

b. *ì bruta is ugly.F ‘She is ugly’

Due to their unstressed phonological nature, SCLs are also called atonic pronouns and are differentiated from tonic pronouns, which behave like fully-fledged pronominal subjects. Atonic pronouns (or SCLs) are not as free as tonic pronouns, but have a fixed syntactic position in proximity to the inflected verb (Benincà 1994). In most NIDs, tonic and atonic pronouns co-occur within the same clause. A single inflected verb can therefore have two pronominal elements: an obligatory subject clitic and an optional tonic pronoun (Renzi & Vanelli 1983). All NIDs have a set of atonic pronouns, but their

55 number and obligatoriness is subject to cross-dialectal variation. Some NIDs present a complete set of subject clitics (one for each grammatical person), while others have only a partial set which always includes the second person singular SCL (Benincà 1994).

Tonic pronouns are common to most languages of the world; SCLs, on the other hand, are less common and relatively recent in the development of Romance (Benincà 1994; Poletto 1993, 2000; Salvi 2004). As for the Null-Subject Parameter, within the Romance family, obligatory subjects are found in French, Franco-Provençal and a very limited number of Ladin varieties spoken in the north-western part of Italy. In French and Franco-Provençal, obligatory subjects can indeed assume the form of SCLs (Kayne 1983). Rizzi (1986) and Brandi & Cordin (1989) have however shown that SCLs in NIDs syntactically differ from those found in French since, among other properties: i) they can co-exist with lexical and pronominal subjects, and ii) they must be repeated in coordinated structures. Nevertheless, NIDs vary significantly with respect to the behaviour of SCLs, making it very difficult to develop robust generalisations on the syntax of these elements. There is nonetheless agreement on treating these elements as rich agreement markers between the overt or null subject and the finite verb (Rizzi 1986; Brandi & Cordin 1989; Poletto 2000). In this respect, NIDs still qualify as null- subject languages.

2.3.2. SCLs across NEIDs: Preliminary Considerations

A fine-grained analysis of SCLs and their syntactic status is beyond the scope of the present research. I will provide a comparative description of the SCL system of Sovramontino and Fornese. I will adopt Renzi & Vanelli’s (1983) framework and, with respect to their empirical generalisations, point out the theoretical hypotheses that derive from it. More importantly, from the existing literature on SCLs, I will spell-out the theoretical views that I will adopt in my thesis. Note that the comparison will only focus on Sovramontino and Fornese. This is for three main reasons: i) Sovramontino and Lamonat SCL systems are very similar, if not identical; ii) Fornese SCL system has been considerably less adequately described than Friulian’s; and iii) the comparison between Sovramontino and Fornese SCL systems illustrates well the full array of Renzi

56 and Vanelli’s (1983) typology. I will nonetheless make reference to Friulian whenever required by the discussion.

In order to study the pronominal system of a NID, Renzi and Vanelli (1983) claim that linguists have to consider the following issues: i) if subject pronouns are tonic or clitic; ii) if pronominalization includes all grammatical persons; iii) the phonetic form of the pronominal elements, whether it is similar to other pronominal forms found in different grammatical persons and/or the verb inflection; iv) if the pronominal elements can co-occur with lexical subjects; v) if the pronominal element can be used with quantifiers such as nobody; vi) if the pronominal element can be used with impersonal and meteorological verbs, or in existential constructions; and, finally, vii), whether or not the pronominal element appears after the inflected verb in interrogative sentences; if it does, then it is also important to investigate its phonetic form, especially in comparison with the form it has in affirmative sentences. The study of SCLs hence involves a multifaceted investigation of different morpho-syntactic environments along with the investigation of the morpho-phonological make-up of the SCLs themselves. In the next subsections, I will consider each point with respect to the clitic system of Fornese and Sovramontino.

2.3.3. SCLs in Fornese and Sovramontino: Empirical Generalisations

Both Sovramontino and Fornese have two sets of pronominal elements: a full set of tonic pronouns and a partial set of atonic pronouns. Table 2 and Table 3 show each set of pronominal elements differentiated by grammatical person in the two NIDs:

Atonic Pronouns FORNESE Tonic Pronouns Affirmative Interrogative Singular 1 jo (i) -ja 2 tu tu -tu 3 masculine lui al -lu feminine li a -la Plural 1 nos i -as 2 vos (i) 3 masculine lour i - ɲu feminine lour as -las

Table 2: Tonic and atonic pronouns in Fornese

57 Atonic Pronouns SOVRAMONTINO Tonic Pronouns Affirmative Interrogative Singular 1 mi -e 2 ti te -tu 3 masculine lu/el el/l -lo feminine ela la/l -la Plural 1 noaltri -e 2 oaltri/vu -o 3 masculine lori i -li feminine lore le -le

Table 3: Tonic and atonic pronouns in Sovramontino

As shown in Table 2 and Table 3, SCLs are divided into two further subsets depending on whether they appear in affirmative or interrogative clauses. In root interrogatives, SCLs encliticise onto the inflected verb. Their phonetic realisation and, hence, morpho-phonological form generally varies across the two different sentence types. The SCLs in brackets are optional. As for the blank cells, it means that there is no SCL form associated with the specified grammatical person: there are one or more gaps in the paradigm.

In Fornese and Sovramontino, the behaviour of tonic pronouns mirrors that of Italian subject pronouns: they are mainly employed to convey particular discourse- pragmatic readings and/or for disambiguation purposes. Tonic pronouns are hence optional and only become obligatory in a limited set of discourse-pragmatic contexts (see Renzi & Vanelli 1983). Conversely, SCLs are obligatory and sensitive to clause typing, assuming different morpho-phonological forms depending on whether the sentence encodes declarative or interrogative force. I will also show that in negative and negative interrogative clauses SCLs may have a further different phonetic realization. This is due to the interaction between the negation and the SCLs in the preverbal domain. I will now outline how SCLs are realised in affirmative clauses in the two NEIDs.

58 2.3.3.1. SCLs in Declarative Clauses

Renzi and Vanelli (1983: 127) identify six different dialect types on the basis of the distribution of obligatory SCLs across the different persons of the verb. They investigated the distribution of those grammatical persons that require an obligatory pronominal element across NIDs. Table 4 summarises the six categories:

Obligatory SCLs Category 1 1SG, 2SG, 3SG, 1PL, 2PL, 3PL Category 2 2SG, 3SG, 1PL, 2PL, 3PL Category 3 1SG, 2SG, 3SG, 3PL Category 4 2SG, 3SG, 3PL Category 5 2SG, 3SG Category 6 2SG

Table 4. Renzi and Vanelli’s (1983) NID categories on the basis of the distribution of obligatory SCLs across the different persons of the verb.

NIDs range from the dialects in the first category that have obligatory SCLs for all grammatical persons to the dialects in the sixth category that only have one SCL for the second person singular. Given these six categories, Renzi and Vanelli (1983: 128) made three generalisations that can be summarized as follows:

I) if a NID has an obligatory subject clitic, it is the second person singular SCL; II) if a NID has two obligatory subject clitics, they are the second and third person singular SCLs; III) if a NID has three subject clitics, they are the second and third person singular SCLs and the third person plural SCL.

As shown in Table 3 above, Sovramontino exhibits three obligatory SCLs in the affirmative form: the second and third person singular SCLs and the third person plural SCL. In Renzi and Vanelli’s (1983) six-way partition, Sovramontino belongs to the group of dialects of the fourth category along with Old Venetian, Paduan, Feltrino,

59 Triestino and Ligurian. The categorization of Fornese is instead less straightforward. Fornese behaves like the NEIDs in the first category (i.e. Friulian and Turinese), but the first person singular and the second person plural SCLs are optional. Renzi and Vanelli (1983) acknowledge the existence of optional SCLs in affirmative clauses across NIDs: for certain grammatical persons, the presence or absence of optional SCLs is completely arbitrary. The two variants (presence or absence of the SCL) are hence in free variation7. In order to account for optional SCLs, Renzi and Vanelli (1983: 129) proposed a fourth generalization:

IV) if SCLs are present in the second and third person singular and in the third person plural, they are never optional. Optional SCLs can only be those of first person singular and plural, and second person plural.

The four generalizations neatly hold for both Sovramontino and Fornese. Renzi & Vanelli (1983) also point out that within a single NID there may be internal variation in SCL distribution with respect to the conjugation of the verbs be and have: more specifically, auxiliary be and auxiliary have versus lexical be and lexical have. Auxiliaries have and be are in fact often accompanied by an extra set of SCLs, better known as auxiliary clitics (Poletto 2000; Pescarini 2016). In Sovramontino, this variation is not found. Have and be nonetheless seem to affect the morpho-phonological form of third person singular SCLs el (masculine) and la (feminine): these two forms tend to be reduced to l before third person singular conjugated be and have, as shown in (17) and (18):

(17) a. Maria la magna pom (Sovramontino) Mary 3SG.F.SCL eat.3SG apple ‘Mary eats apples’

b. Maria l a magnà do pom Mary 3SG.F.SCL have.3SG eat.PTCP two apple ‘Mary ate two apples’

c. Maria l è drio magnar el pom Mary 3SG.F.SCL be.3SG after eat.INF the apple ‘Mary is eating the apple’

7 My data from Fornese show idiolectal variation with respect to the presence or absence of optional SCLs. Further research is needed to determine whether or not free variation is the best way of characteristic this alternation. Nonetheless, at first scrutiny, no particular linguistic constraint could be identified. 60 (18) a. El magna de scondon 3SG.M.SCL eat.3SG of hidden ‘He eats secretly [when no one sees him]’

b. L a magnà LU tut el schith 3SG.M.SCL have. 3SG eat.PTCP he all the fried.cheese ‘He ate all the fried cheese’

c. L è LU drio magnar-lo 3SG.M.SCL be.3SG he after eat.INF-3SG.M.OCL ‘He is eating it’

In Fornese, no special set of SCLs is found with be and have. Note, however, that the third person singular feminine SCL a is omitted in front of present tense third person singular auxiliary and lexical have. This is because SCL a and the third person form of have are homophonous in the present tense: the two identically sounding forms are assimilated, leading to the apparent disappearance of the SCL, as shown in (19):

(19) a. La mula a ì bruta (Fornese) The girl 3SG.F.SCL be.3SG ugly ‘The girl is ugly’

b. La mula a bibût un cafe The girl 3SG.F.SCL=have.3SG drink.PTCP a coffee ‘The girl drank a coffee’

In these circumstances, I assume that the SCL that is assimilated to the following morpho-phonological environment is syntactically present in the inflectional domain of the verb.

2.3.3.2. SCLs in Negative Declaratives and Satellite Clitics

I will now point out how SCLs behave in negative declaratives in Sovramontino and Fornese. I will consider SCLs in negative interrogatives in subsection 2.3.3.5, as a separate discussion is needed. In most NIDs, the preverbal negation precedes the SCL. SCLs rely on the inflected verb for their phonological realisation. A SCL can only be separated from the inflected verb by another clitic of a different type (Benincà 1994). Consider the following examples from Sovramontino in (20):

61 (20) a. No la ghe vede gnent (Sovramontino) NEG 3SG.F.SCL LOC.CL see.3SG nothing ‘She can’t see anything’

b. No i me mena fora NEG 3PL.M.SCL 1SG.OCL take.3PL outside ‘They don’t take me outside’

In examples (20), the preverbal negation no precedes the SCLs la and i. The inflected verb and SCL can only be divided by another clitic: in (20a), the SCL and the verb are divided by the locative-presentative clitic ghe, whereas, in (20b) the first person singular object clitic me separates the verb from the SCL. Note that clitic ghe can also carry a generic dative function for third person singular and plural (see Beninca 2007). As in (21) below, the inflected verb and the SCL can also be divided by a clitic cluster: the generic dative clitic ge and the third person singular feminine object clitic la.

(21) I ghe la magna LURI la polenta (Sovramontino) 3PL.M.SCL DAT.CL 3SG.F.OCL eat.3PL they the polenta ‘They will eat his/her/their polenta’

In the next chapters, we will discuss the syntax-pragmatics of (21). For now, let us only note that, in Sovramontino, non-contrastive narrow focal elements (like luri) are realised in post-verbal position and optional bona fide topics (like la polenta) are not fronted, but, nonetheless, resumed by an agreeing object clitic within the domain of the inflected verb. Going back now to our clitic cluster in (21), with some degree of variation, across NIDs the order of clitics is generally (NEG) > SCL > DAT > OCL > LOC/PRESENTATIVE (Benincà 1994).

The clitic cluster in (21) begs the question: how should this stack of clitics be dealt with morpho-syntactically? A satisfactory answer goes beyond the purpose of the present investigation. Nonetheless, I will provide the reader with the theoretical assumptions I adopt in my research to try to answer this question. First, I adopt the view that if a string of sound forms a single phonological unit, then, very likely, it belongs to the same syntactic head (see Frascarelli 2000). It is undoubtedly true that there is no rigorous one-to-one correspondence between prosodic unit and syntactic unit. I will hence take this as a mere guiding principle. The clitic cluster in (21) very likely has

62 some degree of internal structure, but unquestionably the single clitics rely on the inflected verb for their phonetic realisation8. On this basis they are all satellite clitics of the inflected verb. In their tree representation, I will hence place the stack of clitics under the same node as the inflected verb. The close relationship between these clitics, including the SCL, and the inflected domain is also shown by SCL inversion in root interrogatives, whereby the subject clitic inverts with the inflected verb and the other satellite clitics (excluding negation, which deserves a separate discussion) precede the inflected verb (see section 2.3.3.4). Across NEIDs, clitics belong to the inflectional rather than the verbal domain9. In light of their behaviour, I consider SCLs and other satellite clitics of the inflected verb as elements of T°. In the minimalist programme (Chomsky 1995), TP replaces IP (inflectional phrase). Across Romance, T° is the syntactic position that generally hosts the inflected verb (see Cinque 1999). For some authors, V-to-T movement alone is able to satisfy the EPP warranting their null-subject nature (see Alexiadou & Anagnostopoulou 1998). In the present research, I treat SCLs, the inflected verb and its satellite clitics as elements of T°. Note that I will not analyse the internal syntactic structure of T° (the inflectional domain) as details of its internal structure are not directly relevant to the discussion at hand. Goria (2004) treats SCLs as elements hosted by T° across North-Western Italian Dialects. She hence adapts the previous morpho-syntactic analysis of SCLs (Poletto 1993, 2000) to the minimalist programme. In previous accounts, SCLs were analysed as strong agreement markers hosted in AgrSP above the IP. I will adopt the view that SCLs are still, to some extent, strong agreement markers, but they are hosted in T° along with the inflected verb and its satellite clitics with which they form a single phonological unit.

As for the pre-verbal clitic order put forward by Benincà (1994), namely (NEG) > SCL > DAT > OCL > LOC/PRESENTATIVE, both Fornese and Sovramontino abide by it. In Fornese, the preverbal negation always precedes the SCL, which, as is the case with Sovramontino, can only be separated from the inflected verb by a satellite clitic. An important aspect of the Fornese SCL system is that, in the presence of the pre-verbal negator, optional SCLs become obligatory. In other words, in negative declaratives, the optional SCLs for the first person singular and the second person plural become obligatory. This is due to the fact that, in Fornese, the pre-verbal negation na and the

8 In chapter 4, I will argue that if the inflected verbs is an auxiliary, then it also behaves like a clitic and needs adjacency with the tonic past participle for its realisation. 9 See Kayne (1975, 1989) on the view that clitics are TENSE elements. 63 SCLs are assimilated, creating a set of negating SCLs. This is true for all grammatical persons except for the second person singular, where the two elements, namely negation and SCL, do not merge. In Sovramontino, on the other hand, there is no significant difference between SCL distribution in affirmative or negative declaratives. We only observe the reduction of the third person singular SCL el to l before the negative particle no. Table 5, below, summarises these observations:

SOVRAMONTINO FORNESE

Singular 1 n-i manʤi 2 na tu manʤas 3 masculine no-l maɲa n-al manʤa feminine no la maɲa n-a manʤa Plural 1 n-i manʤon 2 n-i manʤeis 3 masculine no i maɲa n-i manʤa feminine no le maɲa n-as manʤa

Table 5. Negation and SCLs in Sovramontino and Fornese

The fact that optional SCLs surface in the presence of the preverbal negator seems to suggest that, in affirmative declaratives, even if SCLs are not phonologically realised, they are nonetheless present at the morpho-syntactic level of the clause. In section 2.3.5 I will come back to this hypothesis and discuss it further.

2.3.3.3. The Morpho-Phonetic Form of SCLs

In the description of SCLs across NIDs, Renzi and Vanelli (1983) group different dialects according to the phonetic form of their SCLs with respect to the different grammatical persons. On this matter, they propose the following generalization:

V) the second person singular SCL always has a different phonetic form from the SCLs in the other grammatical persons (Renzi and Vanelli 1983: 130)

The generalization holds for both Sovramontino and Fornese. As far as the SCLs in the other grammatical persons are concerned, in Sovramontino, all SCLs are phonetically 64 differentiated. This is typical of the NIDs that belong to category 4 (with three obligatory SCLs – see Table 4). As for Fornese, it belongs to category 1, whereby all grammatical persons have an overt SCL (note that first singular and the second plural SCLs are optional). NIDs that belong to the category 1 have very different SCL realizations across the various grammatical persons. In order to capture these differences or similarities, Renzi and Vanelli (1983) proposed five different subcategories for the NIDs that belong to category 1.

Fornese belongs to the subcategory of dialects in which the first person singular and all the plural persons are identical (SCL i in Fornese). In other words, only the second and third person singular SCLs are phonetically differentiated. The only other documented NID that behaves like Fornese is Istrioto, a nearly-extinct dialect spoken in Rovinj, Croatia (Renzi & Vanelli 1983). It is, however, important to note that, in Fornese, the third person plural feminine SCL as has a different phonetic form from the equivalent masculine SCL i. The feminine SCL as is the plural of the third person singular feminine SCL a, which is derived through the suffixing of the sigmatic plural marker –s: the default mechanism to pluralise feminine and masculine nouns in Fornese. Interestingly, in Fornese, the third person plural tonic subject lour is the same for both grammatical genders; conversely, in Sovramontino, it has two different phonetic forms, one for each , lori masculine and lore feminine.

2.3.3.4. SCLs in Root Interrogatives

I will now discuss SCL behaviour in root interrogatives and, subsequently, in root negative-interrogatives. Note that this sub-section will lay the foundation for the analysis proposed in chapter 3 on wh-questions and T-to-C movement in root interrogatives. For root interrogatives I mean both yes/no questions and content questions introduced by an interrogative pronoun (a wh-word) in the matrix clause10. Across NIDs, root interrogatives can be realised in two ways: i) through the inversion of the SCL with the inflected verb, whereby the SCL attaches to the inflected verb via encliticisation, or ii) through the use of raised intonation, maintaining the syntactic structure of root declaratives. Renzi and Vanelli (1983) identified a relatively restricted

10 I will briefly outline the behaviour of embedded interrogatives throughout the subsequent chapters (chapter 3 and chapter 4). 65 group of NIDs that behave in the latter way: they show no clitic inversion (CI) and the question is simply conveyed through intonation. The rest of NIDs employ CI and the SCL becomes encliticised onto the verb. Across this group of NIDs, CI does not necessarily apply to all grammatical persons. With respect to CI, Renzi & Vanelli (1983) identified five different sub-categories of NIDs.

Sovramontino belongs to the sub-group of NIDs that exhibit obligatory SCLI (subject clitic inversion) in all grammatical persons as shown in Table 6 below:

Atonic Pronouns SOVRAMONTINO Affirmative Interrogative Singular 1 -e 2 te -tu 3 masculine el/l -lo feminine la/l -la Plural 1 -e 2 -o 3 masculine i -li feminine le -le

Table 6: Atonic pronouns in Sovramontino

The interrogative set of SCLs attach to the inflected verb as suffixes. Their morpho- phonological form is different from the one that they have in affirmative declaratives with the exception of the third person feminine singular and plural SCLs, respectively la and le. The surface realization of the first person singular interrogative SCL –e is not always overt. This is because it is homophonous with the first person singular verb ending –e (cf. 22b).11 Its presence, however, becomes transparent with a restricted class of verbs whose first person singular verb inflection is either –o (i.e. present indicative of the verb to have or future indicative) or zero as the first person singular of the verb to be, namely son, as shown in example (22d):

(22) a. Magne MI un panet (Sovramontino) eat.1SG I a bread.roll ‘I’ll eat a bread roll’

11 As mentioned in section 2.2.2, first singular verb ending –e is virtually found across all tenses and moods. 66 b. Magne mi sto panet ultimo qua? eat.1SG=1SG.SCL I this bred.roll last here ‘Shall I eat this last bread roll here?’

c. O magnà MI el panet have eat.PTCP I the bread.roll ‘I ate the bread roll’

d. O-e magnà mi el panet? have.1SG-1SG.SCL eat.PTCP I the bread.roll ‘Did I eat the bread roll?’

Note that Lamonat, instead, exhibits zero marking to a greater extent with respect to the first person singular verb ending i.e. dig ‘I say’, ved ‘I see’. In Lamonat, the surfacing of the first singular interrogative SCL –e is hence more transparent.

Fornese belongs to the sub-group of NIDs that employ CI in all grammatical persons except the second person plural. In this sub-group, there is only another documented NID, namely the Franco-Provençal variety spoken in Valle D’Aosta (Renzi and Vanelli 1983). In Fornese, the second person plural SCL is missing in root interrogatives. Its absence, however, is unlikely to be attributable to the assimilation of two homophonous forms, namely the SCL and the verb inflection. In fact, the second person plural verb inflection ends in a consonant –s throughout virtually all finite tenses. Renzi and Vanelli’s (1983) survey suggests that, across most NIDs, interrogative SCLs tend to differentiate themselves morpho-phonetically from their affirmative counterparts. This is true for all interrogative SCLs in Fornese, as shown in Table 7 (except the second singular SCL –tu).

Atonic Pronouns FORNESE Affirmative Interrogative Singular 1 (i) -ja 2 tu -tu 3 masculine al -lu feminine a -la Plural 1 i -as 2 (i) 3 masculine i - ɲu feminine as -las

Table 7: Atonic pronouns in Fornese

67 A crucial difference between Fornese and Sovramontino is that, in Fornese, SCLI is not obligatory: root interrogatives can be simply realised through raised intonation, maintaining the same syntactic structure of root declaratives. (23) shows the two possible ways of asking the same question in Fornese, namely (23a) through SCLI and (23b) through raised intonation:

(23) a. Lui, a-lu durmît ben? (Fornese) He have.3SG-3SG.M.SCL sleep.PTCP well ‘Did he sleep well?’

b. Lui, al a durmît ben? He 3SG.M.SCL have.3SG sleep.PTCP well ‘Has he slept well?’

Further investigation is needed to establish whether the two interrogation strategies in (23) are in free variation or serve different discourse-pragmatic or syntactic purposes. For now, I can only acknowledge that both strategies, SCLI and raised intonation, are grammatical in Fornese.

2.3.3.5. SCLs in Root Negative-Interrogatives

As for root negative-interrogatives, SCLI is ungrammatical in Sovramontino and optional in Fornese. (24) and (25) show how root negative-interrogatives can be realized with respect to SCLI in Fornese and Sovramontino respectively:

(24) a. Na riva-lu? (Fornese) NEG arrive.3SG-3SG.M.SCL ‘Isn’t he arriving?’

b. Na-l riva? NEG-3SG.M.SCL arrive.3SG ‘Isn’t he arriving?’

(25) a.*No rive-lo? (Sovramontino) NEG arrive.3SG-3SG.M.SCL ‘Doesn’t he arrive?’

b. No-l riva? NEG-3SG.M.SCL arrive.3SG ‘Doesn’t he arrive?’

68 In Fornese, if SCLI takes place the subject clitic encliticises onto the inflected verb and leaves the negation in its bare form na. The optionality of SCLI in root negative- interrogatives is in line with what we found in root interrogatives. In Sovramontino, SCLI is not possible in negative-interrogatives, hence the ungrammaticality of (25a). Sovramontino resorts to intonation to convey negative interrogatives (25b). Benincà (1994) points out that, in Paduan, SCLI in negative-interrogatives is usually absent, but it is preferred in a restricted set of morpho-syntactic contexts: i) when the question is introduced by the interrogative pronoun parcosa, iii) when the question includes the post-verbal negation miga, or ii) when the question appears in a disjunctive coordinate structure. In chapter 3, I will investigate why SCLI is blocked by the negative element in Sovramontino and Lamonat. I will claim that the lack of SCLI in negative- interrogatives is the by-product of a more general structural difference between Lamonat-Sovramontino and Friulian-Fornese, namely the variation in the morpho- syntactic locus in which SCLI takes place.

Lastly, in their description of SCLs across NIDs, Renzi and Vanelli (1983: 139) proposed a further generalisation to capture the relationship between the number of interrogative SCLs and the number of affirmative SCLs with respect to the different grammatical persons:

VI) if a variety employs CI, the number of persons with obligatory SCLs in interrogative sentences is equal or higher to the number of SCLs in the affirmative form.

This generalization holds for both Sovramontino and Fornese. Sovramontino has a complete set of interrogative SCLs, but has only three SCLs in the declarative form. Fornese has one fewer interrogative SCLs with respect to the number of declarative SCLs. Nevertheless, Fornese has a higher number of obligatory SCLs in the interrogative form than in declarative form: only the optional second person plural SCL in the declarative form is missing in the interrogative form. In section 2.3.5, I will further discuss the difference between declarative and interrogative SCLs. More specifically, I will raise the question of whether or not we should consider the two sets of SCLs as containing different elements or the same elements in a different morpho- syntactic distribution.

69 2.3.4. SCLs in Different Syntactic Contexts

Renzi and Vanelli (1983) also describe the behaviour of SCLs across NIDs with respect to different syntactic contexts, namely:

1) existential constructions; 2) meteorological and impersonal verbs; 3) inverted subject constructions; 4) the presence of SCLs with lexical subjects; 5) the presence of SCLs with the quantifiers chi (who) and nessuno (nobody).

Renzi and Vanelli’s (1983) findings have informed the discussion on the morpho- syntactic status of SCLs across NIDs for the last three decades. I will not directly contribute to this debate. However, for the sake of completeness, I will point out the morpho-syntactic behaviour of Sovramontino and Fornese in relation to the five syntactic contexts listed above. In their investigation, Renzi and Vanelli (1983) found a great deal of variation across and within NIDs. They tried to constrain this variation by putting forward the following two generalisations:

VII) if and only if a variety has obligatory subject clitics for all grammatical persons, then it will obligatorily have SCLs in inverted subject, existential and impersonal constructions and with meteorological verbs; VIII) if a variety obligatorily requires SCLs with the quantifiers ‘chi’ and ‘nessuno’, then this variety has obligatory SCLs for all grammatical persons (Renzi and Vanelli 1983: 137)

2.3.4.1. SCLs in Different Syntactic Contexts: Fornese and Sovramontino

As far as existential constructions are concerned, both Sovramontino and Fornese do not exhibit SCLs, but other clitics of a different nature. For instance, Sovramontino exhibits the locative-presentative clitic ge accompanied by the partitive clitic ne and the third singular form of verb be in a construction that literally means ‘there of-it is’, as

70 shown in example (26a). In the interrogative form, however, Sovramontino also requires the presence of a non-agreeing SCL, which resembles the third person singular masculine SCL el. It undergoes CI and is realised as the suffix –lo. (26b) shows that the SCL in question is non-agreeing, as the ‘logical subject’ Maria is a feminine:

(26) a. Ghe-n é tre bocha (Sovramontino) LOC.CL-PRT.CL be.3SG three children ‘There are three children’

b. Ghe-n e-lo Maria? LOC.CL-PRT.CL be.3SG-3SG.M.SCL Mary ‘Is there Maria?’

Fornese exhibits a different strategy in existential constructions. Fornese has no locative-presentative clitic, but an invariant clitic a that, to some extent, resembles an expletive element. In existential interrogatives, clitic a does not undergo CI and in existential negative declaratives, it does not follow the negative element, as shown in (27):

(27) a. A ì tanç bois (Fornese) CL be.3SG many boys ‘There are a lot of boys’

b. A era tantas boisas? CL be.3SG.PST many girls ‘Were there many girls?’

c. A no ì nissun a ciasa CL NEG be.3SG nobody at home ‘There isn’t anyone home’

In this respect, Fornese clitic a does not behave like a SCL. Given its morpho- phonological form and morpho-syntactic behaviour, it may be regarded as a specialised invariant vocal clitic (see Benincà 1994; Poletto 2000; Bernini 2012). Further research is needed to establish the exact morpho-syntactic behaviour of clitic a in Fornese. Nonetheless, empirically speaking, it appears in those syntactic contexts that would require an overt expletive in a non-null-subject language (see Bolinger 1977 for early discussion on expletive typology; Falk 1993; Biberauer & Roberts 2010). This empirical generalisation will become more and more evident in the next few paragraphs

71 As far as meteorological verbs are concerned, Sovramontino, in a similar fashion to many other NIDs, requires the aforementioned non-agreeing SCL el. This is shown in example (28) along with the negative and interrogative form of the same sentence:

(28) a. El piove (Sovramontino) 3SG.M.SCL rain.3SG ‘It rains’

b. *Piove rain.3SG ‘It rains’

c. No-l piove NEG-3SG.M.SCL rain.3SG ‘It doesn’t rain’

d. Piove-lo? Rain.3SG-3SG.M.SCL ‘Does it rain?’

On the other hand, Fornese does not exhibit any SCL, but the presence of clitic a is required for the sentence to be grammatical, as shown in (29):

(29) a. A niviê (Fornese) CL snow.3SG ‘It snows’

b. A na niviê CL NEG snow.3SG ‘It doesn’t snow’

c. A niviê? CL snow.3SG ‘Does it snow?’

Fornese does not license SCLs in existential constructions or with meteorological verbs. Following Renzi & Vanelli’s (1983) generalisations, Fornese should also not license SCLs in impersonal constructions. The prediction is indeed borne out, as shown in (30):

(30) A si fidi-si simpri masa (Fornese) CL IMP trust.3SG-IMP always too.much ‘One trusts people too much’

72 Note that the SCL is absent, but clitic a is present. Sovramontino obligatorily requires the non-agreeing subject clitic el with meteorological verbs, but only licenses it in the interrogative form of the existential constructions. As for Sovramontino impersonal constructions, SCLs are not found, as shown in example (31):

(31) Le patate, se dis le cartufole a Lamon (Sovramontino) the potatoes, IMP say.3SG the potatoes at Lamon ‘Potatoes? They say cartufole in Lamon’

As far as inverted subject constructions (i.e. presentational) are concerned, Fornese, exhibits either i) the invariant clitic a, as in (32), or ii) a SCL agreeing with the inverted subject, as shown in (33):

(32) a. A riva nomi me pari (Fornese) CL arrive.3SG only my father ‘Only my father is coming’

b. A ì muart duç i flors CL be.3SG die.PTCP all the flowers ‘All the flowers died’

(33) a. I en saltât four trei gias 3PL.M.SCL be.3PL jump.PTCP out three cats ‘Three cats jumped out’

b. I en schampâs i cians 3PL.M.SCL be.3PL escape.PTCP the dogs ‘The dogs ran away’

This variation can be constrained by looking at the verb class of the different lexical verbs: verb-subject construction featuring core unaccusative verbs adopt strategy (i), whereas verb-subject constructions featuring non-core unaccusatives adopt strategy (ii). In Lamonat and Sovramontino, the class of core unaccusative verbs includes ‘die’, ‘fall’ and ‘arrive’ (see Levin & Rappaport Hovav 1995). Similarly to Fornese, Sovramontino does not license SCLs in inverted subject constructions featuring core unaccusative verbs. On the other hand, inverted subject constructions featuring non-core unaccusative verbs require an agreeing SCL, as shown in (34):

(34) a. Riva Toni (Sovramontino) arrive.3SG Toni ‘There arrives Toni’

73 b. *El riva Toni 3SG.M.SCL arrive.3SG Toni ‘There arrives Toni’

c. I è scampà i can 3PL.M.SCL be.3PL escape.PTCP the dogs ‘The dogs ran away’

Both Sovramontino and Fornese allow SCLs to co-occur with tonic pronouns and lexical subjects. As shown respectively in examples (35) and (36):

(35) L a magnà TONI la polenta (Sovramontino) 3SG.M.SCL have.3SG eat.PTCP Toni the polenta ‘Tony ate the polenta’

(36) Giovanni al ven diman (Fornese) Giovanni 3SG.M.SCL come.3SG tomorrow ‘Giovanni will come tomorrow’

More precisely, in Sovramontino and Fornese, the presence of SCLs is obligatory with lexical and pronominal subjects (unless they appear in a genuine inverted subject construction).

Finally, let us look at the behaviour of SCLs with respect to the quantifiers nessuno (nobody) and chi (who). In Sovramontino, SCLs are always absent in this syntactic context, as shown in (37) and (38):

(37) a. Nesun magna polenta (Sovramontino) Nobody eat.3SG polenta ‘Nobody eats polenta’

b. Nesun magna polenta? Nobody eat.3SG polenta ‘Does anyone eat polenta?’

(38) No so chi che magna polenta NEG know.1SG who that eat.3SG polenta ‘I don’t know who eats polenta’

Similarly to Sovramontino, Fornese does not license SCLs with the quantifiers nessuno (nobody) and chi (who). In this syntactic context, Fornese once again employs the invariant clitic a, as shown in examples (39) and (40):

74 (39) Nisun a mangia polenta (Fornese) Nobody CL eat.3SG polenta ‘Nobody eats polenta’

(40) N-i sai si c-a mangia polenta NEG-1SG.SCL know.1SG who that-CL eat.3SG polenta ‘I don’t know who eats polenta’

A fine-grained investigation of clitic a in Fornese is needed to (i) fully understand how it interacts with SCLs, and (ii) define its grammaticalisation trajectory. As for the latter point, clitic a seems to be developing into a structural placeholder associated with the subject position.

2.3.5. The Morpho-Syntactic Status of SCLs: Further Assumptions

Brandi and Cordin (1989) and Rizzi (1986) have claimed that the subject clitics in the NIDs are rich agreement markers between a referential pro and the inflected verb. Subsequent research has further supported this view and devised a fine-grained typology of pre-verbal clitics and their morpho-syntactic behaviour across NIDs (Poletto 1993, 2000). According to this line of research, the crucial difference between NIDs and Standard Italian is that the agreement specification in the NIDs receives phonetic content in its syntactic locus, namely AgrSP in pre-minimalist terms.

Brandi and Cordin (1989) show that SCLs do not behave syntactically like nominal or pronominal subjects. SCLs do not occupy the structural subject position (SpecIP or SpecTP), but are part of the verb inflection under the inflection node (I or T) along with the auxiliary verb and any other clitic that separate SCLs from the inflected verb. Brandi and Cordin (1989) reach this conclusion by comparing the syntactic behaviour of SCLs in French12, a non-NSL, with that of SCLs in NIDs, focusing on Fiorentino and Trentino. Despite their apparent resemblance to the SCLs of NIDs, the French atonic pronominal elements behave like lexical and pronominal subjects. First, French SCLs are obligatory for all grammatical persons, whereas, the majority of NIDs, exhibit one ore more gaps in the SCL paradigm. Second, unlike the SCLs of the NIDs, French SCLs cannot co-occur with lexical or tonic pronominal subjects, as shown in

12 French SCLs are in fact considered weak pronominal elements (Cardinaletti and Starke 1999). I label them SCLs here to directly compare them to the SCLs of the NIDs. 75 (41) below. A lexical subject cannot coexist with the atonic pronominal element, as they would compete for the same theta-role and syntactic position.

(41) a. Pierre parle (French) Pierre speak.3SG ‘Pierre speaks’

b. Il parle 3SG.M.SCL speak.3SG

c.*Pierre il parle Pierre 3SG.M.SCL speak.3SG

Rizzi (1986) also investigates the presence of SCLs in coordinate structures. In French, coordination at the IP or TP level can only be predicated of a single SCL that cannot be reiterated in the second part of the coordinated structure, as shown in example (42):

(42) a. Elle parle et dance (French) 3SG.F.SCL speak.3SG and dance.3SG ‘She speaks and dances’

b.*Elle parle et elle dance 3SG.F.SCL speak.3SG and 3SG.F.SCL dance.3SG ‘She speaks and dances’

Across the NIDs, (42a) is not possible: the SCL must also be repeated in the second part of the coordinated structure. This is true for both Sovramontino and Fornese, as shown in examples (43) and (44) respectively:

(43) a. L è rivà e l a magnà (Sovr.) 3SG.M.SCL be.3SG arrive.PTCP and 3SG.M.SCL have.3SG eat.PTCP ‘He arrived and ate’

b.*L è rivà e a magnà 3SG.M.SCL be.3SG arrive.PTCP and have.3SG eat.PTCP ‘He arrived and ate’

(44) a. Al ì rivât e al a mangiât (Fornese) 3SG.M.SCL be.3SG arrive.PTCP and 3SG.M.SCL have.3SG eat.PTCP ‘He arrived and ate’

76 b.*Al ì rivât e a mangiât 3SG.M.SCL be.3SG arrive.PTCP and have.3SG eat.PTCP ‘He arrived and ate’

The evidence provided by Renzi and Cordin (1989) and Rizzi (1986) convincingly shows that there are structural differences between French and the NIDs with respect to the morpho-syntactic behaviour of atonic pronouns.

In the present research, I will hence consider SCLs as strong agreement markers that spell out the agreement relationship between the inflected verb and the subject, be it pro, a pronominal or a lexical subject. This definition seems to accurately capture the morpho-syntactic behaviour of SCLs in the four NEIDs under investigation. Nonetheless, I cannot dismiss the possibility that SCLs do not uniformly behave like agreement markers across the NIDs. Renzi and Vanelli’s (1983) survey has shown that there is a great deal of variation with respect to SCL behaviour across NIDs. This variation is not only found across NIDs, but may also be present within the same dialect, whereby SCLs may behave differently depending on the syntactic context in which they appear. In this respect, Benincà (1994) points out that Paduan SCLs (second singular and third singular and plural) do not indiscriminately behave like agreement markers, but they can also function as fully-fledged subjects. She argues that, in Paduan, if the subject position is already occupied, SCLs cannot be overtly realised. For instance, the obligatory third person singular SCL becomes optional when it co-occurs with a lexical subject and if they both appear in a clause, the lexical subject is always topicalised. The case of Paduan opens the possibility that, within the same NID, some SCLs in certain syntactic environments behave more like weak pronominal elements rather than agreement markers, making it harder to draw any generalisations on their morpho-syntactic status.

Benincà (1994) also challenges the view that across NIDs the obligatory use of SCLs compensates for the morpho-phonetic process of levelling of verb inflections. This view naturally follows from the observation that Null-Subject Languages (NSL) tend to have a richer morphological agreement system in the verbal inflection than non- NSLs (Chomsky 1981). Friulian, however, is a counter-example to this. As shown in Table 8, in Friulian, all verb endings are differentiated, thus Friulian SCLs are very unlikely to have the function of further differentiating these inflections.

77 FRIULIAN (Tricesimo) Singular 1 o mangj-i 2 tu mangj-is 3 masculine al mangj-e feminine e mangj-e Plural 1 o mangj-ìn 2 o mangj-ais 3 masculine a mangj-in feminine e mangj-in

Table 8. Present indicative conjugation of the verb ‘to eat’ in Friulian.

On the relationship between the morpho-phonetic form of SCLs and verb inflections across NIDs, Renzi and Vanelli (1983: 133) proposed the following generalisation:

IX) in a NID, if two or more subject clitics are homophonous, the corresponding verb inflections are well differentiated from one other; viceversa, if the verb conjugation has homophonous verb inflections, the correspondent subject clitics are phonologically different from one other.

The generalisation holds for both Sovramontino and Fornese. In Sovramontino, the verb inflections for third person singular and plural are identical, but the corresponding subject clitics have different forms. In Fornese, the opposite is true: the first singular and plural persons and second plural person have the same subject clitic i.e. i, but the corresponding verb inflections are well differentiated. This is shown in Table 9 below:

SOVRAMONTINO FORNESE

singular 1 i mang-i 2 3 masculine el magna feminine la magna plural 1 i mangi-on 2 i mang-eis 3 masculine i magna feminine le magna

Table 9. SCL forms vs. verb inflection forms in the relevant grammatical persons in Sovramontino and Fornese.

78 On a higher level, SCLs can be seen as one of the manifestations of the resetting of the Null-Subject Parameter. In the second part of the chapter we have seen that, for example, Fornese is developing an expletive-like element that seems to function as a structural placeholder for the subject position, namely clitic a. Along the same line, Sovramontino has developed a non-agreeing third person singular SCL el that surfaces in those syntactic environments in which non null-subject languages would require an expletive. This non-agreeing SCL very likely marks agreement between a non- referential pro (see Rizzi and Shlonsky 2006 for expletive pro) and the inflected verb. In this respect, el functions as an agreement marker, but, by virtue of encoding a non- agreeing relation, challenges the view that SCLs are agreement markers in the first place. Moreover, in addition to the development of expletive-like elements, in the next two chapters, I will show that subject movement is much more constrained in the dialects under investigation than in traditional null-subject languages like Italian and Spanish.

For these reasons, I adopt Cardinaletti and Repetti’s (2010) label ‘non-consistent null-subject languages’ when it comes to categorising the NIDs with respect to the Null-Subject Parameter (Chomsky 1981). The label captures the insight that NIDs are null-subject languages, but exhibit non-canonical syntactic phenomena that are not expected in NSLs. Cardinaletti and Repetti (2010) more radically argue that across NIDs pro is only licensed in those grammatical persons that lack a SCL form. This claim is certainly noteworthy, but perhaps too extreme, as it does not capture the diversity in SCL morpho-syntactic behaviour across NIDs. I will nonetheless adopt Cardinaletti and Repetti’s (2010) hypothesis that NIDs only have one single set of SCLs that surfaces in different morpho-syntactic positions depending on the force of the clause (see Rizzi 1997) and the surrounding morpho-phonological environment. The one-paradigm hypothesis stands in stark contrast with the two-paradigm hypothesis, whereby there exist two distinct sets of SCLs: one for declarative sentences and one for interrogative sentences (Poletto 1993, 2000). The two-paradigm hypothesis is supported empirically, in that: i) the number of interrogative SCLs is higher than the number of declarative SCLs, ii) the morpho-phonological form of interrogative SCLs is generally different from the form of declarative SCLs, and iii) interrogative SCLs have a different morpho-syntactic distribution with respect to declarative SCLs. On the other hand, the one-paradigm hypothesis is sounder in terms of economy: we would, otherwise, have to

79 justify on theoretical grounds why two different sets of the same morpho-syntactic elements are needed in the language system. In adopting the one-paradigm hypothesis, (ii) and (iii) are not overwhelmingly problematic. Depending on the moprho- phonological environment in which the SCL appears, it changes its morpho- phonological form. This is especially true if we bring the diachronic development of SCLs into the equation: SCLs generally developed from tonic subject pronouns (Vanelli 1998). As as far as (iii) is concerned, SCLI must take place from the underlying or default declarative form of the SCL, which inverts with the inflected verb. I will further discuss this point in the next chapter. If there are two different sets of SCLs, then there is no SCLI at the morpho-syntactic level, but simply the merging of the interrogative SCL into the appropriate syntactic position (namely Poletto’s 2000 GroundP). The more challenging point to address is to justify the disparity in number between the number of interrogative and declarative SCLs. Renzi and Vanelli’s (1983: 139) generalisation on the matter states that: if a variety employs CI, the number of persons with obligatory SCLs in interrogative sentences is equal or higher to the number of SCLs in the declarative form. Along the lines of Cardinaletti and Repetti (2010), I will hence assume that if a grammatical person has an obligatory interrogative SCL form, it will also have a corresponding declarative form, which can be either silent or phonetically realised. This assumption becomes more tenable if we briefly review the data from Fornese and Sovramontino with respect to SCL behaviour. In Fornese, optional declarative SCLs become obligatory (hence phonetically realised) in negative declaratives. In Sovramontino, the non-agreeing third person SCL el surfaces in questions featuring core unaccusative verb/subject inversion and existential constructions. Both in Lamonat and Sovramontino, the interrogative set of obligatory SCLs is heavier than the declarative one, suggesting that SCLs more easily surface in questions13. I will hence consider a NID as having gaps in its SCL paradigm, if it lacks one or more interrogative SCLs across its different grammatical persons.

To conclude, the straightforward empirical observation that most NIDs have gaps in their SCL paradigm is enough evidence not to challenge the claim that NIDs are NSLs. However, in a NID, we cannot assume a priori that SCLs behave indiscriminately like agreement markers. Trying to capture the complexity of SCL

13 In chapter 3, I will argue that SCLI is actually necessary to satisfy the structural requirements on the well-formedness of questions (Rizzi 1991) across NEIDs. 80 behaviour across NIDs by forcing them into a pre-established category may, in fact, lead to unfaithful and inaccurate conclusions.

2.4. Conclusion

In this chapter, I provided an overview of the four NEIDs under investigation: Fornese, Friulian, Lamonat and Sovramontino. The discussion of the four NEIDs further weakens the unity and autonomy of Rhaeto-Romance as a language family. The data from Fornese, the Friulian variety of Tricesimo, Lamonat and Sovramontino, in fact, support the view that Ladin, Friulian and their peripheral dialect areas are simply conservative NIDs, whose defining characteristics were once shared by virtually all NIDs. NEIDs can hence be described as a dialect continuum that includes varieties of Friulian, Ladin, Venetan and Gallo-Italic. In the second part of the chapter, I provided a description of a characteristic trait of the NIDs namely their SCL system. I also stated the theoretical assumptions on the morpho-syntactic behaviour of SCLs that I will adopt in the discussion of syntax-pragmatics interface phenomena in the subsequent chapters of this thesis. SCLs are the manifestation of the non-consistent null-subject status of NIDs (Cardinaletti & Repetti 2010). Although a unitary treatment of SCLs across the NIDs is very desirable, it is hardly achievable. Across the NIDs, SCL behaviour may be conceived as a continuum, whereby SCLs behave, on one hand, like weak pronominal elements (Cardinaletti and Starke 1999) and, on the other, like strong agreement markers. This type of variation is not only cross-dialectal, but also dialect-internal depending on the syntactic environment in which the SCL occurs. In a diachronic perspective, subject clitics that developed from tonic subject pronouns (Vanelli 1998) very unlikely developed simultaneously in all syntactic contexts as agreement markers. Instead, it is plausible that in some syntactic contexts, SCLs have retained a weak subject pronoun behaviour. On this matter, Benincà (1994) notes how the status of SCLs in Friulian has changed over time. She argues that in archaic Friulian SCLs were not always obligatory, but they were ungrammatical if used with lexical subjects or with the quantifier nessuno: in this respect, Friulian SCLs have become more and more similar to agreement markers.

81 3. Microparametric Variation in the Locus of Subject Clitic Inversion: Apparent Wh-In-Situ

3.1. Introduction

The present chapter analyses the syntactic behaviour of wh-elements in Lamonat and Sovramontino. In order to strengthen the discussion of the phenomenon, I will also draw upon the data from another NEID, namely Friulian. Lamonat and Sovramontino are two closely related NEIDs spoken in the south-eastern part of the province of Belluno. They are part of the supra-dialectal area of Bellunese. Bellunese has caught the attention of several linguists due to a peculiar syntactic phenomenon that is only found in a handful of NIDs: wh-in-situ (Benincà 1994; Munaro 1998, 1999; Poletto & Pollock 2001, 2004; Manzini & Savoia 2005, 2011). Wh-in-situ is also documented in northern Venetan varieties and in the eastern and Alpine varieties of Lombard (Munaro & Poletto 2002; Manzini & Savoia 2005).

Lamonat and Sovramontino exhibit both canonical and in-situ wh-words, as shown in (1), (2) and (3):

(1) E-lo ndà aonde Toni? (Sovramontino) be.3SG-3SG.M.SCL go.PTCP where Toni ‘Where did Toni go?’

(2) Ond-e-lo ndà Toni? where-be.3SG-3SG.M.SCL go.PTCP Toni ‘Where did Toni go?’

(3) Co che machina e-lo ndà Toni? with which car be.3SG-3SG.M.SCL go.PTCP Toni? ‘With which car did Toni go?’

(1) is an in-situ wh-question, in which the wh-element is realised postverbally, whereas (2) and (3) resemble two canonical wh-questions. In this chapter, I will only partially discuss type (3) wh-questions and account for the alternation between type (1) and type (2) wh-questions. Type (3) is an instance of discourse-linked wh-question (in the sense of Pesetsky 1987): whereby the answer is drawn from a set of entities that were previously introduced in discourse. The answer to a d-linked wh-question is hence

82 discourse-prominent and, as I will show in chapter 5, the d-linked wh-element occupies a discourse-pragmatically salient left-peripheral position. Type (1) and type (2) are instead equivalent in meaning. Ultimately, the different distribution of wh-elements (1) to (3) can be accounted for in light of the interaction between syntax and discourse- pragmatics. More specifically, the present analysis pivots around two basic tenets: (i) wh-elements can serve different discourse-pragmatic functions, and (ii) wh-elements do not form a homogeneous class, but exhibit morpho-syntactic differences. In this respect, XPs encoding different discourse-pragmatic readings (i.e. contrastive vs. non- contrastive) tend to occupy different syntactic positions and XPs exhibiting different morpho-syntactic statuses (i.e. clitic vs. tonic) tend occupy different structural positions (i.e. ° vs. SpecXP). The present analysis and discussion of wh-in-situ in Bellunese will hence depart from the existing accounts of the phenomenon (Munaro 1998, 1999; Poletto & Pollock 2001, 2004, 2009; Munaro & Pollock 2005; Manzini & Savoia 2005, 2011), which seek an explanation on purely syntactic grounds.

Wh-in-situ across NEIDs is only attested in root interrogatives (Munaro 1998, 1999, Poletto & Pollock 2004). As far as the varieties of Lombard that exhibit the phenomenon are concerned, wh-in-situ can also be found in embedded interrogatives (Manzini & Savoia 2011). In Lamonat and Sovramontino, embedded wh-in-situ yields ungrammaticality, as shown by (4b):

(4) a. No se sa che te ö (Lamonat) NEG know.1SG what that 2SG.SCL want.2SG ‘I don’t know what you want’

b .*No se che te ö che NEG know.1SG that 2SG.SCL want.2SG what ‘I don’t know what you want’

The ungrammaticality of (4b) could be considered enough evidence to support the claim that wh-in-situ is not a unitary phenomenon across NIDs. I will in fact argue that, underlyingly, Bellunese wh-in-situ cannot be considered the same phenomenon as Lombard wh-in-situ: the core empirical evidence put forward by Manzini and Savoia (2005, 2011) in their account of Lombard wh-in-situ finds no correspondence in

83 Lamonat and Sovramontino. The present analysis and discussion will therefore only focus on root interrogatives and try to capture the Bellunese-type wh-in-situ.

Drawing on Rizzi’s (1997) split CP-model, I will ascertain how the present account of the behaviour of wh-words shapes the left periphery of root interrogatives in the dialects under investigation. In this respect, the investigation of subject clitic inversion (SCLI) will be essential to establish how the make-up of the left peripheral space allows for different realisations of the wh-element. I will argue that, in Lamonat and Sovramontino, wh-in-situ is only apparent: it is a manifestation of a parametric choice that involves the locus of SCLI. As mentioned in chapter 2, subject clitics (SCLs) are a characteristic trait of NIDs and, especially among NEIDs, SCLI generally arises in an obligatory fashion in root interrogatives (see Renzi & Vanelli 1983; Brandi and Cordin 1989; Poletto 2000), as shown in (5):

(5) a. Al can al a magnà pulito (Lamonat) the dog 3SG.M.SCL have.3SG eat.PTCP well ‘The dog ate well’

b. A-lo magnà pulito l can? have.3SG-3SG.M.SCL eat.PTCP well the dog ‘Did the dog eat well?’

The yes/no question in (5) shows SCLI, whereby the SCL al inverts with the inflected verb a, forming a single morpho-phonological unit, namely a-lo. As the term suggests, the locus of SCLI is the left peripheral projection in which SCLI takes place. I will argue that, across NEIDs, SCLI is necessary for the well-formedness of questions: the verb moves to the left peripheral space to satisfy the interrogative force of the utterance. I will claim that the projection targeted by T-to-C movement in root interrogatives is not the same across NEIDs, but is subject to microparametric variation. Lamonat and Sovramontino wh-in-situ can hence be resolved by postulating that T-to-C movement in the two NEIDs targets a projection higher than that hosting the fronted wh-element. I will show that the locus of SCLI in Lamonat and Sovramontino is ForceP, whereas in Friulian it is FinP. The comparison between Lamonat-Sovramontino and Friulian will reveal that the wh-distribution in (1) and (2) can be accounted for in these terms. As for the preverbal wh-phrase in (3), it will be accounted for by postulating a further constraint on the satisfaction of a salient discourse-pragmatic [EDGE] feature.

84 In Lamonat and Sovramontino, the different distribution of wh-items is hence constrained by both syntactic and discourse-pragmatic constraints. By considering the syntactic distribution of wh-items as a syntax-pragmatics interface phenomenon, the discourse-pragmatic interpretation of wh-items themselves cannot be disregarded. I will provide support to the well-established assumption that wh-items are focal elements (Bianchi 1999), carrying a [FOCUS] feature along with a [WH] feature: wh-items always refer to some novel and/or discourse prominent piece of information that the speaker wants to elicit from the hearer. In my formal model of the left peripheral space of root interrogatives, I will try to account for the discourse-pragmatic nature of wh- elements and I will hence reject any ad hoc wh-operator position. Let us now consider the existing analyses of wh-in-situ in Bellunese and state how and why the analysis put forward in the present chapter is a valid alternative.

3.2. Bellunese Wh-In-Situ: Existing Analyses

Wh-in-situ is a feature of Bellunese that has been long noted. Several scholars have investigated this phenomenon: the two most influential and conflicting accounts are those of Poletto and Pollock (2004, 2009) and Manzini and Savoia (2011)14. I will briefly summarise their analyses and state how and why they differ from mine.

Munaro (1998, 1999) and Poletto and Pollock (2004, 2009) 15 agree that the in- situ position of wh-words is only apparent: the wh-element is not in a TP-internal argument position, but has moved to the C-domain. They claim that the wh-element only deceptively appears in-situ: in reality, it has undergone movement to the left periphery in the same fashion as canonical wh-words in Italian and in other Romance languages. The account of wh-in-situ in Bellunese put forward in the present chapter fully supports this claim, but differs in explaining the alternation between apparently in-

14 Note that two current PhD candidates, Giulia Donzelli and Caterina Bonan are working on wh-in-situ across NIDs at the University of Zurich and at the University of Geneva respectively. They focus on Northern Trevigiano and Lombard. To the best of my knowledge, their research is still in progress. Leaving Lombard aside, the analysis of wh-in-situ for Northern Trevigiano makes use of Belletti’s (2001, 2004) lower periphery of the VP (Bonan p.c.). This analysis departs from the analysis put forward in the present chapter. In chapter 4, I will discuss why, in Lamonat and Sovramontino, Belletti’s VP periphery does not make an ideal analysis of the phenomenon. 15 See also Munaro (1997); Munaro, Poletto & Pollock (2001); Munaro & Poletto (2002); Munaro & Pollock (2005). 85 situ and canonical wh-elements with respect to a number of points.

First, I will argue that in Lamonat and Sovramontino SCLI involves head movement, more specifically T-to-C movement of the verb cluster. The complex T° hosts the finite verb, the past participle (if present) and any satellite clitics. In these two varieties of Bellunese, nothing can intervene between an auxiliary and its past participle except heavily phonologically reduced low (see chapter 4). By considering SCLI as the result of T-to-C movement in questions, we can explain why SCLI is not possible under certain conditions (i.e. negative-interrogatives). Under Poletto and Pollock’s (2004) account, SCLI is instead derived through overt remnant phrasal movement to the C-domain. Poletto and Pollock (2004) argue that after wh-moment takes place, the whole TP layer undergoes remnant movement to the left periphery, as shown in the derivation of the apparent wh-in-situ question in (6):

(6) A-tu parichà che? (Bellunese – Poletto & Pollock 2004: 254) have.2SG-2SG.SCL prepare.PTCP what ‘What did you prepare?’

Input: [IP te a parichà che] Wh-movement: [CP che C° [IP te a parichà che]] Remnant TP Movement: [CP [IP a-tu parichà] C° [CP che C° [IP te a parichà che]]]

In their arguing against Poletto and Pollock’s (2004, 2009) analysis, Manzini and Savoia (2011) point out that, in Bellunese, the only element actually vacating the TP layer is the past participle (if present) along with the inflected verb, as shown more transparently in (7):

(7) G-a-lo dat che Mario al can? (Lamonat) DAT.CL-have.3SG-3SG.M.SCL give.PTCP what Mario to-the dog ‘What did Mario give to the dog?’

The fact that everything else, namely the subject Mario and the oblique prepositional phrase al can, remains in canonical position weakens the remnant movement analysis, whereby we would expect the wh-element to come last in the sentence. Postulating T- to-C head movement hence better captures the data from Lamonat and Sovramontino.

Second, as far as SCLI is concerned, I do not adopt the hypothesis that NIDs exhibit two different sets of subject clitics in declarative and interrogative clauses (see chapter 2). Poletto (1993, 2000) argues that these two distinct sets of SCLs, declarative

86 and interrogative, occupy two different syntactic positions. I instead adopt the one- paradigm hypothesis and treat SCLs as belonging to a complex T° along with the other satellite clitics of the verb cluster. In Lamonat and Sovramontino, the interrogative paradigm is heavier than the assertive one: SCLs in declarative clauses show gaps in their paradigms and have different phonetic realisations with respect to interrogative SCLs, as shown in Table 1:

Atonic Pronouns

Affirmative Interrogative singular 1 -e 2 te -tu 3 masculine el/l/e -lo feminine la/l -la plural 1 -e 2 -o 3 masculine i - li feminine le -le Table 1. Affirmative and interrogative SCLs in Lamonat and Sovramontino

Poletto (1993, 2000) tries to capture these differences by arguing the existence of two distinct sets of SCLs. However, the gaps in the paradigm of assertive SCLs do not necessarily suggest that assertive SCLs are fundamentally distinct from interrogative SCLs. Renzi and Vanelli (1983) note that, across NIDs, if a grammatical person exhibits an obligatory assertive SCL, the same grammatical person also exhibits an obligatory interrogative SCL: it is never the case that a NID has an obligatory SCL in one person of the assertive paradigm and does not have a corresponding interrogative SCL. This empirical generalisation suggests that the two sets are linked, and, very likely, make up a single set of SCLs with different surface realizations. The different morpho- phonological form between the interrogative and declarative set can be accounted for by acknowledging the different morpho-syntactic environment (proclisis vs. enclisis) in which interrogative and declarative SCLs surface. I will put forward the idea that, in synchrony, SCLI cannot be solely explained through T-to-C movement, but also through the ‘morphological rearrangement’ of the complex T° within the locus of SCLI, which is triggered by the presence of a piece of silent functional morphology that encodes information essential to the correct interpretation of the utterance (i.e. the interrogative force or optative/desiderative mood).

87 The question, however, remains of why those grammatical persons lacking SCLs in declarative clauses very often exhibit overt SCLs in root interrogatives. Cardinaletti and Repetti (2010) try to address this issue and postulate that those grammatical persons of the paradigm that do not exhibit a SCL in assertive clauses are equipped with an uninterpretable feature intrinsically associated with T° that is valued by pro. This feature gives null spell-out to the SCL: the SCL surfaces in questions as the inflected verb is moved out of T°. I will not enter into details with respect to their analysis; nonetheless, in terms of economy, positing a feature that assigns null-spell-out to the SCL is not as uneconomical as postulating two distinct sets of SCLs. The crucial point, however, is that, in root interrogatives, SCLs tend to be always overtly realised. SCLI can hence be considered a hallmark of questions across NEIDs: I will argue that SCLI itself licenses the interrogative force of root interrogatives. In this respect, the well- formedness of questions alone is enough to account for why, across NEIDs, the interrogative set of SCLs is heavier than the assertive one.

Third, the analysis put forward in the present chapter is more cost-effective in terms of the C-domain architecture that is necessary to account for the data. I will adopt the split-CP model devised by Rizzi (1997). In their analysis of the alternation between apparently in-situ and canonical wh-elements in Bellunese, Poletto and Pollock (2004) propose the following model of the left periphery:

(8) Operator2P

ForceP

GroundP

TopicP

Operator1P

TP

…

Poletto and Pollock’s (2004) model in (8) depicts the left periphery of root interrogatives across Romance. Even though, to some extent, the model captures the Lamonat and Sovramontino data, the question arises as to whether a more economical account is possible for wh-in-situ in Bellunese. The answer lies in the analysis put

88 forward in this chapter, which, I believe, is a more economical alternative to Poletto and Pollock’s (2004) analysis. Poletto and Pollock (2004) argue that in Bellunese wh- elements appear in Operator2P when they exhibit canonical behaviour, whereas they occupy Operator1P when they are apparently in-situ. In their model, the finite verb is in SpecForceP, the interrogative form of the encliticised SCL in GroundP and, finally, the past participle in TopP. As previously mentioned, their analysis does not involve T-to-C movement, but the finite verb moves to SpecForceP through remnant movement of the TP layer (along with the past participle) after the wh-element is attracted to Operator1P in the C-domain. Their account crucially involves a null restrictor in SpecOp2P that licences the apparently in-situ wh-element in Op1P. They argue that the null restrictor is base-generated in the TP layer with the wh-word, with which it forms a complex DP. In root interrogatives, the complex wh-phrase (the complex DP) is moved to Op1P as a unit and as soon as Op2P is merged the null restrictor is moved from Op1P to Op2P leaving the wh-word in Op1P in apparently in-situ position. If the null restrictor is not licenced, then the question cannot have the wh-word appearing in Op1P: the wh-word itself must move to SpecOp2, yielding a canonical wh-question. Let us exemplify Poletto and Pollock’s (2004) analysis by considering the Bellunese apparent wh-in-situ question in (9a) and its tree representation is (9b):

(9) a. Se-tu ndat andè? (Bellunese) be.2SG-2SG.SCL go.PTCP where ‘Where did you go?’ b. Operator2P

N-rest ForceP

se GroundP

tu TopicP

ndat Operator1P

N-rest andè TP

te se ndat andè N-rest

89 Poletto and Pollock (2004) argue that some Bellunese wh-words have the property of licensing a null restrictor, namely those corresponding to English who, what, where, when and how. The null restrictor is defined as a non-lexical DP that forms a complex wh-phrase with the wh-word. In (9), the null restrictor is base-generated in the T-domain with the wh-element ande as follows: [TP tu se ndat [PRN ande n-rest]]. The whole wh- phrase is then moved to SpecOp1P. Once Op2P is merged in the derivation, it attracts the null restrictor, leaving ande in Op1P in apparently in-situ position. Poletto and Pollock (2004) claim that, across Romance, wh-questions obey this syntactic mechanism. In their account, the key difference between Bellunese and the other Romance languages is that the latter do not exhibit wh-words that can license a null restrictor: hence, as soon as Op2P is merged, the wh-element lacking a null restrictor must move to SpecOp2P to prevent the derivation from crashing. For instance, let us consider the Friulian counterpart of (9) in (10a) and its tree representation in (10b). Note that Friulian does not exhibit wh-in-situ.

(10) a. Dulà se-tu lât? (Friulian) where be.2SG-2SG.SCL go.PTCP ‘Where did you go?’ b. Operator2P

Dulà ForceP

se GroundP

tu TopicP

lât Operator1P

dulà TP

tu ses lât dulà

Under Poletto and Pollock’s (2004) account, the Friulian wh-word dulà cannot license a null restrictor. Once Op2P is merged, dulà must move from Op1P to Op2P to prevent the derivation from crashing. Across NEIDs, apparent wh-in-situ is only present in Bellunese, because Bellunese wh-elements are able to license a null restrictor. The null restrictor hypothesis is based on the empirical observation that, in some NIDs exhibiting wh-in-situ, the apparently in-situ wh-element can be doubled by a wh-clitic in canonical

90 position (Poletto & Pollock 2004, 2009; Manzini & Savoia 2011). Poletto and Pollock (2004, 2009) thus claim that the null restrictor can have a lexicalised form consisting of a wh-clitic. Let us briefly have a look at some data from Lamonat and Sovramontino. To some extent, Poletto and Pollock’s (2004, 2009) account captures the alternation between (11) and (12), but does not explain the unacceptability of (13):

(11) Ond-e-lo ndat Toni? (Sovramontino) where-be.3SG-3SG.M.SCL go.PTCP Toni ‘Where did Toni go?’

(12) E-lo ndat aonde Toni? Be.3SG-3SG.M.SCL go.PTCP where Toni ‘Where did Toni go?’

(13) */?Ond-e-lo ndat aonde Toni? where-be.3SG-3SG.M.SCL go.PTCP where Toni ‘Where did Toni go?’

Under their analysis, (13) should be grammatical, as the wh-clitic ond should act as null restrictor and license aonde in apparently in-situ position. Speakers, however, do not accept the same wh-element simultaneously in canonical and apparently in-situ wh- position. The only exception is what-doubling, similarly to Italian che cosa, which will be discussed later in the chapter. The data from Lamonat and Sovramontino hence challenge the analysis by Poletto and Pollock (2004, 2009). Also, wh-questions like (11) are quite problematic under their account: if ond is a wh-clitic acting as null restrictor and is licensed by the wh-element aonde ‘where’, how come the licensor is not present in the clause? If aonde is absent in apparently in-situ position, what licenses the acting null restrictor ond? Along the same lines, Friulian also has wh-clitic elements that alternate with tonic wh-elements, as shown in (14) and (15) respectively:

(14) La-is-al lât Toni? (Friulian) where-be.3SG-3SG.M.SCL go.PTCP Toni ‘Where did Toni go?’

(15) a. Dulà is-al lât Toni? where be.3SG-3SG.M.SCL go.PTCP Toni ‘Where did Toni go?’

b. *Is-al lât dulà Toni? be.3SG-3SG.M.SCL where go.PTCP Toni ‘Where did Toni go?’

91 The wh-clitic element in (14), namely la, has the same morpho-syntactic status as ond in Sovramontino; nevertheless, apparently wh-in-situ is not possible in Friulian, as shown in (15b). Cross-dialectally the different distribution between clitic and tonic wh- elements further weakens the null restrictor hypothesis. Another important point is that, in Lamonat and Sovramontino, prepositional wh-elements like ‘with what’, ‘to whom’ etc. can also appear in apparently in-situ position: the licensing of the null restrictor would thus be not only limited to a restricted set of genuine wh-elements, as Poletto and Pollock (2004, 2009) claim, but to all wh-phrases.

These empirical observations make Poletto and Pollock’s (2004, 2009) account undesirable for the analysis of Lamonat and Sovramontino: ultimately, it does not fully justify the alternation between apparently in-situ and canonical wh-words. In this chapter, I will propose a revision of Poletto and Pollock’s (2004, 2009) model, which will not resort to the existence of two distinct ad hoc operator positions for wh- elements, but will consider the discourse-pragmatic nature of wh-words. I will, nonetheless, support two important arguments put forward by Poletto and Pollock (2004) that lie at the foundation of the present work.

First, in Bellunese, SCLI involves ForceP in root interrogatives. I will reanalyse their claim saying that ForceP is the locus of SCLI in our two varieties of Bellunese, namely Lamonat and Sovramontino. It follows the apparently in-situ wh-position: the verb crosses the left peripheral wh-position, which in Munaro, Poletto and Pollock (2001) was described as a discourse-pragmatic specialised position encoding new information. Munaro, Poletto and Pollock’s (2001) analysis was, possibly, less formal, but, I believe, more in line with the present analysis of the phenomenon. Second, I adopt Poletto’s (1993) dynamic agreement. Dynamic agreement is responsible for the well- formedness of questions across NEIDs and is based on Rizzi’s (1991, 1996) wh- criterion:

Rizzi (1996: 64) wh-criterion: a) a wh-operator must be in spec-head configuration with an X° [+wh], b) an X° [+wh] must be in spec-head configuration with a wh-operator.

Poletto (1993) proposes a process that is complementary to Rizzi’s wh-criterion, whereby the [WH] feature can be transmitted from the C° position to the corresponding

92 specifier, hence dynamic agreement. Across NEIDs, the syntactic nature of interrogative clauses is also morphologically marked by SCLI: SCLI in fact signals verb movement to C° (Benincà 1994, Munaro 1997). Poletto (1993) argues that through SCLI, C° licenses in SpecCP a silent operator that inherits the feature [WH] from the verbal head: the wh-criterion is then ultimately satisfied. In Poletto’s (1993) view, in Bellunese, SCLI alone satisfies the structural well-formedness of questions. This hypothesis supports the independence of wh-movement and SCLI/verb movement across NEIDs (see Manzini & Savoia 2011). In wh-questions, the wh-item does not need to land in the specifier position of the moved verb, as the structural requirements on interrogatives are satisfied by SCLI alone. Note that I will assume that, in questions, the specifier position of the moved verb is filled by a wh-operator and can only be lexicalised by an XP carrying a [WH] feature. In the present analysis, I will hence reject the hypothesis that Bellunese wh-elements licence a null restrictor (Poletto and Pollock 2004, 2009), but support the view that a silent interrogative operator is necessary for well-formedness of questions via dynamic agreement (Poletto 1993).

Manzini and Savoia (2011) argue against Poletto and Pollock’s (2004, 2009) account in support of an actual in-situ analysis of the phenomenon across NIDs. The Lombard data with which they challenge Poletto and Pollock’s (2004, 2009) analysis, however, yield ungrammaticality when tested in Lamonat and Sovramontino. This is especially true with respect to two points: first, in Lamonat and Sovramontino there is a rigid root vs. embedded asymmetry with respect to wh-in-situ. The phenomenon is only attested in root interrogatives and, conversely to Lombard, wh-in-situ is ungrammatical in embedded interrogatives. Second, unlike Lombard, wh-doubling yields ungrammaticality (except what-doubling) in Lamonat and Sovramontino (cf. 13). The different empirical evidence with respect to wh-in-situ between Bellunese and Lombard suggests that the syntactic mechanism at the basis of the phenomenon is very likely not the same in the two NIDs. Different accounts should be sought out for wh-in-situ across NIDs. In this respect, Manzini and Savoi’s (2011) analysis fails to account for those wh- elements that are not clitic, but are found in canonical wh-position in Lamonat and Sovramontino, as in (16):

(16) Che majon a-tu comprà? (Sovramontino) which jumper have.2SG-2SG.M.SCL buy.PTCP? ‘Which jumper did you buy?’

93 Under Manzini and Savoia’s (2011) account, wh-elements in canonical position are analysed as clitic scope markers as opposed to fully-fledged wh-words. In (16), the wh- phrase che majon is neither clitic nor a scope marker. If we were to adopt Manzini and Savoia’s analysis for Lamonat and Sovramontino, a further constraint would be needed to account for why tonic wh-items can either appear in-situ or in ex-situ position. On the other hand, Poletto and Pollock’s (2004, 2009) account would state that the wh-phrase che majon does not license a null restrictor, hence the wh-phrase itself is attracted to the higher wh-operator position (Op2P). Later in the chapter, I will argue that neither analysis explains the wh-distribution in (13), which must be accounted for in terms of the discourse-pragmatic function encoded by the wh-element. Undoubtedly, Manzini and Savoia’s (2011) account fully captures wh-in-situ in Lombard, but is incompatible with the Lamonat and Sovramontino data. Again, this suggests that the syntactic mechanism at the basis of wh-in-situ in Lamonat and Sovramontino is underlyingly different from that of Lombard. In this respect, Poletto and Pollock’s (2004, 2009) apparently in-situ account better describes the Lamonat and Sovramontino data.

Let us now start our investigation by exploring the different distribution between apparently in-situ and canonical wh-words in Lamonat and Sovramontino.

3.3. The Behaviour of Wh-Words

Lamonat and Sovramontino exhibit both apparently in-situ and canonical wh- elements. Let us consider the following wh-questions in (17) and (18):

(17) A-lo magnà che l can? (Lamonat) have.3SG-3SG.M.SCL eat.PTCP what the dog ‘What did the dog eat?’

(18) Ond-e-la ndà to sor? where-be.3SG-3SG.F.SCL go.PTCP your sister ‘Where did your sister go?’

The question in (17) is an instance of apparent wh-in-situ, whereas (18) resembles a canonical wh-question. For now, let us leave (18) aside and focus on (17).

94 3.3.1 Apparently In-Situ Wh-Words

In (17) the interrogative pronoun follows the main verb and precedes the subject of the clause (el can). In Lamonat and Sovramontino, all instances of apparent wh-in- situ strictly follow the constituent order: main verb, wh-word and subject. The fact that the subject, if expressed and topical in nature16, always follows the wh-element suggests that the latter does not appear in TP internal argument position, but undergoes wh- movement to a higher position in the left periphery. The wh-element hence crosses the subject l can that allegedly occupies SpecTP17. The apparent in-situ position of che in (17) becomes more transparent if compared to the equivalent echo question in (19) below:

(19) El can l a magnà chee? (Lamonat & Sovramontino) The dog 3SG.M.SCL have.3SG eat.PTCP what ‘The dog has eaten what?’

The echo question in (19) is translatable into English as ‘The dog ate what?’. In Sovramontino and Lamonat, echo questions are marked prosodically by an ‘echo intonation’ on the interrogative pronoun that is uttered duplicating its final vowel. In (19), the wh-word chee truly occurs in-situ in its TP internal argument position. The subject el kan precedes the main verb exhibiting a structure that is comparable to that of English echo questions. A crucial difference between (17) and (19) is that the echo question does not exhibit SCLI, whereas, in the apparent wh-in-situ question in (11), SCLI is pivotal for the well-formedness of the question. Thus, the echo question in (19) is not characterised by an interrogative structure, but has the same structure of a declarative clause. In (17), SCLI occurs higher than the left peripheral position occupied by the wh-word, che, and the subject, el can, remains in its canonical subject position. The comparison between (17) and (19) strongly suggests that the wh-element in (17) is only apparently in-situ: it has moved out of the TP-layer and crossed the subject position, landing in a left peripheral position that is lower than that involved in SCLI.

16 Topical in nature and, crucially, encoding background information (a G-topic in the sense of Bianchi & Frascarelli 2010). Discourse-pragmatically, the subject in (17) is not salient. I will further discuss topical background information in chapter 4. 17 See Villa-García (2015) for evidence that in Spanish subjects can occupy SpecTP. In Lamonat and Sovramontino, the relatively rigid constituent order seems to suggest that subjects occupy a dedicated syntactic position (namely SpecTP) when not discourse-pragmatically salient. 95 Wh-movement in Bellunese is detectable through sensitivity to islands, long distance wh-extraction and ‘marginalisation’ of the non-focal background material in the question. Munaro (1999) notes that Bellunese wh-in-situ questions are sensitive to strong and weak island constraints (Munaro, Poletto & Pollock 2001; Poletto and Pollock 2004). In true wh-in-situ languages, wh-islands do not hold (see Huang 1982). The data from Lamonat and Sovramontino conform to Munaro’s (1999) observation. Consider (20), (21) and (22):

(20) *Credee-lo che l fiol de chi no-l gnesse? (Sovramontino) Believe.3SG-3SG.M.SCL that the son of who NEG-3SG.SCL come.3SG ‘*Of whom did he believe that the son would not come?’

(21) *On-e da telefonar prima de ndar aonde? (Lamonat) have.1PL-1PL.SCL to phone.INF before of go.INF where ‘*Where should we call before we go?

(22) *A-tu vedù la femena che la vio aonde? have.2SG-2SG.SCL see.PTCP the woman that 3SG.F.SCL live.3SG where ‘*Where did you see the woman that lives?’

The ungrammaticality of (20), (21) and (22) is respectively due to a complex-NP island, an adjunct island and a relative island respectively 18 (see Munaro 1999 for a more detailed account). Islands constrain the unbounded nature of wh-movement: a wh- element can potentially be an unlimited number of clauses away from its in-situ null copy, with which it establishes a long distant dependency. Let us consider the wh- question in (23):

(23) A-la dit che Flora che Luciano l pensea Have.3SG-3SG.F.SCL say.PTCP what Flora that Luciano 3SG.M.SCL think.3SG.IMP

che l so visin l ese fat? that the his neighbour 3SG.M.SCL have.3SG.SUBJ do.PTCP ‘What did Flora say that Luciano thought that his neighbour did?’

(23) shows that the wh-element che is not in-situ, but has moved to the left periphery of the matrix clause. The apparent in-situ position is hence the result of the verb of the matrix clause crossing the wh-element position.

18 Note that the grammatical counterparts of these sentences would require a complex cleft structure, whereby the wh-item is clefted and the rest of the sentence re-arranged.

96

Poletto and Pollock (2004, 2015) provide another argument in support of the ex- situ nature of wh-questions in Bellunese. They observe that the XPs following the wh- element are de-accented and ‘marginalised’. Consider (24):

(24) G-a-lo dat che Mario al can? (Sovramontino) DAT-CL-have.3SG-3SG.F.SCL give.PTCP what Mario to-the dog ‘What did Mario give to the dog

The subject Mario and the dative complement al can are marginalised: they are de- accented and marked by a falling intonation. In Lamonat and Sovramontino, this prosodic behaviour is typical of topical background information or G(iven)-topics (in the sense of Bianchi & Frascarelli 2010). In chapter 4, I will further discuss G-topics and their properties in the NEIDs under investigation. Poletto and Pollock (2004: 254, 2015: 139) explicitly make reference to the dative complement of a sentence similar to (25):

(25) G-a-lo dat che al can? (Sovramontino) DAT-CL-have.3SG-3SG.F.SCL give.PTCP what to-the dog ‘What did he give to the dog

They claim that the dative complement (al can) is marginalised and separated from the wh-element by a ‘comma intonation’. In chapter 4, I will show that, in Lamonat and Sovramontino, this is true of all XPs following a non-contrastive focal element (not only of dative complements). Given the pitch contour of sentences like (25) (see chapter 4), I argue that it is more correct to talk about ‘falling intonation’, rather than ‘comma intonation’, as a pause is generally absent. Poletto and Pollock’s (2004, 2015) claim that, if the wh-element were in internal TP position, these facts would be hard to account for. In chapter 4, I will argue that the change in intonation marks the boundary between the C-domain and the T-domain.

Now that we have established that the wh-element in (17) undergoes wh- movement, we need to understand which position it targets in the C-domain. In order to do so, I will adopt Rizzi’s (1997) seminal split-CP model of the left periphery. Rizzi (1997) claims that the CP encodes at least two types of information relating to both the ‘outside’ and the ‘inside’ of the clause (hence, the need of splitting the CP into different projections). In his model, ForceP encodes information that looks at the higher structure

97 of the clause and specifies its clausal type (i.e. interrogative, declarative or exclamative force); on the other hand, FinitenessP looks at the inside of the clause, namely at the content of the TP, and is responsible for marking finiteness. Rizzi’s (1997) split-CP model does not only encode force and finiteness, but also discourse-pragmatic information that relates to information structure. He claims that the topic-comment and the focus-presupposition articulations, namely TopicP and FocusP, are sandwiched between ForceP and FinitenessP. It is important to note that I will not adopt Rizzi’s model in its original form, but as it is represented in (26) below:

(26) ForceP

TopicP

FocusP

FinitenessP

…

I abide by Benincà and Poletto’s (2004) claim against the existence of a lower TopicP sandwiched between FocusP and FinitenessP. No lower topic is in fact identifiable in Lamonat and Sovramontino. In chapter 4, I will argue that non-contrastive topics are altogether banned from the left periphery of Lamonat and Sovramontino. Also, in Lamonat and Sovramontino, the available left peripheral topical position is not freely recursive. A more detailed analysis will follow in chapter 4.

Let us go back to the interrogative pronoun che in (17). I repeat (17) in (27) below:

(27) A-lo magnà che l can? (Lamonat) have.3SG-3SG.M.SCL eat.PTCP what the dog ‘What did the dog eat?’

We cannot establish the landing site of che within Rizzi’s (1997) split-CP without first considering the behaviour of the surrounding constituents. Following Poletto’s (1993, 2000) claim that in Bellunese the wh-criterion is fully satisfied by SCLI alone, there is no need for the wh-element to land in the specifier position of the moved verb cluster (a-lo magnà) to fulfil the structural needs of the interrogative sentence. Poletto and Pollock (2004) argue that SCLI involves ForceP. In light of their account, the wh-

98 element must then land in a position lower than ForceP. We can also assume that the de- accented background topical subject el can occupies the canonical subject position, namely SpecTP. In light of this, the landing position of the wh-element must be higher than the TP layer, but lower than ForceP. The next lower position to ForceP is TopP. This position is recursive in nature (Rizzi 1997) and can give rise to a bundle of projections that characterises the Topic Field (Benincà & Poletto 2004). Given the focal nature of wh-elements, it is unlikely that che in (27) lands in TopicP. Bona fide topical elements (i.e. non-contrastive) are optional as they are the presupposed-known portion of the sentence (Lambrecht 1994). By contrast, wh-words always refer to some novel piece of information that the speaker wants to elicit from the hearer. The landing site of apparent in-situ wh-words cannot be located in the Topic Field (in the sense of Benincà & Poletto 2004). The focal properties of wh-elements strongly suggest that the most obvious landing site for che in (27) is FocusP: a position that is above the TP layer, lower than ForceP and not topical in nature. In chapter 4, I will further discuss focus in Lamonat and Sovramontino. For now, it is important to note that focal elements and wh- items share the following properties: uniqueness, lack of a resumptive clitic, prosodic sentential stress and are incompatible with one another (Rizzi 1997; Cruschina 2012). As for the last point, in-situ wh-items are incompatible with both contrastive/corrective and informational focal elements as shown in (28) and (29) respectively:

(28)*AL GAT g-a-lo dat CHE, no al can? (Lam.) to-the cat DAT.CL-have.3SG-3SG.M.SCL give.PTCP what NEG to-the dog ‘What did he give to the cat and not to the dog?’

(29) a.*G-a-lo dat CHE AL GAT (Mario)? DAT.CL-have.3SG-3SG.M.SCL give.PTCP what to-the cat Mario ‘What did Mario give to the cat?’

b. *G-a-lo dat AL GAT CHE (Mario)? DAT.CL-have.3SG-3SG.M.SCL give.PTCP to-the cat what Mario ‘What did Mario give to the cat?’

Contrastive focus is represented in BOLD CAPITALS, whereas informational focus is represented in CAPITALS. In (28), the wh-element che cannot coexist with the contrastive/corrective focal element al gat. In Lamonat and Sovramontino, there can be only one focal element per sentence (see Rizzi 2018 on uniqueness of focus). In chapter

99 4, I will argue that, syntactically, this translates into the inability of an abstract prosodic operator F encoding focal prominence to be recursive. In other words, focal prominence can only be assigned to one constituent per sentence. Uniqueness of focus justifies the ungrammaticality of (28), even though, from a merely linear point of view, the focal contrastive element and the wh-item seem to occupy different syntactic positions. In (29), on the other hand, the wh-element and the informational narrow focal XP al gat seem to compete for same syntactic position19. This is true regardless of (i) the order focal element--wh-word or viceversa, and (ii) the presence or absence of the de- accented topical subject Mario. For the sake of completeness, FinitenessP is also ruled out from the possible landing sites of che, as it is not a suitable hosting position for a focal element (see Rizzi 1997).

Having established the landing position of the apparently in-situ wh-word che in (27), let us briefly explain how the wh-element targets that position in the proposed formal model. Both XPs in informational narrow focus and wh-words can surface in SpecFocP. In addition to a [WH] feature, a wh-element also carries a [FOCUS] feature (see Bianchi 1999). The [FOCUS] feature is primarily responsible for the occurance of the wh-element in SpecFocusP. As for the [WH] feature, it does not need to be satisfied through spec-head agreement: the structural well-formedness of the question (the wh- criterion) is ensured by SCLI alone via dynamic agreement (Poletto 1993). Since ForceP encodes the interrogative force of the sentence, I assume that Force° is equipped with a [WH] feature, but the [WH] feature in Force° is satisfied through SCLI alone and does not need the wh-element to occupy SpecForceP. The wh-element is thus free to surface in SpecFocusP. In Lamonat and Sovramontino, the discourse-pragmatic nature of wh-elements seems to be more prominent than their syntactic status as question operators. The tree in (30) below shows the structure of (27) according to the proposed analysis in Lamonat and Sovramontino:

19 Note that semantic-pragmatic constraints may also be at play in determining the ungrammaticality of (28) and (29).

100 (30) ForceP

a-lo magnà FocusP [WH]

che Focus' [FOCUS] l a magnà TP

el can T'

l a magnà VP

Vº DP magnà che [FOCUS] [WH]

It is important to note that the inflected verb and its past participle move together to Force°. In the next chapter, more specifically in section 4.3.1.1, I will argue that this is the case due to the clitic nature of auxiliaries in Lamonat and Sovramontino, which require adjacency to the past participle for their realisation. T-to-C movement must occur via Foc° in order not to violate the Head Movement Constraint/HMC (Travis 1984). This observation will be important in the next chapter.

3.3.2 “Canonical” Wh-Words

Having proposed a formal model to capture the behaviour of seemingly in-situ wh-elements in Lamonat and Sovramontino, I now need to account for the behaviour of non-d-linked canonical wh-elements like the interrogative pronoun ond in (31) below:

(31) Ond-e-la ndà to sor? (Lamonat) where-be.3SG-3SG.F.SCL go.PTCP your sister ‘Where did your sister go?’

I will argue that, in Lamonat and Sovramontino, canonical wh-words exhibit a different morpho-syntactic behaviour with respect to their Italian counterparts: non-d-linked canonical wh-words do not undergo wh-movement, but attach to the inflected verb that occupies Force°. I will suggest that they are merged directly in Force°. In this regard,

101 canonical wh-elements are not canonical at all. In order to avoid confusion, for the time being, let us simply label this type of wh-element as preverbal wh-words.

First of all, preverbal and apparently in-situ wh-elements exhibit clear morpho- phonological differences, which must be accounted for. In Sovramontino and Lamonat, preverbal wh-elements are generally phonologically reduced and, unlike apparently in- situ wh-elements, cannot receive focal stress: they are ultimately atonic and rely on the inflected verb for their realisation. Let us consider examples (32), (33), (34) and (35) below. They show the distribution of the phonologically reduced preverbal wh-word onde ‘where’ and its apparently in-situ wh-counterpart aonde ‘where’:

(32) Ond-e-lo ndà Mario? (Lamonat & Sovramontino) where-be.3SG-3SG.M.SCL go.PTCP Mario ‘Where did Mario go?’

(33) E-lo ndà aonde Mario? be.3SG-3SG.M.SCL go.PTCP where Mario ‘Where did Mario go?’

(34) Onde/*Aonde va-lo Mario? where go.3SG-3SG.M.SCL Mario ‘Where is Mario going?’

(35) Va-lo aonde/*onde Mario? go.3SG-3SG.M.SCL where Mario ‘Where is Mario going?’

Examples (32) to (35) show that onde and aonde are not interchangeable. Onde can only appear pre-verbally, while aonde can only appear in apparently in-situ position. The allegedly ungrammatical example in (36) below further supports this rigid distribution:

(36) ?*Onde son-e aonde? (Lamonat) where be.1PL-1PL.SCL where ‘Where are we?’

In (36) both wh-positions are lexicalised: onde preverbally and aonde postverbally. This particular sentence was uttered by an old lady during one of our interviews in the municipality of Lamon. She had lost track of the questionnaire item that we were looking at and uttered the sentence in distress. It is hard to tell whether the double

102 lexicalisation of ‘where’ wants to emphasise ‘where’ or is a processing mistake. Given the context, both options are possible. Nevertheless, the simultaneous lexicalisation of the preverbal and postverbal wh-position as a discourse-pragmatic strategy is not common in Lamonat: no other instances were attested. Note, however, that the simultaneous lexicalisation of both wh-positions is possible and judged as grammatical with the wh-word ‘what’ in Lamonat (see section 3.3.4.). As for ‘where’, onde attaches to the inflected verb and, unlike aonde, does not receive focal stress. Onde is hence the clitic form of its focal counterpart aonde. Two additional pieces of evidence can be put forward in support of this claim. First, onde cannot be used in isolation, while aonde can (e.g. ‘where?’ to ask for a precise location). Second, nothing can intervene between the inflected verb and the preverbal wh-element.

Across NIDs, clitic wh-elements are robustly attested (Manzini & Savoia 2005, 2011; Poletto & Pollock 2009, 2015). Poletto and Pollock (2009, 2015) put forward a three-way distinction between tonic, clitic and weak (a là Cardinaletti & Starke 1999) wh-elements. In the present research, I will only distinguish between clitic and tonic wh- words: a two-way distinction is, in fact, sufficient to account for the morpho-syntactic behaviour of wh-elements in Lamonat and Sovramontino (see Manzini & Savoia 2011 for criticism on weak wh-elements). Munaro and Poletto (2014) also note the clitic status of some wh-elements across NIDs, exhaustively discussing the wh-word ‘where’. In their work, they reconstruct the internal layering of wh-words starting from their morphological composition. Munaro and Poletto (2014) argue that, cross-dialectally, ‘where’ is made up of three sub-elements: i) a prepositional formative, which corresponds to either one of the following prepositions in ‘in ’, di ‘of ’, and da ‘from’, or a combination of them; ii) a vocalic formative o/u namely the wh-operator feature that developed from Indo-European wh- formative qw; and iii) a deictic locative element derived from various sources -nd, -v, or -la, expressing locative deixis. Following Munaro and Poletto’s (2014) analysis, in Table 2 below, I show the internal make-up of the tonic wh-word aonde in Lamonat and Sovramontino.

103 a o nd e Aonde locative preposition in o/u formative epenthetic V formative nd

Table 2. The internal makeup of the tonic wh-word aonde in Lamonat and Sovramontino according to Munaro and Poletto (2014).

As for the clitic wh-element onde, it lacks the prepositional layer, resulting in a reduced morpho-phonological form. Also, the epenthetic final vowel -e is absent if the inflected verb to which it attaches starts with a vowel. The clitic wh-form of ‘where’ is also present in Friulian. The same as onde in Lamonat and Sovramontino, Friulian clitic ‘where’ lacks the prepositional layer. Friulian does not exhibit apparent wh-in-situ, wh- elements, in fact, always appear preverbally. The wh-question in (37) is formed with the Friulian clitic form of ‘where’, namely la, whereas (38) shows its tonic counterpart dulà:

(37) La-is-al Jacum? (Friulian) where-be.3SG-3SG.M.SCL Jacum ‘Where is James?’

(38) Dulà is-al Jacum? where be.3SG-3SG.M.SCL Jacum ‘Where is James?’

The Friulian clitic wh-form la exhibits the same behaviour as onde in Lamonat and Sovramontino: it cannot appear in isolation and, phonologically, relies on the inflected verb to which it attaches. It can only appear as such preverbally, in root wh-questions, and, crucially, in embedded wh-questions where it must be followed by an overt neutral complementiser che. The two examples in (39) and (40) below show the identical distribution of the wh-clitics la and onde in two embedded wh-questions equivalent in meaning in Friulian and Lamonat:

(39) No sai di-ti la che e je lade (Frl.) NEG know.1SG say.INF-1SG.DAT where COMP 3SG.F.SCL be.3SG go.PTCP? ‘I am not able to tell you where she has gone’

(40) No se dir-te onde che la è ndaa (Lam.) NEG know.1SG say.INF-1SG.DAT where COMP 3SG.F.SCL be.3SG go.PTCP? ‘I am not able to tell you where she has gone’

104 Note that, in both Lamonat (40) and Friulian (39), the clitic wh-elements onde and la cannot be replaced by aonde and dulà: the clitic wh-forms must appear and be followed by che, tonic wh-forms are banned from embedded questions. The discussion of embedded wh-questions goes beyond the scope of the present chapter and will not be pursued. Nevertheless, the striking similarities with respect to the morpho-syntactic behaviour of Lamonat onde and Friulian la strongly suggest that they are the same element. This begs the question: do onde and la also exhibit an identical morpho- syntactic behaviour in root interrogatives? In section 3.5, I will argue that it is indeed the case. This is important as the analysis of wh-elements put forward in this chapter can, ideally, be adopted for other NIDs.

Besides the onde-aonde distinction, Sovramontino and Lamonat exhibit other wh- items with two wh-forms, namely tonic and clitic, as shown in Table 3 below:

Tonic Clitic postverbal preverbal Where aonde ond(e) What che ch(e) / s(a) When quand quan How come com(e) How much quant quan(t) Table 3. Tonic and clitic wh-elements in Lamonat and Sovramontino

Only clitic wh-forms can appear preverbally. It is important to note, however, that ‘how’, ‘how much’ and, in Sovramontino, ‘what’ have the same form in the tonic and clitic paradigm. They are nonetheless unstressed and tend to be moprho-phonologically reduced. For example, the clitic form of ‘how much’ loses its final –t when the following inflected verb starts with a consonant, but retains it if the verb starts with a vowel, as shown in (41) and (42) respectively:

(41) Quan-coste-lo? (Lamonat & Sovramontino) how.much-cost.3SG-3SG.M.SCL ‘How much does it cost?’

(42) Quant-e-lo? how.much-be.3SG-3SG.M.SCL ‘How much is it?’

105 Similarly, the clitic form of ‘how’, come, loses its final vowel when the inflected verb starts with a vowel. Even though the two sets are not sharply morphologically differentiated, the key difference between the tonic and the clitic set is that the latter is atonic and cannot be assigned focal prominence (see chapter 4). Examples in (43) and (44) show the different forms of the Lamonat wh-word ‘what’. Out of all wh-elements, the tonic and the clitic form of ‘what’ are those that are differentiated the most:

(43) Sa-a-lo fat Toni? (Lamonat) what-have.3SG-3SG.M.SCL do.PTCP Toni ‘What did Toni do?’

(44) A-lo fat che Toni? have.3SG-3SG.M.SCL do.PTCP what Toni ‘What did Toni do?’

The different distribution between tonic and clitic wh-elements begs the question: does the clitic nature of wh-elements also reflect a difference in syntactic behaviour with respect to tonic wh-forms? Simply by making reference to the different syntactic distribution between tonic and clitic wh-elements, the answer is certainly ‘yes’. I hereby claim that, across the NEIDs under investigation, the clitic nature of wh-elements is primarily responsible for their preverbal position.

Clitic behaviour has always being problematic in syntax. Chomsky (1995) points out that clitics are ambiguous categories that share XP and X° properties. In the case of wh-clitics, I will adopt Kayne’s (1991) view that clitics tend to attach to the heads of functional categories. In this regard, the clitic form of a wh-word always attaches to the inflected verb in the locus of SCLI. The wh-clitic attaches to the inflected verb sitting in Force° as shown in (45a) and (45b) below:

(45) a. Quan-si-tu ndà a fong? (Lamonat) when-be.2SG-2SG.SCL go.PTCP to mushroom.PL ‘When did you go picking up mushrooms?’

106 b. ForceP

[WH] Force'

Quand-si-tu ndà FocusP

te se ndà TP

T'

te se ndà VP

V PP ndà a fong

The well-formedness of the question is assured by SCLI, which licenses a silent wh- operator in SpecForceP. It is reasonable to assume that clitic wh-elements are also equipped with a [WH] feature, hence they could lexicalise the wh-operator in SpecForceP. Nonetheless, I will show in chapter 5 that SpecForceP is a dedicated discourse-pragmatic position that can only be occupied by inherently contrastive elements. If SpecForceP is not available to the clitic wh-element, it must occupy Force°. I argue that it is attached as a prefix to the inflected verb with which it forms a single phonological unit. In this respect, this approach is more in line with Kayne’s (1991) view that clitics tend to attach to heads. The other possibility for the wh-element to occur higher than the inflected verb would be in an ad hoc wh-position on the left of ForceP. Arguing, however, in favour of the existence of such position, not only would it be unnecessary, given the clitic nature of this type of wh-elements, but also it would contradict the account put forward for tonic wh-elements.

3.3.3 On the Nature of Wh-Clitics

In this section, I will discuss the nature of wh-clitics. The aim of this section is to shed more light on their morpho-syntactic status. In root interrogatives, I claim that wh- clitics appear as prefixes that attach to the inflected verb in Force°. In embedded interrogatives, they precede the complementizer che, on which they depend in terms of prosodic realisation. This is respectively shown in (46) and (47):

107 (46) Quan-va-li a catar so nona? (Lamonat & Sovramontino) when-go.3PL-3PL.M.SCL to find.INF their grandmother ‘When will they go visit their grandmother?’

(47) No i me a dit quan-che i va NEG 3PL.M.SCL to.me have.3PL say.PTCP when-COMP 3PL.M.SCL go.3PL

a catar so nona to find.INF their grandmother ‘They did not tell me when they will visit their grandmother’

The embedded interrogative in (47) suggests that: (i) the wh-clitic appears in ForceP where che is externally merged, and (ii) the wh-clitic does not undergo wh-movement, but is merged with the complementiser che. As for the second point, the assumption is that, in order to undergo movement, clitics need a host to attach to and ‘ride’ with20. In (46), the host is the inflected verb that has undergone T-to-C movement, however, in (47), there is no host that has moved from the lower portion of the clause: I then assume that the clitic has been externally merged with the complementiser che. If we accept that, in (46) and (47), the wh-clitic quan is the same morpho-syntactic element, it follows that wh-clitics are base-generated in Force°.

Following Baker’s (1970) and Bresnan’s (1972) arguments on clausal typing and the Q-morpheme as well as the subsequent literature on the topic (Chomsky & Lasnik 1977; Huang 1982; Cheng 1997; Bocci 2013 a. o.), I claim that wh-clitics are the lexicalisation of the null question-particle Q (Q-morpheme). Traditionally, Q is contained under the head responsible for the well-formedness of questions in the C- domain. In wh- and yes/no questions, Q attracts the finite verb, which moves from T-to- C to satisfy the structural requirements on interrogatives. Chomsky (1995) argues that Q is affixal in nature and attaches to the overt head that it attracts. Making reference to (46) and (47), Q needs a compatible host to attach to: in (46), in root interrogatives, Q attracts T°, whereas, in (47), Q is satisfied by the external merge of che, hence no T-to- C movement takes place in embedded interrogatives21. Across NEIDs, I propose that Q

20 See Bošković (2001) on the nature of clitics with respect to the syntax-phonology interface. 21 Consider the sentence: No se che majon che l a comprà ‘I don’t know which jumper he bought’. In this case, ‘which jumper’, che majon, is followed by an obligatory complementiser, che, that introduces the embedded interrogative. Che majon is clearly not clitic, and, given its d- linked interpretation, I assume that it can fill the specifier position of the embedded ForceP. Che majon is hence not the lexicalisation of Q: only wh-clitics are. Nevertheless, in Force°, the silent morpheme Q still attaches to the complementiser che because of the interrogative nature of the 108 is merged in the locus of SCLI where the wh-criterion is satisfied: in Lamonat and Sovramontino, Force°. The finite verb moves to the locus of SCLI and Q attaches to it as a phonologically null affix. Lamonat, Sovramontino and Friulian can overtly lexicalise Q with a wh-clitic. The claim that Q can be overtly expressed across Romance through ‘special complementisers’ has already been put forward by several scholars (see Jones 1993; Prieto & Rigau 2007; Cruschina 2008). The novelty of the present hypothesis is that clitic wh-elements themselves can lexicalise Q.

A piece of evidence in support of our claim comes from SCLI itself. I argue that T-to-C movement on its own is not enough to account for SCLI (see Pescarini & Calabrese 2014): what gives rise to SCLI is the presence of the Q affix in the locus of

SCLI that causes the morphological rearrangement Q-Verbfin-SCL. Before further exploring this hypothesis, let us discuss why T-to-C movement alone is insufficient to account for SCLI in Lamonat and Sovramontino. The traditional view that SCLI is the result of the finite verb crossing the base position of the SCL, which, then, in turn, prosodically relies on the moved finite verb through enclisis may be able to capture (48b), but is inadequate for the analysis of (49b)22:

(48) a. L magna polenta volentieri (Sovramontino) 3SG.M.SCL eat.3SG polenta gladly ‘He gladly eats polenta’

b. Magne-lo polenta volentieri? eat.3SG-3SG.M.SCL polenta gladly ‘Does he gladly eat polenta?’

(49) a. Mario l ge a dat an beso a so cusina Mario 3SG.M.SCL DAT.CL have.3SG give.PTCP a kiss to his cousin ‘Mario gave a kiss to his cousin’

b. G-a-lo dat che Mario a so cusina? DAT.CL-have.3SG-3SG.M.SCL give.PTCP what Mario to his cousin ‘What did Mario give to his cousin?’

embedded clause, as shown in the following labelled bracketing: [TP No se [SpecForceP che majon [Force' [Force Q-che] [TP l a comprà]]]]. 22 SCLI must be accounted for as I adopt the one-paradigm hypothesis regarding the different behaviour of SCLs in declarative and interrogative clauses (see chapter 2). Poletto (1993, 2000) supports the two-paradigm hypothesis and hence claims that interrogative SCLs are directly merged in the C-domain. On the contrary, I claim that SCLs move to the C-domain with the inflected verb and all the other satellite clitics since they form a single phonological unit (along the lines of Cardinaletti & Repetti 2010, Calabrese & Pescarini 2014). 109

In (49b), the past participle dat, the wh-element che, and the subject Mario appear in between the inflected verb and the base position of the SCL within the T-domain: if the SCL were in its base position, this would be virtually impossible to account for. The SCL hence moves to the C-domain with the inflected verb and with everything else that is contained under the complex T° (namely the past participle and the satellite clitics). If the complex T° including the SCL moves to the C-domain, it begs the question: how and why does SCLI take place? I claim that the answer lies in the Q-morpheme that causes the morphological rearrangement Q-Verbfin-SCL under the left peripheral functional head that we identified as the locus of SCLI. A formal model that captures SCLI has to take into account the complex structure underneath T°, which undergoes T°-to-Force° movement. Devising such a formal model goes beyond the purpose of the present research; nevertheless, I will briefly outline what, I assume, happens in the locus of SCLI. When Q attracts T° to Force°, the SCL proclitically relies on the finite verb. Once the moved complex T° reaches Force°, the following order of elements is in place:

Q SCL Vfin. Q is affixal in nature and must attach to the inflected verb: the SCL must give way to the Q affix by inverting with the finite verb. Evidence that Q attaches to the moved T° as a prefix (Q-T°) comes from compounds in the dialects: compounds are in fact left-headed23. A plausible conclusion is that SCLs are not compatible with Q. In case other clitics accompany the verb in the order SCL (CL)* Vfin, Q causes the SCL to invert and attaches to the cluster of satellite clitics preceding the finite verb, giving rise to the order Q-(CL)*-Vfin-SCL. The reason why Q can attach to other satellite clitics, but not to SCLs can be perhaps accounted for by making reference to Benincà’s (1994) linear order of preverbal clitics across NIDs: SCL > DAT > OCL > LOC/PRST.

Structurally speaking, the SCL is too far from Vfin for Q to attach to it, whereas the other satellite clitics are still within the orbit of Vfin.

As for wh-questions with tonic wh-elements, Q is also present, but not lexicalised. It is important to note that Q does not need to be lexicalised to trigger SCLI. In fact, SCLI occurs in all root interrogatives, namely wh- and yes/no questions. I assume that this is the case for all NEIDs exhibiting SCLI: in those varieties in which SCLI does not

23 Some compounds in Sovramontino: descantabauchi ‘scarecrow’, ferdafen lit. ‘tool-for-hay’, springafior ‘watering can’, stropacul ‘dog-rose’, strucapatate ‘masher’, picamenestri ‘ladle’, saltamartin ‘grasshopper’.

110 take place, the silent Q-morpheme is morphologically absent, but present at the prosodic level (see Reglero & Ticio 2013). SCLI does not only occur in root interrogatives, but also in a restricted set of root declaratives featuring an inflected verb conjugated in the subjunctive, as shown in (50) and (51):

(50) Clamassi-tu plui dispes! (Friulian) call.3SG.IMPF.SUB-2SG.SCL more often ‘I wish you called more often’

(51) Esse-lo parlà bonora! (Sovramontino) have.3SG.IMP.SUB-3SG.M.SCL speak.PTCP early ‘I wish he had spoken earlier’

In a similar fashion to SCLI in root interrogatives, I propose that in (50) and (51) a parallel syntactic mechanism is in place: SCLI is triggered by a piece of silent morphology, namely the affix S, which encodes subjunctive irrealis mood and needs adjacency to inflected verb (à la Bošković 2001). Affix S encodes optative/desiderative mood, whereas Q that encodes interrogative force, both require a suitable host. In synchrony, subject clitic inversion should hence be regarded as a morpho-syntactic configuration that surfaces in a restricted set of syntactic contexts that involves both verb movement and the expression of an extra piece of semantic-syntactic information (i.e. interrogative force or optative/desiderative mood). SCLI in root declaratives also surfaces in the protasis of hypothetical clauses that feature the imperfect subjunctive, as in (52):

(52) Mangjassi-e di mancut, no sares tant grase (Friulian) eat.3SG.IMP-3SG.F.SCL of less, NEG be.3SG.COND so fat ‘If she ate less, she wouldn’t be so fat’

Enclisis is impossible when (50), (51) and (52) are introduced by a complementiser: if the complementiser is overtly realised, then proclisis is obligatory, as in (53), (54) and (55):

(53) Che te ciamese pi despes! (Sovramontino) that 2SG.SCL call.3SG.IMP.SUB more often ‘I wish you called more often’

(54) Ch-al ves fevelât plui a dore! (Friulian) that-3SG.M.SCL have.3SG.IMP.SUB speak.PTCP more early ‘I wish he had said it earlier’

111

(55) Se la magnese de manco, no la saria tan grasa (Sovr.) if 3SG.F.SCL eat.3SG.IMP of less, NEG 3SG.F.SCL be.3SG.COND so fat ‘If she ate less, she wouldn’t be so fat’

SCLI and the complementiser are therefore mutually exclusive. The data in (50) to (55) suggest that, in the presence of the imperfect subjunctive, C° must be lexicalised: the lexicalisation of C° is dictated by the subjunctive irrealis mood. It takes place through either the direct merge of the complementiser che/se, (53) to (55), or through T-to-C movement, (50) to (52). I claim that, in (50) to (55), the subjunctive irrealis mood is borne by a phonologically null affix S that, in the same fashion as Q, requires a suitable host (see Bošković 2001). S, the same as Q, can be satisfied through either external merge of a complementiser or internal merge of the inflected verb, as shown in the labelled bracketing representations in (56a) and (56b) respectively:

(56) a. [CP subjOPT [C S-che [TP [T la ciamese]]]] b. [CP subjOPT [C S-ciamase-la [TP [T la ciamase]]]]

Similarly to the lexicalisation of Q, the lexicalisation of the subjunctive mood, and other types of irrealis mood is attested in the literature. The lexicalisation is generally linked to the C-domain (Demonte & Fernández-Soriano 2009; Kempchinsky 2009; Rizzi 1997): it may have the form of an operator-like element encoding irrealis (Haegeman 2012) or of a specialised complementiser (Ledgeway 2005; Paoli 2007; Villa-García 2012, 2015).

In Sovramontino and Lamonat, the role of a silent Q- or S-morpheme in SCLI can also be appreciated from the absence of SCLI in root negative-interrogatives and negative-declaratives featuring the subjunctive mood. Let us consider the following negative sentences in (57) and (58):

(57) Mario no l a magnà la polenta (Lam. & Sovr.) Mario NEG 3SG.M.SCL have.3SG eat.PTCP la polenta ‘Mario did not eat the polenta’

(58) No l a magnà Mario la polenta? NEG 3SG.M.SCL have.3SG eat.PTCP Mario la polenta ‘Didn’t Mario eat the polenta?’

Negation in the root declarative (57) is not prosodically marked and allegedly occupies

112 a syntactic position within the TP layer that is below the subject Mario. It hence does not have scope over the subject Mario. On the other hand, the negation in (58) is prosodically marked and has scope over the whole sentence. In this respect, the negation in (58) is an instance of sentential negation whose realisation is rooted in the C-domain above the TP-layer (see Haegeman 1995, Zanuttini 1997; Horn 2001). In order to distinguish the negation in (58) from the negation in (57), I will henceforth refer to the former as strong negation, which must be interpreted as a descriptive label that identifies the negation employed in a negative yes/no question. In (58), the constituent order is altered, yielding T-to-C movement. Given the negative-interrogative nature of the sentence in (58), the strong negation must check its [NEG] feature in the C-domain (Zanuttini 1997; Poletto and Zanuttini 2013). I assume that this is done in PolarityP (see Laka 1990), which is located lower than ForceP and higher than FinP in the left peripheral space. Once the negation has checked its feature, it moves to Force° with the verb cluster where it adjoins to Q. I claim that Q is satisfied by the adjacency with the negative element: Q+NEG SCL Vfin. SCLI is therefore not necessary to ensure that Q is attached to a suitable host, as the strong negation itself is a suitable host for Q. To some extent, the present analysis resembles Zanuttini’s (1997:44) account of the behaviour of negation in root negative-interrogatives in Paduan. In Paduan, the same as Lamonat and Sovramontino, SCLI is blocked by the intervening negation:

I will argue, instead, that the negative marker itself (or, perhaps, its features) moves to C° in negative yes/no questions …for concreteness, I will assume that the yes/no operator is in the specifier of CP and that no can check the features of C° either by adjoining to it in the syntax or by LF-raising of its features (Zanuttini 1997: 44).

The crucial difference with Zanuttini’s (1997) account is that she does not assume T-to- C movement in root negative-interrogatives in Paduan: she claims that only the negation moves to the left peripheral space and the negation alone satisfies the wh-criterion (Rizzi 1991). Lamonat and Sovramontino’s constituent order in negative-interrogatives and the data from negative-interrogatives in Friulian (see section 3.5) seem to suggest that, across NEIDs, T-to-C movement takes place in negative-interrogatives. In sum, in Lamonat and Sovramontino, the strong negation after checking its [NEG] features in PolP becomes part of the verbal cluster and is able to host Q, making SCLI unnecessary. In (58), the intervening negative element hence blocks SCLI, as shown in the arboreal representation (59):

113 (59) ForceP

[WH] Force'

Q-no l a magnà PolP

No l a magnà [NEG] TP

Mario T'

l a magnà VP

V DP magnà la polenta

The strong negation seems nonetheless to be in complementary distribution with both clitic and tonic wh-elements, as shown in (60) and (61) below:

(60) ?No l a fat che Mario? (Lamonat) NEG 3SG.M.SCL have.3SG do.PTCP what Mario? ‘What didn’t Mario do?’

(61) ?Onde no l è ndat Mario? (Sovramontino) where NEG 3SG.M.SCL be.3SG go.PTCP Mario ‘Where didn’t Mario go?’

The ungrammatical or, possibly, infelicitous status of (60) and (61) is not necessarily predicted by the intervening negation. It may have to do with the strong negation having focal-like properties that may not be assigned simultaneously to both the negative element and the wh-element (see Villa-García 2016). Nevertheless, the wh-element in (61) is not assigned focal prominence due to its clitic nature, hence another explanation must be found. I will not further investigate the ungrammaticality of (60) and (61). I want, however, to point out that, in order for these sentences to fully converge at spell- out, the wh-element must appear in a cleft-structure, as shown in (62) and (63):

(Lamonat & Sovramontino)

(62) a. E-lo che che no l a fat Mario?

be.3SG-SCLexpl what COMP NEG 3SG.M.SCL have.3SG do.PTCP Mario? ‘What didn’t Mario do?’

114 b. *E-lo sa che no l a fat Mario?

be.3SG-SCLexpl what COMP NEG 3SG.M.SCL have.3SG do.PTCP Mario? ‘What didn’t Mario do?’

(63) Ond-e-lo che no l è ndat Mario?

where-be.3SG-SCLexpl COMP NEG 3SG.M.SCL be.3SG go.PTCP Mario ‘Where didn’t Mario go?’

Note that clitic wh-elements cannot be clefted (cf 62b), but only tonic wh-elements can (cf. 62a). Wh-clitics appear in the locus of SCLI attached to the dummy construction be.3SG-SCLexpletive. In the next section, I will briefly outline the behaviour of cleft questions in Lamonat and Sovramontino. I will also try to tackle the difference in discourse-pragmatics between clitic and tonic wh-questions.

3.3.4 Cleft-Questions, Clitic-Wh-Questions and Tonic-Wh-Questions

Cleft-structures are frequent in Lamonat and Sovramontino. They generally host either an XP in narrow focus (Lambrecht 1994) or a tonic wh-word. In Lamonat and Sovramontino, two types of clefts can be identified: (i) syntactically motivated and (ii) pragmatically motivated. Syntactically motivated cleft-structures arise as a last resort strategy to focalise XPs whose movement to C-domain would otherwise be illicit. Pragmatically motivated cleft-structures, on the other hand, prompt a marked discourse- pragmatic reading on the clefted focal element that is exhaustive in nature (see Lambrecht 2001). In chapter 4, I will discuss cleft structures in more detail; for now, I will briefly discuss cleft wh-questions.

In the previous section, I showed that strong negation and wh-elements cannot co- occur. For a negative wh-question to converge at spell-out, the wh-element must be clefted, as shown in (64):

(64) E-lo che che Maria no la a fat? (Lam.) be.3SG-SCLexpl what COMP Maria NEG 3SG.F.SCL have.3SG do.PTCP? ‘What didn’t Maria do?’

(64) is an instance of syntactically motivated cleft-structure: the language can only accommodate both the negation and the wh-element in such a structure. In (64) e-lo does not agree with the subject of the question, namely Maria. The dummy construction e-lo always exhibits a masculine third person singular SCL (SCLexpletive) encliticised 115 onto the third person singular form of the verb ‘to be’. If we try to make it agree with the subject Maria, the result is an ungrammatical sentence, as shown in (65):

(65) *E-la che che Maria no la a fat? (Lam.) be.3SG-3SG.F.SCL what COMP Maria NEG 3SG.F.SCL have.3SG do.PTCP? ‘What didn’t Maria do?’

In cleft-questions, Q is satisfied by the lexicalisation of the dummy SCLI construction e-lo. Syntactically motivated clefts also arise when subjects bear narrow focus. In chapter 4, I will provide a more detailed discussion of the phenomenon and reach the conclusion that, in Lamonat and Sovramontino, subjects cannot be moved out of their canonical position. A cleft hence arises to allow the focalisation of an XP that carries the subject function. As for wh-clefts, they are obligatory when the subject of a question is interrogated with ‘what’ or ‘who’, as shown by (66), (67) and (68) below:

(66) E-lo chi che a beù tut al vin? (Lamonat) be.3SG-3SG.M.SCL who COMP have.3SG drink.PTCP all the wine? ‘Who drank all the wine?’

(67) Chi-e-lo che a beù tut al vin? who-be.3SG-3SG.M.SCL COMP have.3SG drink.PTCP all the wine? ‘Who drank all the wine?’

(68) E-lo stà che che a npathinà l mur? be.3SG-3SG.M.SCL be.PTCP what COMP have.3SG dirty.PTCP the wall? ‘What dirtied the wall?’

Munaro (1998) notes that, in Bellunese, when chi and che function as subjects, a cleft construction is generated: the wh-element follows the copular verb and is in turn followed by the complementiser che. As shown in (66) to (68), Lamonat and Sovramontino exhibit the same pattern with subject ‘who’ and ‘what’. ‘Who’, chi, can either appear preverbally as clitic wh-element (cf. 67) or in apparently in-situ position (cf. 66) as tonic wh-element. Crucially, it must appear in a cleft construction in order to be correctly interpreted as the subject of the question. If not clefted, ‘who’ is by default interpreted as the object of the sentence as shown in (69) and (70) below:

(69) #A-lo beù chi l vin? (Lamonat) have.3SG-3SG.M.SCL drink.PTCP who the wine? Lit.‘Who did the wine drink?’

116 (70) #Chi-a-lo beù al vin? who-have.3SG-3SG.M.SCL drink.PTCP the wine? Lit.‘Who did the wine drink?’

Sentences in (69) and (70) literally mean ‘who did the wine drink?’. Thus, the cleft construction arises because of the privileged structural position of the subject. If chi ‘who’ is the intransitive subject of a core unaccusative verb like ‘die’ or ‘arrive’ and, hence, structurally behaves like an object (see Belletti 1988), the wh-element chi is not clefted, as shown in (71) and (72):

(71) Rue-lo chi? (Lamonat)

arrive.3SG-SCLexpl who ‘Who is arriving?’

(72) Elo mort chi? (Sovramontino)

be.3SG-SCLexpl die.PTCP who ‘Who died?’

In Lamonat and Sovramontino, cleft-wh-questions are not always syntactically motivated: they can be used to trigger an exhaustive reading on the focal clefted element (see Lambrecht 2001), as shown in (73):

(Lamonat and Sovramontino) (73) E-lo aonde che Mario l a catà na femena?

be.3SG-SCLexpl where COMP Mario 3SG.M.SCL have.3SG find.PTCP a wife ‘Where did Mario find a wife?’

The wh-question in (73) is an instance of a pragmatically motivated cleft question: no syntactic constraint forces aonde to be clefted, but the clefted wh-element triggers a focal exhaustive reading on the wh-element itself. In Lamonat and Sovramontino, there are hence four possible different ways to ask the same question. Let us consider the sentences from (74) to (77):

(74) A-la catà quand Maria n om? (Lam. and Sovr.) have.3SG-3SG.F.SCL find.PTCP when Maria a husband ‘When did Maria find a husband?’

(75) Quan-a-la catà Maria n om? when-have.3SG-3SG.F.SCL find.PTCP Maria a husband ‘When did Maria find a husband?’

117

(76) E-lo quand che Maria la a catà n om?

be.3SG- SCLexpl when COMP Maria 3SG.F.SCL have.3SG find.PTCP a husband ‘When did Maria find a husband?’

(77) Quan-e-lo che Maria la a catà n om?

when-be.3SG-SCLexpl COMP Maria 3SG.F.SCL have.3SG find.PTCP a husband ‘When did Maria find a husband?’

All four questions can be translated as ‘when did Maria find a husband?’. The difference between clefted (76, 77) and non-clefted (74, 75) wh-question can be straightforwardly resolved in terms of discourse-pragmatics (exhaustive vs. non- exhaustive). The challenge is to account for the difference between wh-questions that make use of tonic wh-elements (74, 76) and wh-questions that make use of clitic wh- elements (75, 77) in discourse-pragmatic terms. In order to tackle this issue, let us acknowledge that there is also a fifth strategy that involves the simultaneous lexicalization of both wh-positions: the clitic and the tonic wh-element. This mechanism, however, is not available with all wh-words, but only to ‘what’, as shown in (80) below:

(78) G-a-li dat che a Simon? (Lamonat) DAT.CL-have.3PL-3PL.M.SCL give.PTCP what to Simone ‘What did they give to Simone?’

(79) Sa-g-a-li dat a Simon? what-DAT.CL-have.3PL-3PL.M.SCL give.PTCP to Simone ‘What did they give to Simone?’

(80) Sa-g-a-li dat che a Simon? what-DAT.CL-have.3PL-3PL.M.SCL give.PTCP what to Simone ‘What did they give to Simone?’

The question in (78) involves a tonic wh-element; (79) exemplifies the behaviour of its equivalent clitic wh-word, whereas in (80) both wh-positions are simultaneously lexicalised. The data in (78) to (80) beg two questions: (i) whether a difference in meaning arises in the use of either type of wh-element, and (ii) what is the discourse- function conveyed by the simultaneous lexicalization of clitic and tonic wh-element.

Let us start with the discussion of (80). Poletto and Pollock (2004) argue that the simultaneous lexicalisation of both wh-positions maximises the interrogative force of

118 the question. They refer to the double lexicalization of ‘what’ with the label emphatic what. Emphatic what is a pragmatic strategy that is primarily used when the speaker wants to convey surprise, incredulity, anger or disappointment while uttering a question. In English, it would translate into a prosodically marked ‘what’. Poletto and Pollock (2004) argue that the double lexicalisation of ‘what’ across NEIDs has a one-to-one correspondence to the Italian complex wh-item ‘che cosa’. In most varieties of northern Italian, the wh-word che cosa maximises the interrogative force of the wh-question. The difference between the two questions in (81) is therefore only pragmatic in nature:

(81) a. Cosa hai detto? (Italian – northern varieties) What have.2SG say.PTCP ‘What did you say?’

b. Che cosa hai detto? What what have.2SG say.PTCP ‘What did you say?!’

In Lamonat and Sovramontino, however, the double lexicalisation of ‘what’ does not produce an emphatic reading compatible to Italian emphatic what. Speakers do not perceive a marked difference between (78) and (80), but do perceive a difference between (79) and (78, 80). The reason, I believe, lies in the status of the clitic wh- element: being the lexicalisation of Q, it simply indexes that we are dealing with a wh- question, in which a constituent ‘what’ is interrogated. Contrary to (78) or (80), in (79) no focus prominence is assigned to the wh-element. I hence claim that the structure of (79) should be interpreted as shown in the simplified labelled bracketing in (82):

(82) [CP Sa-g-ali dat [TP l ge a dat che a Simon]]

The clitic wh-element sa- does not move to the left periphery, but it indexes that a wh- constituent ‘what’ is interrogated in the question, ‘what’ itself, namely the object of the clause, must be realised somewhere in the structure. I believe that it is realised in-situ, within the TP-layer, but it is given null-spell-out (cf. 82 che represented in light grey). Let us consider again (74) and (75), which have been renumbered as (83) and (84) below:

(83) A-la catà quand Maria n om? (Lam. and Sovr.) have.3SG-3SG.F.SCL find.PTCP when Maria a husband ‘When did Maria find a husband?’

119

(84) Quan-a-la catà Maria n om? when-have.3SG-3SG.F.SCL find.PTCP Maria a husband ‘When did Maria find a husband?’

The difference between a wh-question featuring a clitic wh-element (cf. 84) and a wh- question featuring a tonic wh-element (cf. 83) is that the former (cf. 84) can potentially host an element in narrow focus, whereas the latter (cf. 83) cannot, as focal prominence is already assigned to the tonic wh-element (cf. quand in 83). In Lamonat and Sovramontino, given the special role of contrastivess (see chapters 4 and 5), it is virtually impossible to devise a suitable context in which an element in narrow focus coexists with a clitic wh-element in a root question; however, if the focal XP is the subject of the clause, a cleft-question arises where the subject in narrow focus and the wh-clitic coexist, as shown in (85):

(85) Quan-e-lo MARIA che la a catà n om?

when-be.3SG-SCLexpl Maria COMP 3SG.F.SCL have.3SG find.PTCP a husband ‘When did Maria find a husband?’

Further research is undoubtedly needed to capture the fine-grained discourse-pragmatic difference between clitic-wh-questions and tonic-wh-questions.

3.3.5. Preverbal Tonic Wh-Elements: A D-Linked Interpretation

I have not yet proposed an analysis for the type of wh-question in (86) and (87):

(86) Che majon a-tu comprà? (Sovramontino) which jumper have.2SG-2SG.M.SCL buy.PTCP? ‘Which jumper did you buy?’

(87) Che a-lo comprà el? (Lamonat) what have.3SG-3SG.M.SCL buy.PTCP he? ‘What did he buy?’

In (86) and (87) a tonic wh-element is fronted. (86) and (87) seem to posit a challenge to the present analysis of the behaviour of wh-elements in Lamonat and Sovramontino. In fact, it seems that tonic wh-elements can appear in both apparently in-situ and preverbal position. However, in (86) and (87), the preverbal position of the tonic wh-element is systematically justified by its d-linked interpretation (in the sense of Pesetsky 1987). In 120 this respect, crucially, the preverbal position of the tonic wh-element is not identical to the preverbal position of clitic wh-elements. In (86), ‘which jumper’ refers to a limited set of jumpers whose existence and characteristics are known by both interlocutors: they are part of the knowledge shared by the speaker and the hearer. Similarly, in (87), ‘what’ refers to a limited set of items that had to be bought for a communal dinner: implicitly, the interlocutors know what these items are. In chapter 5, I will propose a detailed analysis for d-linked wh-elements. In brief, I will claim that, by virtue of being equipped with a [WH] feature and a [CONTRASTIVE] feature, d-linked wh-elements can occupy SpecForceP, which is a privileged structural discourse-pragmatic position in Lamonat and Sovramontino.

The special behaviour of this type of wh-element in Bellunese is known in the literature, but not fully addressed in terms of discourse-pragmatics. Munaro (1998) notes that, in Bellunese, wh-elements that are structurally complex, like (86), are not licensed in apparently in-situ position, but must appear preverbally. In terms of frequency, d-linked wh-elements are more often part of a complex DP ( ‘which X’), than bare wh-words. In this respect, d-linked wh-elements are often more structurally complex than non-d-linked wh-words; however, this is not always the case, as shown in (87). I claim that the internal structural complexity of the tonic wh-word cannot predict its syntactic position within the clause: what determines the preverbal position of a tonic wh-element is its discourse-pragmatic nature as d-linked (see chapter 5 for further discussion).

3.4. Microparametric Variation in the Locus of SCLI

Having analysed and discussed the behaviour of clitic and tonic wh-elements and the effect of negation on SCLI, it is time to take the investigation a step forward with the following question: why are apparently in-situ wh-elements only present in certain NEIDs and not in others? By comparing Lamonat-Sovramontino with Friulian, I will claim that wh-in-situ is the result of a deeper microparametric variation across NEIDs that involves the locus of SCLI.

121 Let us start our discussion by considering the Friulian yes/no question in (88) and its Lamonat equivalent in (89):

(88) a. No aj-al durmît Mario di besôl?24 (Friulian) NEG have.3SG-3SG.M.SCL sleep.PTCP Mario of alone ‘Didn’t Mario sleep on his own?’

b.*No al a durmît Mario di besôl? NEG 3SG.M.SCL have.3SG sleep.PTCP Mario of alone ‘Didn’t Mario sleep on his own?’

(89) a. No l a durmì Mario da el sol? (Lamonat) NEG 3SG.M.SCL have.3SG sleep.PTCP Mario of him alone ‘Didn’t Mario sleep on his own?’

b.*No a-lo durmì Mario da el sol? NEG have.3SG-3SG.M.SCL sleep.PTCP Mario of him alone ‘Didn’t Mario sleep on his own?’

In (89a), I have claimed that SCLI fails because of the intervening strong negation that acts as host of Q under Force°, namely the locus of SCLI in Lamonat and in Sovramontino. In Lamonat and Sovramontino, SCLI is hence impossible in root negative-interrogatives, as shown by the ungrammaticality of (89b). In Friulian root negative-interrogatives, SCLI exhibits the exact opposite behaviour: in (88a), SCLI takes place regardless of the presence of the preverbal strong negation and, if SCLI does not take place, the sentence yields ungrammaticality (cf. 88b).

I argue that the puzzling data in (88) and (89) can be made sense of if we assume that the locus of SCLI in Friulian is located in a lower C-projection than in Lamonat and Sovramontino. Q in Friulian is not satisfied under Force°, but under Fin°. Taking as points of reference the position of the subject Mario, in SpecTP, and the position in which the strong negation in root negative-interrogatives checks its [NEG] feature, PolarityP, SCLI inversion in Friulian occurs in Fin°. When the complex T° reaches Fin°, Q attaches to it, giving rise to SCLI. The arboreal representation in (90) shows that, in Friulian, the finite verb does not need to rise higher than FinP for SCLI to be licensed:

24 Note that No aj-al Mario durmît di besôl? is also grammatical in Friulian. In contrast to Lamonat and Sovramontino, the auxiliary can be morpho-phonologically independent of the past participle and undergo T-to-C movement on its own. I assume that, if the past participle does not cross the subject position, it stays within the VP: hence, V-to-T movement of the past participle does not take place. I will further discuss this aspect of Friulian in chapter 4. 122 (90) PolP

No Q-aj-al magjât FinP [NEG]

[WH] Fin'

Q -aj-al magjât TP

Mario T'

al a magjât VP

magjât di besôl

Across NEIDs, the presence or absence of SCLI in negative-interrogatives can hence be explained in terms of the microparametric variation in the locus of SCLI: Force° in Lamonat and Sovramontino, and Fin° in Friulian. As for (88a) and (89b), the negation is licensed in PolarityP (Laka 1990), which is located lower than ForceP, but higher than FinP. In Friulian, T° is attracted to Fin° by Q: SCLI takes place before the negation checks its [NEG] feature in PolarityP. In Lamonat and Sovramontino, the negation is licensed in PolarityP before T° reaches Force°: the negation itself is then able to host Q without the need for SCLI. The position of PolarityP, and the difference in the locus of SCLI between Friulian and Lamonat-Sovramontino are schematically shown in (92):

(91) ForceP

Q TopicP Lamonat & Sovramontino FocusP

PolarityP

[NEG] FinitenessP

Q TP Friulian …

If we instead assume an account whereby the negation in negative-interrogatives has focal properties and hence must check a [FOCUS] features in FocusP (a lá Villa-García 2016), the intervening effect of the negation on SCLI can still be appreciated as FocusP, the same as PolarityP, is lower than ForceP, but higher than FinitenessP.

123 Let us now consider the Friulian wh-questions in (92) and (93) and their Lamonat counterparts in (94) and (95):

(92) La-is-e lade to sûr? (Friulian) where-be.3SG-3SG.F.SCL go.PTCP your sister ‘Where did your sister go?’

(93) Dulà is-e lade to sûr? where be.3SG-3SG.F.SCL go.PTCP your sister ‘Where did your sister go?’

(94) Ond-e-la ndaa to sor? (Lamonat) where-be.3SG-3SG.F.SCL go.PTCP your sister ‘Where did your sister go?’

(95) E-la ndaa aonde to sor? be.3SG-3SG.F.SCL go.PTCP where your sister ‘Where did your sister go?’

In section 3.3.2, I claimed that the Friulian wh-word dulà (cf. 93) is a tonic wh-element that bears focus prominence: it has the same morpho-syntactic status as aonde in (95). The corresponding clitic form of dulà is the wh-clitic la in (92): la behaves like onde (cf. 94) in Lamonat and Sovramontino. By assuming a one to one correspondence between Friulian and Lamonat-Sovramontino wh-elements, dulà surfaces in SpecFocP, whereas its clitic counterpart la is attached to the finite verb in the locus of SCLI. If the Friulian wh-elements in (92) and (93) morpho-syntactically behave like their Lamonat- Sovramontino counterparts in (94) and (95), why do they exhibit a different linear order? The difference in linear order is captured by our proposed microparametric variation in the locus of SCLI: Force° in Lamonat and Fin° in Friulian. In Friulian, since SCLI takes place in Fin°, the tonic wh-element dulà and the clitic wh-element la seem to have the same distribution. In reality, the wh-clitic la, the same as onde in Lamonat and Sovramontino, is a lexicalised instance of the Q-morpheme in Fin°, while the tonic wh-word dulà appears in SpecForceP. The identical behaviour of Friulian dulà and Lamonat/Sovramontino aonde as tonic wh-elements in (93) and (95) becomes clearer when the structures of the two sentences are shown under the same arboreal representation. This is shown in (96) below. Note that the Friulian clause is represented in italics, whereas the Lamonat clause in bold:

124 (96) ForceP

Q-e-la ndaa FocusP [WH]

aonde FinP dulà [FOC] Q-is-e lade TP [WH]

to sor T' to sûr

la è ndaa VP e je lade andaa aonde

lade dulà

A similar tree can be drawn to show the identical behaviour of Friulian la (cf. 92) and Lamonat/Sovramontino onde (cf. 94), as in (97) below. Friulian is represented in italics, whereas Lamonat in bold:

(97) ForceP

ond-e-la ndaa FocusP [WH]

FinP

la-is-e lade TP [WH]

to sor T' to sûr

la è ndaa VP e je lade andaa aonde

lade dulà

Another piece of evidence in favour of the microparametric variation in the locus of SCLI across NEIDs comes from the distribution of the particle po with respect to different types of wh-elements. Poletto and Pollock (2004) argue that the particle pa in the Ladin dialect of Fassa marks ‘new information’. They note that, in Fassan, SCLI is only possible when the verb has crossed the position of ‘new information’ marked by pa as shown by (98) and (99) below:

125

(98) a. Ola vas-t pa? (Fassan - Polletto and Pollock 2004: 256) where go.2SG-2SG.SCL FOC ‘Where are you going?’

b. *Ola pa vas-t?

(99) a. Ola pa tu vas? where FOC 2SG.SCL go.2SG ‘Where are you going?’

b. *Ola tu vas pa?

If pa occupies a left peripheral projection lower than FocP, the data in (98) and (99) suggest that the locus of SCLI in Fassan is higher than FocP: in fact, Poletto and Pollock (2004) argue that it is Force°. Let us consider the Friulian data in (100) and (101):

(100) a. Po la-va-tu? (Friulian) FOC where-go.2SG-2SG.SCL ‘Where are you going!?’

b. *La po va-tu?

(101) a. Dulà po va-tu? where FOC go.2SG-2SG.SCL ‘Where are you going!?’

b. *Po dulà va-tu?

Friulian has an understudied comparable particle po that is prosodically stressed and, similarly to Fassan pa, can emphasise new information. Note that (i) I will only consider those instances of po that co-occur with any type of wh-element, and (ii) I will not commit to an exact left peripheral position for po25. I will adopt Poletto and Pollock’s (2004) suggestion that, the same as pa, po marks the ‘boundary’ with new information and I will assume that it occupies an XP below FocusP. As shown in (100a) and (101a), contrary to Fassan, in Friulian the verb does not have to cross the particle po for SCLI to take place. In fact, in Friulian, the locus of SCLI is lower than the position of the particle. If po sits in an XP between Focus and FinP, it precedes the clitic wh-

25 A fine-grained investigation of this particle is needed. I would suggest that it resembles an evidential marker. As for its position, it seems low in the C-domain, possibly an XP between FocusP and FinP. 126 word la in Fin°, and follows the tonic wh-word dulà: this is indeed the case as shown in (100a) and (101a). (100b) is ungrammatical because the wh-element la, like its Lamonat counterpart onde, must attach to the inflected verb: in Friulian, it does so in a lower position than that hosting the particle po. (101b) is ungrammatical because po lexicalises an XP lower than FocusP: it cannot precede the tonic wh-element in SpecFocusP.

In Lamonat and Sovramontino, po is also attested. Speakers claim that their particle po is similar to Trevisan and Venetian particle ʧò. A more fine-grained investigation is needed to assure full comparability between Fassan pa, Friulian po and Sovrmontino/Lamonat po. For the present discussion, let us assume that po sits in the same XP as in Friulian and in Fassan. The wh-questions in (102) and (103) seem to support this assumption:

(102) a. Fon-e come po noaltri? (Lamonat) do.2PL-2PL.SCL how FOC we ‘How do we manage!?’

b. *Fon-e po come noaltri?

(103) a . Quan-va-tu po da to mare? (Sovramontino) when-go.2SG-2SG.SCL FOC to your mother? ‘When do you go to see your mother!?’

b. *Po quan-va-tu da to mare?

In Lamonat and Sovramontino, po occurs after the tonic wh-element come the same as in Friulian (cf. 101). Po cannot precede the tonic wh-element, as in (102b). Similarly to Fassan, SCLI takes place once the verb has crossed the position of new information whose boundary is marked by po. As for the clitic wh-element in (103), po cannot precede it (103b), but must follow the cluster Q-V-SCL, as in (103a). Considering the microparametric variation in the locus of SCLI, the difference between Friulian (100) and Sovramontino (103) can be accounted for in a straightforward manner: in (103), given that the locus of SCLI is Force°, po cannot precede the clitic wh-element as its position is much lower in the structure. It becomes clearer when the sentences in (100) and (103) are represented under the same tree, as in (104) below. Friulian is represented in italics, whereas Sovramontino in bold:

127 (104) ForceP

Quan-va-tu FocusP

XP

po FinP po

la-va-tu TP

te va quand da to mare tu vas dulà

For the sake of completeness and clarity, I will also draw a similar tree for (101) and (102) in (105). Friulian is represented in italics, whereas Lamonat in bold:

(105) ForceP

Q-fon-e FocusP

come XP dulà

po FinP po

Q-va-tu TP

noaltri fon come tu vas dulà

Another piece of empirical evidence in support of the microparametric variation in the locus of SCLI comes from information structure, more specifically, the limitation on the number of preverbal topics in root interrogatives. In the next two chapters, I will discuss this phenomenon in more detail; for now, we are just interested in the number, rather than type of topics that can appear preverbally. Let us consider the Friulian sentence in (106):

(106) Iar tal bosc a-tu viodût to barbe? (Friulian) yesterday in-the woods have.2SG-2SG.SCL see.PTCP your uncle? ‘Did you see your uncle in the woods yesterday?’

128

Given the recursive nature of topics (Rizzi 1997; Benincà and Poletto 2004), Friulian’s C-domain can theoretically host an unlimited number of topics in a yes/no question like (106). On the other hand, in the exact same yes/no question, Lamonat and Sovramontino can only host a single topic preverbally, as shown in (107):

(107) a. *Ier an tel bosc a-tu vedù to barba? (Lam.) yesterday in the woods have.2SG-2SG.SCL see.PP your uncle? ‘Did you see your uncle in the woods yesterday?’

b. Ier a-tu vedù to barba? yesterday have.2SG-2SG.SCL see.PP your uncle? ‘Did you see your uncle yesterday?’

(108) A-tu vedù to barba an tel bosc ier? have.2SG-2SG.SCL see.PP your uncle in the woods yesterday? ‘Did you see your uncle in the woods yesterday?’

Topical elements are not generally licenced in the C-domain of Sovramontino and Lamonat. In Lamonat, the same linear order as (106) would yield an ungrammatical sentence, as in (107a), which becomes grammatical once the presupposed elements are realised in-situ within the TP layer, as in (108). This difference can be accounted for by appealing to the difference in the locus of SCLI across the two NEIDs under investigation: since Friulian SCLI takes place in Fin°, there is room for a more articulated left periphery. By contrast, in Lamonat, the presence of one or more topics would impede the verb in its movement to Force°. In chapter 4, further discussion will follow. In the next chapter, I will also explore the possibility that the present microparametric variation in the locus of SCLI is a reflex of V2: in other words, cross- dialectally, verb movement in root interrogatives directly reflects where, within the C- domain, the V2 constraint was satisfied in the Medieval V2 stage of the NEIDs under investigation (see Benincà 2006; Wolfe 2016).

3.5. Conclusion

In conclusion, I have provided an alternative account for the behaviour of wh- words in Lamonat and Sovramontino. These two NEIDs exhibit two different sets of

129 wh-elements: (i) a tonic set of wh-words that occupy SpecFocP and receive focal prominence, and (ii) a clitic set of wh-elements that are the lexicalisation of the Q- morpheme. I have also claimed that Q is primarily responsible for SCLI in the NIDs. I have shown that, across NEIDs, the apparent difference in the distribution of wh- elements is a ‘side effect’ of a deeper microparametric variation that concerns the locus of SCLI, which can either be Force° or Fin°. I have provided some empirical evidence in support of the effect that this variation has on NEIDs cross-dialectally: the behaviour of negation in root negative-interrogatives, the behaviour of the focal particle po and the limitation on preverbal topic placement in root interrogatives. Furthermore, I have offered a partial description of cleft-questions in Lamonat and Sovramontino. I also provided a partial account of the behaviour of D-linked wh-elements. Both topics will be further explored in the next chapters. Finally, I have only marginally considered the properties of embedded wh-elements, for which further investigation is needed.

130 4. Focus Structure, Pragmatically Motivated V2 Effects and Subjecthood

4.1. Introduction

A fundamental aspect of the discourse-pragmatics of a language is how new or non-presupposed information is unpacked and prompted in discourse. The syntactic organisation of new information in discourse is referred to as focus structure (Lambrecht 1994) and its interaction with the syntax of a language often gives rise to non-canonical or marked word orders. This is the case with Lamonat and Sovramontino, whose non-canonical constituent orders must find an explanation in the interaction between information structure and syntax at the syntax-pragmatics interface. Lamonat and Sovramontino depart from the other Romance languages as the interaction between these two components of grammar is tightly linked to a characteristic syntactic property of these two varieties: the partial retention of a Medieval Romance-like V2 system (see Benincà 1983) that is crucial for the realisation of pragmatically salient constituents. The V2-motivated T-to-C head movement assures both the right pragmatic interpretation and the adequate phonological realisation of the informational units that are unpacked in discourse. In the NEIDs under investigation discourse-pragmatics seems to be a pervasive component of the language, nevertheless, syntax is crucial to enforce the right mapping between the surface position of a constituent and its semantic-pragmatic interpretation.

In Sovramontino and Lamonat, the complex interaction between V2 and information structure determines different specific structural configurations that trigger different discourse-pragmatic readings. In this system, the displaced verb functions as a boundary that marks different discourse functions; in particular, it sharply sets the boundary between contrastive vs. non-contrastive informational units (see Kuno 1976; Kiss 1998; Buring 1999). Contrastiveness as a functional-pragmatic notion will be extensively discussed in the next chapter; in this chapter, I will primarily investigate the interaction between V2 and focus structure.

The interplay between information structure and syntax inevitably shapes the geometry of the left periphery, which Rizzi (1997) proposes as the syntactic locus in

131 which information-structural categories are encoded. For example, I will claim that left dislocation (in the sense of Cinque 1990) is banned in Lamonat and Sovramontino as the head of TopP blocks head movement, which is necessary for the satisfaction of the V2 constraint. Consequently the left periphery of these varieties lacks the TopP layer, but, in turn, features an abundant use of scene-setting topical elements (see Poletto 2000; Benincà & Poletto 2004; Greco & Haegeman 2016 a.o.), which surface in the left-most projection of the C-domain, namely FrameP. The case of Lamonat and Sovramontino suggests that the make-up of the left periphery is, to some extent, language specific and it is shaped by the interaction between information structure and syntax; by the same token, it also shows that the functional projections that make up the left periphery and their hierarchical order are good candidates to be universal features of human language.

The interplay of information structure and syntax also manifests itself on a different level: namely, in the interaction between focus and syntactic function. In this respect, the data from Lamonat and Sovramontino suggest that subjects enjoy special syntactic status over the other arguments of the verb. In fact, in these two NIDs, subjects that encode new information, hence focal subjects, are treated differently by the syntax. I will explore the notion of subjecthood with respect to focus structure and the non-consistent null-subject status of these Romance varieties (see Cardinaletti & Repetti 2010). I will argue that, in Lamonat and Sovramontino, the marked syntactic constructions that arise to accommodate focal subjects are the result of the emergence of a rigid SVO constituent order in which a structural subject position is developing. The fact that focal subjects cannot be realised in the left periphery, in the same way as other arguments of the verb can be read in terms of a ban on particular extraction domains, the impossibility of short-distant movement (see Lasnik & Saito 1992) or the avoidance of the violation of criterial freezing (Rizzi 2006; Rizzi & Shlonsky 2007). I will explore these possibilities, reaching the conclusion that short distance movement is problematic in these two varieties as, along with V2-motivated T-to-C movement, it generates structural ambiguity.

132 4.1.1. Overview and Structure

In the investigation of the interaction between information structure and syntax, the present chapter builds on the previous chapter as it expands on the proposal that the parametric variation in the locus of SCLI across NEIDs is a reflex of Romance Medieval V2. More specifically, I claim that Lamonat and Sovramontino are characterised by the partial retention of V2, which has been reinterpreted as a crucial syntactic strategy to encode salient pragmatic information. This hypothesis helps us account for the pragmatic-syntactic differences that set apart the two groups of NEIDs identified in this thesis: Friulian-Fornese and Sovramontino-Lamonat. In this chapter, I will propose a syntactic analysis that captures the marked behaviour of focus structure in Lamonat and Sovramontino, reaching the conclusion that T-to-C V2-motivated head movement is what ultimately differentiates these two varieties from Friulian and Fornese. In order to convey my arguments more effectively, in section 4.2, I will define and explain the notion of focus structure and its effects on constituent order in Lamonat and Sovramontino. It is very important to note that the conclusions reached in this chapter are solely based on synchronic evidence, as no diachronic data is available for these varieties26. The tacit assumption is that all these varieties were historically V2 systems: the V2 constraint was then lost in Friulian-Fornese, and reinterpreted in Lamonat-Sovramontino as a syntactic strategy to encode pragmatically salient information.

Since Benincà (1983) diachronic research has shown that the Medieval Romance languages were V2 systems (Salvi 2004; Benincà 2006; Ledgeway 2008 a.o). In particular, Poletto (2005, 2006, 2014) showed that Old Venetian, allegedly the main contact language with Lamonat and Sovramontino diachronically, had a fully-fledged V2 system in place. Synchronically, several Rhaeto-Romance varieties, more specifically some Romansh and Ladin dialects (Haiman & Benincà 1992; Poletto 2000 a. o.) still exhibit V2 constituent order. I will especially draw a comparison between Badiotto, a V2 Ladin dialect spoken in the neighbouring region of Trentino-Alto-Adige (Poletto 2000, 2001; Poletto & Zanuttini 2003), and Lamonat-Sovramontino in order to investigate the differences and similarities between the two V2 systems. In Lamonat and

26 No written records of Lamonat and Sovramontino prior to the 20th century can be found. The only NEID under investigation that counts several diachronic linguistic studies is Friulian (see Heinemann & Melchior 2011). Friulian also has its own on-line historical corpus: Dizionario Storico Friulano (Vicario et al. 2009) 133 Sovramontino, the V2 constraint is pragmatically motivated: the V2 system has retained the pragmatic saliency of the preverbal position from Medieval Romance and further developed to become a syntactic means of assuring linear adjacency with focal elements. Benincà (1994) argues that the retention of the Medieval V2 constraint in some Romansh and Ladin varieties is due to sustained language contact with German varieties. In the case of Lamonat and Sovramontino, however, I suggest that this is not the case, as no German variety is spoken in the neighbouring areas. I believe that these two Romance varieties partially retained this syntactic trait due to their isolated mountainous geographic location. As Salvi (2016) points out, the V2 system in place in some Romansh and Ladin varieties is different from that of Medieval Romance. The same is true for Lamonat and Sovramontino’s V2 system. I will argue that the traditional preverbal position only hosts constituents bearing a contrastive reading. In fact, in Lamonat and Sovramontino, it seems that the [EDGE] feature that, in V2 languages, is responsible for the presence of a preverbal constituent (see Vikner 1995) has pragmatically specialised, becoming [CONTRASTIVE].

Lamonat and Sovramontino’s V2 system therefore differs from that of Medieval Romance in that the latter could host most types of pragmatically salient constituents in preverbal position (see Benincà 2006). Lamonat and Sovramontino’s V2 system also differs from that of Badiotto (Poletto 2000, 2002) as (i) no SCLI takes place in Lamonat and Sovramontino, and (ii) Badiotto allows constituents other than pragmatically contrastive ones in preverbal position. In section 4.3, I will explore these differences at length suggesting that Lamonat and Sovramontino’s contemporary V2 system must find an adequate description in light of its tight interaction with information structure. Following Wolfe’s (2016) V2 typology, the synchronic data from Lamonat and Sovramontino suggest that these two varieties were once High-V2 or Force-V2 systems: making reference to Rizzi’s (1997) split CP model, V2-motivated T-to-C movement targeted the highest left peripheral head, namely Force°, and it was accompanied by A’- movement of another constituent in its specifier position. In synchrony, I argue that, while verb movement to Force° is often still in place27, the generalised A’-movement of a preverbal constituent has been lost, but mandatorily retained for a reduced set of pragmatically salient elements: contrastive foci and contrastive topics. In the next

27 In section 4.6, I will argue that V2-motivated T-to-C movement does not occur in the case of in-situ focalisation, more specifically, in the case of predicate or sentence focus (in the sense of Lambrecht 1994). 134 chapter, I will provide an in-depth analysis of the syntactic behaviour of contrastive elements. In this chapter, I will limit the investigation to the interaction between focus structure and V2.

In light of this, as argued in the previous chapter, wh-in-situ in Lamonat and Sovramontino can be captured by postulating that the verb crosses the fronted wh- element by moving as high as Force° due to the V2 constraint. In section 4.3, I will briefly revisit the evidence put forward in the previous chapter in favour of verb movement to Force° in interrogatives and expand on it to include data that suggest that the same verb movement takes places in declarative sentences. Rizzi (1996, 2001), in his discussion of the wh-criterion, argues that, in Romance, evidence of previous V2 systems is indeed often found in verb movement in questions. Along the same lines, Salvi (2016) points out that, in Romance, V2 residues are often found in the constituent order of interrogative sentences, especially among NIDs. Lamonat and Sovramontino’s V2 system becomes more transparent when compared with the clausal-level constituent orders found in the other NEIDs under investigation in the present research, namely Friulian and Fornese. Friulian and Fornese have either lost their High-V2 system or, diachronically, developed a Low-V2 system (targeting Fin°) which was eventually lost, exhibiting, in synchrony, only a minor trace of this system detectable in the verb raising as high as Fin° in questions. Although only a systematic diachronic study can shed light on this issue, what is uncontroversial is that Friulian and Fornese are not V2 systems synchronically. In section 4.5, I will show some Friulian data that suggest that the NEID under investigation exhibits a similar interaction between V2 and focus structure; nonetheless, the bulk of my data clearly indicates that Friulian has lost the V2 constraints.

The loss or reanalysis of V2 among NEIDs was presumably accelerated by language contact with Italian. In the case of Lamonat and Sovramontino, the V2 constraint must therefore be conceived of as a syntactic operation that takes place in a system in which two competing grammars are at work: an archaic V2 system and an Italo-Romance SVO system. The clash between these two systems generates what I define as a pragmatically motivated V2 system, in which the V2 constraint has been retained as a syntactic means to fulfil specific pragmatic needs of the language. T-to-C movement of the verb takes place to license contrastive elements and to assure that

135 elements in narrow focus are assigned sentential stress (see Cruschina 2012), which is crucial for the prosodic realisation of focus prominence (Zubizarreta 1998; D’Imperio 2002; Donati & Nespor 2003; Bocci & Avesani 2006; Frota et al. 2007). On the other hand, in Lamonat and Sovramontino, the underlying default SVO constituent order surfaces in sentences expressing broad focus (see Lambrecht 1994), in which the focal part of the utterance extends to more than one constituent. In section 4.6, I will investigate how broad focus differs from narrow focus and argue that the realisation of broad focus is independent of the V2 constraint. In the two NEIDs, I argue that broad focus is expressed through the means of in-situ focalization: the constituents in focus receive a focal reading in their base-generated position. In these cases, the V2 constraint is overridden by in-situ focalisation and V2-motivated T-to-C movement does not occur. The way in which the syntax treats broad focus suggests that the realisation of narrow focus does not only require the achievement of the right syntactic configuration, but also of the right prosodic contour. Adjacency with the verb cluster plays a key role, as it assures that the focal constituent is assigned focal prominence. In this respect, Lamonat and Sovramontino came to associate the V2 constraint with this requirement.

Following these introductory remarks, the complex interaction between the V2 constraint and focus structure should become more transparent in relation to apparent wh-in-situ in Lamonat and Sovramontino. The SVO grammar pushes the NEIDs under investigation to realise the wh-element in canonical position under FocP in the left periphery, whereas the V2 grammar forces the verb to move to Force° to satisfy the V2 constraint. The result of this composite syntactic operation is apparent wh-in-situ; the wh-element receives focal prominence by virtue of being adjacent to the verb cluster. Apparent wh-in-situ mirrors the generalised behaviour of constituents in informational narrow focus. In fact, in Lamonat and Sovramontino, focus fronting is never possible as shown by the ungrammaticality of (1a) and (2a):

(1) CONTEXT: Who did you see? (Lamonat) a. *NANI mi è vedù Nani I have.1SG see.PTCP ‘I saw NANI’

b. è vedu NANI mi have.1SG see.PTCP Nani I ‘I saw NANI’

136 (2) CONTEXT: Did you see Toni? a. *NANI mi è vedù, no Toni Nani I have.1SG see.PTCP, NEG Toni ‘I saw NANI’

b. NANI è vedù mi, no Toni Nani have.1SG see.PTCP I, NEG Toni ‘I saw NANI, not Toni’

In case of informational focus, in (1b), the constituent in narrow focus must always be shielded by the verb28. Similarly, in the case of contrastive narrow focus, focus fronting is not sufficient per se, but the verb cluster must be adjacent to it. Note that the subject mi can be omitted in all the examples as Lamonat and Sovramontino still qualify as null-subject languages, but, if the subject is uttered as background information, it follows the distribution in (1) and (2).

By the same token, apparent wh-in-situ mirrors the behaviour of constituents in informational narrow focus: wh-fronting, similarly to focus fronting, is banned and the wh-element must be shielded by the moved verb cluster, as shown in (3):

(3) a. *AONDE se-tu ndat? (Sovramontino) where be.2SG-2SG.SCL go.PTCP ‘Where have you been?’

b. Se-tu ndat AONDE? be.2SG-2SG.SCL go.PTCP where ‘Where have you been?’

The only difference between (1b) and (3b) is the presence of subject clitic inversion (SCLI) in the latter, which is due to the interrogative force of (3b). As argued in the previous chapter, SCLI in (3b) takes place independently of the V2 constraint due to the presence of the affixal head Q under Force° that is necessary to convey the interrogative

28 Given the competing SVO system, the focal constituent in (1b) can occasionally be realised in-situ; this type of variation is found especially among the younger speakers of the Lamonat and Sovramontino speech communities. In section 4.4, I will argue that whether the speaker realises informational focus in the left periphery or in-situ depends on whether or not that particular piece of information is pragmatically marked or unmarked in terms of discourse saliency. If it is unmarked, younger speakers will realise it in-situ. Note that the informational status of this type of focus is preserved in both positions; in fact, contrastive/corrective focus is realised altogether in a different position in the left periphery (SpecForceP), that is higher than FocP where marked informational focus is realised. 137 force of the sentence. In sections 4.3 and 4.4, I will suggest that V2 motivated verb movement takes place in both interrogative and declarative sentences; in the former case, Q is satisfied by the V2 constraint itself as, in Lamonat and Sovramontino, both syntactic constraints involve the same left peripheral head, namely Force°.

Through the investigation of focus structure, it is possible to observe the structural effects of the two grammars competing in Lamonat and Sovramontino, namely V2 and SVO constituent order. In fact, whereas the V2 constituent order is employed in the realisation of narrow focus, the SVO constituent order is dominant in the expression of broad focus, suggesting that the latter is the underlying default constituent order of these Romance varieties. In section 4.6, I will provide further evidence in support of a default SVO underlying constituent order by investigating the interaction between focus structure and presentational sentences featuring core unaccusative verbs (Belletti 1988). Along the same lines, in section 4.2, I will investigate cleft structures, which exploit in-situ focalization and mirror the default SVO constituent order of these languages.

4.2. Focus Structure

Traditionally, the part of the utterance that is asserted in contrast to the presupposed or old part of the same utterance is called the focus of the sentence (Lambrecht 1994; Kiss 1998). In other words, focus refers to new or non-presupposed information that was prompted in discourse. In order to investigate the interplay between syntax and information structure in Lamonat and Sovramontino, we first need to understand the hierarchical structure of focal information in an utterance. I will adopt Lambrecht’s (1994) focus structure, which describes the way in which new information is arranged in a sentence, depending on which portion of the utterance enjoys focal status. The focal portion of the sentence is called the focus domain (Lambrecht 1994). An important contrast is between broad focus and narrow focus; in the former, the focus domain extends to more than one constituent and encompasses the whole sentence, whereas in the latter it is limited to one single constituent. In Lamonat and Sovramontino, narrow focus is associated with V2 T-to-C movement, whereas broad focus is generally found with canonical SVO constituent order. Even though Lamonat and Sovramontino exhibit a comparable syntactic behaviour with respect to the 138 interaction between information structure and syntax, Lambrecht’s (1994) two-way distinction of focus is also important to account for the existing micro-variation in the encoding of focus structure between Lamonat and Sovramontino: in fact, the two varieties slightly diverge in the realisation of focal subjects in narrow focus. Note that the notion of broad focus and its analysis will not be dealt with until section 4.6.

4.2.1. Narrow Focus

In narrow focus, the focal domain of the utterance is limited to a single constituent (Lambrecht 1994). Only one constituent per sentence can appear in narrow focus. Differently to topics, foci cannot be recursive: only one focal element is allowed per utterance (see Rizzi 1997; 2018). Narrow focus is incompatible with resumptive clitics and cannot co-occur with wh-elements. Prosodically, it is associated with the main prominence of the sentence (see Cruschina 2012, 2016). In section 4.4, I will show that this last point is crucial in understanding the interaction between focus structure and the V2 constraint. Narrow focus marks the non-presupposed, contrastive or unidentifiable element of an utterance (Prince 1981; Dik 1989; Lambrecht 1994). According to its different discourse-pragmatic interpretation, narrow focus can be further categorised into informational and contrastive29. There is general agreement in favour of two separate dedicated structural positions that encode contrastive and, respectively, informational focus (Kiss 1998; Rizzi 1997; Belletti 2001, 2004: Benincà & Poletto 2004; Bianchi & Francarelli 2010; Cruschina 2012; Bianchi & Bocci 2012; Bocci 2013; Rizzi & Bocci 2017 and many others). Contrastive focus is a special type of narrow focus: a focal element is contrastive when it specifically refers to an information unit that is contrary to the corresponding interlocutor’s presuppositions (Kiss 1998). Informational focus, on the other hand, refers to new non-presupposed information that does not bear a contrastive reading, but contributes towards the advancement of the conversation (see Krifka 2007). Lamonat and Sovramontino sharply distinguish between informational and contrastive focus assigning them different syntactic positions. Both types of narrow focus, nonetheless, must be linearly adjacent

29 Narrow focus can also encode an exhaustive, corrective or mirative reading (see Cruschina 2012 for further discussion). In Lamonat and Sovramontino, corrective focus patterns with contrastive focus. The same holds true for mirative focus, with the exception of mirative focus in questions (see Bianchi et al 2015) that seems to occupy a different syntactic position. Exhaustivity is generally encoded in cleft structures. 139 to the verb cluster: informational focus is characterised by right adjacency with the verb cluster, as in (4), whereas contrastive focus features left adjacency with the verb cluster, as in (5):

(4) CONTEXT: What did Mario drink? (Sovramontino)

a. l a beest EL VIN Mario 3SG.M.SCL have.3SG drink.PTCP the wine Mario 'Mario drank the wine' b. #EL VIN l a beest Mario the wine 3SG.M.SCL have.3SG drink.PTCP Mario 'Mario drank the wine' c. *l a beest ier EL VIN Mario 3SG.M.SCL have.3SG drink.PTCP yesterday the wine Mario 'Mario drank the wine yesterday'

(5) CONTEXT: Did Mario drink the wine?

a. No, LA SGNAPA l a beest Mario NEG the grappa 3SG.M.SCL have.3SG drink.PTCP Mario 'No, Mario drank the grappa' b. #No, l a beest LA SGNAPA Mario NEG 3SG.M.SCL have.3SG drink.PTCP the grappa Mario 'No, Mario drank the grappa' c. *No, LA SGNAPA ier l a beest Mario NEG the grappa yesterday 3SG.M.SCL have.3SG drink.PTCP Mario 'No, Mario drank the grappa yesterday '

In example (4), el vin (the wine) is an instance of informational focus, whereas la sgnapa (the grappa) in (5) is an instance of contrastive focus with a corrective interpretation. Note that, starting with (4) and (5), contrastive focal elements will be represented in BOLD CAPITALS, informational narrow focus in CAPITALS and contrastive topics in bold.

To better understand the difference between informational and contrastive focus, I will introduce the concept of common ground (Stalnaker 1974, Lewis 1979). The common ground (CG) can be defined as the set of propositions that is mutually known by the speech participants and is to be shared in communication. It is made of different subcomponents that characterise the conversational context, and each sentence must be interpreted as a potential update of the CG. Krifka (2007) draws an important

140 distinction between CG content and CG management: the former concerns the truth- conditional information that has been stated up to a point in conversation; the latter concerns the conversational moves of the participants involved in the communicative act. In light of the notion of the common ground, we can now further distinguish informational and contrastive focus. Contrastive focus is a discourse-pragmatic strategy that pertains to CG management, whereas informational focus pertains to CG content (Bianchi & Frascarelli 2010, Krifka & Musan 2012). In chapter 5, I will return to this distinction and further discuss the discourse-pragmatic notion of contrastiveness.

In (4) and (5), both types of narrow focus must be linearly adjacent to the verb cluster, as shown by the ungrammaticality of (4c) and (5c). Contrastive focus exhibits adjacency to the left of the verb, whereas informational focus features right-adjacency to the verb. (4b) and (5b) show that the two narrow focus positions are not interchangeable, but encode different discourse-pragmatic readings. As I pointed out in the previous chapters, Lamonat and Sovramontino qualify as non-null-subject languages, therefore, the subject Mario in (4) and (5) does not need to appear. In (4) and (5), Mario is in fact characterised by a falling prosodic intonation, that generally marks background information (see Bianchi and Frascarelli 2010). The subject cannot be placed between the element in narrow focus and the verb. This is because linear adjacency of the focal element with the verb must be achieved. The verb bears sentence stress, it assigns focal prominence (Zubizarreta 1998; D’Imperio 2002; Donati & Nespor 2003; Bocci & Avesani 2005) to the element in narrow focus through linear adjacency. Focus fronting is therefore not a sufficient mechanism to assure that the focal element receives focal prominence, the verb must also move out of its base- position to land in an adjacent position. In Lamonat and Sovramontino, the realisation of narrow focus is hence a composite syntactic operation, in which linear adjacency is achieved through: (i) movement of the element in narrow focus out of its base-position and (ii) movement of the verb out of its base-position.

4.2.2. Subjects in Narrow Focus: Lamonat vs. Sovramontino

Let us now turn our attention to the interaction between narrow focus and subjecthood. Focus structure is sensitive to argument structure: cross-linguistically, the syntactic manifestation of focus is often affected by the grammatical function of the

141 constituent that bears the focal reading (see Bentley 2008; Cruschina 2012). Subjects in narrow focus enjoy a special syntactic treatment. Lamonat and Sovramontino differ with respect to how the syntax treats focal subjects: Lamonat exhibits two specific syntactic strategies to accommodate focal subjects, namely cleft structures and object topicalisation. Sovramontino, on the other hand, treats focal subjects in the same way as direct and indirect objects, oblique arguments and adjuncts30. In this subsection, I will explore these differences and try to account for the special syntactic behaviour of focal subjects in Lamonat. The discussion will lead me to the following conclusions: (i) the special behaviour of subjects in narrow focus is linked to their need to move out of their canonical position in order to receive a focal reading; and (ii) Lamonat is a step ahead in the process of shifting towards an SVO constituent order with respect to Sovramontino.

Lamonat and Sovramontino differ in the treatment of focal subjects only if they encode informational narrow focus. The two NIDs in fact treat subjects in contrastive narrow focus in the same way. In chapter 5, I will discuss the claim that Lamonat and Sovramontino convey contrastiveness through a specific pragmatic-syntactic strategy, whereby contrastive elements, regardless of their informational status as focus or topic, occupy a privileged position in the left periphery (higher than that occupied by informational focus). For now, this asymmetry shows that informational and contrastive focus occupy two distinct structural positions. The following discussion therefore only concerns focal subjects in informational narrow focus: new non-presupposed information that does not bear a contrastive reading. In both varieties, direct and indirect objects, oblique arguments and adjuncts in narrow informational focus are licensed in postverbal position, as shown by (6) and (7):

(6) a. CONTEXT: Who did Mario see? (Lamonat) L a vedù TO SOR Mario 3SG.M.SCL have.3SG see.PTCP your sister Mario ‘Mario saw your sister’

b. CONTEXT: What did Mario give to Toni? L ge a dat DO OU Mario a Toni 3SG.M.SCL DAT.CL have.3SG give.PTCP two eggs Mario to Toni ‘Mario gave two eggs to Toni’

30 Nevertheless, Sovramontino does reserve a special treatment for focal subjects in wh- questions, in which the focal element bearing the subject function is the wh-element (see chapter 3). 142

(7) a. CONTEXT: Who did Mario see? (Sovramontino) L a vist TO SORELA Mario 3SG.M.SCL have.3SG see.PTCP your sister Mario ‘Mario saw your Sister’

b. CONTEXT: What did Mario give to Toni? L ge a dat DO UVI Mario a Toni 3SG.M.SCL DAT.CL have.3SG give.PTCP two eggs Mario to Toni ‘Mario gave two eggs to Toni’

As previously argued, the postverbal position of the element in narrow informational focus is not its in-situ position. In a similar fashion as wh-items in Lamonat and Sovramontino (cf. chapter 3), I argue that narrow focal elements appear ex-situ in FocusP where they receive focal interpretation. The hypothesis that narrow focal elements do not appear in their in-situ position is supported by: (i) the adjacency requirement between verb cluster and focal element, and (ii) the relative order of constituents in the clause (cf. 6 and 7). As for the latter point, the order of the constituents is fixed regardless of their grammatical function: in the case of narrow informational focus the first element of the clause is generally the verb cluster (made of its clitics, auxiliary or inflected verb and, if relevant, the past participle)31 and the second element is the constituent in narrow focus. Nothing can intervene between the verb cluster and the focal element. Also, the topical constituents of the clause, if uttered, must appear after the element in narrow focus in their fixed hierarchical order (cf. 6b and 7b), as shown by the ungrammaticality of (8a):

(8a). CONTEXT: What did Mario give to Toni? (Sovramontino) *L ge a dat DO UVI a Toni Mario 3SG.M.SCL DAT.CL have.3SG give.PTCP two eggs to Toni Mario ‘Mario gave two eggs to Toni’

(8a) is ungrammatical because the beneficiary, a Toni, appears higher in the clause than the subject Mario. The rigid constituent order in (7) and the ungrammaticality of (8a) suggest that, in these contexts, topical elements are not cases of afterthought or clitic

31 Unless the clause is introduced by a frame-setter or an A-Topic (see chapter 5). Frame-setters and A-topics are nonetheless clause-external and therefore do not count as the first element of the clause. 143 right dislocation32 (see López 2009; Cruschina 2016), but they occur in-situ, whereas the constituent in informational narrow focus is realised ex-situ in the C-domain, namely in FocusP. Assuming that the constituent in informational narrow focus moves to the left peripheral space of the clause means arguing that the verb cluster also appears in the C-domain. I claim that this is indeed the case: the verb moves higher than the focal element to ForceP. This movement is dictated by the V2 constraint at work in these two NIDs. In section 4.3, I will provide further evidence in support of the Force- V2 nature (in the sense of Wolfe 2016) of Lamonat and Sovramontino. By satisfying the V2 constraint, T-to-C movement also ensures that linear adjacency of the verb cluster and the focal element is achieved. The labelled bracketing in (8b) shows the proposed structural representation of (7b):

(8b) [ForceP l ge a dat [FocP DO UVI [TP Mario [T’ [T l ge a dat [VP dat [DP do uvi [PP a Toni]]]]]]]]

The composite movement of the verb cluster and the focal element to the left peripheral space accounts for why these two elements must be adjacent and in a rigid sequential order: verb cluster-focal element (regardless the non-subject grammatical function of the constituent bearing narrow focus).

If the constituent in informational narrow focus is the subject of the clause, Lamonat and Sovramontino exhibit different syntactic behaviours. In both varieties, subjects in narrow informational focus cannot appear in their canonical preverbal position, namely SpecTP (cf. 9a and 10a). In Sovramontino, focal subjects are treated in the same way as any other argument of the verb and adjuncts: they occur in postverbal position after the verb cluster, as in (10b). In Lamonat, on the other hand, this constituent order is ungrammatical: the focal subject cannot follow the verb cluster in clause-initial position, as in (9b):

32 In (7), the agreeing subject clitic and the dative clitic may seem to suggest that the constituents at the right edge of the clause, namely Mario and a Toni, are clitic right dislocated elements. However, this is not the case as l and ge in (7) are not resumptive clitics. In Lamonat and Sovramontino, SCLs are always obligatory (cf. chapter 2), regardless of the discourse- pragmatic status of the subject. Also, the dative clitic ge is obligatory, if the verb selects a beneficiary as argument (regardless its discourse-pragmatic status). 144 (9) CONTEXT: Who killed the dog? (Lamonat) a.*MARIO l a copà l can Mario 3SG.M.SCL have.3SG kill.PTCP the dog ‘Mario killed the dog’

b. *L a copà MARIO l can 3SG.M.SCL have.3SG kill.PTCP Mario the dog ‘Mario killed the dog’

(10) CONTEXT: Who drank the wine? (Sovramontino) a.*TONI l a beest el vin Toni 3SG.M.SCL have.3SG drink.PTCP the wine ‘Toni drank the wine’

b. L a beest TONI el vin 3SG.M.SCL have.3SG drink.PTCP Toni the wine ‘Toni drank the wine’

The constituent order in (9b) is ungrammatical. If the subject appears after the verb cluster, it is interpreted as the object of the clause. The default interpretation of (9b) is that il can, the dog, killed Mario and not viceversa. In Lamonat, the subject hence cannot be focalised through the same syntactic strategy as the direct object. By contrast, in Sovramontino, the constituent order in (10b) is grammatical: speakers seem to distinguish the agent from the patient relying on previous discourse.

Let us now examine the two possible syntactic strategies to accommodate subjects in narrow informational focus in Lamonat, and try to explain why subjects cannot be focalised in postverbal position. The two syntactic strategies that Lamonat can adopt to accommodate subjects in narrow informational focus are shown in (11a) and (11b):

(11) CONTEXT: Who killed the dog? (Lamonat) a. Al lo a copà MARIO l can 3SG.M.SCL 3SG.M.OCL have.3SG kill.PTCP Mario the dog ‘The dog, Mario killed it' b. L è sta MARIO che l a copà l can

3SG.SCLEXPL be.3SG be.PTCP Mario COMP 3SG.M.SCL have.3SG kill.PTCP the dog ‘It was Mario that killed the dog’

145 4.2.2.1. Accommodating Focal Subjects: Object Topicalisation

The first strategy in (11a) is object topicalisation, whereas the second strategy in (11b) is the use of a cleft structure. Let us discuss the first strategy, namely object topicalisation, and focus on (11a). If uttered, the topicalised object must appear after the focal subject as shown in (12):

(12) CONTEXT: Who drank the wine? (Lamonat)

a. L lo a beù MARIO l vin 3SG.M.SCL 3SG.M.OCL have.3SG drink.PTCP Mario the wine ‘Mario drank the wine’

b. *L vin el lo a beù MARIO the wine 3SG.M.SCL 3SG.M.OCL have.3SG drink.PTCP Mario ‘Mario drank the wine’

The fact that the topicalised object cannot appear preverbally (cf. 12b) sets off Lamonat from Italian that also adopts the same syntactic strategy (object topicalisation) to accommodate subject in narrow focus (see Bentley 2008). Given the rigid constituent order, I argue that in (11a) and (12a), the same mechanism of focalisation of non- subject arguments and adjuncts is in place: the focal element moves to FocusP and the verb to ForceP leaving the object in the TP-layer. In (12a), l vin is therefore not clitic right dislocated (Cinque 1990; López 2009; Cruschina 2016), but it appears in-situ. The object is, nonetheless, clearly topicalised: this can be seen by the context provided and the fact that it is resumed by an agreeing pronominal clitic (see Cinque 1990, 1997). The prosodic contour of (12a) is comparable to that of a sentence that contains a non- subject argument in informational narrow focus: it is featured by a rising intonation that culminates on the past participle beù (the prosodic pitch), the focal element Mario is characterised by a high intonational plateau, and, finally, the intonation falls on the in- situ topical elements (see section 4.3.4 for further discussion). There is no audible pause that sets off the focal element (in 12a, Mario) from the topical constituents encoding background information (in 12a, l vin). The rejection of clitic right dislocation is also based on the fact that clitic left dislocation (Cinque 1990) is banned in Lamonat and Sovramontino, if the dislocated topical element encodes given or background information (see Bianchi & Frascarelli 2010): it must however take place, if the topical element bears a contrastive reading (see chapter 5). The ban on clitic left dislocation is 146 reconcilable with the V2 nature of the two NEIDs under investigation. In fact, Rizzi (1997) argues that left dislocation constitutes a barrier for head movement due to its successive cyclic fashion, as Top° is not a suitable host for the transit of T-to-C movement. In section 4.3, I will discuss the V2 properties of Lamonat and Sovramontino.

Let us now go back to object topicalisation in Lamonat as a syntactic strategy to accommodate subjects in informational narrow focus. Further research is needed to adequately explain why object topicalisation is needed to licence the composite movement of the verb cluster and the subject in narrow focus to the left periphery. In particular, it would be worth comparing the same strategy in Lamonat and Italian to achieve a better understanding of the phenomenon. The data in (12) suggest that the underlying syntactic operations at the basis of this strategy are different in Italian and Lamonat. As for why object topicalisation is needed in Lamonat to accommodate subjects in informational narrow focus, I want to tentatively suggest two reasons, which are not necessarily incompatible: (i) since the direct object is resumed by an agreeing pronominal clitic, reduces the ambiguity between the subject and the object, especially if the former appears in canonical object position; (ii) the topicalisation strategy, marked by an agreeing resumptive pronominal clitic, may facilitate T-to-C movement. In V2 Medieval Romance (Benincà 2006), T-to-C movement of the inflected verb accompanied the topicalisation of the object in SpecCP. It may be that, in the contemporary V2 system in Lamonat, if an element is topicalised, the language more strongly resorts to V2 verb movement as a reflex of the original V2 system (even if the movement of the topical element in SpecCP has been lost).

4.2.2.2. Accommodating Focal Subjects: Cleft Structures

Let us know consider the second syntactic strategy employed by Lamonat to accommodate subjects in informational narrow focus, namely the use of a cleft structure, as in (11b) above and in (13) below:

147 (13) CONTEXT: Who is cooking? (Lamonat)

L è I TOSAT che i è dreghe

3SG.SCLEXPL be.3SG the boys COMP 3PL.M.SCL be.3PL behind

a far da magnar to make.INF of eat.INF ‘It is the boys that are cooking’

Clefts are a characteristic trait of Bellunese, the supra-dialectal variety to which Lamonat and Sovramontino belong, and clefts are most commonly associated with focalised subjects (see Munaro 1998). In Lamonat, clefts are used to accommodate subjects in informational narrow focus. This strategy coexists with object topicalisation: nevertheless, it is more frequently employed and complements object topicalisation (for example, with the subject of unergative verbs, which do not have an object). (13) is a cleft structure, in which the focal subject i tosat is clefted: it appears between the dummy third singular copular be, l è, and the complementiser, che, that introduces a relative clause. I tosat loses its subject properties and becomes the complement of copular be. There is no agreement between the third singular subject clitic l and the plural expression, i tosat, showing that the structural subject of the matrix clause cannot be the latter (see Poletto 1993, 2000 for the analysis of subject clitics as agreement markers between subject and inflected verb). The dummy subject clitic l is instead the manifestation of the agreement between the inflected verb and expletive pro (Rizzi 1986; Rizzi & Shlonsky 2006), which is the structural subject of the matrix clause in (13). Through the use of a cleft structure, the subject becomes the structural complement of copular be in a similar fashion to the thematic argument of core unaccusative verbs (see Belletti 1988, Levin & Rappaport Hovav 1995). In (13), i tosat is syntactically treated as an object and focalised in the same way as objects are. I hence argue that i tosat is focalised in the left periphery of the matrix clause. The in-situ null copy of the moved constituent along with the CP it selects form a big DP (à la Munaro 1998)33. This account departs from the standard view that, in clefts, copular be selects a

33 The syntactic properties of the big DP differ from those of a regular DP. For a start, the big DP does not pass most constituency tests. The sentence in (14) shows that the clefted focal element can be excorporated from the big DP and contrastively focalised in the left periphery of the matrix clause. Unless we assume an elaborate apparatus of discontinuous spell-outs, this is rather problematic in the conception of the big DP as a fully-fledged constituent in complement position of copular be. I believe that the big DP must be regarded as such only at the semantic level, as, on purely syntactic grounds, it would not hold. Munaro (1998) primarily adopts the big DP analysis because the SCL after the complementiser is optional. However, this is not true 148 small clause (in the sense of Burzio 1986, Stowell 1982 and Rothstein 2000) and the clefted focal element occupies the focus position of the left periphery of the small clause34. In Lamonat and Sovramontino, it seems that the clefted element does not appear in the C-domain of the small clause. This becomes evident in the case of contrastive narrow focus; in fact, the focal element can ‘move out’ of the cleft, as shown in (14):

(14) CONTEXT: Is your wife cooking? (Lamonat)

No. I TOSAT l è che i è dreghe

NEG the boys 3SG.SCLEXPL be.3SG COMP 3PL.M.SCL be.3PL behind

a far da magnar to make.INF of eat.INF ‘No, it is the boys that are cooking’

Without engaging into a detailed analysis of (14), if i tosat were focalised in the left periphery of the small clause selected by copular be, it would be difficult to explain why it appears higher than copular be itself. (14) is compatible with an analysis of (13) in which the clefted element is either focalised in-situ (complement position of copular be) or in the left peripheral space of the matrix clause. In the case of Lamonat and Sovramontino, I argue in favour of the latter option: as shown in (15), the labelled bracketing representation of (13):

(15)[ForceP l è [FocP I TOSAT [TP proexpl [T l è [VP è [DP i tosat [CP che i è dreghe a far da magnar]]]]]]]]

Similarly to objects, oblique arguments and adjuncts in informational narrow focus, the focalisation of the structural complement of copular be consists of two distinct instances of internal merge: (i) i tosat, from its base-generation position (the complement position of copular be) is moved to FocusP, where it is assigned focal

in Lamonat and Sovramontino cleft structures, in which the SCL is instead obligatory. It is optional, if not ungrammatical, in appositive relative clauses, suggesting that there is no perfect overlap between relative clauses and clefts. 34 An alternative account is that the clefted focal element bearing the subject function is focalised in the vP periphery of copular be (see Belletti 2008). While this claim may hold for Italian, it does not for Lamonat and Sovramontino. As discussed in chapter 3, in Lamonat and Sovramontino, Belletti’s (2001, 2004) vP periphery is not involved in postverbal focalisation. 149 interpretation, and (ii) copular be is moved to ForceP (T-to-C movement) to satisfy the V2 constraint. This analysis allows for a uniform account of non-subject constituents in informational narrow focus.

4.2.2.3. Subjects of Core Unaccusatives

Lamonat and Sovramontino V2 systems are shifting towards a SVO constituent order. Lamonat seems to be ahead of Sovramontino in this process and, in the next section, I will argue that this is why it must cleft subjects in informational narrow focus. Given the two competing grammars, namely V2 and SVO, cleft structures can be reanalysed as instances of in-situ focalisation, in which both copular be and the clefted focal element appear in-situ. Across Romance, in-situ postverbal focalisation is a common strategy to express narrow focus (see Cruschina 2012, 2016). In Lamonat (and possibly Sovramontino), another ambiguous case between in-situ vs. left peripheral focalisation concerns the only thematic argument of core unaccusative verbs (Belletti 1988; Levin & Rappaport 1995), as in (16):

(16) CONTEXT: Who died? (Lamonat & Sovramontino)

è mort DO FEMENE be.3SG die.PTCP two women ‘Two women died’

Unaccusative verbs lack the external argument and their so-called logical subject is base-generated in the verb’s complement position, syntactically behaving like an object. In fact, do femene does not agree with the verb in number and gender, suggesting that the subject position is filled by a null expletive, as shown by: (i) the lack of an agreeing subject clitic (third person plural feminine) and (ii) the lack of agreement between the past participle mort and le femene, which would be in place, if le femene were the subject of the clause. In this respect, the only thematic arguments of unaccusative verbs are structurally identical to clefted elements in relation to copular be: they occupy the same internal argument position, structurally being the complement of the verb phrase. The difference between unaccusative verbs and copular be in clefts is that the former assign a theta role (generally the theme semantic role) to their internal argument,

150 whereas dummy copular be does not, although this difference does not have structural consequences. In the same fashion as clefted elements, I claim that, in (16), do femene is focalised in the C-domain, under FocusP, and the verb moves to ForceP to satisfy the V2 constraint. In most SVO Romance languages, however, (16) would be an instance of in-situ postverbal focalisation. I will address in-situ focalisation in section 4.6 with respect to broad focus.

In Lamonat, it could also be the case that focal constituents like the clefted element in (13) and the core argument of die in (16) have been reanalysed as instances of in-situ focalisation. This is in line with the aforementioned competing grammar model, in which the V2 system is giving way to an emerging SVO constituent order35. In (16), once the logical subject do femene and, hence, the core unaccusative verb to die are reanalysed as occurring in-situ, narrow focalisation becomes increasingly associated with the postverbal in-situ position, mining the existing V2 system. From a speculative point of view, presentational sentences may be considered to be the syntactic bridge through which the reanalysis V2-to-SVO took place under the pressure of Standard Italian. The rise of cleft structures must therefore be seen as a strategy to accommodate subjects in informational narrow focus since their occurrence in postverbal position would be marked in the emerging SVO system. Nonetheless, the SVO nature of the language cannot alone explain the widespread use of clefts36, but a conflicting grammar model would. This is why I adopt a V2 analysis for all the instances of informational narrow focus; in addition, we cannot ignore that, in Lamonat, the constituent order VOS and OVS (the latter with a contrastive object) are frequently attested.

4.2.2.4. Subjecthood in Lamonat and Sovramontino

As for Sovramontino, clefts are less widespread, but still robustly attested. Subjects in informational narrow focus can appear in postverbal position, but they can

35 The pressure from Standard Italian (Northern Italian) is allegedly the force driving the V2-to- SVO shift in the NEIDs under investigation. In fact, virtually all interviewed speakers are bilingual in their dialect and in Italian. 36 In fact, we would otherwise expect cleft structures to be very productive in most Romance SVO languages. This is not the case with the exception of French, which might exhibit a similar grammar to that of Lamonat and Sovramontino. Further comparative research is needed to identify the differences and similarities between the NEIDs under investigation and French (see Poletto and Pollock 2004 for a comparative analysis of wh-in-situ) 151 also be clefted as shown in (17) and (18) below:

CONTEXT: Who drank the wine? (17) l a beest MARIO l vin (Sovramontino) 3SG.M.SCL have.3SG drink.PTCP Mario the wine 'Mario drank the wine'

(18) L è MARIO che l a beest l vin

3SG.SCLEXPL be.3SG Mario COMP 3SG.M.SCL have.3SG drink.PTCP the wine ‘It was Mario that drank the wine’

The main difference with Lamonat is that subjects in informational narrow focus do not have to be clefted or resort to object topicalisation in order to be realised. Following from our previous discussion, this relatively free word order signals that Sovramontino is a step behind Lamonat in the development of an emerging SVO rigid constituent order. Nevertheless, if I were to perform a quantitative analysis on my Sovramontino data, it is very likely that it would reveal that cleft structures are more frequent with subjects in informational narrow focus. This is beyond the scope of the present research, but it is an observable pattern. Sovramontino seems to be heading in the same direction as Lamonat, but it is not there yet, being more conservative than Lamonat with respect to its V2 system. In general, in both NEIDs, virtually any lexical XP can be clefted regardless of its grammatical function, with the exception of certain classes (epistemic, deontic and, more generally, adverbs encoding manner – see Cinque 1999) and, of course, inflected verbs. In fact, clefts do not simply encode plain informational narrow focus, but generally convey an interpretation that is exhaustive (see Lambrecht 2001; Kiss 2010)37. In Sovramontino, (17) and (18) may indeed have a difference in interpretation in the sense that (18) conveys an exhaustive reading, whereas (17) does not. Sovramontino speakers claim that when a cleft is used, the speaker is sure of the truth-value of the proposition. In other words, in (18) the only possible person who could have drunk the wine was Mario. In Lamonat, given the ungrammaticality of the VSO constituent order (except in case of object topicalisation),

37 This seems to be a trend and not a linguistic universal. For instance, in French, cleft structures are used to convey both informational and contrastive narrow focus (see Cruschina 2016)

152 this difference seems to be less marked. Nonetheless, clefts in Lamonat also convey a further pragmatic reading, most probably exhaustive, otherwise we would not be able to explain why virtually any XP can be clefted, regardless of its grammatical function.

Having described the interaction between informational narrow focus and argument structure, the next step is to determine why, in Lamonat, focal subjects enjoy special status. I have argued that clefts arise due to the competing V2 and SVO constituent orders. More specifically, I have claimed that clefts as a syntactic strategy are the result of the emergence of a rigid SVO constituent order in which a structural subject position is developing. As mentioned in the previous chapters, Cardinaletti and Repetti (2010) claim that NIDs are non-consistent null-subject languages in which the subject position is indeed structurally salient. This view is compatible with the special treatment of subjects in informational narrow focus. I will now try to explore the notion of subjecthood in relation to focus structure and its consequences in terms of hierarchical structure.

The special behaviour of focal subjects strongly suggests that constituents in informational narrow focus need to move out of their canonical position in order to check a matching [FOCUS] feature in the specifier position of FocusP and receive a focal interpretation. It is reasonable to assume that clefts arise because focalised subjects are forced to move out of their canonical position, but, in Lamonat, this syntactic operation is barred. The fact that focal subjects cannot be realised in FocusP, in the left periphery, can be due to three different syntactic constraints. First, constraint (i): the possibility that, in Lamonat, there exists a ban on particular extraction domains, more specifically on specifiers. In other words, a constituent can only be extracted from complement position: given that the canonical subject position is SpecTP, subjects cannot be extracted from the specifier position of TP and therefore a cleft arises. Second, constraint (ii), Lamonat may ban short distance movement. For instance, in English, the ban on short distance movement is argued to be responsible for the impossibility of topicalising the subject of a sentence (see Lasnik & Saito 1992). A similar mechanism may be in place in Lamonat: the subject cannot undergo short distance movement and therefore a cleft arises. Third, constraint (iii), the subject may not be moved out of its canonical position due to criterial freezing (Rizzi 2006; Rizzi & Shlonsky 2007). By postulating that an [EPP] feature characterises the canonical subject

153 position (see Chomsky 1995), once that feature has been satisfied by merging an overt subject in SpecTP, the subject cannot be moved again and hence a cleft arises to focalise the subject. All these constraints have one common denominator i.e. that the subject position is structurally different from the other argument positions. For instance, it is at the edge of the core of the clause, in specifier rather than complement position, and, finally, the canonical subject position is allegedly determined by the satisfaction of an edge feature, the [EPP]. As for which of the three constraints is responsible for the rise of clefts, the grammaticality of the sentence in (19) below is enough to exclude the constraints in (i) and (iii):

(19) CONTEXT: Did Toni drink the wine? (Lamonat)

No. MARIO l a beù la sgnapa, no Toni NEG Mario 3SG.M.SCL have.3SG drink.PTCP the grappa, NEG Toni 'Mario drank the grappa, not Toni'

In (19), Mario is contrastively focalised: in case of contrastive narrow focus, focus fronting is licit in both Lamonat and Sovramontino. In chapter 5, I will thoroughly discuss the mechanism behind contrastive narrow focus. In short, given the V2 nature of these NIDs, the verb undergoes T-to-C movement in the C-domain and the contrastive element moves to its specifier position, SpecCP. As previously mentioned, in Lamonat, if the subject is in informational narrow focus, it cannot be focus-fronted, as in (19), but the language resorts to two different syntactic strategies: (i) object topicalisation, as in (20), or (ii) a cleft structure as in (21):

CONTEXT: Who drank the grappa? (20) L la a beù MARIO la sgnapa (Lam.) 3SG.M.SCL 3SG.F.OCL have.3SG drink.PTCP Mario the grappa 'Mario drank the grappa'

(21) L è MARIO che l a beù la sgnapa

3SG.SCLEXPL be.3SG Mario COMP 3SG.M.SCL have.3SG drink.PTCP the grappa ‘It was Mario that drank the wine’

Given (20) and (21), typologically speaking, in (19), it would be extremely marked to consider contrastively focalised Mario as occurring in-situ in its canonical position. In

154 the literature on Romance, there is mounting evidence in support of two distinct structural positions for contrastive and informational narrow focus, in which the contrastive focus position is structurally higher than the informational one (see Bentley 2007, 2008; Cruschina 2012, 2016; Rizzi 2018 for further references). It is therefore safe to assume that, in (19), contrastive Mario is ex-situ. Given that it is contrastively focalised out of its canonical position, constraints (i) and (iii) cannot be responsible for the rise of clefts in the context of informational narrow focus. In fact, Mario is moved from its canonical position SpecTP to an even higher projection in the left periphery, namely ForceP. This means that: it is possible to extract the subject from the specifier position of TP, and, once the [EPP] is satisfied by merging the overt subject, the subject can still undergo movement to the higher portion of the sentence, ruling out criterial freezing effects. The only constraint left is the ban on short distance movement, which is compatible with the sentence in (19). In fact, in (19) Mario is contrastively focalised in ForceP, at the edge of the left periphery, in a position structurally higher than FocusP that instead triggers an informational narrow focus interpretation. In other words, the movement of the subject to ForceP is licit, whereas to FocusP is not. The fact that movement to FocusP of the other arguments of the verb is always grammatical, strongly suggests that short distance movement may indeed be problematic in Lamonat. In this respect, the impossibility for the subject in SpecTP to be moved to SpecFocP in turn suggests that the structure of the left periphery of root clauses in the NIDs under investigation is less elaborate than in the average Romance language. I have already briefly mentioned the lack of TopP in the left peripheral space of declarative and interrogative root clauses in Lamonat and Sovramontino; in addition, given the subject ban on short distance movement, it means that FocusP (the projection that assigns informational focus) directly dominates the TP layer. If this were true it would mean that FinP is also missing. This leaves Rizzi’s (1997) periphery impoverished, with only ForceP and FocusP having an active role in the two NEIDs under investigation. In (22) below the structure of the left periphery that I propose for Lamonat and Sovramontino:

(22) [FrameP [ForceP [FocusP [TP]]]]

Note that, in chapter 5, I will discuss the properties of FrameP (Poletto 2002; Benincà & Poletto 2004; Greco and Haegeman 2016), which contains base-generated scene-

155 setting elements, including A-topics (Frascarelli & Hinterhölzl 2007; Bianchi & Frascarelli 2010) and hanging topics (Cinque 1990). As for the clause-internal portion of the left periphery, the data points towards a system in which the left periphery is reduced to two functional projections, namely ForceP and FocusP.

In Sovramontino, the subject ban on short distance movement does not seem to be in place, and the Lamonat sentence in (20) raises the question of why this ban is lifted in case of object topicalisation (with the subject focalised postverbally in FocusP). Surely, the pressure from standard Italian complicates things, as object topicalisation is a typical Italian strategy to focalise the subject of a clause. This however does not affect by any means the structural importance of the aforementioned functional projections, namely ForceP and FocusP in the NEIDs under investigation: they are essential to the encoding of pragmatic information and seem to encompass all the functions generally fulfilled by the left periphery (i.e. modal specification, clause typing, etc. see Rizzi 1997). As for the rise of clefts in Lamonat, they are employed to accommodate subjects in informational narrow focus because the structural subject position itself is undoubtedly becoming more and more salient (regardless of the presence of a subject ban on short distance movement). In fact, Lamonat seems to be a step ahead in the process of shifting towards a SVO constituent order with respect to Sovramontino that is still predominantly V2. In the development of a rigid SVO word order, the co-occurrence of V2-motivated T-to-C movement and the movement of the subject in a lower focal position is problematic, as the subject ends up in postverbal position: a position that, in a SVO language, is strictly an object position.

In this section, I have tried to formalise this by exploring several properties of the subject canonical position in relation to the impossibility of overt subjects to be focalised in FocusP (in the same way as adjuncts and other arguments of the verb). I believe that the exact formalisation of this aspect of language is still subject to debate. Nevertheless, what is crucial is that, in Lamonat, cleft structures containing a focal subject are not only the result of the interaction between focus structure and syntax, but also the result of the complex tripartite interaction among focus structure, the emerging SVO constituent order and the archaic V2 system. In fact, the V2 system in place imposes the realisation of informational narrow focus in postverbal position, whereas the SVO system imposes a relatively rigid constituent order, whereby the subject is in

156 preverbal position and the object in postverbal position. This clash on the realisation of subjects in informational narrow focus is what determines the use of cleft structures as a syntactic strategy to compromise between the two competing constituent orders. Ultimately, this further supports the claim that a V2 system is still in place in Lamonat, generating the aforementioned syntactic quirks in the languages. In sections 4.3 and 4.4, I will deepen the discussion of the V2 properties of Lamonat and Sovramontino.

4.3. The V2 Nature of Lamonat and Sovramontino

I will now step away from the discussion of information structure in Lamonat and Sovramontino and define their V2 nature. I have amply claimed that the two NEIDs under investigation have partially retained V2, more specifically a continuation of V2 in Medieval Romance (see Benincà 1983, 2006; Salvi 2016). As previously claimed, the V2 system in Lamonat and Sovramontino is different from that of Medieval Romance. Lamonat and Sovramontino have presumably changed over the centuries and the V2 system that is now in place seems to be indissolubly linked to discourse-pragmatics; more specifically, the preverbal XP has been reanalysed as a contrastive XP. This means that in case the verb undergoes V2-motivated T-to-C movement, an XP will not move in front of it, unless it bears pragmatic contrastiveness and, hence, a matching [CONTRASTIVE] feature. I will thoroughly discuss contrastiveness as a discourse- pragmatics notion in chapter 5. In this section, I want to address the following: (i) show that Lamonat and Sovramontino qualify as V2 languages (making reference to V2 as appropriate); (ii) define the typology of V2 in Lamonat and Sovramontino (see Wolfe 2016) by investigating which left peripheral projection is targeted by the V2 constraint; and, finally, (iii) compare and contrast Lamonat and Sovramontino with V2 Medieval Romance, diachronically, and with the Ladin V2 variety of Badiotto synchronically (Poletto 2000, 2002).

4.3.1 V2 Diagnostics

In his seminal work for German and Dutch, den Besten (1983) identifies three syntactic aspects that characterise V2 syntax: 1) subject inversion; 2) the linear restriction on preverbal constituents that determines the inflected verb’s V2 position and

157 3) a root vs. embedded asymmetry, whereby the V2 phenomenon is solely found in root clauses. So far, I have shown that, in Lamonat and Sovramontino, narrow focus, be it informational or contrastive, heavily relies on the adjacency with the V2-moved verb, from which it receives sentence stress and, hence, focus prominence. In the next subsections, I will present further arguments in support of the claim that Lamonat and Sovramontino have, synchronically, partially retained the V2 constraint. Their V2 system is tightly linked to the realisation of narrow focus and contrastive elements. I will ultimately break down the basic notion of V2 and its diagnostics, assessing the extent to which they apply to Lamonat and Sovramontino.

4.3.1.1. What Moves? The preverbal XP and the Clitic Status of Auxiliaries

The V2 constraint involves the movement of the inflected verb to the C-domain accompanied by the movement of a pragmatically salient XP in preverbal position. Let us now see what happens in Lamonat and Sovramontino. I have claimed that the Lamonat and Sovramontino V2 system is sensitive to discourse-pragmatics: it is a means to assure the right syntactic-pragmatic configuration to convey contrastiveness and narrow focus. In these syntactic-pragmatic contexts, the verb always moves to the C-domain. An XP, however, can only be licensed in preverbal position, if it bears a contrastive interpretation, as shown in (23) and (24) below:

(23) CONTEXT: When do you go picking up mushrooms? (Lamonat)

a. Vag DE SABO mi a fong go.1SG of Saturday I to mushrooms ‘On Saturdays I go picking up mushrooms’

b. #DE SABO vag mi a fong of Saturday go.1SG I to mushrooms ‘On Saturdays I go picking up mushrooms’

(24) CONTEXT: Do you go picking up mushrooms on Fridays?

a. DE SABO vag mi a fong, no de viendre of Saturday go.1SG I to mushrooms NEG of Friday ‘On Saturdays I go picking up mushrooms, not on Fridays'

158 b. #vag DE SABO mi a fong, no de viendre go.1SG of Saturday I to mushrooms NEG of Friday ‘On Saturdays I go picking up mushrooms, not on Fridays'

In (24), the preverbal position is filled because the preverbal XP, de sabo, is in a contrastive relationship with de viendre. If a contrastive XP is not available, the verb can still move to the C-domain, but the specifier position of the C° targeted by the verb remains empty (the case of informational narrow focus). This a clear difference with Germanic V2 and certain Medieval Romance V2 languages (see Beninicà 2006). Hence, one characteristic feature of the V2 system of these two Bellunese dialects is that the preverbal XP can only be licensed if it encodes a pragmatic contrastive reading. Formally, I argue this is due to the pragmatic specialisation of the [EGDE] feature that, in V2 languages, is responsible for the presence of an overt XP in the specifier position of the C° targeted by the moved verb. In the two NEIDs under investigation, preverbal XPs have been reanalysed as bearing a contrastive reading and the [EDGE] feature has consequently become [CONTRASTIVE], being only active in case a contrastive XP enters the derivation.

In V2 languages, T-to-C movement concerns the inflected verb that lexicalises T°. In Lamonat and Sovramontino, this is also true, as shown in (23) and (24) above, where the inflected verb vag is moved to the C-domain. However, this is not always the case: a key difference between canonical V2 languages and Lamonat and Sovramontino is that what undergoes movement is not only the inflected verb, but the whole verb cluster, as shown in (25a) below:

(25) CONTEXT: When did Mario kill the dog? (Lamonat)

a. L a copà SABO PASAA Mario el can 3SG.M.SCL have.3SG kill.PTCP Saturday past Mario the dog 'Mario killed the dog last Saturday'

By verb cluster, I mean: i) the inflected verb (lexical or auxiliary), ii) the past participle, if present and iii) any satellite clitics that may prosodically rely on the inflected verb for PF realisation (see Pescarini 2016). This is clear in (25a), in which l a copà is moved to the C-domain to satisfy the V2 constraint, as shown in the simplified labelled bracketing representation of (25a) in (25b):

159 b. [ForceP l a copà [FocP DE SABO [TP Mario [T’ [T l a copà] [VP copà [DP l can [PP de sabo]]]]]]]

In chapter 3, in the discussion of T-to-C movement in root interrogatives, I claimed that all the elements belonging to the verb cluster are structurally realised under a complex T°, forming a single phonological unit (in the sense of Frascarelli 2000). By virtue of creating a single phonological unit, the whole verb cluster undergoes T-to-C movement to satisfy the V2 constraints (or Q in the case of root interrogatives). The reason for such heavy T° lies in the morpho-syntactic status of auxiliary verbs in Lamonat and Sovramontino: auxiliaries have clitic status and rely on the adjacency with the past participle (the lexical verb) for their PF realisation. In Lamonat and Sovramontino, hence, V-to-T movement of the lexical verb, be it inflected or in the past participle form, always takes place. In the case of analytic tenses, the past participle must in fact be raised to T° to support the tensed clitic auxiliary which is base-generated in T°. Let us consider the examples in (26) and (27):

(26) a. Mario l magna despes polenta e conicio (Sovramontino) Mario 3SG.M.SCL eat.3SG often polenta and rabbit ‘Mario often eats polenta and rabbit meat’

b. *Mario despes l magna polenta e conicio Mario often 3SG.M.SCL eat.3SG polenta and rabbit ‘Mario often eats polenta and rabbit meat’

(27) a. Mario l a magnà despes polenta e conicio Mario 3SG.M.SCL have.3SG eat.PTCP often polenta and rabbit ‘Mario has often eaten polenta and rabbit meat’

b. *Mario l a despes magnà polenta e conicio Mario 3SG.M.SCL have.3SG often eat.PTCP polenta and rabbit ‘Mario has often eaten polenta and rabbit meat’

Note that Mario does not bear a contrastive reading here. It belongs to a thetic sencence, encoding broad focus, whereby I assume a default in-situ SVO constituent order. In the two NEIDs under investigation, low adverbs (see Cinque 1999) appear postverbally after the main lexical verb. (26a) shows that the inflected verb magna has moved passed the frequency adverb despes which is base-generated in the specifier position of VP, yielding V-to-T movement. Similarly, in (27) the frequency adverb despes appears after

160 the past participle, signalling that the latter must have vacated its base-generated position. Nevertheless, a restricted class of low adverbs can appear between the clitic auxiliary and the past participle as shown in (28):

(28) a. Mario no l a mai magnà polenta e osei (Sovr.) Mario NEG 3SG.M.SCL have.3SG never eat.PTCP polenta and birds ‘Mario has never eaten polenta and bird meat’

b. Mario no l a magnà mai polenta e osei Mario NEG 3SG.M.SCL have.3SG eat.PTCP never polenta and birds “Mario has never eaten polenta and bird meat”

c. Mario l a gia magnà tut Mario 3SG.M.SCL have.3SG already eat.PTCP everything ‘Mario has already eaten everything’

d. Mario l a magnà gia tut Mario 3SG.M.SCL have.3SG eat.PTCP already everything ‘Mario has already eaten everything’

Only this class of low adverbs can appear between the clitic auxiliary and the past participle; this is because they are phonologically reduced and, generally, unstressed. For instance, an adverb like despes (often) cannot be found between the clitic auxiliary and the past participle (cf. 27b) due to its phonological weight, whereas its phonologically reduced counterpart spes (often) can be found in that position, as shown in (29).

(29) a. On spes cantà in cesa (Sovramontino) have.1PL often sing.PTCP in church ‘We often sang at mass’

b. *On despes cantà in cesa have.1PL often sing.PTCP in church ‘We often sang at mass’

c. On cantà despes in cesa have.1PL sing.PTCP often in church ‘We often sang at mass’

The different distribution suggests that those low adverbs that can be placed between the clitic auxiliary and the past participle also have clitic status and can be stacked under

161 the complex T° with the clitic auxiliary, the subject clitic and any other possible satellite clitics. Clitic adverbs cannot be focalised, behaving like weak elements in the sense of Cardinaletti and Starke (1999), and cannot occur as sentence fragment. As for clitic auxiliaries in Lamonat and Sovramontino, they must not be confused with auxiliary clitics across NIDs (Poletto 2000; Pescarini 2016), which accompany auxiliaries, coexist with subject clitics and generally have a fixed form38. Clitic auxiliaries are found cross-linguistically and, generally, they rely on the past participle for their realisation39.

Since the whole verb cluster contained under T° undergoes V2-motivated T-to-C movement, it impedes the availability of a fairly common phenomenon in other V2 languages: stylistic inversion of the past participle with the inflected auxiliary verb (Mailing 1980; Holmberg 2000, 2006). Hence the ungrammaticality of (30):

(30) *Thercà l a Mario la polenta, no magnà (Sovr.) taste.PTCP 3SG.M.SCL have.3SG Mario the polenta NEG taste.PTCP ‘Mario has tested, not eaten the polenta’

(30) is ungrammatical because the auxiliary and the part participle cannot be separated: the former relies on the latter for its realisation and they must be in a proclitic relation. (30) is ungrammatical even if the fronted constituent conveys a contrastive reading and this type of fronting is pragmatically licit : the PF adjacency requirement between clitic auxiliary and past participle rules out this type of construction, lending further support to the claim that the whole verb cluster undergoes T-to-C movement and the auxiliary cannot be excorporated from the past participle. Not only does the ungrammaticality of (30) make Sovramontino and Lamonat depart from “standard” V2 languages, but also it makes them depart from those Romance languages that allow the focalisation of the past participle through focus fronting (see Cruschina and Remberger 2009 for Sicilian and Sardinian).

38 Lamonat and Sovramontino do not have auxiliary clitics. There may be a link between auxiliary clitics and clitic auxiliaries, which is worth investigating; however, this is beyond the scope of the present research. 39 See Bošković (2001) for a detailed account of the behaviour of clitic auxiliaries in some (i.e. Macedonian and Bulgarian). A key difference with Lamonat and Sovramontino is that, in Macedonian and Bulgarian, past participle and clitic auxiliary give rise to enclisis. This is not the case in the NEIDs under investigation, as the two elements are always in a proclitic relation: it may be due to a parametric difference with respect to head directionality. 162 To conclude, Lamonat and Sovramontino differ from canonical V2 languages with respect to the observation that not only does the inflected verb move to C°, but also the whole verb cluster, including the past participle in analytic tenses. The clitic auxiliary and the past participle form a single phonological unit: only clitic adverbs can intervene between the two elements. Finally, another key difference between canonical V2 languages and the NEIDs under investigation is that an XP can only be licensed in preverbal position if it bears a discourse-pragmatic contrastive reading.

4.3.1.2. Subject Inversion

Another characteristic feature of V2 languages is subject inversion with the inflected verb: the inflected verb moves to the C-domain, leaving the subject in its canonical position. In Lamonat and Sovramontino, if a constituent other than the subject bears narrow focus or is assigned a contrastive reading, the subject (when overt) appears postverbally. As explained in the previous subsection, inversion takes place between the subject and the whole verb cluster. Superficially, this type of inversion distinguishes Lamonat and Sovramontino from fully-fledged V2 languages, although the same underlying phenomenon is in place. A non-focal non-contrastive subject (if overt) appears postverbally: (i) in the presence of an XP in informational narrow focus and (ii) in case an element of the clause bears a contrastive reading, as shown respectively in (31) and (32):

(31) CONTEXT: Where did Mario go? (Lamonat) L è ndà AL OSTARIA Mario 3SG.M.SCL be.3SG go.PTCP to-the bar Mario ‘Mario went to the pub’

(32) CONTEXT: Did Mario go to church? AL OSTARIA l è ndà Mario, no in cesa to-the bar 3SG.M.SCL be.3SG go.PTCP Mario NEG in church ‘Mario went to the pub, not to church’

The high position of the verb with respect to the subject is compatible with a V2 system that is sensitive to discourse-pragmatics: in the case of informational narrow focus in (31), the verb moves higher than the focus position, leaving the subject in its clause-

163 internal position; in (32), the verb moves to the same position, but the contrastive focal element is realised in the specifier position of the C° hosting the moved verb, namely Force°.

It is important to note that subject inversion is not always detectable, as the subject is not always overt. In fact, in the aforementioned syntactic-pragmatic contexts, inversion of the subject and the verb cluster takes place if the subject is uttered as given background information (G-topic in the sense of Bianchi & Frascarelli 2010). Although Lamonat and Sovramontino are non-consistent null-subject languages (Cardinaletti & Repetti 2010), they are still null-subject languages and the subject can therefore be omitted. Perhaps because of their hybrid status as NSL, the de-accented subject is often found in canonical subject position (when not pragmatically salient).

4.3.1.3. Linear restrictions: V1 and V3

In root clauses, Lamonat and Sovramontino exhibit V1, V2 and V3 word orders. The occurrence of the three different word orders is mainly constrained by discourse- pragmatics. More specifically, on the type of elements that can appear preverbally and the pragmatic reading that they encode. In Lamonat and Sovramontino, genuine V4 word order is not attested, suggesting that only up to two constituents can appear in preverbal position, as shown by the ungrammaticality of the data in (33):

CONTEXT: What did Mario give to Nane at school? (Lamonat and Sovramontino)

(33) a. *Mario a scola a Nane l ghe a dat AN LIBRO Mario at school to Nane 3SG.M.SCL DAT.CL have.3SG give.PTCP a book ‘Mario gave a book to Nane at school’

b. *A scola a Nane Mario l ghe a dat AN LIBRO at school to Nane Mario 3SG.M.SCL DAT.CL have.3SG give.PTCP a book ‘Mario gave a book to Nane at school’

c. *A Nane Mario a scola l ghe a dat AN LIBRO to Nane Mario at school 3SG.M.SCL DAT.CL have.3SG give.PTCP a book ‘Mario gave a book to Nane at school’

164 As shown in (33), in Lamonat and Sovramontino, the co-occurrence of three preverbal constituents is always banned regardless of their relative orders.

In the two NEIDs under investigation, constituents in preverbal position trigger specific discourse-pragmatic readings, which depend on the syntactic-pragmatic status of the constituents that make up the clause. For instance, in the case of broad focus, I argued that a SVO constituent order is in place, therefore the discourse-pragmatic interpretation of the two potential preverbal constituents is not the same as that of those preverbal constituents found in a sentence where narrow focus or pragmatic contrastiveness triggers V2-verb movement. In any case, in both SVO and V2 underlying word orders, V4 is banned. In this subsection, I will explore the linear restrictions on preverbal constituents in the syntactic-pragmatic contexts in which V2 is triggered, namely with narrow focus and pragmatic contrastiveness. In section 4.6, I will better discuss broad focus and in-situ focalisation.

By not considering V2 a superficial descriptive label (see Ledgeway 2008, Benincà 2006), we can account for V1 and V3 orders in light of the V2 constraint. In Lamonat and Sovramontino, V2 must be understood as a requirement of the verb to rise to a C- position in the left periphery to support a pragmatically salient constituent that is also moved to the C-domain40. In root clauses, I argue that this type of head movement (T- to-C) is still in place with V3 and V1 word orders featuring a constituent in narrow focus or encoding pragmatic contrastiveness. In fact, V2-motivated T-to-C movement determines: (i) which type of preverbal elements, in discourse-pragmatic terms, can give rise to V3, and (ii) justifies the occurrence of V1. In the next chapter, the higher portion of the left peripheral that hosts those preverbal constituents responsible for V3 constituent order will be thoroughly investigated. For now, I will limit myself to the description of the type of elements (in discourse-pragmatic terms) that can appear in preverbal position. Preverbal elements cannot (co-)occur randomly, but must be in a specific hierarchical order according to the pragmatic reading that they bear.

40 This will be better defined and explored in section 4.4. By “support”, I mean assign sentence stress to the pragmatically salient constituent. In this light, I will propose that, in Lamonat and Sovramontino, V2 is a PF phenomenon along the lines of Bošković (2001) and Richards (2016). 165 In a sentence containing a constituent in informational narrow focus and exhibiting V3 constituent order, the left-most constituent can only be a frame-setter, as shown in example (34) below:

(34) CONTEXT: Did you see my father and my mother? Where? (Sovramontino)

Ier to pare l o vist AL MARCÁ mi, yesterday your father 3SG.M.OCL have.1SG see.PTCP at-the market I

to mare no so onde che la era your mother NEG know.1SG where that 3SG.F.SCL be.PST.3SG ‘Yesterday, I saw your father at the market. As for your mother, I did not know where she was’

Frame-setters are circumstantial in nature and modify the propositional content of the main clause: typically, they are adjuncts that provide temporal and/or modal restrictions to the circumstances of evaluation of the proposition expressed by the main clause (see Greco & Haegeman 2016). In her research on V2-Rhaeto-Romance dialects, Poletto (2000, 2001) shows that frame-setters are characterised by a scene-setting function, which anchors the speech act in terms of locative and temporal deixis, and speech participants (see also Benincà & Poletto 2004; Poletto 2002). In (34), ier, ‘yesterday’, functions as a frame-setter. Frame-setters are clause external, base-generated and generally responsible for V3 constituent orders (Haegeman 2000; Holmberg 2012). For instance, in Old Venetian, diachronically a Force-V2 language, most V3 orders were determined by frame-setters (Wolfe 2016). In this respect, Lamonat and Sovramontino seem to obey the linear restriction on V3 commonly found across V2 languages.

As far as the V2 constituent order itself is concerned, it generally occurs when a pragmatically salient constituent is fronted in the absence of a frame-setter. As previously mentioned, the preverbal constituent must carry a matching [CONTRASTIVE] feature: it can be either a contrastive topic like to pare, ‘your father’ (in bold), in (34) or a contrastive focal element like to pare, ‘your father’ (in BOLD CAPITALS), in (35) below:

(35) CONTEXT: Did you see my mother at the market? (Lamonat) No. TO PARE è vedù al marcà, no to mare NEG your father have.1SG see.PTCP at-the market NEG your mother ‘I saw your father, not your mother at the market’

166 The important difference that sets apart contrastive topics and contrastive foci is that a contrastive topic can co-occur with an informational focal element, whereas a contrastive focal element cannot. Even though informational narrow focus and contrastive narrow focus occupy two distinct positions in the left periphery, they cannot co-occur in the same clause: only one element per clause can receive narrow focus41 (Rizzi 2018). V3 and V2 constituent orders are sensitive to the type of element placed at the left of the moved verb. For the sake of completeness, another frequent V2 constituent order is determined by V1 preceded by a frame-setter, in this case no contrastive element appears in preverbal position.

In Lamonat and Sovramontino, V1 word order is found when a clause features an XP in informational narrow focus and a frame-setter and/or a contrastive topic is absent. A constituent in contrastive narrow focus, on the other hand, always generates a V2 constituent order (in the absence of a frame-setter) and is therefore incompatible with V1 (cf 35). In the presence of an XP in informational narrow focus, V1 emerges because the verb cluster moves to a left peripheral position that is higher than that occupied by the focalised element. The sentence in (36) exemplifies V1 in Lamonat and Sovramontino:

(36) CONTEXT: To whom did Mario give the book? (Sovramontino)

L ghe l-a dat A GIANNI Mario l libro 3SG.M.SCL DAT.CL 3SG.M.OCL-have.3SG give.PTCP to Gianni Mario the book ‘Mario gave Gianni the book’

To conclude, in Lamonat and Sovramontino, V1 and V3 constituent orders are possible and fully accountable for within the discourse-pragmatic sensitive V2 nature of the NEIDs under investigation. By the same token, V4 is always ungrammatical and V3 is always characterised by a left-most frame-setting element which is clause external and base-generated in that position.

41 See Poletto (2002), Poletto & Benincà (2004) for an account in which left peripheral focus can consist of more than one constituent; nevertheless, these constituents must be adjacent to one another (see also Bentley 2008) and make up a focus field. 167 4.3.1.4. The Root vs. Embedded Asymmetry

Before delving into the discussion of root vs. embedded V2 in Lamonat and Sovramontino, it is important to note that not all Germanic V2 languages display a root vs. embedded asymmetry: and Icelandic, for instance, are generalised V2 languages in which the V2 constituent order is also possible in embedded contexts. Diachronically, the root vs. embedded asymmetry is not found in (Fontana 1993), but it was in place in and Medieval NIDs (Benincà 1984). In the two NEIDs under investigation, there is a clear root vs. embedded asymmetry, whereby pragmatically motivated V2 is only found in root clauses.

In embedded clauses, Lamonat and Sovramontino exhibit a dominant SVO constituent order as shown by (37). The VOS word order which is commonly found in root clauses when O is in informational narrow focus is ungrammatical in embedded contexts, as shown in (37b):

(37) CONTEXT: What did Maria tell you that Nani killed? (Sovramontino) a. La me a dit che 3SG.F.SCL 1SG.DAT.CL have.3SG say.PTCP that

Nani l a copà l SO GAT Nani 3SG.M.SCL have.3SG kill.PTCP the her cat ‘She told me that Nani killed her cat’ b.*La me a dit che 3SG.F.SCL 1SG.DAT.CL have.3SG say.PTCP that

l a copà L SO GAT Nani 3SG.M.SCL have.3SG kill.PTCP the her cat Nani ‘She told me that Nani killed her cat’

The ungrammaticality of (37b) becomes clear, if we consider V2 as requirement on a left peripheral functional head that contains a strong feature and for that reason must be lexicalised (see Ledgeway 2008; Poletto 2002). In embedded contexts such a requirement is already satisfied by the presence of the complementiser that checks through external merge the strong feature on the relevant left peripheral head. There is hence no need for the verb cluster to move to C°. By the same token, T-to-C movement cannot take place as the landing position of the verb cluster is already filled by the

168 complementiser. I argue that the V2 requirement is present in both root and embedded clauses, but in the latter case it is immediately satisfied by the presence of the complementiser itself. This claim is supported by the observation that, in Lamonat and Sovramontino, in embedded interrogatives, the wh-complementiser is accompanied by the canonical complementiser che, ‘that’, so that C° is lexicalised, as in (38) below:

(38) No se sa che te ö (Lamonat) NEG know.1SG what that 2SG.SCL want.2SG “I don’t know what you want”

In this respect, in the periphery of the embedded clause, the V2 constraint is fully satisfied: the verb cannot move to the C-domain, determining a SVO constituent order. Interestingly, in the embedded clause selected by bridge verbs, including verba dicendi, it is possible to find a type V2 structure with an element in narrow focus as shown by (39):

(39) CONTEXT: speakers talking about people who have lost weight recently (Sovr.)

La me a dit che l ghe a 3SG.F.SCL 1SG.DAT.CL have.3SG say.PTCP that 3SG.M.SCL DAT.CL have.3SG dit MASSIMO che l è dimagrì sie kili say.PTCP Massimo that 3SG.M.SCL be.3SG lose-weight.PTCP six kilos “She told me that he said that Massimo has lost six kilos”

Embedded clauses like that in (39) are at times found in the NEIDs under investigation, suggesting that, in the clause selected by the bridge verb dir, ‘say’, the verb cluster manages to rise to the C-domain and co-occur with the embedding complementiser. In these embedded contexts, in the presence of a selecting bridge verb, the SVO constituent order is generally dominant, but V2 is still possible. This is not unheard of across V2 languages: in fact, bridge verbs may allow embedded V2 in German, mainland Scandinavian (Vikner 1995) and in the Rhaeto-Romance variety of Val Badia (Poletto 2000); Old Neapolitan could also allow V2 in those embedded clauses selected by a bridge verb (Ledgeway 2008).

A fine-grained investigation of the CP layer selected by bridge verbs is beyond the purpose of the present research. Nevertheless, it seems that bridge verbs are able to select a more articulated CP layer which resembles that of root clauses. For instance,

169 Poletto (2000) shows that, in Badiotto, the embedded periphery selected by bridge verbs seems to tolerate scene-settings adverbs and hanging topics. These elements are generally banned in embedded contexts (unless they are contrastively focalised), being mainly a feature of root clauses (see Bianchi & Frascarelli 2010). The properties of bridge verbs are definitely worth investigating, especially from a cross-linguistic perspective42 with respect to the geometry of the left periphery; nevertheless, I will not pursue the discussion here.

To conclude, it is important to note that: (i) in Lamonat and Sovramontino, there is a clear root vs. embedded asymmetry when it comes to the realisation of V2 in relation to information structure, and (ii) this asymmetry is commonly found across V2 languages and is compatible with the view that V2 requires the lexicalisation of a strong feature on a left peripheral functional head.

4.3.2. The Typology of V2: High vs. Low V2 Languages

In the last two decades, since Rizzi’s (1997) split CP model, researchers have tried to adopt a new perspective in the study of the V2 phenomenon. In fact, in Rizzi’s (1997) left periphery, the V-to-C movement account no longer holds, as the CP layer is conceived of as a number of separate functional projections hosting different elements and checking different semantic-pragmatic features. There seems to be general agreement that the main left peripheral projections involved in the satisfaction of the V2 constraint are FinP and ForceP. The difference between V2 languages and non-V2 languages therefore lies in the requirement of satisfying either a [FORCE] or a [FIN] feature on the relevant left peripheral head.

Diachronic research on Medieval Romance (Benincà 1983; 2006) has shown that both types of V2 languages were attested. Wolfe (2016) suggests that the left peripheral projection in which the V2 constraint was satisfied varied between languages: this variation indeed concerns the ability for either ForceP or FinP to be the locus of V2. Among the Force-V2 or High-V2 languages we find Old Spanish, Old French and Old

42 Spanish embedded clauses allow for recomplementation only if introduced by a bridge verb, more specifically a verbum dicendi (see Villa-García 2015). This is further cross-linguistic evidence that bridge verbs are able to select a more articulated CP-layer. 170 Venetian, whereas Fin-V2 or Low-V2 systems include Old Sicilian, and Old Neapolitan (see Benincà 2004, 2006; Ledgeway 2003, 2008; Salvi 2012; Poletto 2014). In Germanic Fin-V2 languages, it is important to note that the linear restriction on the V2 constituent order is explained by postulating that the movement of other constituents in the higher portion of the left periphery is blocked, giving rise to the so- called bottleneck effect (Haegeman 1996, 2012; Roberts 2004; Cardinaletti 2010 a. o.). The ban on any type of movement to the higher portion of the CP-layer must be interpreted as a consequence of relativized minimality (Rizzi 1990, 2001a): in Fin-V2 systems, in fact, A’-movement and head movement take place to satisfy the V2 constraint blocking any other homologous type of movement. In Force-V2 systems, the postulation of the bottleneck effect is not crucial as the periphery lacks a portion higher than ForceP, except FrameP which, however, as previously pointed out, is associated with base-generation (Poletto 2000, 2002; Benincà & Poletto 2004).

4.3.2.1. Lamonat and Sovramontino: A High-V2 System

In chapter 3, I claimed that the difference in the distribution of tonic and clitic wh- elements between Friulian/Fornese and Sovramontino/Lamonat is determined by microparametric variation in the locus of SCLI. In Lamonat and Sovramontino, T-to-C movement in root interrogatives is higher than in Friulian and Fornese, generating apparent wh-in-situ. Let us consider the Fornese and Sovramontino data in (40) and (41):

(40) a. Dunlà i-lu dzût Mario? (Fornese) where be.3SG-3SG.M.SCL go.PTCP Mario ‘Where did Mario go?’

b. La-i-lu dzût Mario? where-be.3SG-3SG.M.SCL go.PTCP Mario ‘Where did Mario go?’

(41) a. E-la ndaa aonde Maria? (Sovramontino) be.3SG-3SG.F.SCL go.PTCP where Mary ‘Where did Mary go?’

b. Ond-e-la ndaa Maria? where-be.3SG-3SG.F.SCL go.PTCP Mary ‘Where did Mary go?’

171 In Fornese (40b) and Sovramontino (41b), clitic-wh-elements exhibit the same constituent order in preverbal position attaching as prefixes to the inflected verb, whereas tonic wh-elements appear preverbally in Fornese (cf. 40a) and postverbally in Sovramontino (cf. 41a). As discussed in chapter 3, in root questions the verb cluster moves to the C-domain to attach to the null affix Q which is responsible for the well- formedness of questions. The difference in constituent order between (40a) and (41a) reflects the difference in the C-projection that hosts Q and therefore is targeted by T-to- C movement: ForceP in Lamonat/Sovramontino and FinP in Friulian/Fornese. This becomes clearer by considering the distribution of tonic and clitic wh-elements: clitic- wh-elments attach to the inflected verb and move with the whole verb cluster to the C- domain, yielding an apparent identical surface constituent order in Fornese (40b) and Sovramontino (41b), as shown in the labelled bracketing representation in (42b) and (43b). Constrastingly, by virtue of their [WH] and [FOCUS] feature (see Bianchi 1999), tonic wh-elements undergo movement to the specifier position of FocusP, which is higher than FinP, but lower than ForceP, yielding the difference in constituent order between (40a) and (41a), as shown by the labelled bracketing representation in (42a) and (43a) below.

(42) a. [ForceP [IntP/FocP dunlà [FinP Q-i-lu dzût [TP Mario [T’ [T al ì dzût] [VP dzût dunlà]]]]]]

b. [ForceP [IntP/FocP [FinP la-i-lu dzût [TP Mario [T’ [T al ì dzût] [VP dzût ]]]]]]

(43) a. [ForceP Q-e-la ndaa [IntP/FocP aonde [FinP [TP Maria [T’ [T la è ndaa] [VP ndaa aonde]]]]]]

b. [ForceP ond-e-la ndaa [IntP/FocP [FinP [TP Maria [T’ [T la è ndaa] [VP ndaa]]]]]]

I have claimed that, in Lamonat and Sovramontino, T-to-Force° movement in root interrogatives is the continuation of the Medieval Romance V2 constraint (see Benincà 1983; 2006). Across Romance, this is not surprising as verb movement in root interrogatives, and especially across NEIDs, is generally considered as residual V2 movement (see Rizzi 1996, 2001b; Salvi 2016). In this respect, Lamonat and Sovramontino were once fully-fledged high-V2 systems similarly to their main contact language diachronically, Old Venetian (see Wolfe 2016 and references therein), and similarly to the contemporary V2-Ladin dialects spoken in the neighbouring region

172 Trentino- (Poletto 2000, 2002). The pragmatically motivated V2 system in Lamonat and Sovramontino has conserved generalised T-to-Force° movement in root interrogatives and in root declaratives that feature a pragmatically salient constituent (an XP that encodes narrow focus and/or bears a contrastive reading). As discussed in chapter 3, verb movement to Force° in root interrogatives is also reflected in the lack of SCLI in negative-interrogatives. Let us consider examples (44) and (45):

(44) a. No la a beù Maria al vin? (Lamonat) NEG 3SG.F.SCL have.3SG drink.PTCP Mary the wine ‘Didn’t Maria drink the wine?’

b. *No a-la beù Maria al vin? NEG have.3SG-3SG.F.SCL drink.PTCP Mario the wine ‘Didn’t Maria drink the wine?’

(45) a. No aj-al mangjât Mario? (Friulian) NEG have.3SG-3SG.M.SCL eat.PTCP Mario ‘Didn’t Mario eat?’

b. *No al a mangjât Mario? NEG 3SG.M.SCL have.3SG eat.PTCP Mario ‘Didn’t Mario eat?’

The micro-parametric variation in the locus of SCLI manifests itself in the opposite treatment of SCLI in this syntactic context. In Lamonat, SCLI is banned, whereas, in Friulian, SCLI is obligatory, as shown in (44) and (45). In Lamonat, given that the verb moves as high as ForceP to satisfy Q, SCLI is blocked by the negation, which checks its [NEG] feature in PolP before reaching Force°. Q is instead satisfied by the adjacency with the negative element. This does not happen in Friulian, as SCLI takes place in FinP where Q is hosted: a lower projection with respect to PolarityP. In Friulian, the negation cannot host Q, as it still has to check its features in PolP before spell-out. This is shown in the labelled bracketing representation of (44a) and (45a) in (46) and (47) respectively:

(46) [ForceP Q-no la a beù [PolP no[NEG] la a beù [FinP [TP Maria [T’ [T la a beù] [VP beù al vin]]]]]]

(47) [ForceP [PolP No [NEG] Q-aj-al mangjât [FinP Q-aj-al mangjât [TP Mario [T’ [T al a

mangjât] [VP mangjât]]]]]]

173 In Lamonat and Sovramontino, T-to-Force° movement is also reflected in the linear restrictions on the type of constituents that can appear preverbally. V1, V2 or V3 are possible constituent orders in root interrogatives and in root declaratives that feature a pragmatically salient constituent. As far as root interrogatives are concerned, there are mainly two main types of elements that can appear in preverbal position: (i) a scene- setting element that anchors the speech act in terms of locative, temporal deixis and speech participants (Haegeman 2000, 2006, 2007; Benincà & Poletto 2004; Poletto 2002) and d-linked wh-elements (see Pesetsky 1987). Scene setting elements include: A-Topics (Bianchi & Frascarelli 2010), hanging topics (Cinque 1990; Benincà 2001) and circumstantial frame-setters (see Greco & Haegeman 2016)43. They are associated with base-generation and generally set off from the rest of the sentence by a pause. They surface in clause-external position, higher than ForceP where the interrogative or declarative force of the sentence is encoded (Rizzi 1997). As in (48), the presence of scene-setting elements higher than the locus of SCLI is hence compatible with a Force- V2 system:

(48) De so sor, t-a-la contà? (Lamonat) of her sister 2SG.DAT.CL-have.3SG-3SG.F.SCL tell.PTCP ‘About her sister, has she told you about it?

In (48), de so sor (about her sister) is an A-topic and the question must be interpreted in light of the clause external constituent. Conversely to referential-topics, aboutness- topics cannot be omitted without compromising the discourse-pragmatic felicity of the utterance (see Cruschina 2016). In root interrogatives, V3 constituent order is also possible in the case a d-linked wh-element appears between a scene-setting element and the moved cluster verb as in (49):

(49) L an pasà, pede chi a-tu laorà? (Lamonat) the year past with who have.2SG-2SG.SCL work.PTCP ‘Last year, with whom did you work?’

43 A-topics, hanging topics and frame-setters exhibit different morpho-syntactic characteristics. By no means should they be considered to be identical; nonetheless, they belong to the same discourse-pragmatic category, namely scene-setting elements. As I will argue in chapter 5, they are generally limited to one occurrence and occupy the same left peripheral A’-bar position, namely SpecFrameP (see Haegeman 2000, 2007; Benincà & Poletto 2004; Poletto 2002) 174 The prepositional-wh-element, pede chi, in preverbal position is only pragmatically felicitous if it encodes a d-linked interpretation (Pesetsky 1987), assuming the meaning “which of these people or which of the people you previously mentioned” (see chapter 5). Consequently, in (49), pede chi is in a spec-head relation with the verb cluster, with which it checks its [CONTRASTIVE] feature.

In root interrogatives, the satisfaction of the V2 constraint and the satisfaction of Q are undoubtedly related diachronically, but must be regarded as two distinct syntactic properties in contemporary Sovramontino and Lamonat. First of all they serve two different functions: the V2 constraint is crucial for the realisation of contrastive elements (in the case of root-interrogatives, d-linked wh-elements) serving thus a specific discourse-pragmatic reading, whereas the satisfaction of Q is a syntactic requirement necessary for the typing of the force of the clause as interrogative. The satisfaction of Q, nonetheless, also satisfies V2 through the lexicalisation of Force°; in this respect, V2 is always satisfied in root interrogatives. In chapter 3, I supported the argument that SCLI alone through the Q affix is enough to satisfy the wh-criterion (Rizzi 1991; 1996). As discussed in chapter 3, this involves the lexicalisation of the specifier position through a null wh-operator, following Poletto’s (1993) spec-head dynamic agreement with respect to SCLI across NEIDs. A legitimate question arises: if SpecForceP is occupied by the null-wh-operator, how can d-linked wh-elements occupy the same position? The most straightforward answer to this question lies in the proposal that, since d-linked wh-elements are equipped with a [WH] feature, they can replace the wh-operator in the syntactic computation of the sentence, functioning themselves as wh- operators. In fact, it is not case that XPs bearing a contrastive reading are barred in root interrogatives, except d-linked wh-elements. As I will discuss in chapter 5, the contrastive nature of d-linked wh-elements [CONTRASTIVE] along with their wh- nature [WH] license them in preverbal position adjacent to the moved verb cluster, namely SpecForceP.

As for root interrogatives, I have briefly reiterated the theoretical and empirical arguments in favour of a High or Force-V2 system in Lamonat and Sovramontino whereby pragmatically motivated T-to-C movement targets Force°. As far as root declaratives are concerned, pragmatically motivated V2 T-to-C movement takes place in those clauses featuring a pragmatically salient constituent, namely an XP bearing a

175 contrastive reading and/or narrow focus. We can therefore say that whereas T-to-C movement takes place virtually in all root interrogatives, the V2 property is limited to a subset of root declaratives. Let us outline how this type of T-to-C movement is compatible with a High or Force-V2 system. First, the aforementioned root interrogative linear restrictions on V1, V2 and V3 constituent orders are also in place in root declaratives encoding narrow informational focus, as shown in (50) below:

(50) Magne POLENTA mi de sabo (Sovramontino) eat.1SG polenta I on Saturday ‘I eat polenta on Saturdays’

(50) is the canonical way of answering the question “What do you eat on Saturdays?” where the object polenta is in informational narrow focus and mi and de sabo de- accented background information. I have argued so far that in case of declarative sentences featuring a constituent in narrow focus, the constituent undergoes focus movement to the left periphery, namely to FocusP, and the verb cluster moves to a higher left peripheral projection, ForceP: adjacency between the verb cluster and the element in narrow focus is hence achieved, giving rise to VOS or VXS constituent orders. In (51), the labelled bracketing representation of (50) that summarises my proposal:

(51) [ForceP magne [FocP polenta [TP mi [T’ [T magne] [VP magne polenta de sabo]]]]]

As is the case with root interrogatives, the number and type of constituents allowed preverbally is generally two: a scene-setting element and/or a contrastive element, as shown by the grammaticality of (52a) and ungrammaticality of (52b)

(52) a. De sabo, mi magne POLENTA mi on Saturday, I eat.1SG polenta I ‘On Saturdays, I eat polenta’

b. *De sabo, co-l me om, mi magne POLENTA mi on Saturday with-the my husband I eat.1SG polenta I ‘On Saturdays, with my husband, I eat polenta’

176 In (52a), the object polenta encodes informational narrow focus and the verb cluster is adjacent on its left. In preverbal position there is the scene-setting element de sabo and the contrastive topic mi. No other topics can appear in preverbal positions as shown by the ungrammaticality of (52b). The status of de sabo as a scene-setting element is quite straightforward, whereas I want to spend a few words on the status of mi as a contrastive topic. Mi cannot be a contrastive focal element for two reasons: (i) only one constituent is allowed in narrow focus per utterance in Lamonat and Sovramontino (see Rizzi 2018 for the uniqueness of focus), (ii) the mi can be resumed by an identical de-accented copy in the TP-layer. As for the second point, a contrastive narrow focal element cannot be resumed by a proform be it clitic, tonic or an epithet. Resumption of a contrastive topicalised subject through a tonic pronoun is relatively common in Lamonat and Sovramontino: the resumptive pronoun is always de-accented and can be omitted without altering the grammaticality of the sentence. Sentences like (52a) in which the de-accented pronominal subject strictly appears on the right of the focal element, but on le left of any other possible background topical XPs give us some crucial information about sentence structure in Lamonat and Sovramontino: the de- accented pronoun signals the boundary between the C-domain and the T-domain, highlighting left peripheral focus A’-movement and pragmatically motivated T-to-Force movement.

V2 and V3 linear restrictions in place in root interrogatives are hence also in place in the subset of root declaratives that feature T-to-Force movement: only a scene-setting element and/ or a contrastive XP (focal or topical) can appear preverbally, yielding a V2 mechanism compatible with a Force or High-V2 system. A possible counterargument to this claim is the aforementioned bottleneck effect (Haegeman 1996, 2012; Roberts 2004; Cardinaletti 2010 a. o.). Assuming a split-CP model, the bottleneck effect postulates that V2 T-to-C movement in conjunction with A’-movement to SpecCP blocks all types of XP movement to the higher portion of the left periphery as a consequence of relativized minimality. In this respect, the only elements that can appear in preverbal V2 position can only be base-generated regardless of whether the T-to-C movement targets Fin° or Force°, challenging de facto the relevance of V2 and V3 linear restrictions. In the case of Lamonat and Sovramontino, the bottleneck effect does not pose a problem for the view that these two NEIDs are pragmatically motivated Force-V2 (and not Fin-V2) systems. For instance, given the data scrutinised so far, the V2 constraint in place

177 cannot target Fin° as we would not be able to explain why the verb crosses the focal elements hosted in FocusP. It is, in fact, the collapsing nature of the V2 system in place in Lamonat and Sovramontino that unveils its Force-V2 nature: having its preverbal position specialised for contrastive elements, non-contrastive pragmatically salient elements (informational narrow focal elements) must be realised elsewhere in the C- domain. This is indeed achieved in FocP position, exploiting, at the same time, the left adjacency with the verb cluster, which ensures that the focal element attains focal prominence. The contrastive XP nature of the preverbal position in Lamonat and Sovramontino’s V2 system is what sharply distinguishes the partial V2 system in place in these two NEIDs from most other fully-fledged V2 systems. In this model, it is not surprising that regular topics (in the sense of Rizzi 1997) or G-topics (in the sense of Bianchi and Frascarelli 2010) encoding background information cannot appear in left- peripheral position, but are realised in the TP-layer. In fact, not only would the presence of a lexicalised TopP impede the adjacency between the moved verb cluster (in ForceP) and the element in narrow focus in (FocusP), but would also block V2 T-to-Force movement as Top° is not a suitable host for head movement (see Rizzi 1997).

Further evidence in favour of a High-V2 system comes from root declaratives in which the positive or negative polarity of the sentence44 is emphasised through the focalisation of an agreeing polarity item, as shown in (53):

CONTEXT: Was John paid properly? (Sovramontino) (53) a. L a ciapà SI’ Nane i so schei 3SG.M.SCL have.3SG take.PTCP yes John the.PL his money ‘Yes, John did get the money (he deserved)’

b. No-l a ciapà NO Nane i so schei NEG-3SG.M.SCL have.3SG take.PTCP NEG John the.PL his money ‘No, John did not get the money (he deserved)’

Note that, in (53b), since the negation in postverbal position emphasises the negative polarity of the utterance, the preverbal negation is indeed for the grammaticality of the sentence. In (53a) instead sì, ‘yes’, emphasises the default positive polarity of the utterance. Similarly to tonic wh-items and XPs in informational narrow focus, focalised

44 To some extent, this construction can be regarded as an instance of verum focus (see Leonetti & Escandell-Vidal 2009). It is not the inflected verb that is focalised, but the lexicalised polarity value of the utterance. 178 ‘yes/no’ appears in postverbal position followed by the subject Nane and the direct object i schei. Given the striking similarity with the distribution elements in informational narrow focus, I argue that the positive and negative polarity markers indeed attain their empathic reading through focal prominence that is achieved through the adjacency with the verb cluster. The negative or positive polarity marker is generated in PolarityP (see Laka 1990) and then moves to FocusP to check its [FOCUS] feature. The verb cluster, in an identical fashion to tonic wh-elements and XPs in informational focus, appears at the left of the polarity marker in the higher portion of the C-domain. T-to-C movement is hence pragmatically motivated to assign to the adjacent negative polarity item focus prominence and hence sentential stress. Given the constituent order in (53), the verb lands higher than FocusP and PolP, ruling out a Fin- V2 system.

The licensing of negative polarity items (or NPIs – see Ladusaw 1979) can also help us uncover the Force-V2 nature of the pragmatically motivated V2 system in Lamonat and Sovramontino. Consider the following data:

CONTEXT: With whom did Bepa speak? (Lamonat)

(54) a. No-l-a parlà CO GNESSUN Bepa NEG-3SG.F.SCL-have.3SG speak.PTCP with noboby Bepa ‘Bepa did not speak to anybody’

b. *L-a parlà CO GNESSUN Bepa 3SG.F.SCL-have.3SG speak.PTCP with noboby Bepa ‘Bepa did not speak to anybody’

CONTEXT: Bepa told Maria the secret, I can’t believe it!

(55) a. CO GNESSUN no-l-a parlà Bepa with noboby NEG-3SG.F.SCL-have.3SG speak.PTCP Bepa ‘Bepa did not speak to anybody’

b. *CO GNESSUN l-a parlà Bepa with noboby 3SG.F.SCL-have.3SG speak.PTCP Bepa ‘Bepa did not speak to anybody’

The NPI co gnesun is in informational narrow focus in (54a) and in contrastive narrow focus in (55a): in (54a) it appears to the right of the verb cluster and in (55a) to the left of the verb cluster. In both instances, the focal NPI relies on the adjacent verb cluster for its realisation. Only contrastive elements can be licensed in preverbal V2 position,

179 namely SpecForceP, as in (55a); non-contrastive narrow focalised constituents, on the other hand, are realised in FocusP. Note, however, how in Lamonat and Sovramontino the fronting of a contrastive focal NPI requires the preverbal licensing negator no. Standardly, across Romance, postverbal NPIs need to be licensed by a c-commanding licensor like the preverbal negative marker no; however, if the NPI appears preverbally, it does not co-occur with the preverbal negation (see Belletti 1999). The ungrammaticality of (55b) shows that, in Lamonat and Sovramontino, the preverbal negation is nonetheless needed when an NPI bearing contrastive focus is fronted. The data in (55) indicate that, at some point in the derivation, the preverbal negator licensing the NPI was in a c-commanding position. The ungrammaticality of (55b) is also interesting with respect to defining the status of the preverbal negation in these two NIDs. The preverbal negation seems to heavily rely on T° for its realisation, forming one phonological unit with it and its satellite clitics. In this respect, it behaves like a clitic negative element (see Zanuttini 1997). If we consider negation as an affix that requires adjacency with a PF host for its realisation (a là Bošković 2001 et seq.), we could say that, in Lamonat and Sovramontino, the preverbal negation privileges T° as a hosts. In fact, it does not seem to be able to attach to the NPI surfacing in preverbal position. By the same token, we could argue that the negation attaches to T° before the NPI bearing a contrastive reading is fronted. Contrastive narrow focus fronting can hence be seen as a last resort strategy that crucially involves PF linearization. The properties of the preverbal negation in Lamonat and Sovramontino will not be discussed any further, as they are beyond the purpose of the present research.

In this section, I have recapped the empirical and theoretical arguments in support of the High or Force-V2 nature of Lamonat and Sovramontino. T-to-C movement takes place in all root interrogatives and in a subset of root declaratives that feature a contrastive element and/or an XP bearing narrow focus. Let us now compare Lamonat and Sovramontino’s V2 system with that of Badiotto synchronically and with that of Medieval Romance diachronically.

4.3.3. Comparing and Contrasting Romance V2 in Synchrony and Diachrony

In this section, I will compare and contrast Lamonat and Sovramontino’s pragmatically V2 system with: (i) the V2 system of Medieval Romance (Benincà 1984, 180 2006) and (ii), synchronically, with the V2 system of the Ladin variety of Badiotto (Poletto 2000, 2001). The common trait between Lamonat and Sovramontino’s V2 system and those of Medieval Romance and Badiotto is the strong link between the V2 constraint and the packaging of information: most importantly, the V2 constraint seems to be sensitive to the satisfaction of focus in contexts of narrow focus. The main difference between Lamonat and Sovramontino’s V2 and Medieval Romance and Badiotto’s V2 lies in the fact that the former is a collapsing V2 system, in which only contrastive XPs can occupy the left-peripheral V2 preverbal position. Let us start with Medieval Romance.

In her investigation of Old French and Medieval Northern Italian dialects, Benincà (2006; 1984) notes that the Romance languages spoken in the Medieval period share important features in terms of lexical and functional properties relating to sentence structure. These common traits identify what I will refer to as Medieval Romance. Nonetheless, Medieval Romance languages cannot be said to form a unitary language system (see for example Benincà 2006, Ledgeway 2008). Many authors have argued that Medieval Romance, hence Medieval Romance languages, exhibited the V2 constraint (Vanelli 1986; Fontana 1993, 1997; Roberts 1993; Benincà 1995, 2006; Salvi 2004 a. o.). In V2 Medieval Romance, the preverbal position was not a privileged subject position, but a pragmatically salient structural position that licensed topical or focal interpretations (Vanelli 1986, Salvi 2004, Benincà 2006). Benincà (2006) points out that an XP in preverbal position could assume a variety of discourse-pragmatic readings: contrastive or informational focus including wh-words, as well as different types of topic, namely anaphoric, referential and aboutness topics. In Medieval Romance, this portion of the sentence was more easily activated than in most modern Romance languages. The main similarity between Medieval Romance and the two NEIDs under investigation is that the V2 preverbal position serves discourse- pragmatics: this operator position is licensed by V-to-C movement to the left periphery. Lamonat and Sovramontino are however much more selective with regard to what can be realised in preverbal V2 position: only an XP encoding some degree of discourse- pragmatic contrastiveness can be licensed in such a position (see chapter 5). Differently from Medieval Romance, Lamonat and Sovramontino do not allow any other type of pragmatically salient XP in preverbal position: in a diachronic perspective, their V2 system has profoundly changed. This is also true of all the contemporary Romance

181 languages that have maintained the V2 constraint. For instance, some Rhaeto-Romance varieties are still V2, but have no residue of a defining feature of Medieval Romance, namely its clitic syntax (Salvi 2016). Medieval Romance clitic syntax relates to the V2 constraint and is the manifestation of the Tobler-Mussafia law that states that enclisis of a clitic element is obligatory when the verb is in first position as in (56) (see Poletto 2013):

(56) Legge-si di Salamone che… (Novellino 138, r.1, Poletto 2013: 161) read.3SG-IMP.CL of Salomone that ‘One reads about Somolone that…’

It relates to V2 in the sense that proclisis is only possible when V2 is underlyingly satisfied, otherwise enclisis takes place (Benincà 1995). If the verb in C° is preceded by a focal element, enclisis is impossible, whereas if it is preceded by a topical or a scene- setting element, enclisis becomes obligatory45. In this regard, Benincà (1995, 2006) claims that the lower focus field (in the sense of Benincà and Poletto 2004) is the only portion of the left periphery that can host moved elements and structurally satisfy V2, bearing in mind that enclisis is triggered by V-to-C movement. Medieval Romance languages also varied with respect to which left peripheral projection, FinP or ForceP, was targeted by V2 motivated T-to-C movement (see Wolfe 2016) and the requirement on preposing or not an XP preverbally (see Benincà 2006; Ledgeway 2008). Lamonat and Sovramontino have lost the V2 clitic syntax that characterised Medieval Romance, exhibit a pragmatically motivated Force-V2 system and the preposing of an XP is not obligatory: the V2 constraint can solely be satisfied through the lexicalisation of Force° via T-to-C movement. Nevertheless, T-to-C movement is always associated with the lexicalisation of a pragmatically salient constituent in the left periphery: a contrastive focal element, a contrastive topical element or an XP in narrow informational focus. The tight link between the realisation of focal elements and the satisfaction of V2 in Medieval Romance further suggests that, in Lamonat and Sovramontino, verb movement and its adjacency requirement with an XP in narrow focus are diachronically motivated.

45 A property that has been, for instance, maintained in Asturian (Fernandez-Rubiera 2009) 182 In synchrony, both typologically and geographically speaking, Lamonat and Sovramontino’s closest V2 Romance variety is Ladin. Not all Ladin varieties exhibit the V2 constraint (see Salvi 2016), the most well-known V2 Ladin varieties are Gaderano and Gardenese (see Poletto 2000, 2002; Salvi 2016). The variety of Ladin spoken in the Badia valley, namely Badiotto, has received particular scholarly attention (Valentin 1998; Poletto 2000, 2002; Salvi 2001). I will now compare Lamonat and Sovramontino’s V2 system with that of Badiotto, pointing out differences and similarities in the satisfaction of the V2 constraint. At the morpho-syntactic level, the satisfaction of V2 in Badiotto triggers SCLI as shown in (57):

(57) Gonoot vas-t a ciasa sua (Badiotto, Poletto 2002: 8) often go.2SG-2SG.SCL to house his ‘Often you go to his place’

Fronting the XP gonoot triggers SCLI. SCLI is generally associated with verb movement, signalling that the inflected verb has undergone T-to-C movement (see chapter 3). SCLI does not take place in all root declaratives, but only in those root clauses that feature a fronted pragmatically salient constituent. This means that either the V2 constraint is not at work in all root declaratives or T-to-C movement alone is not enough to trigger SCLI. I will not try to answer this question for Badiotto and point the reader to the relevant aforementioned literature. Nonetheless, in case T-to-C movement were to take place only to accommodate a pragmatically salient constituent (and hence signalled by SCLI), Badiotto’s V2 system would be identical to that of Lamonat and Sovramontino, in which V2 motivated T-to-C movement only takes place in a subset of root declaratives, which feature an XP that is pragmatically salient. Note that Badiotto’s SCLI in V2 root declaratives must not be confused with Medieval Romance V2 clitic syntax: the former type of enclisis only concerns SCLs and obligatory takes place with a preposed XP, whereas the latter type of enclisis concerns any clitic element and takes place solely if no XP is fronted. Morpho-syntactically, Lamonat and Sovramontino V2 differs from V2 in Badiotto with respect to SCLI, as no SCLI takes place in Lamonat and Sovramontino in V2 root declaratives. In the absence of diachronic evidence, it is hazardous to claim that SCLI has been lost in this syntactic context; nonetheless, synchronic and cross-dialectal evidence points in that direction. For instance, SCLI was once widespread across NIDs, but has been slowly receding for the last two centuries including in root interrogatives due to the gradual loss of T-to-C movement (Benincà,

183 Parry & Pescarini 2016). In Lamonat and Sovramontino, SCLI is limited to root interrogatives (except negative-interrogatives) and a subset of root declaratives that feature the imperfect subjunctive. As discussed in chapters 2 and 3, SCLI in the NEIDs under investigation is not solely linked to verb movement, but associated with the morphological rearrangement caused by an extra piece of morphology that enters the derivation and needs adjacency to the inflected verb (namely, Q in questions and S for the irrealis subjunctive mood). In the case of V2 T-to-C movement in root declaratives, SCLI does not take place in Lamonat and Sovramontino, but the verb cluster moves with all its satellite clitics in proclitic position. As argued so far, lack of SCLI does not necessarily translate into lack of verb movement. The lack of SCLI may in fact depend on other factors. It may be linked to the status of SCLs in the NEIDs under investigation as becoming more and more similar to agreement markers (see Poletto 1993, 2000) than pronominal weak elements (Cardinaletti & Starke 1999). It may also be related to the reinterpretation of SCLI as involving a further layer of syntactic-semantic meaning, such as encoding the force of root questions or the irrealis mood in root declaratives. V2 motivated T-to-C movement does not provide any further information with respect to the force of the clause or the mood of the verb, but has the function of giving focal prominence to the focal element and/or ensuring that contrastive elements are correctly interpreted in their appropriate syntactic configuration.

In Lamonat and Sovramontino, the only elements that can satisfy the V2 linear constraint are XPs that bear a contrastive reading and are hence equipped with a matching [CONTRASTIVE] feature. In the V2 Rhaeto-Romance variety of the Badia valley, Poletto (2002) claims that the only set of elements that can satisfy the V2 constraint is restricted to focal constituents, scene-setting adverbs and wh-elements. She also points out that hanging topics and left dislocated XPs cannot enter a V2 structure. The difference in what can enter a V2 construction between the NEIDs under investigation and Badiotto is quite straightforward: in Lamonat and Sovramontino, focal constituents and wh-elements can linearly satisfy V2 only if they also bear a contrastive reading46. Also, scene-setting adverbs cannot enter a V2 structure, but tend to give rise to V3 linear orders, unless, of course, they are contrastively focalised. Similarly to Lamonat and Sovramontino, hanging topics cannot satisfy V2, but generally are

46 Wh-elements must be D-linked. See chapter 5 on contrastiveness in Lamonat and Sovramontino.

184 responsible for V3 constituent orders. Left dislocation is also banned from appearing in a V2 structure in Badiotto. As far as this last point is concerned, the NEIDs under investigation and Badiotto are fully in line. In Lamonat and Sovramontino, non- contrastive non-aboutness topics are realised in-situ in their canonical position in the T- domain: they cannot appear in left peripheral position. Rizzi (1997) points out that left dislocation is a barrier for head movement, as Top° is not a suitable host for T movement. In Badiotto and in the NEIDs under investigation, the lack of left dislocated elements in the left peripheral space is a direct consequence of the V2 nature of these Romance varieties. In fact, as Poletto (2000) points out for Badiotto, in the three varieties, the left peripheral projection involved in the satisfaction of the V2 requirement is Force° and not Fin°. The verb must rise to Force° and the lower Top°, if projected, would block such movement. This is another common trait between Lamonat and Sovramontino’s and Badiotto’s V2 structure: they are High or Force-V2 systems. A question arises: if Badiotto is a High-V2 system on a par with Lamonat and Sovramontino, why does not it exhibit apparent wh-in-situ? The answer lies in the nature of the V2 system in these languages. While Badiotto has maintained a fully- fledged V2 system, the V2 system in Lamonat and Sovramontino is a “collapsing one”, in which only contrastive elements can enter a V2 structure. Tonic wh-elements are focal in nature, but crucially not contrastive (unless they bear a d-linked interpretation); they can therefore rise to FocusP, but not any further in the structure as they are not equipped with a [CONTRASTIVE] feature. This mechanism gives rise to wh-in-situ in Lamonat and Sovramontino, whereas in Badiotto the V2 preverbal position is not specialised and the focal element can hence move further in the structure to ForceP to satisfy the V2 constraint. Poletto (2000, 2002) argues that the position targeted by focal elements is indeed SpecForceP. However, she argues that, before getting to its final position, the focal element moves through SpecFocP where it checks its [FOCUS] feature. In principle, I support Poletto’s analysis and believe that the same model should be applied to account for XPs in contrastive narrow focus in Lamonat and Sovramontino, which must check not only a [CONTRASTIVE] feature in ForceP, but also a [FOCUS] feature in FocusP. However, in section 4.4, in order to avoid criterial freezing effects, rather than checking a proper [FOCUS] feature, I will propose the satisfaction of a prosodic marker F hosted in Focus°, which ensures that focal prominence falls on the XP in narrow focus through the adjacency with the verb.

185 Finally, in Lamonat and Sovramontino, the [CONTRASTIVE] specialisation of the V2 preverbal position is what gives rise to the aforementioned marked constituent orders and what ultimately differentiates the NEIDs under investigation from Badiotto. For the sake of completeness, it is worth mentioning that Poletto (2000, 2002) points out that, in Badiotto, there is a difference in the treatment of the V2 constraint and the geometry of the periphery between root interrogatives and root declaratives. Poletto and Pollock (2004) argue that there is a mismatch between the structural position that types interrogative and declarative force, with interrogative force located much lower in the structure; consequently, different linear restrictions are in place between root interrogatives and root declaratives. In Lamonat and Sovramontino, this mismatch is not detected and the model of left periphery proposed here suits well both interrogative and declarative root clauses. From a merely speculative point of view, the difference between interrogative and declarative clauses may be the result of the interaction between the satisfaction of the V2 constraint and the satisfaction of Q in questions. I will not pursue this issue further and leave a fine-grained comparison between the two NEIDs under investigation and Badiotto for future research.

To conclude, I have dicussed the common traits between the V2 system of the NEIDs under investigation and (i) that of Medieval Romance, diachronically, and (ii) that of Badiotto, synchronically. The main difference lies in the collapsing nature of the V2 system in Lamonat and Sovramontino, in which the V2 preverbal position has specialised to host contrastive XPs only. Nonetheless, in the two NEIDs, T-to-C movement has been conserved in association with the fronting of a pragmatically salient constituent. Assuming that Old Lamonat and Old Sovramontino worked the same as Medieval Romance, on a vertical dimension, these diachronic syntactic changes led to the marked constituent orders that characterise the NEIDs under investigation. On the horizontal dimension, the comparison with Badiotto, is particularly telling because it helps us change our perception of the V2 phenomenon. Synchronically, Romance V2 must not be defined by the sole T-to-C movement accompanied by the fronting of any XP, but, rather, it is the fronting of a pragmatically salient XP that dictates T-to-C movement: a mechanism that provides the right structural configuration to pragmatically salient constituents. In this respect, V2 can be seen as a phenomenon pertaining to the syntax-pragmatics interface.

186 4.3.4. Informational Narrow Focus: Against an In-situ Analysis in Lamonat and Sovramontino

In this section, I will discuss why an in-situ analysis of XPs in informational narrow focus a là Romance (see Cruschina 2016) is not suitable for the Lamonat and Sovramontino data and we have to resort to a pragmatically motivated V2 system. Most Romance languages privilege the in-situ position for the realisation of informational narrow focus47, fronting only XPs bearing contrastive narrow focus if this operation is not barred by the grammar of the language48 (see Cruschina 2012, 2016). Could this be the case of Lamonat and Sovramontino? Do we really need the aforementioned V2 machinery to account for XPs in informational narrow focus? An alternative analysis may involve the XP focalisation in-situ and the right dislocation of the following XPs. In this section, I will show that this is not the case and the postulation of a pragmatically motivated V2 system remains a better option to account for the syntactic behaviour of the two NEIDs under investigation.

Let us consider the data in (58), (59) and (60):

(58) G-a-li dat CHE i dotor a Simon? (Lam.) DAT.CL-have.3PL-3PL.M.SCL give.PTCP what the doctors to Simone ‘What did the doctors give to Simone?’

(59) CONTEXT: Who did you greet? (Sovramontino)

O saludhà MARIO mi have.1SG greet.PTCP Mario I ‘I greeted Mario’

47 Two well-known exceptions, namely Romance varieties that can license informational narrow focus in the left peripheral space, are Sardinian and Sicilian (see Jones 1988; Bentley 2008; Cruschina & Remberger 2009; Cruschina 2012). In section 4.5, I will further discuss the properties of Sicilian and Sardinian fronted XPs bearing informational narrow focus. Note that in these languages, informational narrow focus can also occur in-situ: the surfacing position widely depends on the discourse-pragmatic characteristics of the focal XP. In section 4.6, I will show that, in Lamonat and Sovramontino, the in-situ option is the default option in case of broad focus (Lambrecht 1994). 48 In several Romance languages focal constituents bearing a contrastive reading can also appear in-situ (i.e. Italian). Some authors even propose that the unmarked position for contrastive focal elements is the in-situ position (see Samek-Lodovici 2015)

187 (60) CONTEXT: To whom did Toni give the money?

L ghe (li) a dat A MARIO Toni i schei 3SG.M.SCL DAT.CL (3PL.M.OCL) have.3SG give.PTCP to Mario Toni the money ‘Toni gave the money to Mario’

The three sentences exhibit: (i) an element in informational narrow focus, (ii) the verb cluster in first position and (iii) one or more G-topics (Bianchi & Frascarelli 2010) encoding background given information. In our proposal, the verb moves to Force°, namely the projection targeted by V2 T-to-C movement, the focal element checks its [FOCUS] feature in FocusP and the topical background information is realised in-situ within TP. So far, I have provided empirical and theoretical evidence in support of this proposal; nevertheless, would not it be more straightforward to claim that in (59) the XP in informational narrow focus, Mario, is realised in-situ and the G-topic, mi, is right dislocated? The answer is ‘no’, as there are various arguments against an in-situ analysis of Lamonat and Sovramontino’s focal elements. Let us start by ruling out topic right dislocation. By topic right dislocation, I mean placing one or more bona fide topics on the right of the clause. In Romance, this operation normally involves an agreeing clitic in the T-domain, resuming the dislocated element (see Cruschina 2016), hence labelled clitic right dislocation (see Cinque 1990; Lopez 2009). In Lamonat and Sovramontino, the resumptive clitic is non-consistently present in this context, as shown by the optionality49 of the object clitic li in (60). Note that SCLs must not be confused with resumptive clitics: SCLs are in fact obligatory in the two NEIDs under investigation (see chapter 2). Also, in (58), the dative clitic ge does not properly function as a resumptive element, but, in the two NEIDs under investigation, its occurrence is obligatory in association with a beneficiary XP 50.

Clitic right dislocation is generally set off from the rest of the clause by an audible pause. This is not the case in Lamonat and Sovramontino: the topical constituents are not set off by a pause, but instead characterised by a falling intonation, as shown by the pitch contour of (58) and (60) in (61) and (62) respectively:

49 See Cardinaletti (2001) on the status of the resumptive clitic in clitic right dislocation in Italian. She points out a key difference between clitic right dislocation and marginalisation, whereby the latter features an in-situ destressed constituent. Even though an in-depth analysis of the phenomenon is beyond the purpose of the present work, I want to suggest that in-situ G- topics in Lamonat and Sovramontino behave indeed like marginalised XPs in Italian. 50 Clitic doubling whereby the dative clitic obligatorily doubles the beneficiary is common across NIDs (see Benincà, Parry & Pescarini 2016). 188 (61) 400 ) z H (

h c t i G-a-li P dat ke i dotor a Simon G-a-lidatkei dotora Simon

0 5.298 7.739 Time (s) Figure 1. Pitch contour of the root interrogative in (58) featuring an apparently in-situ wh-element.

(62) 400 ) z H (

h c t i P El ge a dat a Mario Toni i skei

0 1.116 3.611 Time (s) Figure 2. Pitch contour of the root declarative in (60) featuring an XP in informational narrow focus.

Bona fide topics (Rizzi 1997) are recursive and can be stacked in a topic field (a là Benincà & Poletto 2004). This is not possible in Lamonat and Sovramontino. Besides the finite number of topical constituents that can appear at the right of the XP in informational narrow focus, their order must be hierarchical; if uttered, the subject must

189 be the first element at the right of the focal XP, as shown by the ungrammaticality of (63) and (64):

(63)*G-a-li dat CHE a Simon i dotor? (Lam.) DAT.CL-have.3PL-3PL.M.SCL give.PTCP what to Simone the doctors ‘What did the doctors give to Simone?’

CONTEXT: To whom did Toni give the money? (Sovramontino) (64)*L ghe (li) a dat A MARIO i schei Toni 3SG.M.SCL DAT.CL (3PL.M.OCL) have.3SG give.PTCP to Mario the money Toni ‘Toni gave the money to Mario’

Instead, topical XPs appearing in a topic field do not need to be in hierarchical order. In Lamonat and Sovramontino, topical elements show instead a rigid constituent order, suggesting that they do not form a topic field at the right of the clause, but are de- accented elements in-situ in their canonical position. Finally, bona fide topics are banned from the left of the clause in the two NEIDs under investigation: only A-topics and C-topics (Bianchi & Frascarelli 2010) are licensed in the left peripheral space of the clause. G-topics hence appear in their canonical position, where they are de-stressed and characterised by a falling intonation that makes them background given information.

Another argument against an in-situ analysis of XPs in informational narrow focus is that any constituent encoding informational narrow focus follows the verb cluster, hence not only XPs carrying the object function. The constituent order can be summarised as V focal-XP, where XP can be a direct or indirect object, an oblique argument, an adjunct and, in Sovramontino, even a subject. If constituents in informational narrow focus were realised in-situ, we would not expect any adjacency requirement with the verb. It is hence quite clear that a different strategy is in place, which, I argue, does not involve the TP-layer, but the CP-layer. It is nonetheless true that an alternative in-situ analysis is possible by adopting Belletti’s (2001, 2004) VP periphery. 51 It would involve placing the element in informational narrow focus in the periphery of the VP. Belletti argues for the existence of a low periphery in the T-domain that mirrors that of the C-domain. Placing XPs encoding informational narrow focus in FocusP of the periphery of the VP would however be still problematic. First of all, we

51 Note that Caterina Bonan is currently pursuing this analysis for wh-in-situ in Trevigiano. She is a PhD candidate at the University of Geneva. To the best of my knowledge, her research is still in progress. 190 would need a mechanism of right dislocation to account for (58), (59) and (60). We have seen that right dislocation is not compatible with the data from the two NEIDs. It would actually be desirable to have the topical elements also realised in the periphery of the VP; however, given that topical constituents always follow the focal XP, adopting this account would result in a typologically problematic periphery in which FocP is higher than TopP. A further problem with the adoption of the periphery of the VP to account for the Lamonat and Sovramontino data is that it would lead to a heavy violation of relativized minimality (see Rizzi 1990, 2001a), which Belletti herself points out as a crucial aspect of her model. Finally, as far as tonic wh-items are concerned, if we were to adopt Beletti’s model, it would be difficult to explain the linear adjacency with the verb cluster, given that SCLI signals that the verb has moved to the C° domain and the focal elements would remain in-situ within TP.

For these reasons, I believe that, in order to account for informational narrow focus in Lamonat and Sovramontino, the pragmatically motivated V2 analysis remains a better option. As far as Belletti’s (2001, 2004) VP periphery is concerned, it is undoubtedly a useful syntactic analysis to account for some T-domain related phenomena in Italian; however, I have provided some empirical arguments to show that that it is not a suitable analysis of informational narrow focus, including wh-in-situ, in the two NEIDs under investigation.

4.4. The Pragmatically Motivated V2 Mechanism

In this section I will try to formalise what I have argued so far: a syntactic model that also considers the pragmatic and phonological nature of the Lamonat and Sovramontino V2 system. T-to-C movement, in fact, enforces the right syntactic- pragmatic configuration, but also ensures that the element in narrow focus receives focal prominence through the adjacency with the verb cluster that carries sentential stress.

In general terms, V2 can be described as a complex movement operation in which the finite verb raises to the C-domain in root clauses and, in turn, a pragmatically salient constituent is fronted to a preverbal position within the C-domain. In Lamonat and

191 Sovramontino’s V2 system, we have seen that the two movement operations do not always occur together: T-to-C movement is obligatory, whereas the fronting of a salient constituent may not take place, yielding V1 word orders. An XP moves preverbally only if it bears a contrastive reading as in (65):

CONTEXT: Is Mario in his room? (Lamonat)

(65) DO NTE LA CANEVA l è ndat Mario down in the cellar 3SG.M.SCL be.3SG go.PTCP Mario ‘Mario went downstairs to the cellar, not to his room’

Differently from canonical V2 languages, Lamonat and Sovramontino exhibit a specialised contrastive preverbal left-peripheral position. V2 T-to-C movement also takes place with the movement of an informational narrow focal XP in the C-domain at the right of the moved verb cluster, as in (66):

CONTEXT: Where did Mario go?

(66) L è ndat DO NTE LA CANEVA Mario 3SG.M.SCL be.3SG go.PTCP down in the cellar Mario ‘Mario went downstairs to the cellar’

In devising a formal model that captures Lamonat and Sovramontino’s pragmatically motivated V2 system, we have to take into account these three different movement operations.

Let us start with T-to-C movement of the verb cluster. I have argued that it is an instance of head movement in which all the lexical material contained under T° moves to the left periphery, targeting Force°. The issue is to define what triggers T-to-C movement and in which syntactic context. T-to-C movement occurs in root declaratives that feature a constituent in informational narrow focus and/or bearing a contrastive reading. The common denominator between contrastive XPs and XPs in narrow focus is that they are realised in the C-domain. I therefore propose that the V2 constraint is activated every time the left peripheral space is projected and lexicalised. As I will discuss in section 4.6, this mechanism would explain why root declaratives in broad focus do not exhibit V2 T-to-C movement and is compatible with the behaviour of embedded clauses in which the V2 constraint is activated and satisfied by the

192 complementiser. Note however that the lexicalisation of FrameP, namely the projection that hosts frame-setters and A-topics, does not trigger the activation of the V2 constraint. Frame-setting elements can in fact co-occur with thetic sentences encoding broad focus without triggering V2 T-to-C movement. As Haegeman (2000, 2007) points out, frame-setters are clause-external and hence cannot be assimilated into the V2 structure. A-topics and frame-setters are independent of the force of the clause and can be considered as self-contained speech acts (Krifka 2001). In this regard, FrameP is part of the left peripheral space, but not part of the C-domain. In (67), I illustrate the proposed language-specific make-up of the left periphery of Lamonat and Sovramontino, showing the boundaries of the Frame, C and T domains:

(67) FrameP

ForceP

FocusP

TP

…

With respect to Rizzi’s (1997) left periphery, the peripheral space of Lamonat and Sovramontino is reduced. Most noticeably, (i) TopP and FinP do not seem to play a role in the left peripheral computation of the clause, and (ii) frame-setting elements are independent of the C-domain. As previously argued, the absence of TopP for bona fide topics (in the sense of Rizzi 1997) is due to the unsuitable nature of Top° for accommodating T-to-C movement to Force°. As far as FinP is concerned, Lamonat and Sovramontino seem to rely on ForceP to fulfil the functions that are generally assigned to FinP52. In chapter 3, I pointed out how IntP (Rizzi 2001b, 2013) is also not involved in the satisfaction of the [WH] feature in Lamonat and Sovramontino. Following a cartographic approach, I do not exclude that further functional heads can be projected in the left peripheral space of Lamonat and Sovramontino, such as Rizzi’s (2004) ModP or Cinque’s (1999) specialised adverb positions. Nevertheless, in order to account for the data collected in the present research, we do not need a particularly elaborated left

52 See the discussion on SCLI and the irrealis operator with the imperfect subjunctive in chapter 3. 193 peripheral geometry: it seems clear that some of the aforementioned functional projections are not part of the make-up of the Lamonat and Sovramontino CP-space. Hence, to resort to the full inventory of left-peripheral projections seems unmotivated.

The V2 constraint is activated when the C-domain is projected and lexicalised: an XP appears in the specifier position of either FocusP or ForceP, bearing a focal and/or a contrastive reading. In these contexts, T-to-C movement takes place to satisfy the V2 constraint on Force°. Haegeman (1996) and Roberts (1999), in their claim that V2 is a property of the lowest head of the left periphery, namely Fin°, argue that verb movement to Fin° is a last resort strategy to check a strong [FIN] feature. The difference between a V2 language and a non-V2 language can therefore be conceived of as a further requirement in the C-domain of checking a [FORCE] or [FIN] feature in the head and specifier position of the relevant C-projection. The feature checked by the V2 constraint is one of agreement, which is achieved through a spec-head relationship (Tommaselli 1990; Vikner 1995), depending on the Fin or Force-V2 nature of the system in place. In the NEIDs under investigation, the feature checking requirement is on Force° and the relevant feature is [FORCE]. As far as the preverbal V2 position is concerned, in Lamonat and Sovramontino’s collapsing V2 system, discourse-pragmatics has become paramount in the V2 structure and the [EDGE] feature responsible for the lexicalisation of SpecForceP has pragmatically specialised, becoming a [CONTRASTIVE] feature and hence hosting only XPs bearing a contrastive reading. The V2 constraint has therefore become a sole requirement on Force° that must be lexicalised when the C-domain is activated and projected. V2 should then not be seen simply as verb raising, but as a requirement of feature checking on a left peripheral head before spell-out through which it is assigned a lexical value (see also Poletto 2002 and Ledgeway 2008). In Lamonat and Sovramontino, the spec-head requirement has been lost and the satisfaction of the V2 constraint is achieved through T-to-C movement.53

Now that I have established which syntactic feature is responsible for attracting the verb to Force° and the syntactic context in which this mechanism takes place, we need to account for the different types of A’-movement that accompany V2 T-to-C movement. As far as contrastive elements are concerned, they surface in SpecForceP in

53 Note that, in embedded clauses, the V2 constraint is still in place, but is instantly satisfied by the external merge of the complementiser. In this regard, V2 can be satisfied both through internal and external merge. 194 the pragmatically specialised preverbal V2 position that characterises Lamonat and Sovramontino’s pragmatically motivated V2 system. In the model proposed here, I postulate that when contrastive XPs enter the derivation of the clause they are equipped with a [CONTRASTIVE] feature that must be checked against the [CONTRASTIVE] feature hosted in ForceP in a spec-head configuration. The model, however, must be able to distinguish between contrastive focal elements and contrastive topical elements. In chapter 5, I will better discuss the notion of contrastiveness and try to set apart topical from focal contrastiveness. I have so far claimed that contrastive elements, be they focal or topical in nature, occupy the same position in the left periphery. We need hence to distinguish between contrastive foci, as in (65) above, and contrastive topics, as in (68) below:

CONTEXT: My crop field was spared by the hail last Thursday, yours? (Sovr.) (68) Doba, l me camp la lo a cresemà Thursday the my field 3SG.F.SCL 3SG.M.OCL have.3SG confirm.PTCP

la tempesta l me camp the hail the my field ‘Last Thursday, the hail devastated my crop field’

L me camp, ‘my field’, is in a contrastive relationship with ‘the field’ of the interlocutor, which was instead spared by the adverse meteorological conditions. It is a topical element and, being the object of the clause, is resumed by an agreeing object clitic (Cinque 1990). Prosodically, the sentence is characterised by the aforementioned falling intonation on the background topical elements: la tempesta, ‘the hail’, and the repetition of l me camp, ‘my field’. The last two elements can in fact be omitted. Note, however, that, very often, contrastive topical elements are also doubled by an identical copy54 in- situ in the T-domain along with the mandatory resumptive pronominal clitic. The lower copy of the contrastive topical element seems to suggest that, conversely to contrastive focal elements, contrastive topical elements are externally merged directly in SpecForceP where their [CONTRASTIVE] feature is immediately satisfied. In the on- going debate on whether contrastive topics are moved or base-generated in the left periphery, I can say that the data from Lamont and Sovramontino seem to support the latter option.

54 The use of this term, namely copy, does not necessarily imply a movement operation. I mean an identical co-referential XP located elsewhere in the clause structure. 195 As for contrastive topics, the two functional features [CONTRASTIVE] and [TOPIC] are split and satisfied in different domains of the clause: the [CONTRASTIVE] feature is satisfied by the XP that is directly merged in SpecForceP, whereas the [TOPIC] feature is satisfied in-situ within TP by the identical lower copy of the XP. The lower in-situ copy behaves like any other bona fide topical element (a G- topic) and hence can also be omitted (by receiving null-spell out). It is hence the compositionality of the [CONTRASTIVE] and [TOPIC] feature that gives rise to the contrastive topical interpretation. In (69), the derivation of (68) as in our proposed model so far:

(69) FrameP

Doba ForceP

l me campi Force' [CONTRASTIVE]

la lo a cresemà TP [FORCE]

la tempesta T'

la lo a cresemà VP

Vº DP cresemà el me campi [TOPIC]

As for contrastive XPs in narrow focus, it is plausible that a similar mechanism of compositionality is in place with the [CONTRASTIVE] and [FOCUS] features. The [CONTRASTIVE] feature is satisfied through internal merge of the focal element in SpecForceP, but what about the [FOCUS] feature? Before delving into the formalisation of the behaviour of XPs in contrastive narrow focus, we hence need to provide a formal account for the syntactic behaviour of XPs in informational narrow focus. Let us consider the sentence in (66), which, for convenience, is repeated in (70) below:

CONTEXT: Where did Mario go? (Lamonat) (70) L è ndat DO NTE LA CANEVA Mario 3SG.M.SCL be.3SG go.PTCP down in the cellar Mario ‘Mario went downstairs to the cellar’

196 The focal element, do nte la caneva, carries the [FOCUS] feature that is satisfied in the left peripheral space in SpecFocP: the focal XP moves to the C-domain, which, by virtue of being activated and projected, triggers T-to-C movement of the verb cluster, l è ndat. Our proposal must account for the adjacency requirement between the verb cluster and the focal element. I have previously claimed that besides the discourse-pragmatic specialisation on the prefinite V2 structural position, pragmatically motivated V2 T-to- C movement has become a mechanism to ensure that the element in narrow focus receives focal prominence through the adjacency with the verb cluster that carries sentential stress. This is visible in the F0 representations that I have previously shown in (61) and (62), in which the pitch contour of the moved verb cluster and the following focal XP consists of a high pitch plateaux-like intonation. In the NEIDs under investigation, thus the pragmatically motivated V2 system does not only concern the syntax-pragmatics interface, but also the syntax-phonology interface. Topical constituents are marked by a de-accented falling tone, whereas focal XPs are marked with a high-pitch tone. Many scholars highlight the link between tonal events and information structure (see Truckenbrodt 1999; Rooth 1996; Frascarelli 2000; Bianchi & Frascarelli 2010 a.o.). Most interestingly, some scholars have also argued in favour of a PF analysis of the V2 phenomenon. For instance, Bošković (2001) argues that Germanic V2 is a PF requirement whereby a phonologically overt element must be realised in front of the verb right adjacent to an I-phase boundary. Along the same lines, Richards (2016) argues that the V2 phenomenon should be interpreted as a stress assignment requirement at the word level: T-to-C movement occurs because the head C is affixal and needs a host, whereas the moved XP in the specifier position satisfies the aforementioned prosodic requirement.

We certainly cannot ignore the strong link between the phenomena illustrated in this chapter and their phonological realisation. Across Romance, phrasal stress and tonal events do play a key role in the expression of focus and in marking, for example, the distinction between broad and narrow focus (Zubizarreta 1998; D’Imperio 2002; Donati & Nespor 2003; Bocci & Avesani 2005; Frota et al. 2007). The adjacency requirement between fronted focal element and inflected verb is, nonetheless, controversial in Romance, but cross-linguistically shows a solid trend (Cruschina 2016)55.

55 Romanian and Spanish seems to be the two main Romance languages that exhibit the most rigid adjacency requirements (Cornilescu 2002; Zubizarreta 1998). 197 Lamonat and Sovramontino’s pragmatically motivated V2 system is primarily a syntax-pragmatics interface phenomenon. Nonetheless, there is a prosodic component to it that closely interacts with the expression of focus and the assignment of focal prominence on the relevant XP through the adjacency with the verb cluster. For this reason, this aspect of the syntax-phonology interface must be taken into account in our formal model. Drawing on Richards’ (2016) idea that head movement is primarily driven by the need for affix support, whereby affixes are attached to hosts, I propose a syntactic representation of the null prosodic operator that ensures the assignment of focal prominence on the focal XP and justifies the verb-focal element adjacency requirement. I will name this null focal operator F: it is base-generated in Foc°, if FocP c-commands a lower XP that carries a [FOCUS] feature, F attracts the verb cluster to Foc° and the [FOCUS] feature is checked by moving the lower focal XP to SpecFocP. Note that in our model the [FOCUS] feature is only assigned to elements in narrow focus, as broad focus is not associated with movement (see section 4.6). The verb cluster that has undergone head movement to Foc° is then moved to Force° to satisfy the V2 [FORCE] feature. The verb cluster moves to Force° along with the null prosodic operator F, yielding right adjacency with the focal element in SpecFocP and ensuring the attainment of focal prominence. The proposed model is shown in (71) which is the arboreal representation of (70):

(71) ForceP

Force'

l è ndat=F FocusP [FORCE]

do nte la caneva Focus' [FOCUS]

l è ndat=F TP

Mario T'

l è ndat VP

Vº PP ndat do nte la caneva [FOCUS]

198 Now that we have a formal model to account for XPs in informational narrow focus, we can expand our model and capture the behaviour of XP in contrastive narrow focus. Let us consider (65), which, for convenience, is repeated in (72) below:

CONTEXT: Is Mario in his room? (Lamonat) (72) DO NTE LA CANEVA l è ndat Mario down in the cellar 3SG.M.SCL be.3SG go.PTCP Mario ‘Mario went downstairs to the cellar, not to his room’

The XP in contrastive narrow focus, do nte la caneva, checks its [CONTRASTIVE] feature in SpecForceP; however, by virtue of being an XP in narrow focus, do nte la caneva also needs to check its [FOCUS] feature. I argue that the checking of [FOCUS] also takes place in SpecForceP. In fact, the verb cluster adjoins to F in Focus° and drags F along on its way to satisfy the V2 [FORCE] feature in Force°. The element in narrow focus and the verb cluster plus F hence find themselves in a spec-head configuration that makes the checking of the [FOCUS] feature possible. Adopting this model along with the focal prosodic operator F avoids criterial freezing problems: for example, in Poletto’s (2002) account of Badiotto’s V2 system, XPs in narrow focus first check their [FOCUS] feature in SpecFocP and then move again to SpecForceP to satisfy the V2 constraint. In this regard, the proposed analysis avoids the requirement that the focal element moves twice violating simultaneously criterial freezing and greed. The proposed formal account for the behaviour of XPs in contrastive narrow focus is shown in (73), which is the arboreal representation of (72):

(73) ForceP

do nte la caneva Force' [FOCUS] [CONTRASTIVE] l è ndat=F FocusP [FORCE]

l è ndat=F TP

Mario T'

l è ndat VP

Vº PP ndat do nte la caneva [FOCUS] [CONTRASTIVE]

199 The proposed model allows us formally to distinguish contrastive foci from contrastive topics. The adoption of a null prosodic operator F at the syntactic level also acknowledges the well-known prosodic differences in the realisation of contrastive focal elements, contrastive topical elements and informational focal elements (Frascarelli 2000; Bianchia & Frascarelli 2010).

For the sake of completeness, let us briefly see how this model would apply to root wh-interrogatives and embedded clauses. As for root wh-interrogatives featuring a tonic wh-element, the same mechanism as in root declaratives with an element in narrow focus would take place. The key difference is that SCLI would take place due to the morphological rearrangement caused by the sequence Q + verb cluster (see chapter 3). The final landing position of the tonic wh-elements depends on its discourse- pragmatic nature: if it is d-linked as in (74) the landing site is SpecForceP, whereas if it is not d-linked, as in (75), the landing site is the canonical focus position, namely SpecFocusP:

(74) Che majon a-tu fat ti? (Sovramontino) which jumper have.2SG-2SG.SCL do.PTCP you ‘Which jumper did you knit?’

(75) Incui, a-tu fat che ti? today have.2SG-2SG.SCL do.PTCP what you ‘What did you do today?’

In chapter 5, I will better explain the discourse-pragmatics nature of d-linked wh- elements (Pesetsky 1987). It is important to note that they behave like contrastive XPs, whereas non-d-linked wh-elements behave like XP in informational narrow focus. In (76) and (77) the arboreal representation of (74) and (75) respectively:

200 (76) ForceP

che majon Force' [FOCUS] [CONTRASTIVE] Q-a-tu fat=F FocusP

te a fat=F TP

ti T'

te a fat VP

Vº DP fat che majon [FOCUS] [CONTRASTIVE] (77) FrameP

Incui ForceP

Q-a-tu fat =F FocusP

che Focus' [FOCUS]

te a fat=F TP

ti T'

te a fat VP

Vº DP fat che [FOCUS]

I hence assume the same model as for XPs in informational and contrastive narrow focus. In questions, the V2 constraint is also satisfied by T-to-Force°, but there is no need for the postulation of a [FORCE] feature as Q alone attracts the verb to Force° (see chapter 3).

Even though the present research focuses on root phenomena, it is worth, at least cursorily, investigating what happens in embedded contexts. I argue that the same syntactic mechanism is in place, but T-to-C movement proves impossible as the highest C projection is occupied by an overt complementiser. I assume that the V2 constraint is still in place as the C-domain is activated and projected, but the [FORCE] feature is

201 satisfied by the external merging of the complementiser directly in Force°. The V2 root vs. embedded asymmetry is therefore determined by the fact that Force° is not available to the verb cluster, as shown by the ungrammaticality of (78c):

(78) CONTEXT: What did Maria tell you that Nani killed? (Sovramontino) a. La me a dit che Nani 3SG.F.SCL 1SG.DAT have.3SG say.PTCP that Nani

l a copà L SO GAT 3SG.M.SCL have.3SG kill.PTCP the his cat ‘She told me that John killed his cat’

b. #La me a dit che L SO GAT l a copà Nani

c. *La me a dit che l a copà L SO GAT Nani

In (78), l so gat is in narrow informational focus: the XP in informational narrow focus is the object of the embedded clause. The VOS order is banned as in (78c) and the focal element must be realised in-situ as in (78a). The verb cannot rise to Force° which is occupied by che yielding the ban on the VOS constituent order. Conversely, in root declaratives with a direct object in narrow informational focus, VOS is the default constituent order. As for (78b), it is pragmatically infelicitous: for some of the speakers, it would be felicitous if l so gat were to bear a contrastive focus. Focus fronting is therefore not invariably excluded in embedded contexts, but seems to be frequent in complement clauses that are dependents of bridge verbs (Benincà 2006; Ledgeway 2008). I believe that in embedded contexts, XPs in contrastive focus are not satisfied by the V2 mechanism in place in root contexts, but in FocusP itself (similarly to focus fronting across Romance). This may explain the adjacency effect in (78b), in case (78b) is given a contrastive interpretation. Whether or not the null prosodic focal operator F is active in embedded contexts is open for debate and will not be investigated in the present research. Again, I note that (78b) shows adjacency between the contrastive focal element and the verb cluster suggesting that F may indeed play a role in the expression of contrastive focus in embedded clauses.

An embedded context in which the V2 constraint manifests itself is in clauses introduced by a wh-element as in (79) and (80):

202 (79) No se parchè che te me mande via (Lamonat) NEG know.1SG why that 2SG.SCL 1SG.OCL send.2SG away ‘I don’t know why you push me away’

(80) Nol me a dit sa che l a NEG-3SG.M.SCL 1SG.DAT have.3SG say.PTCP what that 3SG.M.SCL have.3SG ‘He did not tell me what is wrong’

Across the NIDs, the complementiser che often accompanies the wh-element in dependent interrogatives (Benincà & Poletto 2004). In Sovramontino and Lamonat wh- embedded clauses, the co-occurrence with che is compulsory. It is possibly dictated by the need to satisfy the V2 [FORCE] feature through the lexicalisation of Force°. Embedded wh-elements consist of either clitic, like sa in (80), or d-linked tonic wh- elements, like parchè in (79). As argued for root wh-questions in chapter 3, both types of wh-elements appear in ForceP: clitic wh-elements attach to Vfin in Force°, whereas d- linked tonic wh-elements check their [CONTRASTIVE] feature in the preverbal V2 discourse-salient position, namely SpecForceP. Tonic non-d-linked wh-elements cannot introduce an embedded clause. In order to reach firmer conclusions, further systematic research on focal elements in embedded contexts is necessary. What clearly emerges is that the pragmatically motivated V2 system is not in place and the left periphery of non- root contexts seems to more closely resemble that of Romance SVO systems (see Cruschina 2006).

To conclude, I have formalised the pragmatically motivated V2 system in place in Lamonat and Sovramontino. The proposed model captures the geometry of the left peripheral space in the two NEIDs under investigation. It considers both the syntax- pragmatics and the syntax-phonology interfaces, whereby the right syntactic configuration is given by both the satisfaction of the relevant functional features, namely [CONTRASTIVE], [FOCUS] and [TOPIC], and the expression of the right tonal event. In the next section, I will provide one last empirical argument in support of the discourse-pragmatics nature of Lamonat and Sovramontino’s Force-V2 system by looking at the data from a relatively closely NEID, Friulian, which suggests that residual V2 T-to-C movement takes place to convey a particular type of informational narrow focus, hence for discourse-pragmatics needs.

203 4.5. Three Focal Positions: Friulian Marked Informational Focus

In this section, I will concisely outline the behaviour of narrow focus in Friulian, paying attention to a particular type of narrow focal construction that does not behave like either informational or contrastive focus. The data from Friulian seem to support the claim that Lamonat and Sovramontino’s V2 is pragmatically motivated: in Friulian, V2 has virtually been lost, but has been retained in co-occurrence with this type of narrow focus, which, for now, I will call marked informational narrow focus. The entrenched V2 structure generates the right syntactic configuration needed to appropriately capture the discourse-pragmatic feature linked to the aforementioned non- contrastive non-informational narrow focus. Ultimately, the data point to the existence of three distinct dedicated syntactic positions for the realisation of narrow focus in Friulian and, possibly, cross-linguistically: two left-peripheral (pre- and post-finite) positions and one post-verbal in-situ position. I will finally put forward some wider considerations on the relationship between narrow focus and syntax within the Romance language family. More specifically, I will point out that, although virtually all Romance languages exhibited the V2 property, the resulting distribution of narrow focus as an operator syntactic position is rather diverse, but ultimately linked to the right syntactic- pragmatic and syntactic-prosodic interpretation of focal XPs.

Friulian exhibits a peculiar type of narrow focus, which is exemplified in (81) below:

(81) CONTEXT: Speaker A asks speaker B about her lottery results this week (Frl.) O ai UN NUMAR cjapaat! 1SG.SCL have.1SG a number get.PTCP ‘I got a number!’

The syntactic behaviour of this type of narrow focus differs from that of informational narrow focus in (82) and contrastive narrow focus in (83):

(82) CONTEXT: How many numbers did you get from the lottery? (Friulian) O nd-ai cjapaat UN 1SG.SCL PART.CL-have.1SG get.PTCP one ‘I got one’

204 (83) CONTEXT: Did you get two numbers from the lottery? UN NUMAR o ai cjapaat no do-i a number 1SG.SCL have.1SG get.PTCP not two-M.PL ‘I got one number, not two numbers’

Syntactically speaking, the focal element in (81) is found between the inflected auxiliary and the past participle: this linear order rules out both the syntactic realisation of (i) contrastive narrow focus, which, in Friulian, is realised in pre-verbal position in the left periphery (cf. 83), and (ii) informational narrow focus, which is realised in-situ in post-verbal position (cf. 82).

The first striking difference between Friulian, on the one side, and Lamonat and Sovramontino, on the other, is that, in Friulian, a focal XP can appear between the auxiliary and the past participle. In Lamonat and Sovramontino the equivalent constituent order is ungrammatical: nothing can appear between the auxiliary verb and the past participle (except clitic-like adverbs) due to the clitic nature of auxiliaries in these two varieties (see section 4.3.1.1 above). The fact that the auxiliary verb can be freely detached from the past participle and appear in a position higher than an XP bearing a focal reading, suggests that: (i) in (81), the inflected verb moves to the C- domain, leaving the past participle in its canonical T-domain position; (ii) unlike Lamonat and Sovramontino, Friulian auxiliaries are not clitic in nature. As far as the focal constituent un numar in (81) is concerned, it must also appear in an ex-situ left peripheral focal position, since, being the object of the clause, its canonical in-situ position would actually be after the past participle (cf. informational focus in (82)). I propose that the realisation of Friulian’s marked informational narrow focus relies on the same pragmatically motivated V2 mechanism in place in Lamonat and Sovramontino: the focal XP is realised in SpecFocP, whereas the inflected verb undergoes T-to-C movement to Force° to satisfy a residual [FORCE] feature that is only triggered with this type of narrow focus. In addition, the null prosodic operator F also plays a role in this construction as the inflected verb followed by the focal element exhibits a high-pitched F0 contour. In (85) the formal arboreal representation of (84), which features the XP, l adsl, in marked informational narrow focus:

(84) CONTEXT: The speaker is talking about the innovations that recently took place in her street

205 L an pasât o vin L ADSL no vude the year past 2PL.SCL have.2PL the adsl we have.PTCP ‘Last year we got a fast internet connection’

(85) FrameP

l an pasât ForceP

o vin =F FocusP

l adsl Focus' [FOCUS]

o vin=F TP

no T'

o vin VP

Vº DP vude l adsl [FOCUS]

The past participle does not move in Friulian and only the inflected verb moves to the C-domain. I argue that for this type of focal representation V2 is maintained. V2 retrenchments are common across NEIDs (see Salvi 2016). It is striking that, in Friulian, the data suggest that a similar mechanism of that of Lamonat and Sovramontino is in place to convey a specific type of discourse-pragmatic interpretation. The independent evidence from Friulian strengthens our hypothesis that Lamonat and Sovramontino’s V2 system, despite being more productive, is strongly pragmatically motivated.

Now that I have proposed a syntactic model for this type of narrow focus in Friulian, we need to explore its discourse-pragmatic interpretation. In (86) below, an example of marked informational narrow focus in Friulian:

(86) CONTEXT: while I was away, did anyone visit? (Friulian) E je UNE FEMINE rivade, ti cirive 3SG.F.SCL be.3SG a woman arrive.PTCP 2SG.OCL looking-for.3SG.IMP ‘A woman came, she was looking for you’

206 As the name suggests, I consider the focal structure in (86) an instance of informational narrow focus: it can in fact be used to provide new-non-contrastive information in a question-answer setting. Nonetheless, it has an extra discourse-pragmatic flavour and, at the syntactic level, enjoys a different realisation. In Friulian, postverbal focalisation or in-situ focalisation (see Cruschina 2016) has become the predominant strategy to express XPs in informational narrow focus, in line with most modern Romance languages. I argue that the discourse-pragmatic interpretation of Friulian marked informational narrow focus is similar to informational focus fronting in Sicilian and Sardinian (Jones 1993; Cruschina 2006; Bentley 2008; Cruschina & Remberger 2009). In modern Sardinian and Sicilian the focal nature of an XP is a sufficient condition to trigger fronting. Focus fronting can hence take place with informational narrow focal elements; postverbal (or in-situ) focalisation of XPs in narrow focus is still possible both in Sardinian and Sicilian, yet focus fronting is much preferred (Cruschina 2012). Cruschina (2006) points out that the discourse-pragmatic interpretation of fronted informational narrow focal elements is not the same as that of in-situ informational narrow focal elements: the fronted XPs express surprise/or unexpected information. In addition, Cruschina (2006) points out that the illocutionary force of the sentence containing a fronted informational focal XP is that of an exclamative sentence. This interpretation seems to be in line with that of marked informational narrow focus in Friulian. However, the surprise/or unexpected reading does not always seem to capture the discourse-pragmatic interpretation of marked informational narrow focal elements: it works for (84) and (86), but less so for (85) or (87) below:

CONTEXT: Speaker B’s reply after speaker A compliments speaker B’s wristwatch.

(87) Lu ai A UDIN comprât 3SG.M.OCL have.1SG in Udine buy.PTCP ‘I bought it in Udine’

In the case of marked informational narrow focus in Friulian, for instance examples (85) and (87), it could be better described by assuming that the focal XP encodes some degree of contrastiveness, but not enough to be eligible to the preverbal left-peripheral position. This idea will be further explained in chapter 5. For now, I would argue that Friulian marked informational narrow focus fulfils the highlighting function (see Molnar 2006) whereby a constituent is drawn from a wide set of possible entities and is emphasised in discourse. I will not discuss further the discourse-pragmatic or syntactic

207 characteristics of Friulian marked informational narrow focus56. The two important empirical points are: (i) there seems to be a third focal structural position in the clause where narrow focus can be expressed, besides the in-situ postverbal position and the left peripheral preverbal position; (ii) the residual V2 system in place with marked informational narrow focal elements is pragmatically motivated, yielding a strong link between information structure and the V2 phenomenon. As for the first point, across Romance, there are at least two different focal positions associated with narrow focus: a postverbal position within the VP and a higher preverbal position within the C-domain (Cruschina 2016). Bianchi and Bocci (2012) and Bianchi (2013) show that the preverbal and the postverbal focal positions are not equivalent: in certain syntactic contexts, one position is privileged instead of the other according to two different interpretations of narrow focus. Sicilian and Sardinian provide strong evidence in favour of the existence of at least two different focal positions. These two Romance languages, along with Friulian, actually suggest that there are three different structural positions in which narrow focus can be expressed: two C-domain positions for contrastive focus and fronted informational focus, and one T-domain position for informational focus. Across Romance, the expression of argument focus does not hence neatly divide into contrastive and informational narrow focus, but other discourse-pragmatic shades of interpretation are possible and frequent57. In addition, the encoding of narrow focus often interacts with other syntactic phenomena such as quantifier fronting and negative preposing (Benincà 1988; Cinque 1990; Vallduví 1993; Haegeman 2000; Quer 2002), making its syntactic-pragmatic description more challenging.

To conclude, it is intriguing to see how three relatively distant Romance languages, namely Friulian, Sicilian and Sardinian, exhibit a similar syntactic-pragmatic

56 Note that marked informational narrow focus is also compatible with yes/no questions in Friulian. As shown by the following sentence: Ajal PIERI magjade le mortadele? Literally ‘Has Piero eaten the mortadella?’ The fronting of a focal XP in a yes/no question is also possible in Sicilian: Cruschina (2006) argues that this type of question represents a request for confirmation of some unexpected information or a rhetorical question, in which the exclamative and interrogative illocutionary forces combine. Syntactically speaking, this type of yes/no questions does not necessarily argue against our account of Friulian interrogatives in chapter 3. Q is still satisfied in FinP, but the residual V2 [FORCE] feature makes the inflected verb move further up in the structure. 57 Also, see verum focus fronting (VFF) in Spanish (Leonetti & Escandell-Vidal 2009). Superficially, it resembles Friulian marked informational narrow focus; however, under closer scrutiny, they constitute two different types of narrow focus. Spanish VFF has the purpose of emphasising the positive truth-value of the proposition by preposing a non-contrastive XP. In addition, Leonetti and Escandell-Vidal (2009) point out that VFF is not informational in nature, as it can never be used in a question-answer pair. 208 behaviour when it comes to the expression of narrow focus. For instance, the realisation of informational focus in the C-domain is possible in Friulian similarly to Sicilian and Sardinian informational focus fronting. In Lamonat and Sovramontino, the encoding of informational narrow focus in the C-domain is not only possible, but the default option. Further research on the Italian dialects may reveal similar strategies in the encoding of argument focus into the syntax and similar discourse-pragmatic interpretations of focal XPs. More specifically, it would be desirable to assess the extent to which residual V2 phenomena play a role at the syntax-pragmatics interface as a means of licensing specific syntactic-pragmatic configurations and discourse-pragmatic interpretations.

4.6. Broad Focus and SVO

In this last section, I will complete the investigation of focus structure (Lambrecht 1994) in Lamonat and Sovramontino by looking at how broad focus is encoded in the two NEIDs. As previously mentioned, the expression of broad focus does not trigger the pragmatically motivated V2 system, but its realisation complies with a canonical Romance SVO system. In this respect, a competing grammar model is in place: V2 is shifting to SVO. The default SVO constituent order is manifest in out of the blue thetic sentences. In a minimalist or cartographic framework, the formal representation of broad focus is problematic, as no specific feature valuation is associated with the elements in broad focus. Nonetheless, I reject the possibility that the whole clause in broad focus is moved to FocusP: in Lamonat and Sovramontino, tonic focal elements associated with the FocusP position require adjacency with the verb cluster and receive prosodic stress. If the whole clause were to move to FocusP these two essential requirements would not be met. I hence assume that XPs in broad focus have no need to check a [FOCUS] feature, but they are realised in their in-situ or canonical T-related position. Finally, I will put forward the hypothesis that the present shift from V2 to SVO may be fuelled by the reanalysis of VS unaccusative structures. Let us now define and exemplify broad focus. In broad focus, the focus domain, namely the portion of the clause bearing a focal reading, extends to more than one constituent. Lambrecht (1994) identifies two kinds of broad focus: 1) predicate focus where all the constituents of the clause are focal, except for the subject; in other words, the predicate is the assertion about the subject-topic; and 2) sentence focus where the focus domain extends over the whole sentence. In sentence

209 focus, there is no topical subject. I will first briefly illustrate predicate focus and then sentence focus. The sentences in (88) and (89) are instances of predicate focus in Lamonat and Sovramontino respectively, the focal portion of the sentence is represented in SMALL CAPITALS:

(88) a. CONTEXT: What did Mario do? (Lamonat) Mario L A BEÙ L VIN DO N TE LA CANEVA Mario 3SG.M.SCL have.3SG drink.PTCP the wine down in the cellar ‘Mario drank the wine in the cellar’

b. CONTEXT: What happened to Maria? Maria LA È CADUA Mary 3SG.F.SCL be.3SG fall.F.PTCP ‘Mary fell'

(89) a. CONTEXT: What did Mario do? (Sovramontino) Mario L A BEEST L VIN IN CANTINA Mario 3SG.M.SCL have.3SG drink.PTCP the wine in cellar ‘Mario drank the wine in the cellar’

b. CONTEXT: What happened to Maria? Maria L È CAJESTA Maria 3SG.F.SCL be.3SG fall.F.PTCP ‘Mary fell’

The topical part of the utterance or topic (Lambrecht 1994, adopted from Gundel 1988) is traditionally considered the presupposed-known portion of the sentence that is easily retrievable from previous discourse. In Rizzi's (1997) split-CP model topics are generally hosted in the left peripheral portion of the clause under TopicP. However, this is not the case in Lamonat and Sovramontino as these two NEIDs lack TopP in the C- domain for bona fine topics (see section 4.3.4). Mario and Maria in sentences like (88) and (89) generally behave like frame-setting elements (see discussion in chapter 5). If they do not behave like frame-setting or contrastive elements speakers prefer to omit them from the subject position altogether. 58 All constituents in predicate focus, namely the verb and its arguments are realised in the T-domain and the linear order of the sentence is unmarked: in other words, this is the default manner of answering the

58 If the topical subject functions as a frame-setter the C-domain is not activated and appears in FrameP. If the subject bears a contrastive interpretation and needs to check the [CONTRASTIVE] feature, the C-domain is activated and projected: the verb cluster moves to Force° to satisfy the V2 [FORCE] feature and the contrastive topical subject in SpecForceP. 210 questions in (88) and (89). Frascarelli (2000) shows that in Italian broad focus (predicate and sentence focus) can never constitute a single phonological unit, but phonological boundaries are held among the different constituents in focus. Assuming a model of grammar where syntactic phrases correspond to phonological units (see Frascarelli 2000), this makes it virtually impossible for the whole TP layer to be hosted under FocP. In Sovramontino and Lamonat, the prosody of (88a) and (89a) suggests that the utterance is made up of three phonological phrases (represented as φ in square bracket) as shown below in (90):

(90) CONTEXT: What did Mario do? (Sovramontino)

[L a beest]φ [ l vin]φ [in cantina]φ 3SG.M.SCL have.3SG drink.PTCP the wine in cellar ‘Mario drank the wine in the cellar’

Given the presence of more phonological phrases in the same utterance, I argue against the movement of the whole sentence under FocP, but every single constituent that makes up the focus domain of the utterance is realised in-situ, as is the case with Italian. As previously mentioned, a model whereby the whole TP layer moves to SpecFocP to check a [FOCUS] feature would be syntactically and prosodically incompatible with the type of focal elements that can be hosted in FocusP.

As far as sentence focus is concerned, the focus domain extends over the whole sentence. In sentence focus, there is no topical subject. Sentences in (91), (92), (93) and (94) illustrate sentence focus in Lamonat and Sovramontino:

(91) CONTEXT: What happened? (Lamonat) MARIO L A COPÀ L CAN Mario 3SG.M.SCL have.3SG kill.PTCP the dog ‘Mario killed the dog'

(92) CONTEXT: What happened? È CADÙ MARIA be.3SG fall.PTCP Mary ‘Mary fell’

(93) CONTEXT: What happened? (Sovramontino) MARIO L A COPÀ L CAN Mario 3SG.M.SCL have.3SG kill.PTCP the dog ‘Mario killed the dog' 211 (94) CONTEXT: What happened? È CAJEST MARIA be.3SG fall.PTCP Mary ‘Mary fell’

As shown in (92) and (94), core unaccusative verbs (Burzio 1986) like gner (come), morir (die), cajer (fall) obligatorily undergo verb-subject inversion in sentence focus contexts. A pre-verbal subject would yield an ungrammatical sentence as in (95a):

(95) CONTEXT: What happened? (Lamonat)

a. *Maria è cadù Mary be.3SG fall.PTCP ‘Mary fell’

(92) and (94) show that in a thetic sentence featuring a core unaccusative verb, there is no agreement between the verb and the subject (in this case shown by the lack of gender agreement on the past participle). Conversely, in contexts of predicate focus like (88b) and (89b) the unaccusative verb agrees with the topical null or overt subject. In sentence focus, the establishment of such agreement would be either pragmatically infelicitous, as in (95b), or ungrammatical, as in (95c):

(95) CONTEXT: What happened? (Lamonat)

b. #Maria l è cadua Mary 3SG.F.SCL be.3SG fall.F.PTCP ‘Mary fell’

c. *L è cadua Maria 3SG.F.SCL be.3SG fall.F.PTCP Mary ‘Mary fell’

Lack of agreement is primarily signalled by the absence of the third person singular feminine subject clitic la. SCLs are, in fact, analysed by some scholars as agreement markers (see Poletto 1993, 2000): their absence within the verb cluster indicates that no agreement between the subject and the verb is in place. In (92) and (94), the verb is conjugated in third person singular masculine and does not agree with the post-verbal subject in gender. The agreement is between a null third person masculine expletive subject, namely proexpletive (Rizzi 1986; Rizzi & Shlonsky 2006) and the unaccusative verb (see Bentley and Cruschina 2018 for an alternative account). 212 Core unaccusative verbs have a reduced argument structure: they lack an external argument and their only thematic argument is base-generated in VP complement position, syntactically behaving like an object (Burzio 1986; Belletti 1988). The broad focal status of the internal argument and the unaccusative verb demands that the former be realised in-situ where it is base-generated, namely in VP internal position. Even though broad focus has no specific feature associated with its realisation, it implicitly affects the constituent order by requiring that that all constituents contained in it be realised in-situ, hence the word orders in (92) and (94).

Finally, I want to put forward the hypothesis that the current shift from a pragmatically motivated V2 to an SVO system with respect to the realisation of narrow focus is fuelled by the reanalysis of VS presentational clauses. Let us consider the sentences in (96) and (97):

(96) CONTEXT: What happened? (Lamonat and Sovramontino) É MORT PIERA be.3SG die.PTCP Piera ‘Piera died’

(97) CONTEXT: Who died? É mort PIERA be.3SG die.PTCP Piera ‘Piera died’

The syntactic representation of the two clauses in broad sentence focus in (96) and in informational narrow focus in (97) is rather different, as shown by their arboreal representations in (98) and (99) respectively:

(98) TP

proexpl T'

è mort VP

Vº DP mort Piera

213 (99) ForceP

è mort =F FocusP [FORCE]

Piera Focus' [FOCUS]

è mort=F TP

proexpl T'

è mort VP

Vº DP mort Piera [FOCUS]

Given the exact same surface realisation of the two clauses, speakers may reanalyse the structure in (99) as being the same as (98), namely much lower and T-related. The result is a shift from a pragmatically V2 system to a Romance SVO system where narrow focus is realised in the T-domain. This, in fact, opens to the reinterpretation of subjectless VO sentences in which O is in narrow informational focus. It is, however, true that intonation alone, through the realisation of the null prosodic operator F, should differentiate (96) and (97). Nonetheless, focal prominence is given by the linear adjacency between the verb cluster and the focal element, which is achieved by default in (96) and may also generate a similar prosodic effect. Further research is needed in this direction to investigate this hypothesis.

To conclude, in Lamonat and Sovramontino, broad focus is realised in-situ in the T-domain. There is a difference in the realisation of sentence and predicate focus, especially visible in the realisation of presentational VS clauses. The expression of broad focus is not linked to the satisfaction of any particular discourse-pragmatic feature, but affects word order by dictating the default in-situ position to the XPs that make up the focus domain of the clause. Finally, I have put forward the idea that the on- going shift between a pragmatically motivated V2 system to a Romance SVO system may involve the reanalysis of presentational sentences.

214 4.7. Conclusion

In this chapter, I have investigated how focus structure is encoded in two NEIDs under investigation, namely Lamonat and Sovramontino. The investigation of narrow focus has unveiled a syntactic machinery that is pragmatically motivated and resembles the V2 constraint, yielding T-to-C movement in root clauses featuring a contrastive and/or narrow focal XP. This V2 mechanism has allegedly been retained from Medieval Romance through the discourse-pragmatic specialisation of the preverbal V2 position, which has become a structural position that exclusively hosts contrastive XPs. The pragmatically motivated High or Force-V2 system does not obey strict V2 linear restrictions, but is a syntactic V2 system, whereby a strong C° feature, namely [FORCE], must be checked through verb movement to Force° once the C-domain is activated and projected. This V2 system is by no means canonical, but behaves differently from Germanic, Rhaeto-Romance or Medieval Romance V2. In fact, throughout the chapter, I have often referred to it as a ‘collapsing’ V2 system. Residual V2 movement co-occurs with the leftwards movement of a pragmatically salient constituent.

215 5. Root Contrastiveness and V2: A Supra-Informational Status

5.1. Introduction

The literature on Romance and Germanic V2 (Holmberg 1986, Holmberg & Platzack 1995, Vikner 1995, Vilkuna 1995, Poletto 2002, Benincà 2006 a. o.) hints at a crucial relationship between the V2 phenomenon and information structure, whereby the fronted XP is pragmatically salient by virtue of encoding some specific discourse- pragmatic function. The discussion of the phenomenon is nonetheless often marginal in the literature59 and more research is needed to formalise the interaction between the V2 constraint and information structure. This chapter offers a step in that direction trough the investigation of Lamonat and Sovramontino. As argued in the previous chapter, these Romance varieties exhibit residual V2 constituent order, which serves the sole purpose of unpackaging discourse-pragmatic information. Through Rizzi’s (1997) left periphery, I will show that the formalisation of this type of interface phenomenon not only allows us to acquire a better understanding of the mechanisms at the basis of the syntax-pragmatics interface, but also to uncover the primitive discourse-pragmatic functions that are valued by the syntax-pragmatics interface itself. In fact, in Lamonat and Sovramontino, V2 does not interact with any pragmatically salient XP, but specifically with contrastive XPs, suggesting that the syntactic-pragmatic realisation of contrastiveness is a fundamental language component in the dialects under investigation.

Lamonat and Sovramontino depart from most Romance languages as they retained Medieval Romance V2 (see Benincà 1983) as a crucial strategy for the realisation of contrastive XPs, lending support to the claim that a pivotal bidirectional relation between information structure and the V2 constraint is indeed in place. So far, I have argued that V2-motivated T-to-C movement ensures: (i) adjacency of the verb to the pragmatically salient element that, hence, receives discourse prominence; and (ii) the correct interpretation of contrastive elements, which are realised in the preverbal field. In this chapter, I will focus on the latter point. In Sovramontino and Lamonat, the

59 See Hinterhölzl and Petrova (2010) for the discussion of the development and interaction of V2 and information structure in German. 216 displaced verb functions as a boundary that marks different discourse functions; in particular, it sharply sets the boundary between contrastive vs. non-contrastive informational units (see Kuno 1976, Kiss 1998).

5.1.1. Contrastiveness and V2: An Overview

In due course, I will provide a fine-grained definition of contrastiveness. An XP is contrastive when it specifically refers to an information unit that is contrary to the corresponding interlocutor’s presuppositions (Kiss 1998). In this respect, contrastiveness can apply orthogonally to different informational units that exhibit either focal or topical characteristics. Contrastiveness is nonetheless traditionally considered a property of foci. A considerable amount of research has shown that contrastive focal elements receive a different treatment than regular or informational foci at the syntactic level. There is general agreement in favour of two separate dedicated structural positions that encode either contrastive or informational focus (Kiss 1998; Rizzi 1997; Belletti 2001, 2004; Benincà & Poletto 2004; Bianchi & Francarelli 2010; Bianchi & Bocci 2012; Cruschina 2012; Bocci 2013: Rizzi & Bocci 2017 and many others). There is also amounting evidence that, contrastiveness is independent of focus and also extends to topics: topics can also bear a contrastive reading and occupy a specialised syntactic position (Szabolcsi 1981; Gundel 1988; Frascarelli & Hinterhölzl 2007; Lopez 2009; Neeleman et al. 2009; Bianchi & Francarelli 2010). Erteschik-Shir (2007) points out that cross-linguistically: (i) sentence initial topics tend to represent so- called contrastive or switch topics and consist of a specific type of topic that has focus- like properties, and (ii) sentence initial foci tend to represent restrictive or contrastive foci consisting of a specific type of focus that has topic-like properties. In other words, contrastiveness seems to have the ability of superseding topic or focus status. The NEIDs under investigation, Lamonat and Sovramontino, strictly obey this typological trend, syntactically exhibiting a sharp distinction between contrastive and non- contrastive elements. A constituent bearing contrastiveness is banned from the TP layer and must be realised in the left-peripheral space.

Contrastive elements, be they topics or foci, are limited to one occurrence per

217 sentence and must occur clause-initially60. I argue that the preverbal position occupied by contrastive XPs is filled through the activation of the V2 constraint: more specifically, I argue that, in the two NEIDs under investigation, the [EGDE] feature that, in Medieval Romance V2, was responsible for the presence of an overt XP in the specifier position of the C° targeted by the moved verb, has undergone pragmatic specialisation and become [CONTRASTIVE], being only active in case a contrastive element enters the derivation. My analysis of contrastive elements will treat contrastive topics and foci as occupying the same highest structural position in the clause in the SpecCP position of the C° hosting the moved verb. The present research hence adopts a unitary syntactic analysis of this discourse-pragmatic category in the two Romance varieties under investigation. The present investigation suggests that contrastiveness should therefore be regarded as an independent discourse-pragmatic status that supersedes topic and focus. In this respect, I will show that contrastiveness should not be considered a categorical notion, but a continuum, in which the degree of contrastiveness is determined by the properties of the set containing the contrastive element.

5.1.2 A Unified Treatment of Contrastive XPs: Cross-linguistic Evidence

Before delving into the analysis of the syntactic-pragmatic behaviour of contrastive elements in Lamonat and Sovramontino, let us look at cross-linguistic evidence in support of a unified treatment of contrastiveness as an independent informational status. In several languages, contrastive elements are morpho- syntactically encoded irrespectively of their informational nature as foci or topics, providing evidence in favour of the supra-informational nature of contrast with respect to the topic-focus divide. One such language is Finnish: Vilkuna (1995) argues that contrastive topics and foci target the same dedicated structural position in the C-domain. She claims that, in Finnish, contrast must be analysed as a property of topicality and focusing, but independent from either informational statuses. Contrastiveness therefore enjoys special status and, in Finnish, is encoded at the syntactic level through a

60 They can be preceded by a frame-setting element. Frame-setters are base-generated elements that occupy a dedicated clause-external syntactic position, namely FrameP (see Haegeman 2000, 2006, 2007). In section 5.4, I argue that A-topics (Bianchi & Frascarelli 2010) are also licensed in FrameP and hence behave like frame-setting elements. 218 designated high structural position. Nevertheless, syntax is not the only way in which a language can highlight the special discourse-pragmatic status of contrastive elements: other languages encode contrastiveness through different strategies. For example, as for the familiar case of English, contrastiveness is encoded at the prosodic level and the constituent bearing a contrastive reading receives sentential stress. The same is true for Italian: in case a contrastive element is not realised in the C-domain, but in the TP layer, it receives sentential stress which distinguishes the contrastive focal element from its non-contrastive focal counterpart, as shown respectively in (1) and (2):

(1) CONTEXT: Did you give a T-shirt to the winner? (Italian) No. Gli abbiamo dato UNA MEDAGLIA NEG DAT.CL have.1PL give.PTCP a medal 'No. We gave him a medal.’

(2) CONTEXT: What did you give to the winner? Gli abbiamo dato UNA MEDAGLIA DAT.CL have.1PL give.PTCP a medal ‘We gave him a medal’

In a language like Italian that arguably does not draw a sharp syntactic distinction between contrastive and non-contrastive elements, prosody seems to be the only means to draw this distinction. Note that contrastive focus will be represented in BOLD CAPITALS (cf. 1) and informational focus in CAPITALS (cf. 2). Different languages of the world, to different degrees, seem to adopt different strategies to mark contrastiveness as an independent informational status, like, for instance, the use of specialised syntactic configurations or prosodic patterns (Selkirk 2002, Molnár 2002). Another means of encoding contrastiveness as a supra-informational status is through morpho-syntax, as in the case of Japanese (see Shimojo 2011). Kuno (1973) notes that certain instances of use of the Japanese topic marker -wa is ambiguous between a topic and a focus reading, claiming that the common pragmatic reading suitable for both elements is a contrastive one.

Cross-linguistic evidence suggests that contrastiveness is specially marked in several languages of the world and, in some cases, can supersede the notion of focus and topic. In such cases, the same linguistic strategy marks the target constituent as contrastive regardless of the topic-focus divide: in Lamonat and Sovramontino, this

219 common strategy translates into the activation of the V2 constraint.

5.2. The Data: Contrastive XPs in Lamonat and Sovramontino

NEIDs show a non-homogeneous behaviour with regard to the expression of contrastiveness. Lamonat and Sovramontino are much stricter in encoding contrastiveness than, for example, Friulian and Fornese. Let us start by considering contrastive topics in Friulian (3) and Lamonat (4):

(3) CONTEXT: Did you see my parents yesterday? (Friulian) Iar to pari lu ai viodût tal bosc, [to mari e jere a cjase] yesterday your father OCL have.1SG see.PTCP in-the woods 'As for you father, I saw him in the woods yesterday, your mother was at home.'

(4) CONTEXT: Did you see my parents yesterday? (Lamonat) Ier to pare l è vedù an tel bosc, [to mare l era fora casa] yesterday your father OCL have.1SG seen.PTCP in-the woods 'As for you father, I saw him in the woods yesterday, your mother was at home.'

In (3) and (4) the relevant contrastive topic is in bold. It is a topic as it is resumed by the third person singular object clitic l/lu in the TP space: clitic left dislocation is traditionally considered a diagnostic of topicality across Romance (Cinque 1990, Rizzi 1997). The constituent to pare/to pari is contrastive in nature as it is framed in a contrastive relationship with to mare/to mari. In both Lamonat and Friulian, the frame- setter iar/ier appears in a higher position than the underlined contrastive topic to pari/to pare. Frame-setters are circumstantial in nature and modify the propositional content of the main clause: typically, they are adjuncts that provide temporal and/or modal restrictions to the circumstances of evaluation of the proposition expressed by the main clause (see Greco & Haegeman 2016). In the two NEIDs, despite the identical surface position of the contrastive XP, I argue that the underlying constituent order is fundamentally different. In Lamonat, to pare surfaces in the specifier position of the C° targeted by V2-motivated T-to-C movement and is licenced by the contrastive nature of the XP itself. In Friulian, this is not the case, as shown in the proposed labelled bracketing of (3) and (4), in (5) and (6) respectively:

220 (5) [FrameP Iar [TopP to pari [TP [T’ [T lu ai] [VP viodût tal bosc]]]]]]

(6) [FrameP Ier [ForceP to pare [Force’ [Force l è vedù] [TP [T’ [T l è vedù] [VP vedù an tel bosc]]]]]]]

In the rest of the chapter, I will show that this analysis perfectly captures the marked constituent orders found in Lamonat and Sovramontino along with their syntactic requirement on the left-peripheral realisation of contrastive XPs. If the proposal in (5) and (6) were truthful, speakers would be able to insert an overt subject between the contrastive XP and the verb in Friulian, but would not be able to do so in Lamonat. The prediction is borne out, as shown in (7) and (8):

(7) CONTEXT: Did you see my parents yesterday? (Friulian) Iar to pari jo lu ai viodût tal bosc, [to mari …] yesterday your father I OCL have.1SG see. PTCP in-the woods 'As for you father, I saw him in the woods yesterday, your mother was at home.'

(8) CONTEXT: Did you see my parents yesterday? (Lamonat) a. *Ier to pare mi l è vedù an tel bosc, [to mare …] yesterday your father I OCL have.1SG seen. PTCP in-the woods 'As for you father, I saw him in the woods yesterday, your mother was at home.'

b. Ier to pare l è vedù mi an tel bosc, [to mare …] yesterday your father OCL have.1SG seen. PTCP I in-the woods 'As for you father, I saw him in the woods yesterday, your mother was at home.'

In Lamonat, the subject, if overt as background information (G-topic in the sense of Bianchi & Frascarelli 2010), must be placed post-verbally (cf. 8b): nothing can intervene between the contrastive XP and the verb cluster. As argued in chapter 4, this constituent order is typical of languages exhibiting the V2 constraint.

Further evidence that, in Lamonat and Sovramontino root clauses, contrastive XPs must be realised in the C-domain is provided by the behaviour of contrastive foci. Let us consider examples (9) and (10):

221 (9) CONTEXT: Did you give a T-shirt to the winner? (Friulian) Po no no! Iar UNE MEDAE o vin dât al vincitôr EXCL NEG yesterday a medal 1PL.SCL have.1PL give.PTCP to-the winner 'No. We gave a medal to the winner yesterday.'

(10) CONTEXT: Did you give a T-shirt to the winner? (Sovramontino) No. Ier NA MEDAJA ghe on dat al vincitor NEG yesterday a medal DAT.CL have.1PL give.PTCP to-the winner 'No. We gave a medal to the winner yesterday.'

In (9) and (10), the focal element une medae/na medaja corrects the interlocutors presuppositions, hence this type of C-focus is corrective in nature. In (9) and (10) the same linear constituent order is in place, but the underlying constituent order is different. In Sovramontino, focal contrastive XPs must surface in the C-domain, whereas, in Friulian, contrastive elements can also surface in-situ in the TP layer, as shown respectively in (11) and (12):

(11) CONTEXT: Did you give a T-shirt to the winner? (Friulian) Po no! Iar o vin dât UNE MEDAE al vincitôr EXCL NEG yesterday 1PL.SCL have.1PL give.PTCP a medal to-the winner 'No. We gave a medal to the winner yesterday.'

(12) CONTEXT: Did you give a T-shirt to the winner? (Sovramontino) #No. Ier ghe on dat NA MEDAJA al vincitor NEG yesterday DAT.CL have.1PL give.PTCP a medal to-the winner 'No. We gave a medal to the winner yesterday.'

The sentence in (12) is pragmatically infelicitous. It would be a felicitous answer to the question ‘what did you give to the winner?’, given that the post-verbal focal position is associated with informational focus (see chapter 4). The fact that contrastive elements must surface in the C-domain in Sovramontino further suggests that different requirements are in place for the satisfaction of different discourse-pragmatic functions in the two NEIDs, hence the proposed labelled bracketing analyses of (9) and (10), in (13) and (14) respectively:

222 (13) [FrameP Iar [FocP une medae [Foc’ [Foc o vin] [TP [T’ [T o vin] [VP dât une medae al vincitôr]]]]]]

(14) [FrameP Ier [ForceP na medaja [Force’ [Force ge on dat] [TP [T’ [T ge on dat] [VP dat na medaja al vincitor]]]]]]

The proposed labelled bracketing in (13) and (14) captures what seems a strong adjacency requirement across all NEIDs between the verb and the focal XP: it is in fact not only a characteristic of Lamonat and Sovramontino’s pragmatically-motivated V2 system, but is also found across all NEIDs under investigation61. This is however not surprising, as focal elements often require adjacency to the verb cluster to ensure that focal prominence is associated with them (see Cruschina 2012, 2016 for further discussion). What makes Lamonat and Sovramontino remarkably different from the other NEIDs is the sharp distinction between contrastive and non-contrastive elements, as further shown in Lamonat vs. Fornese, in (15) and (16):

(15) CONTEXT: Did you dink grappa? (Lamonat) a. #No. É beù VIN, no sgnapa! NEG have.1SG drink.PTCP wine NEG grappa 'No. I drank wine, not grappa!'

b. No. VIN è beù, no sgnapa! NEG wine have.1SG drink.PTCP NEG grappa 'No. I drank wine, not grappa!'

(16) CONTEXT: Did you dink grappa? (Fornese) a. No. I ai bivût VIN, no sgnapa! NEG 1SG.SCL have.1SG drink.PTCP wine, NEG grappa 'No. I drank wine, not grappa!'

b. No. VIN i ai bivût, no sgnapa! NEG wine 1SG.SCL have.1SG drink.PTCP, NEG grappa 'No. I drank wine, not grappa!'

61 Allegedly, as a reflex of their diachronic V2 nature. T-to-C movement associated with operator positions (root interrogatives and narrow focus adjacency) is fairly common across NIDs. The is general agreement in considering these phenomena residual V2 properties (see Rizzi 1996, Salvi 2016). 223 Fornese, the same as Friulian and Italian, allows C-foci to appear in the TP layer, if they are marked with a special prosodic intonation. In Lamonat and Sovramontino, this is not possible. The contrastive element cannot appear in the TP layer and must surface at the left-edge position of the clause. Note that frame-setters are hereby considered clause- external elements (see Haegeman 2006, 2007). As for contrastive topics, the same distinction is not found. In fact, their inability of occurring in the TP layer is a characteristic of all NEIDs under investigation. This suggests that C-topics are intrinsically associated with the C-domain. By adopting Rizzi’s (1997) split CP model, I argue that the position contrastive topics occupy in the C-domain is subject to variation across the NEIDs under investigation. In Lamonat and Sovramontino, this position is higher than in Friulian and Fornese, namely ForceP. In the next section, I will briefly restate why ForceP is the projection targeted by contrastive elements in Lamonat and Sovramontino: the explanation that I provided in chapter 4 is tightly linked to the pragmatically V2 nature of these two NEIDs.

In sum, the evidence from cross-dialectal comparison shows that Lamonat and Sovramontino syntactically encode contrastiveness in a stricter way than Friulian and Fornese. The latter group encodes contrastiveness in a similar fashion to Italian, the former, instead, exhibit the following characteristics: a contrastive element, be it a topical or focal constituent, (i) must surface at the left-edge of the clause, and (ii) can never be realised in-situ in the TP layer. Cross-dialectally, C-topics and C-foci retain the hallmarks of their respective informational status. For instance, C-foci must be adjacent to the verb and, if they are arguments of the verb, C-topics must be resumed by an agreeing pronominal clitic. Ultimately, Lamonat and Sovramontino differ from the other NEIDs with respect to how contrastive elements are realised syntactically.

5.3. The V2 Nature of Lamonat and Sovramontino

In chapter 4, I claimed that the two NEIDs under investigation, namely Lamonat and Sovramontino, are pragmatically motivated V2 systems: continuations of V2 Medieval Romance (see Benincà 1983, 2006, Salvi 2016) in which the V2 constraint has been reinterpreted as a tool to encode specific discourse-pragmatic functions, namely contrastiveness and/or narrow informational focus. I argued that the V2 system in place differs both from that of Medieval Romance and that of contemporary and 224 reasonably closely related NIDs (see Poletto 2002): the preverbal position can only be occupied by an XP bearing discourse-pragmatic contrastiveness. This means that even if the verb undergoes V2-motivated T-to-C movement, an XP will not move in front of it, unless it bears a contrastive reading, hence, a matching [CONTRASTIVE] feature. Note that T-to-C movement also takes place to accommodate narrow focus in Lamonat and Sovramontino, whereby it assures linear adjacency between the narrow focal element and the verb cluster, as shown in (17):

(17) CONTEXT: What did Maria do to her husband? (Sovramontino) La ghe a dat AN BESO Maria al so om 3SG.F.SCL DAT.CL have.3SG give.PTCP a kiss Maria to-the her husband ‘Maria gave a kiss to her husband’

V2-motivated T-to-C movement only occurs to accommodate these two discourse pragmatic configurations, hence the label ‘pragmatically motivated V2 system’. The V2 constraint does not seem to be active in clauses featuring broad focus (Lambrecht 1994). Generally speaking, the V2 constraint is actived by the fronting of a pragmatically salient constituent. Note that in this chapter I will only discuss contrastive XPs (for a discussion of informational focal elements see chapter 4).

5.4. The Preverbal XP: The Contrastive Field

Besides the proposal of a unitary syntactic analysis of contrastive foci and contrastive topics that can account for the behaviour of contrastive elements in Lamonat and Sovramontino, I also want to shed light on the discourse-pragmatic characteristics of the XPs that are licensed in V2 preverbal position. I will call the higher portion of the left periphery that hosts contrastive elements the contrastive field. Before delving into the discussion of this portion of the left periphery and the XPs that it can host, let us first provide a more fine-grained definition of the notion of contrastiveness.

5.4.1. The Hierarchy of Contrast

I firstly define contrastiveness by adopting Kiss’s (1998) definition, which can be

225 roughly paraphrased as follows: an element is contrastive when it specifically refers to an information unit that is contrary to the corresponding interlocutor’s presuppositions. This definition captures the immediate discourse-pragmatic context in which we would find a constituent bearing contrastiveness, but it is not adequate for a thorough description of all the instances in which we find a constituent marked as contrastive in Lamonat and Sovramontino. Referring to contrastive focal elements, Molnár (2002, 2006) argues that contrast is best described as a scalar notion. She proposes a hierarchy of contrast made of a set of salient criteria that define contrastiveness as a continuum, capturing the different discourse-pragmatic shades that characterise a contrastive relationship and, hence, contrastiveness:

Hierarchy of Contrast (Molnár 2006: 211) Highlighting > Dominant contrast > Membership in a set > Limited set of candidates > Explicit mentioning of alternatives

This hierarchy captures well all the cases in which a constituent would be assigned a contrastive reading in Sovramontino and Lamonat.

5.4.1.1. Dominant Contrast

Let us now discuss the actual discourse-functional criteria put forward by Molnár (2006). Dominant contrast is the most straightforward pragmatic context in which a contrastive element can appear. Let us consider the examples in (18) and (19):

(18) CONTEXT: Has Mario eaten? (Lamonat & Sovramontino) No. MARIA l a magnà NEG Maria 3SG.F.SCL have.3SG eat.PTCP ‘No, MARIA has’

Contrastiveness in (18) is an example of dominant contrast: Maria is in a contrastive relation with Mario that has been previously mentioned in discourse. In other words, a dominant contrast is in place if the contrasting piece of information is overt in discourse. (18) is an example of dominant contrast within the domain of focus: a C- focal element. Nevertheless, dominant contrast is also possible in relation to C-topics,

226 as shown in (19) below:

(19) Tasi! Mi son to mare e ti te se me fiol (Sovr.) Shut-up.IMP I be.1SG your mother and you 2SG.SCL be.2SG my son ‘Shut up! I am your mother and you are my son.’

In (19), the two stressed subject pronouns are in a dominant contrastive relationship as mi has an overt contrasting element in the same utterance, namely ti. The two subject pronouns are not regular subjects, but two instances of contrastive topics. Differently to bona fide topical subjects, they are stressed and cannot be omitted (see Ziv 1994 for a detailed discussion of the relationship between stressed pronominal elements and contrastiveness). C-topics and C-foci occupy the same syntactic position at the leftmost clause-internal position of the left periphery, namely the contrastive field in SpecForceP.

5.4.1.2. Membership in a Finite Set of Entities

The last three discourse functions in Molnár’s (2006) hierarchy of contrast, namely Membership in a set > Limited set of candidates > Explicit mentioning of alternatives, will be collapsed into a single discourse-pragmatic function: Membership in a finite set of entities. The pragmatic context of this revised discourse-function is one for which a contrastive element emerges from an explicit choice among a finite set of alternatives where the presupposed contrasting informational unit has not been previously mentioned in discourse (see also Kiss 1998). Let us consider the example in (20):

(20) No sta preocupar-te! MARIO l va in botega (Lamonat) NEG stay.INF worry.INF-2SG.OCL Mario 3SG.M.SCL go.3SG in shop ‘Don’t worry! Mario will go shopping.’

In (20), Mario is marked as a contrastive focal element. It is an instance of C-focus as the adjacency requirement with the verb prevents any constituent from intervening between the focal element and the verb. As discussed in chapter 4, in Lamonat and Sovramontino, subjects in informational narrow focus are banned from the preverbal position (they have to appear in a cleft structure). On the other hand, subjects encoding

227 contrastive focus are grammatical and pragmatically felicitous in preverbal position, as they satisfy the [CONTRASTIVE] feature linked to the V2 constraint. In (40): the contrastive element Mario belongs to the finite set of alternatives of people that can potentially go shopping that, at least, consists of the speaker, the hearer and an absent third party Mario. In (20) Mario does not have an explicitly contrastive antecedent that has been previously mentioned in discourse, but is in a contrastive relation with the other members of the set. The same is possible with C-topics as shown by (21) below:

(21) No sta preocupar-te! El to camp (*da la tempesta) (Sovr.) NEG stay.INF worry.INF-2SG.OCL the your field (by the hail)

no l è sta ruinà da la tempesta NEG 3SG.M.SCL be.3SG be.PTCP damage.PTCP by the hail ‘Don’t worry! Your field was not devastated by the hail.’

In (21), the element el to camp is in a contrastive relationship with the other members of the set in which it belongs: the crop fields of the village. The context of (41) must be thought of as if, after hailing, speaker A, who had just come back from a walk across the crop fields of the village, reassures speakers B, who enquires about the condition of the fields due to the hail. El to camp is an instance of C-topic and occupies the contrastive field (SpecForceP): it cannot be separated from the verb cluster, as shown by the ungrammaticality generated by the intervening constituent da la tempesta.

The discourse function membership in a finite set of entities easily blends in with another discourse function marking contrastiveness put forward by Molnár (2006), namely highlighting. In (21), for instance, the speaker highlights a specific crop field out of all the crop fields of the village. The crucial difference between membership in a finite set of entities and highlighting is the characterisation of the set from which the contrastive element is drawn. In membership in a finite set of entities the set is finite and all members of the set are known to the speakers, whereas, in highlighting, the set of entities does not have to be finite and one of the speakers may not have immediate access to the alternatives of that set. The divide between these two functions can be indeed blurry, with membership in a finite set of entities that, to some extent, can be thought of as a sub-function of highlighting.

228 5.4.1.3. Highlighting

The function of highlighting gives pragmatic saliency to an informational unit by selecting an entity out of an indefinite set of alternatives when these alternatives are not immediately accessible to the interlocutor. The chosen entity ends up in a contrastive relationship with the other entities of the set that could have been highlighted. Lambrecht (1994) argues that highlighting is a sufficient condition for contrast. Nevertheless, there is general agreement that the expression of contrastiveness through highlighting generates a weak contrast (see Cruschina 2012 for a detailed discussion). This is true in Lamonat and Sovramontino, as highlighting, to some extent, can also be performed by informational focus, frame-setters and A-topics. Not surprisingly, the function of highlighting is very rarely found in association with contrastive focus. It is, in fact, more common with C-topics. The semantic-pragmatics of such topical elements anchors the interpretation of following clause, which must be interpreted in light of the left-edge topical element.

A type of C-topics that seems to heavily draw from the highlighting discourse- function is stage topics (in the sense of Erteschirk-Shir 2007). Stage topics do not belong to a presupposed set of entities or have an explicit contrastive counterpart in discourse, but they are anchored to the spatio-temporal dimension of the communicative act. Gundel (1974) and Erteschirk-Shir (1997) argue that this type of information unit refers to the particular spatio-temporal situation or setting in which it is asserted. In other words, every act of communication has a stage, in which information can be accessed by virtue of belonging to the here and now. Let us consider example (22):

(22) CONTEXT: The speaker is touching a rotten mushroom, which is lying on a table Sta brisa l è mejo che no te la magne (Lamonat)

this mushroom 3SG.SCLexpl be.3SG better that NEG 2SG.SCL OCL eat.2SG ‘This mushroom, it’s better if you don’t eat it.’

In example (22), sta brisa is a stage topic. It is contrastive in the sense that serves the function of highlighting. It does not contrast with an explicitly uttered piece of information, nor belongs to a finite set of entities that is immediately accessible to the interlocutor. Its contrastive nature can be interpreted as follows: out of all the entities present in the setting of the communicative act, the speaker highlights a specific one,

229 the mushroom; therefore, the contrastive relationship in place is between that particular informational unit and all the other potential stage topics.

The contrastive element in (22) exhibits topic-like features as it is resumed by an agreeing object clitic; nevertheless, it does not identify old or given information and is prosodically marked. In Frascarelli and Hinterhölzl (2007) and Bianchi and Frascarelli (2010) this type of topic closely resembles a switch or A(boutness)-topic. A-Topics newly propose or reintroduce a topic in discourse; they are generally clause-external and anchor the interpretation of the following clause, which must be interpreted in light of the left-edge A-topic. This suggests that A-topics may also be eligible candidates for the preverbal contrastive field (SpecForceP) in Lamonat and Sovramontino, since their highlighting function makes them de facto a type of contrastive element. Nevertheless, similarly to Italian, A-topics can co-occur with bona fide C-topics with the A-topic preceding the C-topic (Bianchi & Frascarelli 2010) as in (23):

(23) (Sovramontino) Le nosele Titin l le ciapea e Titele l le magnea the hazelnuts Titin 3SG.M.SCL OCL pick.3SG.IMP and Titele 3SG.M.SCL OCL eat.3SG.IMP ‘The hazelnuts, Titin used to pick them up and Titele used to eat them’

In (23), the A-topic le nosele is (re)introduced. It is resumed by an agreeing object clitic and is compatible with a reading that puts Titin and Titele in a contrastive relationship.

Similarly to A-topics, frame-setters are a particular type of topical elements that are clause external, and encode the spatio-temporal dimension of the speech act. They modify the propositional content of the main clause by providing temporal and/or modal restrictions to the circumstances of evaluation of the proposition expressed by the main clause (Greco & Haegeman 2016). In this respect, frame-setters are also good candidates for the fulfilment of the discourse-function highlighting. The speaker highlights one aspect of the spatio-temporal or modal dimension of the speech act, which ends up being in a contrastive relationship with all the other potential aspects that the speaker could have highlighted of such dimension. Frame-setters, similarly to stage topics, have a scene-setting function, which anchors the speech act in terms of locative, temporal deixis and speech participants (Haegeman 2000, 2006, 2007, Benincà & Poletto 2004, Poletto 2002). Given their highlighting function, they can be regarded as

230 contrastive in nature and, potentially, suitable candidates for the contrastive field in Lamonat and Sovramontino. Krifka and Musan (2012) also point out the pragmatically- syntactic common characteristics of contrastive topics and frame-setters by putting them in the same category as delineators: they express that, in the communicative act at a specific point of discourse, the given element is only applicable to a limited portion of discourse that is anchored to the delineator.

5.4.2. Topical and Focal Contrastiveness: The Revised Hierarchy of Contrast

Given their distribution with respect to the discourse-function they serve, we can divide contrastive elements into two general categories: focal contrastive elements and topical contrastive elements. Focal contrastiveness is encoded though contrastive focus. In Lamonat and Sovramontino, virtually all Molnár’s (2002, 2006) discourse-functions can be fulfilled by a C-focal element, with the exception of highlighting. On the other hand, topical contrastive elements can be split into frame-setters, A-topics and C-topics. They tend to fulfil different discourse-functions within contrastiveness. Nevertheless, A-topics and frame-setters can coexist with C-topics. In those cases, they genuinely behave like clause external anchoring elements and do not occupy the contrastive field, but a higher projection, namely FrameP. Table 1. below summarises the typology of contrastive elements that I have just illustrated:

Membership in Dominant Highlighting finite set of entities contrast Focal N/A C-Foci C-Foci Contrastiveness Topic A-Topics A-Topics C-Topics Contrastiveness Frame-setters C-Topics

Table 1. The typology of contrastive elements in Lamonat and Sovramontino with respect to Molnár’s (2002, 2006) revised discourse-functions of contrastiveness.

I have previously mentioned that highlighting generates a weak contrast. In fact, this function is not performed by contrastive focal elements, as C-foci are generally associated with an overt contrast in discourse that is corrective or exhaustive in nature

231 (see Szabolcsi 1981, Cruschina 2012). The behaviour of C-foci is more in line with the discourse-functions of membership in a finite set of entities and dominant contrast. Dominant contrast, especially, generates a strong contrast, in which two explicit elements are contrasted. Following Molnár’s (2002, 2006) claim that contrastiveness should be interpreted as a continuous notion, I can now place highlighting and dominant contrast at either side of the continuum, with membership in a finite set of entities laying in the middle. Membership in a finite set of entities is more similar to either highlighting or dominant contrast depending on the characteristics of the set of alternatives from which the contrastive element is drawn. If the set is limited with explicit alternatives, the discourse function membership in a finite set of entities partially overlaps with that of dominant contrast.

On the other hand, if the set is finite, but its members are not immediately accessible to the interlocutor, the function of membership in a finite set of entities tends to overlap with that of highlighting. Table 1 suggests that focal contrastiveness serves strong contrastive discourse-functions, whereas topical contrastiveness better serves weak contrastive discourse-functions. In discourse-pragmatic terms, in Lamonat and Sovramontino, it is more common to find dominant contrast fulfilled by a contrastive focal element and highlighting fulfilled by a Frame-setter or an A-topic. In this regard, highlighting is seldom expressed in the contrastive field, but more often in FrameP. Table 2 below summarises what I have just claimed as a revised version of Molnár’s (2002, 2006) hierarchy of contrast:

WEAK Highlighting > STRONG CONTRAST Membership in finite set of entities > CONTRAST Dominant Contrast Set of alternatives >

TOPICAL Finite set of alternatives > FOCAL CONTRAST Finite set of explicit alternatives > CONTRAST Two-way contrast

Table 2. Revised Hierarchy of Contrast for Lamonat and Sovramontino

232 The common denominator of focal and topical contrastive elements is indeed contrast: more specifically, different degrees of contrastiveness that are fulfilled by different discourse-functions. In light of this common denominator, Lamonat and Sovramontino treat all contrastive elements in the same way at the syntax-pragmatics interface. Contrastive XPs are syntactically assigned a single surface position in the specifier position of ForceP, in a spec-head configuration with the verb cluster which moves to Force° to satisfy the V2 constraint. Topical contrastive elements differentiate themselves from focal contrastiveness and, hence, from contrastive focus in a number of ways: i) as opposed to contrastive focus, they do not require adjacency with the inflected verb62; ii) they can be externally merged, whereas contrastive focus is always associated with internal merge; iii) they are resumed by an agreeing clitic when arguments of the verb. This shows that despite the different morpho-syntactic characteristics of contrastive topics and contrastive foci, the V2 system in place forces all pragmatically salient-constituent bearing contrastiveness to surface in the contrastive field. A possible reason for why contrastiveness is a salient discourse-pragmatic feature in the two NEIDs under investigation is that it affects the common ground (Stalnaker 1974, Lewis 1979) of the communicative act. The common ground (CG) can be defined as the set of propositions that are mutually known by the interlocutors and are to be shared in an act of communication. The CG is made up of different subcomponents that characterise the conversational context. Each new sentence must be interpreted as a potential update of the CG. Krifka (2007) draws an important distinction between CG content and CG management: the former concerns the truth- conditional information that has been stated up to a point in conversation, whereas the latter concerns the conversational moves of the participants involved in the communicative act. In light of the notion of common ground, we can now propose a common property shared by C-foci, C-topics, A-topics and Frame-setters: they all pertain to the dimension of CG management (Bianchi & Frascarelli 2010, Krifka & Musan 2012). For instance, Krifka & Musan (2012) claim that contrastive focus signals that the CG content contains a proposition with which the current sentence must be contrasted. On the other hand, A-topics signal that a new piece of information is added or reintroduced in the CG content (Bianchi & Frascarelli 2010). Both A-topics and C-

62 C-topics are always adjencent to the verb cluster as they ends up in spec-head realtion with it. Nonetheless, I have claimed that weak contrastiveness can also be encoded by bona fide frame- setters and A-topics: in such cases, adjacency is not in place. 233 foci hence pertain to CG management and are not strictly associated with CG content itself, but with its management. The same is underlyingly true for frame-setters and C- topic. I will not take this discussion on the properties of the CG any further, but I want to propose that discourse saliency in Lamonat and Sovramontino is encoded in those elements that pertain to CG management rather than CG content. At the syntax- pragmatics interface, these contrastive elements are assigned the preverbal position of the V2 system in place, which effectively facilitates the flow of information exchange.

5.5. The Case of D-Linked Wh-Elements

In light of this analysis, D-linked wh-elements are intrinsically contrastive in Lamonat and Sovramontino. They occupy the same syntactic position as topical and focal contrastive elements, namely the contrastive field in SpecForceP. D-linked wh- elements are in fact in complementary distribution with contrastive XPs, as shown in (24) below:

(24) *Mario CHE CASA a-lo comprà? No Toni (Lamonat) Mario which house have.3SG-3SG.M.SCL buy.PTCP NEG Toni 'Which house did Mario buy and not Toni?’

In (24) the d-linked wh-element che casa and the C-topic Mario cannot coexist in the same sentence, unless Mario is demoted to background information63 (G-topic) and placed after the past participle comprà. D-linked wh-elements, in the same fashion as contrastive elements, must surface at the left-edge of the clause and are limited to a single occurrence.

Before delving into the discussion of d-linked wh-elements, let us first define them: the answer to a d-linked wh-element must be drawn from a set of entities that were previously introduced in the discourse. The answer to a d-linked wh-question is thus in a contrastive relation with the other members of the set from which it is drawn (see Pesetsky 1987). The term d-linked wh-elements refers to those wh-phrases that establish a link with the existing discourse and, more specifically, with a finite set of

63 Mario must hence be de-accented and placed clause finally in canonical subject position

234 alternatives. In Rizzi’s (1990) and Cinque’s (1990) terms d-linked wh-elements are referential, meaning that they are associated with given information that is presupposed in discourse.

In terms of Molnár’s (2002, 2006) revised hierarchy of contrast, I claim that d- linked wh-elements serve the membership in a finite set of entities discourse-function and hence qualify as contrastive elements. D-linked wh-elements must in fact be kept apart from regular wh-phrases: in Sovramontino and Lamonat, this distinction is reflected syntactically as d-linked wh-elements and regular wh-elements occupy two distinct structural positions that are not interchangeable, as shown in (25) and (26):

(25) a. A-lo beù CHE Mario? (Lamonat) have.3SG-3SG.M.SCL drink.PTCP what Mario ‘What did Mario drink?’

#b. CHE a-lo beù Mario? what have.3SG-3SG.M.SCL drink.PTCP Mario ‘What did Mario drink?’

(26) a. CHE MAJON a-lo comprà Mario? (Sovramontino) which jumper have.3SG-3SG.M.SCL buy.PTCP Mario 'Which jumper did Mario buy?’

*b. A-lo comprà CHE MAJON Mario? have.3SG-3SG.M.SCL buy.PTCP which jumper Mario 'Which jumper did Mario buy?’

Examples (25) and (26) show that tonic wh-items can occupy two distinct positions in the left peripheral space depending on their informational status. Regular wh-phrases occupy the specifier position of FocusP (see chapter 3). On the other hand, d-linked wh- elements occupy the contrastive field, namely SpecForceP. The position they occupy syntactically satisfies a specific discourse-pragmatic function: heralding contrastiveness. This is further supported by bare or prepositional wh-elements that receive a contrastive reading and, hence, surface in a different syntactic position within the C-domain, as shown in (27) below:

235 (27) (Sovramontino) a. Laorei-tu CO CHI l an pasà? Non-d-linked interpretation work.PST.2SG-2SG.SCL with who the year passed 'With whom did you work last year?’

b. CO CHI laorei-tu l an pasà? D-linked interpretation with who work.PST.2SG-2SG.SCL the year passed 'With whom did you work last year?’

The question in (27b) is uttered seeking the following piece of information: ‘who of an already mentioned set of the people worked with you last year’. In sum, there are two distinct discourse-pragmatically motivated structural positions in Lamonat and Sovramontino for tonic wh-items: SpecForceP for d-linked wh-elements and SpecFocP for their non-d-linked counterparts.

A similar pattern is also found in another Romance language: Romanian. Romanian has received relatively a high degree of attention, as it allows multiple wh- fronting. Comorovski (1996) argues that wh-questions with multiple fronted wh- operators display a crucial property: the higher wh-element must be discourse linked. Dobrovie-Sorin (1994) and Cornilescu (2002) point out that the d-linked interpretation of the higher wh-element is necessary for the grammaticality of the question. In other words, if the leftmost wh-word is not d-linked the resulting question is ungrammatical. Similarly to Lamonat and Sovramontino, the position of the wh-element depends on the discourse-pragmatic status of the wh-element itself. In Lamonat and Sovramontino, multiple wh-fronting, or better, in our case, the presence of two wh-elements, one d- linked and the other non-d-linked, in the C-domain is ungrammatical as shown by (28):

(28) *CO CHI si-tu ndaa AONDE? (Lamonat) with who be.2SG-2SG.SCL go.PTCP where? 'With whom did you go where?’

The ungrammaticality of (28) seems to suggest that in Lamonat and Sovramontino the nature of d-linked wh-phrases is focal and hence cannot coexist with the non-d-linked focal wh-element aonde. Lamonat and Sovramontino in fact observe the general

236 tendency across Romance that focal elements are limited to one single occurrence per sentence (Belletti 2004, Rizzi & Bocci 2017, Rizzi 2018).

Lastly, another key difference that sets apart non-d-linked wh-elements and their d-linked counterpart is the fact that d-linked wh-elements cannot be clefted, as shown in (29b). On the contrary, regular wh-elements can appear within a cleft, as shown in (30):

(29) a. CHE CASE e-lo che i a vendù? (Lamonat)

which houses be.3SG-3SG.SCLexpl that 3PL.M.SCL have.3PL sell.PTCP 'Which houses did they sell?'

b. *E-lo CHE CASE che i a vendù?

be.3SG-3SG.SCLexpl which houses that 3PL.M.SCL have.3PL sell.PTCP 'Which houses did they sell?'

(30) E-lo CHI che a beù al vin?

be.3SG-3SG.SCLexpl who that have.3SG drink.PTCP the wine? ‘Who drank the wine?’

The difference in their syntactic distribution signals that the two different types of wh- elements belong to two distinct syntactic-pragmatic categories. The fact that d-linked wh-elements cannot appear in a cleft strengthens the view that these elements are intrinsically contrastive: they cannot be realised in-situ or in informational focus position (see section 4.6 of previous chapter), but they must be moved to the C-domain where they check their matching [CONTRASTIVE] feature in SpecForceP.

To conclude, I have shown that in Lamonat and Sovramontino, d-linked wh- elements are intrinsically contrastive and their pragmatic-syntactic behaviour is captured well by the analysis put forward in the present chapter. They occupy the same syntactic position as topical and focal contrastive elements, namely SpecForceP, and are, in fact, found in complementary distribution with them. In terms of Molnár’s (2002, 2006) revised hierarchy of contrast, d-linked wh-elements serve the membership in finite set of entities discourse-function and their pragmatic-syntactic behaviour seems to resemble that of contrastive foci.

237 5.6. Conclusion

In conclusion, in Lamonat and Sovramontino root clauses, contrastiveness must be conceived of as a supra-informational and continuous pragmatic notion, which can be encoded by both topical and focal elements. The pragmatically motivated V2 nature of Lamonat and Sovramontino only allows contrastive XPs in preverbal position in the contrastive field. Given the Force-V2 system in place in Lamonat and Sovramontino, contrastive XPs must appear at the left-edge of the clause and are limited to a single occurrence. The contrastive field can host different elements that serve different discourse-functions and hence encode different degrees of contrastiveness: C-topics, C- foci, but also A-topics, Frame-setters, and d-linked wh-elements. C-topics, A-topics, frame-setters encode topical contrastiveness, whereas C-foci and d-linked wh-elements encode focal contrastiveness. The common denominator of focal and topical contrastive elements is the establishment of a contrastive relationship within a set of entities that can vary in size and accessibility to the interlocutor: the smaller the set of entities that are contrasted, the stronger the contrast. Focal contrastiveness generally serves strong contrast, whereas topical contrastiveness serves weak contrast. Even though topical and focal contrastive elements syntactically surface in the same structural position, they exhibit the characteristics of their respective informational status. At the syntax- pragmatics interface, all contrastive elements share a [CONTRASTIVE] feature that must be checked in the contrastive field. Lamonat and Sovramontino hence mark contrastiveness at the syntactic level. Ultimately, this is not uncommon across the languages of the world, which adopt different strategies to transversally mark the difference between contrastive and non-contrastive elements regardless of their informational status as topics of foci.

238 6. Conclusion

This thesis outlines several syntax-pragmatics interface phenomena across North- Eastern Italian Dialects (NEIDs). It sheds light on (i) the syntactic mechanisms adopted to encode different types of discourse-pragmatic information and (ii) the properties of the portion of the clause that syntactically encodes such information. In this chapter, I will summarise the main findings put forward in this thesis. I will do so by following the structure of the thesis. Subsequently, I will highlight some points that would benefit from future research. I shall end with some concluding remarks.

6.1. Summary of Findings

In chapter 2, I provided a partial dialectological description of the four NEIDs discussed in this thesis, namely Fornese, Friulian, Lamonat and Sovramontino. In terms of genetic affiliation, the categorization of Friulian and Fornese is relatively more straightforward than that of Lamonat and Sovramontino. These two last NEIDs exhibit a mixture of features, which are typical of Venetan, Ladin and Gallo-Italic varieties. Friulian and Fornese instead neatly display the characteristic traits of the putative Rhaeto-Romance language family. Making reference to the relevant literature, I nonetheless argued against Rhaeto-Romance as a unitary and independent language family. The features that distinguish NEIDs from the other NIDs are indeed mostly shared retentions rather than shared innovations. In broad terms, I provided evidence in support of the view that considers Rhaeto-Romance to be a sort of crystallised pre- Gallo-Italic, whereby different NEIDs represent different stages of the process that led to contemporary Gallo-Italic. I also described the morpho-syntactic behaviour of SCLs in Sovramontino and Fornese. SCLs behaviour across NEIDs does not particularly differ from that of the rest of NIDs. I argued in support of the claim that considers SCLs to be strong agreement markers; nonetheless, I acknowledged the possibility that, in some particular syntactic environments, SCLs may also resemble the behaviour of weak pronominal elements.

In chapter 3, I argued that wh-in-situ in Lamonat and Sovramontino can be accounted for by treating it as a syntax-pragmatics interface phenomenon. I showed that

239 Lamonat and Sovramontino wh-in-situ is only apparent: the verb crosses the fronted wh- element and appears as the first element of the clause. In Friulian and Fornese root interrogatives, T-to-C movement targets instead a lower left peripheral projection, namely FinP, not giving rise to apparent wh-in-situ. I treated wh-elements as narrow focal elements that occupy the same syntactic position across the NEIDs: the specifier position of FocusP. The investigation of root wh-interrogatives uncovered a fundamental syntactic constraint that distinguishes Friulian and Fornese from Lamonat and Sovramontino: the microparametric variation in the locus of subject clitic inversion (SCLI). It consists of the difference in the left-peripheral functional projection that satisfies the structural requirements on questions and hosts SCLI. I argued that in Lamonat and Sovramontino this functional projection is ForceP, whereas in Friulian and Fornese it is FinP. I also showed that, in Lamonat and Sovramontino, the difference in the distribution between canonical (pre-verbal) and apparently in-situ (post-verbal) wh- elements is constrained by: i) the morpho-syntactic status of the wh-element itself, clitic (pre-verbal) vs. tonic (post-verbal), and (ii) the discourse-pragmatic information it encodes, d-linked (preverbal) vs. non-d-linked (postverbal) interpretation.

In chapter 4, I claimed that in Lamonat and Sovramontino, the non-canonical constituent order found in root-interrogatives is also found in root declaratives that feature a pragmatically salient XP. Specifically, I argued that residual V2 T-to-C movement takes place: hence, the key difference between, on one side, Lamonat and Sovramontino, and the majority of the other NIDs is that residual V2 is not restricted to root interrogatives in the mentioned dialects, but also takes place in a subset of root declaratives. The emergence of non-canonical constituent orders in sentences that exhibit a discourse-pragmatically salient XP yields that the V2 system in place is pragmatically motivated: it is a means to encode specific discourse-pragmatic information. In fact, I claimed that, in root declaratives, T-to-C movement is associated with the movement of a narrow focal element in the C-domain or the fronting of a contrastive XP. As for SVO constituent order, this order is found in root declaratives encoding broad focus and in embedded declaratives. I hence argued that V2 and SVO are competing word orders in the two NEIDs: SVO is the default constituent order, whereas V2 has been reinterpreted as a syntactic tool which assigns the adjacent XP (i) discourse-pragmatic prominence and (ii) the right syntactic-pragmatic interpretation. More specifically, I claimed that the verb cluster (the material under the moved T°)

240 constitutes a boundary between contrastive and non-contrastive elements: contrastive XPs are realised preverbally, in V2 preverbal position (SpecForceP), whereas non- contrastive narrow focal XP are realised post-verbally (SpecFocP). Contrastive elements include both contrastive topics and contrastive foci. As far as non-contrastive topics are concerned, I argued that they are not allowed in the left peripheral space of Lamonat and Sovramontino, as their lexicalisation is incompatible with T-to-C movement.

I showed that the pragmatically motivated V2 system of Lamonat and Sovramontino exhibits most features of V2 languages: subject inversion, root vs. embedded asymmetry, and linear restrictions on V1, V2 and V3. Nevertheless, the data are not always straightforward due to the null-subject nature of Lamonat and Sovramontino: the subject can be omitted, if it encodes background information. Another point of departure between the two NEIDs and canonical V2 languages concerns the clitic status of auxiliaries. In the two NEIDs auxiliaries require adjacency with the past participle. I argued that the V2 system of Lamonat and Sovramontino differs form Medieval Romance V2. In synchrony, it also differs from the V2 system in place in Gardenese and Badiotto. The key difference concerns the discourse-pragmatic specialisation of the preverbal position in Lamonat and Sovramontino. Once T-to-C movement takes place, the pre-verbal position can only be occupied by a contrastive element: if no contrastive element is available the position can be empty (e.g. non- contrastive narrow focus). Contrary to Gardenese and Badiotto, subject clitic inversion (SCLI) does not mark T-to-C movement in declaratives. I argued that, in synchrony, verb movement alone is not enough to justify SCLI, but an extra layer of information must be encoded in the clause in order for SCLI to take place (i.e. interrogative force or an optative/desiderative reading). Lack of SCLI also suggests that Lamonat and Sovramontino are T-languages, in the sense that the T-domain is the most prominent part of the clause, rather than C-languages (like most Germanic V2 languages). Clitics are merged into T° and the complex T° undergoes T-to-C movement.

In chapter 5, I further explored the notion of discourse-pragmatic contrastiveness. In the pragmatically motivated V2 system uncovered in my investigation, the preverbal position can only be occupied by a contrastive element. It can be occupied by both a focal or topical XP, which bears a [CONTRASTIVE] feature. I put forth the idea that

241 this data distribution is the result of the discourse-pragmatic specialisation of the Medieval Romance [EDGE] feature responsible for the presence of a V2 preverbal constituent. Even though contrastiveness supersedes the topic-focus divide, contrastive topics and contrastive foci exhibit the characteristics of their informational status. I argued that contrastiveness should be considered to be a continuous, rather than a discrete notion. The degree of contrastiveness of an XP is determined by the properties of the set of contrasting elements from which it is drawn: most importantly, size and explicitness of the alternatives. Building upon the relevant literature, I identified three main semantic-pragmatic categories that define a contrastive relation: highlighting, member of a finite set, dominant contrast. Highlighting generates a weak contrast, whereas dominant contrast generates a strong contrast. The main correlation found is that dominant contrast is virtually only expressed through contrastive/corrective focus, whereas highlighting never resorts to contrastive/corrective focus. I claimed that highlighting does not by default lead to the filling of preverbal position, but it can also be expressed through informational narrow foci, A-topics, or frame-setters. I finally claimed that d-linked wh-elements obey this mechanism by virtue of being contrastive elements. In fact, they are the only tonic wh-elements that are found in canonical position in Lamonat and Sovramontino (preverbal or V2 preverbal position).

6.2. Challenges, Future Objectives and Concluding Remarks

The analysis put forward in this thesis fosters future research with respect to both the synchronic and diachronic dimension of the NIDs. As far as synchronic work is concerned, the syntactic-pragmatic study of other NIDs could strengthen or weaken the theoretical and empirical generalisations put forward in this thesis: particularly, the study of other NIDs exhibiting wh-in-situ and subject clitic inversion. Along with the investigation of new NIDs, the creation of robust corpora for Fornese, Friulian, Lamonat and Sovramontino would allow for quantitative analysis, which could lead to more sound and accurate generalisations and, most importantly, encourage future research on these NEIDs. The socio-biological clock of these languages is ticking and, unless profound changes tear apart the very fabric of Italian society, which is unlikely, they are destined to disappear in approximately two generations of speakers. Beyond Romance, the investigation of the relationship between V2 and information structure across contemporary Romance and Germanic V2 languages would provide evidence on

242 the extent to which V2 is dependent on information-structure and, hence, is a syntax- pragmatics interface phenomenon.

As far as diachrony is concerned, the systematic investigation of Medieval NIDs and Medieval Italo-Romance could reveal whether, at some stage of the transition from V2 to SVO, a pragmatically motivated V2 system like that of Lamonat and Sovramontino was in place. At least two main challenges can be identified with this type of research: (i) the lack of diachronic evidence for a large number of NIDs and Italo-Romance varieties, and, for those dialects that count historical written texts, (ii) identifying and categorising the relevant discourse-pragmatic categories on the basis of just the written texts. I believe that making further progress in the study of information structure through this line of research is of paramount importance, as it allows us to assess the extent to which discourse-pragmatics is the driving force determining syntactic change. In the case of Lamonat and Sovramontino, different types of discourse-pragmatic information pattern with different basic constituent orders (V2 vs. SVO), suggesting that the on-going syntactic change is indeed sensitive to discourse- pragmatics.

To conclude, until further research is carried out on the NIDs, the claims put forward in this thesis must be considered to be specific to the NEIDs under investigation and cannot be taken for granted across all NIDs. As the present research has shown, each NID is an independent language system and must be investigated as such, hopefully, bringing us a step closer to the understanding of the mechanisms behind human language.

243 Bibliography

Abbi, A. 2001. A manual of linguistic field work and structures of Indian languages. Munich: Lincom Europa.

Alexiadou, A. & E. Anagnostopoulou. 1998. Parametrizing Agr: word order, V- movement and EPP-checking. Natural language & linguistic theory 16: 491- 539.

Ascoli, G. I. 1883. L’Italia dialettale. Archivio glottologico italiano 8, 98-128.

Austin, P. 2006. Data and language documentation. In J. Gippert, N. Himmelmann & U. Mosel (eds.), Essentials of language documentation, 87-112. Berlin: Mouton de Gruyter.

Baker, C. L. 1970. Notes on the description of English questions: the role of an abstract Q morpheme. Foundations of Language 6, 197-219.

Belletti, A. 1988. The case of unaccusatives. Linguistic inquiry 19 (1), 1–34.

Belletti, A. 1999. “Inversion” as focalization and related questions. Catalan working papers in linguistics 7, 9–45.

Belletti, A. 2001. Inversion as focalization. In A. Hulk & J. Pollock (eds.), Subject inversion in Romance and the theory of universal grammar, 60-90. Oxford/New York: Oxford University Press.

Belletti, A. 2004. Aspects of the low IP area. In L. Rizzi (ed.), The structure of IP and CP. The cartography of syntactic structures 2, 16-51. Oxford/New York: Oxford University Press.

Belletti, A. 2005. Extended doubling and the VP periphery. Probus 17 (1), 1-35.

Benincà, P. 1983/84. Un’ipotesi sulla sintassi delle lingue romanze medievali. Quaderni patavini di linguistica 4, 3-19.

Benincà, P. 1988. L’ordine degli elementi della frase e le costruzioni marcate. In L. Renzi (ed.), Grande grammatica italiana di consultazione 1. : Il Mulino, 129-194.

Benincà, P. 1994. La variazione sintattica. Studi di dialettologia romanza. Bologna: Il Mulino.

244 Benincà, P. 1995. Complement clitics in Medieval Romance: the Tobler-Mussafia Law. In A. Battye & I. Roberts (eds.), Clause structure & language change, 325–44. Oxford/New York: Oxford University Press.

Benincà, P. 2001. The position of topic and focus in the left periphery. In G. Cinque & G. Salvi (eds.), Current studies in Italian syntax. Essays offered to Lorenzo Renzi, 39-64. Amsterdam: Elsevier.

Benincà, P. 2006. A detailed map of the left periphery of medieval Romance. In R. Zanuttini (ed.), Crosslinguistic research in syntax and semantics: negation, tense and clausal architecture, 53-86. Georgetown: Georgetown University Press.

Benincà, P. 2007. Clitici e ausiliari: gh ò, z é. In D. Bentley & A. Ledgeway (eds.), Sui dialetti italo-romanzi. Saggi in onore di Nigel Vincent. The Italianist 27, 27-47.

Benincà, P. & C. Poletto. 2004. Topic, focus and V2: defining the CP sublayers. In L. Rizzi (ed.), The structure of IP and CP. The cartography of syntactic structures 2, 52–75. Oxford/New York: Oxford University Press.

Benincà, P. & L. Vanelli. 2016. Friulian. In A. Ledgeway & M. Maiden (eds), The Oxford guide to the Romance languages, 139-153. Oxford: Oxford University Press.

Beninca, P., M. Parry & D. Pescarini. 2016. The dialects of Northern Italy. In A. Ledgeway & M. Maiden (eds.), 185-205

Bentley, D. 2007. Relazioni grammaticali e ruoli pragmatici: siciliano e italiano a confronto. In A. Ledgeway (ed.), Sui dialetti italo-romanzi: saggi in onore di Nigel B. Vincent, 48-62. King's Lynn, Norfolk: Biddles.

Bentley, D. 2008. The interplay of focus structure and syntax: evidence from two sister languages. In R. D. Van Valin (ed.), Investigations of the syntax-semantics- pragmatics interface, 263-284. Amsterdam: Benjamins.

Bentley, D. & S. Cruschina. 2018. The silent argument of broad focus: typology and predictions. Glossa: a journal of general linguistics 3(1), 118, 1-37.

Bernini, G. 2012. Il clitico a nell'italo-romanzo settentrionale: osservazioni metodologiche. In O. Vincenzo (ed), Per Roberto Gusmani. Linguistica storica e teorica. Studi in ricordo, 269-282. Udine: Forum.

Berruto. G. 1987. Sociolinguistica dell’Italiano Contempoaraneo. Roma: Carocci.

245 Berruto, G. 1989. On the typology of linguistic repertoires. In U. Ammon (ed.), Status and function of languages and language varieties, 552–569. Berlin–New York: de Gruyter.

Berruto, G. 1993. Varietà diamesiche, diafasiche, diastratiche. In A. Sobrero (ed.), Introduzione all'italiano contemporaneo. La variazione e gli usi. Bari: Laterza.

Berruto, G. 2005. Dialect/standard convergence, mixing, and models of language contact: the case of Italy. In P. Auer, F. Hinskens & P. Kerswill (eds.), Dialect change. Convergence and divergence in European languages, 81–97. Cambridge: Cambridge University Press

Bianchi, V. 1999. Consequences of antisymmetry: headed relative clauses. Berlin: Mouton de Gruyter.

Bianchi, V. 2013. On focus movement in Italian. In M. V. Camacho-Taboada, A. Jiménez Fernández, J. Martín-Gonzáles & M. Reyes-Tejedor (eds.), Information structure & agreement, 193-216. Amsterdam & Philadelphia: John Benjamins.

Bianchi, V. & G. Bocci. 2012. Should I stay or should I go? Optional focus movement in Italian. Empirical issues in syntax and semantics 9.

Bianchi, V. & M. Frascarelli. 2010. Is topic a root phenomenon? Iberia 2 (1), 43–88.

Bianchi, V., G. Bocci, & S. Cruschina. 2015. Focus fronting and its implicatures. In E. Aboh (ed.), Romance languages & linguistic theory 2013: selected papers from going Romance, 1–19. Amsterdam: John Benjamins.

Biberauer, T. & I. Roberts. 2010. Subjects, tense and verb-movement. In T. Biberauer, A. Holmberg, I. Roberts & M. Sheehan (eds), Parametric variation: null subjects in minimalist theory 263-303. Cambridge: CUP.

Bidese, E. & Tomaselli, A. 2007. Diachronic development in isolation: the loss of V2 phenomena in Cimbrian. Linguistische Berichte 210, 209–228.

Bidese E. & A. Tomaselli. 2016. V2 and nominative case: the view from a German- based language-contact dialect in Italy. Rethinking verb second workshop, Cambridge, 22-24/3/16.

Bidese, E. & A. Tomaselli. 2018. Developing pro-drop. The case of Cimbrian. In: F. Cognola & J. Casalicchio (eds.), Null subjects in generative grammar. A synchronic and diachronic perspective, 52-69. Oxford/New York: Oxford University Press.

Bird, S. & G. Simons. 2003. Seven dimensions of portability for language 246 documentation and description. Language 79, 557-582.

Bocci, G. 2013. The syntax-prosody interface: A cartographic perspective with evidence from Italian. Amsterdam: John Benjamins Publishing.

Bocci, G. & C. Avesani. 2005. Focus contrastivo nella periferia sinistra della frase: un solo accento, ma non solo un accento. Proceedings of 2° convegno nazionale associazione italiana di scienze della voce.

Bolinger, D. 1977. Meaning and form. London: Longman.

Bošković, Ž . 2001. On the nature of the syntax-phonology interface: cliticization and related phenomena. Amsterdam: Elsevier.

Brandi, L. & P. Cordin. 1989. Two Italian dialects and the null subject parameter. In O. Jaeggli & K. Safir (eds.), The null subject parameter, 111-142. Dordrecht: Kluwer.

Bresnan, J. 1972. Theory of Complementation in English Syntax. Ph.D. Dissertation, MIT.

Brunetti, L. 2004. A Unification of Focus. Padova: Unipress.

Büring, D. 1999. Topic. In P. Bosch & R. van der Sandt (eds.), Focus - linguistic, cognitive, and computational perspectives, 142-165. Oxford University Press, New York.

Burzio, L. 1986. Italian syntax. A government and binding approach. Dordrecht: Reidel.

Calabrese, A. & D. Pescarini. 2014. Clitic metathesis in the Friulian dialect of Forni di Sotto. Probus 26 (2), 275-308.

Cardinaletti, A. 2001. A second thought on emarginazione: destressing vs. right dislocation. In G. Cinque & G. Salvi (eds.), Current Studies in Italian Syntax. Essays offered to Lorenzo Renzi, 117-135. Amsterdam: Elsevier.

Cardinaletti, A. & M. Starke 1999. The typology of structural deficiency. A case study of the three classes of pronouns. In H. van Riemsdijk (ed.), Clitics in the languages of , 145-233. Berlin: Mouton De Gruyter.

Cardinaletti, A. & L. Repetti. 2010. Proclitic vs enclitic pronouns in northern Italian dialects and the null-subject parameter. In R. D’Alessandro, A. Ledgeway & I. Roberts (eds), Syntactic variation. The dialects of Italy, 119-134. Cambridge: Cambridge University Press. 247

Cardona. G. 1986. Storia universale della scrittura. Milano: Mondadori.

Chafe, W. L. 1976. Givenness, contrastiveness, definiteness, subjects, topics, and point of view. In C. N. Li (ed.), Subject and topic, 25–55. New York: Academic Press.

Chelliah, S. 2001. The role of text collection and elicitation in linguistic fieldwork. In P. Newman and M. Ratliff (eds.), Linguistic Fieldwork, 152-165. Cambridge: Cambridge University Press.

Chelliah, S. & W. de Reuse. 2011. Handbook of descriptive linguistic fieldwork. Berlin: Springer.

Cheng, L. 1997. On the Typology of wh-questions. New York/Londen: Garland Publishing.

Chomsky, N. 1981. Lectures on government and binding. Dordrecht: Foris.

Chomsky, N. 1986. Knowledge of language: its nature, origins and use. New York: Praeger.

Chomsky, N. 1993. A minimalist program for linguistic theory. In K. Hale & S. J. Keyser (eds.), The view from building 20, 1-52. Cambridge, MA: MIT Press.

Chomsky, N. 1995. The minimalist program. Cambridge, MA: MIT Press.

Chomsky, N. 2000. New horizons in the study of language and mind. Cambridge: Cambridge University Press.

Chomsky, N. 2001. Derivation by phase. In M. J. Kenstowicz (ed.), Ken Hale: a life in language, 1–52. Cambridge, MA: MIT Press.

Chomsky, N. 2005. Three factors in language design. Linguistic inquiry 36, 1–22.

Chomsky, N. & H. Lasnik. 1977. Filters and control. Linguistic inquiry 8, 425-504.

Cinque, G. 1990. Types of A’-dependencies. Cambridge, MA: MIT Press.

Cinque, G. 1994. On the evidence for partial n-movement in the Romance DP. In G. Cinque, J. Koster, J. Pollock, L. Rizzi & R. Zanuttini (eds.), Paths towards Universal Grammar. Studies in Honor of Richard S. Kayne, 85–110. Washington: Georgetown University Press.

Cinque, G. 1997. 'Topic’ constructions in some European languages and ‘connectedness’. In E. Anagnostopoulou, H. van Riemsdijk, & F. Zwarts (eds.), 248 Materials on left dislocation, 93–118. Amsterdam/Philadelphia: John Benjamins.

Cinque, G. 1999. Adverbs and functional heads. Oxford/New York: Oxford University Press.

Cinque, G. 2002. The structure of DP and IP. The cartography of syntactic structures 1. Oxford University Press, New York.

Clackson, J. 2016. Latin as a source for the Romance languages. In A. Ledgeway & M. Maiden (eds.), 3-13.

Comorovski, I. 1996. Interrogative phrases and the syntax-semantics interface. Dordrecht: Kluwer.

Cornilescu, A. 2002. Rhematic focus at the left periphery: The case of Romanian. In C. Beyssade, R. Bokbennema, F. Drijkoningen, & P. Monachesi (eds.), and Linguistic Theory, 77-91. Amsterdam: John Benjamins.

Corrà, L. 2001. Il dialetto di Lamon : cultura nelle parole. Lamon : di Lamon.

Corrà, L. 2001. Note sul dialetto di Lamon. Studi linguistici alpini in onore di Giovan Battista Pellegrini, 184-195. Firenze, Istituto di Studi per l’Alto Adige.

Cruschina, S. 2006. Informational focus in Sicilian and the left periphery. In M. Frascarelli (ed.), Phases of interpretation, 363-385. Berlin: De Gruyter Mouton.

Cruschina, S. 2008. Discourse-related features and the syntax of peripheral positions. A comparative study of Sicilian and other Romance languages. Ph.D Dissertation, University of Cambridge.

Cruschina, S. 2012. Discourse-related features and functional projections. Oxford/New York: Oxford University Press.

Cruschina, S. 2016. Information and discourse structure. In A. Ledgeway & M. Maiden (eds.), 596-608.

Cruschina, S. & E. Remberger. 2009. Focus fronting in Sardinian and Sicilian. Proceedings of the XXXV incontro di grammatica generativa (Siena 26– 28/2/09). STiL-CISCL working papers 3,118–30.

D’Imperio, M. 2002. Italian intonation: an overview and some questions. Probus 14,1. 37-69.

249 De Bortoli, L. 2016. Sovramonte. Guida al territorio. Santa Giustina (BL): Alpinia Itinera.

De Cia, S. 2013. La vitalitât dal furlan te provincie di Udin. Un studi sociolinguistic. Udine: Provincia di Udine.

De Mauro, T. 1991. Storia linguistica dell’Italia unita. Bari: Laterza.

Demonte, V. & O. Fernandez-Soriano. 2009. Force and finiteness in the Spanish complementizer system. Probus 21 (1), 23–49.

Den Besten, H. 1983. On the interaction of root transformations and lexical deletive Rules. In W. Abraham (ed.), On the formal syntax of the Westgermania, 47-61. Amsterdam/Philadelphia: John Benjamins.

Dik, S. 1989. The theory of functional grammar. Part I: the structure of the clause. Dordrecht: Foris.

Dobrovie-Sorin, C. 1994. The syntax of Romanian. Comparative studies in Romance. Berlin: Mouton de Gruyter.

Donati, C. & M. Nespor. 2003. From focus to syntax. Lingua 113, 11, 1119-42.

Dorian, N. 1977. The problem of the semi-speaker in . International Journal of the Sociology of Language, 23-32.

Dorian, N. 1981. Language death: the life cycle of a Scottish Gaelic dialect. Philadelphia: University of Pennsylvania Press.

Erteschik-Shir, N. 1997. The dynamics of focus structure. Cambridge: Cambridge University Press.

Erteschik-Shir, N. 2007. Information structure: The syntax-siscourse interface. Oxford/New York: Oxford University Press.

Falk, C. 1993. Non-referential subjects and agreement in the history of Swedish. Lingua, 89, 143-80.

Ferguson, C. A. 1959. . Word 15: 325-340.

Fernández Rubiera, F. J. 2009. Clitics at the edge: clitic placement in western Iberian Romance languages. Ph.D. dissertation, Georgetown University.

Finco, F. 2007. Fonetiche e fonologjie. In F. Fari (ed.), Manuâl di lenghistiche furlane, 47-82. Udine: Forum. 250

Fontana, J. 1993. Phrase structure and the syntax of clitics in the history of Spanish. Ph.D. Dissertation, University of Pennsylvania.

Fontana, J. 1997. On the integration of second position phenomena. In: A. van Kemenade & N. Vincent (eds.), Parameters of morphosyntactic change, 207- 249. Cambridge: CUP.

Frascarelli, M. 2000. The syntax-phonology interface in focus and topic constructions in Italian. Dordrecht: Kluwer.

Frascarelli, M. & R. Hinterhölzl. 2007. Types of topics in German and Italian. In S. Winkler & K. Schwabe (eds.), On information structure, meaning and form, 87– 116. Amsterdam: John Benjamins.

Frota, S., M. D’Imperio, G. Elordieta, P. Prieto & M. Vigário. 2007. The phonetics and phonology of intonational phrasing in Romance. In P. Prieto, J. Mascaró & M.-J. Solé (eds.), Prosodic and segmental issues in (Romance) phonology, 131-153. Amsterdam/Philadelphia: John Benjamins.

Gartner, T. 1883. Raetoromanische grammatik. Heilbronn, Henniger.

Gelb. I. 1993. Teoria generale e storia della scrittura. Fondamenti della grammatologia. Milano: Egea.

Givón, T. 1976. Topic, pronoun and grammatical agreement. In C. N. Li (ed.), Subject & topic, 151–188. New York: Academic Press.

Goria, C. 2004. Subject clitics in the Northern Italian dialects. Dordrecht: Kluwer.

Greco, C. & L. Haegeman. 2016. Framesetters and the micro-variation of subject-intial V2. lingbuzz/003226.

Grenoble, L. A. & L. J. Whaley. 2006. Saving languages: an introduction to language revitalization. New York: Cambridge University Press.

Gundel, J. K. 1974. The role of topic and comment in linguistic theory. Ph.D. Dissertation, University of Texas at Austin.

Gundel, J. K. 1988. Universals of topic-comment structure. In M. Hammond, E. Moravczik & J. Wirth (eds.), Studies in syntactic typology, 209-239. Amsterdam: John Benjamins.

Haegeman, L. 1995. The syntax of negation. Cambridge: Cambridge University Press.

251 Haegeman, L. 1996. Verb second, the split CP and null subjects in early Dutch finite clauses. Geneva Generative Papers 4, 133–75.

Haegeman, L. 2000. Inversion, non-adjacent inversion and adjuncts in CP. Transactions of the Philological Society 98, 121–160.

Haegeman, L. 2006. Argument fronting in English, Romance CLLD and the left periphery. In R. Zanuttini, H. Campos, E. Herburger, & P. Portner (eds.), Negation, tense and clausal architecture: Cross-linguistic investigations, 27-52. Georgetown: Georgetown University Press.

Haegeman, L. 2007. Operator movement and topicalisation in adverbial clauses. Folia linguistica 41, 279-325.

Haegeman, L. 2012. Adverbial clauses, main clause phenomena, and the composition of the left periphery. Oxford: Oxford University Press.

Hagège. C. 1985. L’homme de paroles. Contribution linguistique aux sciences humaines. Paris: Arthème Fayard.

Haiman, J. & P. Benincà. 1992. The Rhaeto-Romance languages. London: Routledge.

Hale, K., M. Krauss, L. J. Watahomigie, & A. Yamamoto. 1992. Endangered languages. Language 68:1-42.

Himmelmann, N. 1998. Documentary and descriptive linguistics. Linguistics 36, 161- 95.

Himmelmann, N. P. 2006. Language documentation: What is it and what is it good for? In J. Gippert, N. P. Himmelmann & U. Mosel (eds), Essentials of language documentation. Berlin: Walter de Gruyter.

Hinterhölzl, R. & S. Petrova. 2010. From V1 to V2 in West Germanic. Lingua, 120 (2), 315-328.

Holmberg, A. 1986. Word order and syntactic features in the Scandinavian languages and English. Ph.D. Dissertation, University of Stockholm.

Holmberg, A. 1997. Scandinavian stylistic fronting: movement of phonological features in the syntax. Working papers in Scandinavian syntax 60. Department of Scandinavian languges. University of Lund.

Holmberg, A. 2000. Scandinavian stylistic fronting: how any category can become an expletive. Linguistic inquiry 31, 445–484.

252 Holmberg, A. 2012. Verb second. In T. Kiss & A. Alexiadou (eds), Syntax: an international handbook of contemporary syntactic research. Berlin: Walter de Gruyter.

Holmberg, A. & C. Platzack. 1995. The role of inflection in Scandinavian syntax. Oxford/New York: Oxford University Press.

Horn, L. R. 2001. Flaubert triggers, squatitive negation, and other quirks of grammar. In J. Hoeksema (ed.), Perspectives on negation and polarity items. Amsterdam: John Benjamins.

Huang, C. T. J. 1982. Logical relations in Chinese and the theory of grammar. PhD dissertation, MIT.

Jones, M. 1993. Sardinian syntax. London: Routledge.

Jones, M. A. 1988. Sardinian. In M. Harris & V. Nigel (eds.), The Romance languages, 314-350. London/Sydney: Croom Helm.

Kamp, H. & U. Reyle. 1993. From discourse to logic. Dordrecht: Kluwer.

Katz, L. & R. Frost. 1992. The reading process is different for different orthographies: the orthographic depth hypothesis. In. L. Katz & R. Frost (eds.), Orthography, phonology, morphology & meaning, 67–84. Amsterdam: Elsevier North Holland Press.

Kayne, R. 1975. French syntax: the transformational cycle. Cambridge, Mass.: MIT Press.

Kayne, R. 1983. Connectedness and binary branching. Dordrecht: Foris.

Kayne, R. 1989. Null subjects and clitic climbing. In: O. Jaeggli & K. Safir (eds), The null subject parameter, 239-261. Dordrecht: Reidel.

Kayne, R. 1991. Romance clitics, verb movement, and PRO. Linguistic inquiry 22, 647- 686.

Kempchinsky, P. 2009. What can the subjunctive disjoint reference effect tell us about the subjunctive? Lingua 119 (12), 1788–1810.

Kiss, E. 2010. An adjunction analysis of quantifiers and adverbials in the Hungarian sentence. Lingua 120, 506–26.

Kiss, K. 1995. Discourse-configurational languages. Oxford: Oxford University Press.

253 Kiss, K. 1998. Identificational focus versus information focus. Language, 74, 245-73.

Krifka, M. 2001. For a structured account of questions and answers. In C. Fery & W. Sternefeld (eds.), Audiatur vox sapientiae, Berlin, Akademie-Verlag.

Krifka, M. 2007. Basic notions of information structure. In C. Fery & M. Krifka (eds.), Interdisciplinary studies of information structure, 6. Potsdam: Universitätsverlag Potsdam.

Krifka, M. & R. Musan. 2012. Information structure. overview and linguistic issues. In M. Krifka & R. Musan (eds), The expression of information structure, 1-44. Berlin, Boston: De Gruyter Mouton.

Kuno, S. 1973. The structure of the Japanese language. Cambridge: Massachusetts Institute of Technology Press.

Kuno, S. 1976. Subject raising. In M. Shibatani (ed.), Syntax and semantics 5: Japanese generative grammar, 17-49. New York: Academic Press.

La Fauci, N. 1988. Oggetti e soggetti nella formazione della morfosintassi romanza. Pisa: Giardini.

Ladefoged, P. 2003. Phonetic data analysis: an introduction to fieldwork and instrumental phonetics. Oxford: Blackwell

Ladusaw, W. A. 1979. Polarity sensitivity as inherent scope relations. Ph.D. Dissertation. Austin: University of Texas.

Laka, I. M. 1990. Negation in syntax: on the nature of functional categories and projections. Ph. D. Dissertation. MIT.

Lambrecht, K. 1986. Pragmatically motivated syntax: Presentational cleft constructions in spoken French. In Proceedings of the twenty-second regional meeting of the Chicago Linguistic Society. Papers from the parasession on pragmatics and grammatical theory, 115-126. Chicago: Illinois.

Lambrecht, K. 1987. Sentence focus, information structure, and the thetic-categorical distinction. In Proceedings of the thirteenth annual meeting of the Berkeley Linguistics Society, 366-382. Berkeley: California.

Lambrecht, K. 1994. Information structure and sentence form: Topic, focus, and the mental representation of discourse referents. Cambridge: Cambridge University Press.

254 Lambrecht, K. 2000. Prédication seconde et structure informationnelle. La relative de perception comme construction présentative. In Langue française, 127, 49-66.

Lambrecht, K. 2001. A framework for the analysis of cleft constructions. Linguistics 39, 463-516.

Lasnik, H. & M. Saito. 1992. Move α: Conditions on its applications and outputs. Cambridge, MA: MIT Press.

Ledgeway, A. 2003. L’estensione dell’ausiliare perfettivo avere nell’antico napoletano: intransitività scissa condizionata da fattori modali. Archivio glottologico italiano 88. 27–71.

Ledgeway, A. 2005. Moving through the left periphery: the dual complementiser system in the Dialects of . Transactions of the Philological Society 103, 336-396.

Ledgeway, A. 2008. Satisfying V2 in early Romance: Merge vs. move. Journal of linguistics, 44, 437-470.

Ledgeway, A. & M. Maiden (eds.). The Oxford Guide to the Romance languages. Oxford: Oxford University Press.

Leonetti, M. & V. Escandell-Vidal. 2009. Fronting and verum-focus in Spanish. In A. Dufter & D. Jacob (eds.), Focus and background in Romance languages, 155- 204. Amsterdam: John Benjamins.

Levin, B. & M. Rappaport Hovav. 1995. Unaccusativity. Cambridge. MA: MIT Press.

Lewis, D. 1979. Scorekeeping in a language game. Journal of philosophical logic 8, 339-359.

López, L. 2009. A derivational syntax for information structure. Oxford/New York: Oxford University Press.

Loporcaro, M. 2011. Syncretism and neutralization in the marking of Romance object agreement. In M. Maiden, J. C. Smith, M. Goldbach & M. Hinzelin (eds.), Morphological autonomy: perspectives from Romance inflectional morphology, 328-357. Oxford: Oxford University Press.

Maiden, M. & M. Parry (eds), 1997 The Dialects of Italy. Routledge, London

Maling, J. 1980. Inversion in Embedded Clauses in Modern Icelandic. Íslenskt mál og almenn málfræði 2, 175-193.

255 Manzini, M. R. & L. Savoia. 2005. I dialetti italiani e romanci. Morfosintassi generativa. Alessandria: Edizioni dell’Orso.

Manzini, M. R. & L. Savoia. 2011. Wh-in situ and wh-doubling in Northern Italian varieties: Against remnant movement. Linguistic Analysis.

Marcato, C. 1986. Forme verbali bicomposte (“surcomposées”) nelle parlate del . In M. Cortelazzo (ed.), Guida ai dialetti veneti 8, 45-60. Padova: Cleup.

Martín-González, J. 2002. The Syntax of Sentential Negation in Spanish. Ph.D. Dissertation, Harvard University.

Melchior, L. 2012. Tra esperienzialità e iteratività: il ‘passé surcomposé à valeur spéciale’ in francese (e in altri idiomi romanzi). Revue de linguistique romane, 301-302, 65-98.

Molnár, V. 2002. Contrast in a contrastive perspective. In Hilde Hasselgård, Stig Johansson, Bergljot Behrens & Cathrine Fabricius-Hansen (eds.), Information Structure in a Cross-Linguistic Perspective, 147-161. Amsterdam/New York: Rodopi.

Molnár, V. 2006. On different kinds of contrast. The Architecture of Focus. Berlin, Boston: De Gruyter Mouton.

Moseley, C. 2012. The UNESCO Atlas of the World's Languages in Danger: Context and Process. http://www.unesco.org/languages-atlas.

Munaro, N. 1997. Proprietà distribuzionali dei sintagmi interrogativi in alcuni dialetti veneti settentrionali. Quaderni di lavoro ASIS 1, 63-74.

Munaro, N. 1998. Wh-in situ in the Northern Italian Dialects. In Olga Fullana & Francesc Roca (eds), Studies on the Syntax of Central Romance Languages, 189- 212. Girona: Universitat de Girona Convegno: III Symposium on the Syntax of Central Romance Languages.

Munaro, N. 1999. Sintagmi interrogativi nei dialetti italiani settentrionali. Padova: UNIPRESS

Munaro, N. 2011. Toward a hierarchy of clause types. In P. Benincà & N. Munaro (eds.), Mapping the Left Periphery, 125-162. Oxford/New York: Oxford University Press.

Munaro, N., C. Poletto & J. Pollock. 2001. Eppur si muove! On Comparing French and Bellunese Wh-Movement. Linguistic Variation Yearbook 1, 147-180. Amsterdam/Philadelphia: John Benjamins.

256

Munaro, N. & C. Poletto. 2002. La tipologia dei wh in situ nelle varietà alto-italiane. Quaderni patavini di linguistica 18, 79-91.

Munaro, N. & C. Poletto. 2014. Synchronic and diachronic clues on the internal structure of 'where' in Italo-Romance. In P. Benincà, A. Ledgeway & N. Vincent (eds.), Diachrony and dialects, 279-300. Oxford: Oxford University Press.

Munaro, N. & J. Y. Pollock. 2005. Qu'est-ce-que (qu)-est-ce-que? A case study in comparative Romance interrogative syntax. In G. Cinque & R. Kayne (eds.), The Oxford Handbook of Comparative Syntax, 542– 606. Oxford: Oxford University Press.

Nathan, D. 2006. Thick interfaces: mobilizing language documentation with multimedia. In J. Gippert, N. Himmelman & U. Mosel (eds.), Essentials of Language Documentation, 363-379. Berlin: Mouton de Gruyter.

Nathan, D. 2010. Sound and unsound practices in documentary linguistics: towards an epistemology for audio. In P. Austin (ed.), Language Documentation and Description 7, 262-284. London: SOAS.

Neeleman, A., E. Titov, H. van de Koot & R. Vermeulen. 2009. A syntactic typology of topic, focus and contrast. In Jeroen van Craenenbroeck (ed.), Alternatives to Cartography, 15-51. Berlin: Mouton de Gruyter.

Oliviéri, M. & P. Sauzet. 2016. Southern Gallo-Romance (Occitan). In A. Ledgeway & M. Maiden (eds.), The Oxford Guide to the Romance Languages, 319-349. Oxford: Oxford University Press.

Oliviéri, M. 2004. Paramètre du sujet nul et inversion du sujet dans les dialectes italiens et occitans. Cahiers de Grammaire Questions de linguistique et de dialectologie romanes 29, 105-120.

Paoli, S. 2007. The fine structure of the left periphery: COMPs and subjects. Evidence from Romance. Lingua 117 (6), 1057–1079.

Pellegrini G. B. 1979. I dialetti ladino-cadorini. Archivio per l’Alto Adige 72, 245-265.

Pellegrini, G. B. 1972. Saggi sul ladino dolomitico e sul friulano. Bari: Adriatica.

Pellegrini, G. B. 1977. Carta dei dialetti d’Italia. Pisa: Pacini.

Pellegrini, G. B. 1992. Studi storico-linguistici bellunesi e alpini. Belluno

Pellegrini, G. B. 1991. La genesi del retoromanzo (o ladino). Tübingen: Niemeyer. 257

Pescarini, D. 2016. Clitic Pronominal Systems: Morphophonology. A. Ledgeway & M. Maiden (eds.), 742-757.

Pescarini, D. 2016. Clitics: morphophonology. In A. Ledgeway & M. Maiden (eds.), 742-757

Pesetsky, D. 1987. Wh -in-situ: Movement and unselective binding. In Eric J. Reuland and Alice G.B. ter Meulen (eds.), The Representation of (In)definiteness, 98– 129. Cambridge: MIT Press.

Poletto, C. 1992. La sintassi del soggetto nei dialetti italiani settentrionali. Ph.D. thesis. Padua: University of Padua.

Poletto, C. 1993. Subject clitic-verb inversion in north eastern Italian dialects. In A. Belletti (ed.), Syntactic Theory and the Dialects of Italy, 204-251.: Rosenberg and Sellier.

Poletto, C. 2000. The Higher Functional Field. Evidence from Northern Italian Dialects. Oxford/New York: Oxford University Press.

Poletto, C. 2001. Complementizer deletion and verb movement. In G. Cinque & G. Salvi (eds.), Current Studies in Italian Syntax. Essays Offered to Lorenzo Renzi, 265–86. Amsterdam: Elsevier.

Poletto, C. 2002. The left-periphery of V2-Rhaetoromance dialects: A new view on V2 and V3. In S. Barbiers, L. Cornips & S. van der Kleij (eds.), Syntactic microvariation, 214-242. Amsterdam: Meertens Instituut.

Poletto, C. 2006. Parallel phases: a study on the high and low left periphery of old Italian. In M. Frascarelli (ed), Phases of interpretation. Studies in generative grammar 91, 261-294. Berlin: Mouton de Gruyter.

Poletto, C. 2009. Doubleauxiliaries,anteriorityand terminativity. Journal of comparative Germanic linguistics 12:31‐48.

Poletto, C. 2013. On V2 Types. In S. Luraghi & C. Parodi (eds.), The Bloomsbury Companion to Syntax, 154–164. London: Bloomsbury.

Poletto, C. 2014. Word order in old Italian. Oxford: Oxford University Press.

Poletto, C. & J. Pollock. 2004. On the left periphery of some Romance Wh-Questions. In L. Rizzi (ed.), The Cartography of Syntactic Structures Vol. 2, The Structure of CP and IP, 251-296. Oxford: Oxford University Press.

258 Poletto, C. & J.Y Pollock. 2009. Another look at wh-questions in Romance: the case of Medrisiotto and its consequences for the analysis of French wh-in-situ and embedded interrogatives, 199-258. Amsterdam: John Benjamins.

Poletto, C. & J.Y. Pollock. 2015. Arguing for Remnant Movement. 135-178. Berlin: Mouton de Gruyter.

Poletto, C. & R. Zanuttini. 2003. Making imperatives: evidence from central Rhaetoromance. In C. Tortora (ed.), The syntax of Italian dialects,175-207. Oxford University Press.

Poletto, C. & R. Zanuttini. 2013. Emphasis as reduplication: evidence from si che/no che sentences, Lingua 128, 124-141.

Prieto, P. & G. Rigau .2007. The Syntax-Prosody Interface: Catalan interrogative sentences headed by que. Journal of Portuguese Linguistics 6 (2), 29–59.

Prince, E. 1981. Topicalization, focus movement and Yiddish movement: a pragmatic differentiation. BLS 7, 249–64.

Quer, J. 2002. Edging quantifiers. On QP-fronting in western Romance. In C. Beyssade, R. Bok-Bennema, F. Drijkoningen & P. Monachesi (eds.), Romance linguistics & linguistic theory 2000, 254-270. Amsterdam, Philadelphia: John Benjamins.

Reglero, L. & E. Ticio. 2013. A unified analysis of wh-in-situ in Spanish. The Linguistic Review 30, 1-47.

Reinhart, T. 1981. Pragmatics and linguistics: An analysis of sentence topics. Philosophica 27, 53-94.

Renzi, L. & A. Andreonese. 2015. Manuale di linguistica e filologia romanza, Nuova edizione, Bologna, Il Mulino.

Renzi, L. & L. Vanelli. 1983. I pronomi soggetto in alcune varietà romanze. In Scritti linguistici in onore di G.B. Pellegrini, 121-145. Pisa: Pacini.

Richards, N. 2016. Contiguity theory. Cambridge, MA: MIT Press.

Rizzi, L. 1986. On the Status of Subject Clitics in Romance. In O. Jaeggli & C. Silva- Corvalan (eds.), Studies in Romance Linguistics. Dordrecht: Floris. 137-52

Rizzi, L. 1990. Relativized Minimality. Cambridge, MA: MIT Press.

Rizzi, L. 1991. Residual Verb Second and the Wh-criterion. Geneva Generative Papers, 2.

259

Rizzi, L. 1996. Residual Verb Second and the Wh Criterion. In A. Belletti & L. Rizzi (eds.), Parameters and Functional Heads, 63-90. Oxford University Press.

Rizzi, L. 1997. The fine structure of the left periphery. In L. Haegeman (ed.), Elements of Grammar, 281–337. Dordrecht: Kluwer.

Rizzi, L. 2001a. Reconstruction, weak island sensitivity, and agreement. In C. Cecchetto, G. Chierchia & M. T. Guasti (eds.), Semantic Interfaces, 145–176. Stanford: CSLI.

Rizzi, L. 2001b. On the position of Int(errogative) in the Left Periphery of the Clause. In G. Cinque & G. Salvi (eds.), Current Studies in Italian Syntax, 287-296. Amsterdam: Elsevier.

Rizzi, L. 2004. Locality and left periphery. In A. Belletti (ed.), Structures and beyond: The cartography of syntactic structures 3, 223–251. New York: Oxford University Press.

Rizzi, L. 2006. On the form of chains: criterial positions and ECP effects. In L. Lai- Shen Cheng & N. Corver (eds.), Wh-movement: moving on, 97–134. Cambridge, MA: MIT Press.

Rizzi, L. 2013. Notes on cartography and further explanation. Probus 25, 97-226.

Rizzi, L. 2018. Uniqueness of left peripheral focus, “further explanation”, and Int. In L. R. Bailey & M. Sheehan (eds.). Order and structure in syntax I: Word order and syntactic structure, 333-343. Berlin: Language Science Press.

Rizzi, L., & G. Bocci. 2017. The left periphery of the clause – Primarily illustrated for Italian. In Blackwell Companion to syntax, II edition. Blackwell Publishers.

Rizzi, L. & U. Shlonsky. 2006. Satisfying the subject criterion by a non subject: English locative inversion and heavy NP shift. In: M. Frascarelli (ed.), Phases of interpretation, 341-361.Berlin: Mouton de Gruyter.

Rizzi, L. & U. Shlonsky. 2007. Strategies of subject extraction. In H. M. Gärtner & U. Sauerland (eds.), Interfaces + Recursion = Language? Chomsky’s minimalism and the view from syntax-semantics, 115–160. Berlin: Mouton de Gruyter.

Roberts, I. 1993. Verbs and diachronic syntax: a comparative history of English and French. Dordrecht: Kluwer.

Roberts, I. 1999. Verb movement and markedness. In Michel deGraff (ed.), Language creation and change, 287. Cambridge, Mass.: MIT Press.

260

Roberts, I. 2004. The C-system in Brythonic ,V2 and the EPP. In L. Rizzi (ed.), The structure of CP and IP. The cartography of syntactic structures 2, 297–328. Oxford: OUP.

Roberts, I. & Holmberg, A. 2009. Introduction: Parameters in minimalist theory. In: T. Biberauher, A. Holmberg, I. Roberts & M. Sheehan (eds), Parametric variation: null subjects in minimalist theory. Cambridge: Cambridge University Press.

Rogers, H. 2005. Writing systems. A linguistic approach. Oxford: Blackwell

Rooth, M. 1995. Indefinites, adverbs of quantification and focus semantics. In G. N. Carlson & F. J. Pelletier (eds), The generic book, 265-291. Chicago: Chicago University Press.

Rooth, M. 1996. Focus. In S. Lappin (ed), The handbook of contemporary semantic theory, 271-298. Oxford: Blackwell.

Ross, J. R. 1967. Constraints on variables in syntax. PhD dissertation, MIT.

Rothstein, S. 1995. Small clauses and copular constructions. In A. Cardinaletti & M. T. Guasti (eds.), Small Clauses, 27–48. (Syntax and Semantics). San Diego: Academic Press.

Rothstein, S. 2000. Secondary predication and aspectual structure. In E. Lang, C. Fabricius-Hansen & C. Maienborn (eds.), ZAS papers in linguistics: papers from the Oslo conference on adjuncts. ZAS, Berlin.

Salvi, G. 1997. Ladin. In M. Maiden - M. Parry (eds.), The dialects of Italy, 286-294. London: Routledge.

Salvi, G. 2004. La formazione della struttura di frase romanza. Ordine delle parole e clitici dal latino alle lingue romanze antiche. Tübingen: Niemeyer.

Salvi, G. 2004. La formazione della struttura di frase romanza. Ordine delle parole e clitici dal latino alle lingue romanze antiche. Tübingen: Niemeyer.

Salvi, G. 2012. On the nature of the V2 system of Medieval Romance. In L. Brugé, A. Cardinaletti, G. Giusti, N. Munaro & C. Poletto (eds.), Functional heads: the cartography of syntactic structures 7, 103–111. Oxford: Oxford University Press.

Salvi, G. 2016. Word Order. In A. Ledgeway & M. Maiden (eds.), 997-1012.

261 Samek-Lodovici, V. 2015. The interaction of focus, givenness, and prosody. A study of Italian clause structure. Oxford: Oxford University Press.

Seifart, F. 2006. Orthography development. In J. Gippert, N. Himmelmann & U. Mosel (eds.), Essential of language documentation, 275-299. Berlin: Mouton de Gruyter.

Selkirk, E. 2002. Contrastive FOCUS vs. presentational focus: Prosodic evidence from right node raising in English. In B. Bel & I. Marlien (eds.), Speech Prosody 2002: Proceedings of the 1st International Conference on Speech Prosody, 643- 646. Aix-en-, : Laboratoire Parole et Langage.

Shimojo, M. 2011. The left periphery and focus structure in Japanese. In Wataru Nakamura (ed.), New Perspectives in Role and Reference Grammar, 266-293. Cambridge Scholars Publishing.

Stalnaker, R. 1974. Pragmatic presuppositions. In M. K. Munitz & P. K. Unger (eds.), Semantics and Philosophy, 197-214. New York: New York University Press.

Stalnaker, R. 1978. Assertion. In Syntax & semantics 9. New York: Academic Press.

Stalnaker, R. 2002. Common ground. Linguistics & philosophy 25, 701–21.

Stowell, Tim. 1982. Subjects across categories. Linguistic review 2 (3), 285–312.

Szabolcsi, A. 1981. The semantics of topic-focus articulation. In J. Groenendijk, T. Janssen & M. Stokhof (eds.), Formal Methods in the Study of Language, 513- 541. Amsterdam: Matematisch Centrum

Tomaselli, A. 1990. La sintassi del verbo finito nelle lingue germaniche. Padua: Unipress.

Travis, L. 1984. Parameters and Effects of Word Order Variation. PhD Dissertation, MIT.

Truckenbrodt, H. 1999. On the relation between syntactic phrases and phonological phrases. Linguistic inquiry 30, 219-255.

Tuttle, E. Veneto. In M. Maiden & M. Parry (eds), The dialects of Italy, 263-70. London: Routledge.

UNESCO. 2003. Language Vitality and Endangerment. Paris: International Expert Meeting on UNESCO Programme Safeguarding of Endangered Languages.

262 Valentin, D. 1998. L’ordine dei costituenti nella frase del ladino della Val Badia, tesi di laurea, Università di Padova.

Vallduví, E. 1993. A preverbal landing site for quantificational operators. Catalan working papers in linguistics 2, 319-343.

Van Valin, R. D. 1993. A synopsis of Role and Reference Grammar. In R. D. Van Valin Jr. (ed.), Advances in Role and Reference Grammar. Amsterdam: John Benjamins.

Van Valin, R. D. 1999. A typology of the interaction of focus structure and syntax. In E. V. Rachilina & J. G. Testelec (eds.), Typology and linguistic theory from description to explanation: For the 60th birthday of Aleksandr E. Kibrik, 511- 524. Moscow: Languages of Russian Culture.

Vanelli, L. 1986. Strutture tematiche in italiano antico. In Harro Stammerjohann (Ed.), Tema-Rema in Italiano, 249–73. Tübigen, Germany: Gunter Narr Verlag.

Vanelli, L. 1998. I dialetti italiani settentrionali nel panorama romanzo. Studi di sintassi e morfologia. Roma: Bulzoni.

Vanelli, L., L. Renzi & P. Benincà. 1986. Tipologia dei pronomi soggetto nelle lingue romanze medievali. Quaderni Patavini di Linguistica 5, 49–66.

Vicario, F. 2005. Lezioni di linguistica friulana. Forum: Udine

Vicario, F. et al. 2009. Dizionario storico friulano. http://www.dizionariofriulano.it/easyne2/SZN.aspx?ID=19&CODE=DZFR. Diachronic Corpus of the Friulian Language

Vikner, S. 1995. Verb movement and expletive subjects in the Germanic languages. Oxford/New York: Oxford University Press.

Vilkuna, Maria. 1995. Discourse configurationality in Finnish. In K. É. Kiss (ed.), Discourse-Configurational Languages, 244-268. Oxford/New York: Oxford University Press.

Villa-García, J. 2012. Recomplementation and locality of movement in Spanish. Probus 24 (2), 257–314.

Villa-García, J. 2015. The syntax of multiple-que sentences in Spanish : along the left periphery. Amsterdam/Philadelphia: John Benjamins.

263 Villa-García, J. 2016. TP-ellipsis with a polarity particle in multiple-complementizer contexts in Spanish: on topical remnants and focal licensors. Borealis: An International Journal of Hispanic Linguistics 5 (2), 135-172.

Wolfe, S. 2016. On the Left Periphery of V2 Languages. Rivista di Grammatica Generativa 38, 287-310.

Woodbury, A. 2003. Defining documentary linguistics. In P. K. Austin (ed), Language documentation and description, 35.

Wright, R. 2016. Latin and Romance in the Medieval Period. In A. Ledgeway & M. Maiden (eds.), 14-23.

Zamboni, A. 1974. Veneto. In M. Cortelazzo (ed.), Profilo dei dialetti italiani. Pisa: Pacini.

Zamboni. A. 1984. I dialetti cadorini. In G. B. Pellegrini & S. Sacco (eds), Il ladino bellunese. Belluno: Istituto bellunese di ricerche sociali e culturali.

Zamboni. A. 1998. Dal latino tardo al romanzo arcaio: aspetti diacronico-tipologici della flessione nominale. In P. Ramat & E. Roma (eds.), Sintassi storica. Atti del XXX congresso internazionale della società di linguistica italiana. : Bulzoni, 127-46.

Zanuttini, R. 1997. Negation and Clausal Structure: A Comparative Study of Romance Languages. New York: Oxford University Press.

Ziv, Y. 1994. Left and right dislocations: Discourse functions and anaphora. Journal of Pragmatics. 22, 629-645.

Zorzi, A. 2001. La Repubblica del Leone. Storia di Venezia. Milano: Bompiani.

Zubizarreta, M. L. 1998. Prosody, focus, and word order. MIT Press, Cambridge, Mass.

264